Hi. Newbie here. I'm designing a control circuit for a TEC device. Super fast response is not a criteria, but it needs to be very low noise.
Attached is a 3 stage low pass filter design with an output of 2.5v to a logic level n channel mosfet. The source is 490hz Arduino PWM.
The reason for 3 stages is to eliminate continual switching of the mosfet gate and to prevent the mosfet gate from intermittently going low when operating at low voltage.
Newbie questions
With this design do I need an RC filter? I'm guessing no.
Will the response time be too slow to control a thermal mass with low inertia - that is, it wont change temperature too rapidly.
That design is just three RC filters cascaded.
I am sorry but this is a very poor design. Plus the fact I don't know why you need it. If all you are doing is controlling a Peltier cooler then I don't see why you need to smooth the PWM signal in the first place. The problem you will have is by smoothing it you are going to operate the FET in the linear region so that means it will get very hot and that is assuming that you get a FET with a gate threshold of 2.5V. You will never be able to turn the FET fully on because that takes 5V or 10V depending on the FET type.
That means over half the PWM's output is going to give you no change in current.
Do you know what current your Peltier device takes?
You don't need to smooth the PWM output to drive a Peltier device, in fact it is a bad ideas for the reason that Mike describes. But you should increase the PWM frequency. See item 41 at http://www.tellurex.com/technology/peltier-faq.php.
On reflection I see what you mean. Current is dependent on the gate voltage. In that case an RC snubber is needed. The TEC is 8 amps. The FET is a 30v 120A RDSon 1.4mohm at 4.5v.
No, in this case average current is dependent on PWM width.
You want the MOSFET full on for low Rds or full off for open.
Anything in the middle is higher Rds and potentially high heat.
If you were making an amplifier and wanted output to reflect linearly, then you'd want current to depend on gate voltage.
dc42:
You don't need to smooth the PWM output to drive a Peltier device, in fact it is a bad ideas for the reason that Mike describes. But you should increase the PWM frequency. See item 41 at http://www.tellurex.com/technology/peltier-faq.php.
PWM on a Peltier devices a moot point. There was a thread here a few years ago where someone said they had used them in industry for over 40 years and used PWM on them.
There is no doubt that DC would be better but it is harder to produce at those sorts of currents and for a project I don't see why PWM will not do. The project should concentrate on the control algorithm rather than the rather difficult aspect of generating a variable 8 amps.
Note to geoland a FET is normally limited by thermal issues and not the maximum current given in the data sheet. It is very rare you can get anywhere close to the data sheet maximum current in practice.
The link I posted to saying that PWM is OK for a Peltier device is from a manufacturer of these devices, so they should know. On the other hand, the link that MarkT posted is correct: when you are not running the device close to full power, it is more efficient to feed the device with DC rather than PWM, if you can produce that DC efficiently (not by using a mosfet in linear mode). Near or at full power, it makes little or no difference.
So the question is whether it is important to you for the system to be efficient when it is running well below full power. If no, then use simple PWM. If yes (solar powered fridge?), then use high frequency PWM with a series inductor and Schottky flyback diode.
I need to minimize RF and effectively control set point - derivative control I think it is called. Having controlled one basic TEC set up with PWM, it's not suitable around sensitive electronics and the switching inductance and capacitance issues associated with the mosfet switching was a pain to filter out properly. Something I never achieved.
Tellurex also provide this discussion on TEC control, PWM and Linear. Tellurex also state that frequencies above 60hz is preferable, elsewhere they state 120hz.
Well switching 8A with PWM is asking for RFI so you are back to DC control. However, using a FET in linear mode is not easy due to the thermal considerations mentioned before. This circuit was for DC control of a fan but it will scale up with the appropriate heat sinks and FETs if you can get them.
if you use a series inductor and Schottky flyback diode, then you should be able to confine the high frequency components of the PWM current to the mosfet, diode and decoupling capacitor, where you can contain and if necessary shield them. Use of a high enough inductor and PWM frequency will also smooth out the current, thereby increasing the efficiency of the Peltier device at low loads.
Thanks GM and dc42. Now that PWM and the whole mosfet switching thing is clearer to me now, I see the opportunity for a little R and D. The series inductor works at 1nh smoothing out the bottom of the wave form.
One last question. Other than contolling the mosfet switching inductance and capacitance interference, will modifying the PWM waveform by, "raising the floor", so that voltage is close to, but never zero, improve the noise issue; that is, no on/off, rather, a slightly out of square RC modified waveform?
No, it won't. What matters is how much the current changes (which is fixed by the Peltier device and the decision to use PWM), how fast it changes (which, for the current that actually goes to the Peltier device, you can control with the series inductor), and the area enclosed by the circuit in which the changing current flows.
This is the mosfet switching circuit I came up with. 2 rdc rate-of-rise voltage clamping snubbers - one for the mosfet, the other around the TEC represented by a resistor. The RC values are estimated, based on fully charging the capacitor
Interesting, but overcomplicated IMO. There is no point in having 2 snubbers to slow down the rise and fall time of current in the same circuit. All you really need to slow down the rise/fall time of the Peltier current is the inductor, plus a Schottky diode from the drain of the mosfet to +12V. Also missing from your circuit and absolutely essential is a good decoupling capacitor across the 12V supply.
See attached schematic for my suggestion. Keep the circuitry inside the dashed circle as compact as possible - that is the bit that will generate noise. L2 may not be needed, or a ferrite bead may suffice - its purpose is to persuade the switching current to take the path to ground through C1 instead of via the wires to the power supply. C1 needs to have a low ESR.
An additional capacitor was needed to remove inductance spikes with L1 at 100uH. The power supply is a 0 - 24v 15A 9mv ripple smps - do I still need the decoupling cap. L2 removed...
That added 1uF capacitor is bad for the mosfet - you will get high peak currents which will heat the mosfet (and the capacitor may not last long either). Without the capacitor, you will get a fast rise time at the mosfet drain when the mosfet switches off, however that is not what causes EMI provided that you keep the mosfet, inductor, Schottky diode and main capacitor close together (and preferably shielded). What will cause EMI is fast rise or fall times in the wiring external to that circuit, i.e. in the wiring to the Peltier and the power supply. So you should concentrate on getting those currents to have slow rise/fall times, and not worry about the rise time of the voltage on the mosfet drain. If you really must include a capacitor across the mosfet, place a resistor (e.g. 1 ohm) in series with it, or use a diode feeding a parallel RC network like you had before. But IMO it's not necessary, provided you lay out the circuit sensibly.
The diode in your schematic looks a little odd - I'm not sure whether you think it is a zener diode (which would explain the "10V" annotation) or a Schottky diode. It should be a Schottky diode.
To prevent the power supply leads radiating EMI, I think you will need a much larger main decoupling capacitor. I also I think you will need the second inductor, but it can be much lower than 100uH, maybe 1uH will be enough. Use your simulation to look at the current in the power supply leads, not the power supply voltage.
