Dark Current bad performance

cacella

Hi,

In testing the cooled 174MM I had observed a really BAD dark current. In the order of 0.47 electrons per second with a -20C temperature. An SBIG ST8 with -10C has a 0.045 electrons per second and it is considered a very NOISY CCD. This means that for exposure longer than 30 seconds the fact that is cooled does not really make a difference in total noise when comparing to a CCD. It is like is not cooled at all !

I have an ST8 and the dark never accumulates much even with very long exposures. With ZWO 174MM the total noise it s 10 times higher than an ST8 with a 900 seconds exposure. Then the better QE does not payoff either.

In short, it seems that it is a LONG way from a cooled CCD even if we are talking of very OLD brands.

Let me know if my reasoning is faulty in some way.

I am using for spectroscopy and results have been nice except that when I go fainter the noise dominates wildly. My expectation was that at the target price it could be a competitor to ATIK314, but it seems light years away.

Regards

james1ca

Is it possible that what you are measuring is amp glow? Cooling won't help to eliminate amp glow and its effects will just grow larger as your exposure time increases. I'm pretty sure all of these cooled CMOS cameras from ZWO have amp glow. In fact, the ASI174 may have the highest amp glow within the entire lineup.

Sam@ZWO

James is right

you need use the center place to measure the dark current because of the amp-glow

the amp-glow is the difference between old cooled CCD

cacella

Hi all,

Unfortunately is not the case. I use ISIS to measure and then yesterday I have decided to measure some my CCDs to compare.

The results are abysmal. I've selected the darkest area, well away from the ampglow. Dark noise is 7 to 10 times worse than in a old ST8.

In fact this makes ASI174MM useless for dim signals even if with bright stars it works OK in spectroscopy. The consequence for normal deepsky imaging is obvious. It means much more background noise.

Regarding ampglow yes, it is there, it is big, but does not affect the area that I work with in spectroscopy.

Take a look below and see the difference in offset(bias) mean level and dark mean level. In SBIG it is small as dark current is much less. In ZWO 174 it grows rapidly. Darks were 120 seconds exposures. Notice also that SBIG is at much warmer -10C than ZWO -20C and this means that the difference is much worse than Ive pointed, if at same temperature.

============

SBIG ST at 5C temperature

Camera gain (inverted) : 2.544 e-/ADU

Read Out Noise (RON) : 18.18 électron(s)

Offset mean level : 1018.1 ADU

Dark mean level : 1027.2 ADU

Dark signal : 0.1922 e-/s

SIG ST at -10C temperature

Camera gain (inverted) : 2.532 e-/ADU

Read Out Noise (RON) : 18.73 électron(s)

Offset mean level : 1041.5 ADU

Dark mean level : 1043.6 ADU

Dark signal : 0.0442 e-/s

ZWO174MM high gain at -20C temperature

Camera gain (inverted) : 0.150 e-/ADU

Read Out Noise (RON) : 4.20 électron(s)

Offset mean level : 155.8 ADU

Dark mean level : 627.1 ADU

Dark signal : 0.5881 e-/s

ZWO 174MM low gain at -20C temperature

Camera gain (inverted) : 0.511 e-/ADU

Read Out Noise (RON) : 6.17 électron(s)

Offset mean level : 161.0 ADU

Dark mean level : 221.6 ADU

Dark signal : 0.2581 e-/s

My conclusion is that there are three possible explanations :

1 - The temperature it shows is not the CCD but from another place and the CCD is at a much higher temperature

2 - The CMOS chip has a VERY HIGH dark current killing longer exposures

3 - My camera is faulty (what I do not believe because Ive got a 178 cooled and it has the same behavior)

Please use this software http://www.astrosurf.com/buil/isis/isis.htm or any other and do your own measures to see if this is true OR my CCD is defective in some way. It is very simple to do. Go to the "Misc" tab and get two bias, two darks and two flats. Set working folder in settings and write the name of files you got without extension. Press GO and that is it.

Here you see in details http://www.astrosurf.com/buil/isis/noise/eval.htm

And here data from some new and old CCDs. http://www.astrosurf.com/buil/isis/noise/result.htm

Compare dark signal fro ZWO and from the CCDS, mainly ATIK 314. Staggering difference. To the point that a friend with a 14 inches telescope can see in details things that I cannot have a hint with a 20 inches scope. The difference is enormous to the point that I am thinking to buy an ATIK as the price difference is not big considering the difference in performance.

Before buying Ive searched all specs , datasheets etc to look for dark current level in 174MM cmos but found it nowhere. This was certainly a DARK SIGNAL :)

Regards

Paulo

riklaunim

You can check also http://www.skyinspector.co.uk/zwo-cmos-digital-video-cameras

I have Atik 314L+ and ASI178MM Cool and I did some basic dark frame comparison - and Atik has quite clear and low on noise dark frame while ASI has some additions of amp glow (less than in 174). Bit close, but still not there yet.

cacella

Hi ,

There is something odd here when you say that is similar to atik314. Ive read your article, very nice indeed, but it focus the usage of the camera more on very bright objects.

Ill recheck now if I had some issue with my measurements.

I do not care about read noise(it is similar to atik) as exposures are much longer and being dominated by dark current noise.

What I am saying is that the dark current is VERY HIGH compared even with very old CCDs and the final noise is very high as a consequence, drowning details and dimmer stars. In terms of noise It is light years away from an ST8, from an atik is megaparsecs away. To the point that I think is not really usable for dim objects.

I've posted the numbers. If you can test yours and post them (dark current comparison and final noise level in ATIK and 174MM cooled) please do it for me to know IF my camera is faulty of if the 174MM noise is really BIG.

If possible, please check the following :

1 - Put atik and take a bias exposure and see the median level

2 - Take a longer exposure like 120 seconds and check the level again. It will be almost the same meaning that the total final noise will be small and almost constant not regarding the exposure time.

3 - Do the same with cooled 174. I expect you will see a really big difference caused by HIGH dark current and, as a consequence the level of noise grows sharply with exposure.

If this is true the longer is the exposure the worse is 174MM compared with ATIK314.

Please let me know for me to check here too.

Regards

riklaunim

I didn't wrote that article. I have 178 not 174. What you see as dark current is likely amp glow (where 174 has it way more than 178) or really something weird. You can download ASI174 cooled dark frame from http://astronomy-imaging-camera.com/products/asi-cooled-cameras/asi-174mm-cool/ so you can compare it with your camera.

cacella

Hi,

Ive downloaded the dark frame and it is there ! Huge noise level. The average pixel level is 270 in the middle. This is huge.

Ive compared with my darks with ST8 and the difference in levels is staggering. It seems that the cooling make not that much difference. While in ST8 a decline in temperature form 5C to -10C (15 degrees in total) gives a total noise that is 8 times smaller and dark noise 15x smaller, the ZWO174 cooled gives only 3x less total noise with temperature drop from 20C to -20C !!!!

Even if the cooler is faulty the noise is gigantic.

Ive bought also a 178MM cooled Ive tested minutes ago. The same thing. It seems clear that cooling basically do not work as expected.

The decline in noise is VERY SLOW with temperature change.

If possible do the following :

1 - Get 178MM at ambient temperature. Get a bias and dark exposures of 10, 30, 60 and 120 seconds

2 - Turn on cooling. Do the same.

What I have observed is the following : There is a dominance in noise related with exposure time not with temperature. Then cooled or not the noise at 60 seconds is much higher than at 30 seconds. This is not expected to happen.

I see two explanations possible :

- These CMOS chips do not decline the dark current with temperature in a sensible way (what means that cooling is almost useless)

- Or there is a design fault in cooling and temperature shown is not the CMOS chip temperature by far.

I will wait for people at ZWO to explain this better;

bdeclerc

While I won't go into the technicalities as I haven't performed any objective measurements, I have done observations with quite few CCD-camera's and I've used an uncooled ASI174 camera to achieve this - https://flic.kr/p/BQDhSv

This image reaches limiting magnitude somewhere between 21.0-22.0 from a moderately light polluted location using a 30cm Newtonian scope.

So the statement "To the point that I think is not really usable for dim objects." you've made is incorrect for imaging.

It is true that the ASI174 suffers from quite a bit of horizontal banding, which can be difficult to control and which will influence measurements on single raw frames, I'm using some software "tricks" to mostly neutralise the issue, but I can imaging that for scientific observations where precision is paramount, that might not be good enough.

I'm now using an ASI178MM-Cool and it does not suffer from this banding effect.

cacella

Hi,

If you used thousands of short exposures this is OK. What I am saying is about the use of longer exposures from 120 seconds upwards.

I am seeing that to overcome this problem people uses thousands of short exposures. In this case the "cooling" has not much use as read noise will dominate. It is important for longer exposures.

I want to see stack of , at least, 120 seconds exposures with these cameras as it is done rotinely with CCDs

james1ca

I've done some 96s long exposures with the uncooled ASI174MM and those produced some fairly nice images, but the amp glow was hard to control.

I've also done some 180s subs on a cooled ASI178MM and those also worked fairly well.

In both cases, however, you needed very well matched darks and bias frames.

Actually, with the ASI174MM the amp glow is both pretty strong and diffuse. In fact, with a long exposure I wouldn't be surprised if it affected the background level across the entire sensor.

cacella

Hi,

The point is not what you are seeing in the image. Even if you have a bad CCD you can see nice images. The problem is the background noise. The rate of noise with exposure time does not diminishes the way it happens with a CCD. In that way, for long exposures, it seems that the signal is engulfed by background noise that is high.

You can easily see this doing the following. Take a dark shot of 30seconds uncooled. Take another with cooling. Compare the level of background. It is lower, but much higher than in a CCD. In this case you can have 3 times less noise, but a CCD would have at least 128x less noise. This is a world of difference.

When you have very short exposures the read noise dominates but when we are talking about long exposures this happens ONLY if dark current is small, as it happens with a CCD. In 174MM case it seems that longer exposures are dominated by dark current no matter what temperature you set. And this means that it reaches values much higher than read noise. In an ATIK314 the background noise total is the read noise plus a very small dark noise. For you to have an idea we are talking about numbers of 5 and 0.0001 per second. In case of 174MM we are talking about 5 and 0.25 per second. This means that a 20 second exposure doubles the background noise and goes to 5+0.25*20 =10, while in ATIK it goes to 5+0.0001*20 = 5.002. With longer exposures the problem goes worse than this.

In practice this means that faint signals disappear behind the background noise.

bdeclerc

cacella wrote:
Hi,

If you used thousands of short exposures this is OK. What I am saying is about the use of longer exposures from 120 seconds upwards.

I am seeing that to overcome this problem people uses thousands of short exposures. In this case the "cooling" has not much use as read noise will dominate. It is important for longer exposures.

I want to see stack of , at least, 120 seconds exposures with these cameras as it is done rotinely with CCDs

One advantage of lowered read noise is that you don't *need* to do long exposures to register faint objects, I'm not taking them to overcome the Dark Noise "problem", I'm doing it that way because it places far lower constraints on the rest of the equipment, reduces the impact of single "lost" frames due to tracking errors, wind effects, even airplanes crossing the FOV, the lowered read noise allows this - CCD and CMOS are different kinds of devices, and on CCD a lot of engineering has gone into reducing certain types of problems, with CCD's being designed for astronomy or for work with simular requirements - on CMOS this hasn't been the case, but it's becoming more and more so, I fully expect CMOS to catch up to, and probably leapfrog over CCD in the years to come, simply due to the amount of R&D being invested in it, while CCD is "dying" in the long term.

CMOS readouts are measured in milliseconds, CCD readouts are measured in seconds, all the while CMOS retains lower read noise than CCD. If you treat the one like the other, you may get less optimal results, and I can imagine that for certain purposes, current CMOS sensors aren't really any good, but a blanket "not useful for faint objects" is simply incorrect.

merlin66

For AP I understand your argument.

To obtain scientific useful spectral data we need the ability to stack exposures into the hours ( 20 x 600sec exposures) and obtain a spectral image which can be repeatably processed for Darks/ Flats etc. to extract as much real data as possible.

SNR is very "objective" and "flexible" in AP but a real and significant factor in scientific spectroscopy.

cacella

Sorry to disagree but it is bad for dim objects. It is a simple math problem. A single 300 seconds exposure has much better snr than 30x10 seconds. It seems the problem is that dark current is so bad ( and this makes cooling a very suspect addition) that you need very short exposures to maintain noise near read noise.

In fact this problem makes cmos useless for scientific work like spctroscopy, unless the objetc is bright. It is not manageable to have 200x10 seconds exposure. This only makes sense if you are getting a pretty picture of a galaxy and stay there the whole night storing gigabytes of information in something that simply could be plug and play.

All limitations you mentioned like mount, guiding is meaningful only if you have a very limited equipment.

It seems for me that for science these chips are megaparsecs away from a real CCD, unless Sam do the test and show to us that dark current is, in fact, very low, and we need to interpret data in another way.

Regards

riklaunim

For scientific there are sCMOS chips, and for more common people there are advanced CMOS sensors, like in this one case: http://qhyccd.com/QHY42.html

IMX174/178/224 aren't designed for "scientific" or microscope camera work. They are car vision, monitoring cameras, and are very good at planetary/lunary/solar imaging. Adding cooling allows some specific DS imaging, but not in the same way as typical DS CCD cameras.

bdeclerc

cacella wrote:
Sorry to disagree but it is bad for dim objects. It is a simple math problem. A single 300 seconds exposure has much better snr than 30x10 seconds. It seems the problem is that dark current is so bad ( and this makes cooling a very suspect addition) that you need very short exposures to maintain noise near read noise.

I'm sorry, but you're wrong here, a single 300s exposure does not have "much better SNR" than 30x10s through simple math (which you're not actually explaining...)

Simplified, but in good approximation, there are three sources of noise:

1) Shot noise, which is purely a function of total exposure time - the shot noise for 30x10s is exactly the same as for 1x300s or even 3000x0.1s

2) Readout noise, which is purely a function of how many frames you've taken.

3) Dark current (for simplicity, I'll assume amplifier glow is part of the dark current, given how its effects are similar and a good dark frame will remove it from the image) - it is dependent on temperature and total exposure time.

Readout noise is the only one of these three noise sources which is directly dependent on the number of frames taken. Reduce readout noise and you can take more frames of less exposure time and achieve very similar results to less frames of more exposure time.

While technically, you are *always* increasing readout noise influence when taking more frames vs longer exposures, if the readout noise is as low as it is on the newest ASI camera's, the effect is virtually negligeable when comparing 300s or 30s exposure stacks

So you could technically say "a single 300s has better SNR than a stack of 10x30s", saying it is *much* better is simply incorrect (note : for the newer ZWO ASI camera's, not for CCD's with 10x the readout noise)

I have a really good mount (Mesu-mount, +-2" non-periodic error, easily neutralised with autoguiding) and I've done the tests - going from 300s to 10x30s doesn't make a damn bit of difference, even on an *uncooled* ASI174, it might even be a bit better since I can dither between the 10 30sec frames to remove any vestigial traces of pattern noise that might be present.

Going from 30 seconds to 1 second exposure is still a jump too far, the readout noise again becomes a major factor and you start to lose the fainter data, though for bright subjects the increased sharpness afforded by lucky imaging can be worthwhile - for an example check out this superb shot by Emil Kraaikamp : http://www.astrokraai.nl/dump/20151010_2100_NGC7331.jpg

Using an EMCCD, which can have readout noise again ten times better than the ZWO camera's, it becomes possible to go to exposures of 1s or even less with very little readout noise, which allows one to improve on the lucky imaging.

And if we're talking SNR, beating the seeing *will* give you significantly better SNR, since the actual signal will be less spread out, while the noise remains the same.

Look : I'm not saying the ZWO camera's are good for everything, but your blanket statements are simply not correct, or at the very least the lack a lot of nuance. Quite a number of folks are recording faint objects with many short exposures with a lot of success, under seeing conditions where such fine detail was considered impossible to achieve in the past, using equipment that is technically a lot less precise (and therefore typically a lot cheaper) than was the case.

It's absolutely true that this is mostly for "pretty pictures" rather than for scientific research (for deepsky at least) but the improvements shown in CMOS technology over the last few years at this level are quite astounding and I have no doubt that if we give it a few more chip generations, we will start getting CMOS-chips with capabilities CCDs will never reach.

And let's not forget, Dark Current itself is removed by using dark frames, so it's mostly an issue that can be corrected for (okay, I know, not entirely, if dark current is a significant fraction of the total signal, the "shot noise" inherent to the dark current also comes into play), and part of that compensation might be to take advantage of the lower readout noise to take more, shorter exposures and stack those together.

bdeclerc

At any rate, I think we mostly agree - the current ZWO CMOS camera's are not a better (or even equal) solution to CCD's for scientific work like faint object spectrography - I don't think even Sam would dream of arguing that is the case.

What I'm protesting (perhaps a little too much, grin) is blanket statements like "not useful for faint objects" or "much better SNR in 300s than in 30x10s" - those statements are *not* backed by the evidence, a bit more nuance would probably help.

cacella

That is a good discussion :). Lets separate the points to be clear on anyone reading.

1 ) dark current level

This is as simple as it gets. I want to know what is the dark current and if it changes in temperature the way it changes in a CCD. In a CCD it is around 2X less for every 5C temperature drop. That way a dark current of 1 will be 1/256 in 40C drop allowing longer exposures (that is the final objective of cooling !!!).

In my perception there is no trick. This number is crucial to evaluate the camera. An ATIK314 is not that much more expensive in terms of dollars than an 174MM.

2 ) ZWO cameras

I've never said that ZWO promised anything here. As the difference in performance of CCDs and CMOS is blurring with time I think is important for prospective users to know if we reached a point where we have an equivalent option. I was thinking that this would be sooner or later but, as correctly pointed out, they are surveillance chips and they do not care much about dark current for video. My opinion is that we are a bit far from that if dark current is at these outrageous levels and if they are not that sensible to temperature drop at the sensor ( what would make cooling not that much better) .The fact that they (SONY) will have no interest on this may delay the usage as an alternative for real CCDs in scientific astronomy. The camera at QHY has a big pixel and cannot be used in my application.

3 ) less snr in 1x300 than 10x30s

The readout noise of a CMOS is not better than a modern CCD chip like in ATIK314. Even if we are talking of a SBIG ST8 we are talking of a difference from 5e- to 15e-. Because of the dark noise the total expected noise is the same for an 174MM and an ST8 with some 30 seconds exposure or so. For longer exposures the difference would be enormous with advantage for the ST8. This means, without any doubt, that a longer exposure in 174MM will lead to somewhat less snr than even an old CCD. If I am right we are restricted to short exposures up to 30 seconds to have the same thing as in an old CCD (not considering QE that would be a 20% or so advantage for 174MM depending of wavelength)

For those who want to understand better noise take a look at http://www.photometrics.com/resources/technotes/pdfs/snr.pdf. I recommend also here https://starizona.com/acb/ccd/advtheoryexp.aspx

Because when you average frames the noise increases 0.7 for each 1 in light this is different from a long exposure where noise is dominated by dark current (negligible in ATIK). I cannot see the magic here and 1x300 sec has much better snr than a 10x30s no matter the read noise level. When you compare images obtained only with 174MM you are right ! Because of the outrageous dark current levels there is no sensible advantage in longer exposures. But when you compare different cameras this will make the world of difference. That is why CCD is much better for longer exposures. In a ST8 1x300s has a much better snr than a 10x30s. What seems to be at faulty here is the 174MM at 300 seconds because of dark current levels, that is why you see no difference when you stack.

4) dim objects

what happens here is that because we are limited to short exposures we would take ages to have a good snr. It will not appear at all. Usually we overcome this with longer exposures.

My problem is exactly that one. My signal is weak. When I shot a 30 seconds exposure it does not appear no matter how many frames I get. With longer exposures kind of 300 seconds, they appear but total noise is horrible compared to a CCD because of dark current. And it seems that cooling does not make a difference at all like it makes with a CCD.

Then we have 2 Problems

a) dark current seems to be high

b) dark current lowers in a very gentle way with temperature killing the possibility of real long exposures compared with a CCD.

I will be extremely happy if I am wrong because this would mean that we have a more affordable option to a CCD and likely more guys on the boat of a scientific astronomy.

See, what I am looking here are simple answers like that :

a) what is the dark current ?

b) how dark current changes with temperature ?

c) if it is bad is there a way that we can overcome this or Ill need a real CCD ?

ps: if want to take a look on my observatory see here http://cacella.tachyonweb.net/blog/

Thanks for the good discussion.

Regards

Sam@ZWO

this is a very interesting topic!

we wish our camera can beat these CCDs on market

but seems the cooled cameras available right now is hard to beat with long exposure

I think the dark current of 8300 is worst among these CCDs

so 8300 will be our target to beat with our next camera

you are right the dark current is far worse than these CCD from SONY as these sensors are not designed for astronomy to do long exposure

anyway we will try to measure it and provide these info