QSI Arrives

[Paul Haese]

Well I finally have the QSI in my possession. Interesting looking unit and quite heavy for its apparent size.

I was a bit distressed to learn that I had to install the LRGB filters myself. I had thought that I was going to receive the unit with them installed. QSI does this in a clean room if you want it done. I did ask for it to be done but something got lost in translation.

I had a bit of trouble working out how to do darks in Maxim, but I have worked this out now. There are quite a few tutorials I came across that helped.

Connecting the unit and installing software was a breeze. Every thing is self explanatory in that regard. It is just Maxim that is not very intuitive.

Any tips on this steep learning curve would be much appreciated. There is going to be quite a lot to learn.

[toryglen-boy]

looks excellent Paul, cant wait to see your first images with it


:thumbsup:

[Paul Haese]

No doubt the first images are going to look pretty ordinary. This is just a whole new way of thinking. I am sort of accustomed to it with using RGB on my planetary stuff, but this looks slightly more complicated.

I will work it out in the end though. I want to make the next step in imaging.

[renormalised]

Nice looking unit there, Paul. I hope it lives upto your expectations and you get some cracker piccies from it:D:D

[gbeal]

Green with envy, like most others no doubt, congrats Paul.
To my way of thinking it won't be all that different to the RGB planetary.
Use Maxim (I presume you will??) and simply shoot the three colours then use Maxim to process.
Baby steps for a while, but knowing you it will be a piece of cake.
Gary

[Bassnut]

Nice piece of kit right there, you will eventually get CCD stack, its a given, and it looks like its the guide-before-the-filter model, which is a smart purchase. Your headed for some awesome pics there Paul, RGB aint that hard, youll laugh in a few weeks. Now you can indulge in 3nm Ha filters and the like, and really cream it.

[gregbradley]

Great looking camera.

I recommend you standardise your image lengths. 10 minutes is a good length.

Also the temperature. In winter you will be able to get a camera down to -25C without too much trouble. In summer it may only get to -15C. Not sure how much your camera will cool.

So take 12 darks at -25C 10minutes each 1x1 binning. Make a master dark in CCDstack and use sigma reject for combining them.

Now take 8 bias frames and make a master out of them.

Now do the above again but for 2x2 binning. That is for your RGB exposures where most people use 2x2 binning.

When you are ready to image have your setup focused and camera in the position you will image. Now take some dusk flats. Put a white T-shirt over the end of your scope, point the scope 10 degrees above the horizon towards the west where its evenly illuminated. When its twilight take 3 shots each of at least 3 seconds long and so you get a reading of around 21,000 from the image (about 1/3rd saturated).

Do this rapidly for each LRGB. So Luminance is usually imaged at 1x1 binning so do those flats at 1x1. Do the RGB flats individually and the camera set to 2x2.

Make masters out of your flats and subtract the bias you made above from it. You don't need the bias again after that.

When you are ready to start the image at night, do 10 minute subexposures at 1x1 for luminance and 10 minutes at 2x2 for RGB.

A common exposure time is 6x10 for luminance, 3 x 10 each for RGB.

Then you process these in CCDstack. callibrate by dark subtract and flats. Then data reject removing hot/cold pixels then register to align and then save as a master file for that LRG or B.

Then when you have masters for each of the LRGB you do a colour combine.

Then you save and switch to Photoshop where the first step is to use levels/curves to bring the image up.

Sounds like a lot but its not really.

Greg.

[Peter Ward]

Paul....you only need darks, and flats.

A bias frame = a zero second dark exposure.

Hence *any* non zero dark has the bias there already....sigh...

I'm sure you will enjoy the new toy.

[Paul Haese]

Thanks so much Fred, Greg and Peter for the advice. It will give me a good start. Much appreciated.

Gary, I am not as certain as you of the speed at which I will pick this up.

[strongmanmike]

He's gettin the kit alright :thumbsup:

Exciting times Paul

Mike

[gregbradley]

You should mind your manners Peter - no need to be rude.

To subtract a bias from your flat is standard procedure in CCDstack and in Wodaski's Zone book. If you feel otherwise you may well be right but
you could express that politely.

Greg.

[strongmanmike]

I have to agree with Peter here and I didn't take his comment as rude Greg :shrug:.. slightly obtuse perhaps :P but not rude :)

[DavidU]

Nice ! I had a look at the site, great cam's for sure
What model did you get? I can't read it off the photo

[Peter Ward]

Easy does it Greg...you were not the subject of my post.

But as you've mentioned it, Ron is by no means the source of CCD gospel. There are a number of areas where I, and *many* others, do not have the same world view as Ron.

Subtracting a bias from a flat makes no sense unless you are operating at cryogenic temperatures...i.e. not taking darks.

For the rest of us darks are a necessary evil, and by defintion contain bias information, which is automatically taken out of a flat during dark frame subtraction.

For Paul: Mate, it's easy to calibrate images. All you need is a good library of matching dark frames (ie exposure and temp the same as the data), and flat frames for each filter....but I think you already knew that :)

[gregbradley]

I posted the question on another group and Stan Moore the author of CCDstack answered. Here is his answer about whether to subtract a bias from your flats when making a master flat or not.

One thing to keep in mind is that flats are divided and darks are subtracted so it is a different mathematical operation.

I must say though in my experience as I have done flats without bias the difference is not great but there is a subtle improvement when a bias is correctly done and subtracted from the flats before using them.

"
For the flat divide to work properly, the camera's bias (or pedestal) must be
removed (subtracted) from the flat prior to flat fielding (normalization and
division), otherwise the flat fielding will under-correct (e.g. vignetting or
dust donuts will not be completely removed). There are 3 ways to remove the
bias:

1) Dark subtract the flat as if it was a normal exp (in fact it is).

or

2) Subtract a Bias from the flat (instead of a dark). Flat exps are commonly
very short and there is not enough time for significant dark current to
accumulate (e.g. there is very little difference between a 2 sec dark and a
bias). What little difference there may be (between dark and bias) is totally
insignificant due to the flat's high intensity, which completely buries any dark
current. This is very convenient because you can create a high quality bias
(e.g. made from 30+ frames) once and use it repeatedly, regardless of flat temp
and exp time.

or

3) Subtract a constant (equal the avg bias level) from the flat. This usually
works just fine unless the cameras has a very strong bias pattern and the flat
is very weak.

Stan
http://www.stanmooreastro.com/ (http://www.stanmooreastro.com/)"

I suspect the above is more important if you have a lot of dust donuts in your image or if your camera is pretty noisy or if your cameras' bias pattern drifts around and is not stable.

Greg.

[rally]

Thanks for the informative discussion Greg.

[Dennis]

Hey Greg

After seeing references to CCDStack in your posts and from a few of the other imagers on Ice In Space, I just recently purchased a copy of CCDStack (30 day free trial) and most recently, the Adam Block DVD tutorials (separate purchase) which arrived yesterday and I’m just munching my way through the learning curve.:)

I really like CCDStack as it seems tailored specifically to what I need to do with my ST7 CCD images. Previously, I had to use functionality from CCDSoft, MIRA AP, Maxim DL and AIP4Win and now, I can do all of this is a single, tailored environment!:thumbsup:

Cheers

Dennis

[gregbradley]

CCDstack is a great program although it does not handle memory too well and larger file sizes cause disruptions. But apart from that its great.

Here is another response to this question of bias from Richard Crisp who is a CCD electronics engineer and been the head of various groups about CCD research and is super knowledgeable on the subject:

"You have to remove the offset from both the flats and the image in order for the flat fielding operation to function correctly

Usually flats are of short duration so there is not much on the way of dark fixed pattern noise, so biases are as good as darks for creating your flats

Image exposures are usually longer than flats so the tend to have more dark spikes (dark fixed pattern noise) so darks are appropriate for calibration

If you don't remove the offset from your darks by subtracting bias, then subtracting the dark will remove the offset from the image

If you keep a library of bias subtracted darks then you need to subtract biases when calibrating with such darks"
Richard.

Greg

[Peter Ward]

Gosh, darn it, looks like Stan and I agree :)

I use the KIS principle....(keep it simple). While Stan does mention using just bias with ultra short flats, I'd also point out you will get an iris shadow pattern from the resulting uneven exposure seen in some shutters..eg Melles Griot... which can cause even more grief as the shadow variation, varies with the exposure time (Hey Bert, this would be a challenging differential equation ! )

I've seen this many times with these shutters and the only fix is to take long exposure flats which by definition will require method (1)

[Peter Ward]

Greg, sure, Richard knows his onions, but this is only true with equi-illumination shutters....as I mentioned in my earlier post.....you *will* get uneven shadowing with any other type of shutter....and while dark noise goes down with a decrease in flat field exposure, shadowing increases, totally negating the point of taking a flat frame.

[rat156]

Now, I do take bias frames and subtract them from my darks. I do this so that the darks are scaleable.

My theory is that I only need one darkframe for each bin level, this is what CCDStack is telling me. So I have one 30minute darkframe for 1x1 and 2x2 at -15C. I can go down further in temperature in winter, by not in summer, the darks subtract out nicely so I like to keep one temperature. Apparently when the darks are made in this manner they are scaleable for both time and temperature.

I assume that what CCStack is doing is subtracting the bias from the dark so that what you're left with is "pure" dark noise. As the bias doesn't change in intensity over time, but the dark does, it scales the dark then adds the bias to the dark before subtracting. Put another way the scaled dark has lost the bias information (as it has been subtracted in the master making process), so it has to be subtracted as well as the dar from the light.

Does this make any sense?

Oh BTW Paul, welcome to deep sky imaging with a dedicated CCD. I look forward to your results, the camera looks like a cracker!

Cheers
Stuart

[Peter Ward]

Yes, you can do this, but a hack like me doesn't because.

The dark noise pattern changes slowly over time
Dark noise is not completely linear
The undelying bias pattern can be (and is often) slightly random

In effect my dark library (ie Bias + dark) is constantly updated to provide the cleanest calibrated images possible.

It also doesn't work with back illuminated chips that have spectrally dependent fringing patterns (translation: flats change with colour! arrgh!)

[rat156]

Lucky I don't use one of them fancy back illuminated chips...

Cheers
Stuart

[Peter Ward]

:lol:

It was such a pain in the proverbial, Random bias with a low level ripple, and the fringing was so bad even a 5 degree camera rotation required a new set of median *sky* flats (dome flats proved useless) for each filter.

So.. I don't anymore either. It went back to the manufacturer!

[Octane]

Peter,

Can you explain what advantage the sky flats had over the dome flats? I'm assuming by dome flats you mean projecting a light source onto the roof of your observatory dome and imaging that?

Would you suggest a pauper with a DSLR (like myself) to use sky flats over artificially created ones?

Regards,
Humayun

[Peter Ward]

There is nothing wrong with DSLR's. They are a great tool for wide field work.

Dome flats/T-shirt flats/sky flats are variations on the theme of providing a perfectly blank field that can be used to calibrate the camera's images.

Sky flats are easy. They have to be taken just before dawn or after sunset, adjusting the exposure to give about 30% well saturation of the sensor. (you want a high S/N but don't want ABG to kick in). Take several frames near the zenith, moving the scope around between each frame to minimize any sky gradient. Median combining them will remove any random signal.

The problem you will have with a DSLR is sky flats will not be monochrome. (Easily fixed in Photoshop!), but if done correctly you will have a nice map of the optical vignetting of your telescope (or lens) and any dust doughnuts.

Hope that helps.

[RobF]

Thanks for showing the new baby in all its glory Paul. I'll be hanging out like the others to see what delights you show us in months to come.
(I'll just go find something to wipe this drool off my screen now....)

[AlexN]

Lots and lots of great information in this thread... Even if it has wandered slightly off the original topic...

No more bias frames for me!!

Its funny.. Once Peter said "any non-zero dark will contain the bias" it seemed painfully obvious... but prior to my reading that, I would never have thought about it... haha..

Peter - While we are on the subject of calibration.. What are your thoughts on the amount of darks/flats to be used? Do you use a proportionate amount of calibration frames to your lights? or go with the old "more is better" approach? I've been using a 1:2 ratio of Darks:Lights.. 6 darks to 12 lights and results have been good.. however there is still some noise (which may be lack of decent exposure time more than anything else..)

Thanks to everyone for their input into this thread.. Its been quite a good read..

Alex.

[Alchemy]

looks like an interesting camera, how many megapixels, what size chip, and/pixel size.

will be waiting to see some results, nothing like some new gear to spark the interest to the next level.

[AlexN]

The QSI583 that Paul bought is a KAF-8300 chip, same physical chip size to a QHY8, 8.3mp, and 5.4um Pixel size (from memory) Same chip as used in the FLI ML8300, QHY9 etc..

I can not wait to see what Paul can make this camera do!!

[Alchemy]

ta alex,

quick question, given its the same chip in all those cameras does the electronics give better performance in any of the models ? if so which ?

[AlexN]

The answer to that depends greatly on who you ask... Electronics wise, they are much the same... Cooling capacity and download speed is where you see the FLI cameras taking the lead... the QSI and QHY are much the same in that respect.. The QSI is initially more expensive, however with its integrated filter wheel, and only requiring 1.25" filters you do save a fair bit of the difference in the long run.. The QHY9 still comes out cheaper I think... The QSI also has an integrated OAG that guides infront of the filters... This is definitely nice... however not for everyone.. I'd go the ML8300 if I had the money and inkling to go to a mono sensor again...

[Paul Haese]

Clive,

just going to confirm that the camera is 8.3 mp and 5.4 um for pixel size. QSI electronics are very high grade. QSI has just entered the astronomy field but is best known for its imaging equipment for science. I would imagine all the top of the line camera's have very similar electronics and standards. Proof will be in the pudding though.

[gregbradley]

Not entirely true Paul. Different cameras have different quality electronics.

For example Mikes FLI PL11002 is the same chip as an STL11.
Mike posted one of his darks and I posted same temp and exposure from an STL11000M I had. The STL was way higher noise. Same chip. Difference is the read noise. The SBIG was way higher. I was surprised and I think Mike was too about how clean his FLI PL11002 is.

Perhaps the new STX will close the gap that has occurred between SBIG and Apogee and FLI and perhaps QSI (I don't know much about QSI but it looks quality). Perhaps not if they use the old electronics in a new body. I'd like to see a dark from an STX16803 before I would consider one. New guiding methods are appealling but at the end of the day you want low noise more than anything and reduced cosmetic defects which high cooling also does and the KAF series sensors tend to have (vertical lines from hot pixels and the reading process which fade with cooler temps for example).

Same between Apogee and FLI. The FLI has lower read noise of about 5 electrons compared to the Apogee of about 9 electrons.

So when you compare cooling and noise you have to factor in the read noise which varies from manufacturer to manufacturer.

Greg.

[Peter Ward]

I disagree. (surprise surprise :) )

I've probably tested 2-3 dozen KAF11000 series sensors (no prizes as to which manufacturer) and have seen significant variation in the noise from chip to chip....ie all chips are *not* the same.

I have noticed Kodak seem to be continually improving the noise performance of their sensors, eg. 2009 sensors look much cleaner than those manufactured in 2006.

However, the only accurate way of testing "off chip" noise variations across manufacturers would be to place the same chip into differing heads, then measure the result.

[AlexN]

Far too much effort - Not to mention voiding the warranty on two or three $10K+ CCDs. Better option is to take good calibration frames within 1°c of the light frames, then noise is non-existent anyway.

I know with both my ST-10XEi and ST-9E, noise in a 5 minute frame was considerably high.. however after applying a few dark frames taken within 1°c of the light frames, noise disappeared...

Everyone likes to stand up and argue the positives of their CCD over someone elses, this is all good and fair, who wouldnt want to defend the reason they bought a $10k AUD camera? At the end of the day.. Marcus, Tom Davis and Peter both take sensational images with their STL-11000M's. Mike S. Takes sensational images with his FLI PL-110002M, Greg takes sensational images with his FLI ML-8300 and Apogee U16M, Fred and Steven take sensational shots with their ST-10XME's, Ejcruz (Eddie) takes sensational shots with his QHY9, and many users (Clive, Doug, Louie etc.) take sensational images with their QHY8's

At the end of the day. CCD brand, and chip (both model and individual unit) make little difference. Its more about experience both in capture, calibration and processing. Good mount and optics always play a big part too...

Lets not turn this into a "Who's is better (or dare I say, Bigger) than who's... Rather, lets all kick back and rejoyce in the fact that there are many fantastic options available today for CCD imaging ranging in price from a meager $1800Au right the way through to whatever amount of money you're willling to spend..

Some have this option, some have that option. at the end of the day, as stated above, in the right hands, any good CCD can take great images in the right hands.

Paul's QSI will give great results, as he's made sure he has a good mount, good optics, and a great understanding of capture, guiding, processing behind him..

Thats enough from me, But I hope for once, everyone can agree on something...

Alex.

[Peter Ward]

Agreed Alex, and it was not my intention to get into a bigger/better comp...there are indeed many good choices available, but I'd only add they be made on a well informed basis.

You pays your money and takes your chances :)

[MrB]

Not that I have any experience with the CCD cam's, but I do have an electronic background and would expect that there would indeed be differences between camera brands using the same Kodak sensors, and the quality of their support electronics.
I suspect the quality of the ADC(Analog to Digital converter (http://en.wikipedia.org/wiki/Analog-to-digital_converter)), or anything else in the analogue signal path(like amplifiers etc) would be just as improtant as that of the CCD itself.
As with CCDs, there are different manufacturers of ADCs and different versions of ADC from each manufacturer, some are better than others, some are extremely expensive, some reasonably cheap.
The design of the support circuitry for the ADC would also have an effect on image quality... the accuracy and stability of the ADCs reference voltage for example.

Links in a chain.

[gregbradley]

At the end of the day. CCD brand, and chip (both model and individual unit) make little difference. Its more about experience both in capture, calibration and processing. Good mount and optics always play a big part too...

Alex.[/QUOTE]

I agree that is basically true. The differences between high end gear definitely get narrower and narrower. Like APO refractors. The last little bit of improvement comes at great cost. However when you are paying big dollars the little differences become bigger issues for deciding and they do tend to be little differences. You would be hard pressed to notice anything different between the same imager using an Apogee U8300, a FLI ML8300, A QSI583, a QHY9 all using the Kodak 8300 chip on the same scope and same exposure times.

Some cameras make the job a little easier though. I know imaging last weekend at -40C didn't require darks or flats really. The more noise the more you rely on accurate callibration and if that is even a bit off you have to use Photoshop etc. to try to fix the image whihc can be difficult.

There is tremendous choice now than there was when I started about 4 years ago. Now we have FLI, Apogee, QSI, Starlight Express, QHY and Sbig. The Yankee Robotics Trifid cameras seemed to drop out of the game for some reason.

Greg.

[Octane]

Peter,

Woops! I should have mentioned that I take flats, already. When using lenses I would use my LCD screen set to white in a fullscreen Photoshop document and use that for my flats. For a scope, I've built my own lightbox. Another option I've used is to peg a piece of foam core board about 10cm from the front of the refractor and taken flats in the middle of the day, ensuring my histogram peaks between 1/3 and 1/2 away across.

I was really just wondering as to whether taking sky flats had any obvious advantages over the, dare I say it, synthetic methods.

From what you've mentioned, though, it seems as though either way is fine. :thumbsup:

Re: bias frames. I've never used them in DSLR work. I just don't see the point considering the bias is included in the darks for the lights as well as the darks for the flat lights. In my experience, they've just introduced further noise into the final image.

Getting back on track, I've just looked up the specs and details on the QSI and it looks to be a fantastic system. Kudos to you, Paul. Now, looking forward to what that coupled with the RC can produce.

Thanks.

Regards,
Humayun

[mithrandir]

They are similar but different:

QHY8
Sony ICX-453
Total pixel : 3110 x 2030
Active pixels: 3032 x 2016
Pixel Size: 7.8um x 7.8um square
Effective sensor area: 23.65mm x 15.72mm (28.4 mm diagonal)

QSI583
Kodak KAF-8300
Total pixel : 3348x2574
Active pixels: 3326x2504
Pixel Size: 5.4µm x 5.4µm
Effective sensor area: 17.96mm x 13.52mm (22.48 diagonal)

[AlexN]

I mentioned that the QSI was the same as the QHY9 not QHY8.
The QHY9 uses the same KAF-8300 sensor.

[AlexN]

oh, and effective chip area and chip physical size are not the same measurement... effective chip area is the physical chip minus the are occupied by the pixels used for the antiblooming gate from what I can discern... They are both APS-C sized sensors...

[gregbradley]

Re: bias frames. I've never used them in DSLR work. I just don't see the point considering the bias is included in the darks for the lights as well as the darks for the flat lights. In my experience, they've just introduced further noise into the final image.

Humayan


Do you subtract them from your flat when you are making a master flat or have you subtracted them from each individual light exposure?

The point here is subtracting them from your flats not your light exposures. If you subtract a dark or use ICNR when taking your flat then you are doing what IU am saying here which is to subtract bias when making a flat master not subtracting bias from a light exposure.
Flats are normalised and then divided into the image, darks are subtracted. So mathematically you can see that if you didn't remove the wrong numbers in the pixel grid caused by bias noise then some numbers that is divided into the image when applying a flat would be falsely too high or too low. The point of a flat is to bring about evenness of illumination of the light. This then would reduce its effectiveness. How much less effective? Perhaps not a lot. Probably depends on how noisy your bias is I suppose. Some cameras are cleaner than others.

However I agree subtracting a bias from an image can cause more noise and unwanted effects as you are subtracting the same thing twice if you subtracted a dark. The use referred to here is subtracting it from the flat images before making a master flat. If you subtracted a dark from the flats then you wouldn't subtract the bias (it'd be much the same anyway - a 3 second dark would look the same as a bias).

I think personally it is also a minor improvement and nothing important in the scheme of things. I have also noticed a slight improvement in noisy images by subtracting a pedestal (100 subtracted from the pixel count assumed to be bias type noise) instead of a bias. Again its subtle but when processing a galaxy image every little bit helps.

I prefer to keep my camera and filters clean and to run the camera as cold as possible to reduce noise so all this callibrating is less important in the fist place.

My TEC180 for example is very evenly illuminated in the first place. So is the AP140. Flats then are almost not needed.


Greg.

[Octane]

Hi Greg,

I calibrate each of my flat lights against a flat master dark (flat darks having been median combined and cosmetic defects noted), effectively removing the crud from each flat light. Only once the individual flats have been calibrated, do I median combine them into a master flat.

That has worked for me from day dot. I initially did use bias frames but got rid of them quick fast when they were leaving my images in artifactland.

So, in essence, no, I don't use bias frames to remove any bias from flat lights or darks (noting that the bias is included in the flat dark, just as it is in the dark for the lights).

Once I move to CCDland, it might be a different story. But, that won't be happening for many years as I need to exhaust what I currently have. :)

Regards,
Humayun

[gregbradley]

Then we are basically doing the same thing. I am not sure how a bias is done on a DSLR anyway - is it a really short exposure?

I once had CCDsoft allow me to do auto dark reduction on flats. It normally doesn't come up as an option and I haven't been able to fluke it since but that was handy.

Greg.

[rat156]

We seem to have strayed somewhat off topic, but I'll just push us alittle further away...

I do twilight flats, CCDSoft allows me to do auto darks for the lights, so I get calibrated flats.

This is when I come into problems. The exposure is so that the initial lightframe is about halfway to full, but once the darkframe is removed the maximum pixel count can be as low as 5000. Which exposure do I use, the initial or the dark subtracted.

Cheers
Stuart

[Octane]

Greg,

Yep, spot on. It's whatever short exposure the camera will allow. In some of the older cameras, this was 1/4000sec, nowadays, the higher end cameras allow you to do 1/8000sec. :)

Regards,
Humayun

[Peter Ward]

Initial.