A different flat option for IMX294 based camera owners

[admin]

Hi,

yes, it's not entirely simple to follow what is going on... I'm afraid I won't be putting together a diagram, but will do my best to explain...

There are two things you can change on the 294 sensor (and the 492 mono version) to affect the gain

1) An analog gain control which can go from 0 to at least 27dB - that's a brightness range of a factor of least 22 times

2) the LCG/HCG changeover switch, which gives an additional gain of about 4.5x (about 13dB) when turned to HCG as compared to LCG. As we know, the HCG mode has the advantage of significantly lower read noise.

ZWO have taken the decision to hide this complexity from the end user behind a single gain control, which has worked well for other sensors as it switches automatically to the lower noise of HCG mode as soon as possible when the gain is being increased.

In detail, what happens is this

* Between gains 0 and 119, the camera is in LCG mode and the analog gain is used to increase the brightness as the gain is altered

* At gain 120, the camera is placed into HCG mode and the analog gain is set back to minimum (zero). You may notice a slight brightness jump between gain 119 and 120 due to this, although this will depend on the particular camera as the amount of gain in the LCG->HCG change has slight variations from sensor to sensor

* Above gain 120, the camera stays in HCG mode and the analog gain is again increased - so for example with the gain value set to 220, the analog gain will be set to 10dB (for the difference between the current 220 and the 120 switch over point).

Unfortunately on the 294/492 sensor, this exposes you to the gain range that Sony warn against using - that is from gain 120 when HCG is switched on to about gain 200 or so (where the Analog gain will reach 8dB, roughly the value recommended by Sony as the minimum for HCG mode).

This all leads to the recommendation to avoid the gain range of 120 to 200 on these ZWO cameras, in particular keep well clear of the lower end of this range where the problem will be worst.

Going back to the initial reason for the thread, it seems to be flat correction that is most badly affected if you end up in the 'bad' gain range, since you can have pixels that are actually saturated but do not *seem* to be saturated (ie they have less than 100% brightness level). This is bad from a flat correction point of view, as those pixels will definitely not be contributing valid flat correction data.

Hope this helps,

Robin

[MarMax]

Thank you Robin, your patience always amazes me. Although I still do not see where this 120 to 200 gain non-linearity exists, I'm just going to heed the advice and shoot flats at either gain 0 or 200. I even made this graph in hopes the light bulb would go on, but it did not. I'm just visually guessing at the data points based on the published ZWO graphs since I don't have the formulas to know the correct values.

EDITED to add that I noticed in another 294 reply you said there may also be an issue at low gains in 47 megapixel mode. Does this mean to avoid zero gain flats or just observe the same "avoidance range" of 120 to less than 200 gain?


EDITED to add that this is based on the 14-bit A/D conversion.

[admin]

Hi,

on the subject of the Bin1 mode, I only have experience with other manufacturer cameras, where I have seen the same problem with failure to generate a saturated image when overexposed at low gains. The same test - massively overexpose and check that all pixels hit maximum brightness will work to detect situations with any problems.

cheers,

Robin

[timh]

Thank you all so much for this thread. Very useful and very relevant to me -- may explain a few odd things I think I have seen with poor flats at gain 124 using filters.

I suppose that this saturation problem will be less of a problem with lights -- because it is only some pixels and only will affect bright high SNR parts of the image?

Tim

[MarMax]

This is a bit off topic from flats but I'm still confused regarding the 294MM operating in Bin1 (46 Megapixel; 8288x56440). Much of the discussion in this thread is regarding the 294MC in Bin1 (11 Megapixel - 4144x2822) which is likely equivalent to the 294MM Bin2. This still leaves a lot of unknown regarding the 294MM Bin1.

My first test of the "failure to saturate" anomaly resulted in vertical spikes on the right side of the histogram for zero (0) gain, 120 gain and 200 gain. Afterwards I ran a Sensor Analysis with the a good steady light source. I ran the 294MM Sensor Analysis in Bin1 mode and the results just confuse me. In Bin1 the sensor is operating in HCG mode and the gain range is -8.07 to 27 dB (0 to 270 per the ZWO graph), yet the sensor is operated from gain 0 up to gain 500 and the results are in the text file.

I'm confused with the ability to operate at gains above 270 in Bin1. Also regarding the 294MM I don't understand if Sensor Analysis needs to be run twice. Once for Bin1 (46 Megapixel; 8288x56440) and again for Bin2 (11 Megapixel - 4144x2822).

[admin]

Hi,

the problem is that I can't be sure exactly how ZWO have programmed the sensor in hires (bin1) mode. I have the mono sensor Altair camera, and can hit issues with certain gains, but perhaps ZWO have worked around that or locked out those gains in bin1 mode. You will just have to trust the measurements that you have made, which sound like everything is OK if you can get a fully overexposed image from gain 0.

The sensor analysis *does* have to be run twice - once for 11 Megapixel and once for 46 Megapixel. That's why SharpCap forces the modes to be distinguished in that way rather than by binning number - to allow for the separate sensor data sets.

cheers,

Robin