Regulator for Peltier Element

[Irina F]

For my thesis, I have to design a regulator which controls a Peltier Element (12V, 5 A) to control the temperature in a mini fridge.
I was told I have to measure the temperature inside the fridge with pn junction, compare it to a given value and use the resulted difference to make a PWM command to the Peltier.
I designed a circuit, but I can't manage to simulate it in Spice, so I can't know if the components are dimensioned correctly.
Could you please tell me if I should modify anything?
Thanks a lot!

https://obrazki.elektroda.pl/2378262300_1428993369.png

 

[betwixt]

You need to ground the source pin of M1.

Note that you are not strictly using PWM in that design. You are generating a rectangular waveform from the NE555 and simply gating it on and off with the output of your comparator. A true PWM system would leave the oscillator running all the time and adjust the ratio of the on to off time to control the average output current.

Your values are OK and it should oscillator at around 280Hz which is fine. You do not need to make R9 so high in value, if you drop it to say 470 Ohms it will slightly reduce the storage time in M1 and improve efficiency.

Brian.

 

[KlausST]

Hi,

i fully agree with Brian.

Additionally:

Q1 is the temperature sensor. Giving about 580mV at 25°C and -2mV/K change.
This is amplified by gain of 9.
Output is about 5.2V at 25°C.
A capacitor from pot wiper to GND should stabilize voltage.
Your pot range is about 2.4V ... 7.2V, this is gives a temperature range of about 250K. You can work with this, but maybe you need to adjust the pot range.

Then there is a comparator. R7 should create some hysteresis - but it does not, because there is no source resistance (I hope you don´t count on the OPAMP output impednce)
Mind to use an OPAMP that is able to work as comparator (common mode input range, differential input voltage range, saturated input stage, phase reversal...)

Klaus

 

[FvM]

For optimal Peltier element efficience, you won't use PWM. You have the problem that the peltier heat transport is roughly proportional to Iav while peltier losses depend on Irms.

 

[Irina F]

Hi,

i fully agree with Brian.

Additionally:

Q1 is the temperature sensor. Giving about 580mV at 25°C and -2mV/K change.
This is amplified by gain of 9.
Output is about 5.2V at 25°C.
A capacitor from pot wiper to GND should stabilize voltage.
Your pot range is about 2.4V ... 7.2V, this is gives a temperature range of about 250K. You can work with this, but maybe you need to adjust the pot range.

Then there is a comparator. R7 should create some hysteresis - but it does not, because there is no source resistance (I hope you don´t count on the OPAMP output impednce)
Mind to use an OPAMP that is able to work as comparator (common mode input range, differential input voltage range, saturated input stage, phase reversal...)

Klaus

Thank you very much to both of you, Brian and Klaus!

Klaus, do you think 10 nF would be a good value for the capacitor from pot wiper to GND?

Do you think the LM 741P opamp would be OK to use as a comparator? Would the problem of R7 not creating some hysteresis be solved only by changing the opamp?

 

[Irina F]

Thank you! I've done some research on PWM and I understood that it's not the best solution for controlling a Peltier element. Unfortunately, this is a project for the university and I was told to use PWM...

 

[Irina F]

Thank you very much!

Do you think 10 nF would be a good value for the capacitor from pot wiper to GND?

Could the problem of R7 not creating any hysteresis be solved only by changing the OPAMP? Would LM741P be appropriate to use as comparator?

Thanks a lot again!

 

[betwixt]

I would choose say 1uF, the value is not critical. A smaller value will make it more prone to noise (bear in mind you are switching high current rapidly so it will tend to create interference), a higher value will make the control less responsive as you adjust it.

The hysteresis via R7 is good but to a large extent it is being cancelled by the output current of U1. To make it more effective you need to limit the input current so the current through R7 has more influence. I would suggest adding a resistor in the horizontal line between R4 and R7 on your schematic. A value between about 1K and 10K should be sufficient. You may have to experiment to find the best value.

The LM741 will work in this application but by todays standards it has very poor performance. A CMOS or JFET amplifier will work better.

If you want to use real PWM, despite it not being the best method of control, the best solution would be to use the amplifed voltage from the sensor (Q1) to directly adjust the pulse width. You can do that by controlling the charge and discharge times on the NE555 in the same way as the pot between D1 and D2 does at the moment or you can remove D1, D2 and the pot, replacing them with a single resistor so the NE555 free runs in astable mode then feeding the voltage across C1 to the comparator. The comparator would then feed M1 instead of the NE555. The idea is to feed a ramp signal from the capacitor and compare it with the sensor voltage, depending on how far along the ramp the sensdor voltage reached before matching the ramp voltage would set the pulse width.

Brian.

 

[KlausST]

Hi,

i agree with Brian.

To PWM the peltier, you could connect an inductor in series with the peltier and a freewheeling diode across both.
I don´t know the HF impedance of a peltier, to be on the safe side you could connect a capacitor in parallel to the peltier.
You then need a much higher pwm frequency.


Klaus

 

[FvM]

The peltier impedance is purely resistive. So if you think about a low-pass filter to smooth PWM ripple, L or LC are both an option. Reducing the ripple current below several 10 % will already serve the purpose of getting almost negligible peltier AC current losses.

If the peltier design has sufficient cooling power overhead, or if the duty cycle is near to 100 % anyway, PWM operation won't be a problem.

 

[dick_freebird]

I wonder just what timescale the "PWM" has to be on.
Because after all, a refrigerator thermostat acts as a
hysteretic PWM controller (just on a very slow time
scale). If you can power the Peltier module at a known
safe voltage (hence a known safe current) then a
bang-bang hysteretic control will produce a PWM
waveform - not fixed frequency, and mighty low
frequency, but still.

I see not much value to a linear PWM loop in this
application, whose sample needs to be held to sub-
degree temperature accuracy really?

 

[betwixt]

That is quite true Dick, in fact it should be possible to completely remove the NE555 in the original design and use the comparator to directly drive the MOSFET. I think this is a design exercise though, to show understanding of how pulse width can be used to provide linear control (post #6).

Brian.