Darks have changed dramatically

[ChrisR Oz]

I collected a new library of Dark frames, 17ms to 5m, averaged over 30-50 frames for fast and 20 frames for longer exposures, using a ZWO ASI 294MC Pro camera at -5C acquired in RAW16 in SharpCap Pro 3.2 (glad to support the cause) on a Celestron EdgeHD 8 covered with dark cloths in a dark room (almost) at night.

Looking at them in detail, they seem reasonable, showing a flat distribution with a low standard deviation, and the longer ones progressively showed some amp glow in the top-right corner (minor to the left too).

But checking the average values across a frame, they are around 3130 (with maximum values up to 65532 - the 14 bit ADC values have been scaled to 16 bits in SharpCap). But similar data from some months ago show an average around 440 (still with maxima up around 65532).

The puzzle is - why the change? I did have problems with light leakage into the scope last week when I tried this. Hence, all the covers in a dark room at night (except for a couple of LED power indicators and the dimmed laptop screen facing away).

Before I continue a war against any light leakage, can anyone tell me if there is a way to get a change like this by changing settings, or the complexities of the workflow, in SharpCap?

For example, are the Darks acquired using "Capture Dark" affected by Histogram stretches, etc.? I had used Live Capture, so that may include some existing settings of stretch on that histogram as well, though it was not open at the time. The way the Live Stack histogram is linked to the small histogram is kind of cool - still getting used to these features. But at times like these, it's not clear what may be going on behind the scenes.

Any tips or insight (e.g. into the internals of SharpCap) would be appreciated.
Cheers, Chris.

[admin]

Hi Chris,

this sounds as though you have changed the value of the 'brightness' control at some point. ZWO brightness is what other manufacturers often call 'Offset' or 'Black Level' and it basically just adds a constant value to all pixels as you turn it up. It's useful to ensure that the left hand side of the histogram peak separates from the zero point when using higher gains, but will have just this effect on mean pixel values in dark frames.

cheers,

Robin

[ChrisR Oz]

Hi Robin,

Bingo! Thought this was some post-processing control in SharpCap. But now capturing a short Dark set at 60s exposure clearly shows the pedestal at Brightness=1 (average 106), and greatly reduced at Brightness=0 (average 34).

Interestingly, the ASI Studio has a Brightness control, which like a Gamma, and colour balance, etc. are set at their centre values of 1.0, like something just for viewing, not for Raw capture.

The moral of the story, no doubt, is that ignoring a control that you don't understand the action of (to look into later) will bite you in the end.

Interestingly, the Brightness=0 Dark image has almost no banding, while the Brightness=1 one has noticeable banding.

However, the very left shifted histogram seems to confuse SharpCap in Live Stack mode. Auto stretch works fine. But if I touch the Black Level marker and move it, the image goes black, even if I return it to the far left.

Thanks for the insight. It all makes sense now.

Cheers, Chris.

[ChrisR Oz]

Hi Robin,

As a follow-up question, do you know what settings do not alter the raw data using the ZWO 294MC? Are these correct for "true raw"?

White Bal (B)=50
White Bal (R)=50
Brightness=0

I did find a short reference to Brightness in the ZWO software manual:
Brightness or Offset: This is an offset value added to the output data to avoid any data negative. You may need to turn it up for DSO imaging.

This leaves me wondering how it could go negative?

Thanks again for the tip.
Cheers, Chris.

[admin]

Hi Chris,

the white balance values are correct for true RAW for ZWO cameras as far as I know.

Adjusting the brightness upwards is less of an issue than messing with the white balance. You can still be RAW and have any brightness you like. The best way to set it is to take dark frames at your normal exposure and gain - adjust the brightness until the histogram is not touching the left hand side and then you are done - leave the brightness set at that level for all captures (dark, light, flat, dark flat, etc).

cheers,

Robin

[ChrisR Oz]

Thanks Robin.

Got a good set of Darks last night, using Brightness=0. We'll see if not having a non-zero offset causes me any grief in future.

For short exposure (e.g. 8-67 ms), I see a background average Bias of 2.0 in the Dark images. This seemed to match well SharpCap's sensor analysis for my 294MC cooled to -5 C, which showed a read noise for gain = 121 of 1.67 e at e/ADU = 0.84, and hence an ADU value of 1.67/0.84 = 1.99. All good I thought.

However, the ZWO "ADU" is for 14 bits, and my Dark image has been scaled to 16 bits by SharpCap. So the average of 2.0 is really only 0.5 in 14 bit land.

So, sorry to ask yet another question, but do the "ADU" values quoted in SharpCap's sensor analysis correspond to a scaled 16 bit data range?

Cheers, Chris.

[admin]

Hi Chris,

good question - I had to go and check! It's the 16 bit scaled ADU data that is written out to the sensor analysis data files.

cheers,

Robin

[ChrisR Oz]

Hi Robin,

OK, that tallies with my Dark image. All good perhaps?

But I'm left puzzling about the sensor analysis. The analysis table (pre my Pro license) tallies almost exactly with the ZWO data sheet graphs for the 294MC, which is nice. But the ZWO sheet is no doubt talking 14 bit ADU.

Sensor analysis snapshot - Nov 2020.jpg (70.43 KiB) Viewed 1448 times

From your sensor table, the ratio of Full Well depth (e) over (e/ADU) gives around 16,000, which tallies with 14 bit too.

Still left pondering 14 versus 16 bit values in the Dark measurements ...
Cheers, Chris.

[admin]

Hi,

ok, so the ADU used in the graph display are 'true ADU' - ie 14 bit in the case of your sensor. However, if you go looking at the contents of the saved sensor analysis data files then the ADU are scaled to 16 bit values.

If you have done the analysis and have the histogram showing in SharpCap, then hover the mouse over the histogram area - you get a readout that shows for the mouse position

The pixel value (16 bit scaled ADU)
The true ADU (12/14 bit as per the sensor bit depth)
The number of electrons required to generate that level of signal
The electron rate in e/s/square micron

cheers,

Robin