ZWO ASI220MM sensor analysis - unexpected results

[ONIKKINEN]

Purchased an ASI 220MM in hopes of solving an issue with dropped guide frames that is either the fault of my ASI 120MM or a USB connection gremlin somewhere, and naturally i chose the new camera first as a good excuse to buy new astro toys. (and also replaced all cables etc)

Below my sensor analysis done with sharpcap pro:

And then the graphs from ZWOs website:

You will notice that these 2 are in disagreement as to the amount of read noise measured (amongst other things). I dont think it performs worse than advertised, its just different which is just as annoying if you were to take the ZWO reported values as is and set the gain to 106 expecting 1e- read noise while you should be aiming between 150 and 200 for that. Anyway thought it interesting as someone is bound to ask a question one day as to what gain you should use for guiding with one of these. Personally i think i will be using somewhere in the 119-200 range with mine.

And why care, its just a guide camera? I have an awkward combination of kit where the only place to put a filter is on the coma corrector, so the guide camera in an OAG is also filtered which makes guide frames very noisy when shot through a narrowband filter. The old 120MM technically works but exposures are longer than i would like and still very noisy, so the low read noise is welcome in this application.

[vlaiv]

Everything about those results disagrees with published specs.

At gain 0, SharpCap measured:

4.12 e/ADU, 16K full well capacity while camera claims that it has only 8K full well and that 2.1 e/ADU.

Specs say that unity gain is at gain 68 while sharp cap places it somewhere above 121

[ONIKKINEN]

1 hour ago, vlaiv said:

Everything about those results disagrees with published specs.

At gain 0, SharpCap measured:

4.12 e/ADU, 16K full well capacity while camera claims that it has only 8K full well and that 2.1 e/ADU.

Specs say that unity gain is at gain 68 while sharp cap places it somewhere above 121

I suppose its possible something about the measurement is wrong, but i have used the same tool to measure my Rising Cam and my 678MC, which both more or less agreed with the reported graphs. Just find it odd that there is such a difference.

[ONIKKINEN]

It occurred to me that ZWOs own capture software ASIcap has an electron per ADU readout that i believe is read straight from the driver as its not doing any measurements.

I find that minimum gain is reported as 4.96 e-/ADU, unity gain is at gain 139, 0.5e-/ADU at 199, 0.25e-/ADU at 259. So its looks like none of this agrees with any other method. ASIcap is closer to my measurements than the reported ones, but its still not close 🤔.

[david_taurus83]

How would you translate gain value to PHD anyway? Pretty sure it only goes up to 100. Is that 400 in real terms?

[andrew s]

Not directly related to your issue but ZWO released new drivers  etc. in early Jan that solved the issues I was having with my ASI 294mm pro. There is also a firmware update for the ASI120MM. 

Regards Andrew 

[ONIKKINEN]

7 hours ago, david_taurus83 said:

How would you translate gain value to PHD anyway? Pretty sure it only goes up to 100. Is that 400 in real terms?

I could definitely set it to at least 200, did not check how high it could be set as the very highest gain settings make little sense for guiding.

By the way there is another confusion here, the gain goes up to 600 even though the sharpcap measurements and ZWO diagrams end at 400. Go figure.

[vlaiv]

1 minute ago, ONIKKINEN said:

I could definitely set it to at least 200, did not check how high it could be set as the very highest gain settings make little sense for guiding.

By the way there is another confusion here, the gain goes up to 600 even though the sharpcap measurements and ZWO diagrams end at 400. Go figure.

Why don't you perform your own measurement?

Measuring of the gain depends on simple fact - shot noise is related to the level of signal if both are expressed in electrons.

Say you have 100e of signal - you should measure 10e of noise (minus read noise and other detail that we won't get into now).

Say you have some e/ADU value which you don't know - let it be 3.7 in this case.

you will get

370ADU and standard deviation of 37ADU

You need to use those two numbers to get your coefficient.

sqrt(370/C) = 37 / C

370/C = 37^2 / C^2

370 = 37^2 / C

370 C = 37^2

C = 37 * 37 / 370 = 3.7

There you go.

You only need to take "flat" at certain gain setting and measure average ADU value and stddev in ADUs to derive gain constant.

[ONIKKINEN]

2 hours ago, vlaiv said:

Why don't you perform your own measurement?

Measuring of the gain depends on simple fact - shot noise is related to the level of signal if both are expressed in electrons.

Say you have 100e of signal - you should measure 10e of noise (minus read noise and other detail that we won't get into now).

Say you have some e/ADU value which you don't know - let it be 3.7 in this case.

you will get

370ADU and standard deviation of 37ADU

You need to use those two numbers to get your coefficient.

sqrt(370/C) = 37 / C

370/C = 37^2 / C^2

370 = 37^2 / C

370 C = 37^2

C = 37 * 37 / 370 = 3.7

There you go.

You only need to take "flat" at certain gain setting and measure average ADU value and stddev in ADUs to derive gain constant.

Ill give this a try later, never thought about how its actually done.

I am assuming offset/dark subtraction and a 16-bit to native 12-bit conversion has to be done so that the measurements and results make sense and only real signal is measured?

[vlaiv]

3 minutes ago, ONIKKINEN said:

Ill give this a try later, never thought about how its actually done.

I am assuming offset/dark subtraction and a 16-bit to native 12-bit conversion has to be done so that the measurements and results make sense and only real signal is measured?

Yep.

That skews results just a tiny bit, and in principle can be accounted for with multiple iterative measurement (of the top of my head).

When you subtract dark - you inject back a bit of read noise - which should be removed, but you can only remove it if you can measure it - and in order to measure it, you need to have e/ADU - see the problem?

If you want to be really pedantic about it - this is procedure:

1. Measure light signal and noise like above and derive gain

2. use derived gain to measure / calculate read noise and subtract that from measurement of step 1 - return to step 1 to get better e/ADU value

(repeat several times).

You would measure read noise (combined with dark current noise) like this:

Take number of darks that you are going to use to calibrate lights and split into two groups - stack first group and stack second group and subtract the two and measure standard deviation - then use that to calculate noise level. If you used sum stack - divide measured stddev value with square root of number of subs. If you used average - it's a bit more complicated - divide with square root of two then multiply with square root of number of stacked subs.

 

 

[ONIKKINEN]

8 hours ago, vlaiv said:

Why don't you perform your own measurement?

Measuring of the gain depends on simple fact - shot noise is related to the level of signal if both are expressed in electrons.

Say you have 100e of signal - you should measure 10e of noise (minus read noise and other detail that we won't get into now).

Say you have some e/ADU value which you don't know - let it be 3.7 in this case.

you will get

370ADU and standard deviation of 37ADU

You need to use those two numbers to get your coefficient.

sqrt(370/C) = 37 / C

370/C = 37^2 / C^2

370 = 37^2 / C

370 C = 37^2

C = 37 * 37 / 370 = 3.7

There you go.

You only need to take "flat" at certain gain setting and measure average ADU value and stddev in ADUs to derive gain constant.

I did some of the measurements (did not take into account darks or the effects of read noise, the results seem within acceptable limits without) and found that at Gain 110 i get an e-/ADU conversion of 0.63e-/ADU, close to the ZWO values and within the margin of error for sure (close enough for me). At Gain 40 i get 1.38e-/ADU, again more or less in the ballpark.

Here is where it gets interesting, i had to switch from the native ZWO driver to the ASCOM one because the camera kept freezing and i had difficulty actually capturing the frames to be measured and got annoyed at having to reconnect the thing. Then after having calculated the 2 different gain values i decided to run sensor analysis on the ASCOM driver, just in case and look at this:

COMPLETELY different result, as if taken with another camera completely when compared to the native driver measurements. This is really bizarre, almost like the native driver works in some kind of higher fullwell low conversion gain mode where as the ASCOM driver version looks like one would expect from the ZWO graphs.

Then after some struggles i managed to make a measurement on the native driver at Gain 131 and measured it to be 0.72e-/ADU which is close enough to the original measurements.

So in conclusion, the camera has wildly different behavior depending on the driver used.

And also, finally answered my original question as to what gain to use. 106, like reported by ZWO, for the HCG mode to activate in the ASCOM driver. Not a clue whats going on in the other driver, im going to stay away from that.

And here is what i mean on which driver:

The ZWO ASI220MM Mini option results in the higher fullwell results, the ASI Camera (1) option results in the expected result. Also, PHD2 will only connect to the ASI Camera(1) option, so no option to use the "wrong" one even if i wanted to.

[vlaiv]

17 hours ago, ONIKKINEN said:

COMPLETELY different result, as if taken with another camera completely when compared to the native driver measurements. This is really bizarre, almost like the native driver works in some kind of higher fullwell low conversion gain mode where as the ASCOM driver version looks like one would expect from the ZWO graphs.

Completely forgot about that - but yes, I've been saying for quite some time that people should really use ASCOM drivers for everything except for planetary imaging. I noticed long time ago that data differs between the two and that ASCOM behaves as one would expect.

[AngryDonkey]

The ASCOM driver is really just an abstraction of the native camera driver and will use the same libraries. There is obviously something different, probably some setting that is applied differently or maybe there are different version of the native driver being used. Are there any settings that are exposed though the ASCOM setup panel?

[Ian McCallum]

On 30/01/2023 at 09:49, andrew s said:

Not directly related to your issue but ZWO released new drivers  etc. in early Jan that solved the issues I was having with my ASI 294mm pro. There is also a firmware update for the ASI120MM. 

Regards Andrew 

I see the ASCOM driver update for the ASI120MM mini, but no firmware update other than v1.0.

Edited January 31, 2023 by Ian McCallum

[ONIKKINEN]

3 hours ago, AngryDonkey said:

The ASCOM driver is really just an abstraction of the native camera driver and will use the same libraries. There is obviously something different, probably some setting that is applied differently or maybe there are different version of the native driver being used. Are there any settings that are exposed though the ASCOM setup panel?

Does this make sense to you?

Few of those options are mysterious to me. Gain and offset pretty self explanatory (also measured offset to be 3200 in 16-bit value so 200 in 12-bit), but the others not so much.

[andrew s]

2 hours ago, Ian McCallum said:

I see the ASCOM driver update for the ASI120MM mini, but no firmware update other than v1.0.

Try FW Udate tab https://astronomy-imaging-camera.com/software-drivers

Regards Andrew

[Ian McCallum]

48 minutes ago, andrew s said:

Try FW Udate tab https://astronomy-imaging-camera.com/software-drivers

Regards Andrew

I did, but it's still showing v1.0.

[andrew s]

31 minutes ago, Ian McCallum said:

I did, but it's still showing v1.0.

That's all there is then. Sorry if I misled you. Regards Andrew 

[Ian McCallum]

9 hours ago, andrew s said:

That's all there is then. Sorry if I misled you. Regards Andrew 

No worries.

[AngryDonkey]

12 hours ago, ONIKKINEN said:

Does this make sense to you?

Sorry I meant what do you see when you select the camera in the ASCOM chooser and then click the 'Properties' button. For me this is what I get. The presets dropdown would probably be the most obvious thing you could investigate if you have it:

 

[ONIKKINEN]

1 hour ago, AngryDonkey said:

Sorry I meant what do you see when you select the camera in the ASCOM chooser and then click the 'Properties' button. For me this is what I get. The presets dropdown would probably be the most obvious thing you could investigate if you have it:

 

Ah this menu, looks about the same, without the bayer pattern and MonoBin options since the camera is mono. Nothing unusual there really, just a gain setting with low,high and manual presets.