OpenNEXEN · CRYO3eng

Portable cooling,
built for extreme heat.

CRYO3eng is OpenNEXEN's first hardware prototype. This project materialized from a recent incident involving one of my nieces; she was hospitalized due to a heatstroke, where traditional evaporative cooling (fans) no longer cut.

Prototype · 4 revisions so far
Build log ↓ Read the documentation ↓
Power input 12V · ~23.8A 285W measured, 30A supply
TEC array 4× TEC1-12706 parallel, shared cold-side heatsink
Hot-side rejection 240mm radiator closed loop + reservoir
Cold-side airflow 120mm axial · 0.7A buck-converter throttled, fin-parallel flow
Measured outlet temp 24-25°C ambient-to-outlet ΔT ≈ 5-8°C
Modeled COP ≈ 0.9 see documentation below
Why

Cooling without the grid

Rural provinces like Kampong Thom see routine power outages exactly when heat waves make cooling matter most. CRYO3eng started as a question: could a self-contained, low-power unit take the edge off extreme heat for a household without depending on grid-scale AC infrastructure?

The current prototype is solid-state: four TEC1-12706 thermoelectric modules sandwiched between a shared cold-side heatsink and individual hot-side water blocks. A 120mm fan (throttled by a buck converter) pushes air into one end of the cold heatsink and along the length of its fins — a push, axial-flow configuration — and it exits the far end as the cooled airstream. The hot side is liquid-cooled: thermal paste couples each TEC to a water block, and the closed loop runs from the water blocks to a 240mm radiator, into a small reservoir, through a water pump, and back to the water blocks.

The whole assembly runs off a 12V, 30A switching supply and measures 285W at the wall — a lot of power for a household device whose entire premise is staying usable when the grid isn't reliable. That tension is the real subject of the documentation below.

Build log

Revision history

Rev Status Change
A Done
Concept + TEC array Sourced 4× TEC1-12706 modules, mounted to a shared cold-side heatsink, bench-tested each module individually against a small heatsink and fan.
B Done
Liquid hot-side loop Replaced individual hot-side heatsinks with thermally-pasted water blocks feeding a closed loop through a 240mm radiator, reservoir, and water pump. Added a buck converter to bring the 120mm cold-side push fan down to a usable speed.
C Current
Efficiency audit Measured system draw (285W @ 12V, ~23.8A) against outlet temperature (24-25°C) and found the numbers don't support the project's low-power goal. Built the thermodynamic model below to quantify exactly how inefficient the TEC approach is, and whether a small vapor-compression loop closes that gap without giving up too much portability.
D Next
Instrumentation Add thermocouples at the cold plate, hot plate, and airstream outlet to replace the modeled assumptions below with calibrated measurements.
Documentation

The thermodynamics

A thermoelectric module doesn't have a fixed "efficiency rating" the way a compressor does — its performance depends on the current driven through it and the temperature difference it's fighting against. Below is the model used to understand why this prototype draws 285W for a modest temperature drop, and what that implies for the next revision.

Contents
  1. Governing equations
  2. Extracting the module parameters
  3. Steady-state performance
  4. Transient response of the cold plate
  5. What the numbers say

1. Governing equations

A Peltier module is characterized by three lumped parameters: its Seebeck coefficient \(S\) (volts per kelvin), internal electrical resistance \(R_m\), and thermal conductance \(K_m\) between its two faces. Given a drive current \(I\), hot-side temperature \(T_h\), and cold-side temperature \(T_c\) (all in kelvin), the heat pumped from the cold side and rejected at the hot side are:

$$ Q_c = S I T_c - \tfrac{1}{2} I^2 R_m - K_m (T_h - T_c) $$ $$ Q_h = S I T_h + \tfrac{1}{2} I^2 R_m - K_m (T_h - T_c) $$
Cold-side heat absorbed and hot-side heat rejected. The middle term is resistive (Joule) heating, split evenly between both faces; the last term is passive back-conduction through the module.

The electrical power drawn by the module is the difference between these two, which must equal \(V \cdot I\):

$$ P_{in} = Q_h - Q_c = S I (T_h - T_c) + I^2 R_m $$
Energy conservation check: heat rejected always equals heat absorbed plus electrical power in.

Coefficient of performance for cooling is simply the useful heat moved per watt spent moving it:

$$ \mathrm{COP}_{cooling} = \frac{Q_c}{P_{in}} $$

2. Extracting the module parameters

TEC1-12706 datasheets don't publish \(S\), \(R_m\), or \(K_m\) directly — they publish four rated numbers instead: maximum voltage, maximum current, maximum heat pumped, and maximum achievable \(\Delta T\), all measured at a reference hot-side temperature \(T_{h,ref}\). Typical published figures for this part:

Rated value Symbol Typical figure
Max voltage Vmax 15.4 V
Max current Imax 6.4 A
Max heat pumped (ΔT = 0) Qmax 92 W
Max ΔT (Qc = 0) ΔTmax 66 K
Reference hot-side temp Th,ref 298 K (25°C)

These are commonly published figures for the TEC1-12706 — actual values vary by manufacturer and batch, so this should be treated as a representative model, not a spec for this exact unit.

These four ratings correspond to two known operating points: \(Q_c=0\) at \(\Delta T = \Delta T_{max}\), and \(\Delta T = 0\) at \(Q_c = Q_{max}\), both driven at \(I_{max}\). Solving the governing equations at both points simultaneously gives closed-form expressions for the three module parameters:

$$ S = \frac{Q_{max} + \tfrac{1}{2} I_{max} V_{max}}{I_{max}\left(T_{h,ref} + \tfrac{1}{2}\Delta T_{max}\right)} $$ $$ R_m = \frac{V_{max} - S\,\Delta T_{max}}{I_{max}} $$ $$ K_m = \frac{S I_{max} T_{c,ref} - \tfrac{1}{2} I_{max}^2 R_m}{\Delta T_{max}} $$
Tc,ref = Th,ref − ΔTmax = 232 K.

Plugging in the datasheet numbers:

Parameter Value
Seebeck coefficient, S 0.0667 V/K
Internal resistance, Rm 1.72 Ω
Thermal conductance, Km 0.967 W/K
Sanity check Feeding these three values back into the governing equations at both original reference points reproduces Qmax = 92 W and ΔTmax = 66 K to within rounding — confirming the three parameters are self-consistent with the datasheet ratings before using them anywhere else.

3. Steady-state performance

The prototype doesn't yet have calibrated thermocouples on the hot and cold faces (that's Rev D), so the numbers here use a representative operating point rather than measured data: \(I = 5.85\,\mathrm{A}\) per module (consistent with ~23.4A total across 4 parallel modules, close to the measured 23.8A), cold face at \(T_c = 283\,\mathrm{K}\) (10°C) and hot face at \(T_h = 303\,\mathrm{K}\) (30°C) — a 20 K spread across the module, a reasonable middle-of-the-road assumption for a loaded TEC with an actively cooled hot side.

$$ Q_c = (0.0667)(5.85)(283) - \tfrac{1}{2}(5.85)^2(1.72) - (0.967)(20) \approx 61.7\ \mathrm{W} $$ $$ P_{in} = (0.0667)(5.85)(20) + (5.85)^2(1.72) \approx 66.7\ \mathrm{W} $$ $$ \mathrm{COP}_{TEC} = \frac{61.7}{66.7} \approx 0.93 $$
Quantity Per module × 4 modules
Heat absorbed, Qc 61.7 W ≈ 247 W
Electrical power, Pin 66.7 W ≈ 267 W
COPTEC ≈ 0.93

267W modeled at the module level leaves about 18W unaccounted for against the measured 285W at the wall — a plausible split between the 0.7A fan (≈8.4W), the small water pump (≈4–5W typical for this class), and buck converter / wiring losses. Dividing the cooling power by the full wall-plug power instead of just the TEC power gives a more honest system-level number: \(\mathrm{COP}_{system} = 247/285 \approx 0.87\). That's the number that actually matters for the "low-power" goal.

4. Transient response of the cold plate

The electrical time constant of a TEC module is on the order of milliseconds — negligible next to the thermal mass of the aluminum heatsink it's bolted to. So the cold plate's temperature over time can be modeled as a lumped thermal capacitance receiving constant cooling power \(Q_c\) while losing heat back to the incoming airstream by forced convection:

$$ m c_p \frac{dT}{dt} = -Q_c + hA\,(T_\infty - T) $$
m = heatsink mass, cp = specific heat of aluminum (897 J/kg·K), hA = convective conductance to the airstream, T∞ = ambient intake temperature.

This is a first-order linear ODE with a standard exponential solution:

$$ T(t) = T_{ss} + \left(T_0 - T_{ss}\right) e^{-t/\tau}, \qquad T_{ss} = T_\infty - \frac{Q_c}{hA}, \qquad \tau = \frac{m c_p}{hA} $$

Backing \(hA\) out of the measured 5–8°C ambient-to-outlet drop (using the low end, 5.5 K, against the modeled \(Q_{c,total} \approx 247\,\mathrm{W}\)) gives \(hA \approx 44.9\,\mathrm{W/K}\). Assuming the shared cold heatsink is roughly 0.6 kg of aluminum:

$$ \tau = \frac{(0.6)(897)}{44.9} \approx 12.0\ \mathrm{s} $$
The cold plate should reach ~63% of its final temperature drop in about 12 seconds, and ~95% within roughly three time constants (~36 seconds) of power-on.
Testable prediction This is a falsifiable number — clocking time-to-steady-state with a thermocouple and stopwatch on the actual unit is a direct check on whether the assumed heatsink mass and convective coefficient are close to reality. Flagged for Rev D.

5. What the numbers say

A modeled COP of roughly 0.9 means the array is moving somewhat less heat than it consumes in electrical power — not unreasonable for a thermoelectric device fighting a 20 K spread, but it's roughly 3–4x below what a small vapor-compression loop typically achieves (COP of roughly 2–4, even with a compressor in the 30–60W class). At 285W continuous draw, this also directly undercuts the project's founding premise: a device meant to help during grid outages shouldn't need a small generator to run.

There's also a design detail worth flagging independent of the TEC-vs- compressor question: the cold-side fan pushes air along the heatsink fins — an axial-flow path — rather than through them, the more common cross-flow layout. Letting the boundary layer grow uninterrupted along the full fin length instead of being repeatedly disrupted likely lowers the effective convective coefficient compared to a conventional through-flow heatsink — a plausible reason the measured 5–8°C outlet drop is smaller than the ~20 K the model predicts at the cold face itself.

Whether a small vapor-compression system closes the efficiency gap enough to justify its added complexity — refrigerant handling, compressor noise and vibration, reduced portability — against a COP roughly 3–4x higher and a fraction of the power draw, is the open question for Rev D onward.