calibration frames with asi1600mm cooled

[tooth_dr]

After reading this thread it’s putting me off the ASI1600! Seems like a lot of faffing about that I mightn’t be able to master?!

[andrew s]

10 minutes ago, tooth_dr said:

After reading this thread it’s putting me off the ASI1600! Seems like a lot of faffing about that I mightn’t be able to master?!

Please don't let this debate put you off. It is no harder to use than any other CCD or CMOS camera. 

The conclusion to Christian Buil study was "In summary, all here considered products are great. The optoelectronic performances of the Atik Horizon and ZWO ASI1600MM/PRO cameras are quite similar, and it is hard to select one model or another on this basis." 

They are being used to produce high quality spectra (by Buil and others) and I now have one in a spectrograph I am commissioning - tests in progress. It is just as simple to use as my SX and FLI cameras.

Regards Andrew

[Wirral man]

wow this is way over my head i wish i just got the 1600mc instead now l!!

just to confirm can i setup at any given time and take darks etc..as long as cool the camera to the same as i did my lights and same exposure?

if so i will do them today as its raining here..

[brrttpaul]

4 hours ago, kens said:

Shoot all calibration frames at the same gain, offset and temperature as your lights

Dark flats (or more correctly flat darks) are darks that calibrate your flats. So same gain, offset, temperature and exposure as your flats

Hi Ken, I always have shot my darks, bias etc at the same settings its just that I read a previous poster saying he shoots his darks with zero gain, (unless i misunderstood it) maybe perhaps he shoots his lights at zero gain also , and i misread it

[brrttpaul]

1 minute ago, Wirral man said:

wow this is way over my head i wish i just got the 1600mc instead now l!!

just to confirm can i setup at any given time and take darks etc..as long as cool the camera to the same as i did my lights and same exposure?

if so i will do them today as its raining here..

All I done was set my temp to -25 and took a load of darks at various exposure times and made a master out of each one, eg master dark 1 min master dark 2 min etc

[andrew s]

I would keep gain the same for all frames. Otherwise the results will not be consistent.

Regards Andrew

[Wirral man]

is there any way via a app or programme for my windows 10 laptop to make the screen white for flats?

[geordie85]

1 hour ago, Wirral man said:

wow this is way over my head i wish i just got the 1600mc instead now l!!

just to confirm can i setup at any given time and take darks etc..as long as cool the camera to the same as i did my lights and same exposure?

if so i will do them today as its raining here..

Yes but I'd suggest doing them at Night unless you already have a very dark room to do them so you don't get any light leakage.

1 hour ago, Wirral man said:

is there any way via a app or programme for my windows 10 laptop to make the screen white for flats?

Can't you use notepad? 

That's what I used to do until I got a cheap tracing board from Amazon.

[vlaiv]

2 hours ago, andrew s said:

If this is the case then I think there might be something wrong with your camera.  I can't think Christian would have missed that as it would have shown up in his dark current measurements.

Regards Andrew

It might be my sample, but I've noticed similar thing with my ASI178mcc. Actually I have not been able to use bias on any of CMOS cameras I've been using so far (QHY5LIIc, ASI185mc, ASI178mcc, ASI1600mmc).

I remember QHY5LIIc having inconsistent bias between power cycles - two sets of bias files taken between when any of parameters of camera change and then you return to original parameter values cause camera to recalibrate bias internally and you get slightly different value. If you just continue shooting without changing anything you get consistent bias.

With ASI1600 I've tried on numerous occasions, either by examining mean values of darks, or by trying simple frame arithmetic like (1minute - bias) x2 == (2 minute - bias) and it does not work for me.

Actually anyone can try to see how their ASI1600 behaves - you need a set of bias, set of one minute darks and set of two minute darks (let's say couple dozen of each). If bias works properly following should hold:

Mean pixel value of bias stack < Mean pixel value of 1minute dark < Mean pixel value of 2minute dark

and

Mean of  2 x (1 minute stack) - bias stack - 2 minute stack = 0 (within noise error, so really small value roughly around 0.0005 or so).

For anyone concerned about calibration of ASI1600 - just use calibration scheme without bias - darks, flats and flat darks to calibrate lights and it calibrates out great (you can use this scheme even if there is no issue with bias and it works if there is issue with bias - in both cases it does proper calibration) - remember to shoot appropriate subs at same settings and you will be fine.

[wimvb]

1 hour ago, brrttpaul said:

Hi Ken, I always have shot my darks, bias etc at the same settings its just that I read a previous poster saying he shoots his darks with zero gain, (unless i misunderstood it) maybe perhaps he shoots his lights at zero gain also , and i misread it

Darks should be captured at the same settings as the light frames they are correcting. Dark flats (flat darks) at the same setting as the flats they are correcting. I take light frames at various gain settings and exposure time, depending on target, area in the sky, moon phase, etc. I always shoot flats (and dark flats) at zero gain, because otherwise the exposure time becomes too short. My capture software (ekos) determines ecposure time for flats to get average pixelvalues at 25000 adu. You can of course shoot flats at the same gain as your lights, if that works for you. But it's not necessary.

Again, flats correct the optical issues in your setup: vignetting and shadows. Bias and darks only correct the electronical issues of the camera. That's why you can't change anything in the light path when you shoot flats, and equally you can't change anything in the electrical "path" when you shoot darks or bias. But you can change the other way around. You can shoot flats at other camera settings, and you can shoot darks at orher optical settings, including detaching the camera from the scope. Actually, it is better to remove the camera from the scope when shooting darks, because it's easier to avoid light leakage this way.

As for shooting darks during day time; just nake sure your camera cap is truly opaque. Plastic caps are almost transparent to ir light. I always wrap the front end of my capped camera in aluminium foil and put it in a black box, before shooting darks. I only shoot darks at night. Have learned this the hard way.

[Wirral man]

7 minutes ago, geordie85 said:

Yes but I'd suggest doing them at Night unless you already have a very dark room to do them so you don't get any light leakage.

Can't you use notepad? 

That's what I used to do until I got a cheap tracing board from Amazon.

geordie when you said you use apt for flats do you just run it with the settings it shows or do you change them also found a web page called white page which seams perfect 

[wimvb]

2 hours ago, tooth_dr said:

After reading this thread it’s putting me off the ASI1600! Seems like a lot of faffing about that I mightn’t be able to master?!

 

1 hour ago, Wirral man said:

wow this is way over my head i wish i just got the 1600mc instead now l!!

just to confirm can i setup at any given time and take darks etc..as long as cool the camera to the same as i did my lights and same exposure?

if so i will do them today as its raining here..

No need. Cmos are still the best cameras in terms of real estate per buck (or €, £). The calibration process is the same, regardless of camera type. It's just that people have found that their conditions allow for variations. Eg @ollypenrice has found that he can replace darks with bias and bad pixel maps, and use just bias to calibrate flats. That works for his ccd based setup. If/when he gets into cmos imaging, he will most likely have to change his workflow, and rethink the calibration process. That's just "how the cookie crumbles", to quote Jim Carrey.

[vlaiv]

2 hours ago, wimvb said:

Dust directly stuck on the sensor. Where they completely block any light, no calibration frames can correct them.

Yes, you are right, but I guess dust particle should be between sensor and sensor cover glass in this case. I don't know how big dust particles are, but if they are bigger than single pixel then absolutely you can't correct that since there is effectively no light falling on pixels behind those. Though I haven't seen such case yet. I get small dust particles on sensor cover glass and they produce tiny doughnuts but I guess they are so small that they can effectively be calibrated out with flats.

[vlaiv]

2 hours ago, andrew s said:

If this is the case then I think there might be something wrong with your camera.  I can't think Christian would have missed that as it would have shown up in his dark current measurements.

Regards Andrew

I've just revisited Buil's page - his method of measuring dark current rate is via subtracting two matching darks and then measuring remaining noise and subtracting read noise from it (previously measured). He in fact is not calibrating his darks with bias, but rather working on the premise that dark current noise and dark current are related (which they are) by square root.

Using this approach he would have easily missed the fact that there seems to be bias offset that depends on exposure duration.

[wimvb]

1 minute ago, vlaiv said:

Yes, you are right, but I guess dust particle should be between sensor and sensor cover glass in this case.

In that case, it would be best to return the camera, since the manufacturer must have an issue in the production or quality control process. If a dust particle or other obstruction creates a shadow so large that the umbra, rather than the antumbra, falls on the chip, no calibration can remove it. This is simply because light is completely blocked in the umbra, but not the antumbra. If the light frame and flat frame are ideally dark calibrated, and you divide the light signal by the flat signal, in the umbra you have 0/0, but in the antumbra you don't.

There was a thread not so long ago about just this. The op had a dust particle on the sensor that could not be calibrated out. He was recommended to clean the sensor.

[geordie85]

1 hour ago, Wirral man said:

geordie when you said you use apt for flats do you just run it with the settings it shows or do you change them also found a web page called white page which seams perfect 

I open flats aid, tell it I want an ADU of around 25000, change the shortest exposure time to 0.01, set the first exposure to 1.0 and click run. It will then take a series of images until it finds the best exposure time for my desired ADU. 

Once it finds it I click run and it'll run my desired amount of images.

Once it's finished, I cap my scope and cover it all with something dark and run the same program again for my dark flats.

[andrew s]

Hi Vlaiv - 

2 hours ago, vlaiv said:

Actually anyone can try to see how their ASI1600 behaves

Just what I have been doing. The gain was set at 200, temp -18 throughout. These are for a subsection of the chip 600 - 1000 on y axis 0 4652 x axis as this is what I use in my spectrograph

I did two sets of 8 Bias/Offset frames with a power cycle between them and waiting until the temperature had stabalised.

Bias        Ave           STD

A            21.354       8.668

B            21.353       8.683

I then did 8 each of 60, 300 and 600s Darks with the following result.

While not totally linear (I would put this down to there being a number of dead, hot and warm pixels in the sample) it does seem to be as expected. 

Regards Andrew

[andrew s]

For fun and learning to use Excel I have plotted the data and added error bars based on just the Bias value. This show the trend line is linear to within this tolerance.

The pixel to pixel variation is not Gaussian and show both hot and cold pixels as well as 1/f noise common to CMOS.

Regards Andrew

[vlaiv]

19 minutes ago, andrew s said:

Hi Vlaiv - 

Just what I have been doing. The gain was set at 200, temp -18 throughout. These are for a subsection of the chip 600 - 1000 on y axis 0 4652 x axis as this is what I use in my spectrograph

I did two sets of 8 Bias/Offset frames with a power cycle between them and waiting until the temperature had stabalised.

Bias        Ave           STD

A            21.354       8.668

B            21.353       8.683

I then did 8 each of 60, 300 and 600s Darks with the following result.

While not totally linear (I would put this down to there being a number of dead, hot and warm pixels in the sample) it does seem to be as expected. 

Regards Andrew

While I believe your findings are not inconsistent with mine - bias is not "true bias" if there is no linear dependence of dark current to time (that level of non linearity is a bit suspicious to be explained by dead / hot-warm pixels, especially since I don't have a single dead pixel in mine sample), I will redo measurements on my sample and post it here as soon as I find time for it (probably tonight).

I've just checked my latest master flat of 1 minute exposure - mean ADU value (at unity gain) is ~46.825 at offset value of 50. I don't have any bias files but I do remember that they are never quite exact offset value for my sample - usually around 1-2 ADU below, so I would expect bias to be around 48 which is greater than 46.8 (and it should not be).

Not to speculate further, I'll post my results later so we can compare.

 

[vlaiv]

8 minutes ago, andrew s said:

For fun and learning to use Excel I have plotted the data and added error bars based on just the Bias value. This show the trend line is linear to within this tolerance.

The pixel to pixel variation is not Gaussian and show both hot and cold pixels as well as 1/f noise common to CMOS.

Regards Andrew

This kind of negates my previous statement of graph being non linear, but those error bars seem way too high?

Mind you, read noise is somewhere around 1.7e (for unity gain, you are using slightly higher value) - but you are averaging 8x400x4652 values - so I would expect error bars to be around 4.4e-4 not the one you used - 8.6. Well now that I come to think about it why did you use that value in the first place? It really tells nothing if you don't subtract bias level from bias frames, just setting bias to different values in settings will change it.

No, what I meant to say is that 8.6 stdev is standard deviation of both bias signal and bias noise, shifting by constant offset will not change it, but removing bias signal will.

[andrew s]

4 minutes ago, vlaiv said:

bias is not "true bias" if there is no linear dependence of dark current to time

See my later post where I think it show it is reasonably linear ( I see you have). I agree that a zero length exposure is only an approximate estimate of the electrical offset but the ASI1600MM does not have overscan so it is the best estimate I have.

4 minutes ago, vlaiv said:

Well now that I come to think about it why did you use that value in the first place?

I used the value reported by Maxim DL 4 from its information box. The reason it is not the low figure you estimate is that there are a range of pixel values with a Max of 624 down to a minimum of 16 (note my gain is 200). Averaging the 8 frames will reduce the read noise but not 1/f or noise due to warm/hot pixels. 

The reason I think I have at least one dead pixel is that the minimum value stays at 16 for all the frames I took. (This may be the best estimate of the Bias!)

Here is the histogram of a Bias frame

Regards Andrew

[andrew s]

One final graph. I have added error of one SD of error to the dark average of the frames assuming Gaussian noise and reduced by sqrt(8) for the 8 stacked frames. Given the large bias of the hot pixels, which is not reduced by averaging the frames, I am satisfied the performance is linear. 

(I did the calculation in electron and the converted back to ADU using the figure from the Buil document.)

Regards Andrew

[wimvb]

3 hours ago, wimvb said:

There was a thread not so long ago about just this. The op had a dust particle on the sensor that could not be calibrated out. He was recommended to clean the sensor.

This is the thread I was thinking of:

 

[Wirral man]

2 hours ago, geordie85 said:

I open flats aid, tell it I want an ADU of around 25000, change the shortest exposure time to 0.01, set the first exposure to 1.0 and click run. It will then take a series of images until it finds the best exposure time for my desired ADU. 

Once it finds it I click run and it'll run my desired amount of images.

Once it's finished, I cap my scope and cover it all with something dark and run the same program again for my dark flats.

thanks really helpful and simplified 

[ollypenrice]

4 hours ago, wimvb said:

 

No need. Cmos are still the best cameras in terms of real estate per buck (or €, £). The calibration process is the same, regardless of camera type. It's just that people have found that their conditions allow for variations. Eg @ollypenrice has found that he can replace darks with bias and bad pixel maps, and use just bias to calibrate flats. That works for his ccd based setup. If/when he gets into cmos imaging, he will most likely have to change his workflow, and rethink the calibration process. That's just "how the cookie crumbles", to quote Jim Carrey.

What, change, at my age????

OMG.

Horrors.

(Actually I hope to have a CMOS camera coming soon....)

?lly

By the way, my CCD workflow isn't original, it's widely discussed on the net and used by many. It's entirely supported by AstroArt and explained in their instructions.