Turning a commissary freezer into a semi-blast chiller: what the Tech Ingredients videos get right, what they get wrong, and what to build

Written 2026-08-30 for MOT Restaurant Group. All prices and specifications were observed on 2026-08-30 at the URL given unless another date is stated. Where a page could not be loaded the line says "not confirmed" and no number is invented. Section 9 lists every source. The report was revised twice: once against an internal three-lens review (thermodynamics, refrigeration, procurement) and once against an independent Codex red-team of 30 findings, and then given a final consistency pass that reconciled the fan-kit, BOM and recharge figures across sections. Both rounds are logged at the end, in "Review log" and "Codex review log", with one row per finding and the change made.

Scope. The commissary cools stocks, sauces, cooked grains and braises by putting them straight into a commercial freezer. Nothing here needs a certified blast chiller under the health code; the goal is to cool faster and to stop beating up the freezer. Freezer type is unconfirmed, so both a commercial reach-in upright (forced air, electric defrost) and a commercial chest freezer (static, manual defrost) are covered. One fact reported by the owner after the research started changes the design more than anything else and is treated as a design input throughout: the working freezers hold a stable -15 to -18 C with no load. A second fact arrived after the Codex review and is now the center of the PCM half: the owner has bought three Arctic Ice Tundra XX-L packs (10 lb each, -15 C). They are the PCM to test and use first, in a swap bank with a dedicated recharge chest freezer (section 6.3, Tier D); the -23 C U-tek pack the owner asked about earlier is the higher-driving-force variant of the same scheme and needs a colder recharge box.

1. Summary

• Video 1 (fan) claims a 12 V radiator fan zip-tied inside a freezer halves the time to chill 0.8 kg of water from 50 C to 0 C (120 min to 60 min) and triples the freezing rate (300 min to 100 min). The numbers are internally consistent and the physics is right. The "thousandfold" increase in heat transfer is not; at the 2 to 4 m/s a radiator fan gives over a pot, the air-side coefficient goes up 2x to 3x, and the product's own resistance eats most of that.
• Video 2 (goop) claims a 705 g water / 295 g ammonium chloride mix is a eutectic melting at -15.4 C. The temperature is right; the ratio is not. The NH4Cl-water eutectic is 19.7 wt%, the recipe is 29.5 wt%, and 122 g of every 295 g of salt is dead weight. Effective latent heat is about 240 to 265 kJ/kg (estimate; 250 is used in the tables) instead of 290 to 320. Use about 265 g of salt per 1,000 g of water (21 wt%, a hair over the eutectic so a few salt crystals are present to seed the freeze).
• The "700 kJ to freeze 1 kg of warm water" line is 25 percent high for the 50 C test (543 kJ) and right for a pot straight off the stove (87 C). The 5 h plateau in room air implies 11 W of heat gain per bottle, which is plausible for a sheltered, frosted bottle.
• What hot food does to a healthy freezer, per the manufacturers' own manuals, is filed under "does not refrigerate well": moisture that frosts the coil, longer defrosts, inventory warming, and extra run time. Hoshizaki and Avantco say outright not to put hot food in a holding cabinet, and Turbo Air files it as a cause of poor refrigeration. The compressor risks (high suction, protector trips, hot restarts) are real mechanisms whose size for this box is unknown until a technician measures one batch; a dirty condenser or a hot room makes each of them worse. The cheapest thing in this report for freezer life is a monthly condenser cleaning, not a PCM bank. A PCM bank at the bottle counts in section 6.2 cannot cap the box air during a chill (section 4.2 works the rate); it stores latent capacity that the box pays back overnight, and it spares the frozen inventory part of the excursion. It does nothing for the evaporator, the frost or the defrost load; only covering pans and using 2 in pans does that.
• The owner now has three Arctic Ice Tundra XX-L packs (10 lb / 4.54 kg each, 13.6 kg total, -15 C phase point, rigid HDPE, 10.5 x 10.5 x 3.5 in per the maker's dimensions graphic, https://arctic-ice.com/products/tundra-series, seen 2026-08-30). They are the PCM for this project and the first thing to test. They cannot live in the working freezers: a box cycling -15 to -18 C gives them 0 to 3 K of margin, so they never recharge there and lose 10 to 15 percent of their charge a day at the warm end of the cycle and through door openings. They belong in a swap bank: stored and recharged in a dedicated chest freezer whose logged mean at the pack position is -21 C or colder, moved onto the shelf in the fan stream when a batch goes in, and taken back when the batch reaches 5 C. Any commercial chest freezer rated to -23 C or colder does this. The -23 C U-tek pack the owner asked about earlier fits the same swap bank and absorbs heat two to four times faster per square meter while the box is healthy, but only with a recharge freezer verified at -28 C or colder, which no ordinary commercial chest freezer is rated for; left in the working box it self-discharges in half a day to a day in still air and in 4 to 10 h with the fan on, so it can never be a bank stored there. Section 6.3 Tier D sizes both. Neither maker publishes latent heat; the report carries 220 / 260 / 300 J/g cases for the Tundra packs (1.0 / 1.18 / 1.36 MJ per pack) and a one-day calorimeter test in Tier D pins it down. Before spending anything more, clean the condenser and log the working box for 24 h. A reach-in holding -15 to -18 C empty is at the warm edge of most models' rated ranges (0 F is -18 C) and 5 to 8 K warm for a True T-23F-HC; read the model's manual and log it before calling it faulty, and clean the condenser either way (section 6.1, step 0).
• A booster fan helps in either freezer type. In a static chest freezer it is worth 4x to 6x on the surface coefficient; in a reach-in that already has an evaporator fan it is worth 2x to 3x. Expect about 1.3x to 1.6x on chill time in a reach-in (its baseline already has 400 to 560 CFM of forced air) and about 2x in a chest freezer, which is where the video's own 2x came from since its freezer's stock fan moved tens of CFM.
• Buying instead: the cheapest new blast chiller that plugs into a 115 V outlet is the Avantco SF-3 at $2,089 (15 lb per 90 min) and the SF-5 at $3,789 (31 lb); the Atosa EBF-05 does 55 lb for $5,707.00 list at CKitchen (a $5,187.60 promotional price showed beside it on 2026-08-30; confirm on a written quote). Used 5-pan Traulsen and Beverage-Air units run $1,899 to $4,500 on eBay. A dedicated unit is the only option that takes the load off the storage freezer instead of shifting it to overnight.
• Four build tiers, all before shipping and tax. Tier D is the default now that the packs are bought: the three XX-L packs (sunk), a recharge chest freezer ($299 for a Galaxy CF7 that must pass a 24 h log at -21 C or colder, about $506 to $543 for an Atosa MWF9007 rated to -10 F, $1,369 for a Galaxy Ultra Cold UCCF8 rated to -50 F; prices seen 2026-08-30, section 6.3), and the fan kit ($240 to $345 for the rated electrical stack in section 6.3a, one or two fans, or about $195 to $205 for the video-copy fan and PWM on a dedicated cabinet the owner accepts as altered). The -23 C U-tek variant of Tier D (two sets of twelve, four cases of six, $352, plus a recharge box verified at -28 C or colder: $288 for a household -40 F chest freezer that must prove itself on a logged recharge, $1,369 for the UCCF8, $3,376 for a 5 cu ft lab chest) is the higher-driving-force upgrade at about $905 to $1,045 for one 10 kg batch a day, not the first purchase. Tiers A to C stay as fallbacks for a 10 kg batch: Tier A copies the videos in labeled lab bottles ($465 as printed, of which $76 is unconfirmed placeholders), Tier B uses FCC-grade KCl in the same bottles ($442), for a dedicated chilling cabinet only, because its -10.7 C plateau cannot hold frozen inventory below -12 C, and Tier C buys Engel -15 C packs, whose latent heat Engel does not publish, and a sealed fan ($445). With the rated fan of section 6.3a, which is driven from its own PWM lead so the RioRand drops out, those are $487, $464 and $427. At 30 kg per batch Tiers A to C run $652 to $693 as printed.
• Recommended path: clean the condenser and log the working box first. Buy the fan kit (section 6.3a; no DIY fan goes into an R290 or R600a cabinet without the maker's written approval, and a reach-in has no owner-installable cable route, section 6.5) and the recharge chest freezer, and log the chest freezer for 24 h at its coldest setting before keeping it. Run the two pack tests in section 6.3 Tier D (calorimeter for latent heat, melt test for conditioning time), then the airflow test and the real-batch test in section 6.7 with the three packs as run D. Expect the packs to take 0.4 to 1.0 MJ of a 10 kg batch during the chill (20 to 45 percent) and to repay it in the recharge box, not the working box. One set of three is one 10 kg batch every other day until the melt test proves an overnight recharge; a second set of three makes it one batch a day. Add packs 4 and 5 for surface area, not capacity, and consider Engel XL 5 lb packs for those slots because a thinner pack has twice the skin per kilogram.
• Go/no-go for the whole experiment: the pot test in section 6.7 only positions the fan (three fan-off and three fan-on runs, alternated, after the box has equilibrated overnight). The decision is a logged real batch at the SOP's mass, pans and fill depth: 21 C (70 F) by 2 h and 5 C (41 F) by 6 h with margin, the frozen-inventory sentinel never above -12 C, the compressor not short-cycling, packs charged on schedule, and the result repeated on a second day. Fail any of those and this freezer does not chill that batch: halve it, pre-cool it, dedicate a cabinet, or buy the Avantco SF-5. The pot removes about 190 kJ and a 10 kg batch about 2,200 kJ; there is no multiplier that turns one into the other.

2. What the two videos actually did

2.1 Video 1: the fan retrofit (2026-07-09)

"We turned a household freezer into a high-speed blast chiller!" (https://www.youtube.com/watch?v=G1Syjh319VU, 19 min 41 s, 124,153 views on 2026-08-30). The presenter's argument is that a blast chiller is a fan plus a big compressor, that a residential freezer's compressor is too small to matter, but that the 20 to 50 kg of hard-frozen food already in the freezer is a heat sink big enough to absorb a batch if a fan moves the heat into it fast enough.

Hardware from the description:

• Fan: GESEXI 16 in slim push/pull radiator fan, 12 V, 120 W, 2-pack, Amazon B0DHKQS3DC. The video says "$50 for a pair"; the listing showed $59.99 for two on 2026-08-30 (per the fan survey in this research; the price was not visible on the page when refetched for this report).
• Power: Jesverty SPS-3010 0 to 30 V, 0 to 10 A bench supply, Amazon B09YSJQWRG, $47.49 on 2026-08-30.
• Install: fan blade removed (one nut), flipped, refitted, and the leads reversed so the fan blows through its flat face. Zip-tied flat on a wire shelf near the bottom of an upright residential freezer with the tabs tucked so the shelf stays usable.
• Operating point: the presenter swept voltage against an anemometer and found "at about 8 to 9 volts this thing will move about 80 percent of its maximum air flow and use less than half of the power." The test ran at 9 V.
• Sizing rule: fan cross-section about 40 percent of the freezer cross-section so air can return around it.

Measurements from the transcript:

The presenter notes that air temperature ran "consistently higher" during the fan test because the same heat came out of the pot in a shorter time. That observation is the whole story of this project: the fan does not add refrigeration, it moves heat faster into whatever cold mass is already there.

2.2 Video 2: the eutectic goop (2026-08-29)

"Why This Ugly Black Goop Is BETTER Than Ice" (https://www.youtube.com/watch?v=IGlZ61oCcEY, 40 min 26 s, 212,858 views on 2026-08-30). The problem stated at the top: relying on the frozen food as the heat sink means the freezer has to be full, and repeated batches warm the food until it thaws. The fix is a phase change material that melts below the food's freezing point, so it takes the heat instead of the inventory.

Recipe from the video description, per bottle batch:

Method: water preheated to about 50 C, salt dissolved in a stand mixer with a heated bowl (dissolving is endothermic; the mix dropped to 34 C after half the salt), bentonite sprinkled in, carbon fiber added with the mix under 40 C so the fiber's sizing does not gum up, then xanthan. Filled to 90 percent in PET sports-drink bottles (Powerade) through a funnel with a dowel, later with a vacuum-assisted filler. Claimed melting point -15.4 C.

Test: three bottles from a -20 C freezer left in room air, probes in each. Plain water and the contents of a commercial gel pack (sodium polyacrylate) both plateaued at 0 C. The goop plateaued near -15 C for about 5 hours, then warmed as a liquid. The presenter's conclusion that gel packs are "just encased ice" is correct and the graph supports it.

3. Physics check

3.1 The eutectic composition (the one real error)

The NH4Cl-H2O eutectic is at 19.7 wt% NH4Cl and -15.4 C. The composition is stated in Peppin, Huppert and Worster, "Steady-state solidification of aqueous ammonium chloride", J. Fluid Mech. 599, 465 to 476 (2008), https://doi.org/10.1017/S0022112008000219 ("the eutectic concentration of 19.7 wt%", seen 2026-08-30); the temperature is the standard phase-diagram value and the video's own plateau measurement lands on it. (An earlier draft cited MDPI Materials 2024, https://doi.org/10.3390/ma17112768; that paper is a dendrite-kinetics study that uses NH4Cl-H2O as an alloy analogue and does not state the eutectic numbers, so it is no longer the source.) Wikipedia's solubility of NH4Cl at -15 C, 244 g per litre of water, is 19.6 wt%, which is the liquidus point just above the eutectic (https://en.wikipedia.org/wiki/Ammonium_chloride).

The recipe is 295 g in 705 g, which is 29.5 wt%. That is, to within rounding, the "NH4Cl-70 wt% H2O" solution that metallurgy labs use as a transparent stand-in for hypereutectic alloy castings, which is probably where the number came from. What happens on cooling, using the NH4Cl solubility table (https://en.wikipedia.org/wiki/Solubility_table):

So from about 31 C down the salt crystallises out continuously, and at -15.4 C the remaining 878 g of liquid (705 g water plus 173 g salt, exactly 19.7 wt%) freezes as the eutectic. The plateau the video measured is real. The 122 g of surplus salt and the 38 g of additives are passive. The "hairy blobs" the presenter saw when the fiber went in at 32 to 35 C were partly salt coming out of solution, since the mix was already at saturation.

Latent heat. Ice is 333.5 kJ/kg at 0 C (Osborne, NIST 1939, https://nvlpubs.nist.gov/nistpubs/jres/23/jresv23n6p643_A1b.pdf). Corrected to -15.4 C that is about 300 kJ/kg; NH4Cl crystallising releases 14.78 kJ/mol, 276 kJ/kg (LibreTexts, https://chem.libretexts.org/Courses/Howard_University/General_Chemistry:_An_Atoms_First_Approach/Unit_4:__Thermochemistry/09:_Thermochemistry/Chapter_9.05:_Enthalpies_of_Solution). Mass-weighted for the eutectic that is 296 kJ/kg; this is a calculated (Kirchhoff) figure, not a measured one, and no PCM-grade measurement of the NH4Cl eutectic's latent heat or cycling life was found. PCM Products lists its commercial -15 C salt eutectic E-15 at 320 kJ/kg with a density of 1,060 kg/m3, which matches a 20 wt% NH4Cl brine (https://www.pcmproducts.net/files/PlusICE%20Range%202021-1.pdf). So the true eutectic stores 290 to 320 kJ/kg.

The fix is about 265 g NH4Cl per 1,000 g water (21 wt%). At exactly 20.0 wt% the first salt crystals only appear at about -14 C (liquidus by interpolation between 22.7 wt% at 0 C and 19.7 wt% at -15.4 C), 1.4 K above the eutectic, too late to be useful as a seed. At 21 wt% they appear from about -9 C, and the 16 g of surplus salt per kg costs under 2 percent of latent heat (about 291 kJ/kg instead of 296).

Bentonite as a nucleator is weak evidence either way, and supercooling is the one thing the recipe must not ignore. Montmorillonite nucleates ice in microlitre droplets of pure water at -31 to -35 C for typical sites and about -16 C for the rare best sites (Atmos. Chem. Phys. 2012, https://acp.copernicus.org/articles/12/5859/2012/); those are droplet experiments, and a 1 L gelled bottle with a wall, a few grams of salt crystals and any dust nucleates far more easily than a droplet. But salt-water eutectics in bulk routinely supercool several kelvin before ice forms, and the PCM literature treats supercooling as the main defect of salt-based PCMs. Its nucleator data (borax at 0.3 to 1.2 wt%, SrCl2, diatomite) come from salt hydrates such as Glauber's salt and sodium acetate trihydrate, where NH4Cl appears only as a melting-point additive, not from the NH4Cl-water eutectic itself (MDPI "Advances in the Stabilization of Eutectic Salts as PCMs", https://www.mdpi.com/2504-477X/9/12/667; OSTI, https://www.osti.gov/servlets/purl/2588542). No source shows that a xanthan-gelled 20 wt% NH4Cl brine with no particulates nucleates within 2 to 3 K of -15.4 C, which is all the driving force a -18 C box offers. So the corrected recipe keeps the video's 4 g/L of bentonite as cheap insurance, plus the surplus salt from the 21 wt% ratio, and adds nothing else. No borax: nothing shows it works in NH4Cl-water, nothing shows it is compatible with xanthan or the bottle, and it would be a second unlabeled chemical in a food area. If a jug still supercools after a full night at the recharge setting, the answer is not another additive; stop and move to a characterized commercial PCM (Tier C, the owner's Tier D packs, or the rigid packs in section 5.2). A jug reading colder than -15.4 C that is still liquid is supercooled; shaking it should snap it to -15.4 C (SOP step 1).

Two density claims are off. The presenter said the brine is "about 1.2 g/cm3"; a 24 wt% NH4Cl brine is 1.067 at 20 C (https://advancedthermo.com/electrolytes/density_NH4Cl.html) and the eutectic about 1.06. Carbon fiber is 1.76 to 1.93 g/cm3 on Toray's data sheets, not 1.6 (https://www.toraycma.com/wp-content/uploads/T300-Data-Sheet.pdf, https://www.toraycma.com/wp-content/uploads/M60J-Data-Sheet.pdf). Both errors point the same way: the fiber is 65 to 80 percent denser than the brine and will settle, so the xanthan is justified. Xanthan is salt-tolerant and freeze-thaw stable per its maker (https://www.jungbunzlauer.com/en/products/biogums/xanthan-gum); sodium polyacrylate would collapse in this ionic strength, as the video says.

3.2 Carbon fiber conductivity

The presenter's reference numbers for water (0.6 W/mK), stainless (about 20) and aluminum and copper ("a couple of hundred") are right. "High modulus carbon fiber at 60 or 70 W/mK" is right if the fiber is Toray M40J class (39 to 69 W/mK depending on which Toray document you read, https://www.toraycma.com/wp-content/uploads/M40J-Data-Sheet.pdf). The common cheap chopped tow is standard modulus T300/T700 class at about 10 W/mK, below stainless steel, and an Amazon "high modulus" listing gives no way to tell. "Ultra high modulus up around diamond" is overstated: pitch UHM fibers reach 620 to 800 W/mK, diamond is 1,000 to 2,000.

More important is what 2 wt% of random short fiber does to the bulk. Thermal conductivity in a filled composite does not jump at a percolation threshold the way electrical conductivity does, because the conductivity contrast between fiber and brine is only about 100x, so the useful estimate is the measured linear scaling. Frusteri et al. measured random carbon fiber in a salt-hydrate PCM and found conductivity rising linearly with loading, quadrupling at about 7 wt% (https://www.sciencedirect.com/science/article/abs/pii/S1359431104002984); scaled to 2 wt% that is roughly a 1.8x to 2x gain, from about 0.55 to about 1 W/mK. Real but modest, and the frozen eutectic is already an ice-and-salt composite at about 2 W/mK on its own. The fiber helps mainly during melting, when a liquid layer forms at the wall, and the 1.2 percent xanthan gel takes away the natural convection that would otherwise do the same job. For a commissary the fiber is not worth the black dust and the contamination argument (section 7).

3.3 The 700 kJ claim and the "thousandfold" claim

"Take a kilogram of hot liquid or warm water and cool it off and turn it into ice, you have to remove about 700,000 joules." From the test's own 50 C start it is 4.19 x 50 + 333.5 = 543 kJ; chilling the ice on to -20 C adds 41 kJ. 700 kJ corresponds to starting at about 87 C, which is a pot straight off the stove. The order of magnitude and the conclusion (45 W of spare compressor cannot do this in a useful time) are correct. The aside that melting ice absorbs enough heat to bring the same water "all the way up to boiling" gets to 80 C, not 100 C.

"We can increase that rate of heat transfer a thousandfold." Natural convection in air on a 9 cm pot wall with a 70 K difference is about 7.5 W/m2K; radiation from a shiny pot adds 0.5 to 1.6 W/m2K, more for a dark one. The 16 in fan's claimed 1,000 CFM through its own disc is 3.6 m/s, about 3 m/s at the 9 V operating point, and 2 to 3 m/s by the time it reaches a pan (section 5.3). Air at 2 to 4 m/s over a 16 cm pot gives about 15 to 25 W/m2K (Engineering Toolbox, h = 10.45 - v + 10 v^0.5, gives 23 to 26 W/m2K at 2 to 4 m/s; a cylinder-in-crossflow correlation gives about 17 at 3 m/s; https://www.engineeringtoolbox.com/convective-heat-transfer-d_430.html). That is 2x to 3x on the outside film. The measured 2x chill and 3x freeze are what you get once the water inside the pot, the growing ice shell, the 1 kg of stainless (12 percent of the sensible load) and the warming freezer air all count. The claims are plausible; the mechanism description is exaggerated.

Energy check: each pot test dumps about 460 kJ into the freezer (193 kJ to chill, 267 kJ to freeze). In 30 kg of frozen food at 2 kJ/kgK that is 7.7 K, against the presenter's "five or six degrees," and a 90 W residential compressor clears it in about 70 min of extra run. Consistent.

3.4 Plateau sanity check

A 28 oz bottle at 90 percent holds about 0.80 kg of mix, 200 kJ at 250 kJ/kg. Holding -15 C for 5 h in a 22 C room means 11 W of heat gain. A free-standing textbook estimate (natural convection plus radiation, 9.3 W/m2K over 0.065 m2) gives 23 W and a 2.5 h plateau; three bottles shielding each other on an insulating bench, under a frost layer that forms within minutes, account for the factor of two. The water bottle in the same test computes to a 5.3 h plateau at 0 C, and the video shows water and gel still on their plateau when the eutectic finished. The numbers hang together.

3.5 Recharge, the number that matters most

The eutectic freezes at -15.4 C. To refreeze it the freezer air must sit below that by a margin large enough to move heat at a useful rate. Single free-standing 28 oz bottle, 0.065 m2 of skin, 200 kJ latent plus sensible from 20 C. The temperature that matters is the logged mean air temperature, not the thermostat's cut-out; a box that swings -15 to -18 C averages about -16.5 C. The fan columns assume nucleation happens as soon as the bottle passes -15.4 C, which section 3.1 says is not guaranteed at 1 to 3 K of subcooling.

Bottles packed together are slower: standing shoulder to shoulder in a solid-walled bus tub, only the tops see moving air, and the packed set takes two to three times the single-bottle figure. The video's samples sat "a whole day" in a -20 C freezer, which fits. These are screening estimates. They assume the heat-transfer coefficient in the column header, they assume the bottle nucleates as soon as it passes -15.4 C (section 3.1 says it may not), and the two-to-three-times factor for a packed tub is a judgment, not a measurement. The recharge time that goes on the rack card is the one measured in this freezer, on this rack, on three separate nights (section 6.6 step 1 and section 6.7). The point for this project: the owner reports the working freezers hold -15 to -18 C. At -15 C the NH4Cl goop never refreezes; at a mean of -16.5 C it effectively never comes back in one night; even at a steady -18 C a packed set at 6 V is a two-day job. Section 6.1 deals with this.

The same arithmetic for the owner's 10 lb Arctic Ice XX-L packs (section 6.3, Tier D) is worse per pack because a 3.5 in slab has about 0.20 m2 of skin for 4.54 kg, a third of the skin per kilogram of a 28 oz bottle: from 5 C in still air, 29 to 37 h at -23 C and 38 to 49 h at -21 C, and a fan in the recharge box roughly halves that. That is why the packs get their own chest freezer and a 48 h rule (36 h only at -25 C or colder, or once test (b) in Tier D proves it).

3.6 Claim scorecard

4. Why this matters for a commissary freezer

4.1 What hot food does to the machine

Manufacturers do not treat a hot load as normal holding use. Hoshizaki's reach-in manual, in its warning list and again in its troubleshooting table: "Do not put warm or hot foods in the cabinet. Let them cool first, or they will raise the cabinet temperature and could deteriorate other foods in the cabinet or overload the appliance" (https://hoshizaki.com/docs/manuals/Commercial_S_Series_inst.pdf). Turbo Air's M3 troubleshooting table files "The unit contains hot food or too much food" under "The unit does not refrigerate well", in the same row as "The condenser is clogged" (https://turboairinc.com/wp-content/uploads/2020/01/M3-Reach-Ins_HC.pdf). Avantco's product page answers the question directly: "reach-in refrigerators are intended for keeping cold food chilled, not chilling hot food" (https://www.webstaurantstore.com/avantco-a-23f-hc-29-solid-door-reach-in-freezer/178A23FHC.html, seen 2026-08-30). True's manual has no hot-food line; it asks for the unit to run empty for 24 hours before loading (https://www.webstaurantstore.com/documents/pdf/tseries_manual.pdf). A compressor does pull an empty cabinet down from room temperature at install, but that is one event on an empty box with a clean coil; a 2.2 MJ batch every day on top of a full inventory is a different duty, and no maker on this list rates a holding cabinet for it. What follows is the chain by which a hot load can hurt the compressor. Read it as the list of things to have a refrigeration technician measure during one representative batch: suction pressure, compressor current, discharge line temperature, superheat, condensing temperature, and whether the overload protector opens. Each link is more likely to close with a dirty condenser or a hot room, which is why the condenser comes first, but only the measurement says where this box sits.

Suction pressure climbs out of the design envelope. Tecumseh's service handbook gives the low-temperature compressor band as -40 C to -12 C evaporating and says compressors "operating outside their design evaporator temperature range can be expected to have poor pumping efficiency and experience motor problems" (https://www.tecumseh.com/userfiles/documents/marketing-brochures/tecumseh-service-handbook_na_en.pdf, Table 2-2). Embraco: running above class causes "Overheating of the housing, with the consequent tripping of the thermal protector" and "Oil heating and lubricity loss, leading to increased compressor wear" (https://www.embraco.com/en/blog/technical-information/lbp-mbp-and-hbp-applications.html). A reach-in coil normally runs 8 to 12 K below box air; box air has almost no thermal mass (0.65 m3 is under 1 kg), so with a hot load the air heads for 0 C and the coil for -10 C, at the edge of the envelope.

Motor current rises with suction pressure, because a fixed-speed compressor moves denser gas (Jamie Kitchen, Danfoss, https://www.linkedin.com/pulse/interpreting-current-readings-compressors-jamie-kitchen). Tecumseh's troubleshooting table lists "Suction pressure too high", "Return gas too warm" and "Discharge pressure too high" as causes of "starts and runs, but short cycles on thermal protector."

Discharge and oil temperature rise. True and Turbo Air reach-ins are capillary-tube systems that cannot open up to feed a suddenly hot evaporator; the refrigerant boils off early and the return gas that cools the motor comes back hot. Copeland's limit: 225 F on the discharge line "is desirable for reasonable life expectancy," 250 F "is usually considered a danger level" (https://copeland.custhelp.com/app/answers/detail/a_id/4578/).

If the protector trips mid pull-down, the box is still hot, so the compressor restarts as soon as the protector cools and trips again. Copeland: each start flashes crankcase oil into foam "with the frequent result that a large percentage of the crankcase oil is pumped out of the compressor," and the accumulated cycles "must shorten the life expectancy of both the contactor and the motor" (AE17-1262, https://webapps.copeland.com/online-product-information/Publication/LaunchPDF?Index=AEB&PDF=1262). Small PTC-start compressors need 3 to 10 minutes to cool between attempts; True's manual says wait 5 minutes before restarting.

Summing up: a hot pull-down pushes suction, current and discharge temperature up for 30 to 90 minutes, by an amount this report cannot compute for an unknown compressor and that a technician can read off gauges and a clamp meter in one batch. Tecumseh's short-cycling causes are wiring faults, shorted windings, low voltage, and high suction or discharge pressure, not hot product as such; the high pressures come from a hot load only when the condenser cannot reject the heat. That is why the condenser comes first (section 6.1, step 0): True's manual says condenser coils "require cleaning every 30 days or as needed" and that a dirty coil "can result in non-warranteed repairs and/or cabinet failure" (https://www.webstaurantstore.com/documents/pdf/tseries_manual.pdf); Turbo Air says do not install where the room is over 100 F (43 C). The retrofit removes no heat; at best it changes the conditions under which the heat is removed, and section 4.2 shows how little of that a bottle bank can do.

Moisture is the second cost, and it is a frost and defrost cost, not a big coil-load cost. How much an uncovered pan evaporates is not published for these foods. r5's rough estimate for 60 C stock under a fan is 1.5 to 3 percent of its mass, 150 to 300 g per 10 kg, and it will differ for rice, a braise or a sauce that skins over. Measure it: weigh each pan on the bench scale going in and coming out, log the difference with the item, the depth and whether it was covered, and after a week of batches the number for this kitchen is in the log. The latent heat that vapour carries (2,358 kJ/kg at 60 C) comes out of the food, which is why uncovered pans chill faster; when the vapour deposits on the coil as frost it hands back that same latent heat plus the fusion and ice-subcooling part, about 400 to 630 kJ per kg evaporated. Energy balance for 10 kg from 60 C to 5 C, worked at an assumed 300 g evaporated: the food gives up 2,090 kJ either way, 707 kJ of it as vapour and 1,383 kJ by convection; depositing the vapour to ice at -25 C releases 898 kJ; the sinks see 2,281 kJ, an extra 190 kJ or 9 percent (about 5 percent at 150 g). The real price of that frost is elsewhere: every extra defrost is 0.4 to 1.1 MJ of heater energy dumped into the box, and a furred coil loses airflow and capacity until it is cleared. Hoshizaki's reach-in manual: "Do not put warm or hot foods in the cabinet. Let them cool first, or they will raise the cabinet temperature and could deteriorate other foods in the cabinet or overload the appliance" (https://secure.hoshizakiamerica.com/docs/manuals/Commercial_S_Series_inst.pdf). Clock-initiated defrost (True mechanical timers, Traulsen six per day, Turbo Air every 6 h) does not bring the next defrost forward when frost builds faster, so ice can build day over day; run-time-initiated units (Hoshizaki, Beverage-Air) self-correct. Reach-in defrost heaters are several hundred watts (the one True replacement listing found is a 950 W, 220 V element for a large multi-door cabinet, https://www.amazon.com/dp/B06XQKVD4C, $509.49 on 2026-08-30, which says nothing about the 115 V single-door T-23F; its spec sheet gives no heater wattage, only "time-initiated, temperature-terminated" defrost, https://d3ld6frh4bdurh.cloudfront.net/pdf/598-T23F.pdf). At 500 W and 20 minutes an event is 0.6 MJ. The box runs its scheduled defrosts whether or not it chills food, so the cost that belongs to the hot pans is the increment: a temperature-terminated defrost that runs 10 minutes longer is about 0.3 MJ, and an extra defrost brought forward on a run-time-initiated unit is a whole event. Log the defrost length (display "dEF" on to off) on chill days and non-chill days for a week and charge the difference, not the daily total, to the retrofit; that heat goes into the box and the compressor removes it afterwards.

The third cost is the inventory. Frozen food at 1.8 to 2.3 kJ/kgK means 50 kg absorbs about 100 kJ per K. A 10 kg batch is 2,200 kJ, so even with the compressor taking half, the surfaces of the stock see 10 K or more. Codex says quick-frozen food should never be warmer than -12 C in the warmest package (CXC 8-1976, https://www.fao.org/input/download/standards/285/CXP_008e.pdf); the IIR estimates storage at -15 C instead of -18 C cuts shelf life by about 30 percent (https://iifiir.org/en/news/iir-position-paper-in-ijr-frozen-at-18-c-or-15-c-a-critical-review-of-subzero-storage-quality-and-regulation). Temperature swings drive recrystallization, which is a one-way texture loss.

4.2 The numbers

Design loads: food at 60 C to 5 C, cp 3.8 kJ/kgK (ASHRAE Refrigeration Handbook ch. 8, https://consultancyengineering.com/wp-content/uploads/2023/12/CH-08-Thermal-Properties-of-Foods.pdf), plus 5 percent for pans.

The run times are illustrative. Tecumseh's Quick Select Guide lists, at its -10 F low-temperature rating point, a 1/3 HP AE2413 at 1,300 Btu/h (381 W), a 1/2 HP AJA2425 at 2,200 Btu/h (645 W) and a 3/4 HP AHA2435 at 3,400 Btu/h (996 W) (https://www.sidharvey.com/documents/Tecumseh-Catalog_Broc.pdf); r5 took a 1/2 HP R404A AJA2419Z at 1,865 Btu/h (547 W). Horsepower does not fix capacity: the same nameplate HP spans 1,300 to 2,200 Btu/h inside one guide, and what the owner's box delivers depends on its refrigerant, evaporating and condensing temperature, return-gas temperature and cabinet airflow, none of which are known. The Avantco CF8 is 1/7 HP at 101 W input and publishes no cooling capacity (https://www.webstaurantstore.com/documents/specsheets/178cf8_specsheet.pdf; its Webstaurant product page returned 404 on 2026-08-30, so the model may be superseded by the Galaxy CF line, section 6.3); if it delivers about 150 W, a 10 kg batch is about 4 h of continuous running and a 30 kg batch about 12 h (6,600 kJ / 0.15 kW = 44,000 s), before the cabinet's own baseline load, so 30 kg/day is not a duty for that class of chest freezer. An earlier draft's line about a 1/3 HP chest freezer at 70 percent mixed the two sizes and is withdrawn. On a 1/2 HP class reach-in, 10 kg/day adds roughly 5 to 7 percent duty if its capacity is near the guide's figure; that "if" is the point, and the way to remove it is to measure.

Measure it with a plug-in kWh meter between the freezer plug and the wall (P3 Kill A Watt P4400: cumulative kWh, 0.2 percent accuracy per P3, https://www.p3international.com/products/p4400.html; $29.99 and in stock at https://www.circuitspecialists.com/p4400_kill_a_watt_energy_meter on 2026-08-30; check the meter's amp rating against the freezer nameplate before plugging a freezer through it). Baseline: 24 h with no hot loads and normal door traffic, read kWh, divide by 24 for baseline average watts. Batch: note the kWh when the pans go in, read it again when the box is back at setpoint and the compressor has returned to its normal cycle, and subtract baseline watts times the elapsed hours. What is left, times 3,600, is the kJ of electricity the batch cost; the 2,200 kJ heat load divided by that is the box's effective COP at this duty, and the elapsed hours are the recovery time. Also read the watts every 5 minutes during the chill: a compressor that runs without a break for more than about 90 minutes on a 10 kg batch, or that cycles off and on within minutes, is the signal to stop and get a technician's readings (section 4.1). Two batches on different days bound the answer better than one. Those hours are the same with or without the retrofit; what changes is their character. Freezing the food in the same box is a second load larger than the chill, and it lands in the same overnight hours the PCM needs for recharge; on a freezing night the bank does not come back, and the daily-duty figures above for the chill alone do not apply.

The fan is a heater. The 16 in fan at 9 V dumps about 50 W into the box (cube law from the 120 W nameplate; a brushed motor does not follow the cube law exactly, and at the report's 6 A running limit the input is 54 W, so treat 50 W as an estimate). The 6 V figure of about 15 W has no measured current behind it yet. The validation test measures amps at 6 V and 9 V, running and at cold start; put those watts into this paragraph and into the fuse choice before the SOP is posted. For a 90 min chill that is 270 kJ, 12 percent of the 10 kg load. Left at 9 V for 24 h it is 4.3 MJ, twice the entire 10 kg/day food load, so 9 V is for the chill only. The 6 V recharge setting is about 15 W: 0.65 MJ over a 12 h night, about 30 percent of a 10 kg batch or 10 percent of a 30 kg batch, added to the compressor at its coldest, least efficient suction, and it warms the very air it is meant to freeze the jugs with. That is the price of an overnight recharge with a brushed radiator fan; two rated Sanyo fans idling at 30 to 40 percent PWM (section 6.3a; a few watts each, unmeasured) make the same night a fraction of that. One rule, stated once: overnight fan only on nights jugs were used, at 6 V, never 9 V.

The PCM does not cap the box air at the bottle counts in section 6.2. Sizing on stored kJ says nothing about the rate at which the bottles can take heat, which is UA x (T_air - T_melt). Thirteen 28 oz bottles have 0.85 m2 of skin; packed in a bus tub perhaps 60 percent of it sees air, so at the 9 V fan-on h of 25 W/m2K the bank's UA is 8 to 21 W/K, and at 3 to 5 K above its plateau it absorbs 25 to 105 W. A 10 kg batch releases 400 to 700 W at the start (2,200 kJ over 1 to 1.5 h, front-loaded). The rest goes where it goes today: to the coil and the frozen inventory, whose 50 kg of bags expose about 2.5 m2. A simple sink balance with the report's own numbers (food UA about 10 W/K at 60 C, PCM UA 12 W/K at -15.4 C, inventory UA 25 W/K at -18 C, coil capped at a 1/2 HP compressor's 550 W) settles the air near -12 C with the PCM taking about 35 W, 5 percent of the load, and the compressor the rest. Put a weak 150 W compressor in the same balance and the air goes to about -4 C with the PCM taking 140 W, 20 percent. So the -12 to -10 C figure in an earlier draft was where the compressor and inventory put the air, not a cap set by the PCM, and the earlier claim that the PCM keeps the evaporator inside its envelope is withdrawn. To actually hold air near -10 C against 60 C food you need about 3 to 4 m2 of PCM surface in the fan stream, 40 to 50 free-standing bottles on open racks, or flat GN eutectic plates. What the bank does do at 11 to 13 bottles: it stores 2.2 to 3 MJ of latent capacity at -15.4 C that the box pays back overnight, so the frozen inventory absorbs a smaller share of the batch and its excursion is shallower. Conduction would move several times the heat per unit area that an air film does, but setting hotel pans on frozen packs is not a validated method and this report does not recommend it. Engel rates its packs for keeping cooler contents frozen and publishes no static load, contact temperature or puncture rating (https://engelcoolers.com/products/engel-5-f-15-c-freezer-packs, seen 2026-08-30); the U-tek pouches are flexible shipping refrigerants, not pan supports. A -15 C surface under a steel pan can also freeze a layer at the pan bottom while the center is still warm, which a single center probe misses. If conduction is ever tested, it is with a rigid cold plate made for foodservice (GN eutectic plates), with the plate maker's written approval of the contact temperature and load, and with top, center and bottom food temperatures logged in the first batch.

4.3 Verdict on equipment life

Not proven for the compressor. Section 4.2 shows a bus tub of bottles takes 5 to 20 percent of a chill's heat; it does not hold the evaporator in its envelope. What does protect the compressor is a clean condenser, a cool room, a healthy charge and no short cycling, and those are checks, not purchases. The retrofit's honest benefits are faster chilling (the fan) and a shallower inventory excursion (the PCM), both bought with 10 to 30 percent more compressor minutes overnight. "Helps the compressor" would need a technician's readings during a real batch (section 4.1) and a kWh comparison (section 4.2), not this report. The conditions for the benefits it does have: PCM latent heat at least equal to the day's load and exposed area per the rate rule in section 6.2, sentinels below the plateau before the first batch; fan at 9 V on a timer for the chill, 6 V only on recharge nights; fan and bottles clear of the evaporator's air path; 2 in pans, loosely covered; chills timed away from the defrost clock (section 6.5).

No for the evaporator, the frost and the defrost system. The PCM does nothing there. Covering pans and using shallow pans is the lever, and it is free.

Worse than today if: the bottles are soft and the batch goes in anyway (now the box carries 12 to 35 kg of liquid to refreeze on top of the food); the fan is left at 9 V; the bottles or fan block the coil discharge; the cable entry leaks air or grows frost; a bottle leaks brine onto aluminum fins; or any add-on electrical device is run inside an R290 or R600a cabinet without the maker's written approval (sections 6.1 and 7). The honest caveat: the evidence that excursions age compressors more than rated hours is application-note grade, not a controlled life test of this retrofit.

5. Commercial options, buy not build

5.1 Blast chillers

Capacities are the maker's "lb per 90 min chill / lb per 240 min freeze" and start temperatures differ by maker (Avantco and Beverage-Air rate from 194 F, Thermo-Kool and American Panel from 160 to 165 F, Traulsen CLBC from 135 F over 4 h), so they are not strictly comparable. All prices seen 2026-08-30. Where a reseller and the maker disagree, the maker's current spec sheet is used and the reseller figure is noted; buy on a written quote that names the model, the capacity basis and the warranty term.

Used market, same date:

Used caveats: every US unit older than about 2024 is R-404A, legal to run and service but the refrigerant is under the AIM Act phasedown (https://www.epa.gov/hfcs/technology-transitions-hfc-restrictions-sector); no warranty; probes are dealer wear parts; the Thermo-Kool and Beverage-Air countertop lines are discontinued.

What a real blast chiller does that the retrofit cannot: staged air temperature (Irinox hard chill runs air at -20 C until the core hits +15 C then -1 C to +3 C, so the surface never freezes; https://caterlink-web.s3.ap-southeast-2.amazonaws.com/public/partdocs/Irinox-EasyFresh-Blast-Chiller-User-Manual.pdf), a probe-controlled end of cycle, automatic switch to hold mode, scheduled defrost around cycles, HACCP export on the branded units, and real refrigeration capacity: 90 to 150 BTU/h per lb of 90 min capacity, 60 to 100 W of cooling per kg. Leasing runs $20 to $30 per $1,000 per month on 48 to 60 months (https://www.crestmontcapital.com/blog/commercial-kitchen-equipment-leasing-a-smart-financing-strategy-for-foodservice-businesses), so a $3,789 SF-5 is roughly $75 to $115 a month.

5.2 Phase change products

Recharge rule first, because it rules out most of the catalogue: the freezer must hold several kelvin below the PCM's melting point for many hours. Arctic Ice says freeze its -15 C Tundra pack "below -5 F" (-20.5 C) for 24 to 36 h (https://arctic-ice.com/products/tundra-series).

Cost per kJ, for scale: DIY NH4Cl at 21 wt% is $0.005 to $0.008 per kJ for the salt alone (20 lb in 5 lb bags, grade not stated, $56.99 or $6.28/kg in stock, https://www.amazon.com/dp/B00L2158N0; Wintersun 50 lb feed grade $250.00 or $11.02/kg in stock, https://www.amazon.com/dp/B00N48Q7XY; the 1 kg food-additive listing at https://www.amazon.com/dp/B01EZ921NA read "Currently unavailable" on 2026-08-30 and its $19.19 to $22.79 strings are stale offers, not a price) and about $0.012 to $0.020 all in with bottles and gum; Engel packs about $0.044; Fisher bricks $0.21. Buying is 2 to 4x the cost of mixing, and it buys a rigid, labeled, drop-rated container and no chemistry in the prep area.

5.2.1 Rigid low-temperature packs (the Tier D question)

If the swap-bank scheme goes ahead, the packs in the working freezer will get hot pans set near them, be handled twice a day and be wiped down; a rigid or shape-stable pack is easier to rack, clean and inspect than a pouch. The owner's Arctic Ice XX-L packs are rigid and -15 C. The question for the colder variant is whether any US supplier sells a rigid -20 to -25 C pack with both a published latent heat and a public price. On 2026-08-30 the answer is no. What exists:

Nordic Cold Chain, Pelican BioThermal, Cold Chain Technologies and TCP were checked the same day; none lists a -20 C class rigid pack with a public price or a latent heat figure. ThermoSafe's own PCM brick and bottle lines (TPB and B series) are +4 C and +20 C on the maker's page.

What that means for the tiers. A rigid -21 C bottle with a public price exists (PureTemp -21 at Fisher) but it costs five to six times a U-tek pouch per kilogram, its latent heat is not on the page, and like the U-tek it needs a recharge box at -26 C or colder under the 5 K rule and self-discharges in a -15 to -18 C working box. The only frozen-range rigid product found with a published latent heat is Temprecision's TEMP-16 (333 J/g at -15.6 C), which is a -15 C class material: it fits the same slot as the owner's Arctic Ice packs and the Engel packs in Tier C, with better data and no price. Before any further pack purchase: get the latent heat curve (DSC), the conditioning temperature and time, and the static load rating in writing from the maker, then buy one case and run the section 6.7 test on it. Do not size a bank from a shipping refrigerant's mass alone.

5.3 Fans and paddles

Fans with an actual environmental rating, versus the video's fan:

An earlier draft offered a Noctua array for hydrocarbon cabinets; withdrawn, because the True prohibition covers any electrical appliance, not only brushed motors (section 6.1).

What "rated" buys and does not buy. No add-on fan sold today carries an NSF food-zone listing or a reach-in maker's approval for retrofit; every fan in this report, rated or not, makes the cabinet an unlisted, altered appliance (section 6.1). The Sanyo 9WV1212P1J001 is the closest thing to a fit part: sealed to IP68, rated to -20 C by its maker, brushless with its own PWM input, 12 V at 3 A so a listed 60 W brick runs one and a 102 W brick runs two. Two of them move about 450 CFM for 72 W at full speed, or about 300 CFM for roughly 25 W at 70 percent PWM (cube law, unmeasured), which is the GESEXI 16 in at 9 V on paper with a temperature rating and a seal. Its -20 C rating is 3 K warmer than a -23 C setpoint; during a chill the box air is -12 C or warmer, so the fan runs inside its rating and only idles outside it, but that is the honest gap. Delta's IP68 120 mm fans are rated -10 to +60 C and do not qualify. SPAL's brushless line is IP68 and IP6K9K but starts at 130 mm and 20 A and has no confirmed US price; ebm-papst's 4114N/2H8P is 24 V, 120 W and 78 dB(A). The GESEXI and SPAL brushed fans stay in the report as the video copy for a dedicated cabinet the owner accepts as altered; they are not the first purchase for a cabinet holding inventory.

Power for the 12 V side: Mean Well GST series, UL 62368-1 listed, -30 to +70 C working temperature, hiccup-mode overload protection (spec sheets https://www.meanwell.com/Upload/PDF/GST60A/GST60A-SPEC.PDF and https://www.meanwell.com/Upload/PDF/GST120A/GST120A-SPEC.PDF, seen 2026-08-30). GST60A12-P1J, 12 V 5 A 60 W, $24.86 at https://www.trcelectronics.com/products/mean-well-gst60a12-p1j (stock "Contact Us"; DigiKey listed $18.60 on backorder in search results, page not reachable), enough for one rated 3 A fan or the 12 in GESEXI at 9 V, not for the 16 in fan. GST120A12-R7B, 12 V 8.5 A 102 W, $57.27 at https://www.trcelectronics.com/products/mean-well-gst120a12-r7b ($52.05 listed at Wellforces in search results), enough for two rated fans or the 12 in GESEXI at any setting; the 16 in GESEXI (10 A nameplate) exceeds it at 12 V and its brushed-motor start surge can trip the 8.9 to 12.75 A hiccup point, so bench-test the start before relying on it. The R7B ends in a 4-pin DIN plug on a 16 AWG x 4-conductor UL2464 lead (two conductors per pole); cut the plug off and land the pairs on the enclosure terminals, or buy the mating pigtail. Neither supply is marked Class 2 (section 7). The Jesverty SPS-3010 bench supply and the SHNITPWR 12 V 10 A brick are withdrawn: no listing, and the brick's 5.5 x 2.5 mm barrel plug is a 5 A connector. The RioRand 10 A PWM ($9.99, https://www.amazon.com/dp/B007TH4EN6) stays for brushed fans only; never chop the supply of a brushless fan with it, use the fan's own PWM lead.

A reach-in's own evaporator fan moves about 400 to 560 CFM through a back-wall plenum, but cabinet studies put shelf-level velocity mostly under 0.2 m/s away from the discharge (https://www.sciencedirect.com/science/article/abs/pii/S0140700714000334). A blast chiller (Alto-Shaam QC3-40) moves 1,750 CFM through its cavity, 1.3 to 2.2 m/s (https://cdn.beedash.com/AltoShaam/3042ef891b127d41b813f388d0e8c998af3afca9.pdf), and design guides call for 1.5 to 3 m/s over packaged product (https://ingener.by/refrigeration-systems/food-processing-refrigeration/frozen-food-processing/blast-freezers/air-velocity-requirements/). The booster at 9 V gets 2 to 3 m/s over a pan 6 to 18 in away. The stock evaporator fan stops whenever the door opens (Turbo Air M3 manual, https://turboairinc.com/wp-content/uploads/2020/01/M3-Reach-Ins_HC.pdf); the booster should do the same.

Cooling paddles are what commissaries use today and they put the cold inside the pot, where a fan cannot reach: San Jamar Rapi-Kool 4 L (128 oz) $54.99, 2-pack $109.99; Vigor Polar Paddle 2 L $28.49; Vollrath Insta Chill 4 L 2-pack $106.99 (https://www.webstaurantstore.com/3071/cooling-paddles.html). A 4 L paddle of ice, if it melts completely and warms to 5 C, absorbs about 1.6 MJ (4 kg x (36 + 334 + 21) kJ/kg), up to about 70 percent of a 10 kg batch's 2.2 MJ. How much of that it takes in 30 minutes depends on the fill, the stirring, the pot, and the product's viscosity, and nobody has measured it for this kitchen. The paddle's 1.6 MJ also has to be refrozen somewhere; in the same freezer that is the same overnight load as PCM, only at 0 C, where any freezer can do it. So treat the paddle as its own measured step: log the batch temperature at 0 and 30 minutes with the paddle in, compute the heat removed (mass x 3.8 kJ/kgK x drop), and reduce the PCM count for that batch type by the measured amount only. Recharge the paddle in a box whose night budget includes it. No foodservice supplier sells a retrofit blast-chill fan for reach-ins; the market gap is real and the video's approach is the retrofit.

6. The home-made version

6.1 Three things to do before ordering anything

Step 0, read the nameplate and the manual, then service the box. A commercial freezer holding -15 to -18 C (5 to 0 F) with no load is at the warm end of what these cabinets are rated for, not proof of a fault. Rated ranges seen 2026-08-30: True T-23F-HC "maintains -10 F (-23.3 C)" (https://www.truemfg.com/products/t-23f-hc); Turbo Air M3F24-1-N -10 to 0 F, -23.3 to -17.8 C (https://www.turboairinc.com/products/m3f24-1-n/); Atosa MBF8501GR -8 to 0 F, -22.2 to -17.8 C (https://www.equippers.com/atosa-mbf8501gr-19-1-cu-ft-1-door-bottom-mount-reach-in-freezer-27-w); Avantco A-19F-HC and A-23F-HC -8 to -1 F, -22.2 to -18.3 C (https://www.webstaurantstore.com/avantco-a-23f-hc-29-solid-door-reach-in-freezer/178A23FHC.html); Beverage-Air HBF23HC-1 "down to -10 F" (https://beverage-air.com/HBF23HC-1-Horizon-Bottom-Mount-Solid-Door-Reach-In-Freezer); Arctic Air AWF25 -7 to 10 F, -21.7 to -12.2 C on Webstaurant, -10 to 10 F on other reseller pages (https://www.webstaurantstore.com/arctic-air-awf25-white-single-door-reach-in-freezer-25-cu-ft/123AWF25.html). Chest freezers: Atosa MWF9016GR -10 to -1 F, -23.3 to -18.3 C (https://www.equippers.com/atosa-mwf9016gr-15-9-cu-ft-solid-top-chest-freezer-60-1-8-w); Avantco CF8 -13 to 0 F, -25 to -17.8 C, manual dial (https://www.webstaurantstore.com/documents/specsheets/178cf8_specsheet.pdf). A box cycling -15 to -18 C has its cut-out at the warm edge of most of those bands and its cut-in 2 to 3 K above them; an Arctic Air sits inside its band the whole time. A stable 3 K swing usually reflects the controller differential and the setpoint someone chose, not a failing machine. On a True T-23F-HC the same log would be 5 to 8 K warmer than the model's stated holding temperature and would justify a service call. So, in order: find the model on the nameplate, read its rated range and setpoint procedure, log the air and a product simulant (a 500 mL water bottle among the inventory) through several compressor cycles, and compare the logged mean with that model's range before diagnosing. If it is warm for its model, the causes to check are a dirty condenser, a hot location, an iced evaporator, a failed gasket, low charge, and the setpoint. Before buying anything: clean the condenser coil (True: "require cleaning every 30 days or as needed"; a dirty coil "can result in non-warranteed repairs and/or cabinet failure", https://www.webstaurantstore.com/documents/pdf/tseries_manual.pdf), check the door gaskets with a dollar bill all the way round, look at the evaporator for ice, confirm the setpoint, measure the room air at the condenser intake (Turbo Air: not above 100 F / 43 C, https://turboairinc.com/wp-content/uploads/2020/01/M3-Reach-Ins_HC.pdf), then re-log for 24 h. A box whose logged mean sits at the warm edge of its rated band has little reserve for hot loads; whether that is a service problem or a setpoint choice is what the manual and the log settle, and it changes nothing about the PCM arithmetic below, which runs on the logged mean. Do not turn the thermostat to -23 C on a box that cannot get there; that makes the compressor run 100 percent indefinitely, which in a hot commissary with a fouled condenser is the real compressor killer. Monthly condenser cleaning goes in the SOP; it does more for freezer life than anything else in this report.

Cabinet temperature, logged for 24 h with a probe in the air. A min/max fridge thermometer or a USB data logger does this for the one-off test; the YoLink YS8004 probe sensor is $35.49 (https://shop.yosmart.com/products/ys8004) but its page says "a YoLink Hub or SpeakerHub is required", which is not in that price, so it is not a standalone option. The owner reports -15 to -18 C. What matters is the logged mean, not the cut-out, because that swing reads like a thermostat band (cut-in -15, cut-out -18) with a mean near -16.5 C. That is the single most important number in this report:

• At -15 C the NH4Cl eutectic (-15.4 C) has no driving force and never refreezes. Even sitting idle it slowly melts whenever the compressor is on its off-cycle and the air drifts above -15.4 C.
• At a mean of -16.5 C (a box cycling -15 to -18 C) a single free-standing bottle needs about 55 h at the 6 V recharge setting and about 100 h in still air, and it partly melts during every cut-in period above -15.4 C (section 3.5); a packed tub is two to three times slower. Not viable with one set, or with two. Even a box that holds a steady -18 C is about 22 h per free-standing bottle at 6 V, a two-day job for a packed set. An NH4Cl bank needs a box whose mean air is -20 C or colder.
• A -23 C PCM (U-tek, PureTemp -21, Rubitherm SP-21) cannot be stored in this box. It will not freeze there, and if frozen elsewhere and left there it self-discharges, because the box is 5 to 8 K warmer than its melting point: a 48 oz pack standing on edge in still air has a UA of about 1.06 W/K (0.133 m2 at 8 W/m2K, the same figure the U-tek recharge table in Tier D uses) and loses 5.3 to 8.5 W, so 340 kJ of latent heat is gone in 11 to 18 h, half a day to a day, and in 4 to 10 h with the fan running; packs lying flat in a stack expose one face each, about 0.5 W/K, and last about twice as long. Used as a swap bank (recharge freezer to batch to recharge freezer, never stored in the working box) that same 5 to 8 K is the working margin, and the pack absorbs heat from the first second of a chill. That scheme is Tier D in section 6.3, where the U-tek is the colder variant of the owner's Arctic Ice packs.
• If, after step 0, the thermostat will hold -23 C (True T-23F-HC is rated for -10 F; Avantco CF8 chest is rated -25 to -18 C, https://www.webstaurantstore.com/documents/specsheets/178cf8_specsheet.pdf), set it there for the trial. If it will not, the PCM half of the project needs a dedicated recharge freezer for a -15 C PCM, and that is the default for these boxes, not the fallback. A warmer PCM (KCl at -10.7 C, Tier B) recharges in them but is only for a cabinet that holds no frozen inventory during the chill (section 6.3). The Avantco CF8's product page returned 404 on 2026-08-30 (its spec sheet still loads); the current Webstaurant equivalents are the Galaxy CF7 and Ultra Cold UCCF8, priced in Tier D.

The swap-bank scheme the owner raised (recharge spent packs in a separate dedicated chest freezer, rotate sets) is physically sound and is how blast-chill trucks and cold-plate carts work. The owner has since bought the packs for it: three Arctic Ice Tundra XX-L, 10 lb each, -15 C. Two rules make the scheme work. The recharge box must hold at least 5 K below the pack's melting point at its warmest spot around the clock: -21 C or colder for the -15 C packs (Tier D's rule for the Arctic Ice packs is 48 h in a still-air box logging -21 C or colder, 36 h only at -25 C or colder or once test (b) proves it; take a logged mean of -24 C or colder at the pack position for Engel packs, whose reviewer reports slush at -20.6 C), which every commercial chest freezer rated to -23 C or colder can do at its coldest setting, and -28 C or colder for a -23 C PCM, which no ordinary commercial chest freezer is rated for but a -40 F household chest, the Galaxy Ultra Cold UCCF8 or a -40 C lab chest is (Tier D lists them, $288 to $6,555 on 2026-08-30). And the packs go into the working box only for the chill and come out with the batch; they are never stored there, because the working box is at or above their melting point for half of every cycle and cannot put the charge back. Melting in the working box during a chill is the intended discharge, not a fault (the first Codex red-team finding), so the working box's temperature does not constrain the PCM choice. Under that rule the -23 C U-tek is the higher-rate material, because during a chill the working box's air is 11 to 23 K above its plateau and only 3 to 15 K above a -15 C pack's; the price is the colder recharge box. With the -15 C packs already in hand, the U-tek is the upgrade to consider after the Tundra packs have been tested, not a purchase to make now. The cost of the scheme is the second freezer, priced in Tier D.

Refrigerant on the nameplate. Current True and Turbo Air reach-ins and most new chest freezers are R290 (propane) or R600a; the Atosa MWF9016GR used as the chest example in section 6.5 is R290 (3.2 oz) and its smaller siblings are R600a (Atosa manual, https://www.atosausa.com/manuals/Manual_MWF9007_MWF9010GR_MWF9013GR_MWF9016GR_Atosa.pdf, seen 2026-08-30). True's current T-Series manual (TEC_TM_121 REV. D, https://www.truemfg.com/wp-content/uploads/2023/12/T_SERIES_ENGLISH.pdf, seen 2026-08-30) says on page 5 "DO NOT modify or alter the appliance. Improper alterations can result in electric shock, personal injury, fire, or death" and "DO NOT use electrical appliances inside the food storage compartments of the appliance unless the appliances are approved by the manufacturer", and repeats the second sentence inside the page 6 flammable-refrigerant DANGER block. That sentence covers every fan in this report. A brushless motor does not spark at the brushes, but IP67 is a water rating, not an ignition-protection rating, and the controller, connectors, relay and cable are not ignition-safe either. Rule: no add-on electrical device of any kind goes inside an R290 or R600a cabinet without the maker's written approval. The DIY fan in this report is for a cabinet on R404A, R134a, R448A or R449A, or for a dedicated cabinet the owner accepts as altered, out of warranty and out of its listing. Read the nameplate before ordering anything; if it says R290 or R600a and the maker will not approve a fan in writing, the fan half of this project stops here and the choice is a blast chiller (section 5.1) or the swap bank without a fan.

6.2 Sizing rule of thumb

Heat per batch, cp 3.8 kJ/kgK plus 5 percent for pans, to 5 C:

The freezing row is there so nobody sizes the PCM or the compressor night on the chill row alone; it is not a PCM design case.

PCM mass = batch kJ x 1.3 margin / latent heat. The margin covers pan moisture, fan heat, door openings and the last 10 percent of a bottle that melts slowly.

Two rules, and the second is the one that bites. Energy: 1 kg of -15 C PCM per 1 kg of hot food at 60 to 70 C, margin included; that is the table above. Rate: the bank absorbs only area x h x (air minus plateau), so the exposed area sets how much of the chill it takes while the chill is happening. Required area at the booster's h of about 25 W/m2K:

Area per container, all faces: 28 oz bottle 0.065 m2; 1 L wide-mouth lab bottle about 0.07 m2; Engel XL 0.175 m2; Arctic Ice XX-L about 0.20 m2 effective; U-tek 427 pouch 0.133 m2. Those figures count only if every container stands on an open wire rack with 1 in gaps in the fan stream; in a solid tub about 60 percent of the skin sees air, and stacked flat, one face. So 13 lab bottles spaced are 0.9 m2 and take about 70 W on a healthy box; 39 spaced bottles (three shelves) reach the 2.7 m2 that takes a third; the 40 to 50 bottles in section 4.2 is the same arithmetic. Twelve -23 C U-tek packs on edge are 1.6 m2 and take 260 to 720 W (Tier D). Buying more kilograms without more exposed area slows the recharge and does nothing for the peak. Staff version: the container count on the rack card is the count that passed the validation run for that batch size (section 6.7), every container on the wire rack with a gap, none in the tub during a chill. Do not size for freezing the food; freezing adds about 320 kJ/kg, more than the chill itself (section 4.2 table), and if the food freezes in this box the PCM does not recharge that night.

That rule sizes energy. For the owner's three XX-L packs energy is not the limit; skin is. Three packs hold 3.0 to 4.1 MJ (220 to 300 J/g), more than a 10 kg batch needs with margin (2.9 MJ at 60 C, 3.4 MJ at 70 C), but their 0.60 m2 of skin in the fan stream takes only 90 to 225 W at the start of a chill and 0.4 to 1.0 MJ over 90 min. The packs come out of a 10 kg batch 20 to 45 percent spent. Section 6.3, Tier D, has the full table. The staff version for Tier D: all three packs on the shelf for any batch up to 15 kg; do not add pans to reach 30 kg on the strength of the packs, because at 30 kg they move 6 to 15 percent of the heat and the inventory does the rest.

Per container at 90 percent fill: 28 oz PET Powerade bottle 0.80 kg (232 kJ corrected, 200 kJ video recipe); 32 oz HDPE jug 0.90 kg (261 / 225 kJ); 1,000 mL lab bottle at 900 mL 0.95 kg (277 / 239 kJ); 1/6 hotel pan 4 in deep 1.5 kg (435 / 375 kJ). A 10 kg batch needs 9.9 kg of corrected PCM, 11 lab bottles at 954 g each (12 to 13 with the video recipe). That count is a capacity figure; section 4.2 shows it absorbs only 5 to 20 percent of the chill as it happens and pays back overnight. It does not cap the air. Sized on energy alone it takes 5 to 20 percent of the chill as it happens; the rate rule above says why, and the fix is area in the air stream, not mass in a tub.

Batches per day are limited by recharge, not by chilling. Bottles recharged in a bus tub at 6 V take roughly two to three times the single-bottle figures in section 3.5 (an estimate until measured), so the working rule is one batch per set per day, recharged overnight on an open wire shelf, on a night when nothing is freezing in the box, and the second batch gated on step 1 of the SOP: validated conditioning time, both sentinels below the plateau, no soft container. On a 1/2 HP reach-in held at -23 C that is one 10 kg batch a day on one set or two with two sets; on a 1/3 HP box one; on a 1/7 HP chest freezer one, and only with the fan overnight; on a box whose mean air is -18 C or warmer, the NH4Cl bank does not come back overnight and needs a recharge freezer (the KCl brew recharges there, but only belongs in a cabinet with no frozen inventory). At 30 kg per batch only the 1/2 HP or larger reach-in at -23 C manages one a day, with two sets.

6.3 Four build tiers

Prices are for the 10 kg batch case; the 30 kg delta follows each table. Items common to all tiers are listed in Tier A. Tier D, after Tier C, is the default now that the owner has bought packs; A to C remain as fallbacks. Every total is a line-item sum before shipping and sales tax (NYC 8.875 percent applies to most lines). "Placeholder" lines are guesses with no price seen. There is no total for an R290 or R600a cabinet, because no DIY fan goes into one (section 6.1); the BOM totals table after Tier D separates firm lines from placeholders.

Tier A, the video copy: radiator fan, a listed 12 V supply in place of the video's bench PSU, home-made NH4Cl goop, plus what a commercial box needs (door interlock with a relay to carry the fan current, fuse, timer, secondary tray, labels). One departure from the video in every DIY tier: no beverage containers. Rinsed Powerade bottles and milk jugs full of black brine in a food freezer look like something to drink, and freezer-tape labels frost over and fall off at -20 C. Use amber wide-mouth HDPE lab reagent bottles with a molded label panel, engraved or paint-marked "COOLANT, NOT FOOD", in a color-coded tub. Beverage containers are acceptable only in a dedicated non-food chest freezer.

30 kg delta: four 10 kg batches (about 42 bottles at 954 g, four 12-packs), a second 20 lb bag of NH4Cl ($56.99; two 20 lb bags are cheaper than the Wintersun 50 lb at $250.00), a second pound of fiber, the 5 lb xanthan pail at $60.49, a second bus tub. About $652.

Verdict: the closest copy, and no longer the cheapest, because labeled lab bottles replace free Powerade bottles. Its weaknesses are the salt ratio (15 percent of latent heat wasted), a brushed fan that cannot go in an R290 or R600a cabinet at all, and a chemical mixing step that needs its own tools and a regulator's answer before the bottles can share a freezer with food (sections 6.4 and 7). The open-vented bench supply of the video is gone; the listed Mean Well supply replaces it, and the RioRand PWM gives back the two settings (9 V chill, 6 V recharge) that the bench supply's knob used to provide.

Tier B, the FCC-grade brew, for a dedicated chilling cabinet only: the rated stack of 6.3a (or Tier A's video-copy fan on a dedicated altered cabinet), a salt sold to the Food Chemicals Codex (FCC) specification in the same labeled HDPE lab bottles, no fiber, bentonite kept, xanthan stays. The grade matters for what the salt is, not for what happens if it leaks: GRAS status under 21 CFR 184.1622 attaches to potassium chloride that meets the FCC specification and is made under food GMP. USP is a pharmacopeial grade, a different document, and a listing that says "USP" has not shown FCC compliance; require an identified manufacturer, a lot certificate of analysis, an SDS and a stated FCC claim. None of that makes leaked coolant acceptable on food. A bottle also holds xanthan, bentonite and whatever the container sheds, in unknown amounts, so food touched by any leaked coolant is discarded (SOP step 13), whatever the salt's grade. Morton's 40 lb potassium chloride pellets (https://www.amazon.com/dp/B008Z27QIY, filed under Health & Household, sold alongside water-softener salt, no Amazon buy box, third-party offers $39.99 to $66.95 on 2026-08-30) are a water-softener product, not sold to that spec, and can carry anti-caking and resin-cleaning additives. They would work as a coolant at about a third of the price, but then every "food-grade" and "harmless on food" statement in this tier has to go. Which salt depends on the box. NaCl at 23.3 wt% freezes at -21.1 C (Yilmaz et al., https://proceedings.ises.org/conference/eurosun2010/papers/eurosun2010-0314-Medrano.pdf) and needs a verified -25 C or colder box, which reach-ins do not hold and the owner's -15 to -18 C boxes certainly do not. KCl at 19.5 wt% freezes at -10.7 C (253 to 284 kJ/kg, https://www.researchgate.net/publication/350336961) and recharges in any freezer including a -15 C one. Its plateau rules it out of a storage freezer: section 4.2 puts box air near -8 to -5 C during a chill with a -10.7 C bank, the go/no-go in section 6.7 rejects any run in which the inventory sentinel rises above -12 C, and a -10.7 C bank cannot hold a sentinel below -12 C. So KCl belongs only in a cabinet that holds no frozen inventory during the chill, a dedicated chilling box, and only after the section 6.7 batch test with a simulant sentinel in place. In a storage freezer use Tier D or Tier C with a recharge freezer instead, or drop the PCM half. Both salts are GRAS when they meet the FCC specification (21 CFR 182.1 and 184.1622).

If milk jugs are used anyway (dedicated non-food freezer only): the Webstaurant 1 qt HDPE squat milk jug (500mbsq32, $114.99 per case of 162; the $3.89 figure is the single "sample" SKU 999jbmqtea) has a 38-DBJ dairy neck and its listing names the closure, "Compatible with tamper-evident 500JBC lids (sold separately)", 100 per pack $9.99 (https://www.webstaurantstore.com/1-quart-32-oz-translucent-hdpe-squat-milk-jug-with-handle-162-case/500mbsq32.html; https://www.webstaurantstore.com/search/500jbc.html). The TableCraft 3838 cap an earlier draft specified is a 38-400 continuous-thread squeeze-bottle cap and will not seal on a DBJ neck. The 500JBC tamper band doubles as the seal check.

30 kg delta: three more bottle 12-packs ($119.97), a second 5 kg KCl bag ($95.99; four batches need 9.2 kg), the 5 lb xanthan pail, a second bus tub. About $693.

Verdict: the only DIY brew that recharges in a -15 to -18 C box, and the one that cannot protect frozen inventory, because its -10.7 C plateau sits above the -12 C sentinel line. A tier for a dedicated chilling cabinet, with FCC-grade salt from an identified maker with a lot certificate and SDS. If there is no such cabinet, drop it.

Tier C, buy the PCM, build the airflow: Engel -15 C packs plus a sealed fan.

30 kg: 13 Engel XL packs ($389.87), two bus tubs, about $690. Packs occupy about 30 L of shelf.

Verdict: within $21 of Tier A as printed, and it removes the mixing, the carbon fiber and the labeling argument entirely. This is the tier for a reach-in you will not put chemistry in. On a reach-in still under warranty it is packs only, no fan, with the swap bank and paddles (section 6.5). It still needs a recharge box whose logged mean is -24 C or colder at the pack position, and a full night is enough only with a fan or at -25 C (an XL pack in still air at -24 C takes about 12.6 W against 683 kJ, roughly 15 h plus sensible); the owner's -15 to -18 C boxes do not qualify. Before buying five: one XL pack, the Tier D calorimeter test scaled to the pack (8 kg of water at 25 C in a closed cooler, latent heat from the water drop) and a logged recharge in the intended box. Pass: a plateau within 2 K of -15 C on the way down and up, measured latent heat within 20 percent of 301 J/g (a drop of roughly 20 K in 8 kg of water for a 2.27 kg pack, before the blank and sensible corrections), and the pack hard after a logged recharge of the intended length in the intended box. Fail on any of those and the tier is the corrected NH4Cl recipe or Tier D.

Tier D, the swap bank (the owner's proposal): the Arctic Ice packs already bought, with a -23 C U-tek variant

What it is. Charged packs live in a dedicated non-food recharge chest freezer. When a hot batch goes into the working freezer, one set goes in with it, on an open wire shelf in the fan stream, and comes out when the batch reaches 5 C. It goes straight back to the recharge freezer. The packs never live in the working box. The working box's -15 to -18 C does not matter to this scheme: a pack colder than the box air is discharging, and that discharge is the job. Section 6.1's rule that the working box must hold at or below the PCM melting point applies to a bank stored in the working box (Tiers A to C), not to this one. The owner has bought the packs for it, so the first half of this tier is about those packs; the second half prices the colder -23 C U-tek variant the owner asked about first, which needs a colder recharge box and is the upgrade to consider after the packs in hand have been tested.

The packs in hand: Arctic Ice Tundra Series XX-L.

The owner has three. What the maker publishes (https://arctic-ice.com/products/tundra-series and the dimensions graphic linked from it, https://cdn.shopify.com/s/files/1/0011/1647/1361/files/AITundraSeriesPacksDimensionsGraphic.png, both seen 2026-08-30): 10.5 x 10.5 in face, 3.5 in thick, 10 lb (4.54 kg), "cools at 5°F (-15°C)", rigid UV-stabilized HDPE with a spin-welded cap, "NON TOXIC", and the charging instruction "cool Tundra Series packs at a temperature below -5°F or colder ... for 24-36 hours or until solid". The FAQ (https://arctic-ice.com/pages/faqs) says the Tundra Series "works best when cooled for at least 36 hours at the lowest setting available on consumer freezers". Not published: the fill, its latent heat, a conditioning curve, or an SDS. A 2013 trade article named the fill as PureTemp (r3 memo); PureTemp -15 is 301 J/g, density 1.03 liquid and 1.13 solid, k 0.55 liquid and 2.34 solid, GHS H302 harmful if swallowed (https://puretemp.com/puretemp-15technicaldatasheet/). Treat 301 J/g as the ceiling and measure the real value with test (a) below.

Price and stock, seen 2026-08-30: brand site 2-pack $85.99 (SKU 1262-2PK, $43.00 per pack), sold out in every size; Pelican Supply single $46.29, sold out (https://www.pelicansalessupply.com/products/arctic-ice-tundra-series-10-lb); Walmart shows $43.99 single and $89.95 for the 2-pack in search results, but the product pages served a CAPTCHA and neither price is confirmed; the Amazon listings (B088ZH89LG, B09GWRNGQ4) returned HTTP 500. What owners say about charging, same product family: "Freezer set to -5 F and they still are slushy inside" (Gordon K., 2-star review on the Engel 5 F pack page, https://engelcoolers.com/products/engel-5-f-15-c-freezer-packs) and "If you dont have a deep freezer rated to temps well below freezing, this is not your best choice" (Chad, same page); an Amazon Q&A thread on the Tundra packs carries "In the freezer for 7 days but still not solid" and "take a week to freeze solid" in a bait freezer (https://www.amazon.com/ask/questions/Tx2ARLP4RKIZR38/, quoted from the search result; the page itself returned 503). Those are consumer freezers near -18 C, which is the owner's working box. Expect the same there.

Geometry. The bounding box gives 0.071 m2 per face, 0.095 m2 of edge, 0.237 m2 in all, and 6.3 L of volume. The fill is 4.54 kg at 1.03 to 1.13 kg/L, so 4.0 to 4.4 L; the rest of the envelope is chamfered corners, the handle slot and headspace. Use 0.20 m2 of effective skin per pack, 0.60 m2 for three. For comparison: a 48 oz U-tek pouch is 0.133 m2 for 1.36 kg (0.098 m2/kg), an Engel XL 5 lb pack 0.175 m2 for 2.27 kg (0.077 m2/kg), the XX-L 0.044 m2/kg. The thick pack is built for capacity, not rate.

Capacity, three cases for the unknown latent heat:

Against the batches in section 6.2 with the 1.3 margin: 10 kg at 60 C needs 2.86 MJ, 10 kg at 70 C 3.38 MJ, 30 kg at 60 C 8.58 MJ, 30 kg at 70 C 10.1 MJ. On energy alone the three packs cover 10.5 / 12.4 / 14.3 kg of food at 60 C (8.9 / 10.5 / 12.1 kg at 70 C) for the 220 / 260 / 300 cases. Each further pack adds 3.5 / 4.1 / 4.8 kg. Five packs reach 17 to 24 kg; 30 kg needs seven packs at 300 J/g and nine at 220 J/g. Those are the honest energy numbers, and they are not what limits the set.

Rate. Once the batch is in, the working box air sits at about -5 to 0 C for the first hour on a small box and near -12 C on a healthy one (section 4.2), so the packs see 3 to 15 K. Three packs standing on edge, 2 in apart, on an open wire shelf in the booster fan stream:

Those are starting rates at 10 to 15 K; on a healthy box whose air stays near -12 C the driving difference is only 3 K and the same three packs take 27 to 45 W. As the faces melt, the liquid layer (k 0.55 W/mK) adds resistance; at 10 mm it roughly halves the rate. Call it 0.4 to 1.0 MJ over a 90 min chill on a small or tired box, 20 to 45 percent of a 10 kg batch at 60 C and 6 to 15 percent of a 30 kg batch, and much less on a box whose compressor holds the air near -12 C. The batch releases 400 to 700 W at the start; the packs take 90 to 225 W of it, the compressor and the frozen inventory the rest. For scale: the 11 to 13 NH4Cl bottles in a bus tub (section 4.2) take 25 to 105 W, so the three packs are 2 to 4 times that; six U-tek -23 C packs at an 18 K driving difference would take 215 to 359 W, about twice the three XX-L, because they are thin and colder. The packs come out of one 10 kg batch 20 to 45 percent spent. That is the shock absorber: a quarter to a half of the batch moved off the inventory and onto a store the recharge freezer pays back. It is not a cap on the box air, and it does not shorten the chill by much (the fan does that).

Packs 4 and 5 therefore buy skin, not storage. Two Engel XL 5 lb packs ($29.99 each, in stock at https://engelcoolers.com/products/engel-5-f-15-c-freezer-packs) weigh what one XX-L weighs and expose 0.35 m2 against 0.20 m2; they are the better fill for those slots if XX-L stock does not return.

Self-discharge if left in the working box. The box cycles -15 to -18 C. At the -18 C end the pack refreezes at about 3 K x 8 W/m2K x 0.20 m2 = 5 W; at the -15 C cut-in end there is no driving force at all; and every door opening and every hot batch pushes the air well above -15 C. Still air, h 8, a pack above its melt point for the warm half of each cycle: 2.4 W for 12 h is about 100 kJ a day, plus about 50 kJ for twenty 60 s door openings at 10 C air; 100 to 200 kJ a day is 10 to 15 percent of the pack's charge, and with the evaporator fan running it is double. Whatever it loses it cannot get back there: at a net 1.5 K of subcooling one pack needs 6 to 8 days to refreeze. Conclusion: charged packs live in the recharge freezer and enter the working box only with the batch.

Recharge time in the dedicated chest freezer. Energy to put back from 5 C is the latent heat plus 60 kJ/kg to cool the liquid to -15 C (cp about 3 kJ/kgK) plus 12 kJ/kg to subcool the solid 6 K (cp about 2): 1.33 / 1.51 / 1.69 MJ per pack for the 220 / 260 / 300 cases. Skin 0.20 m2, packs on edge and spaced so both faces see air:

Packs laid flat and touching, so only one face sees air, run 60 to 70 percent longer (46 h at -23 C in still air for the middle case). Conduction through the frozen shell is not the main limit: with k about 2 W/mK and a 44 mm half-thickness the Biot number is 0.18 in still air and 0.33 with a fan, which adds roughly 3 to 4 h at -23 C on top of the film time. Nucleation is not the problem it is for salt brines; the PureTemp-class organics do not supercool much, which is one reason to prefer them to the video's NH4Cl. The maker's 24 to 36 h at "-5 F or colder" matches the fan column or a box at -25 C; a still-air -21 C box is a two-day recharge. The fan column is there for a box whose maker approves one; no fan of any kind goes inside an R290 or R600a recharge box without the maker's written approval (section 6.1), and the three chest freezers priced below are R-600a or R-290, so plan on the still-air column. The rule: packs spaced 2 in apart, standing on edge in a wire basket, 48 h in a still-air box; 36 h only at -25 C or colder, or once test (b) shows the 36 h hold within 15 percent of the 48 h hold; verify with test (b) and then trust the clock, not the feel. Three packs from 5 C are 4.0 to 5.1 MJ over 48 h, an average of 23 to 30 W, well inside a 1/5 HP chest freezer, but the first hour is 0.8 MJ of sensible load, so put warm packs in the recharge box straight away rather than letting them warm further on a counter.

Two tests to run this week with the packs in hand.

(a) Calorimeter, one pack, measures the latent heat.

1. Charge one pack 48 h at the coldest setting of any freezer that logs -21 C or colder (borrow a chest freezer if the recharge box is not bought yet). Do not use the working box.
2. Put 20.0 kg of tap water (weigh it; 20 L) at 30 C into a 48 to 60 qt hard cooler. Lid on. Log the water temperature every 15 min for 30 min with the probe mid-depth. This is the blank: it gives the cooler's own drift per hour, which you subtract later.
3. Take the pack from the freezer, wipe frost off, and lower it into the water within one minute. Stand it on two spoons so water reaches both faces. Tape a second probe to the center of one face if you have one. Lid on.
4. Stir 10 s every 5 min for the first hour, then every 15 min. Log the water temperature at every stir. Note the time the face probe (if used) rises above -10 C.
5. Stop when the water has stopped dropping: two readings 15 min apart within 0.2 K. Expect 4 to 8 h. Record the final water temperature.
6. Arithmetic. Water heat = 20 x 4.19 x (30 - T_final) kJ, corrected by the blank drift (add the drift the cooler would have shown over the same hours; about 100 to 150 kJ for a 50 qt cooler over 5 h at 22 C room). Pack sensible = 4.54 x [12 + 3 x (T_final + 15)] kJ (solid from -21 to -15 C, liquid from -15 C to T_final). Latent heat in kJ/kg = (water heat minus pack sensible) / 4.54.
7. What to expect: a pack at 220 J/g takes about 1.43 MJ and drops the water 17 K, to about 13 C; at 260 J/g, 1.59 MJ and 19 K, to 11 C; at 300 J/g, 1.75 MJ and 21 K, to 9 C. A 1 K reading error is 18 J/g; the assumed liquid cp (2 or 3 kJ/kgK) moves the answer by about 25 J/g. The test separates 220 from 300; it will not separate 260 from 280, and it does not need to.

(b) Melt test, verifies the conditioning time. Room air, no fan, h about 9 W/m2K, 37 K between a 22 C room and the -15 C plateau: 0.20 m2 x 9 x 37 = 67 W. A full pack (1.0 to 1.36 MJ) melts completely in 4.2 to 5.7 h.

1. Tape a probe to the center of one flat face and cover it with a 2 in square of foam or folded towel so it reads the pack, not the room.
2. Take the pack from the recharge box after exactly 36 h of conditioning, stand it on edge on a wire rack in a room you have logged (write down the room temperature), and log the face probe every 10 min.
3. The face reads -18 to -21 C on arrival (subcooling; a pack with a liquid core shows little or none), climbs to -15 C in the first 15 to 20 min, sits within 2 K of -15 C while solid remains under the probe, then climbs. Hold time is the minutes from pull to the face reading -10 C. Expect 3 to 5 h for a full 10 lb pack; a half-charged pack holds about half as long.
4. Repeat with the same pack after a fresh 48 h conditioning. If the 48 h hold time is within 15 percent of the 36 h hold time, 36 h is enough and goes in the SOP. If 48 h holds longer, the rule is 48 h.
5. Write both hold times on the basket label. A pack that later holds less than 80 percent of the label figure was not charged, or the recharge box has drifted; log the box for 24 h.

Operating rules for the packs (these replace the jug steps in section 6.6 for Tier D).

• One set now (three packs), a second set of three when stock allows. With one set, the recharge clock (48 h until test (b) says 36) is longer than a day, so one 10 kg batch every other day; a set that is only 20 to 45 percent spent after one batch may come back overnight in a box at -25 C, but only test (b) run the next morning proves it. Two sets make it one batch a day without conditions.
• Packs go from the recharge freezer straight onto the shelf when the batch goes in: standing on edge in the fan stream, 2 in apart, faces parallel to the airflow, on an open wire shelf. Never under a pan, never lying flat, never in a solid-walled tub, never within 4 in of the evaporator housing.
• Back to the recharge freezer when the thickest pan reads 5 C, or after 3 h, whichever is first. Log time out and time in.
• A dedicated wire basket per set, labeled "COOLANT PACKS, SET 1, NOT FOOD, RECHARGE __ H (from test b)", and the basket goes into the recharge freezer with the packs on edge and spaced. Write the conditioning start time on a wipe-off tag.
• Before use: no slosh when tilted, and the flat face does not flex under a thumb. Either means back in the recharge box; there is no partial credit and no kg deduction (Codex finding 19).
• Cut, cracked, weeping or bulging pack: gloves; bag it; discard any food or pan the fill touched; wipe the shelf with paper towels, then warm soapy water, then the quat bucket. The fill is undisclosed and "non-toxic" is a marketing claim, not an SDS; PureTemp -15, the likely fill, is H302 harmful if swallowed. Email Arctic Ice for the SDS before the first production batch and file it in the HACCP binder with the section 7 position (sealed operating equipment in its own bin, not a chemical in food storage). A pack that has leaked once is retired.
• Weekly: wipe, inspect the cap weld and corners, run test (b) on one pack from each set every month or after any recharge-box alarm.

Cost. The packs are sunk ($43 to $46 each at the prices seen 2026-08-30, all sold out at the brand site and Pelican Supply). The incremental buy:

Incremental total: about $525 to $675 with the Galaxy CF7 gamble (video-copy fan kit to two rated fans, plus the logger; baskets extra), about $730 to $920 with the Atosa, about $1,595 to $1,745 with the Ultra Cold, each plus $60 to $92 if packs 4 and 5 are added. Every option is under the $2,089 Avantco SF-3, and none of them takes the load off the storage freezer; they move a quarter to a half of it onto the recharge box.

Verdict on the packs in hand: the tier to run first, because the PCM is already on the shelf and it is the right temperature for a box that cannot recharge it. Its weakness is the pack shape (capacity the chill cannot use, skin it needs more of), which is why packs 4 and 5 should be thin, and its recharge clock, which is why a second set is the real upgrade.

The -23 C variant: ThermoSafe U-tek 427 packs, for a recharge box verified at -28 C or colder.

Why a -23 C pack rather than a -15 C one. During a chill the working box's air sits where the compressor and inventory put it: about -12 C on a healthy 1/2 HP box, -8 to 0 C on a small or tired one (section 4.2). A -15 C pack sees 3 to 15 K of driving difference in that air; a -23 C pack sees 11 to 23 K, two to four times the heat per square meter, and it works from the first second even in a box whose air never rises above -15 C. That is the case for the U-tek. The case against buying it now is the recharge box: -28 C or colder at the warmest spot, which no ordinary commercial chest freezer is rated for.

Pack data. Published, from ThermoSafe's product table (https://wordpresssonoco.azurewebsites.net/products/refrigerants/pcm/ and https://www.thermosafe.com/products/refrigerants/pcm/, seen 2026-08-30): model 427, 48 oz (1,361 g), -23 C (-10 F), 10.50 x 7.80 x 1.20 in, laminated metallized film pouch. Fisher 035257G: "Freezing Point -23°C", 1361 g, 10.5 x 7.75 x 1.18 in, "Laminated, Metallized Film", $88.00 per pack of 6 (https://www.fishersci.com/shop/products/thermosafe-u-tek-phase-change-materials-6/035257G); Amazon B00C2T41CK $83.22 in stock from a third-party seller, other sellers from $75.88, non-returnable (seen 2026-08-30). The 2013 MSDS (https://www.thermosafe.com/wp-content/uploads/2021/03/MSDSU-tek-10F010113.pdf): specific gravity 1.16 at 75 F, 100 percent water soluble, pale yellow odorless "eutectic gel", eye contact may cause mild irritation, "Ingestion of large amounts may cause stomach upset and thirst", no composition given, and this handling note: "Careless handling of frozen U-tek products may damage the film pouch material, resulting in leakage of the eutectic gel during thaw. Before refreezing, examine film material for tears, punctures, etc." Not published anywhere found on 2026-08-30: latent heat, enthalpy curve, conditioning temperature and time, cycle life, static load limit, or a current GHS SDS with composition. The one Amazon review says a previous set lasted "5+ years before they finally got a puncture". Every capacity number below is therefore given at three assumed latent heats, 200, 250 and 300 kJ/kg, and the first purchase, if any, is one case, not four, until the melt test at the end of this variant is done.

Mass and area. One pack 1.361 kg; a case of six 8.17 kg. Area per pack, both faces plus edges, 0.133 m2 (faces alone 0.105 m2); six packs 0.80 m2, twelve 1.60 m2. That area counts only if the packs stand on edge in a wire rack or basket with 1 in gaps, in the fan stream. Stacked flat in a tub, count one face per pack and about a third of the rate.

Capacity per case at 200 / 250 / 300 kJ/kg: 1.63 / 2.04 / 2.45 MJ latent, plus about 0.1 MJ of sensible cold because the packs arrive at -30 C or below, not at -23 C. A 10 kg batch at 60 C is 2.2 MJ; with the report's 1.3 margin, 2.86 MJ. Packs per 10 kg batch with margin: 11 at 200, 9 at 250, 7 at 300. Two cases (12 packs) per batch cover the energy at any plausible latent heat. One case covers 74 to 111 percent of the bare batch and leaves no margin. Two sets in rotation means four cases, 24 packs, 32.7 kg.

Rate, the number Tiers A to C fail on. Absorption at the start of a chill is area x h x (air minus -23 C). With h 15 to 25 W/m2K (open rack in the fan stream) and air -12 to 0 C:

A 10 kg batch releases 400 to 700 W at the start (section 4.2). Twelve spaced U-tek packs can take most of that; six take a third to a half; the -15 C options take little while the box is healthy, because their plateau is within 3 K of the air. Sink balance with section 4.2's own inputs (food UA 10 W/K at 60 C, inventory UA 25 W/K at -18 C, pack UA at h 20): on a healthy 1/2 HP box (coil 550 W) twelve U-tek packs pull the first-hour air to about -17 C and take about 190 W; six packs give about -15 C and 130 W; the 13-bottle NH4Cl tub gives -12 C and 35 W. On a weak box (coil 150 W) twelve packs hold the air near -11 C and take about 380 W, more than half the food's output; six packs -7 C and 250 W; the NH4Cl tub -4 C and 140 W. These are screening estimates with an assumed h, not measurements; the validation runs below replace them.

Self-discharge if a set is left in the working box. Still air, pack on edge, UA about 1.06 W/K (0.133 m2 x 8 W/m2K, the recharge table's figure below), 5 to 8 K warm: 5.3 to 8.5 W per pack, 340 kJ of latent heat gone in 11 to 18 h (the "half a day to a day" in section 6.1); packs stacked flat expose one face each, about 0.5 W/K, and last about twice as long. With the fan at 6 V (h about 15): 10 to 16 W per pack, gone in 6 to 10 h. With the fan at 9 V (h about 25): 17 to 27 W per pack, gone in 4 to 6 h. Six packs left in the box with the fan running are a 60 to 160 W load on the compressor until they are empty. That is why the packs leave the working box with the batch.

Recharge freezer for the U-tek. The packs need air at least 5 K below -23 C at the warmest spot around the clock, so -28 C is the floor and -30 to -35 C is the working target. Most commercial chest freezers cannot do it: the Avantco CF8 is rated -25 to -18 C (spec sheet in section 6.1); the Galaxy CF7 and Atosa MWF9007 above stop at -22 to -23 C; Smad's Amazon listings say 14 F to -11.2 F (-10 to -24 C); the Excellence HBD-8HC dipping cabinet -15 to 15 F (-26 to -9 C). What can, seen 2026-08-30, all 115 V:

The Northair is the only one under $1,000 and it is a household appliance with no stated current, refrigerant or cooling capacity. It is worth a $288 trial only with a logger at the warmest interior spot and a full 12-pack recharge before any more packs are bought. The Galaxy Ultra Cold is the commercial answer under $1,500 and is the same box that recharges the Arctic Ice packs with margin to spare. Twelve packs on edge with 1 in gaps occupy about 1.3 cu ft; the charged second set, which can sit tight, about 0.7 cu ft; the 3.5 cu ft box holds both, the 5 cu ft boxes hold both with room for the rack.

Recharge time, per pack on an open rack, latent 340 kJ at 250 kJ/kg (408 kJ at 300) plus about 105 kJ of sensible from a -5 C liquid pack, which adds about 1 to 2 h at these rates:

Two limits the table does not show. First, the box's compressor: twelve packs bring about 5.3 MJ (1.5 kWh of heat) per night, and a 3.5 cu ft box at -35 C evaporating may deliver under 100 W of cooling, which alone is 15 h. The box's rated energy (Northair: 0.45 kWh per 24 h, from a search snippet of its site, not the product page) is the idle figure, not capacity. Second, warm packs put in together lift the box air; the log will show how long the air stays above -28 C after loading. Both are settled by one measured recharge, below. No fan of any kind goes inside a recharge box on R290 or R600a without the maker's written approval (section 6.1; the Elcold and the Galaxy Ultra Cold are R290, the Northair does not say): space the packs and use the still-air column.

Cost of the U-tek variant.

Per 10 kg of daily capacity that is $352 of packs plus the box. Cost per kJ of latent capacity at the Fisher price, at 200 / 250 / 300 kJ/kg: $0.054 / $0.043 / $0.036, the same band as Engel packs at their assumed 301 J/g. At 30 kg per batch it is 36 packs per set, 72 packs in rotation, 12 cases, $1,056, plus a 9 cu ft recharge box ($3,900 to $4,200): about the price of the Avantco SF-5 blast chiller ($3,789), which takes the load off the storage freezer instead of moving it. Tier D in either form is a 10 kg design.

Operating rules for the U-tek variant are the pack rules above with three changes: the recharge floor is -28 C on the log for the whole conditioning time (the time measured in the recharge run below plus 2 h), the sentinel pack's face reads -25 C or colder (2 K below the plateau, SOP step 1; a pack that flexes is not charged), and the pouches, being shipping refrigerants, are inspected for tears before they go in and before they go back, per the MSDS; a pack that weeps is retired, and food it touched is discarded, as for any coolant.

Validation before buying the second, third and fourth cases.

• Melt test, one pack, for the latent heat ThermoSafe does not publish: 8.0 kg of water at about 25 C in a closed foam cooler, probe in the water, stirred every minute. Drop in one pack straight from the recharge box with its surface logged at -30 C or colder. Run until the pack is fully liquid and the water stops falling (30 to 60 min). Blank run first, no pack, for the cooler's own loss. Latent heat per kg = (8.0 x 4.18 x water drop, minus the blank loss, minus the pack's sensible heat, 1.361 x [2.0 x (-23 minus pack start temperature) + 3.5 x (water end temperature + 23)]) divided by 1.361. At 250 kJ/kg expect a drop of about 15 K; at 200 about 13 K; at 300 about 17 K. Put the result in the capacity paragraph above and buy the remaining cases only if it supports the count.
• Recharge run, one full set: 12 packs from about 0 C, on the rack, logger at the warmest spot. Record the time for the air to return below -28 C and the time for the sentinel pack's face to read -25 C or colder. That time plus 2 h is the conditioning time. If it is over 20 h the box cannot turn a set a day: the second set becomes the every-other-day set, or the box is too small.
• Chill run: section 6.7 run D with the U-tek basket instead of the Arctic Ice packs, then the first real 10 kg batch. Pass: air peak lower and sooner back below -15 C than run B, inventory sentinel never above -12 C, and the basket comes out with every pack still partly firm. If all are fully liquid before the batch hits 5 C the set is undersized for that batch; use 12.
• Data request, before any of the above if possible: ThermoSafe customer service 800-323-7442 (from the MSDS), or through the Documents tab on Fisher's 035257G page. Ask for the latent heat or enthalpy curve of the U-tek -23 C gel; the recommended conditioning temperature and time for the 48 oz pack; cycle life; static load and puncture limits; a current GHS SDS with composition; and whether the pouch is suitable for use in a food-storage cabinet. Until the latent heat arrives, the variant's capacity numbers are the 200 / 250 / 300 band.

Verdict on the variant: the better swap-bank material on rate, two to four times the Arctic Ice packs per square meter, and the wrong first purchase, because the owner already holds 13.6 kg of -15 C PCM and the -28 C recharge box is the expensive part. Buy it only after the Arctic Ice tests show the -15 C packs cannot take enough of the chill on this box, and only with a recharge box that has logged -28 C or colder at its warmest spot.

BOM totals, firm versus placeholder (all seen or computed 2026-08-30; before shipping and sales tax; no total for an R290 or R600a cabinet because no DIY fan goes into one, section 6.1). The "printed" column is the sum of the rows as printed in this section after the Codex round, which replaced the bench supply with the Mean Well GST120A12-R7B, the reed switch with the GRI contact, added the flyback diodes, and replaced the flat cable with 14/2 cable and a lid gland, and the final pass, which put the RioRand PWM ($9.99) back into Tiers A and B so that the 9 V and 6 V settings the SOP uses exist. The last column swaps the video-copy fan for the rated Sanyo fan and a 120 mm guard and drops the RioRand, because the Sanyo is driven from its own PWM lead (section 6.3a).

Arithmetic: Tier A = 59.99 + 57.27 + 9.99 + 56.99 + 15.00 + 30.99 + 1.99 + 79.98 + 9.39 + 7.29 + 8.49 + 3.09 + 9.99 + 1.50 + 11.99 + 38.22 + 8.00 + 15.00 + 5.29 + 3.79 + 31.20. Tier B = 263.81 common hardware + 95.99 + 30.99 + 39.99 + 9.39 + 1.99. Tier C = 149.95 + 99.94 + 57.27 + 9.99 + 128.07. The 30 kg rows add the deltas stated under each tier. Before this round the same tiers summed to $442.81, $419.53 (with the bentonite the table had omitted) and $406.08; the supply (+$9.78), door switch (-$0.86), diodes (+$1.50), cable and gland (+$2.22) and the RioRand PWM (+$9.99, final pass) account for the difference in Tiers A and B, and the supply (+$36.28) plus the same switch, diode and cable changes for Tier C. The earlier Noctua substitution for R290 cabinets ($437.94 at 10 kg, $683.15 at 30 kg) is withdrawn. Rated fan delta: Sanyo in place of the GESEXI 2-pack +$41.34 (A and B) or in place of the SPAL +$1.39 (C), 120 mm guard placeholder $5 in place of $15, RioRand PWM dropped (-$9.99, all three tiers, the Sanyo uses its own PWM lead); a second Sanyo fan adds $101.33 plus a $5 guard.

6.3a The rated electrical stack

This is the parts list that replaces the fan, supply, door switch, cable and fuse lines in every tier when the fan goes into a cabinet that holds inventory. It is still an unlisted modification of the cabinet (section 6.1); it is the closest a retrofit gets to rated parts, not an approval.

Conductors and fuse. The smallest conductor in the run is the supply's own 16 AWG output lead (spec sheet: "UL2464 16AWGx4C" on the 12 V model, two conductors per pole). NEC 240.4(D) limits 16 AWG to 10 A and 14 AWG to 15 A (the small-conductor rule; verify against the NYC 2025 code copy cited in section 7), so the fuse is 10 A, and the pass-through and in-box run are 14 AWG so that nothing downstream of the fuse is smaller than the fuse. What protects what: the supply's hiccup limit (105 to 150 percent of 8.5 A, so 8.9 to 12.75 A for a fraction of a second, then shutdown and retry) is the overload protection for the whole run; a current-limited supply will never open a 15 A fuse, and may not open a 10 A one either. The 10 A fuse is the backstop for a supply that fails high or for a fault between the enclosure and the fan when the supply is delivering its limit, and it must be there because the 16 AWG lead cannot take a failed supply's output. Motor start: a brushless San Ace fan has its own soft start and draws well under 3 A on start-up (rated 3 A), so one fan sits inside the GST60A12's 5 A and two inside the GST120A12's 8.5 A with margin. A brushed GESEXI or SPAL fan is a different case: its stalled-rotor current is several times the running current and can trip the hiccup point, so bench-test the start at the intended setting with the bench meter before trusting a listed brick with it; the GST120A12 does not run the 16 in GESEXI at full voltage (10 A nameplate).

Speed. Do not chop the supply of the Sanyo fan with the RioRand PWM controller; that is a brushed-motor chopper and a brushless fan with internal electronics does not want it. Run the fan at 12 V full speed (224 CFM, 36 W each) for the chill, or drive its PWM lead with a logic-level 25 kHz PWM signal (a $5 signal-generator module, not priced here; check the San Ace PWM input spec on the datasheet) and idle it at 30 to 40 percent between chills so it never cold-starts. Full speed on two fans adds 72 W of heat to the box, 260 kJ over a 60 minute chill; at 70 percent PWM the cube law says about 25 W for about 300 CFM, unmeasured, so log the supply's input watts on the plug meter and use the measured figure in the section 4.2 budget.

Wiring in one paragraph. Supply, relay, diode, fuse holder and terminals live in a small enclosure on the wall bracket above mop height, plugged into the GFCI receptacle through the countdown timer. Supply positive to the 10 A fuse; fuse to one lead of the GRI door switch and to relay terminal 30; the other switch lead to relay terminal 86; terminal 85 to supply negative; the 1N4001 across 85 and 86 with its band on 86; relay 87 to the cable positive; cable negative to supply negative; cable through the lid gland to the fan's positive and negative leads with crimped, heat-shrunk terminals; the fan's PWM and tach leads back through the same gland if speed control is used. Door closed: magnet at the switch, contact closed, coil energized, fan runs. Door open: contact opens, coil drops, fan stops within a second. For a brushed fan (Tiers A and B, or the SPAL in Tier C) the RioRand PWM sits between relay 87 and the cable positive, inside the enclosure, with its knob marked at the 9 V and 6 V settings (section 6.7 Part 1); the Sanyo fan gets no RioRand.

Weekly fail-safe check. Open the door with the fan running and watch it stop within a second; a fan that keeps running means the door switch has welded or the relay has stuck, and the fan is out of service until the part is replaced. Close the door and confirm it restarts. Pull the fuse and confirm the fan stops. Look at the gland for frost and the cable at the lid edge for chafe. Log all four checks.

What is still unlisted. The fan is rated for cold and wet, the supply is listed, the switch is UL-recognized, and none of that makes the cabinet a listed appliance again. No add-on fan sold today carries an NSF food-zone listing or a reach-in maker's approval for retrofit (searches 2026-08-30). The cabinet's own listing and warranty are gone the day the fan goes in, and on an R290 or R600a cabinet the fan does not go in at all without written approval (section 6.1).

6.4 Recipe, video ratio versus corrected, per 10 kg batch (about 9.4 L)

Mixing. This is chemical work, not food prep, and it gets its own tools whether or not carbon fiber is in the recipe: a 5 gal HDPE bucket with lid, a drill-mount paddle mixer or a long whisk, a scoop, a scale tray and the wide-mouth funnel, all marked "COOLANT ONLY" in paint marker and stored with the chemicals, never on a food shelf. The 20 qt production mixer and its bowl stay out of it. Mix in a corner away from open food and food-contact surfaces, with the bus tub under the bucket, gloves and safety glasses on, N95 for the bentonite (section 7). Bucket and drill paddle prices are not confirmed; both are stock items at any restaurant or hardware supply. Steps:

1. Heat 7,900 g of water to 45 to 50 C. Dissolving 2.1 kg of NH4Cl absorbs about 580 kJ and drops 10 kg of solution 14 K, so it finishes near 31 C without a heated bowl.
2. Drill paddle at low speed, or whisk. Add the salt in four scoops over two minutes, mix five minutes until clear.
3. Probe; anything from 25 to 40 C is fine. Sprinkle in the bentonite. (If copying the fiber step, this is where the under-40 C rule bites.)
4. Sift the xanthan over the surface in a slow stream over 60 seconds with the paddle running, mix five minutes, scrape the wall, three more minutes.
5. Fill warm; it thickens as it cools. Ladle into an 18 in pastry bag, snip a 20 mm corner, pipe through the wide-mouth funnel (a commenter's tip on video 2; faster than the dowel and much cheaper than the vacuum filler).
6. Fill to 90 percent: 900 mL, which is about 954 g at 1.06 kg/L, in a 1,000 mL lab bottle; a 10 kg batch fills about 10 bottles. Weigh the first filled bottle on the kitchen scale (tare the empty bottle) and write both figures on the recipe card, "900 mL / 950 g"; if the scale says something else, the gram figure rules and the mL figure gets corrected. The eutectic is 80 percent ice by mass and expands 6 to 8 percent.
7. Cap hand tight on a dry thread, freezer tape over the seam, and write on the molded label panel with paint marker or engrave: "COOLANT, NOT FOOD. DO NOT OPEN. Ammonium chloride. Leak: rinse with water, NO BLEACH. Discard any food it touched." Tape alone is not a label at -20 C. Number the bottles, write the count on the tub.
8. Rinse the bucket, paddle, scoop and funnel in the utility (mop) sink, not the prep sink and not the three-compartment sink. Collect the first rinse in the bus tub and dispose of it as the SDS and local sewer rules say; detergent, rinse, dry, back to the chemical cabinet. No chlorine on anything that touched the brine. These tools never go back to food use.
9. First freeze: bottles spaced 1 in apart on an open wire shelf (not in the tub) with the fan at 6 V, 24 h in a -23 C box, longer in a warmer one; check both sentinels are below -15.4 C, every jug is hard, none has bulged or wept, and write down how long it took: that is the first data point for the conditioning time. A jug that reads below -15.4 C and is still liquid is supercooled: shake it and watch it snap to -15.4 C and go hard. If any jug is still liquid below -15.4 C after a full night at 6 V, or has to be shaken to freeze on two recharges running, do not add nucleators or change the recipe. Retire the brine and use a characterized commercial PCM.

6.5 Install plan

Reach-in upright (True T-23F class, 22 7/8 x 23 1/4 in shelves). Bottom shelf: the jugs stand on the wire shelf with 1 in gaps during the chill and during recharge, tub underneath as a drip tray, the set pushed forward so the rear 4 in stays clear of the evaporator housing (True and Turbo Air clearance rule). Jugs packed in a solid tub see moving air on their tops only; they absorb at about a third of the spaced rate during a chill and take roughly two to three times longer to recharge (estimate, section 3.5). Tier D packs take the same shelf, on edge, 2 in apart, faces parallel to the flow. Next shelf up: the 16 in fan lying flat, flat face up, blade reversed and leads swapped per the listing, blue threadlocker on the nut, zip-tied at the four shroud corners with a strip of closed-cell foam tape under it. The fan is about 35 percent of the cabinet cross-section, close to the video's 40 percent. Shelf above the fan: hot pans 4 to 8 in over the blade; that shelf is the finger guard. Shallow 21 in boxes take the 12 in fan. Cable entry. There is no owner-installable route into a reach-in. A cable under the magnetic gasket is a permanent air leak on a door that cycles all day, and the gasket cannot seal to the frost it will grow ("Gaskets can't seal to ice or frost", https://www.gasketguy.com/blog/heater-wires); the flat Starlink-type cables sold for this are 18 or 20 AWG (Amazon B0H395T36Y, seen 2026-08-30), so that route is withdrawn. The liner is foamed in place with refrigerant and heater wiring in the walls, and True's manual (page 5, seen 2026-08-30) says "Contact TRUE Manufacturing to locate refrigerant lines and electrical wiring before drilling, cutting or puncturing interior or exterior walls". So the reach-in version of this build exists only if the maker identifies a safe spot in writing and a refrigeration technician installs a sealed bulkhead gland there. Without that, the reach-in gets the swap bank and the paddles, not the fan. Supply outside on a wall bracket above mop height, plugged into a GFCI receptacle through the countdown timer, with the relay, fuse holder and terminals in a small enclosure beside it. Fuse and conductors per section 6.3a: 10 A blade fuse on the positive lead, matched to the smallest conductor in the run (the supply's 16 AWG lead, 10 A under NEC 240.4(D)), with the pass-through and in-box run in 14 AWG. Crimped, heat-shrunk terminals at the fan; no splices inside the box. Door interlock. A Bosch-style 12 V 30/40 A relay coil draws 150 to 175 mA (Bosch 0332209150: "Coil Operating Current: 160 mA", "Equivalent Coil Resistance: 69 - 80 Ohms", https://www.texasindustrialelectric.com/relays_0332209150.asp; TEMCo CN0170: 1.8 W coil, https://temcoindustrial.com/bosch-style-automotive-relay-cn0170-1-qty-12-v-30-40-amp-spdt/; both seen 2026-08-30). The True MODS kit does not publish its coil current, so assume the same. The Adafruit 375 reed switch is rated 100 mA and cannot drive that coil either; it is dropped. Use a GRI 2020-12 magnetic contact instead: hermetically sealed Form A reed, 0.5 A and 10 W switching rating, 200 VDC, "plated contacts resist sticking or freezing" (datasheet https://www.grisk.com/wp-content/uploads/mdocs/2020%20Series_.375%20In.%20Recessed.pdf; $3.09 each at https://www.jmac.com/GRI_2020_12_B_p/GRI-2020-12-B.htm, seen 2026-08-30). The coil draws a third of its rating. It is a 3/8 in press-fit part, so do not drill the cabinet for it: press the switch into a drilled block of HDPE stuck to the outer cabinet face beside the door edge, and the magnet into a matching block on the door, both outside the gasket and the wet zone. Wiring: +12 V from the fused supply lead to one switch lead; the other switch lead to relay terminal 86; terminal 85 to supply negative; a 1N4001 diode across 85 and 86 with its band (cathode) on the 86 side (Adafruit 755, 10 for $1.50, seen 2026-08-30) so the coil's collapse discharges through the diode instead of across the reed contacts; fan positive through relay 30 to 87. Door closed, magnet near, switch closed, coil on, fan runs; door open, switch opens, coil drops, fan stops, the same as the stock evaporator fan. An open failure of the switch stops the fan (safe); a welded switch leaves the fan running with the door open (unsafe and silent), which is why the weekly check below exists. If a lower-current switch is ever substituted, put an opto-isolated 12 V relay module between it and the coil (trigger current about 5 mA, contacts 10 A at 30 VDC, HiLetgo B01LW15D1M or equivalent; price not confirmed on 2026-08-30) rather than running coil current through it. Relay, diode and terminals live outside the box in the same enclosure as the fuse. Defrost clock: reach-ins defrost on a timer, typically every 6 h (Turbo Air: "every 6 hours (4 times per a day)"; "the compressor does not run during the defrost cycle"), with a several-hundred-watt heater on the coil and the stock fan off so the heat stays in the housing. A 2 h chill has about a one in three chance of overlapping one, and the booster would pull that heat out across the pans and inventory. Record the unit's actual defrost schedule over at least one full day before the first chill ("dEF" or "def" on the display, or the heater click, with start and end times). Defrost length is model-specific and often temperature-terminated, so there is no universal wait: True ends it "when a specific evaporator coil temperature is reached or by a time duration" (https://www.webstaurantstore.com/documents/pdf/tseries_manual.pdf); Turbo Air runs one every 6 h with the compressor off throughout (https://turboairinc.com/wp-content/uploads/2020/01/M3-Reach-Ins_HC.pdf); Hoshizaki starts one after 5 h of cumulative compressor run, ends it at 45 F (7.2 C) or 1 h, then holds the compressor 3 minutes and the evaporator fan 7 minutes before restarting them (https://hoshizaki.com/docs/manuals/Commercial_S_Series_inst.pdf). Start a chill only after a defrost has finished and the stock evaporator fan is running again, never inside 30 minutes of the next scheduled one. If a defrost starts mid-chill: booster off, keep logging product temperature every 15 minutes, and if the log shows 21 C (70 F) by 2 h is at risk, move the pans to a validated alternate method (ice bath, cooling paddle, another cabinet) rather than waiting. Booster back on only when the display shows a temperature again and the stock fan is running. A chill run puts more frost on the coil, so the following defrost runs longer. PWM knob at the 9 V mark for the chill (8 to 9 V across the fan leads, 36 to 50 W), at the 6 V mark for recharge (5 to 6 V, about 15 W, also keeps the bearing warm and the blade frost-free); the two marks are set on the bench per section 6.7 Part 1.

Chest freezer (Atosa MWF9016GR class, interior about 54 x 21 x 27 in). Note that the MWF9016GR itself is R290, so this layout applies only to an older R134a chest or to a hydrocarbon chest with the maker's written approval. The 16 in fan stands vertically on the compressor step against the end wall, zip-tied to a piece of epoxy-coated wire shelf leaning on the wall, blowing along the length of the box; the end-wall cross-section is about 3.9 sq ft so the fan is 36 percent of it. Floor-mounted blowing up only works with two fans because the floor is 7.9 sq ft. A wire rack or upturned basket 12 to 24 in downstream at mid-height carries the pans; jugs standing 1 in apart on a wire rack on the floor at the far end, the bus tub under the rack as a drip tray; Tier D packs on edge on the same rack, 2 in apart, faces along the flow. Cable entry through the lid, not the drain and not the walls. A chest freezer's condenser tubing runs "just inside the outside wall" and "wraps around the back, sides and front", and the evaporator tubing runs inside the walls against the liner (GE, https://products.geappliances.com/appliance/gea-support-search-content?contentId=18308, seen 2026-08-30); the lid is foam between two skins with no tubing, which is why every chest-freezer conversion drills the lid ("no pipes to drill / cut through", https://www.jimsbeerkit.co.uk/forum/viewtopic.php?t=76490). The drain is out: Atosa's manual says "replace the drain plugs when defrosting and cleaning are completed. If the drain is left open, warm air may enter the freezer", and it is where the defrost water runs. Drill a 15.5 mm hole through the lid near the hinge side, well inside the gasket line, fit a PG9 IP68 nylon cable gland for 4 to 8 mm cable (Essentra CG-PG9-1-BK, $1.22 at DigiKey per search listing, page not reachable on 2026-08-30; uxcell 10-pack B01MQWU2NM, price not confirmed), bed both faces in silicone, and run 14/2 round cable (Southwire 55213242 type, section 6.3a) with a drip loop outside and slack for the lid swing. This is still an unauthorized modification under Atosa's warranty terms, so it is for a dedicated cabinet the owner accepts as altered, and on an R290 or R600a chest it also needs the maker's written approval (section 6.1). Fit the wire guard: the open blade face points where hands go. A lid switch matters more here because a chest freezer holds its cold air until a fan pumps it out. A chest freezer has no defrost heater, so the fan moves pan steam onto the walls and lid gasket; always cover pans and add wall frost to the weekly check. In a static box the fan is worth 4x to 6x, so 8 V across the fan leads is enough.

Cold start. The GESEXI fan carries no temperature rating. If it hums and does not spin at -25 C, cut power at once (a stalled brushed motor at 10 A heats fast), open the door and free the blade by hand with the supply off; frost on the hub is the usual cause. Do not kick it with 12 V: that is the stalled-rotor case, and it either blows the fuse or drops the brick into protection. The 6 V overnight setting on recharge nights keeps the blade frost-free; on other nights the fan is off and may need a hand start in the morning. The Sanyo fan is rated to -20 C and has its own soft start; if it fails to spin at -23 C it is outside its rating, not faulty, and the fix is to leave it idling at a low PWM setting so it never cold-starts.

6.6 Standard operating procedure

Post beside the freezer. One log row per batch: item, pan depth and mass, time in, product temperature at 0, 1, 2, 4 and 6 h, freezer air low and high, jug count in and out, initials. Keep 90 days. For Tier D (Arctic Ice packs) use the pack versions of steps 1, 2, 4, 9 and 10 below; the jug versions apply to Tiers A to C.

1. Charge check, fail closed. The set is charged only if all three hold: (a) the recharge log shows box air below the set's recharge floor (-20 C for NH4Cl, -24 C for Engel -15 C packs, -28 C for U-tek; Tier D packs use the pack version below) for the whole validated conditioning time, which is the time measured in this freezer on this rack on three separate nights (section 6.7 recharge run) plus 2 h, written on the rack card; (b) both sentinels read below the plateau by at least 1 K and more than the probe's stated accuracy; (c) three containers from different rack positions are hard and none is bulging or weeping. If any of the three fails, the set is not charged: chill without the bank and log it. No partial credit, no counting soft jugs, no cutting the batch by feel. Sentinels: two per set, one at the warmest rack position (door side, top) and one at the coldest (nearest the evaporator). For DIY bottles the probe sits in a thermowell, not in the brine: a closed-end polypropylene or HDPE tube (a 10 mL PP centrifuge tube or 6 mm tubing) through a hole in the cap, sealed on the outside with epoxy, a few drops of mineral oil in the bottom for contact, tip at the bottle's center. For Engel or U-tek packs there is no thermowell; tape the probe tip to the center of a face under a 2 in square of closed-cell foam, with a second pack face to face over it for the last hour, and treat the reading as surface temperature: pass is 2 K below the plateau plus the validated time. Do not use an IR gun for any of this: a frosted plastic surface reads air and frost, not core state. A container that reads below the plateau and squeezes soft is supercooled liquid; shake it, watch it snap to the plateau and go hard, and treat it as uncharged today. Log every supercooled jug by number; two recharges running with supercooled jugs ends the recipe (6.4 step 9). - Tier D (packs). Take the set from the recharge freezer only when the pans are portioned and ready to load. Check the tag: at least the tested time on the basket label (48 h until test (b) in section 6.3 shows that 36 h is enough). Tilt each pack: no slosh, no flex of the flat face. Any pack that fails goes back on the tag's clock; do not use it and do not trade it for extra pans. Do not use an IR gun or a squeeze.
2. Count jugs against the tub and add one tally mark per jug on its label panel. Retire any cracked, bulging, weeping or crazed jug and any jug with a distorted cap or damaged thread; when the first jug of a set fails, cap the set at 80 percent of that cycle count (section 7). - Tier D (packs). Count three packs against the basket label. Retire any pack that is cracked, weeping, bulging or has a damaged cap weld.
3. Portion into hotel pans 2 in deep or less, filled to 1.5 in. Never a 4 in pan, never a stockpot. Cooked grains and braises at 1 in if they must reach 5 C in under 3 h.
4. Use the full validated set for the batch size on the rack card, every container on the wire rack with a gap. If the set did not pass step 1, chill without it and log that. - Tier D (packs). All three packs on the shelf for any batch up to 15 kg. The packs do not license a bigger batch; section 6.2 sets the batch size.
5. Cover each pan loosely (film not sealed at one corner, or a lid ajar). Uncovered is faster (by how much is not measured for these foods; r5's guess was 20 to 30 percent) and frosts the coil. Until it is measured here: loosely covered, and uncover only for the first 10 to 15 minutes while it steams. To measure it, run the same item covered and uncovered on different days, weigh the pans in and out, log the time to 21 C and the length of the next defrost, and set the rule from that.
6. Load on the shelf above the fan (reach-in) or the rack downstream (chest), 2 in gaps, nothing against the back wall, nothing within 4 in of the evaporator housing, nothing on the jug tub, no stacking. Tier D: packs on edge in the fan stream, faces parallel to the flow, 2 in apart, never under a pan.
7. Check the recorded defrost schedule (section 6.5): start only after a completed defrost with the stock fan running, never within 30 minutes before the next one. Probe in the center of the thickest pan. Door shut. PWM knob to the 9 V mark, timer to 2 h. If "dEF" appears during the chill: booster off, keep logging every 15 minutes, and if 21 C (70 F) by 2 h is in doubt, move the pans to the ice bath, the paddle or another cabinet. Booster back on only after the display shows a temperature again and the stock fan is running (section 6.5).
8. Log at 0, 1, 2, 4 and 6 h. Checkpoints: 21 C (70 F) by 2 h, 5 C (41 F) by 6 h total (FDA Food Code 3-501.14; NYC Health Code 81.09(e) counts 4 h after the first stage). Design targets: 21 C at 1 h, 5 C at 3 to 4 h. Miss the 2 h checkpoint: reheat to 165 F and restart, or discard, and log which.
9. Pull pans when the center reads 5 C, not when the timer ends. The design load ends here: move the pans to a refrigerator, or to a different freezer if the item is to be frozen. If the item must freeze in this box, that is a second load of about 320 kJ/kg (3.2 MJ per 10 kg, larger than the chill): log it as a freezing run, keep the pans covered, put no PCM on recharge in this box that night, and count the set as unavailable until it has had a full recharge night after the pans read -18 C. Turn the PWM knob down to the 6 V mark (recharge setting) only on a recharge night and note the time. - Tier D (packs). Pull pans when the center reads 5 C. Packs go back to the recharge freezer at the same time, or at 3 h, whichever is first: on edge in their basket, 2 in apart. Write the time on the tag. Then the PWM knob to the 6 V mark, or the fan off; the working box has nothing to recharge.
10. Recharge: jugs out of the tub and spaced 1 in apart on the wire shelf, tub underneath. Overnight at the 6 V mark in a -23 C box, verify by step 1 (log, sentinels, hardness) in the morning; a -18 C box or a static chest is two days or more. One batch per set per day; a second batch the same day only on a second set that passed step 1. - Tier D (packs). Recharge in the dedicated chest freezer only, at its coldest setting, with the box's own 24 h log showing -21 C or colder at the basket. 48 h in a still-air box, which all three priced boxes are; 36 h only at -25 C or colder, or once test (b) has shown the 36 h hold within 15 percent of the 48 h hold. One set is one batch every other day; a second set makes it daily. Never put packs in the working box to "top up".
11. End of day: fan at the 6 V mark overnight only on nights jugs were used (about 0.65 MJ of fan heat, roughly 30 percent of a 10 kg batch's heat, on the compressor that night), off if not. Never leave it at the 9 V mark.
12. Weekly: pull every jug, wipe, inspect base and cap thread, squeeze, check the label and cap tape. Wipe frost off the fan blade, check the lid gland for frost and the cable for chafe at the lid edge, check the coil face for more than a light fur, check the drain pan after defrost, check chest-freezer walls for ice, and open the door with the fan running to confirm the fan stops within a second; if it keeps running the door switch has welded or the relay has stuck, and the fan is out of service until the part is replaced. Monthly: clean the condenser coil (brush and vacuum, power off), and log the empty-box temperature for a night.
13. Leak: gloves. Jug out in the bus tub. Discard any food the brine touched. Thaw, scoop, rinse three times with warm water, dry, quat sanitiser. Never the chlorine bucket. If coolant reached the evaporator housing, the drain, the stock fan or any wiring: unplug the cabinet, tag it out of service, move the food to another unit and hold anything the coolant could have touched for discard, and call a refrigeration technician. Do not rinse the coil. A spray does not show that chloride is out of a fin-and-tube coil (NH4Cl is soldering flux, section 7), it pushes liquid into the fan, the drain and the connectors, and a repair decision does not belong in a spill step. Collect the cleanup liquid, wipes and thawed coolant in a lidded bucket and dispose of them under the SDS accidental-release and disposal sections and local rules (BASF SDS, https://download.basf.com/p1/000000000030046480_SDS_GEN_US/en/Ammonium_chloride_S_foodpharma_grade_Safety_Data_Sheet_USA_English.pdf); nothing goes to the floor drain. Restart only after the technician has inspected the coil, wiring and drain, and the cleanup is written in the log. Arctic Ice pack: bag it, discard food and pans the fill touched, paper towels then warm soapy water then quat, retire the pack, and note it on the SDS request to Arctic Ice.
14. If the compressor clicks on and off within minutes during a chill, stop loading and check the condenser and the room: the overload protector is tripping. Box air above -8 C an hour into a 10 kg chill is expected with 11 to 13 jugs on a small compressor and is not by itself a jug problem; what it does mean is that the inventory is doing the work, so check the sentinel (step 8 log) and cut the next batch if it went above -12 C.

6.7 Validation: an airflow test, then a real batch

Two tests, and only the second one decides anything.

Part 1, airflow. Purpose: find the fan position and confirm the fan reaches the product. It does not say whether the freezer can take a real batch. Same rig as video 1: 800 mL of water at 50 C in a 1 kg stainless pot, probe in the water; a second probe in the box air at pan height; a third probe in a sentinel bottle of water standing among the frozen inventory; the kWh meter on the freezer plug (section 4.2). Before the first run, set the two fan marks on the bench: meter across the fan leads, turn the RioRand knob until the meter reads 9 V and mark the knob, then the same at 6 V (brushed fans only; the Sanyo fan is driven from its own PWM lead, section 6.3a). Write the fixed conditions on the log sheet before the first run and keep them for every run: pot position and shelf, fan position and PWM mark, water mass and start temperature (50 C plus or minus 1), door-open time (open, place, close, timed), same starting inventory, no defrost due inside the run (section 6.5), packs out of the box or the same pack set in the same place every time. For Tiers A to C, jugs in the box with the fan at 6 V the day before. For Tier D, run the two pack tests in section 6.3 first (calorimeter and melt test); they take one day each with the packs in hand and need nothing that is not already bought except a cooler and 20 kg of warm water. Packs conditioned 48 h in the recharge box before run D.

Runs: A (fan off) and B (fan at the 9 V mark), three of each, alternated. Air back at setpoint is not the box back at the same state; the liner, the coil, the inventory and any packs keep absorbing for hours (sections 3.5 and 6.2). So one run per evening, or a second run the same evening only if the sentinel probe has returned to within 0.5 K of its pre-run reading and the compressor is cycling normally. Change the order each day (A then B, B then A, A then B) so neither condition always follows the other. Log every 5 minutes: pot temperature, air, sentinel, watts. For each run record the time from 50 C to 0 C, the air peak, how long the air takes to get back below its pre-run value, sentinel minimum and maximum, whether the compressor ran without a break, and for B runs the fan volts and amps at the 6 V and 9 V marks, running and at cold start (volts x amps is the fan heat in section 4.2 and the load the fuse must carry), and whether the fan started on its own from cold. With the rated Sanyo fan the test is simpler: it must start at -23 C at the idle PWM setting and its tach lead must report speed.

Result: the mean and spread of the three B times against the three A times, written in the log as this box's number. Expect roughly 1.3x to 1.6x faster in a reach-in and about 2x in a chest freezer (section 4.2 and r4). If the three B runs are not all faster than the three A runs, the fan is not reaching the pot: move it and repeat. There is no pass percentage. A run C (fan plus jugs) or run D (fan plus the three Arctic Ice packs on edge in the fan stream, 2 in apart, straight from the recharge box) with the pot is optional and shows only whether the bank changes the air peak against a 190 kJ load. With the 800 mL pot the three packs are enormous relative to the load (1 MJ or more of store against 0.19 MJ of pot), so run D flatters the packs; it says nothing about a 2,200 kJ batch and must not be used to size or buy packs. What run D can show: the face probe on one pack stays within 2 K of -15 C for the whole run; if it climbs past -10 C during an 800 mL run the pack was not charged.

Part 2, the real batch. This is the go/no-go. One product class at a time, at the mass, pan count, fill depth and start temperature the SOP will use, for example 10 kg of stock at 60 C in two 2 in pans filled to 1.5 in. Probe in the center of the thickest pan; in the first run of each product add a second probe at the pan bottom. Sentinel in the inventory. kWh meter on the plug. Everything else per the SOP: packs charged and checked the night before (step 1), defrost schedule checked (step 7), door shut, fan at the 9 V mark, timer set. Log at 0, 1, 2, 4 and 6 h at minimum, every 30 minutes if a second person is available.

Go on all of these, no-go on any one:

• Product: 21 C (70 F) by 2 h and 5 C (41 F) by 6 h total (FDA Food Code 3-501.14; NYC Health Code 81.09(e)), with margin, meaning the design targets in step 8 (21 C by 1 h, 5 C by 3 to 4 h) are met, so that a slower day still passes the legal checkpoints. Rice, beans, braises and thick sauces each need their own run; a liquid result does not transfer to a solid.
• Inventory: sentinel never above -12 C at any point during the run or the recovery.
• Compressor: no on-off cycling shorter than 10 minutes during the run; back to its normal cycle and the box at setpoint within 60 minutes of pulling the pans. Continuous running for more than about 90 minutes on a 10 kg batch is the cue for a technician's readings (section 4.1) before the next run.
• Packs, if used: every pack hard and the pack sentinel below its plateau the next morning on the recharge schedule the SOP allows (step 10), without stretching the recharge night beyond what the box gave in the test. Tier D adds two checks. The pack face probe should still read near -15 C when the pans reach 5 C; if it has climbed past -10 C before the pans reach 5 C, the packs are being spent on a batch they cannot cover, so cut the batch or add thin packs. And the recharge box's 24 h log reads -21 C or colder at the basket (-28 C for the U-tek variant), and the melt-test hold time after 36 h is within 15 percent of the 48 h figure; otherwise the rule is 48 h, and if the box cannot log -21 C it goes back to the seller.
• Repeat: two passing runs on different days before a product class goes into the SOP.

Recharge run, Tiers A to C, three nights before the bank is used for a real batch: full set from room temperature on the rack, logger at the warmest and coldest rack positions, record the time until both sentinels are below -15.4 C by 1 K or more and every sampled jug is hard next morning after a night at 6 V on the open shelf. The longest of the three times plus 2 h becomes the conditioning time on the rack card. If the set does not come back on three nights the box is too warm for this PCM: recharge freezer, or drop the PCM half (KCl only in a cabinet with no frozen inventory). For Tier D the equivalent is test (b) in section 6.3 and the recharge box's own log.

No-go means this freezer, with this fan and this pack count, does not chill this batch safely. In order: halve the batch and re-run; shallower pans (1 in for grains and braises); pre-cool with the paddle or an ice bath before the box; a dedicated chilling cabinet with no frozen inventory; a blast chiller (section 5.1). Do not extrapolate from the pot: it removes about 190 kJ (0.8 kg x 4.19 kJ/kgK x 50 K plus about 25 kJ for the pot), a 10 kg batch removes about 2,200 kJ, and depth, internal convection, pan metal and compressor response do not scale with that ratio.

7. Safety and contamination

Ammonium chloride. GRAS as a direct food ingredient (21 CFR 184.1138, https://www.law.cornell.edu/cfr/text/21/184.1138), E510, the salt in salty licorice. GHS: harmful if swallowed (H302, oral LD50 rat 1,410 mg/kg), eye irritant, not a skin irritant, not a carcinogen (BASF SDS, https://download.basf.com/p1/000000000030046480_SDS_GEN_US/en/Ammonium_chloride_S_foodpharma_grade_Safety_Data_Sheet_USA_English.pdf). A bottle holds 295 g, a few rodent LD50-equivalents for an adult, so label it as a chemical container even though a trace on a pot is a flavor problem. The one reaction to worry about: ammonium salts plus bleach make chloramine gas (https://doh.wa.gov/community-and-environment/contaminants/bleach-mixing-dangers). Spills get water then quat, never the chlorine bucket, and that rule goes on the label. Keep it away from alkaline degreasers too.

Corrosion. NH4Cl is soldering and galvanizing flux; compatibility charts rate copper, brass and cast iron D (severe), 304 stainless C, 316 B, aluminum B only to 72 F (https://www.bellows-systems.com/chemical-compatibility-metals/). A reach-in evaporator is aluminum fin on copper tube; shelving is often galvanized. So: no bottle above or beside the evaporator, HDPE not metal, bus tub as secondary containment, bottom shelf only.

Containers. PET does not fail at -25 C by itself. Its glass transition is 69 to 85 C (https://ctherm.com/resources/newsroom/thermal-analysis-labs/determining-glass-transition-temperature-of-pet/), so it is no more glassy at -25 C than at 20 C, and the PET industry body says there is "no inherent danger" in freezing PET bottles (https://napcor.com/frequently-asked-questions/, seen 2026-08-30). What a Powerade bottle lacks is a rating for anything this job does to it: repeated freeze-thaw cycling of a brine that expands 6 to 8 percent, drop impact while frozen, a cap liner sitting in NH4Cl brine at -15 C, and no published cycle life for any of it. Add that a black-brine sports bottle looks like a drink. Those are the reasons beverage containers stay out of the food freezer, not the polymer. The DIY tiers use amber wide-mouth HDPE lab reagent bottles with a molded label panel (section 6.3): polyethylene "remains impact resistant at low temperatures" (https://www.usplastic.com/knowledgebase/article.aspx?contentkey=954), HDPE is a container material the BASF SDS names, and the bottle cannot be mistaken for food. The same usplastic page lists polypropylene as brittle below freezing, and most lab bottles, the Kritkin 12-pack included, ship with polypropylene caps: check the cap material on the lot you buy, prefer a polyethylene or PE-lined cap if the supplier offers one, and handle cold bottles as if the cap will crack when dropped, because it may. Fill to 90 percent. Mark with paint marker or engraving; freezer tape frosts over and falls off at -20 C. Retirement: number every bottle, tally each freeze-thaw cycle on the label panel, inspect at every thaw, and retire on the first crazing, bulge, weep, cap distortion or thread damage. When the first bottle in a set fails inspection, cap the rest of that set at 80 percent of its cycle count from then on. Until a set has produced that number, this report claims no cycle life for the bottles, because none is published. Beverage containers belong only in a dedicated non-food chest freezer, and only with the same tally and retirement rule.

Non-food in a food freezer. The Food Code defines poisonous or toxic materials as "substances that are not intended for ingestion" and its third category is "substances necessary for the operation and maintenance of the establishment such as nonfood grade lubricants and personal care items that may be deleterious to health" (FDA Food Code 2022, 1-201.10). A home-mixed brine in a reagent bottle is in that category. Four rules then apply (chapter 7 text at https://www.c-uphd.org/documents/eh/2022-FDA-Food-Code-Chapter-7-Poisonous-or-Toxic-Materials.pdf, seen 2026-08-30). 7-201.11: they "shall be stored so they can not contaminate food, equipment, utensils, linens, and single-service and single-use articles by: (A) Separating the poisonous or toxic materials by spacing or partitioning; and (B) Locating the poisonous or toxic materials in an area that is not above food, equipment, utensils, linens, and single-service or single-use articles." 7-202.11(A): "Only those poisonous or toxic materials that are required for the operation and maintenance of a food establishment, such as for the cleaning and sanitizing of equipment and utensils and the control of insects and rodents, shall be allowed in a food establishment." 7-202.12(A): they shall be used according to "law and this Code", "manufacturer's use directions included in labeling" and "additional conditions that may be established by the regulatory authority." 7-101.11 wants a legible manufacturer's label on the container and 7-102.11 the common name on working containers. NYC 81.17(g) wants chemicals "prominently labeled" and stored where food is not stored; an inspector can score 04E/08C at up to 28 points (https://raw.githubusercontent.com/nychealth/Food-Safety-Health-Code-Reference/main/Violation-Health-Code-Mapping.csv). A commercial pack has a manufacturer's label and use directions. A home-mixed bottle has neither, so it cannot meet 7-202.12(A)(2) on paper, and "required for the operation" under 7-202.11 is a judgement the inspector makes, not the operator. The case that a sealed, labeled coolant bottle is operating equipment like a commercial ice pack, kept in its own bin on the bottom shelf with the recipe sheet and SDSs in the HACCP binder, is an argument. It is not an approval, and nothing in this report is one. Close it one of two ways before the first brine bottle goes into a freezer that holds food: ask the local regulator (NYC Department of Health and Mental Hygiene for a city commissary) in writing whether sealed home-mixed coolant containers may be kept in a food freezer under these conditions, and file the answer with the SDSs; or keep every chemical PCM in a dedicated non-food chilling cabinet where food is never stored. Until one of those is done, Tiers A and B are a plan, not a practice. Commercial packs with a manufacturer's label (the owner's Arctic Ice packs in Tier D, Tier C, and the rigid packs in section 5.2) remove the labeling half of the problem and still have to meet the separation and not-above-food rules.

Other ingredients. Bentonite SDSs list 0 to 5 percent crystalline silica, an inhalation carcinogen; wear an N95 when scooping, wet-wipe the dust. Xanthan's SDS warns that a wet spill is "Slippery, can cause falls if walked on"; mop at once with lots of water. Carbon fiber filaments are electrically conductive and "may cause a short circuit in electrical equipment if filaments become airborne" (HiTemp CF51 SDS, https://hitemp.ca/wp-content/uploads/2018/07/CF51_Carbon-Fiber.pdf.pdf); never open the bag near the freezer's control board or the PSU, and the fibers are a black foreign-material contaminant if a bottle fails. Omit both for a commissary build.

Electrical. The 12 V side is not a shock hazard, but it is not a Class 2 circuit either. NEC Chapter 9 Table 11(B) caps a Class 2 DC source at 100 VA and 5 A at 30 V or less, and the source must be listed and marked Class 2 (PULS note: the supply "must be a Listed Class 2 power supply" and "clearly marked as a Class 2 Power Source", https://www.pulspower.com/wp-content/uploads/2023/10/The-PULS-Advantage-Issue-11-NEC-Class-2-Power-Supplies-Aug-2025.pdf; 100 VA, 5 A, 60 V DC per https://www.superbrightleds.com/blog/what-is-class-2-a-guide-to-ul-nec-power-supplies-strip-lights.html; both seen 2026-08-30). At 12 V that is 5 A and 60 W. A 12 V 10 A supply cannot be Class 2, and the Mean Well GST60A12 (12 V 5 A) and GST120A12 (12 V 8.5 A) carry UL 62368-1 marks, not a Class 2 mark (spec sheets, seen 2026-08-30). So treat the fan feed as a low-voltage power circuit: a listed supply from the first trial, an enclosure for the relay, fuse and terminals, and conductors and fuse chosen as in section 6.3a. The earlier statement that the 12 V side "needs nothing" is withdrawn; have an electrician confirm the wiring method for the pass-through and the enclosure. The 120 V side lives outside the box on a wall bracket above mop height, plugged into a GFCI receptacle, which NEC 210.8(B) requires in commercial kitchens and NYC's 2025 Electrical Code (2020 NEC with amendments, in full effect 21 December 2025, https://www.nyc.gov/site/buildings/codes/electrical-code.page) adopts. The Jesverty bench supply carries no UL, ETL or other listing and is dropped from every tier; a commissary is not a laboratory, and the listed supply costs $25 to $57. Do not tap the freezer's internal 120 V wiring for anything. Fit a wire guard wherever the open blade face is reachable. Do not add any electrical device inside an R290 or R600a cabinet without the maker's written approval (True T-Series manual pages 5 and 6, seen 2026-08-30); brushless and IP-rated fans are not exempt.

Frozen inventory. During a chill the box air goes where the compressor and the inventory put it, from about -12 C on a healthy 1/2 HP box to above -5 C on a small one, and the PCM bank shaves only a few kelvin off that (section 4.2). The inventory is the heat sink either way. Keep ice cream, fatty fish and anything with a long planned shelf life in a different freezer, or dedicate a box to chilling.

8. Decision

Recommended tier, in order of what to do:

0. Service the box before buying anything (section 6.1, step 0): read the nameplate and the manual, clean the condenser, check gaskets, look for an iced coil, confirm the setpoint, check the room temperature at the condenser intake, then log the box for 24 h. A reach-in holding -15 to -18 C empty is at the warm edge of most rated ranges, not a proven fault; the model manual and a logged mean decide that (section 6.1, step 0). Put monthly condenser cleaning in the SOP; it is the cheapest thing in this report for freezer life.
1. Fan kit and recharge chest freezer together. The fan is the part that shortens the chill in any freezer, including one holding -15 to -18 C, because the frozen inventory is still the heat sink for a test load. Buy the rated stack of section 6.3a: one Sanyo 9WV1212P1J001 ($101.33), the Mean Well GST120A12-R7B ($57.27), the door switch, relay and diodes ($14.58), timer ($11.99), 14/2 cable and gland (about $38, not confirmed), fuse and holder ($8 placeholder), a 120 mm guard ($5 placeholder): $236.39, call it $240, or $342.72 (about $345) with a second fan and guard for a reach-in; add a bus tub and freezer tape ($9.08) if the shelf needs a drip tray. The GESEXI video fan ($39.59) is only for a dedicated cabinet the owner accepts as altered, and only on a non-hydrocarbon box. On an R290 or R600a cabinet there is no DIY fan; get the maker's written approval or go to the blast chiller line at the end of this section. A reach-in also has no owner-installable cable route: the maker locates the lines in writing and a technician fits a bulkhead, or the reach-in gets the swap bank and the paddles without a fan (section 6.5). The recharge chest freezer is what makes the packs usable at all: Galaxy CF7 at $298.99 if you are willing to log it and return it, Atosa MWF9007 at about $506 to $543 (price not confirmed on a live page on 2026-08-30) as the box rated to the line, Galaxy Ultra Cold UCCF8 at $1,369.00 if you want no doubt, and it is the one commercial box under $1,500 whose rating also covers the -28 C floor of the U-tek variant (the $288 Northair is household grade and must prove itself on a logged recharge). Log the box 24 h at the coldest setting with the basket in it; keep it only at -21 C or colder.
2. Pack tests, before any purchase beyond a cooler: the calorimeter (latent heat) and the melt test (conditioning time) in section 6.3, Tier D. They settle the two numbers Arctic Ice does not publish and cost a day each.
3. Validation: the airflow test in section 6.7 (runs A and B, three each, alternated), run D with the three packs, then the real-batch test in the same section before any product goes through the box on the SOP.
4. Then the swap bank in production: one 10 kg batch every other day on one set, daily once a second set of three is bought (about $130 to $140 when the XX-L is back in stock). Packs 4 and 5 for skin, not capacity: two Engel XL 5 lb at $29.99 each. The -23 C U-tek variant of Tier D is the upgrade if the packs measure well but take too little of the chill on a healthy box, and only with a recharge box that has logged -28 C or colder at its warmest spot: about $905 to $1,045 with the $288 Northair box, about $1,985 to $2,125 with the UCCF8; buy one case, run the variant's melt test and the logged recharge run, and buy the other three cases only on those results. The U-tek packs are never stored in the working box. Fallbacks if the Arctic Ice packs measure under 220 J/g or the recharge box cannot hold -21 C: Tier C (Engel packs, one XL at $29.99 first for the melt test, then four more, $149.95 in all) on a reach-in you do not want chemistry in, packs only if the warranty matters, with a recharge box logging -24 C or colder at the pack position, or corrected-recipe NH4Cl in labeled HDPE lab bottles (Tier A hardware with lab bottles, salt about $13 to $23 per 10 kg batch, about $140 for bottles, salt, gum and bentonite) on a box you do not mind altering, and only once the regulator question in section 7 has a written answer or the box is a dedicated non-food cabinet. Tier B with KCl recharges anywhere, but its -10.7 C plateau cannot keep the inventory sentinel below -12 C, so it is only for a cabinet that holds no frozen inventory during the chill.
5. Cover pans, use 2 in pans, and use the cooling paddles for the first 30 minutes regardless of tier. A full 4 L paddle can take up to about 1.6 MJ of a 10 kg batch's 2.2 MJ; how much it takes in 30 minutes is a measurement, not a promise, so log the drop per batch type (section 5.3) and let the measured figure reduce the PCM count and the freezer's load. The pans then go in cooler and steam less, which is where the frost saving comes from. It costs nothing you do not already own, apart from the paddle's own refreeze.

Cost of the recommended path: nothing but an hour for step 0; about $240 to $345 for the rated fan stack (or about $195 to $205 with the video-copy fan and PWM on a dedicated altered non-hydrocarbon cabinet) plus $299 to $1,369 for the recharge box in step 1, about $525 to $1,745 all in with the logger (Tier D cost table); two days and a cooler for step 2; $60 to $140 for step 4's extra packs. The packs themselves are sunk. For the fallbacks, Tiers A to C on the printed BOMs are $442 to $465 for a 10 kg batch and $652 to $693 for 30 kg, before shipping and tax, or $427 to $487 and $672 to $714 with the rated fan of section 6.3a; a second rated fan adds $101.33 plus a $5 guard. Tier D with the U-tek variant is about $905 to $1,045 for 10 kg with the $288 household -40 F box, or about $3,995 to $4,135 with the So-Low lab chest; at 30 kg Tier D costs about what the Avantco SF-5 does and the SF-5 wins.

Go/no-go for the experiment as a whole. The pot test settles only where the fan goes. The decision is the real-batch test in section 6.7: with the fan at 9 V and the packs charged, a batch at the SOP's mass, pan count and fill depth reaches 21 C (70 F) by 2 h and 5 C (41 F) by 6 h with margin, the inventory sentinel never rises above -12 C, the compressor does not cycle shorter than 10 minutes, the bank is charged again on the schedule the SOP allows (Tiers A to C: SOP step 1 passes the next morning, meaning the recharge log held below the set's floor for the validated time, both sentinels read at least 1 K below the plateau, -16.4 C for NH4Cl, and the sampled jugs are hard; Tier D: the packs had the label's tested recharge time, 48 h, or 36 h if the melt test after 36 h matched the 48 h figure within 15 percent), and the run passes twice on different days. No-go on any of those means this freezer does not chill that batch; the fixes in order are a smaller batch, shallower pans, a paddle or ice bath before the box, a dedicated chilling cabinet, or a dedicated unit: the Avantco SF-5 at $3,789 (31 lb per 90 min, 120 V, ships in a day) or the Atosa EBF-05 at $5,707.00 list (55 lb chill and 22 lb freeze on the maker's 2026 spec sheet, R-290, 3 year parts and labor per the 2026 manual; CKitchen showed a $5,187.60 promotional price on 2026-08-30), or a used 5-pan Traulsen at $3,500 to $4,500. Those are the only options that take the load off the storage freezer instead of shifting it to overnight, and if daily volume grows past what section 6.2 allows, that is where this ends up anyway.

Review log

Three adversarial reviewers (thermodynamics, refrigeration, procurement) returned 34 findings on the 2026-08-30 draft. Each was recomputed or re-fetched on 2026-08-30 before the change was made. Numbering follows the order the findings arrived in. None was rejected outright; where a finding was only partly right the entry says so. The same information is in review-log.json.

Codex review log

A second, independent red-team by Codex returned 30 findings on the once-revised draft (redteam-codex.md: 5 blocker, 17 major, 4 minor, 4 defensible). Each was recomputed or re-fetched on 2026-08-30 before the change was made; "accepted in part" rows say what was rejected and why. Between the Codex review and this revision the owner also bought three Arctic Ice Tundra XX-L packs, which is why Tier D is built around a -15 C pack in hand and treats the -23 C U-tek of finding 1 as its colder variant. The same rows are in codex-review-log.json. A final consistency pass on 2026-08-30 then cleared two residuals (finding 2 in the chest-freezer layout, finding 19 in the section 8 go/no-go), put the RioRand PWM back into Tiers A and B so the 9 V and 6 V settings exist, reconciled the fan-kit and BOM totals, set one Engel recharge floor (-24 C logged mean) and made 48 h the printed pack-recharge default; rows 2, 19 and 23 note it.

9. Sources

All seen 2026-08-30 unless noted.

1. Video 1, fan: https://www.youtube.com/watch?v=G1Syjh319VU
2. Video 2, eutectic: https://www.youtube.com/watch?v=IGlZ61oCcEY
3. Fan listing, GESEXI 16 in 2-pack: https://www.amazon.com/dp/B0DHKQS3DC
4. GESEXI 16 in single: https://www.amazon.com/dp/B0BNZT17TL
5. GESEXI 12 in single: https://www.amazon.com/dp/B0BNZRBP98
6. Jesverty SPS-3010 PSU: https://www.amazon.com/dp/B09YSJQWRG
7. SHNITPWR 12 V 10 A brick: https://www.amazon.com/dp/B07PWZQ33N
8. RioRand PWM controller: https://www.amazon.com/dp/B007TH4EN6
9. PHc 9 V 5 A adapter: https://www.amazon.com/dp/B006QZ7YRU
10. SPAL 30100467: https://www.amazon.com/dp/B019XEAIUU and https://www.spal.com/en/axial-fans
11. Noctua NF-A14 iPPC IP67: https://www.amazon.com/dp/B00KFCR0PQ (withdrawn per red-team finding 4; kept for the record)
12. ebm-papst iQ 3612 motors: https://www.amazon.com/dp/B07Y8TY2GQ ($117.31, only 20 left); B06XQLC9N5 was "Currently unavailable" on 2026-08-30 and is no longer used for a price; datasheet https://ebmpapst.se/sv/dat/media_manager/pid/2970/product-datasheet/datasheet.pdf
13. Mean Well GST60A (12 V 5 A) and GST120A (12 V 8.5 A) spec sheets, UL 62368-1, hiccup overload, no Class 2 mark: https://www.meanwell.com/Upload/PDF/GST60A/GST60A-SPEC.PDF, https://www.meanwell.com/Upload/PDF/GST120A/GST120A-SPEC.PDF; prices $24.86 and $57.27 at https://www.trcelectronics.com/products/mean-well-gst60a12-p1j and https://www.trcelectronics.com/products/mean-well-gst120a12-r7b (2026-08-30)
14. NH4Cl eutectic composition 19.7 wt%: Peppin, Huppert, Worster, J. Fluid Mech. 599 (2008), https://doi.org/10.1017/S0022112008000219. (The MDPI Materials 2024 paper at https://doi.org/10.3390/ma17112768, cited in an earlier draft, is a dendrite-kinetics study and does not state the eutectic values.)
15. NH4Cl-70 wt% H2O solidification: https://www.sciencedirect.com/science/article/pii/S001793102033502X
16. NH4Cl solubility and properties: https://en.wikipedia.org/wiki/Ammonium_chloride and https://en.wikipedia.org/wiki/Solubility_table
17. Heat of fusion of ice, Osborne 1939: https://nvlpubs.nist.gov/nistpubs/jres/23/jresv23n6p643_A1b.pdf
18. NH4Cl enthalpy of solution: https://chem.libretexts.org/Courses/Howard_University/General_Chemistry:_An_Atoms_First_Approach/Unit_4:__Thermochemistry/09:_Thermochemistry/Chapter_9.05:_Enthalpies_of_Solution
19. PCM Products PlusICE range (E-15): https://www.pcmproducts.net/files/PlusICE%20Range%202021-1.pdf
20. Montmorillonite ice nucleation (droplet immersion-freezing data): https://acp.copernicus.org/articles/12/5859/2012/; supercooling of brine PCMs and nucleating agents: https://www.mdpi.com/2504-477X/9/12/667
21. NH4Cl brine density: https://advancedthermo.com/electrolytes/density_NH4Cl.html
22. Toray fiber data sheets: https://www.toraycma.com/wp-content/uploads/T300-Data-Sheet.pdf, https://www.toraycma.com/wp-content/uploads/M40J-Data-Sheet.pdf, https://www.toraycma.com/wp-content/uploads/M60J-Data-Sheet.pdf
23. Percolating carbon fiber PCM composite: https://www.sciencedirect.com/science/article/abs/pii/S0306261915006522
24. Frusteri et al., carbon fibers in salt-hydrate PCM: https://www.sciencedirect.com/science/article/abs/pii/S1359431104002984
25. Xanthan in brine: https://www.jungbunzlauer.com/en/products/biogums/xanthan-gum
26. Convective heat transfer coefficients: https://www.engineeringtoolbox.com/convective-heat-transfer-d_430.html
27. Metal compatibility with ammonium chloride: https://www.bellows-systems.com/chemical-compatibility-metals/
28. Tecumseh Hermetic Compressor Service Handbook: https://www.tecumseh.com/userfiles/documents/marketing-brochures/tecumseh-service-handbook_na_en.pdf
29. Tecumseh Quick Select Guide (capacities at -10 F): https://www.sidharvey.com/documents/Tecumseh-Catalog_Broc.pdf
30. Embraco, LBP/MBP/HBP applications: https://www.embraco.com/en/blog/technical-information/lbp-mbp-and-hbp-applications.html
31. Copeland AE17-1262, short cycling: https://webapps.copeland.com/online-product-information/Publication/LaunchPDF?Index=AEB&PDF=1262
32. Copeland discharge line temperature limits: https://copeland.custhelp.com/app/answers/detail/a_id/4578/
33. Jamie Kitchen (Danfoss), compressor current: https://www.linkedin.com/pulse/interpreting-current-readings-compressors-jamie-kitchen
34. True T-series installation manual: https://www.webstaurantstore.com/documents/pdf/tseries_manual.pdf
35. True T-23F-HC: https://www.truemfg.com/product/t-23f-hc/ and spec sheet https://d3ld6frh4bdurh.cloudfront.net/pdf/598-T23F.pdf
36. Turbo Air M3 manual: https://turboairinc.com/wp-content/uploads/2020/01/M3-Reach-Ins_HC.pdf
37. Hoshizaki S-series manual: https://secure.hoshizakiamerica.com/docs/manuals/Commercial_S_Series_inst.pdf
38. Traulsen G-series manual: https://assets.katomcdn.com/raw/upload/v1616526910/products/206/206-G10000/206-G10000_manual.pdf
39. Beverage-Air HF manual: https://beverage-air.com/content/manuals/809-143A_MANUAL_INSTALLATION_AND_OPERATION_BEV_AIR_HF.pdf
40. True defrost heater listing (950 W, 220 V, multi-door cabinet part, $509.49; not the T-23F heater): https://www.amazon.com/dp/B06XQKVD4C
41. Avantco CF8 chest freezer spec: https://www.webstaurantstore.com/documents/specsheets/178cf8_specsheet.pdf
42. ASHRAE Refrigeration Handbook ch. 8: https://consultancyengineering.com/wp-content/uploads/2023/12/CH-08-Thermal-Properties-of-Foods.pdf
43. Codex CXC 8-1976: https://www.fao.org/input/download/standards/285/CXP_008e.pdf
44. IIR -18 C vs -15 C position paper: https://iifiir.org/en/news/iir-position-paper-in-ijr-frozen-at-18-c-or-15-c-a-critical-review-of-subzero-storage-quality-and-regulation
45. Gasket frost: https://www.gasketguy.com/blog/heater-wires
46. Display cabinet shelf velocity: https://www.sciencedirect.com/science/article/abs/pii/S0140700714000334
47. Alto-Shaam QC3-40 spec: https://cdn.beedash.com/AltoShaam/3042ef891b127d41b813f388d0e8c998af3afca9.pdf
48. Blast freezer air velocity guide: https://ingener.by/refrigeration-systems/food-processing-refrigeration/frozen-food-processing/blast-freezers/air-velocity-requirements/
49. Avantco SF-3: https://www.webstaurantstore.com/avantco-sf-3-25-countertop-blast-chiller-freezer-15-lb-9-lb/449SF3.html
50. Avantco SF-5: https://www.webstaurantstore.com/avantco-sf-5-30-countertop-blast-chiller-freezer-31-lb-26-lb/449SF5.html
51. Avantco SF-10: https://www.webstaurantstore.com/avantco-sf-10-31-reach-in-blast-chiller-freezer-88-lb-77-lb/449SF10.html
52. KoolMore 3 pan: https://www.amazon.com/dp/B0DMWPWCYR
53. Atosa EBF-03: https://www.nishienterprise.com/products/atosa-ebf-03-blast-chillerfreezer
54. Atosa EBF-05: https://www.ckitchen.com/p/atosa-ebf-05-blast-chiller-freezer.html; maker's page https://atosausa.com/product/ebf-05-blast-chiller-freezer/ 54a. Atosa 2026 blast chiller spec sheet: https://atosausa.com/wp-content/uploads/2025/11/ATOSA-USA-SPEC-SHEET-2026-Blast-Chillers.pdf 54b. Atosa EBF series manual (2026-08-14): https://atosausa.com/wp-content/uploads/2026/08/EBF-Series-Manual-08142026-1.pdf
55. Techfrost JOF23: https://www.webstaurantstore.com/techfrost-jof23-24-blast-chiller-freezer-15-lb-115-120v/774JOF23.html
56. Techfrost E5: https://www.webstaurantstore.com/techfrost-e5-33-blast-chiller-freezer-44-lb-208-230v/774E5.html
57. Infrico IBC-ABT51L: https://www.webstaurantstore.com/infrico-ibc-abt51l-33-countertop-blast-chiller-freezer-115v/18CBCABT51L.html
58. Alto-Shaam QC3-3: https://www.webstaurantstore.com/alto-shaam-qc3-3-27-quickchiller-reach-in-commercial-countertop-blast-chiller-36-lb/131QC33A.html
59. Traulsen CLBC4: https://www.webstaurantstore.com/traulsen-clbc4-l-c-series-54-1-8-left-hinged-undercounter-blast-chiller-80-lb-115v/882CBC4L.html
60. Traulsen TBC5-50: https://www.webstaurantstore.com/traulsen-tbc5-50-spec-line-undercounter-5-pan-blast-chiller-right-hinged-door-with-6-casters/882TBC550.html
61. Beverage-Air RBS-054-SA-HC: https://www.webstaurantstore.com/beverage-air-rbs-054-sa-hc-33-1-2-countertop-blast-chiller-freezer-208-240v/185RBS54SAHC.html
62. WebstaurantStore blast chiller category (Irinox, Delfield, Techfrost prices): https://www.webstaurantstore.com/13657/commercial-blast-chillers.html
63. eBay Beverage-Air CF031AG: https://www.ebay.com/p/14027587147
64. eBay Traulsen RBC/TBC: https://www.ebay.com/p/1938694989
65. MS Restaurant Equipment Sales, Traulsen RBC50: https://www.msrestaurantequipmentsales.com/products/traulsen-rbc50-zwm12-50lb-capacity-undercounter-blast-chiller-tested-works
66. BidOnEquipment Master-Bilt: https://www.bid-on-equipment.com/freezers-and-coolers/batch-freezers-and-coolers/freezing-and-cooling-boxes/135196~mb-master-bilt-blast-chiller-freezer.htm
67. EPA HFC technology transitions: https://www.epa.gov/hfcs/technology-transitions-hfc-restrictions-sector
68. Irinox EasyFresh manual: https://caterlink-web.s3.ap-southeast-2.amazonaws.com/public/partdocs/Irinox-EasyFresh-Blast-Chiller-User-Manual.pdf
69. Crestmont Capital leasing guide: https://www.crestmontcapital.com/blog/commercial-kitchen-equipment-leasing-a-smart-financing-strategy-for-foodservice-businesses
70. Engel 5 F / -15 C packs: https://engelcoolers.com/products/engel-5-f-15-c-freezer-packs
71. Arctic Ice Tundra: https://arctic-ice.com/products/tundra-series and https://www.amazon.com/dp/B0DZK515BK
72. Sonoco/Fisher PureTemp -20 C brick: https://www.fishersci.com/shop/products/sonoco-puretemp-phase-change-materials-foam-brick-phase-temp-20-deg-c-l-x-w-x-h-10-75-x-9-25-x-0-5-in/03600110
73. PureTemp FAQ and -15 datasheet: https://puretemp.com/home/faqs/ and https://puretemp.com/puretemp-15technicaldatasheet/
74. Cooler Shock: https://www.amazon.com/dp/B086ZY99JC
75. Polar Tech Ice Brix: https://www.webstaurantstore.com/polar-tech-48-oz-ice-brix-low-temperature-cold-pack-case/620XB48CS12.html
76. Nordic Ice: https://www.webstaurantstore.com/49203/hot-cold-packs.html?vendor=Nordic-Ice
77. YETI Ice: https://www.tylerstx.com/yeti-ice/; Pelican Ice: https://www.foamerica.com/products/pelican-2lb-icepack; Igloo Maxcold: https://www.igloocoolers.com/products/25201-maxcold-ice-large-freeze-block-blue; Fit & Fresh: https://fit-fresh.com/products/cool-coolers-xl-ice-multicolor
78. Rubitherm SP range: https://www.rubitherm.eu/en/productcategory/anorganische-pcm-sp
79. PLUSS savE range: https://pluss.us.com/save-pcm-product-range/
80. Frigolab eutectic plate: https://www.frigolab.eu/gb/option-and-accessories/842-eutectic-plate-cold-gn-11-18c.html
81. Kangabox plates: https://www.kangaboxuk.com/hot-and-cold-eutectic-plates-24-c.asp
82. NaCl eutectic measurements, Yilmaz et al.: https://proceedings.ises.org/conference/eurosun2010/papers/eurosun2010-0314-Medrano.pdf
83. KCl eutectic for cold storage: https://www.researchgate.net/publication/350336961
84. Ammonium chloride 20 lb in 5 lb bags, $56.99, in stock: https://www.amazon.com/dp/B00L2158N0; 50 lb Wintersun feed grade, $250.00, in stock: https://www.amazon.com/dp/B00N48Q7XY; 1 kg food-additive grade, "Currently unavailable" on 2026-08-30: https://www.amazon.com/dp/B01EZ921NA
85. Potassium chloride, food grade USP, 5 kg $95.99: https://www.amazon.com/dp/B0F8W9GJ43 (listing states USP, not FCC; no manufacturer, CoA or SDS shown; page returned an error on 2026-08-30 re-check); 1 kg $23.99: https://www.amazon.com/dp/B0F8WQR18Q; Morton 40 lb water-softener pellets (not FCC grade, third-party offers $39.99 to $66.95, no buy box): https://www.amazon.com/dp/B008Z27QIY
86. 21 CFR 184.1138 (NH4Cl GRAS): https://www.law.cornell.edu/cfr/text/21/184.1138; 184.1622 (KCl): https://www.law.cornell.edu/cfr/text/21/184.1622; 182.1 (NaCl): https://www.law.cornell.edu/cfr/text/21/182.1
87. BASF ammonium chloride SDS: https://download.basf.com/p1/000000000030046480_SDS_GEN_US/en/Ammonium_chloride_S_foodpharma_grade_Safety_Data_Sheet_USA_English.pdf
88. Bleach and ammonia: https://doh.wa.gov/community-and-environment/contaminants/bleach-mixing-dangers
89. Bentonite SDS: https://www.westernclay.com/wp-content/uploads/2017/01/SAFETY-DATA-SHEET.pdf
90. Xanthan SDS: https://sds.chemtel.net/docs/Asbury%20Carbons%20Inc-0001931/Site%20SDS/CP%20Kelco%20U.S.%20Inc_Kelzan%20(GHS)_Unknown_06-27-2017_English.pdf
91. Carbon fiber SDS: https://hitemp.ca/wp-content/uploads/2018/07/CF51_Carbon-Fiber.pdf.pdf
92. Freezer-safe plastics: https://www.usplastic.com/knowledgebase/article.aspx?contentkey=954
93. PET glass transition: https://ctherm.com/resources/newsroom/thermal-analysis-labs/determining-glass-transition-temperature-of-pet/
94. FDA Food Code 2022: https://www.fda.gov/media/164194/download
95. NYC Health Code Article 81: https://www.nyc.gov/assets/doh/downloads/pdf/about/healthcode/health-code-article81.pdf
96. NYC DOHMH violation mapping: https://raw.githubusercontent.com/nychealth/Food-Safety-Health-Code-Reference/main/Violation-Health-Code-Mapping.csv
97. NEC 210.8(B) text: https://www.electricallicenserenewal.com/Electrical-Continuing-Education-Courses/NEC-Content.php?sectionID=1425
98. NYC Electrical Code: https://www.nyc.gov/site/buildings/codes/electrical-code.page
99. PULS Class 2 note: https://www.pulspower.com/wp-content/uploads/2023/10/The-PULS-Advantage-Issue-11-NEC-Class-2-Power-Supplies-Aug-2025.pdf (quoted in section 7: Class 2 source must be listed and marked; 100 VA and 5 A at 30 V or less)
100. ThermoWorks ThermoPop 2: https://www.thermoworks.com/products/thermopop-2
101. YoLink YS8004: https://shop.yosmart.com/products/ys8004
102. Cooling paddles: https://www.webstaurantstore.com/3071/cooling-paddles.html
103. GRI 2020-12 magnetic contact (0.5 A, 10 W, 200 VDC): datasheet https://www.grisk.com/wp-content/uploads/mdocs/2020%20Series_.375%20In.%20Recessed.pdf, $3.09 at https://www.jmac.com/GRI_2020_12_B_p/GRI-2020-12-B.htm; Bosch 0332209150 coil 160 mA, 69 to 80 ohm: https://www.texasindustrialelectric.com/relays_0332209150.asp; TEMCo CN0170 1.8 W coil: https://temcoindustrial.com/bosch-style-automotive-relay-cn0170-1-qty-12-v-30-40-amp-spdt/; True MODS 12 V 30/40 A relay kit, $9.99: https://www.amazon.com/dp/B01N66W2XF; 1N4001 10-pack $1.50: https://www.adafruit.com/product/755. Adafruit 375 (100 mA) withdrawn: https://www.adafruit.com/product/375
104. BN-LINK timer: https://bn-link.com/search?q=countdown+timer
105. MoreWine bentonite (2 oz $1.99; 8 oz and larger available; drop-shipped, 1 to 4 weeks; free shipping only over $79, so a $1.99 order pays shipping): https://morewine.com/products/bentonite-volclay-kwk-krystal-klear
106. WebstaurantStore search pages for xanthan, jugs, caps, bus tubs, cable ties, freezer tape, pastry bags, funnel, bulk salt: https://www.webstaurantstore.com/search/xanthan-gum.html, https://www.webstaurantstore.com/search/1-quart-hdpe-jug-with-cap.html, https://www.webstaurantstore.com/search/38mm-milk-jug-cap.html, https://www.webstaurantstore.com/search/bus-tub.html, https://www.webstaurantstore.com/search/cable-ties.html, https://www.webstaurantstore.com/search/freezer-tape.html, https://www.webstaurantstore.com/search/piping-bag-18-inch-disposable.html, https://www.webstaurantstore.com/search/plastic-funnel-wide-mouth.html, https://www.webstaurantstore.com/search/bulk-salt-50-lb.html
107. Arctic Air AWF25 reach-in: https://www.webstaurantstore.com/arctic-air-awf25-white-single-door-reach-in-freezer-25-cu-ft/123AWF25.html
108. Atosa MWF9016GR chest freezer: https://therestaurantwarehouse.com/products/mwf9016gr
109. Kritkin amber HDPE 1,000 mL wide-mouth lab bottles, 12-pack $39.99: https://www.amazon.com/dp/B0DP618KRP; Nalgene 1 L HDPE wide-mouth, case of 50, $255.00: https://www.amazon.com/dp/B0055V3R2M
110. ThermoSafe U-tek -23 C gel pack, 48 oz, pack of 6, $88.00: https://www.fishersci.com/shop/products/thermosafe-u-tek-refrigerant-packs-3/035257G
111. Webstaurant 1 qt HDPE milk jug (38-DBJ neck) and 500JBC tamper-evident caps: https://www.webstaurantstore.com/1-quart-32-oz-translucent-hdpe-squat-milk-jug-with-handle-162-case/500mbsq32.html; https://www.webstaurantstore.com/search/500jbc.html

112. TableCraft 3838 cap (38-400 thread, does not fit the DBJ jug): https://www.webstaurantstore.com/tablecraft-3838-solid-white-storage-cap-for-squeeze-bottles-with-a-38-mm-opening/8083838.html

113. ThermoSafe PCM product table (model 427, 48 oz, -23 C; no latent heat published): https://www.thermosafe.com/products/refrigerants/pcm/ and https://wordpresssonoco.azurewebsites.net/products/refrigerants/pcm/

114. ThermoSafe U-tek 2013 MSDS (SG 1.16, water soluble, no composition): https://www.thermosafe.com/wp-content/uploads/2021/03/MSDSU-tek-10F010113.pdf
115. ThermoSafe U-tek 427 on Amazon, $83.22 third party, from $75.88, non-returnable: https://www.amazon.com/dp/B00C2T41CK; current Fisher URL: https://www.fishersci.com/shop/products/thermosafe-u-tek-phase-change-materials-6/035257G
116. Northair 3.5 cu ft 14 F to -40 F chest freezer, $287.98: https://www.amazon.com/dp/B09PBDPHZY; $289.90 to $299.90: https://northaircool.com/products/3-5-cu-ft-low-temperature-freezer
117. So-Low CH40-5 ($3,376.00) and CH40-9 ($4,200.00) -40 C lab chest freezers: https://glesales.com/products/so-low-ch40-5-40c-mini-freezer-5-0-cu-ft-115v; https://glesales.com/products/so-low-ch40-9-40c-mini-freezer-5-0-cu-ft-115v
118. Elcold 41 in -40 Extra Cold chest freezer, R290, $3,900.00: https://freezerplanet.com/products/elcold-41-solid-chest-freezer-40-extra-cold
119. FreezerLink -40/-45 C chest freezers, $5,880.00 and $6,555.00: https://www.freezerlink.com/chest-freezer--40c-45c-super-cold.html
120. Master-Bilt IHC-27 hardening cabinet, used, $8,995.00: https://www.visionequipment.com/product/master-bilt-ihc-27-ice-cream-hardening-cabinet-2/
121. Global Refrigeration T30HSP-L hardening cabinet, lease $355.16 per month: https://leasetaurant.com/t30hsp-l-31-ice-cream-hardening-cabinet/
122. Arctic Ice Tundra XX-L dimensions graphic (10.5 x 10.5 x 3.5 in, 10 lb) and FAQ (36 h conditioning): https://cdn.shopify.com/s/files/1/0011/1647/1361/files/AITundraSeriesPacksDimensionsGraphic.png; https://arctic-ice.com/pages/faqs
123. Pelican Supply, Arctic Ice Tundra 10 lb single, $46.29, sold out: https://www.pelicansalessupply.com/products/arctic-ice-tundra-series-10-lb
124. Arctic Ice Tundra Amazon Q&A (page returned 503; quoted from the search result): https://www.amazon.com/ask/questions/Tx2ARLP4RKIZR38/; Amazon listings B088ZH89LG and B09GWRNGQ4 returned HTTP 500
125. Galaxy CF7 chest freezer, 7 cu ft, R-600a, -8 to 0 F, $298.99: https://www.webstaurantstore.com/galaxy-cf7-commercial-chest-freezer-7-cu-ft/177CF7.html
126. Galaxy Ultra Cold UCCF8 ($1,369.00) and UCCF5 ($1,109.00), -50 to -4 F, R-290: https://www.webstaurantstore.com/ultra-cold-chest-freezer-8-cu-ft-steel-and-plastic-white-powder-coated/177UCCF8.html; https://www.webstaurantstore.com/galaxy-ultra-cold-uccf5-5-1-cu-ft-white-chest-freezer/177UCCF5.html
127. Atosa MWF9007 chest freezer, -10 to -1 F, R600a: https://www.gofoodservice.com/p/atosa-mwf9007 ($506, "no longer available for purchase"); https://therestaurantwarehouse.com/products/mwf9007 (specs, no price)
128. Atosa MWF9007 / MWF9010GR / MWF9013GR / MWF9016GR manual (R600a and R290, drain plug, liability exclusions): https://www.atosausa.com/manuals/Manual_MWF9007_MWF9010GR_MWF9013GR_MWF9016GR_Atosa.pdf
129. True T-Series manual TEC_TM_121 REV. D (pages 5 and 6: no modification, no electrical appliance inside, contact True before drilling): https://www.truemfg.com/wp-content/uploads/2023/12/T_SERIES_ENGLISH.pdf
130. Sanyo Denki San Ace 9WV1212P1J001 splash-proof fan: https://products.sanyodenki.com/en/sanace/dc/splash-proof-fan/9WV1212P1J001/; $101.33 at https://www.sager.com/product/9wv1212p1j001-sanyo-denki/3916291
131. ebm-papst 4114N/2H8P at DigiKey, $221.26 backorder, IP rating not confirmed: https://www.digikey.com/en/products/detail/ebm-papst-inc/4114N-2H8P/1860290
132. Super Bright LEDs Class 2 guide: https://www.superbrightleds.com/blog/what-is-class-2-a-guide-to-ul-nec-power-supplies-strip-lights.html
133. Southwire 14/2 low-voltage cable 55213242: https://www.southwire.com/wire-cable/pump-irrigation/landscape-low-voltage-150-volt-lighting/p/55213242
134. Essentra CG-PG9-1-BK PG9 gland at DigiKey ($1.22 in a search listing, page not reachable): https://www.digikey.com/en/products/detail/essentra-components/CG-PG9-1-BK/3811770
135. GE support, chest freezer condenser and evaporator tubing location: https://products.geappliances.com/appliance/gea-support-search-content?contentId=18308
136. Jim's Beer Kit forum, chest-freezer lid drilling: https://www.jimsbeerkit.co.uk/forum/viewtopic.php?t=76490
137. Flat Starlink-type DC cable, 20 AWG (withdrawn as a cable route): https://www.amazon.com/dp/B0H395T36Y
138. HiLetgo opto-isolated 12 V relay module (price not confirmed): https://www.amazon.com/dp/B01LW15D1M
139. True T-23F-HC current product page: https://www.truemfg.com/products/t-23f-hc
140. Turbo Air M3F24-1-N (-10 to 0 F): https://www.turboairinc.com/products/m3f24-1-n/
141. Atosa MBF8501GR (-8 to 0 F) and MWF9016GR (-10 to -1 F) at Equippers: https://www.equippers.com/atosa-mbf8501gr-19-1-cu-ft-1-door-bottom-mount-reach-in-freezer-27-w; https://www.equippers.com/atosa-mwf9016gr-15-9-cu-ft-solid-top-chest-freezer-60-1-8-w
142. Avantco A-19F-HC and A-23F-HC (-8 to -1 F; FAQ on hot food): https://www.webstaurantstore.com/avantco-a-19f-hc-29-solid-door-reach-in-freezer/178A19FHC.html; https://www.webstaurantstore.com/avantco-a-23f-hc-29-solid-door-reach-in-freezer/178A23FHC.html
143. Beverage-Air HBF23HC-1 (down to -10 F): https://beverage-air.com/HBF23HC-1-Horizon-Bottom-Mount-Solid-Door-Reach-In-Freezer
144. Hoshizaki Commercial S Series manual (hot food warning, defrost logic; same document as source 37): https://hoshizaki.com/docs/manuals/Commercial_S_Series_inst.pdf
145. P3 Kill A Watt P4400: https://www.p3international.com/products/p4400.html; $29.99 at https://www.circuitspecialists.com/p4400_kill_a_watt_energy_meter
146. FDA Food Code 2022, chapter 7 text: https://www.c-uphd.org/documents/eh/2022-FDA-Food-Code-Chapter-7-Poisonous-or-Toxic-Materials.pdf
147. FDA Food Code 2022, full text (definition 1-201.10): https://resources.finalsite.net/images/v1677677402/westhartfordctgov/mjnko1bu5at9njpwuueb/2022FDAFoodCode.pdf (the fda.gov/media/164194 link returned 404 on 2026-08-30)
148. OSTI, stabilizing a low temperature phase change material based on Glauber's salt: https://www.osti.gov/servlets/purl/2588542
149. NAPCOR FAQ on freezing PET bottles: https://napcor.com/frequently-asked-questions/
150. Sonoco ThermoSafe PureTemp -21 rigid bottles at Fisher, B27V $1,195.00 per 16 and B38V $1,079.00 per 12: https://www.fishersci.com/shop/products/puretemp-phase-change-materials-pcm/03600180; https://www.fishersci.com/shop/products/puretemp-phase-change-materials-pcm/03600181
151. Temprecision ThermaBrick TEMP-16 TDS (333 J/g) and frozen-range PCM page: https://www.temprecision.com/wp-content/uploads/PCM-16C3-1.pdf; https://www.temprecision.com/phase-change-materials/frozen-temperature-pcms-life-sciences/
152. Cryopak 20 Below and Sub-Zero: https://www.cryopak.com/solutions/refrigerants/phase-change-materials/20-below/; https://www.cryopak.com/solutions/refrigerants/phase-change-materials/sub-zero/
153. Insulated Products Corp -20 C PCM packs: https://ipcpack.com/products/gel-packs/20c-pcm-packs/
154. savENRG HS26N: https://savenrg-pcm-pouch.com/products/savenrg-ice-pack
155. Codex red-team review, 30 findings (this directory): redteam-codex.md

Not confirmed on 2026-08-30: chopped carbon fiber price, 16 in and 120 mm fan guards, inline fuse holder, Southwire 14/2 price (Cerrowire 50 ft equivalent $37.00 in a Home Depot search snippet), PG9 gland price (Essentra CG-PG9-1-BK $1.22 in a DigiKey search snippet), Mean Well UL file number, HiLetgo relay module price, ebm-papst 4114N/2H8P IP rating, SPAL 30100467 amperage and IP rating (catalogue figures, model data sheet not reachable), the Atosa MWF9007 live price, Arctic Ice XX-L prices at Walmart, the Tier D wire baskets, and the Tier D mixing tools. None of these moves a tier total by more than $40; the recharge chest freezer is now priced (Galaxy CF7, UCCF8, Northair) and is the line that moves Tier D.