A different flat option for IMX294 based camera owners

[Borodog]

I haven't been able to find the full datasheet, unfortunately.

Could you post a screenshot of the relevant section perhaps?

Thanks for your time.

[oopfan]

This was discovered at the ZWO website by a sharp-eyed young man (not me, I'm as blind as a bat):

2. I’ve heard there might be strange patterns on some ASI294MC Pro stretched images, does this mean the camera is defective?

NO, it is not a defect, and it is just a characteristic of the sensor. The pattern calibrates out of the light image with proper flat frames. A proper flat frame is a capture using an evenly illuminated light source, dimmed down to achieve an average ADU value of 24,000 (this is up to the user and their setup – recommended the histogram peak at 1/3 to 1/2 from the left, as shown in the figure below), a 5-second exposure, captured with the same gain and focus point as the light frames.

You can find it here:
https://astronomy-imaging-camera.com/pr ... 6#comments

Scroll down to the Q&A section. Perhaps it is best to direct your questions at ZWO. If you're still dissatisfied, demand a refund.

[RoscoeD]

Hi,

it's in the Sony datasheet for the IMX294/492 - unfortunately I can't share a copy here, but you may be able to find a copy online somewhere if you search a bit. The spec seems to specify that the normal gain should be at least ~8dB before moving to HCG, but most Astro cameras can get away with about 6dB (factor of 2) with no obvious ill effects. If you happen to find the document, search for 'Conversion Gain' inside it.

cheers,

Robin

None of the charts that I've seen use DB but rather linear gain, so I'm struggling at the math conversion (not to mention that in my world 3db is a factor of 2, and 6db is a factor of 4)

[admin]

Hi,

what you are looking for is the full Sony datasheet for the sensor, not the ZWO one. The copy I have is under NDA, so I cannot share, but others have been less cautious - if you search on a popular search engine for

Sony Sensor IMX294 Full Datasheet

you should find a copy in PDF format fairly quickly.

Please note that I am 100% speculating here - I have not tested ZWO294 colour or mono cameras in this regard, but I have worked with that sensor from other manufacturers which is how I know about this quirk of the sensor. As I pointed out earlier there is an easy test - set the gain you want to use and turn up the exposure to a very large value. If you fail to get the whole image to 100% pixel value then you have an issue. If it all goes to 100% pixel value then you should be fine.

cheers,

Robin

[Borodog]

Ok; thanks Robin. After the hint I believe I found what you are referring to, a 166 page PDF. On page 52 there is a table listing the minimum analog gain for 14 bit HGC mode as +8.07 dB.

[ngc1977]

If you are using a camera based on the IMX294 or IMX492 sensor, test the gain that you intend to use before committing to image with it. The way to test is to set up the camera at the gain you want and significantly overexpose the image (long exposure, lots of bright light, etc). If you can get the histogram to be a single spike at the far right hand side then you are OK. If you cannot get the histogram to go to that single spike at the right in spite of massive overexposure then *do not use that gain value* - try a higher one if you want to take advantage of the lower read noise in HCG mode or a lower one if you want a bigger full well depth.

cheers,

Robin

Apologies for replying to an old thread, but this could potentially solve a 3+ year old mystery for me and I wanted to confirm what you're saying here with what I'm seeing in SharpCap.

Here is a screenshot I took while putting my camera through this test.



At Gain 120 the histogram would not move any further to the right no matter how long I set the exposure. On either side of "unity gain" (120) I could get a single spike starting at 119 and 139 respectively.

If I am understanding your post here correctly, this shows definitively that 120 is a gain setting I should avoid on this camera. Should I assume this applies to both lights and flats?

If this solves my hard-fought challenges with trying to calibrate and process images, you'll have a SharpCap customer for life.

[admin]

Hi,

yes, you have definitely got some indication of failure to saturate caused (I believe) by the analog gain being outside the range recommended by Sony for using HCG mode. It's not as bad as you tend to see on the mono version of the camera at gain 120, or as I have seen on other brands of camera using the 294C sensor. Maybe that is an indication of the workaround claimed by ZWO doing at least a partial job of dealing with the problem. You can see on your histogram that the total number of pixels that are less than saturated is moderately small (probably in the 10s of thousands out of 12 million or so), but still noticeable.

I would certainly aim to avoid imaging in the gain range that shows the behaviour, ideally for all types of frames

cheers,

Robin

[ngc1977]

And sure enough, here it is in my ASI294MM-Pro at Gain 120 as well.



Now I'm very eager to test all this out with a set of real data and see how that affects calibration and processing.

[couchcaptain]

Not trying to hijack the thread and this is unrelated to the 294 sensor, but this is a similar behavior I have with the Meade DeepskyImager mono IV (discontinued) camera which has the Panasonic MN34230PLJ sensor found in the 1600 series and many Touptek and other camera spinoffs.

The camera doesn't have a native driver for Sharpcap or anything else beside Meade's copy of a Touptek imaging software, which I don't use and also severely outdated an lacks functions.
The ASCOM driver sorta-kinda works, but noticed that there is a LCG - HCG mode the camera does, but disabled in ASCOM.
Just to confirm, the camera is capable producing amazing images with the right settings, but horrible ones with the wrong settings.
What I noticed is, that after taking 12 seconds sky-flats for one of the color channels as the Sun was setting, that was the only channel that worked, while the other flats that were 2-3 seconds long did not.
I don't think it was the length of exposure that did the trick, but the level of saturation of sensor array. Again, the histogram looks similar to the 1 second one, or even worse since the average value was only 19000ADU and not 32k or more, so it was that exposed to that 2/3rd of the histogram. Also noticed that doing narrowband flats suspiciously work ok, but not with RGB filters. The difference? The length of exposure it took to saturate the flats, but also the subframes as well.
Not totally sure how to avoid making bad flats, but my usual procedure would be taking very long flats and just like with the 294 sensor, picking the appropriate gain settings.

[MarMax]

Robin, my apology for continuing to beat this to death. I was always the guy in the classroom that would raise his hand when the instructor asked if there were any questions and the whole class would cringe.

I'm one of those visual folks that can understand a graph (like in your Post #8) but can not follow what you are saying in Post #3 with the conversions of db to analog gain. I've been following this thread with interest since I may have the pattern shown in Post #1 and as discussed on CN. I've also searched up the Sony datasheet and did not know if it's appropriate to discuss here so feel free to delete this if it's not allowed.

Here is my translation from the LCG>HCG mode switching as provided under Conversion Gain in the datasheet which I found via Internet search and downloaded the PDF and typed up this information to present here.

When using conversion gain high (MCOVGAIN=1), there is following restriction on setting range of analog gain.

---Conversion-----|-A/D conversion-|-Readout---------|-Setting range of-|-Setting range of-|
---gain switching-|-bits---------------|-mode No.-------|-register PGC-----|-analog gain [db]-|
===============|==============|==============|===============|===============|
MCOVGAIN=0-----|---------------------|--------------------|-0h to 7A5h-------|-0 to 27-----------|
MCOVGAIN=1-----|-14-----------------|-0------------------|-4D8h to 7A5h----|-8.07 to 27-------|
MCOVGAIN=1-----|-12-----------------|-1, 1A, 1B, 3, 4--|-4A5h to 7A5h----|-7.54 to 27-------|
MCOVGAIN=1-----|-10-----------------|-2, 2A, 5 to 11---|-608h to 7A5h----|-12.17 to 27------|

Followings are other notations when changing conversion gain by the register.
1.Power consumptions not vary with the conversion gain principally.(But there can be analog/digital power consumption change by the change of output value.)
2.Saturation signal gets smaller according to MCOVG in “Image Sensor Characteristics” on page 31.
3.Noises in dark condition don’t change principally.So usually set MCOVGAIN=0→1 makes S/N ratio higher in according to MCOVG, but please confirm thischaracteristics in your environment since S/N ratio cannot be guaranteed.

And I know you have tried several times to explain this but is there a simpler decipher ring or "explanation for dummies" that can translate the above in regards to the setting range of analog gain [db]? You have said to avoid gains from 120 to less than 200, and that gain 0 is the best choice for flats, but I keep staring at the numbers and do not see it.