High value pixels in flat frame

[Jonk]

I was calibrating and stacking 50 x 3nm ha flats and I noticed a bright pixel in the resulting stack - not seen this before. On closer inspection there are a handful but also a darker pixel or 2.

I get that screen stretches aren't really telling the whole story, so I hovered and checked closer. As can be seen, the brighter pixel is significantly more than the surrounding pixels.

 

It can also be measured for the darker pixel, albeit the difference isn't as much in terms of value %

The flats were taken with a Gerd Neumann panel, so there shouldn't be any outliers, and calibrated with master bias / dark prior to stacking in the normal way.

Here's a screen stretched master flat:

And I've just found the worst pixel, nearly 100%:

Any ideas?

[Jonk]

I might add that these high and low pixels do exist in all of the original subs, but still confusing - I've set the percentile clipping values as low as possible, but this doesn't work.

Setting the percentile clipping values higher also does not affect the result. In fact, the values set anywhere don't seem to have an effect. Have I made a mistake?

I'm wondering if I should do them all again, see what happens. They are a few months old so the stuck pixels might not exist now.

[vlaiv]

1 hour ago, Jonk said:

I've set the percentile clipping values as low as possible, but this doesn't work.

I guess it is just hot pixel of sorts.

In theory, using matching flat dark instead of bias should remove hot pixel - unless it saturates - then it will be just 0 (max_value - max_value = 0).

Using any sorts of sigma / percentile rejection will not work on hot pixels for flats.

It works on lights - only if you dither. That way hot pixel will move around (it will stay the same with respect to sensor, but will move around once you star align the lights). When hot pixel moves around - it get stacked against other normal pixels and percentile rejection sees odd one and removes it.

When you stack flats - hot pixel is just in the same spot every time. There is no odd pixel in the stack - as all are hot in that place - and algorithm can't tell the difference if it's normal or abnormal.

You can use hot pixel map to remove such pixels from flats - but if you dither and use sigma / percentile rejection on your lights - then you don't really need to bother. Those pixels will be wrongly calibrated - but they will be rejected from the stack because of their weird value (or at least they should be rejected if all is good - dithered and all).

[ollypenrice]

Couldn't you dither by moving the panel relative to the scope? Just rotate it a tiny bit between subframes.

Olly

[vlaiv]

3 hours ago, ollypenrice said:

Couldn't you dither by moving the panel relative to the scope? Just rotate it a tiny bit between subframes.

Olly

That won't do anything useful.

Flat field should be the same always - regardless of flat source used - as it is used to correct system response.

We don't align flat frames - so trying to dither them won't make much sense.

[ollypenrice]

38 minutes ago, vlaiv said:

That won't do anything useful.

Flat field should be the same always - regardless of flat source used - as it is used to correct system response.

We don't align flat frames - so trying to dither them won't make much sense.

No, you're right, the hot pixels aren't on the panel. It was rather early in the morning and I guess my brain was still in bed!

Olly

[Jonk]

Understood.

I’m thinking of running the defect map tool on the flats before calibration? I don’t use a master dark on the lights as I have the defect map.

Being an 8300 ccd sensor it’s always a snowstorm but the defect map sorts it out in one click.

I wonder if this will be fine for flats, I’m guessing so.

[vlaiv]

2 minutes ago, Jonk said:

I wonder if this will be fine for flats, I’m guessing so.

Yep, it will do the same.

[Jonk]

Ok I’ll try later and update in case anyone else wonders the same.

[Jonk]

I've just been through this again, but this time applying the defect map tool on the uncalibrated lights (original flats).

The theory worked just fine, I no longer have any 'bright' pixels in the resulting master flat. However, the resulting 'repaued' pixel is darker than its neighbours. Something to look at.

Next up is to apply the master flat to the lights as part of the normal calibration and stacking, and have a very good pixel peeping session!

So far, so good although I will do all the master bias / defect map and flats again, as I'm not confident that they're 'in date' and this might account for the not so perfect pixel repair.