ADU What does it really mean?

[PKay (Peter)]

Hi Clever Ones.

I have tried and tried to understand this concept...and failed.

1: I know (or think) that ADU is an acronym for Analogue Digital Unit.

2: We can calculate the ADU value for a camera (using Gain, Bias, Read Noise).
Hear at IIS this value specific to a camera is called the 'Target-ADU'.

3: We can use a test image and find out the average 'Background Sky Level in ADU per second (Say: using the statistics tool in Pixinsight).

So I don't have to type all that again call it BSL-ADU

4: We then have to chose a sub exposure length that allows the BSL-ADU to be greater than the Target-ADU

OK, I still don't get it
Can some one explain what the above all means in plain English?
If there has to maths involved, well so be it.
I am just after the conceptual understanding, the maths can come later.

[ChrisV (Chris)]

Okay here’s a “maths free” go from a fellow AP newb and not necessarily clever – but I work with digitizing biological signals, and the concept aren’t too disimilar. And if this all sounds like nonsensical crap then Craig Stark, Starizona have some great intros to it all.

A camera takes in photos – these are quantal in nature but they sum over your exposure to give an analog signal. So the exposure could have from 0 (well not this low really) to 20,000 electrons in a pixel (this is well size for your asi1600mc per pixel). So its an analog signal = the number of electrons in each pixel.

Your computer has to digitize it to work in the binary world. It does this with an analog to digital converter. This will have a certain gain, eg unity gain converts one electron into one bit (ADU) plus an offset. So the camera will have a certain range of values. Many cameras seem to have AD converters which convert the analog signal into 12 or 14 bits. So the computer can convert your ‘analog’ electron signal into a computer value between 0 – 2^12 = 0 -4,096 ADUs for your asi1600mc (this actually then gets converted to 16bit so 0 -65,536 ADUs).

You work your camera (gain/offset/exposure time) settings so that you stay in that range. If its goes too low (0 ADUs) you’ll clip and lose low level data (this is bad as its where a lot of our dim targets sit). If it goes too high you saturate (16,384 ADUs, or usually 65,536 as most convert to 16 bits in the end). Looking at your background sky ADU level lets you check this.

Adding to this all signals have noise associated with them (target and background, equal to the SQRT(signal level)), plus the camera introduces noise (read/dark noise etc). So, your measured sky background (or target) will have a certain noise. What this does is spread the signal around its mean level. So in effect you don't want to go “0 + noise” or “max – noise”. This can all be calculated – there’s some great info on IIS about this by Shiraz etc.

If you look at a histogram of you sky background it will have
- a mean level = background photons (plus camera dark current) – converted into ADUs by the gain and offset
- variance around that determined by sky background noise, read noise, dark current noise

You don’t want the lower tail of the histogram near zero. Equally if you go to high then more stars and possibly your target will start to saturate.

[PKay (Peter)]

thanks mate, will have to think about this for a few days...

[glend (Glen)]

I recommend reading Ray's (Shiraz's) posts on ADU.