Histogram

[Anthony1979]

Where does this histogram need to be for me to take flat frames 

[symmetal]

Where it is is fine. As long as the peaks are well away from the left and right edges is all you need to do.

Alan

[david_taurus83]

What camera is that? You want to be around half the full well depth of the sensor and the histogram wont tell you that.

[Anthony1979]

Nikon d3500

[Anthony1979]

Just to let you no i have never taken flat frames before so im still learning 

[symmetal]

1 hour ago, david_taurus83 said:

You want to be around half the full well depth of the sensor and the histogram wont tell you that.

As long as the histogram peaks are on a linear portion of the sensor transfer characteristic is all that matters I would have thought.  With modern cameras this is pretty much anywhere as long as it's not black or white clipped. Also of course not clipped by the ADC, so away from the edges of the histogram satisfies both requirements . I use my ASI1600 at unity gain always so ADC full well is 4096 e- (electrons). To get half sensor full well of 10000 e- it would be white clipped, unless I switched to minimum gain on the camera. I just take flats at the same gain (or ISO on a DSLR) as the lights. Saves having too many variables.

With Anthony just starting out using flat frames this should make it easier.

Alan

[Anthony1979]

So am i on the right track 

[david_taurus83]

If I use the Histogram on my camera or APT even with its flats aid the flats are under exposed each time. I always open a test flat in Pixinsight to inspect. Using Av mode on the camera gives me a flat of around 3k ADU (bias alone is 2047 ADU!) at ISO1600 when a saturated value is around 15k, no higher. I expose to get around 8k ADU. On the Histogram this is way over to the right. When I had a 600D I used to use Av mode but with an exposure value of +2.

All I would suggest is take a flat of around 10s on a bright target and and inspect it in your processing software to see what the fully saturated value is. On a Canon DSLR I would aim to be around half this value.

[symmetal]

David, your first reply reply confused me as you said 'half the full well depth of the sensor'. Your 6D has a full well depth of around 80,000 electrons at its lowest ISO.

If this is ISO 100, then at ISO 1600 the ADC saturates at about 5000 electrons.  I think you meant to say half the saturation value of the ADC.  The DSLR camera histogram display is often not linear (like the default linear histogram on most capture software) and stretches the dark half more. So half ADC saturation is more towards the right hand side on the camera display.

However, the flat frame calibration maths is

Calibrated (x,y) = Image (x,y) / Flat (x,y) * (Average Flat pixel value)

As long as the flats histogram is not left or right clipped you get the same calibrated output values irrespective of the actual ADU values of the flat, assuming a linear response of the sensor.

If the sensor response is not linear it would actually be more accurate for the average flat exposure ADU value to be more similar to the actual image average ADU value.

We assume a linear sensor response to make things easier and so just need the flats to be not clipped. Putting them around the middle of the histogram is convenient to ensure this but is not necessary. In the Craig Stark presentation around the 50 min mark be says something like as long as the flat exposures aren't 'banging into the endstops' they'll be fine. 

19 hours ago, Anthony1979 said:

So am i on the right track 

I would say yes you are. Good luck! 

To avoid the camera shutter or a 'flickering' flat light panel possibly causing exposure problems just ensure the exposure is at least around a second or so. I don't see any reason not to use a lower ISO for the flats in order to get a longer exposure more easily, as long as you take a separate set of 'bias' frames at that ISO for the flats calibration.

Alan

 

Edited January 24, 2020 by symmetal

[david_taurus83]

@symmetalYes your right, i meant about half the available ADU count. Just seen the OP has a Nikon so possibly completely different to my experience with Canon. But for sake of others who do use a Canon and may read this, don't rely on the histogram for accurate exposure of flat frames.

[Anthony1979]

I dont understand this ADU number and how do i no if there the right numbers.. This is all new to me.... So how will i no i have the right flat frame... Its all puzzling i thought you just take a lot of light images

[symmetal]

CCD astro cameras give a 16 bit output so a pixel value can be one of 65536 values (2^16 = 65536). These values are called ADU (Analogue Digital Unit).

The histograms on capture software display these ADU values, 0 is on the left up to 65535 on the right. Ideally you want the histogram of your important image data to be within these values. To see the faint areas of the image you have to expose long enough that the sky background is above ADU value 0 which means the left end of your histogram is not clipped off at the left hand edge. Bright stars will then invariably be too bright to be correctly exposed and will be clipped at the right edge of the histogram with an ADU value of 65535. Clipped values are lost forever and can't be recovered.

CMOS cameras generally have fewer than 16 bit Analogue to Digital converters, 12 or 14 bit is common. The astro CMOS camera will normally add extra bits of value zero to create a 16 bit value output so that the capture software histogram display still shows 16 bit ADU values. DSLR cameras, I believe, output a 14 bit value in their raw format so the capture software adds 2 zero value bits to make 16 bits again.

This means histogram displays can be compared no matter what the camera bit depth is.

If your flat frame peak is in the middle of the capture software histogram it has a value around 32000 ADU. Usually you can hover your mouse over your captured image and the captute software will display the ADU value under the cursor. So when people talk about a certain ADU value, it has the same meaning independant of the camera producing it.

Your DSLR most likely doesn't have a linear display histogram as mentioned previously, so a value of 32000 ADU as displayed in the middle of the capture software histogram will likely display more over to the right on the camera display. David likes the flat frame peak around this 32000 ADU value, while I'm happy as long as the edges of the bump are not clipped at the sides of the histogram. 😀

Alan

[david_taurus83]

Hi Anthony, sorry if I've confused things talking about a different camera!

As Alan mentioned, most software can tell you the ADU value of the pixel under the mouse pointer. Your camera is 12 bit (2¹²) which means the pixel values can be from 1 to 4096. Think of these values as a simple brightness level of a pixel. 1 being the darkest black and 4096 being the brightest white. In an ideal world, the histogram on the camera should correlate with the ADU values. In your example above, the peaks are just to the left of centre so you would expect the ADU values to be just under 2000. That would be perfect for your flat frames. The point I was trying to make though, is that it may not correlate with the ADU values. Not on my Canon DSLR anyway. So I need to inspect my flat frames in an image processing software to make sure I am taking the correct exposure.

Theres a few things you can do during the day to find out how your raw files are output. Firstly, what image processing software will you be using for your images? Are you shooting in raw mode on the camera?

[wimvb]

On 24/01/2020 at 00:48, Anthony1979 said:

Where does this histogram need to be for me to take flat frames 

 

On 24/01/2020 at 03:28, Anthony1979 said:

So am i on the right track 

Yes. The level looks good. 2 things to add:

1. Use a neutral object to point your camera at when you take flats. That is, as neutral as is possible. This makes subsequent processing a bit easier.

2. (And more important) make sure that object is evenly illuminated. The only variation in your flats should be due to vignetting and noise in the imaging train. Otherwise you're in trouble. Sky flats may not always work.

[Anthony1979]

I put a t shirt over my scope and point it out of my windows would that be fine

[wimvb]

There's really only one way to make sure: test the flats with your images.

I think you will get more repeatable results if you point your scope at a (white) wall with indoor lights. LED lights if possible. Incandescent lights can give a too warm colour. Otoh, if you get good results with your flats, stick with your method.

[Anthony1979]

Probably have another practice at it tomorrow