'Dual Stage Thermoelectric' simulation project by buzzing_info


#1

I created a new simulation project called 'Dual Stage Thermoelectric':

A 2 stage thermoelectric cooler with water jacket.


More of my public projects can be found here.


#2

Overview

Dual stage thermoelectric (peltier) temperature control plate. Objective is -40C plate temp with 12C chiller water. Ends up drawing about a kilowatt. The simulations are testing the effects if different plate designs, materials and thicknesses, temperature sensor placement, as well as estimating dTmax.

Thermoelectric Peltier Device Simulation

The thermoelectric devices are simulated as a negative thermal flux on the top surface (the Qc, or cooling power), and a positive flux on the bottom surface (Qh, the hot side power, sum of Qc and Qe, the electric power in). Peltier devices loose efficiency the greater the temperature differential between the sides, this is simulated by setting the thermal conductivity so Qc is conducted completely through at dTmax (the maximum temperature difference).

We’ll run them at 8V, since that’s more efficient than the maximum rated 12V (thus less waste heat to be transported by the second stage). From the plots in the manufacturer’s datasheet, that draws ~7.5A (and Qe = 8V x 7.5A = 60W), has a maximum Qc of 93W, and dTmax of 58C. The device surface area is 0.00175 m^2, and thickness is 0.003 m.

From Laird PC12,139,F1,3550,TA,W6 Thermoelectric Datasheet:

Thermoelectric parameters in the model:

Top Surface Flux = - Qc / A = -93 / 0.00175 = 53143 W/m2
Bottom Layer Flux = (Qc + Qe) / A = (93 + 60) / 0.00175 = 87429 W/m2
Conductivity = Top Flux * Thickness / dTmax = 53143 * 0.003 / 58 = 2.75 W/(m K)

Water Exchange Cold Plates

First type uses a single Aavid Thermalloy 416101U00000G water chilled cold plate. The second type (R1) uses 2 Lytron CP15G01 cold plates, to get lower thermal resistance from the water to the thermoelectrics. This improves total dTmax by ~5-10C. Most simulations were run by setting a boundary condition holding the copper pipes to the water temperature of 12C (285K), to visualize hot and cold spots.

Since SimScale doesn’t simulate the energy transfer to the flowing water in the pipes, a better dTmax estimation comes from the “BottomT Ctrl” simulations, which set the back of the cold plates to 12C, and set the thermal conductivity using the manufacturer’s data sheets, similar to the calculation for the thermoelectric device conductivity. The CP15G01 datasheet resistance from water to cooling surface is 0.009 C/W, or a conductivity of 111 W/C.

Plate Conductivity = (Thickness / Area) * Conductivity = 0.008 / (.095 * 0.15 ) * 111 = 62 W/(m K)

Simulation Results

Final performance (assuming a normal convective heat load on the cooling plate surface) is dTmax = 71K, or a minimum temperature of -59C. The water bath will get up to ~20C. This is similar to direct calculations. Also important is that reaching -40C is simulated to take just 3 minutes from 25C. Building it from copper instead of aluminum provides relatively little advantage. A critical factor in actual performance will be limiting thermal losses through hardware and surrounding structures.

Cut through middle plate - showing concentration of heat from 3 upper stage devices in the middle


#3

@buzzing_info - this is a great project, thanks for sharing! For what do you use this device?


#4

@dheiny - Thanks :slight_smile:. It’s for a test system that tests device performance across a temperature range from -40C to 70C. The slots in the top plate are for a vacuum to hold the parts under test in good contact with the plate. The main challenges are getting to -40C in a reasonable amount of time, and thermal cycling quickly - most commercial systems are built to maintain a constant temperature.