Flats: Simple question - but I can't find an answer

[hughgilhespie]

I want to take some flats with my ASI1600MM Pro. I will be using a Flip-Flat on the scope. My simple question is - Do I need to have the camera cooled to the same temperature that the lights were taken at or can I just take the flats without having the cooling on?

I really have looked for an answer and haven't managed to find one!

Regards, Hugh

[wxsatuser]

Don't know the answer for sure but I do mine with cooling on.

[Owmuchonomy]

For flats I would have thought not.  For darks obviously but I await a more authoritative response.

[thomasv]

It may not be necessary to take the flats at the same temperature as the lights, but why not reduce the noise by taking them with the cooling on. As far as I know noise in any calibration frames will transfer to noise in the calibrated image, that's why people take many cal frames rather than just 1?

 

[Cozzy]

Hugh.

found this maybe of some use to you.

 

Uranium235 comments about halfway down the page, takes about temps and ADU.

 

Tim

[drjolo]

You may not need to, but for CMOS cameras it is good to calibrate flats with dark flats, then both must be done for the same temperature, gain and offset settings, then it is better to have it cooled at some regulated temperature. 

[david_taurus83]

Flats are used to remove optical imperfections from the light frames, like vignetting and dust etc. Technically, the flats don't need to be temp matched to the lights. However, with the 1600MM it is generally recommended not to try and use bias frames but to use a dark frame to subtract noise from the flat frame prior to using on the light frames!!

 

So the flat dark needs to match the exposure length AND temp of the flat frame. The flat darks are used on the flat frames to remove thermal noise the same way your long exposure darks are used to remove thermal noise from your light frames.

 

You still with me?!?

 

Is there a reason why you can't do everything at the same temp? I shoot lights, flats, darks and flat darks all at -25°C

[AbsolutelyN]

I made flats for my 1600 yesterday and thought best to keep everything the same. Does anyone know of a really good tutorial (for beginners) on how to make flat darks?

[hughgilhespie]

Hi,

+1 for a really good tutorial!

I am happy enough to keep the camera cooled when taking flats and definitely for flat darks. It can't hurt and is probably always a good thing to keep the conditions used for calibration frames as close as possible to those for the light frames. I was just surprised that I couldn't seem to find a definitive answer amongst the reams of information on the web. One factor seems to be that the newer CMOS cameras need to be treated somewhat differently from CCD cameras and much of the available information only applies to   CCD and DSLR cameras.

All a bit academic as I now realise that I have cocked up my calibration frames anyway.  I failed to match the camera gain and offset settings. So, attempt two coming up - all good practice!

Regards, Hugh

[david_taurus83]

1 hour ago, AbsolutelyN said:

I made flats for my 1600 yesterday and thought best to keep everything the same. Does anyone know of a really good tutorial (for beginners) on how to make flat darks?

Yes. Presumably your flat frames will be a short exposure, depending on your light source. So if they are around 1 or 2 seconds long, then you simply shoot another set at the same exposure and temp but with the lens cap on. And voila, you have a set of flats and matching flat darks! Then stack the flat darks to get a master flat dark, and then subtract that master flat dark from all your flats. Then stack the calibrated flats into one nice thermal noise free master flat!

[AbsolutelyN]

Brill thanks, a lot to learn! 

[Laurin Dave]

With my ASI1600 I've used matched flats and flat darks of about 0.5 second exposure (to get ADU of c 30000 on 16bit basis) and they've worked well, same gain setting as my lights.  Also I use matched darks and flats (ie no bias) to calibrate lights.  I've found I need to take the camera off the scope to take the flat darks and darks to mitigate light leaks which mess with the darks.  I have built a library of darks and flat darks to use with different exposure flats so that I don't mess my flats up by taking the camera off!

HTH

Dave

[tomato]

Apologies if this is a daft question, but how can you take a flat dark? By definition a flat needs a flat light source to evenly illuminate the sensor, and a dark is taken with light excluded from the sensor. So how can you take a flat dark? Does it mean you take a separate set of darks to match the exposure time of the flats, and use these to calibrate the flats? My flats are usually about 2 secs exposure so how much noise would the sensor pick up in that exposure time?

[tooth_dr]

2 hours ago, tomato said:

Apologies if this is a daft question, but how can you take a flat dark? By definition a flat needs a flat light source to evenly illuminate the sensor, and a dark is taken with light excluded from the sensor. So how can you take a flat dark? Does it mean you take a separate set of darks to match the exposure time of the flats, and use these to calibrate the flats? My flats are usually about 2 secs exposure so how much noise would the sensor pick up in that exposure time?

Yes exactly as you say, a flat dark, is a dark taken for calibrating the flat.  Lens cap on and same exposure time as the flat.

[tomato]

Thanks for the clarification, that makes sense, but do they have much impact on the calibration given the usual short exposure time for the flat (1-2 secs)?

[Stub Mandrel]

I find that unless my (DSLR) control frames are cooled I get background noise. it's worst for darks and bias, but I cool for flats too.

[vlaiv]

In principle you really don't need to do it, except, like pointed out by @drjolo , you want to keep your sensor temperature the same for flats and flat darks.

Flats are usually taken at very small exposure length, and dark current noise at those exposure lengths simply can't even match read noise, let alone be anything significant compared to signal (which you aim to be at least half histogram).

Same temperature between flats and flat darks is important, because you want to calibrate out dark current signal (noise is not important as it is very small compared to signal). If there is residual dark current signal (or bias - flat darks remove both) - your flats can under correct.

[hughgilhespie]

OK - the consensus seems to be better to cool than not cool but not absolutely necessary.

So, if you don't mind, a supplementary question. This is specific to the ASI1600 - but will also apply to the other CMOS cameras that allow the gain (e-/ADU) to be varied and have less than 16 bit Analogue to Digital converters.

The question is: What gain setting should I use for taking my ASI1600 flats?

The first answer that springs to mind is - Use the same settings that were used to take the light frames.

But - from my reading about flats in general, it is recommended to expose to achieve a certain number of ADU, usually in the range of about 20,000 to 32,000 ADU. I suspect that these figures all come from experience with CCD chips with 16 bit ADCs so they can be expressed as about 31% to 50% of the chip's Full Well Capacity.

Now for my ASI1600 I have a Full Well Capacity of about 20,000 electrons. So, I want to expose to get between 31% and 50% of the FWC filled with electrons, say for argument 8000 sites with electrons out of my 20,000 available sites. This will make sure I am nicely in the middle of the linear range of the chip.

Here's the problem. If I want to expose until I have got my 8000 electrons I need to be able read out the ADUs that correspond to 8000 electrons. But my ADC is only 12 bits so I need to set the gain so that I will have a reading less than 65535. Using the data from ZWO about the ASI1600 camera, these are the 'gain' figures quoted by ZWO in 0.1 decibel units and the corresponding ADU that is read out from the camera for 8000 photosites occupied.

ZWO 'Gain'    ADU at 8000e-

0                      25,600

50                    46545

80                    65360

These figures take into account that the output from the ADC (0 to 4095) is multiplied by 16 in the driver (I think) so that the reported ADC values are always scaled to 0 to 65535.

So, the simple problem is now. Do I set and use a very low 'gain' value  of say 50 and expose until I get around 46000 ADU  which means I have filled 8000 of my 20000 photosites or if I set the gain value to 200, which is the highest gain setting I am using for use with narrowband filters, and expose to say 32000 ADU - as is often recommended. If I take the second course I will only have 1000 electrons instead of the desirable 8000.

Does any of this matter? I guess that if the camera response (electrons vs exposure time) is linear at 1000 electrons collected then probably not. But it would be nice to know if this is right?

Regards,

Confused Hugh

 

 

[vlaiv]

ADU recommendation for flats stems from two things:

- linearity of sensor

- SNR for flats

First one is just saying that you should stay in linear region. ASI1600 is pretty linear, so your only concern should be clipping - you don't want any clipping "to the left" or any clipping "to the right" - meaning no anomalous values either cut off at small values or cut off at large values.

Second means that you want as much signal per flat to maximize SNR of flats in order to introduce minimal noise into calibrated sub.

I personally use unity gain for everything, both lights and flats, and I don't particularly pay attention to exact ADU value, but rather look at histogram. If histogram peak is somewhere around 75% and there is no clipping to the right (no bright clipping) - I'm happy with that. Depending on your capture software, ADU value will either be around 49000 (if values from camera are read out in 16bit format and software does not convert it to 12bit), or around 3000 in 12 bit format. Difference between being factor of exactly x16.

You can also use other gain values, but bare in mind that going with lower gain increases read noise (this is not important for flats, as flats have high SNR as is - provided that you get enough signal - peak to the right on histogram), while using higher gain does impact your SNR - you will collect less photons before reaching 75% - SNR will be lower.

[hughgilhespie]

@vlaiv,

Interesting! I hadn't considered the SNR aspect for taking flats. By my (dodgy) calculations, at unity gain (e-/ADU) an ADU value of 48,000 when the apparent ADU range is 0 to 64k would correspond to 3000 electrons which seems to be a respectable number.

So, am I right in saying that there is no need to match the gain used to take flats with the gain used to take the light frames? If that's right it does make life a lot easier. I would then do as you do and take all flats at say unity gain and just make sure that the light panel brightness produces exposure lengths that are (a) not too short, say at least 0.5 seconds and (b) gives a (scaled) ADU value of shall we say 32000 to 48000 (2000 to 3000 electrons)? 

Regards, Hugh

 

[vlaiv]

Yes, if for example you get 3000e per flat - SNR of your flat is somewhere around 50. Stack enough of them and you get very good SNR - error well below 1%.

Flats noise is added modulatory rather than additively, so SNR calculations are not as easy, but contribution over other noise sources is very low.

I personally use very short flat subs often as short as 50ms for Lum, and never had any issues (I did have issue with dodgy power connector so led panel was flickering - not visible by eye, but at such short subs it showed on flats). My flat panel is really strong.

[hughgilhespie]

Excellent! Thanks for that, it's always pleasing to find anything in astrophotography that makes life a bit easier, rather than even more complicated!

This is particularly helpful as the acquisition software I use, Sequence Generator Pro, has a very nice Flats Calibration Wizard but I could not see how to use it with variable camera gain settings. Now I can use it to do get my flat calibration exposures across the board using a single camera gain setting. Nice!

Regards, Hugh

[ollypenrice]

Just a thought: I don't know how the cooled CMOS behaves but, if using very short flats values, it might be an idea to build in a delay of 2 or 3 seconds between flats captures. The chip on my CCDs warm up quite quickly on fast repeated captures with an implication for the noise.

Olly

[emyliano2000]

Hi.

I just had a read through this and now I'm thinking that the exposure of my flats might be way too short. I'm using a qhy183m and for example the exposure time of the Luminance flats is 0.03612sec. I'm using APT's CCD flats aid to get close to 32000ADU.

I'm using my daughter's tablet with an A4 white paper in front of it and the flats are not correcting as they should.

Emil

[kirkster501]

I have spent much time tinkering and experimenting with this.  And it depends on the exposure time of the flats.

So in my experience (which amounts to little compared to some others on here, I readily admit), having the cooler on for your flats makes hardly any difference at all compared to it not being on.  Certainly, with any camera with an electronic shutter (Atik 460 etc), it will make no difference whatsoever to take your flats without the cooler being on because the exposure will be so short - as long as your panel is not blazingly bright.

The time when you may need to consider putting the cooler on is when you have a mechanical shutter (KAF sensor cameras, etc.).  In this case, in order to avoid a gradient in your flats, you need to keep your flat exposures at about about 2-5 seconds.  If your light panel is dim, and your flat exposure is getting on for 10 seconds or soemthing like that (which, with the Ha and narrowband filters it may be), then you need to consider having the cooler switched on.  There is a balance between getting the panel dim enough to get your exposure to about 2 seconds to avoid a gradient from the mechanical shutter.  If you can get the exposure time down, you can take flats at ambient and forget about the cooler.