Project: Poor Man's Calorimeter

Thread Starter


Joined Sep 9, 2010
World's Smallest Refrigerator? Temperature-Controlled TEC Cooler
Objective: Build a small, thermostatically controlled chamber, that can be used in experiments to measure heat production at temperatures below ambient. Production of heat is estimated by measuring the amount of cooling called for by the controller. Such a device is often referred to as a PMC, or poor man's calorimeter. It is not meant to be a commercial device.

Strategy overview: Because of size limitations and the need to go to temperatures below ambient, cooling by Peltier thermoelectric cooling (TEC) was selected. These devices are horribly inefficient , but cheap, simple and effective when minimal cooling below ambient is needed. A heater per se was not required, as it might be in a "real" calorimeter, because all the experiments I intend to perform will generate heat. For development work, I have used 2, 10Ω 10W resistors in parallel to simulate my experimental load.

An internal, thermally-conductive chamber is surrounded by insulation except for a "window" containing the TEC and associated heat sinks, fan, etc.

Electronics to control temperature are nearby. Control is by a simple on-off thermostatic control of the TEC. The circuits include a high-temp cut-out of the load current (for instance if heat production exceeds the cooling capacity), detecting the presence of the load, and a delayed shut-off of the fan so that the fan runs a little while after the TEC shuts down. All functions are indicated with LEDs.

Data collection is performed using data acquisition, details of which are beyond the scope of this write-up. If I had it all to do over, I'd put my data acquisition equipment to better use (eg. PWM control of the TEC). The current PMC design operates completely independently of the DAQ.

Design Parameters: I need to hold the chamber in the range of 10-20°C, and measure heat production of less than ~0.5W. So, this design is not immediately useful for significantly lower or higher temperatures. And, the cooling capacity of the current design limits the amount of heat production that can be used.

TEC I started this project by buying a couple of TEC 12709 modules, mostly just to play with. They each demand up to 10A at ~12v to run at full capacity. Supplied 5v, the TEC 12709 draws ~4.5A and consumes ~25 watts. It moves less than 1 watt out of the chamber, so it's not terribly efficient.
Power Supply Because of the high current demand of the TEC, I chose an old computer PSU to be the power supply for this project. The 12v supply was not up to that current demand, and I was worried about the TEC getting too hot at high current, so I decided to use the 5v supply for powering the TEC. With the TEC on the 5v supply, I chose to run everything else off the 12v supply. Thinking that I might someday want this all to operate off a car battery, I chose NOT to use the -12v supply, although this would have solved some issues. Only later did I learn what many here already know - that some loading of the 5v side is needed to bring the 12v supply under regulation. So my supply varies from maybe 9 to 16v when the TEC is off. When on, the 5v and 12v supplies both hold very steady. I'm considering putting a light bulb or something on the 5v supply to steady it out.
MOSFETs I chose the N-channel IRF540N because it could safely switch more than the full 10A demanded by the TEC, and because it was easy to find cheaply. The VN10KM was simply a small N-channel MOSFET I had in a parts drawer. It only switches a 60mA fan.
ICs The LM35 thermometer is cheap, widely available and a good fit for this project. I was hoping to use the LM34, since it produces a higher voltage difference between temperatures (10mv per °F, versus °C) AND it produces a voltage from a single supply all the way down to 0°F (instead of 0°C), but it's much harder to find. The LM339 quad comparator was chosen because I could walk into Radio Shack and buy one.

The LM358 dual op-amp was chosen because it can sense down to the negative rail. An early breadboard circuit with another op-amp was failing at lower temperatures because the LM35 voltage, 0.15v at 15°C, was falling too close to the rail and the op-amp could no longer "see" it. Using a negative supply fixed this, but I wanted to stick to a single supply.

Schematics The first post is the cooler and control. The second picture is the constant current (heater) control.
Picture 1.png
Picture 3.png

This is an early version of the thermostat that controlled to ±0.6°C or so. That's not bad but I decided I wanted a tighter control.
Comp Only.png

Here's another variation without the fan circuit, but with the op-amp added back.
Picture 1 12-05-38.png

Description: All circuitry except the TEC and its LED indicator, which are powered by the 5v supply, are powered off the 12v supply from the PSU. A 7805 regulator is used to provide voltage reference, and is also used to power all ICs. This is an arbitrary choice that limits the upper and lower temperatures that can be used. It would not be difficult to make changes to expand the working temperature range.

Chamber temperature 1-25°C is sensed by IC1 and amplified 18X by the first op-amp. The range is limited on the low end by the LM35 (produces 0v at 0°C) and the use of a single supply for the LM35 and the op-amp (can’t sense below the negative rail). The upper end is limited by the 5v upper rail voltage. 25°C gives 0.25v from the LM35, times 18 for the gain is 4.5v, which may be near or above the upper range of the op-amp.

The amplified temperature is compared to a set point established by R9, a multi-turn Bourns precision pot, allowing very nice fine tuning. The comparator employs minimal hysteresis provided by the 10M resistor. I recommend reducing this to 5M or 3.3M. The impact on temperature swing is minimal but switching chatter is reduced. I tried several values and 10M was working OK on the breadboard. It works less well in the finished prototype. Comparator 1 output directs the MOSFET control of the TEC, and energizes an RC tank on comparator 2, turning on the fan and its indicator LED.

When the TEC turns off, the fan is held on a few seconds while C1 drains through R8. For simplicity, comparator 2 uses the same voltage reference as comp-1. So the fan delay is slightly affected by the temperature setting; higher temp equals less delay. Note that the fan turns ON immediately with the TEC. In fact it always comes on a bit before the TEC.

The HEATING part of the circuit is simply a resistive load, such as two 10Ω, 10W resistors in parallel, with a constant current supply. Op-amp 2 compares the voltage on a shunt resistor to a pre-set established by R15 (another Bourns pot). The darlington arrangement under op-amp control thus controls current through the load and shunt resistor. The op-amp connection to the shunt is overwhelmed, and current thereby cut off, when comparator 3 sees too high a voltage output from op-amp 1 (the amplified temperature). Comp-3 compares the output of op-amp 1 to a reference established by R16, a cheap, single-turn pot. So if temperature rises above a coarsely set maximum, the load (and heating) are shut off.

Presence of continuity through the load is detected by comp-4 and indicated with yet another LED. (Can you tell how much I love the pretty little flashing lights?!)

Thread Starter


Joined Sep 9, 2010
Here are some pictures of this project.
Preparation of the thermometer. After soldering the leads, I added some hot glue as a sort of filler, then slid the heat shrink tubing on over that. The wire I attached was 4-conductor headphone cable from cheap airline headphones, with the two "ground" wires connected together.
Cutting of the chamber insulation from 1" thick pink styrofoam construction insulation. The blocks are about 6" square. This stuff is easy to cut, and a bread knife (straight, serrated edge) is a great way to shave and sculpt it. Not shown is the bottom layer, which is just a slab with no cutouts.
The aluminum chamber was made from 1" angle aluminum, 1/8th inch thick. Each side is 3 and 1/8" long, so the center window is about 1" square and the inner chamber is almost 9 cubic inches. The sides were attached with JB Weld. The sides were assembled on a piece of glass, with the small window down on the glass, so that the 4 pieces would be as aligned as possible on that face plane.

Thread Starter


Joined Sep 9, 2010
Another view of the aluminun box

The box inserted into the styrofoam. Very handy that both are exactly 1" thick.

Plastic tile separators cut used as a frame to hold the TEC

The TEC in its frame

...and covered by the top layer of insulation

...with the heat sink and fan in place

Thread Starter


Joined Sep 9, 2010
Here are two pictures of the components placed on the perfboard. These pictures were taken BEFORE all the interconnecting wires were added and obscured the view.


Here are the component and interconnect maps.
Picture 2.png

Picture 4.jpg

Parts List
Can someone tell me how to get the tab stops to look better on this table?

R Description Ω
1 Stabilize LM35 output 220
2 TEC hysteresis 5.6K
3 TEC hysteresis 10M
4 TEC gate pull-up 3.3K
5 FAN gate pull-up 4.7K
6 TEC LED D2 current limit 330 blue
7 FAN LED D3 current limit 470 white
8 RC time delay 100K ~10s
9 Temp adj. voltage divider 0-10K
10 Temp adj. voltage divider 4.7K
11 Op-amp negative feedback 180K
12 Op-amp negative feedback 10K
13 POWER LED D4 current limit 2.2K green
14 Current adj. voltage divider 39K
15 Current adj. voltage divider 0-10K
16 Charge temp. voltage 470K
17 CHARGE LED current limit 1K red
18 Charge temp. hysteresis 3.3M
19 Charge temp. hysteresis 10K
20 Charge drive 470
21 Charge feedback 100K
22 Charge current shunt 0.47 for <2A
23 Connect voltage divider 100K
24 Connect voltage divider 16K
25 Connect LED D7 current limit 2.2K green
26 Connect input pull-down 100K
27 Darlington current limit 2 X 82 2W

C Description µF
1 RC time delay 47 16v
2 comparator quieting 0.1
3 5v regulator 0.1
4 5v regulator 0.47
5 op-amp bypass 0.1 not shown

D Description
1 RC time delay charge, 1N4148
2 LED, blue, TEC ON
3 LED, white, FAN ON
4 LED, green, PCB Power ON
5 LED, red, CHARGE enable
6 Charge enable signal, 1N4148
7 LED, green, Connect

IC Description
1 LM35 thermometer, 10mv/°C
2 N/A, removed
3 LM339 quad comparator
4 LM358 dual op-amp

Q Description
1 Charge current control, 1st stage, C1815
2 Charge current control, 2nd stage, 2N3055

Not listed: The power supply, fan and heat sink are old computer parts. The TEC is a 12709.

Thread Starter


Joined Sep 9, 2010
A better fan and heat sink

Detail of the heat sink INSIDE the chamber, attached to the TEC by conductive grease. There is no forced air in the chamber.

The completed module. The board is attached to an aluminum box using screws and plastic tubing cut to make standoffs. You can see the large power resistor shunt underneath the platform. The heat sink around the 2N3055 is a couple pieces of leftover aluminum angle.

The underside of the platform, with the power shunt resistor, a connection panel and the two resistors to make R27, the current limiting on the Darlington stage 1 transistor.

The connection to my data acquisition device, a LabJack U3-HV


Joined Apr 2, 2009
Not bad...Never seen this thread before.

By the way, Like to know how this performed compared to u know a fridge.

Thread Starter


Joined Sep 9, 2010
Sorry, didn't notice your post til just now.

In short, it's terrible as a fridge. It can barely keep the insulated chamber maybe 15°C below ambient. It's limited by the fan and heat sink ability to cool the hot side of the TEC, and I think by the fact that the heat sink is also a path back into the chamber when the TEC and fan are off. It warms up faster than it should. Saying that, I just thought of putting a little flap over the fan that closes when the fan goes off. Hmmm...

TECs are terribly inefficient, removing one heat unit from the cold side while consuming 9 and delivering 10 to the hot side. So you buy 9 to move 1, and have to dissipate 10. Your home A/C moves something like 10 for each 1 you buy, and you have to dissipate 11. About 100X better.

BUT, as a calorimeter to measure power consumption to maintain temperature, this rig works nicely.

Thread Starter


Joined Sep 9, 2010
The TEC datasheets address this, but I think I recall that the efficiency goes up if the ∆T across the TEC is smaller, so yes it would be better with better exterior cooling. But I don't think a TEC ever moves even 1 heat unit for each 1 consumed. If you have water cooling available, you might as well just ditch the TEC.