BIAS frames darker on one side in HDR mode. - Forums

I am building a library of master calibration frames with my Touptek ATR585M. Since I bought it last year I only used HCG mode, now I decided to try HDR. My bias frames are darker on the right, I wonder why and wether it’s a problem or not. I don’t have this issue in HCG mode or on the Dark frames in HDR, only bias frames.

📷 Capture d’écran, le 2026-02-15 à 07.50.25.png

I don’t see that in my ATR585 biases. They look pretty even. Still, I wouldn’t worry if you’re not having calibration issues.

Could be a very slight lightleak, who knows. You can try to shoot Flatdarks instead of Bias frames and see if something changes.

Here is one of my ATR585M one second flat dark masters for comparison (default STF applied):📷 FlatDark_1s.jpg

If you can, try rotating your camera and see if the dark side moves.

I suspect a tilt issue to have your aim off. Based on my experience before I hard mounted my imaging train. For years I used a slide in adaptor for my imaging train. I got rid of tilt when I rigidly screwed everything together, and got rid of the slip-fit booger.

But I used to experience a “dark side” like that. Now, just a teeny bit of dark in each corner, but even. Don’t see any ill effects to my images.

When I put my rotator in, I wanted my imaging train as even and straight as possible. So it is hard attached (threaded) from the focuser tube back.

Eric Gagne:
I am building a library of master calibration frames with my Touptek ATR585M. Since I bought it last year I only used HCG mode, now I decided to try HDR. My bias frames are darker on the right, I wonder why and wether it’s a problem or not. I don’t have this issue in HCG mode or on the Dark frames in HDR, only bias frames.

📷 Capture d’écran, le 2026-02-15 à 07.50.25.png

How did you take them? Camera taken off scope and sensor cap screwed on?

the whole train assemmbled with the cap on the scope. I don’t think it’s a light leak because as I said in my post, it’s only in HDR mode and only the bias. In HCG mode they look just fine and the 300s darks are good in both modes. The picture I posted shows the master bias at the top and the master dark at the bottom, all taken with the same settings.

I would test it again with the camera screwed off to be 100% sure.

If you're suspecting the HDR mode, try updating the firmware. Touptek had to improve the HDR mode several times in the past.

I would add that there’s nothing that says that the bias frames have to be flat and smooth looking on visual inspection. If they are accurately showing the bias signal of the sensor in that mode then all will calibrate.

I have nothing to try them on right now. I made a big mistake last fall and lost all my subs and stacks……but that’s another story.

I haven’t had a single usable night since November so I’ll just wait and see. At least I didn’t get a unanimous “oh man you’re screwed” or anything 😀

LOL, I know what you mean! The HDR mode is an unusual beast that has a few odd aspects but it works extremely well.

Eric — try taking bias frames with a 0.2 second exposure time. I wrote about this phenomenon in an article, but TLDR: Some sensors enter a different mode when the exposure length is very short, meaning that bias frames taken at the minimum exposure time will not properly correct your lights. Taking bias frames with a ~0.2 second exposure time avoids this issue, so long as you ensure that there’s no light leaks.

Noah Tingey · Feb 15, 2026 at 09:57 PM
Eric — try taking bias frames with a 0.2 second exposure time. I wrote about this phenomenon in an article, but TLDR: Some sensors enter a different mode when the exposure length is very short, meaning that bias frames taken at the minimum exposure time will not properly correct your lights. Taking bias frames with a ~0.2 second exposure time avoids this issue, so long as you ensure that there’s no light leaks.

Thanks. I will try that, I took them at 0.0001 seconds, I always thought they should be taken at the shortest exposure the camera allows.

Eric Gagne · Feb 15, 2026, 10:08 PM
Noah Tingey · Feb 15, 2026 at 09:57 PM
[…]
Thanks. I will try that, I took them at 0.0001 seconds, I always thought they should be taken at the shortest exposure the camera allows.

That’s the common advice, and it works fine for most cameras. But yeah for whatever reason some of them behave weirdly when asked to take exposures that short.

I read your article, it’s very interesting, thanks for the link. The examples you have there look exactly like mine. And to my surprise, in my frames I have exactly the same “woven fabric” pattern that you show in your zoomed in ones. I found it strange last night but I thought it was related to the HDR mode.

I’ll post back tomorrow with the results of 0.25s shots.

I couldn’t wait…had to know :D

0.25, all other settings are the same as last night…….no gradient.

My understanding of HDR is that it does both LCG and HCG and kinda merges them, maybe 0.00s is just too short for the camera to do its thing.

My K values are 0.0078 all across now.

📷 image.png

Thanks again Noah !

Eric Gagne · Feb 15, 2026, 12:54 PM
now I decided to try HDR

The question is : what is the process that allow to boost dynamic range ? Is it a succession of two consecutive frames shot with different gain then “merged” together ?

The problem may lay into the merging process itself.

Personally, i would “HDR” with the good old approach : sequence of subs grouped by exposure duration at “magic” HCG gain value. This allows to apply adequate calibration, dynamic recombination may be performed at a later stage in the processing sequence.

The process is called “Clear HDR” and Sony describes it like this:

Clear HDR
When the Clear HDR feature is on, the image sensor captures two images simultaneously, one with a low gain level set to the bright region and the other with a high gain level adjusted to the dark region*2. The images are then synthesized.(Source: https://www.sony-semicon.com/en/technology/security/index.html)

There is nothing broke about the way Touptek implemented the HDR mode. It works very well and sensor problems while shooting extremely short BIAS frames is not reserved to the IMX585 only. Other CMOS sensors might experience banding issues for example.

Once Eric shot “longer” BIAS frames, the “problem” disappeared. Me, for example never experienced any problems ever with any of my cameras and calibration because I only shoot flat darks instead of BIAS.

alpheratz06 · Feb 16, 2026, 01:41 PM
Eric Gagne · Feb 15, 2026, 12:54 PM
now I decided to try HDR
The question is : what is the process that allow to boost dynamic range ? Is it a succession of two consecutive frames shot with different gain then “merged” together ?
The problem may lay into the merging process itself.
Personally, i would “HDR” with the good old approach : sequence of subs grouped by exposure duration at “magic” HCG gain value. This allows to apply adequate calibration, dynamic recombination may be performed at a later stage in the processing sequence.

It takes a single frame but uses both high gain and low gain amps at the same time to create the data sets that are then combined and manipulated in camera for the final 16 bit HDR result. Rather clever really…

Sony’s Clear HDR technology is truly some sort of magic. I don’t understand how the same electrons can be read out simultaneously at two different gains. I have zero image sensor engineering expertise and haven’t taken a quantum mechanics course beyond a couple classes in my undergrad, but when my uneducated brain thinks about measuring the same electron twice makes my brain throw up all sorts of questions about “what about the no-cloning theorem?”, etc.

If anyone has some sources on in-depth technical explanations of how Clear HDR works, I’d be super interested to read them. Though I can’t find much detail with a cursory search; I suspect Sony intentionally doesn’t reveal too many secrets.

Noah Tingey · Feb 17, 2026, 04:10 PMM

It’s simple really. The basic output from the chip is read out and then fed through two different amplifiers giving a high gain signal and a low gain signal. These are then combined via some fancy processing and Bob’s your uncle…:)

Another way to look at is imagine you’ve taken an image of M42, now make two copies, one with a normal stretch that shows all the faint stuff and one with a mild stretch that preserves the bright core. Combine the too and you now have an image that has higher dynamic range and shows both the faint stuff and detail in the core. Same original signal but different processing.

Hope that helps…