Affiliation
Variable Stars South (VSS)
Sun, 01/26/2020 - 08:14
Ed Wiley suggested that this new topic be created, because there was a lot of general discussion about CMOS cameras in his post from last year "Call for Action: CMOS Photometry", in which Ed suggested that it would be useful to have a manual or guide for photometry with CMOS cameras.
The last post in "Call for Action: CMOS Photometry" prior to Ed's suggestion to create this new topic, concerned gain settings and binning in CMOS cameras.
I copied that post into a document which I've attached. No-one (as of a few minutes ago) has yet posted a resonse to it.
Roy
File Upload
Alva
Yes, the principles of DSLR photometry translate to colour CMOS astronomical cameras, and of course you have the big advantage of a cooled sensor.
Roy
Hi,
It seems ASI183MC color filters have significant leek in infrared (especially green filter)
https://astronomy-imaging-camera.com/product/asi183mc-pro-color
So it would be nesessary to use infrared cut-off filter.
Best regards,
Max
Hello! Just curious how you determined the red leak in order to characterize the camera. I'm thinking of adding DSLR or OSC to my photometry. Best regards.
Mike
See manufacturer's data by the link in my previous message. There are filter responce curves in one of the plots. It is seen that green filter has significant responce in infrared area.
There is a similar problem in DSLR, cameras have special cut-off filter (wich is removed in astrophoto-modded cameras). So unmodified cameras are better for DSLR photometry.
Regards,
Max
Thank you!
The colored dyes making up the Bayer mask on most one-shot color camera sensors do have a red leak. However, the ZWO ASI183mc has a red cutoff filter as the entrance window into the camera, and so you should be fine. As Max mentions, this is similar to what is done for DSLR cameras. If you ordered the mm (monochrome) version, then they replace the entrance window with clear, AR coated, glass.
I have an ASI294mc camera here that will be sky-tested sometime in the next few weeks. I bought it specifically to test photometry with a OSC, but the project got sidelined because of the AAVSOnet upgrades.
Arne
Thank you for the information! It would be very interesting to see ASI294mc results. By the way, is it a cooled version?
Max
This is really good to know. Many thanks!
FWIW. Many years ago I wrote a script to automate stacking in MaxImDL. It used to be available though the MexImDL website but I'm not sure that is still the case. I still use it with MaximDL 6.11 and Win 7.
You place the script in a folder. Then drag and drop the folder to the folder containing the images to be stacked. Pick the number you want in each stack and go have a cup of coffee.
If you want want, let me know off list at nt7t at nt7t dot com. Just don't ask me to make changes to it :)
Two years ago I observed short-period pulsator AE Uma (SXPHE variable) with DSLR camera (Canon EOS 600D) attached to a Sky Watcher 150/750 Newtonian. Now I made observations using cheap uncooled monochrome CMOS camera ZWO ASI120MM-S. This camera positioned as a planetary and guiding camera. It has a small 1/3" chip, which is its main drawback. Yet a field of view of the setup (~16'x22') turned to be sufficient for many interesting variables, such as T UMi, RX UMa, AE UMa, etc.
Here is a comparison of data for AE UMa obtained using ASI120MM camera + Baader photometric V filter with the data obtained using Canon EOS 600D camera (see attached figures).
In both cases, data were transformed: for ASI120MM+Vfilter one-filter transformation was performed (using the previously defined Tv_b-v coefficient and an average color index of the variable); for Canon EOS 600D two-filter transformation using green and blue channels was done (again, using previously defined Tv_b-v, Tb_b-v, and Tbv coefficients).
Exposure per point was 45s for ASI120MM (for each point 3 frames by 15s were stacked) and 30s for Canon EOS 600D. So integration times are roughly compatible.
It is seen that the ASI120MM camera shows better photometric performance (lesser scatter).
The data reduction process with ASI120MM is simpler and much faster.
So in cases where the wider field of view is not required, ASI120 is probably a better choice. I plan to use it in parallel to DSLR for photometry.
The data shown in the figures are available through AAVSO database (https://www.aavso.org/webobs), observer code PMAK.
Max
I'm considering investing in a ZWO ASI 178 MM (uncooled CMOS monochrome) for three purposes:
The thing that draws me to this camera is how many different purposes it can serve at very low cost.
As I understand things, the main tradeoff betweeen cooled and uncooled CMOS is signal-to-noise due to the free variable of sensor temperature. This may be a clueless question to ask such a group of dedicated enthusiasts, but:
Can I obtain (with appropriate care) what you would consider to be "good photometry data" from an uncooled CMOS camera such as the ZWO ASI 178MM? Or should I save my pennies for a ZWO ASI 183MM Cooled or equivalent (which has been discussed specifically on this thread)? What are the tradeoffs? Many thanks in advance.
QHY has a cooled version of 178 if you like the format.
Dennis
I'm using the ZWO ASI 178mm (uncooled) on a 203mm F/5 Newton for photometry. I find the results in good agreement with others', I like the fact that it has a 14 bit ADC and the ca 25 x 17 ' field that I get from this relatively small sensor on my scope is still big enough. But if your focal length is much longer, the field might be too small to be convenient for brigher objects (with comp stars further apart on average).
The disadvantage of the uncooled version is not just the fact that it's...well... not cooled per se, but that the sensor's temperature will vary over the night with the ambient temperature, which makes darkfield calibration more challenging. If the temperature change is very huge (e.g. for observers in a desert-like environment), you might want to make extra sets of darks "in-between". Also this model has quite a bit of amp glow, which is manageable with dark subtraction (see above) and affects only the corners of the frame anyway.
The camera has an amazing high framerate which can be useful for lunar & planetary imaging or occultation observations etc.
The uncooled version is very light-weight and can be used even on rather small telescpopes/mounts/flimsy focusers and you won't need to worry about the extra drain on batteries from the cooling if you are working "in the field".
There are versions with the same sensor and regulated cooling (just not from ZWO anymore) for roughly twice the price tag.
Cheers
HB
Many thanks for the info. My scope is a 152mm (6") f/5 Newtonian, with an even shorter focal length than yours. I live in an area in which nighttime temperature shifts are not extreme, though I didn't think about darkfield calibration until you brought it up. That's a great point!
I have been catching up on this discussion with a view to understanding the proper settings with regard to the new ZWO ASI 6200 which has a 16 bit ADU. In this case the range of ADUs exceeds the bit depth of the camera, which is 51,400 electrons.
As far as I can see, changing the gain and offset would require a new library of calibration frames for every setting – not something I would like to do. So here is my reasoning and I open it up for critique.
One rule of thumb would be to divide the full well by the maximum number of ADUs. In other words, 51,400 divided by 65,535 or 0.78 electrons for one ADU. However, since there is no such thing as a fraction of an electron, a unity gain would be the most practical setting. Full well would be achieved at 51,400 ADUs, but the smallest possible increment in signal would be one electron at a time, so nothing would be lost there. What purpose would an increase in gain, or an offset serve? Perhaps if the sky background was zero, an offset would “up” the smallest possible increment above the sky background. But in reality, the sky background would not be zero, and the sky background would also be “upped” as well.
I find the gain settings a bit confusing. Whether using the “native” driver or the ASCOM driver, gain settings go from zero to 100, or “highest dynamic range” or “unity gain” or “lowest read noise”. I assume “unity gain” refers to 1 e/ADU, but I am not sure how the numbered gain works. Does 50 gain mean 50e/ADU? That would be strange, would it not?
I think I am past trying to sound intelligent on this matter. :-)
That's a bit weird since according to vendor info, the gain range doesn't even include unity gain;
https://astronomy-imaging-camera.com/product/asi6200mm-pro-mono
??
Also "mind the step" in the read noise plot.. so it would really be useful to know how the gain setting in the driver translates to what is given in the plots.
HB
Cheers
HB
I assumed, which is something that I shouldn't do, is that the software would round the nearest whole ADU for the given number of electrons collected by the pixel. Using the ASI6200 as an example, if the pixel had 18,712 electrons and the gan was 0.78 electrons per ADU, then the exact number would be 23,989.7435... but it would be rounded to 23,990 ADUs. It is impossible to go to unity gain with the ASI6200. The Gain at 0 is 0.78 electrons/ADU and the number can't go any higher than that. So you can never have an actual gain of 1 electron/ADU with the ASI6200. The ASI178 (14-bit) is the same way
Actual gain = 15000 ÷ (214-1)
Actual gain = 0.916 electrons/ADU
Unfortunately ZWO does not offer the cooled version of the ASI178 any more :(
The actual gain (electrons/ADU) can found by this equation
Actual gain = [Full-well capacity ÷ (2Bit depth-1)] ÷ (10ZWO gain setting ÷ 200)
For example, you want to know what the Gain is if the setting is 50 on the ZWO camera software. We can plug those numbers in and find out.
Actual gain = [51400 ÷ (216-1)] ÷ (1050 ÷ 200)
Actual gain = [51400 ÷ (65536-1)] ÷ (100.25)
Actual gain = (51400 ÷ 65535) ÷ 1.778
Actual gain = 0.784 ÷ 1.78
Actual gain = 0.441
We can test this by plugging in the gain setting given by the ZWO website for the unity setting for other ZWO cameras.
According to ZWO, the unity setting for the ASI183MM-Pro is 120 dB (In reality, it's actual 12 dB but ZWO it's not not decibels but rather centibels because each unit is really 0.1 dB rather than a decibel). The bit depth is 12-bits and the FWC is 15,000 electrons.
Actual gain = [15,000 ÷ (212-1)] ÷ 10120 ÷ 200
Actual gain = [15,000 ÷ 4095] ÷ 100.6
Actual gain = 3.66 ÷ 3.98
Actual gain = 0.92 electrons/ADU
Well, it's close but the actual unity gain should really be 113. When 120 is replaced with 113, the value of the actual gain is 0.997 electrons/ADU. Maybe they do something with the offset to make it work out at 120. ZWO never gives a satisfactory answer except that the gain setting is in units of 0.1 dB.
I hope this helps.
I don't think you can argue quite like that. The translation from electrons to ADUs is not done by software or anything that can do "rounding", but by an Analog-to-Digital Converter, after (!) the electrons drained from the sensor are run thru an amplifier, and the amplifier is configured by the gain setting. At least that's my understanding.
CS
HB
Greg,
You wrote the following in post #114:
"As far as I can see, changing the gain and offset would require a new library of calibration frames for every setting – not something I would like to do."
For ordinay 'long exposure' photometry, I wonder how much you would change the gain. The minimum gain is about 0.78 e-/ADU which is greater than unity gain. Any change from that would be amplifying the signal and noise. If your targets are bright stars that go beyond the linearity of your sensor, you would need to shorten the exposure and/or defocus the image.
So, if you stay at minimum gain, you could I presume scale your callibration files for various exposures.
You also wrote: "I assume “unity gain” refers to 1 e/ADU, but I am not sure how the numbered gain works. Does 50 gain mean 50e/ADU?"
Yes, unity gain refers to 1 e-/ADU. The numbered gain is in units of 0.1 dB. To see the relationship to gain in e-/ADU see the manufacturer's web site and manual for your camera. The data in graphical form from the manual is attached.
Roy
You would have an incentive to use higher gain values tho to get lower readout noise. However, since this model has a step-function like dependency of readout-noise vs gain (a regime where readout noise is ca 3.5 e- and one where it is only ca 1.5 e-), there are really only two actually useful gain settings that you would need to build calibration libraries for: one that optimizes dynamical range at the expense of read-out noise (max gain) and one that takes you in the low-readout noise regime .
CS
HB
The problem with using the higher gain settings is a lower dynamic range since the ADU count would be full before the reaching a full well.
Yes, but remember that when you are going after faint targets, you will collect only a few photons per second. to fill the massive well of this model, you would have to do exposures of many minutes to even an hour or so, risking to lose the observation in case of tracking errors etc. If you need to collect photons for an hour, you'll want to split that session into several exposures , without ever using the full well capacity. In that case, it makes sense to move to higher gain , to get less than half the read noise , and still not using the full well capacity.
CS
HB
Low read noise in CMOS cameras has been mentioned frequently in this thread. It is achieved at higher gain settings (fewer e- per ADU). Is anyone actually doing photometry at low read noise settings? If so it would be interesting to hear about the results.
Roy
It works out pretty well for any star brighter than magnitude 14.
For exoplanet observations, I wouldn't want to lose any more dynamic range by raising the gain so high the difference in magnitude between transits and the rest of the time the star is lost.
Hi Tim,
Thanks. From your light curve I can find the data in the AID. The cadence was ~2.03 minutes. Were your exposures 2 minutes?
What type of 'scope and aperture? Camera? Bit depth? Binning? I presume you used well-focussed images? Gain setting (e-/ADU)?
Roy
N/p
Yes
The ES Comet Hunter, a Maksutov-Newtonian with a 6-inch aperture at f/4.8
ASI183MM-Pro. Bit depth is 12-bits converted by the ASICAP software to a 16-bit .FIT file.
2x2
Yes, while the ASI183MM is only 12-bits, the pixels are only 2.4-µm so the star's FWHM is always within the minimum 2-3 pixels across easily.
The gain setting is 3.66 e/ADU (15000 ÷ 4095) in its native 12-bit bit depth. After converting to a 16-bit .FIT, it's more like 0.2289 e/ADU.
Tim,
Thanks. You wrote the following:
"The gain setting is 3.66 e/ADU (15000 ÷ 4095) in its native 12-bit bit depth. After converting to a 16-bit .FIT, it's more like 0.2289 e/ADU."
See attached graphic from the ASI183 manual, on which I have marked a red circle around what I believe your gain setting was. If that's correct, looking at the two panels below the gain graph shows that your setting corresponded to the highest dynamic range and highest read noise.
My question sought to find any observers who used low read noise settings in these cameras for photometry.
Roy
I do this with my 178mm . I hate to do more dark frames than necessary, so if the temperature is more or less constant (I have an unclooled model), I try to shoot several fields with the same Gain and exposure time.
1) I first pick the maximum exposure length I want to use (e.g. if I want to include pretty pictures and it's a bit windy, this could be as low as 10s, but mostly 16sec or 32 sec (only!)). My mount is operating at the upper end of it's weight capacity, so sometimes it get's a bit wobbly. Also for some comets etc you need to avoid longer exposures because of their movement when guiding on fixed stars.
2) based on the brightest star that I'll want to measure/image that night without saturation, I then pick the gain. The value I enter in SharpCap usually comes out as either 150 or 250, but I doubt this is the same units (0.1 dB) as in ZWO ASI spec diagrams. But anyway this is in the high gain/low readout noise regime you were asking about (the 14 bit ZWO ASI 178 has about unity gain at it's lowest (zero) gain setting already, so ANY non-zero gain value takes you into the reduced-read-noise & reduced-dynamical-range regime!
The improvement in read noise is not spectacular (less than a factor of 2) but I see no reason for picking a gain value that leads to higher read noise in a situation where the exposure time you want to use is so short that you have a choice because you don't need the full dynamical range per exposure.
HB
Thanks HB. You wrote:
"I then pick the gain. The value I enter in SharpCap usually comes out as either 150 or 250, but I doubt this is the same units (0.1 dB) as in ZWO ASI spec diagrams."
The true gain in e-/ADU should appear in the FITS headers.
With an uncooled camera, my guess is that small changes in read noise may make no difference because they may be swamped by thermal electrons.
Roy
Unfortunately, SharpCap doesn't include this (EGAIN) in the header. Given some discussions in the SharpCap forum, my impression was that camera drivers are all over the place in the way they let users specify the gain and there is no way for SharpCap to know the true physical e-/ADU gain. I should mention tho that SharpCap has an extensive sensor characterization functionality, so if I wasn't that lazy I could use it to figure out the true gain by experiment....
Depends, I mean I live in Germany so even an "uncooled" camera is not necessarily getting too warm :-)
I could not find data for the IMX178 sensor in a quick search, but for the IMX183, at 5°C sensor temperature (I reach that in winter sometimes) , the dark current is as low as 0.01 e-/s/pix (see [1]), so less than the read noise for short exposures like discussed above ( << 1 min). Now, in summer nights, let's say the uncooled sensor temperature in operation is more like 25 °C sometimes, and the IMX183 then has dark current 0.06 e-/s/pix , still in the same ballpark as the read noise for those short exposures. So my take-away from this is that for modern CMOS sensors, for short exposures, the dark current does not necessarily dominate over the already low read noise and there might be an advantage in deliberately selecting gain to minimize read noise, even for uncooled sensors.
Cheers
HB
[1] https://astronomy-imaging-camera.com/product/asi183mm-pro-mono
I use an ASI183MM camera with MaximDL and the issue is more complex. The ASCOM driver returns an EGAIN that MaximDL uses for the FITS header, in my case ~3.2 e-/ADU when the gain is 0dB. But this is the gain for each bit in a 12-bit ADC, The ASCOM driver shifts the 12 bits left to make a 16-bit value in MaxinDL but the EGAIN remains at ~3.2 e-/ADU. Any program that uses the FITS file is going interpret the EGAIN for a bit in 16-bit, and over estimate the number of electrons by a factor of 16.
I have used Maxim's scripting to correct this to 0.226... e-/ADU in every FITS file after calibration. A bit of a nuisance. Then I found that Maxim does not correct the EGAIN if one averages several FITS files, so again the EGAIN must be adjusted.
Be careful, other CMOS camera drivers and other programs may do something different.
Mark
This is a great thread and I hope to see additions to the current CCD guide for CMOS cameras or a separate CMOS guide.
I noticed something odd yesterday in all of my images (dark, bias, and light frames). All of the counts differ by multiples of 32 when we expected them to differ by multiples of 16 (65,535 + 4,096 since this is a 12 bit camera). Out of curiousity, I tried various gain settings and as expected the results were the same - pixels differed in counts by multiples of 32. There must be something simple I'm missing, has anyone else noticed this? I don't have another 12 bit camera to compare it to but I'd be very interested to know if other 12 bit CMOS cameras had pixel value differences in multiples of 16 or something else.
Thank you!
In post #126 above, Tim wrote (for a ZWO ASI183MM Pro):
“The gain setting is 3.66 e/ADU (15000 ÷ 4095) in its native 12-bit bit depth. After converting to a 16-bit .FIT, it's more like 0.2289 e/ADU.”
I’m just trying to get my head around this. See the 2nd panel from the top in the attachment from the ASI183MM manual. Take two examples: setting the gain to 1 e-/ADU (120 dB units) and 3.6 e-/ADU (0 dB units). At 1 e-/ADU, that’s 4096 e- for a full well (12 bit camera). At 3.6 e-/ADU that’s 3.6x4096 = ~15,000 e-/ADU (see top panel, where full well is shown as 15k ~3.6 e-/ADU.
I think the 1 e-/ADU and 3.6 e-/ADU (assuming the manufacturer’s specifications are accurate) represent the actual gain.
It doesn’t make sense to me to use the expression ‘15000 / 4096’.
Taking this a bit further, the FITS header in my images (from a 12 bit ZWO ASI1600MM Pro) at full well shows the ‘ADUs’ to be about ~65,500 which is 4096 x 16. I’ve never worked out why the software or driver should do this, but it does. However, I believe it is not correct to say the true gain is 3.6 / 16 e-/ADU. Why? Because the 65,500 is just a ‘count’. It is achieved by multiplying the true ADUs by 16, but it doesn’t change the true ADUs nor does it change the actual e-/ADU.
Roy
Sorry, that attachment I planned for the previous post will not upload. See the ASI183MM Pro manual online.
Roy
Hmmm. Upload worked after all.
Roy
I wrote:
"It doesn’t make sense to me to use the expression ‘15000 / 4096’."
I retract that. 15.000 divided by 4096 is about 3.6, the gain according to the specifications, which is correct
What I meant to write refers to: "After converting to a 16-bit .FIT, it's more like 0.2289 e/ADU." 0.2289 is 3.66 e-/ADU divided by 16, which I don't believe is a valid calculation, because the value of 3.66 e-/ADU is the true value of the gain.
Roy
When I submitted .FIT files taken with an ASI183MM, the ADU count can go up 65535 per pixel. Now, the FWC of the ASI183MM doesn't change and it's still at 15,000 electrons per pixel.
Gain = 15,000 electrons ÷ 65535 ADU
Gain= 0.2289 electrons/ADU
I used the same steps but in a different order to come up with the same solution and to show that the Gain, after being converted to a .FIT file, really is 0.2289 electrons/ADU.
Yes, 15,000 divided by 65,535 is 0.2289, but I argue that is not 0.2289 e-/ADU because of the following.
This (ASI 1600MM Pro) is a 12 bit camera. You can't actually have 65,535 ADU, you can only get 4096. With my 12 bit ASI camera, if you take a saturated image at unity gain and look at the FITS header you see ~65,000 ADUs, but as I said in a recent post, software multiplies the true ADUs by 16 to get the ~65,000, which are only 'counts', and the number does not represent what is happening in the sensor. The graphs in the manual for Gain and FW (Full Well) show what is actually happening.
Roy
BTW, SharpCap has an option in its settings to allow you to switch off the automatic scaling to a full 16 bit value range, so if peole like to work with native ADU counts instead for experimentation, you could use that option. it should really just be a multiplication with 16 or 4 for 12 or 14 bit models, but some people might prefer to actually check this themselves.
CS
hb
This topic has been discussed previously in this forum, but I have a new set of results to contribute from my ZWO ASI1600MM Pro CMOS 12 bit camera through an 80mm f/7.5 refractor. For brights stars down to about 9th magnitude, the best precision seems to be achieved at unity gain (1 e-/ADU) by defocussing the images. For fainter stars of 10th and 11th magnitude, the best precision is seen by imaging with the gain up to 5 e-/ADU with well-focussed images.
I'm not surprised that defocussed images of bright stars gives good precision, but for well-focussed images getting better precision with fainter (rather than brighter) stars was not expected.
Roy