OGMA AP08c sensor analysis questions

[jbnelson]

Hi Robin,

I recently acquired an OGMA AP08c camera and have performed several RAW8 & RAW16 sensor analyses in HDR mode using both the native and the ASCOM OGMAVision drivers and have found discrepant results. For example, in RAW16, while the FWC is similar for both drivers (13k @ gain 100), the dynamic range and read noise are quite different (ASCOM: 15.15/0.36, native: 7.83/56.49 respectively). The ASCOM results are more similar to the published results which were also done using the ASCOM driver although the FWC was quite higher in the published results (28k @ gain 100). As you can see in the attached graphs, my sensor linearity was poor across the board but are consistently reproducible even when varying the light source from an iPad screen light with the camera capped with a translucent cap to ambient indoor daylight (evenly lit white wall) with my Newtonian covered with a white T shirt. The camera firmware is current as is the latest version of SharpCap which, as I read, included an updated SDK for OGMA cameras. Once again, my sensor analyses were done in HDR mode, not LCG or HCG modes, since I purchased the camera to use the HDR mode capability.

I guess my primary questions are:

1) Should I scrap the sensor analyses due to the lack of linearity (any thoughts as to why?) or rely on the ASCOM driver to perform a smart histogram evaluation and also using it to image instead of the native driver?

2) Why are the native driver results so discrepant from the ASCOM driver?

3) Is there anything that would suggest a problem with the sensor itself?

Thanks in advance for any insights or suggestions.

Jim

[admin]

Hi Jim,

to be honest, sensor analysis in HDR mode isn't something that I've tried or tested - it's available to run because there is nothing in the code to explicitly block it from running (as I block analysis in RGB modes), but I have a suspicion that the results will be unreliable...

Sensor analysis relies on getting the unaltered ADU readout values from the camera - this is why it is blocked in RGB mode (the pixel values are already mixed together as part of the debayering process) and why best practice is to disable all colour balance adjustments or other image tweaks like gamma when doing the analysis.

By the very nature of HDR, it will be mixing data non-linearly to bring the short and long exposure data together into a single value - this will confuse the sensor analysis code and is probably leading to the anomalies that you are seeing - for instance the lack of linearity is probably to be expected with a camera in HDR mode.

My belief is that the HDR will be a digital processing step after the pixels have been read out in either LCG or HCG mode, but I cannot be 100% sure (I don't have access to the IMX585 sensor datasheet to check). If I am correct, then the right thing to do from a sensor analysis point of view is to run it in LCG/HCG mode (not HDR) as those results determine the raw behaviour of the pixels before the data is passed to the HDR process. However, what is unclear is exacltly how the HDR works - for instance if you set the exposure to 100ms and choose HDR, what goes on the in the background to get your HDR image? Does the camera capture a shorter and longer exposure? Is 100ms used as the short or the long time? Is it something different...

Sadly it's all rather unknown, and that is, I suppose, because HDR can mean a lot of different things when you look at the detail, and we don't really know what it means in this case...

cheers,

Robin

[jbnelson]

Thanks Robin for the thoughtful reply. I did perform sensor analyses in both LCG & HCG modes (RAW16) using the ASCOM driver and did obtain much more accurate results that were very similar to the published results. However, the results with the native driver (normal mode) were again quite off similar to what I found in HDR mode (the ASCOM driver was at least in range of the published results).

From a practical standpoint:

1) Should I use the ASCOM driver rather than the native driver for imaging?

2) Try performing a smart histogram analysis with the ASCOM driver in HDR mode and see if the results seem reasonable or perform the analysis in HCG mode and apply those results to image in HDR?

Otherwise it seems it will be an empiric approach when using HDR of setting the preferred gain & offset and using the histogram to determine exposure.

Thanks again for your help!

Jim

[admin]

Hi Jim,

I wonder if what is happening with the native driver is that SharpCap is auto-handling the LCG/HCG changeover for you as part of the gain adjustment. This is my preferred approach when SharpCap has the necessary data to deal with the changeover, as many camera users will not be aware of the consequences of choosing LCG or HCG, or under what circumstances they should choose one or the other.

If the analysis you are getting with the native driver shows a large step drop in read noise at about gain 1300 to 1400 then this is what is happening - nothing to worry about, it's just SharpCap switching automatically to HCG as soon as it is possible to do so.

In general I always recommend using the native driver unless something prevents it from working properly - it will let you access more functionality of the camera and tends to be better at recovering from small glitches in operation.

cheers,

Robin

[jbnelson]

Yes that is exactly what it does when using the native driver in normal mode. That drop off in read noise is of course at the expense of a significant decrease in FWC:
*Gain 100: RN 6.52/FWC 39239 =>LCG
Gain 1355: RN 5.72/FWC 4752
Gain 1365: RN 1.81/FWC 4908=>HCG

*my data, ASCOM driver, RAW16

The theoretical advantage of HDR is it maintains a high FWC with very low RN:
**Gain 100 (my data): RN 0.36/FWC 13,077
Gain 100 (published data): RN 1.92/FWC 28,510

**Both analyses done with ASCOM driver, RAW16

So if I use the the native driver, normal mode I’ll need to make a choice between RN vs FWC based on gain settings whereas theoretically with HDR at low gain of 100 I still get fairly decent FWC with even lower RN than HCG.

If I’m interpreting your recommendations correctly, despite the erroneous sensor analysis results with the native driver in HDR mode, it would be better & still provide reliable results to use it for imaging rather than the ASCOM driver HDR mode, correct?

Jim

[admin]

Hi Jim,

one more thing to throw into the mix is that considering FWC on its own is not quite right - you should consider FWC * number of sub frames.

Let's suppose that the conditions you are imaging under require 120s sub frames at gain 100 (LCG) - that means you take 60 subs per hour, to a total FWC of 1,117,170 electrons.

Now, switch to HCG at gain 1365, the reduction in RN from 6.52 to 1.81 is a factor of 3.6, which means that the sub length can be reduced by a factor of 12.9 (read noise ratio squared) and still get well exposed subs of 9.2s. You will get 388 of those in your 1 hour of imaging, with a total FWC of 1,904,304 - ie you get a higher total FWC in the stacked image from the higher gain HCG mode because you can take more, shorter, subs while still exposing them appropriately.

Now, of course, 9s sub exposures might be a bit short from the data processing point of view, but the key point is not to consider the FWC of a single frame (unless you don't stack images of course) - consider the total FWC of all the images you take in a particular time period.

On the subject of ASCOM vs Native - in native you can adjust things like the black level, USB Speed, Fan settings and colour space without needing to go into the ASCOM setup dialog, so there is a convenience improvement as well as an expected improvement in performance and reliability. Behind the scenes, both will be setting up the camera via the same SDK, so should give the same frame data.

cheers,

Robin

[jbnelson]

I had not thought in those terms but it makes complete sense. BTW in the meantime I repeated the sensor analysis by setting the appropriate offset based on desired gain & exposure which I had forgotten to do before and at least for the ASCOM driver in both LCG & HCG modes the results showed >90% linearity & 60% linearity with the native driver in normal mode with the expected conversion to HCG at gain 1365.

Overall, by your calculations, imaging in normal mode at the higher gain ie HCG yields even higher FWC in the final stacked images with significantly improved RN. Still not sure how the HDR mode factors in based on your earlier comments although the offset I arrived at for HDR at gain 100, 60s exp was 2350. I suppose I can try these values and compare the images to the those taken in HCG. I suspect they may not be significantly different with adequate integration time.

Thanks again so much for your thoughtful analysis!

Jim