Cooling any ZWO camera - really really easy. update and opinions please on if a video would be useful?

[Astro Noodles]

39 minutes ago, vlaiv said:

I just wanted to point out what sort of benefits this mod will provide. If it does not provide set point cooling - well, it's just like imaging in winter. Do you see great difference when imaging in winter with non cooled cameras? They can easily end up being in single temperature digits if outside temperature is around zero or maybe slightly below - I'm sure most of us imaged in those conditions.

So you're saying that the only benefit of a cooled camera is the ability to set the sensor temperature? That correct calibration will subtract any noise as long as you can maintain the correct same temperature for the lights and calibration frames? 

So there is  point whatsoever having a camera which will cool to -30deg below ambient?

I have read all over the internet that the cooler the sensor, the better the SNR. Surely then there would be far less noise to have to deal with if the sensor was cold. If, for example, the calibration frames were a couple of degrees out, it would still be a better image because there was less noise in the first place.

Edited July 30, 2021 by Astro Noodles

[tomato]

I think retro fitting Peltier coolers to CMOS cameras can be a good thing, but in my experience you need to be careful. Full on unregulated cooling fitted externally to the casing can easily lead to moisture or ice formation which can get on the sensor window, especially if the camera is pointing near the vertical. 
I use ASI178 cameras retrofitted with regulated set point coolers, so they run at a reasonably stable set point (+/- 2 deg C) so more along the lines that @vlaiv describes.  My preferred set point is above freezing, 5 deg C.

 

[vlaiv]

11 minutes ago, Astro Noodles said:

So you're saying that the only benefit of a cooled camera is the ability to set the sensor temperature? That correct calibration will subtract any noise as long as you can maintain the correct same temperature for the lights and calibration frames? 

So there is  point whatsoever having a camera which will cool to -30deg below ambient?

No I'm not saying that.

Correct calibration will remove signal - never what is true noise. Difference is in randomness.

When you don't cool your camera and have bunch of hot pixels - dark sub that looks noisy - it is in fact not noise - it is signal that is present on all dark frames and it calibrates out. There is some increase in noise but it is not nearly as bad as people think.

Point in cooling to -30°C or even -45°C below ambient is in very special circumstances (otherwise it is just set point temperature that is important):

1. You have high dark current camera (old CCD sensors)

2. You image in such conditions that thermal noise is dominant - like very long exposure that swamps read noise and you don't have LP that causes other type of noise

Camera manufacturers publish dark current data - I linked above one from ASI183 Pro. You can easily check if dark noise is going to be a problem for you or not.

I'm going to show you that dark current noise is not significantly increased in 30°C difference. 

There is something called dark current doubling temperature and it is around 6°C for most sensors (just a bit above actually, but lets go with 6°C even if that gives us worse results). This means that for each 6°C sensor is warmer - dark current increases by double.

30°C / 6°C = x5 - there is five times doubling of dark current signal in 30°C difference.

Two to the power of 5 is 32, so dark current is about 32 times larger when sensor is 30°C hotter.

Dark current noise is equal to square root of dark current. If dark current is x32 larger, then thermal noise will be square root of 32 times larger - and that is ~x5.66. Yes, thermal noise is only x5.66 larger for 30°C difference. If cooled thermal noise is below 1e, then x5.66 that will be comparable to read noise (meaning 3-4e).

25 minutes ago, Astro Noodles said:

have read all over the internet that the cooler the sensor, the better the SNR. Surely then there would be far less noise to have to deal with if the sensor was cold. If, for example, the calibration frames were a couple of degrees out, it would still be a better image because there was less noise in the first place.

In principle yes - but main point is proper calibration and not temperature difference. Cooled sensor will have smaller dark current and hence smaller dark current noise - but it is already small to begin with if we look at modern sensors.

If calibration was out by couple of degrees - then you would not remove whole signal and signal is not uniform and any non uniformity of signal we see as noise. Take for example amp glow. People dread amp glow - not because of the noise it makes, but because sometimes they can't calibrate it out.

Amp glow is really not that big a deal - it is x3-4 stronger dark current - which means that dark current noise is less than x2 in that area. If dark current noise is small to begin with - it really does not make a difference in presence of higher noise - often dark current noise is even small in comparison to read noise.

 

[Astro Noodles]

OK, Thanks Vlaiv.

So, for example, when I do an imaging session with my old DSLR, the temperature can vary by as much as 5 deg C or more. I only take one set of calibration frames per session. Would you agree that the SNR would be better if the sensor was 20 deg C colder? The calibration frames would still be as much as 5 deg C different but there would be less noise in the first place so the images would be cleaner. Not perfect of course.

If that is the case then Powerlord's experiment is perfectly valid for those who don't want to drop £1000+ on a cooled camera.

 

[vlaiv]

Just to show how even on very hot sensor calibration does wonders, here is my ASI185 running now at 40°C (reported in SharpCap) - I took 60s darks.

Dark itself is disaster

Amp glow, countless hot pixels ...

Here is histogram:

Very spread out - and one might think that stars would get lost in there, but look what happens when I subtract two such one minute subs:

Nice flat almost uniform noise (you can actually tell where amp glow is around the corners as noise is just a tad higher there).

Look at histogram:

Very nice bell shaped curve.

In the end - since this camera is not thermally stable, let's look at something interesting. Fits header reports following sensor temperatures:

39.1°C
39.5°C
39.7°C
40.0°C
40.2°C

And look at mean value of frames:

That is why it is important to have set point cooling and maintain temperature with precision of a fraction of degree C.

 

Edited July 30, 2021 by vlaiv

[vlaiv]

13 minutes ago, Astro Noodles said:

If that is the case then Powerlord's experiment is perfectly valid for those who don't want to drop £1000+ on a cooled camera.

Experiment is valid regardless if one has or does not have funds for cooled camera.

Please don't get me wrong - I like the idea of DIY cooled camera even if this version is not set point temperature. Maybe next iteration will have means to stabilize temperature somehow (I don't know what mechanism is used to cool the camera in the first place but maybe it can be controlled).

15 minutes ago, Astro Noodles said:

So, for example, when I do an imaging session with my old DSLR, the temperature can vary by as much as 5 deg C or more. I only take one set of calibration frames per session. Would you agree that the SNR would be better if the sensor was 20 deg C colder? The calibration frames would still be as much as 5 deg C different but there would be less noise in the first place so the images would be cleaner. Not perfect of course.

I don't agree.

Thermal signal is not uniform - look at above example. It can have amp glow or some sort of gradients, etc ...

If you take darks at different temperature - then you are not removing dark signal - you are just scaling it. If you have something that is not uniform and you scale it by some factor - it will still be present in the image. Look at that amp glow in my example above - if we can remove it completely - that is perfect.

But even tiny change in temperature will result in not removing the signal perfectly - look at those mean values - they need to be equal. If they are not equal, when you subtract images - you'll be left with very small signal - like 1-2e in above case, but that 1-2e will not be uniform - it will be scaled version of amp glow.

We can't see it in calibration that I performed because it is smaller than noise when using only two subs, but when you stack 50 or 100 subs - noise goes down and signal stays the same and that amp glow will resurface.

[JP-S]

Enlightening thread - thanks vlaiv. Still, I am curious to see the results.

[powerlord]

Well, I'm afraid it's all moot anyway. It fell at the final hurdle (and possibly many previous ones as Vlaiv points out).

On the bench I was getting 30C reduction, no lens condensation, etc.

Today  I tried screwed onto telescope. With the extra conductivity to the scope, reduction is down to around 23 degrees, which isn't terrible. And there is no condensation on the lens. Maybe with a push fit it would be a bit less conduction to scope and get a bit more cooling.

However, the fan is causing tiny vibrations that I cannot remove. I have tried all sorts of vibration isolation mounts and none are entirely successful. So I'm afraid with that, it is as much use as a chocolate fireguard.

As I say, I'm doing nothing new here, it was just that I found a kit on ebay which had pretty much everything you need, and just happened to have screw holes in the heatsink that miraculously were perfect fit for ZWO cameras. Nothing here was trying for innovation - just ease of fitting and cost. The screw hole fitting making it a literal 5 min job was what made me think it was worth sharing.

This one here (price went up for that seller now, I paid 12 quid):

https://www.ebay.co.uk/itm/265111254730

Big heatsink, big slow fan and powerful peltier.

It just bolts straight onto the back as if it was made for it.

Yes yes yes it's a terrible way to cool.. I'm cooling a big block of aluminium, etc, etc. I ain't daft. But we were going for simplicity here. Not taking everything to bits and trying to rearrange in some sort of custom new case where I could cool only the CMOS.

Idea is - don't worry about efficiency - cool that case down. The inside will then cool down. Yes, you will get condesation all over the case - don't care - case it mental. so what. The small amount of air inside case with condense some water out too, but it will condense onto the inside of the case - it's the coolest bits. So not on electronics or lens.

All of the above proved to be true. In fact, when screwed onto the telescope, you make the lens even better protected from condensation, as the M42 mount or whatever cools, and condensation forms their preferentially.

I proved all of that in tests. It drew about 4a @ 12v, and took around 20 mins to cool to 23 below ambient while attached to scope.

On the bench when my study was cooler I had it sitting there covered in ice, lens still clear, temp -10.

However.. the fan, even though I could feel no vibrations through my fingers, DID cause vibrations when I got to actually try it on a scope. With it on, the asi224 output was jelly. fan off, fine.

I tried about 6 types of antivibration mount (I used to build my own quadcopters to have spares boxes full of options). I could cut it down to nearly nothing, but not nothing.

It is worse case, it's a tiny asi224 sensor fitted on a mak102 with a FL of 1300. But frankly any vibration makes the whole thing a non starter imho.

I then tried no fan - it is a big heat sink after all - but that limited the cooling to only 7 degrees below ambient.

So with that, I'm calling it a day. Sure, you could stand a fan next to it and cool it down, or do some elaborate fan on an arm mounted to the tripod or something.. but any of that and the whole point of simplicity is thrown out the window. So.. it was an interesting experiment, and if anyone want to try you can find lots of sellers selling the same kit - just search for the keywords 'Peltier  Cooling  Kit'

but for me, zee war is over.

Here's some pictures:

 

[Lee_P]

Points for effort!

[vlaiv]

I have to say that I'm surprised that you managed to find kit that fits ASI casing at such low price.

This makes me wonder if there is also option to add some sort of regulation for not much more money.

Even if camera is not cooled much below ambient - I think that it would be beneficial of having it at set temperature that one can replicate later?

Some sort of Arduino board with temp sensor and power driver for Peltier. It does not need to be complicated - maybe even two leds - green and red. Red means temp not reached, green means - go for it!

 

[Ags]

That fan looks much too powerful, and if replaced with a smaller and better fan would easily clear up the vibrations for not too much money. Doesn't matter if the fan screws dont match - happy to mess up the heat sink as long as the ASI stays pristine.

The heat sink is too big for me, but you have me thinking. If the fan makes a 15°C difference, set point cooling can be achieved by regulating the fan...

[vlaiv]

2 minutes ago, Ags said:

The heat sink is too big for me, but you have me thinking. If the fan makes a 15°C difference, set point cooling can be achieved by regulating the fan...

Not sure if that is such a good idea.

There are 3 different temperatures in this cooling setup.

1. Cold Peltier side

2. Hot Peltier side

3. Ambient air

For given current input to Peltier - we have only one thing set - that is temperature difference between points 1. and 2.

If you don't have fan on, then hot Peliter side will get rather hot. Fan helps move away heat from hot Peliter side into environment. With controlling fan speed - you would control temperature at point 2 and then thru set differential temperature at point 1.

There are three things that are potential issue with this.

- You are using 100% power on Peltier all the time (PWM regulation of current to Peltier is much more efficient and uses power only as needed)

- You will get hot running Peltier - which is not good thing - these can burn if you don't provide heat sink on hot side

- you will constantly change speed of fan - and I guess that is more likely to create vibrations than constant fan speed? (not sure about this, but it sort of seems logical).

What you could do instead is drop Peltier element and just go with set point above ambient. You can never reach below ambient this way - but you might still have benefit of having set point above ambient. Maybe small difference of few degrees.

When I gave above example, ambient temp was just below 30°C (really hot these days) and sensor temperature went to 40°C easily. Aluminum casing helps with heat dissipation, but I'm sure large heat sink and active convection will help greatly. I just wonder if fan speed gives you enough control to stabilize temperature.

 

[vlaiv]

Just out of interest, I wondered how much premium does ZWO charge for having cooling added to their cameras. Here are few models that are likely to be modded:

ASI183mc

Regular: $550, Cooled: $800, Difference $250

ASI294mc

Regular: $1000, Cooled: $700, Difference $300

ASI1600mm

Regular: $900, Cooled: $1080, Difference $180

Not sure what to think of that. I do feel that more than $200 difference is both justified and probably too steep.

Justified in sense that you can't get that level of cooling and comfort in DIY for $200.

Compactness, included desiccants and sealed chambers. Heated window so that frost does not form on it. Decent deltaT and very good temperature stability.

It also means added complexity of servicing and higher potential for fail in warranty period.

On the other hand - fact that there is such difference between premium needed to be payed suggest that price is not only formed by labor, parts and R&D but that there is also allure and market demand factor? Parts and labor can't really cost that much as we have seen - maybe just R&D costs, but I doubt that will raise price that much as well.

[powerlord]

temp  control is easy electronics wise - just 4 quid:

https://www.ebay.co.uk/itm/184140691488

But.. you are going to have quite a bit of hysteresis. And gets invasive if you want to start fitting that probe inside the case.

Yes, the fan is big, but that peltier is using 60w of power. without it, the heat sink gets too hot to touch. A smaller fan you just not do it imho. Well, not with that peltier - of course if you wanted to only use something like a 10w peltier and settle for a few degrees cooling, different ball game. That was the problem as I saw it with other people's attempts, they'd stick a tiny fan and heatsink on the case - that just won't work. With a Peltier cooling system the hot sink needs to be MUCH bigger than the cool sink - look at the kit - I'm using the hotsink. the cool sink if it was being use as a fridge is the tiny square heat sink. Here's its the already considerably larger asi case. To be honest, if you wanted to do this with custom parts, I'd be looking for a massive, ideally passive heatsink from a CPU or GPU. Possibly heat pipe based:

https://www.scan.co.uk/products/nofan-copper-icepipe-cpu-cooler-80w-fanless-for-all-intel-and-amd-cpus

But again, complexity, cost, and yer not gonna get that lucky screws match coincidence.

The fan is cheap and unbalanced for sure - but again, the idea was cheap and easy. But if someone wants to try with a balanced ballbearing fan and some vibration isolation, I'd say it has a good chance of working.

add the temp controller above and you may get somewhere, I'd probably just put the peltier on the controller, but keep fan all the time.

If you wanted to get more complicated and vary the voltage to the peltier and fan, then yes - arduino (well esp32 better - cheaper at about 4 quid, tiny, arduino compatible, much faster and wifi out the box so could monitor remotely), but again - it is getting complicated when as you say, $200 bux isn't too bad premium for the work to do all that well.

interestingly I noted that my asi2224 has the cmos on a seperate board like my asi1600 - it makes sense as there was a cooled version. Maybe this is just for cameras that have a cooled version ? So you can see basically that the proper cooling is just to move the bottom board to the side, fit small peltier and heatsink directly to bottom of cmos board. But of course, that isn't going to fit in the case anymore. But then you just need tiny peltier, small heatsink, etc. AND less troubles with condensation.

 

[Tomatobro]

asi178 mono. Fan and heatsink from Artic (CPU cooler kit)

Peltier cooler

Temperature control

 

144109625008

 

[vlaiv]

Anyone tried mirrorless mod like this?

That is probably much more interesting.

For example, Canon M200 is $550 with kit lens (and price will probably go down with time, not to mention second hand availability):

https://www.bhphotovideo.com/c/product/1508687-REG/canon_3699c009_eos_m200_mirrorless_digital.html

That is 28mm diagonal of APS-C sensor vs 17mm diagonal of 1" sensor (ASI183mc vs M200).

Much more real estate on sensor - translates into much faster setup (if utilized properly). Here is comparison between two popular setups:

SW 80ED + matching FF/FR and ASI183mm vs SW 150PDS + M200

Price wise - two same setups really, maybe even 150PDS + M200 has slight edge there, but speed will be different 80mm vs 150mm - almost x4 more light gathered to almost same the FOV. If one bins images to same resolution - advantage for second setup is obvious.

I'd much more like cheap set point temp solution for mirrorless camera than for ASI models that already have cooling solution.

 

[tomato]

The artic fans as used by @Tomatobro are vibration free, certainly down to an imaging scale of 0.47” per pixel.