How to drop Voltage without reducing current for Peltier Cooler Pad

[eyeow]

After understanding the behavior of Peltier, now would like to find the best way improve the circuit design to drive the Peltier at desired voltage and current to protect the Peltier from damaged as well as the Peltier could do the cooling to a desired temperature with best cost-effective way.

Loading:
Peltier rating: 100W, Vmax 20V, Imax 8.5A. After some experiments, the desired input voltage is 15-16 V, expected current drawn will be 4-5 A.

Existing circuit Problem:
A microcontroller is used to output pwm signal to drive the mosfet as shown in the diagram. PWM setting duty cycle 40%, Freq 2016 Hz. DMM Voltage reading: 9.5 V, Clamp Meter Current reading: 3.14 A. The readings look ok. We can control the duty cycle to desired voltage level 15-16V. But based on scope measurement, the Vmax and Vin actually exceeds the Peltier operating voltage range as shown in the photo which we may think it will damage Peltier slowly.





Constraint/Condition:
At this moment we cannot lower the power module voltage rated output 24V.
We also don't have option to change to higher voltage rating Peltier due to cost issue.

Question/Solution:
1. we got this proposal 1 as below:





4 diodes + 4 heatsinks: To drop the 24V DC to 17-18 V and disipate heat
LC filter circuit: to filter and smooth spikes at Vmax and Vmin

It solves our problem and we can tune the duty cycle to get back desired voltage, as long as the diode stops the rated voltage at 17-18V, but those components take extra space, cost is higher to use 4 extra heatsinks to dissipate the diodes' heat and required fan as well. Inductor can be optional as Peltier is seems like a resistive loading only.

2. Voltage divider? High power rating resistors + heatsinks may be required, and it also limits the current as well, right?

3. Buck DCDC Converter? Any good recommendation?

4. Other solution/ suggestion to improve existing PWM control/mosfet?

Thanks.

 

[dick_freebird]

At such a low frequency and without an inductor in
the power path your PWM consists of applying an
overvoltage, and no voltage, by turns. It's a pretty
poor control scheme for reliability.

Why can't you lower (or ideally, control) the module
output? Inside there is almost certainly some fairly
traditional PWM IC and using the uC to bend its
feedback or reference would be (a) elegant and
(b) eliminate the MOSFET and the poor PWM action.

 

[BradtheRad]

The buck converter is a good option. It adds an inductor in series with the switching device.

My simulation has essential components only. The waveforms are typical for your specs.



If you look around you might find a commercially-manufactured converter, which has adjustable output voltage. Or else it's up to you whether you wish to build your own.

 

[FvM]

The question title suggests that you did not fully understand the behavior of a Peltier element but at least realized that the unfiltered pwm circuit must not be used.

The simple fact behind the observed problems is that cooling power of the Peltier element is ruled by Iavg while it's internal power disspation is proportional to squared Irms. So current pulsation above 20-30% will considerable reduce it's efficiency.

we got this proposal 1 as below

Don't know if you actually got this suggestion. It's a flawed buck converter, it can achieve smooth output current but dumps the stored inductor energy to the switching transistor due to wrongly connected diodes. Can be turned into a working buck converter by a slight modification. Similar to the buck converter circuit posted by Brad.



Your previous thread suggests a comment about Peltier current rating. It's a theoretical number only achievable with a perfect heat sink at the hot side. As your experiments show, the optimal current with a specific heat sink can be considerably lower.

 

[c_mitra]

1. Peltier cooling effect depends on the current only. The voltage that appear on the module is "similar" to a diode (Seebeck) voltage only. You control the current and the device will select the voltage accordingly.

2. Although the top and bottom surfaces are metals, the sandwich is a rather poor conductor. Nevertheless, heat will be flowing from the hot side to the cold side happily. You can "easily" get a temp difference of 20-30C.

Can you just pass 1A (just put some high wattage resistor in series) through the pad, how much cooling you are getting? Current and cooling should be approx linear for small currents.

 

[andre_luis]

In addition to the smooth control suggested above, there are other factors that appoint the PID control as a suited method for this application, instead of making a simple instantaneous error correction of On/Off driving in loop back control:

• The first aspect is that the mass of the heatsink where the sensor is supposed to be fastened, give an intrinsic capacity of accumulate heat, giving a result quite different of the actual temperature of the cell.
• The second aspect is that considering that you’ll probe the temperature based on a contact thermometer, there will happen some delay between the temperature of the device under test, giving an inaccurate result.

If you want really make something that simple, at least you could manage the PWM duty-cycle in such a way that increase the pulse-width more smoothly, in order to give time to temperature become a bit more steady.