ASI294 (IMX492) sensor quirks

[wimvb]

Much has been written about the sometimes odd behaviour of the sensor inside the ASI294MM, the IMX492.

To get the best dynamic range (highest full well depth), I have so far only used it at its lowest gain. On the other hand, I've always wanted to take advantage of the lower read noise at high conversion gain. Before just adopting a new setting, I've done some homework and read up on this as well as doing my own very unscientific testing.

Robin Glover (SharpCap) has commented on this sensor on various occasions, for example here:

https://forums.sharpcap.co.uk/viewtopic.php?t=5561&sid=0bc6040e8c7a98eb6cba4f0ae1215f51

I decided to test the behaviour of the sensor regarding saturation and found disturbing results.

I set the temperature to 2 degrees C, offset to 5, bin 2, and used 5 and 20 seconds exposures with my flat panel at its lowest setting. At both exposure times, the sensor was saturated (20 s made sure of that)

First a few images, one with low conversion gain (119) and one with high conversion gain (121)

The same stretch was applied to both images

The hcg image shows noise and some banding (maybe difficult to see in the jpeg).

Next I measured the minimum pixel value of overexposed frames versus gain. Here's what I found

The graph shows the lowest pixel value in a frame for 5 and 20 seconds exposures, vs camera gain setting. Gain 120 is where the high conversion gain kicks in.

This is a disturbing behaviour because it means that not all pixels are saturated at gain just above the high conversion gain threshold.

At gain 139 and above, the sensor behaves as it should, but between the switch to hcg (120) and 138, the sensor is not well behaved.

This behaviour is documented in the IMX492 spec sheet.

Tonight will be a clear night, although wiindy, and I'll try to do some test exposures at 0 gain and 139 gain, just to compare lcg mode with hgc mode.

[Elp]

Are you referring to the residual pattern which is left on the image after calibration (it's shown in the link above)? I've never really managed to get rid of it on my uncooled OSC and I normally use gain 200 with this camera. Reading about it I believe it had something to do with bias not being removed from flats correctly at lower gains, as the sensor works differently. A lot of people don't see the issue, you generally won't if you don't push the data hard, after a background extraction in Siril the issue is amplified further.

 

[wimvb]

1 hour ago, Elp said:

Are you referring to the residual pattern which is left on the image after calibration (it's shown in the link above)? I've never really managed to get rid of it on my uncooled OSC and I normally use gain 200 with this camera. Reading about it I believe it had something to do with bias not being removed from flats correctly at lower gains, as the sensor works differently. A lot of people don't see the issue, you generally won't if you don't push the data hard, after a background extraction in Siril the issue is amplified further.

 

I've only used 0 gain for all my imaging, and flat fielding hasn't really been a problem for me. What I see here is first and foremost a subtle banding and noise at high signal, both of which shouldn't be there. What my tests also show is that this anomaly disappears for gain 139 and higher.