1100D cold finger sensor cooling with TEC and water cooling

[Gina]

I've now removed the LM317T from the circuit board and attached it to the water cooler on the outside and wired it to the board. Hole and connections sealed and insulated with hot melt glue and insulated from the WB with the usual kit.

Three photos :-

1. Front cover showing LM317T attached to WB, with thermal pad and insulator, rubber strip between connections and WB and sealing/insulating hot melt glue.
   
2. Inside showing wiring to circuit board.
   
3. Inside with several packs of silica gell ready for sealing up.
   

[Gina]

Had a thought... May put a window in the front cover so I can see the silkica gel pack to check for colour change.

[Gina]

New bench test successful and I've put the camera on the scope and connected it up. Just waiting for the rain to stop or the easterly wind to drop so that I can drop the flap and view the eastern horizon to carry out a full test.

[Gina]

Managed a short test in the obsy. Camera working fine until I turned up the cooling then - strangely - the camera voltage regulator started overheating in spite of a virtually perfect heat sink. The WB was cold but the chip was hot. The camera stopped working but the cooling works. The only difference from the bench test was having the 5v supply on as well as the 12v. The 5v supply goes to the Peltier TEC only. Time to disconnect everything and bring it back indoors for investigation, I guess.

[Gina]

Did a quick test before disconnecting everything. The LM317T regulator is still working and so is the camera with the PWM switched off - just power on the TECs/MOSFETs but with the MOSFETs held in the off state. Then while watching the live picture in live view mode in EOS Utils, I switched on the manual PWM and noticed interference in the picture so I need to address that. As I turned up the pulse width of the PWM control the interference got worse and the LM317T started heating up. I then switched off the PWM and the LM317T cooled off leaving the camera running happily.

Seems to me that the MOSFETs and their wires need screening as well as decoupling with capacitors. I think the MOSFETs want to go as close to the TECs as possible. A slight redesign of the layout is in order. In fact I'm thinking of placing the MOSFETs right beside the TECs with copper sheet shielding. The MOSFETs could be thermally connected to the screen with thermal pads between drain and screen.

[Auntystatic]

If the regulator is on the same supply as the pwm, you have a known and documented problem. Its the same problem that plagues a lot of people when running a telescope off the same supply as some dew heaters, the telescope starts to act strangely with its guiding etc.

[yesyes]

Gina, could you try running the 5V TEC only and see if you still get interference? Would be interesting to know if the interference feeds back through the 12V power line or whether it's RF.

[Auntystatic]

RF from a 555? its more likely to be power pulses, from the sharp square wave on/off nature of the pwm switching the high current load, a very large reservoir capacitor might help.

I think I'm going to get me one of these 20A supplies or one of these 30A supplies as I'm currently using my amateur radio supply and it is needed elsewhere.

[yesyes]

didn't mean directly from the 555... no idea what frequency Gina is running it at...

if it's not RF, then shielding the wire won't help either...

[Gina]

I think the 555 runs at about 500Hz but being a square wave there are harmonics going up into the RF range. The MOSFETs are square wave switching currents up to 5A and voltages of 5v and 12v. The present wiring is a few inches long and unscreened. I think I was rather remiss in my design - should have known better

[RogerTheDodger]

We are all guilty of that Gina, you wouldn't think we'd been trained in electronics sometimes

[Gina]

If the regulator is on the same supply as the pwm, you have a known and documented problem. Its the same problem that plagues a lot of people when running a telescope off the same supply as some dew heaters, the telescope starts to act strangely with its guiding etc.

The regulator is well decoupled. Anyway all was well with the 12v supply - I turned up the PWM all the way up for a short periuod. The trouble seems to be with the 5v.

[yesyes]

would it help to "soften" the wave a bit (not sure what the correct term for that is...)

[Gina]

Gina, could you try running the 5V TEC only and see if you still get interference? Would be interesting to know if the interference feeds back through the 12V power line or whether it's RF.

Well I could have done but I've unsoldered the MOSFETs now ready to put them next to the TECs. Actually, I've had an idea. The 5v TEC is the one that's on the cold finger and that in turn is against the sensor and presently isn't earthed No problem when I was running off a fixed DC supply, of course. I guess I should earth the cold finger Alternatively, I guess I could run the smaller TEC at a constant voltage - maybe with a manual switch to go between 3.3v and 5v and just control the 12v TEC with PWM. I think this could work - I'd planned to control both TECs to share the temperature drop but is this really necessary? I wonder. I think I'll try set point control by actually controlling just one TEC automatically. Much simpler.

[yesyes]

That's one of the reasons why I use 12V for both TECs. Both controlled by a single MOSFET.

[Gina]

RF from a 555? its more likely to be power pulses, from the sharp square wave on/off nature of the pwm switching the high current load, a very large reservoir capacitor might help.

I think I'm going to get me one of these 20A supplies or one of these 30A supplies as I'm currently using my amateur radio supply and it is needed elsewhere.

Yes, agreed. I have 100nF bypass capacitors plus electrolytics across the camera supply. It might want more supply decoupling - thanks for the suggestion

[yesyes]

That 20A one is the one I'm using at the moment...

[Gina]

For power I'm using a PC PSU. 12v @ 15A and 5v @ 30A.

I'm using a 5v 20W and a 12v 50W TEC so both take around 4A at full input power. I had thought of connecting them in series but due to the back EMF this results in instability. Can you see anything wrong in running one TEC from a fixed voltage and controlling the other, Chris?

[yesyes]

Not sure, really... But if I would do that, I would most likely control the one on the cold finder and keep the 2nd one running on full as a heat sink.

[Gina]

Yes, I think you're probably right. It's important to take the heat away from the hot side so that it doesn't overheat - the cold side doesn't matter. Means I'll have to earth the cold finger - maybe make a new one with a tab I can connect an earth wire to. Or I suppose I could put a thin copper sheet between TEC and cold finger and earth that. The thin copper sheet should come tomorrow with luck.

[nightvision]

Thoughts and observations:

Reading the blurb on TECs it appears that minimising the temperature gradient (and stress) across each TEC prolongs its life so it makes sense (from that perspective) to adjust both to share the gradient, with your temperature monitoring at each stage you can already see the local condition.

I hadn't noticed any mention of back emf, is that as a result of the TEC inherent inductance and the switching transients from the PWM approach to control?

I know that PWM control is recommended against longer term cycling/switching (which can stress the TEC) but what I haven't seen is any mention of a linear approach with soft starts and gradual fine control of the applied voltage.

[yesyes]

I think that's because it's more effort to control the voltage than it is to use PWM at full voltage, especially with the relatively high currents involved here. Depending on how you do it, regulating voltage generates more or less heat. Switching a MOSFET full on or full off barely generates any heat.

I seem to remember reading that TECs are a purely resistive load. Are you saying it also has inductive properties?

I also seem to remember reading that a TEC works the other way around as well, i.e. having a temperature differential between its sides generates a voltage. Could this generate interference during the PWM off cycle when there's quite a temperature difference, feeding a voltage back to the contoller?

[nightvision]

That makes sense Chris, it would be a challenge to control at these currents. It was Gina mentioning back EMF that got me thinking about where it was coming from, the TEC generating it directly from local temperature effects might be the cause, interesting challenges! I suspect it's the transient spikes across the whole circuit that is the culprit for back emf and instability.

If you view the use of PWM as effectively an un-rectified and or un-smoothed switched-mode-psu output then the lack of appropriately tuned smoothing (targeting the resulting output frequencies and very low output impedance) would mean the TEC is being switched off and on rapidly, with good smoothing it should see a gradually varying voltage (with no heat problems).

[Auntystatic]

That 20A one is the one I'm using at the moment...

Yes I have seen yours and others power boxes, they are very impressive, don't you have a separate psu for your camera connection? I have 2 different types of battery and so will need 2 psu's in the box, I have just ordered the bits to do the automatic dew heaters that's in the yahoo groups files, although I have no idea at this stage of how an arduino works, I'll attempt Gina's cold finger mod as soon as I pick up enough courage.

[Gina]

Thoughts and observations:

Reading the blurb on TECs it appears that minimising the temperature gradient (and stress) across each TEC prolongs its life so it makes sense (from that perspective) to adjust both to share the gradient, with your temperature monitoring at each stage you can already see the local condition.

Yes, I remember reading that now too - thanks for the reminder That was why I decided to share the control in the first place. Back to Plan A then

I hadn't noticed any mention of back emf, is that as a result of the TEC inherent inductance and the switching transients from the PWM approach to control?

I'm talking about the Peltier effect whereby an EMF (voltage) is generated in the device by the temperature difference.

I know that PWM control is recommended against longer term cycling/switching (which can stress the TEC) but what I haven't seen is any mention of a linear approach with soft starts and gradual fine control of the applied voltage.

As with switch mode power supplies, we are dealing with high currents (in general) and analogue control of high currents results in a large heat dissipation in the controller/regulator. The difference between switch mode control of TECs and switch mode PSUs is that in the PSU, use is made of inductors to smoothe the output whereas with TECs there is no such smoothing using the standard circuit.

This however, puts another thought in my mind - that of smoothing the rise and fall of current with a choke (inductor). Taking this further (because we want to reduce the weight at the camera) we could use variable switch mode PSUs to control an analogue voltage to each TEC. Food for thought - more complicated but one way of solving the interference problem.