CCD gain, CCD readout noise & CCD dark current per second

Affiliation
American Association of Variable Star Observers (AAVSO)
Wed, 08/02/2023 - 17:56

Hello All,

To calculate photometric errors, AstroImageJ requires you (in Aperture Photometry Settings) to give it certain values, these are:

- CCD gain [e-/count]

- CCD readout noise [e-]

- CCD dark current per sec [e-/pix/sec]

I can easily get the precise estimate of the first value from the FITS Header (EGAIN keyword). However, I have some questions about the second and the third one.

1. I am using Bin4 (GAIN 120) so how does it change the CCD readout noise? If it does, where should I get the right value from? I suspect that the 1.2 e- is given for the Bin2 mode only - https://i.zwoastro.com/zwo-website/manuals/ASI294_Manual_EN_V2.2.pdf (camera manual).

2. The same question is with the CCD dark current per second. The approximate value of 0.005 was calculated by me using a 600s master dark and a 1s master dark (simple math). According to the camera manual, page 13, it should be around 0.004 e-/s/pix (-15°C sensor temperature).

So I got pretty close (I will use the value given by the manufacturer, of course) and because of that I think the answer is that it should not - that it remains the same no matter the bin one might be using. However, I am not sure and therefore I am seeking advice from you.

Krzysztof Zieliński

Affiliation
American Association of Variable Star Observers (AAVSO)
Binning, noise, gain, and dark current

Krzysztof:

There are two ways to do binning, and I can't tell from the ZWO manual which they use. One way uses a simple "add" to combine all the pixels being binned. In your case, binning 4x4 would simply add the 16 pixels together. The other way uses an "average": the 16 pixels are averaged together. The easiest way to tell which is being done in your system is to take two images of a star field, using the same exposure time. One image should be unbinned and one should be binned 4x4 (or 3x3 or whatever binning you're going to use). Now compare the background pixel values between the two images. If the values are about the same, then your system is using averaging. If the values are consistently higher in the binned image, then your system is doing an "add".

Gain: If your system is adding, then binning does not change the gain compared to an unbinned image. This means that you can use the manufacturer's gain curves with no adjustments. If your system is averaging (instead of adding), then the gain will be higher than the unbinned value by a factor of (binning)^2. 

Readout Noise: If your system is adding, then readout noise increases by (binning) -- not (binning)^2. If your system is averaging, then readout noise decreases by (binning). (This happens because noise adds in "quadrature": the square root of the sum of the squares.) Use the values in the manufacturer's curves and then multiply or divide by your binning factor (4, in your case).

Dark Current: If your system is adding, then dark current increases by (binning)^2. If your system is averaging, then dark current also increases by (binning)^2. (And I would trust your measured value of dark current -- assuming you've done the correct conversion from counts to electrons, remembering that the gain may be different  -- over the manufacturer's curves for modern cameras.)

- Mark M

 

Affiliation
American Association of Variable Star Observers (AAVSO)
CCD gain, CCD readout noise & CCD dark current per second

Mr. Munkacsy,

Thank you so much for your detailed explanations that are so simple at the same time. I do not have this camera with me at the moment, but as soon as it happens I will do as you advised me.

 

Krzysztof Zieliński

Affiliation
American Association of Variable Star Observers (AAVSO)
Readout Noise

I have checked which method of binning ASI294MM Pro uses - averaging. I tried to find the readout noise value for Bin4 using the curves from the manual (as you have advised me). However, I have encountered another problem.

There are two curves for the read noise (its values are given in e-rms) in the manual. One is for Bin2 and the second one is for unlocked Bin1 mode. Based on these curves, for gain 120 the read noise is about 1.9 e-rms for Bin2 and 1.6 e-rms for Bin1. This seems to contradict what you have said - 1.6 divided by 2 is 0.8 and not 1.9. Or perhaps this is an exception. So what should I do?

And I am even more confused by what the manufacturers are saying in the document:

As you can see, the Readout Noise of the ASI294 camera is extremely low when compared with traditional CCD cameras. In addition, it is with Built-in HCG mode, which can effectively reduce read noise at high gain and keep the same wide dynamic range for this camera as at low gain. When the gain is 120, the HCG mode will be automatically turned on. Additionally, the read noise is as low as 1.2e while the dynamic range can still be close to 14bit.

I do not know what is the difference between "e-rms" and "e-" values of read noise - this is probably the source of my confusion. Could somebody enlighten me?

 

Krzysztof Zieliński

Binned and unbinned images

The following is tangential to the thread but it is posted because part of the discussion above relates to the effects of binning.

I use ZWO ASI1600MM and ASI294MM cameras. I used to bin my images (2×2), but ran into problems with saturated pixels.   Because the binned ADUs are averages, some pixels may be saturated, but the display of the peak ADU value in a star image will not tell you that. Of course, it is possible to check peak ADUs in an unbinned image prior to taking your binned science images, but I tired of setting then resetting the value for binning.

I decided to avoid the issue completely by always taking unbinned images. That gives rise to two other issues. First, the unbinned file is four times the size if the binned one. Second, depending on the equipment, images may be oversampled.

To deal with the larger file size issue, I found that I could set the ROI (Region of Interest) to 1/2 but still fit the target and comps I want into the smaller FOV. Of course, that may not always be possible.

Concerning oversampling, I know that it is AAVSO advice to avoid it, but in fact any adverse effects are minimal.

Roy

Affiliation
American Association of Variable Star Observers (AAVSO)
ASI294 Quirks

Krzysztof:

Easy question first: Read noise is always described as a root-mean-square (rms) value. And it's always described in units of electrons. So, "e-rms" and "e-" should be read as being exactly equivalent.

Second: It's easy to measure read noise and gain yourself. (There are many online references; here's one: https://www.cloudynights.com/articles/cat/column/fishing-for-photons/signal-to-noise-part-3-measuring-your-camera-r1929). No matter what you conclude from manufacturers' data sheets, there's a lot of value in making your own measurements -- it eliminates all ambiguity and it isn't hard to do.

Third: The ASI294MM has some quirks. One is that the so-called "unlocked" mode that is also called "bin 1" is different at the hardware level than the default mode "bin 2". (One of the differences is that the sensor shifts between a 14-bit ADC in the default mode to a 12-bit ADC in the unlocked mode.) Bin 1 and Bin 2 on this sensor are not going to be related to each other using the simple relationships that I put in my prior post, which assumed software-level binning.

Because of this behavior for the ASI294MM, it isn't at all obvious how vendors would implement Bin 3 and Bin 4 modes -- there's too much ambiguity. It may also be different as you move from one software package to another. And so I strongly suggest that you measure read noise and gain yourself with your particular configuration.

Another thing you should measure yourself is the linearity curve. Roy posted here his experience with non-intuitive saturation limits on the ASI294MM as you change binning. (The AAVSO CCD/CMOS Guide touches on this topic briefly (but not specific to the ASI294MM) and provides some guidance that is only relevant when the hardware readout process remains the same during binning: that may or may not apply with the ASI294MM depending on which binning modes you compare.) This becomes hugely important with the possibility of shifting between 12-bit and 14-bit pixel values as you change binning modes. It's really important that you have a good handle on what the linearity limit is for the particular gain, readout mode, and binning for your particular setup.

Fourth: It's fairly common for today's CMOS camera vendors to internally shift readout mode at certain camera gain settings. For the ASI294MM, there's a mode change that happens at a gain setting of 120. When you set up for an observing run, you will choose a very specific gain setting. (There's some guidance in the AAVSO CCD/CMOS Guide on how to select gain.) You should use that exact same gain setting (and bin setting) for all your calibration frames as well (bias, dark, flat), and your system gain, sensor read noise, and linearity threshold will be specific to that one gain/bin setting.

- Mark

Affiliation
American Association of Variable Star Observers (AAVSO)
Gain, readout noise, dark current per second

I would like to thank you Mr. Munkcasy and Mr. Axelsen for your advices and insights. 

More than a week ago, I have written an email to ZWO about the issues. In short, I have asked them for system gain, readout noise and dark current per second values for ASI294MM Pro Bin4 mode (120 gain). Nevertheless, I have also measured them on my own. Only measuring the system gain is something I currently cannot do and because of that, I have to trust the EGAIN Fits Header keyword value. The problem is that I still have not recieved the desired answer from the manufacturer, but I was told that my question is with the software team and that they will reply when they can. My plan was that this post will contain a comparison of values measured by me and the "official ones", but it seems that for now, I am limited to my own measurements. In the end, I decided that I will publish here a new post with the ZWO's values if I get them. The reason is becuase I have left you without any response for more than a week and I do not know when to expect the reply from them. So here are my results (I used methods from the sample link given by you Mr. Munkcasy):

Gain: 1.00224268436432 e-/count (according to the FITS Header)

Readout Noise: 3.75 e- (~3.74 StdDev multiplied by Gain)

Dark current per second (for -15°C): 0.016 e-/pix/sec if I use entire images; 0.005 e-/pix/sec if I use areas that are not affected by amp-glow only (CMOS sensor) (question which technique is the correct one; I have heard that amp-glow is said to increase nonlinearly with the exposure time)

 

The system gain [e-/count], as I have said before, according to the FITS Header, remains the same regardless of the bin mode. What is interesting, the camera manual says that the readout noise for Bin2 mode for 120 gain is about 1.9 e-rms, while my measurements for Bin4 say that it is 3.75 e-rms. So that is approximately twice the value for Bin2. Moreover, the curve for dark current per second in the manual indicate 0.004 e-/pix/sec as the value for -15°C for Bin2 while my measurements for Bin4 indicate 0.016 e-/pix/sec (amp-glow included) for the same temperature and that is 4 times (2^2 times) the value for Bin2. It is possible that these relationships are a coincidence.

The calibration frames I used here were captured using ASIAIR, but because now I primarily use Astro Photography Tool I should check whether there is any difference between images taken by these two.

 

Krzysztof Zieliński