Thanks for your very fast answer.
Operating system is Arch linux. I control the ccds via indi_setprop in a little python script.The lines are visible.
Now I looked at the files with varoius zoom levels and found that the indi frames have an additional horizontal line at y=1527 (first pixel is 1/1).

I compared mean biases from two ccds: 460 and 420 and computed the ratio of the means captured with artemis and with indi.
The results are
Indi420/Artemis420: mean=1.000124, min=0.974, max=1.026, sigma=0.00533
Indi460/Artemis460: mean=1.000201, min=0.682, max=1.188, sigma=0.01267

Three laptos were involved: Indi420 and Indi460 on an 9 year old Dell Inspiron, Artemis460 on a Lenovo X220 runnig win7. I don't know which laptop was used to make Artemis420, runnig XP.

Summary: Indi produces a horizontal line on 460, vertical lines coincide in both softwares. Biases from 420 are identical.

I will continue with flats to see if there are issues at higher ADU values.

cherno

Hmmm, if the only difference will be the additional horizontal line I can imagine, that the issue is max packet size. Default value is 8MB and total transfer is 12MB in case of 460. This artifact is in 1/3 of the image and it is not clear if the image is flipped or not. This looks suspicious, it is worth to try to increase packet size. I'll make new build and let you know.

What puzzles me is that also vertical line looks different... If you are using different machines for tests linux vs. windows, it also possible, that 460 is more sensitive than 420 for something related to USB controller on this notebook. Peter

Vertical lines are the same on all three machines involved. But image captured on windows machines are flipped compared to the indi frames. But that doesn't matter.

cherno

OK, I'll make new build tonight. Peter

Hi cherno, can you try this version (with 16mb buffer size):

download.cloudmakers.eu/atikccd-1.3-i386.deb
download.cloudmakers.eu/atikccd-1.3-i386.rpm
download.cloudmakers.eu/atikccd-1.3-amd64.deb
download.cloudmakers.eu/atikccd-1.3-amd64.rpm

Thank you, Peter

Perfect!
The horizontal line is gone.

You've done a great job.
Thank you very much!
cherno

Sorry to say. The horizontal line is back at row 1527. I made one hundred new bias frames with atikccd-1.15-i386.deb on my Arch laptop and find the horizontal line in mean bias frames, and mean darks independent of exposure time. The line is completely removed with bias subtraction.
The horizontal line was absend in 1.3.
Who done it?
cherno

Strange, there is 8MB buffer again and I can't find any tracks that it was bigger in 1.3... Nevertheless, I increased it to 24MB, can you try version 1.16? Peter

The horizontal line is gone.
Thank you!

cherno

I tried 1.16 and the download problem came back... EKOS just keep "Downloading...." and does not move after the capture is done on the 1st frame

The only change between 1.15 and 1.16 is the chunk size. What platform is it? Peter

pi3 ubuntu.. arm