D1. DIY semi-blast chiller build design for the commissary

Written 2026-08-30. Inputs: the two Tech Ingredients transcripts (transcript2.txt, fan; transcript.txt, eutectic) and research memos r1 to r6 in this folder. Prices are listing prices seen on 2026-08-30 at the URL given; where no page could be loaded the line says "not confirmed" and the number is a placeholder to be checked before ordering. Nothing below assumes a health-code blast chiller; the target is faster cooling of prepped food and a compressor that runs inside its envelope.

Freezer type is unconfirmed, so every section covers both a commercial reach-in upright (forced air, electric defrost, 1/3 to 1 HP) and a commercial chest freezer (static, manual defrost, 1/7 to 1/3 HP). Two facts that change the design more than anything else, and that the owner should confirm first:

1. Cabinet holding temperature, logged for a day. -23 C (True T-23F-HC class) versus -18 C decides how fast the PCM refreezes and whether a NaCl eutectic is even possible (r1 section 2, r3 recharge rule).
2. Refrigerant on the nameplate. Current True and Turbo Air reach-ins and most new chest freezers are R290 (propane) or R600a. True's manual forbids unapproved electrical appliances inside the compartment; a brushed automotive fan inside a hydrocarbon cabinet is an unapproved ignition source and a warranty loss (r5 section 4). On an R290 box use the brushless fan option in section 3.0, an older R404A/R134a unit, or a cheap dedicated chest freezer.

1. Sizing

1.1 Heat to remove per batch

Food cp 3.8 kJ/(kg K) (ASHRAE, r5), plus 5 percent for stainless pans. Target 5 C.

Fan heat while chilling: about 50 W at 9 V for the 16 in fan (cube law, r4), so 270 kJ over a 90 min 10 kg chill and 540 kJ over a 3 h 30 kg chill. Small next to the food but not zero, and 4.3 MJ per day if someone leaves it on.

Optional freeze after chilling, per kg of watery food: 5 C to 0 C 19 kJ, freeze 0.8 x 333.5 = 267 kJ, 0 C to -18 C 36 kJ, total about 320 kJ/kg. Freezing 10 kg adds 3,200 kJ and 30 kg adds 9,700 kJ, more than the chill itself. Do not size the PCM for freezing; let the compressor do that overnight after the pans are at 5 C, or move the pans to a refrigerator.

1.2 Effective latent heat of the PCM options

1.3 PCM mass per batch

Mass = batch kJ x 1.3 margin / latent heat. The 30 percent margin covers pan moisture, fan heat, door openings and the last 10 percent of a bottle that melts slowly.

Rule of thumb: 1 kg of -15 C PCM per 1 kg of hot food at 60 to 70 C, with the 30 percent margin already inside that number (0.99 to 1.17 kg/kg for the corrected NH4Cl brine). For the video's ratio use 1.2 kg/kg. For NaCl use 1.3 to 1.5 kg/kg. Staff version: one hard 1 L bottle per kg of food going in, plus one extra bottle for every 5 kg if the food is above 65 C.

Per container, using a 90 percent fill:

So a 10 kg batch at 60 C with the corrected recipe needs 11 HDPE jugs or 13 PET bottles; a 30 kg batch needs 33 jugs or 37 PET bottles.

1.4 Chilling time, so the SOP has a number to log against

Two full-size 2 in pans holding 10 kg of stock present about 0.8 m2 to the air. Log-mean temperature difference from 60 C to 5 C against -12 C air is 38 K.

For solids (cooked grains, braises) the food's own conductivity limits the rate: a 2 in pan of solid food needs 2 to 3 h to reach 5 C at the center regardless of fan speed, and a 4 in pan needs most of a day. The fan buys the most on liquids and on shallow pans. Video benchmark for calibration: 800 mL water in a 1 kg pot, 50 C to 0 C in 60 min with the fan versus 120 min without, in a static residential freezer (transcript2.txt).

Design targets, from r6: 60 C to 21 C in 60 min or less, 5 C in 3 to 4 h total, frozen inventory never above -10 C at the warmest package.

1.5 Recharge time on the freezer's compressor

Two things limit how often you can run a batch: how fast a bottle refreezes (heat transfer at the bottle surface, 8 K driving difference in a -23 C box, 2.6 K in a -18 C box) and how much spare compressor capacity exists overnight.

Per 0.9 kg jug, 280 kJ to refreeze from -10 C (r5 section 5.4, r1 section 2):

Compressor budget: PCM kJ plus overnight fan heat (15 W x 10 h = 540 kJ) against spare cooling capacity at -23 C evaporating (rated capacity minus the 40 to 50 percent duty the envelope already takes; ratings from r5 section 5.2).

1.6 Maximum batches per day

Take the slower of the two limits above, with one PCM set unless noted.

The pattern that works on a single 1/2 HP reach-in: batch at 8 am, bottles back in the tray by 10 am, fan on low until 4 pm, second batch at 4 pm, fan on low overnight, hard by 6 am.

2. Three build tiers

Prices are for the 10 kg-per-batch case unless a line says otherwise; the 30 kg case is given as a delta at the end of each tier. Items common to all tiers (thermometer, timer, tape, tub) are listed in Tier A and referenced after that.

Tier A: video copy

Radiator fan, bench PSU, home-made NH4Cl goop in PET sports-drink bottles, as in the two videos, with the additions a commercial box needs (door interlock, fuse, timer, secondary tray, labels). BOM uses the video's ratio so it is a faithful copy; section 4 says why to drop the salt to 20 wt%.

30 kg per batch delta: four 10 L batches (about 45 bottles, two bus tubs). Replace the six 1 kg salt bags with one Wintersun 50 lb feed-grade bag at $250.00 (https://www.amazon.com/dp/B00N48Q7XY, r3, 2026-08-30), add a second pound of carbon fiber ($15 placeholder), buy the 5 lb xanthan pail at $60.49 instead of the 1.1 lb ($29.50 more) and a second bus tub ($5.29). Total about $590.

Verdict: cheapest way to prove the concept and the closest to the videos, but the video's salt ratio wastes 15 percent of its latent heat, PET goes brittle at -25 C, the bench supply is an open-vented instrument in a kitchen, and the brushed fan is not for an R290 cabinet.

Tier B: food-safe brew

Same fan hardware. The brine is a food-grade salt in HDPE jugs (or, second choice, lidded 1/6 hotel pans). No carbon fiber, no bentonite; xanthan is food grade and stays.

Which salt. NaCl at 23.3 wt% freezes at -21.1 C. A commercial freezer that holds -18 to -23 C gives it 0 to 2 K of driving difference, and a compressor that cycles between -23 and -20 C spends half its time above the NaCl plateau. It will not refreeze overnight in that box; Arctic Ice's own rule for a -15 C pack is a cabinet at -20.5 C or colder for 24 to 36 h (r3). So NaCl is only an option if a logged thermometer shows the box holding -25 C or colder around the clock, which some chest freezers do (Avantco CF8 is rated -25 to -18 C) and reach-ins do not. KCl at 19.5 wt% freezes at -10.7 C and recharges in any freezer, but caps the box air at about -8 to -5 C during a chill instead of -12 to -10 C. The chill runs about 10 percent slower at the start and 25 percent slower in the last stretch below 15 C, and the frozen inventory sees a warmer excursion than the Codex -12 C guideline (r5 section 3). Pick KCl for a reach-in or any box that does not verifiably hold -25 C; pick NaCl only for a verified -25 C chest freezer. Neither is as good a match to a -23 C box as NH4Cl at -15.4 C; that is the price of a brine you can spill on a brisket.

Purity: Morton water-softener KCl pellets are 99 percent plus KCl but not sold as food grade; Morton Purex is a food-grade all-purpose NaCl. Food-grade KCl (salt substitute) costs 10 times more per kg and is not worth it for a sealed jug; treat a KCl leak like a salt leak (rinse, discard the food it touched because of the unknown quantity, not because of toxicity).

Container. HDPE stays impact resistant at freezer temperatures and is a listed container material for chloride brines; PET and PP caps go brittle (r6 section 3). Lidded stainless 1/6 pans work as cold plates (1.5 kg each, big flat faces for the fan) but the lid is not sealed, so a thawed pan sloshes brine, the pan looks exactly like food, and 304 stainless is only "fair" in chlorides over many cycles. If pans are used: dedicate them, use the slotted lid position for a riveted tag, never stack them over food.

30 kg per batch delta: buy the jug case ($114.99 for 162, use 40) instead of 12 samples, 40 caps ($59.60), second bus tub; the 40 lb KCl bag covers four batches. Total about $500. With NaCl in a verified cold chest freezer, subtract $25 and add 3 jugs per 10 kg.

Verdict: the only tier whose bottle contents are harmless on food, at the cost of a warmer plateau (KCl) or a recharge that most freezers cannot do (NaCl). Right choice if an inspector's reaction to unlabeled chemistry in a food freezer worries you more than 20 percent slower chills.

Tier C: buy the PCM, build the airflow

Commercial -15 C packs plus a sealed fan. Highest reliability, nothing to mix, packs are drop-rated, spin-welded HDPE and labeled non-toxic (still H302 if swallowed; a leak still condemns the food, r3).

Fan choice depends on the refrigerant. On an R404A or R134a cabinet, the SPAL 30100467 12 in brushed puller is the only 12 V fan with real airflow and an IP68 motor. On an R290 or R600a cabinet, use four brushless Noctua NF-A14 industrialPPC IP67 fans on a 12 x 12 in polycarbonate board: 430 CFM total, 6 W of heat, no brushes, no spark.

30 kg per batch: 13 Engel XL packs ($389.87), two bus tubs; total about $640. Packs occupy about 30 L of shelf.

Verdict: costs about the same as Tier A because retail ammonium chloride is not cheap in 1 kg bags, and it removes the mixing, the carbon fiber, the PET fatigue and the labeling argument. This is the tier to run in a warranted reach-in.

Used and new blast chillers for comparison (r2, all seen 2026-08-30)

A dedicated unit is the only option that takes the load off the storage freezer instead of shifting it to overnight. The retrofit is a $400 to $650 way to chill 10 to 30 kg a day faster and inside the compressor's envelope; it is not a substitute if volume is going to grow past what section 1.6 allows.

3. Fan install plan

3.0 Gate: refrigerant and fan type

Read the nameplate. R404A, R134a, R448A: any tier's fan is fine. R290 or R600a: no brushed motor inside the cabinet. Use the Noctua array (Tier C) or put the retrofit on a dedicated cheap chest freezer whose warranty you do not care about. Either way the PSU and the 120 V side stay outside the box.

3.1 Reach-in upright (True T-23F class, 22 7/8 x 23 1/4 in shelves)

Sketch in words: looking into the open door, the bottom shelf holds a black bus tub of upright bottles on the left two thirds, pushed to the front so the rear 4 in stays clear of the evaporator housing. The next shelf up carries the 16 in fan lying flat, flat face up, zip-tied at its four shroud corners to the wire shelf, with a strip of closed-cell foam tape under the shroud. The shelf above the fan carries the hot pans 4 to 8 in over the blade; that shelf is the finger guard. The fan cable runs down the hinge-side wall, under the gasket at the hinge-side bottom corner, to the PSU on a wall bracket beside the cabinet. A reed switch sits on the door frame at the top hinge side with its magnet on the door.

• Mount: fan on the shelf second from the bottom, flat face up (blade reversed, leads swapped as the listing describes, blue threadlocker on the nut). Air goes up through the pan shelf, spreads at the top, returns down the front and sides. The 16 in fan is about 35 percent of the cabinet cross-section, close to the video's 40 percent rule (r4 section 6b). Shallow 21 in boxes: use the 12 in fan.
• Do not block the evaporator: nothing within 4 in of the evaporator housing, nothing against the back wall, fan not blowing across the coil face (True and Turbo Air clearance rule, r5).
• Voltage: 8 to 9 V for the chill (about 75 percent airflow, 36 to 50 W), 5 to 6 V for overnight recharge (about 15 W). On the bench PSU set the voltage knob and leave the current limit at 8 A. With the brick and PWM, mark the knob at the 75 percent and 40 percent positions after measuring with the PSU once.
• Cable: flat two-conductor cable under the magnetic gasket at the hinge-side bottom corner, a loop of slack so the door does not tension it, silicone over the gasket gap. Check for frost at that spot weekly for a month; frost there means the gasket is no longer sealing (r4 section 4.1). Do not drill a foamed liner.
• PSU: outside, on a wall bracket above mop height, plugged into a GFCI receptacle (NEC 210.8(B); NYC 2025 Electrical Code) through the countdown timer. Inline 7.5 A blade fuse on the positive lead at the PSU. Crimped, heat-shrunk spade terminals at the fan tabs; no splices inside the box.
• Interlock: reed switch in series with the fan positive lead so the booster stops on door-open, the same as the stock evaporator fan (Turbo Air manual). Cook opens the door, fan stops, steam stays out of the coil.
• Guard: the pan shelf covers the blade face. If the fan is ever run on the top shelf with the blade exposed, fit the 16 in wire guard.
• Bottles: bottom shelf in the bus tub, never above or beside the evaporator, count written on the tub.

3.2 Chest freezer (Atosa MWF9016GR class, interior about 54 x 21 x 27 in)

Sketch in words: looking down into the open lid, the compressor step is at the right end. The 16 in fan stands vertically against the right end wall on the step, flat face toward the left, zip-tied to a piece of epoxy-coated wire shelf that leans on the wall, blowing along the length of the box. A wire rack or upturned basket sits 12 to 24 in downstream of the fan at mid-height; hot pans go on it. Bottles stand in a bus tub on the floor at the far left end. The cable runs down the end wall and out through the defrost drain in the floor, sealed with a grommet and silicone, to the PSU on the wall.

• Mount: end wall above the compressor step, blowing along the box. The end-wall cross-section is about 3.9 sq ft and the fan is 1.4 sq ft, 36 percent, so the return path is the rest of the end wall and under the rack (r4 section 6a). Floor-mounted blowing up only works with two fans because the floor is 7.9 sq ft.
• Voltage as 3.1. In a static chest freezer the fan is worth 4 to 6x on the surface coefficient, so 8 V is enough.
• Cable through the defrost drain plug, grommet plus silicone. You lose the drain until you unseal it at the next manual defrost.
• PSU on GFCI, fuse, timer, terminals as 3.1.
• Interlock: reed switch on the lid rim; or a plain switch mounted where whoever opens the lid can see it. The lid interlock matters more here because a chest freezer dumps cold air slowly and the fan running with the lid open pumps it out.
• Guard: the fan's open blade face points into the box where hands go. Fit the wire guard, or mount a wire basket in front of it.
• Frost: a chest freezer has no defrost heater, so the fan moves pan steam onto the walls and the lid gasket faster. Cover the pans loosely, always, and add the wall frost to the weekly check.

3.3 Cold start

The GESEXI fan carries no temperature rating. If it hums and does not spin at -25 C, give it 12 V for two seconds then drop to 9 V; if it does not start within five seconds cut power (a stalled brushed motor at 10 A heats fast). The simplest avoidance is the 5 to 6 V overnight setting, which keeps the bearing warm and the blade frost-free for under 10 W (r4 section 4.5).

4. Goop recipe at 10 L

4.1 The two ratios

Video description, scaled x10 (about 10.4 kg, 9.6 L): water 7,050 g, ammonium chloride 2,950 g, sodium bentonite 40 g, chopped carbon fiber 200 to 250 g, xanthan gum 120 g. This is 29.5 wt% NH4Cl on the salt-plus-water basis.

Corrected eutectic, 10 L (about 10.4 kg): water 8,000 g, ammonium chloride 2,000 g (20.0 wt%, a hair above the 19.7 wt% eutectic on purpose), xanthan gum 60 to 120 g, and optionally bentonite 40 g and carbon fiber 200 g if you want the full video copy.

Use the corrected ratio. Reasons, from r1 section 1:

• The NH4Cl-water eutectic is 19.7 wt% at -15.4 C. At 29.5 wt% about 122 g of every 295 g of salt precipitates as plain NH4Cl between 31 C and -15.4 C and stores nothing at the plateau; only 85 percent of the video batch freezes eutectically. Effective latent heat 250 versus about 290 kJ/kg, so 15 percent more bottles for the same job and a third more salt per bottle.
• The video ratio is above solubility below 31 C (37 g per 100 g water at 20 C against 41.8 g needed), so 33 g of salt per bottle never dissolves at room temperature and the mix has to be held warm to get it into the bottle. The 20 wt% ratio (25 g per 100 g water) dissolves fully at any temperature above about -10 C, so the mix can be made at 30 C and pours.
• Plateau temperature is identical either way; the video's -15 C plateau is the eutectic showing up in both cases.
• Keeping 20.0 rather than 19.7 wt% leaves a few grams of salt to crystallise out below about -10 C in each bottle, which seeds the eutectic freeze and limits subcooling, the job the video assigned to bentonite. Bentonite (montmorillonite) only nucleates ice at far larger subcooling and is a thickener in PCM practice, not a nucleator (r1 section 2).

What to do about the additives in a kitchen:

• Xanthan: keep it, at 60 g (0.6 percent) for a pourable gel or 120 g (1.2 percent) for the video's yield-stress paste. Its real value here is that a cracked bottle yields a blob you can scoop, not a puddle that runs into the evaporator. It is food grade and salt tolerant. The video needed it to hold the fiber and (unknowingly) the precipitated salt in suspension; a true eutectic has nothing to settle.
• Carbon fiber: omit for the commissary build. The gain is roughly 2x on the brine's 0.55 W/mK, mostly during melting, and only if the Amazon fiber really is high-modulus grade (r1 section 3); against that it is an electrically conductive dust near the freezer's control board, a black foreign-material contaminant if a bottle fails, and the reason the mixer bowl has to be quarantined (r6 section 2.4). The frozen eutectic is an ice-plus-salt composite at about 2 W/mK on its own.
• Bentonite: omit, for the reason above and the quartz dust note in r6.

Ingredient cost per 10 L corrected batch: 2 kg NH4Cl at $19.19/kg retail = $38.38, or $22.04 from the 50 lb bag; xanthan 120 g from the 1.1 lb Sosa pack = $7.44. About $30 to $46 per 11 jugs, or $3 to $4 per jug, against $29.99 for a 2.27 kg Engel pack (about $12 per equivalent kg of latent heat).

4.2 Mixing in a 20 qt mixer with the paddle

A 20 qt bowl takes 10.4 kg of brine with a hand's width of freeboard. Schedule it when the mixer is free for 40 minutes, in a corner away from open food, gloves and safety glasses on (NH4Cl is an eye irritant; nothing here fumes or heats, r6).

1. Weigh 8,000 g of water (tap water is fine; the video used distilled to be tidy) and heat it to 45 to 50 C in a pot. Dissolving 2 kg of NH4Cl absorbs about 550 kJ and will drop 10 kg of solution by 14 K, so 45 C water finishes near 31 C. The video needed a heated bowl only because its ratio does not dissolve below 31 C.
2. Pour the warm water into the 20 qt bowl, fit the flat beater (paddle, not the whip; the whip aerates and sprays brine), lock the bowl at working height.
3. Speed 1. Add the 2,000 g of ammonium chloride in four scoops over two minutes. Run 5 minutes. The solution goes clear.
4. Probe the solution. Anything between 25 and 40 C is fine. (If you are copying the video's fiber step, this is where the under-40 C rule bites; sizing on the fiber gums up above that.)
5. Still at speed 1, sift 60 to 120 g of xanthan over the surface in a slow stream over 60 seconds so it does not clump. Speed 2 for 5 minutes. Stop, scrape the bowl wall and paddle, speed 2 for 3 more minutes. Result: a translucent, pourable gel at 0.6 percent, a thick paste at 1.2 percent.
6. Fill (section 4.3) while it is still warm; it thickens as it cools.
7. Wash bowl, paddle and probe in the three-compartment sink with plain detergent, rinse well, then sanitise with the usual quat step. Do not use the chlorine bucket on anything that has touched the brine (chloramine, r6 section 2.2). If carbon fiber was used, the bowl and paddle should be a dedicated set; do not return them to food service.

4.3 Fill, containers, count

• Fill volume: 90 percent (r1: the eutectic is 80 percent ice by mass and expands 6 to 8 percent on freezing; 10 percent headspace covers it). 850 mL in a 32 oz HDPE jug, 745 mL in a 28 oz Powerade bottle.
• Method: ladle the gel into an 18 in pastry bag, snip a 20 mm corner, pipe into the jug through the wide-mouth funnel. About 30 seconds per jug at 0.6 percent xanthan; the 1.2 percent paste needs the dowel or the bag squeezed hard.
• Count per 10 L batch: 11 HDPE jugs at 0.9 kg (recommended, Tier B and the corrected Tier A), or 12 PET bottles at 0.8 kg with about a litre left over. Batches needed: 10 kg per batch at 60 C, one batch (11 jugs, one spare); 30 kg per batch, three batches (33 jugs).
• Cap: TableCraft 38 mm solid cap on the milk jug, hand tight, wipe the thread dry first. A ring of freezer tape over the cap seam shows tampering and seepage.
• Label every container in freezer tape and marker, front and back: "NOT FOOD. DO NOT OPEN. Ammonium chloride coolant. Leak: rinse with water, NO BLEACH. Discard any food it touched." Number the jugs 1 to N and write the set count on the bus tub. Keep the recipe sheet and the BASF SDS in the HACCP binder (r6 section 3 on FDA 7-102.11 and NYC 81.17(g)).
• First freeze: 24 h in the box with the fan at 6 V before the first use; check every jug is rock hard and none has bulged or wept at the cap.

5. 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.

1. Before loading, squeeze three jugs from different parts of the tub. All must be rock hard. An infrared reading on the jug surface colder than -18 C after the compressor's off cycle also means charged; a reading of -15 C means the jug is on its plateau and only partly frozen (r5 section 7). If any jug is soft, cut the batch by 1 kg per soft jug and write it on the log; if more than a third are soft, do not chill in this box today.
2. Count the jugs against the number on the tub. Any cracked, bulging, weeping or crazed jug comes out and is retired (step 12).
3. Portion the food into hotel pans 2 in deep or less, filled to 1.5 in. Never a 4 in pan and never a stockpot. Cooked grains and braises at 1 in if you need them at 5 C in under 3 h.
4. Weigh or estimate the batch. Rule: one hard jug per kg of food, plus one extra per 5 kg if the food is above 65 C. Do not exceed the hard-jug count.
5. Cover each pan loosely: film not sealed at one corner, or a lid set ajar. Uncovered is faster (20 to 30 percent) but frosts the coil; if you uncover, do it only for the first 10 to 15 minutes while it steams, then cover.
6. Load the pans on the shelf above the fan (reach-in) or the rack downstream of the fan (chest), 2 in gaps between and above pans, nothing against the back wall, nothing within 4 in of the evaporator housing, nothing on top of the jug tub. No stacking or nesting.
7. Put the probe thermometer in the center of the thickest pan. Close the door. Set the PSU to 9 V (or the PWM mark), press the countdown timer for 2 h. The fan only runs with the door shut (interlock).
8. Log product temperature and freezer air at 0, 1, 2, 4 and 6 h. Checkpoints: 70 F (21 C) by 2 h, 41 F (5 C) by 6 h total; the design target is 21 C at 1 h and 5 C at 3 to 4 h. If 2 h passes without 21 C, reheat to 165 F and restart, or discard, and log which (NYC 81.09).
9. Remove pans when the center reads 5 C, not when the timer ends. Move to the refrigerator, or leave in the freezer if the item is to be frozen. Turn the fan down to 6 V (do not turn it off: this is the recharge setting) and note the time.
10. Recharge before the next batch: 6 to 8 h at 6 V in a -23 C reach-in, 14 h or more in an -18 C box, a day or more in a static chest without the fan. Next batch only when step 1 passes again. Two batches a day on one set of jugs is possible on a 1/2 HP reach-in at -23 C (8 am and 4 pm); otherwise use a second set or wait until the next day.
11. End of day: fan at 6 V overnight if jugs were used, off if not. Never leave it at 9 V.
12. Weekly, whoever does the temperature log: pull every jug, wipe it, look for cracks at the base and the cap thread, squeeze, check the cap tape. Check the fan blade for frost and wipe it, check the gasket at the cable entry for ice, check the coil face (reach-in) for more than a light fur of frost, check the drain pan after defrost, and check the chest-freezer walls for ice. Log jug count and any retirement.
13. If a jug leaks: gloves on. Take the jug out in the bus tub. Discard any food the brine touched. Thaw and scoop the gel, rinse the area three times with warm water, dry, then quat sanitiser. Never the chlorine bucket (chloramine gas). If brine reached the reach-in evaporator: unplug, thaw, rinse the coil with a spray bottle of warm water several times, dry with the door open, restart; no coil cleaner. Retire the jug; make a replacement from the next batch. Write it in the log.
14. If the compressor clicks on and off within minutes during a chill, or the box air is above -8 C an hour into a chill, stop loading: the jugs are spent or undersized, or the overload protector is tripping (r5 section 1.4).

6. One-afternoon validation test

Same test as video 1: 800 mL of water at 50 C in a 1 kg stainless pot, probe in the water, second probe or the YoLink sensor in the freezer air, third probe in a sentinel bottle of water standing among the frozen inventory. Log every 5 minutes. Do this before the first real batch and again after any change (fan voltage, pan type, new jug set).

Preparation the day before: jug set in the box, fan at 6 V, so the jugs are hard on the day. A plug-in watt meter on the freezer plug for the whole afternoon (not priced; any Kill A Watt class meter).

Runs, back to back, freezer allowed to return to setpoint between runs (30 to 60 minutes):

Also record for each run: freezer air maximum and the time it returns below -15.4 C after the pot hits 0 C; sentinel bottle minimum and maximum; compressor watts and whether it ran continuously; fan amps at 9 V from the PSU display; a cold-start check at the start of run B (fan starts at 9 V from a cold soak of at least an hour).

Go/no-go:

• Go on airflow: run B reaches 0 C in 65 percent or less of run A's time (video: 50 percent in a static box; a reach-in already has some forced air, so 65 percent is the floor). No-go: better than 80 percent means the fan is not reaching the pot; reposition before the next test.
• Go on PCM: in run C the air peak is below -10 C and the air is back below -15.4 C within 30 minutes of the pot reaching 0 C. No-go: air above -8 C means the jugs were not hard or the set is too small.
• Go on inventory: sentinel bottle never above -12 C in any run. No-go on a reach-in used for storage: pick a less sensitive box or reduce the batch.
• Go on the compressor: no on-off cycling shorter than 10 minutes during any run, and the box returns to setpoint within 60 minutes after run C. No-go: overload protector is tripping; this freezer cannot take the load.
• Go on the fan: starts at 9 V from cold; current at 9 V is 6 A or less (16 in) so a 7.5 A fuse and the 10 A supply have margin. No-go: does not start, or draws over 8 A; swap to the 12 in fan or the SPAL.
• Go on recharge: next morning every jug is rock hard after a night at 6 V. No-go: this box cannot turn the set around daily; halve the daily load or add a second set and alternate days.
• Scale-up sanity: multiply run C's time by about 1.5 for a real 10 kg liquid batch in two 2 in pans (section 1.4) and by 3 for solids, then confirm on the first real batch with the SOP log.

7. Risks and mitigations

Sources not already cited inline

• Transcripts: research/transcript2.txt (video 1, fan, 2026-07-09), research/transcript.txt (video 2, eutectic, 2026-08-29); recipe from video2.description.
• r1 to r6 in this folder, all dated 2026-08-30, for every physical constant, product spec and code citation used above.
• WebstaurantStore search pages fetched 2026-08-30 for pans, lids, jugs, caps, xanthan, salt, bus tubs, cable ties, freezer tape, pastry bags, funnel.
• MoreWine bentonite: https://morewine.com/search?q=bentonite (2026-08-30).
• Adafruit reed switch: https://www.adafruit.com/product/375 (2026-08-30).
• BN-LINK timer: https://bn-link.com/search?q=countdown+timer (2026-08-30).
• Not confirmed today (Amazon returned HTTP 500/503 on every product and search fetch; DigiKey, Mouser, Ace, Tractor Supply, Fibre Glast blocked or empty): chopped carbon fiber price, 16 in fan guard, inline fuse holder, Mean Well GST60A12 price, Volclay bentonite pack size, Morton KCl current price. Check these in a browser before ordering; none of them moves a tier total by more than $40.