Jump to content


Photo

Review of QHY10 camera

  • Please log in to reply
19 replies to this topic

#1 Charlie Hein

Charlie Hein

    Voyager 1

  • *****
  • Administrators
  • Posts: 12333
  • Joined: 02 Nov 2003
  • Loc: 26.06.08N, +80.23.08W

Posted 16 June 2012 - 05:44 AM

Review of QHY10 camera

By Craig Stark

#2 rmollise

rmollise

    Hubble

  • *****
  • Posts: 15704
  • Joined: 06 Jul 2007

Posted 16 June 2012 - 10:37 AM

Excellent review, and the camera sounds impressive...but...

For most of us a DSLR is just a better choice. A 60da is considerably less expensive, will return excellent Pretty Pictures, and is useable for terrestrial picture taking. CCDs? I think for folks going after the dimmest of the dim they are still a good choice. Ditto for those wanting to do science. Even then, I think the best choice is a monochrome camera, however. ;)

#3 Craig

Craig

    Vendor (Stark Labs)

  • *****
  • Vendors
  • Posts: 768
  • Joined: 16 Sep 2003
  • Loc: Irvine, CA

Posted 17 June 2012 - 09:59 PM

Rod,

I've got to do another CCD vs. DSLR comparison at some point, but when I've looked in the past, the CCDs have some real boosts to consider. When going mono, LRGB's L is a huge win for your SNR when you look at the same total time imaging.

OSC CCDs, of course are on the same footing there, so it comes down to bit depth, linearity, cooling, and package. For many, the package size won't matter but for the Hyperstar crowd it does. Bit depth - well, you're talking 14 vs 16 but really, given the noise stats and the well capacity, 14's not really hurting in a big way.

Cooling and linearity interact. The sensors are nicely linear but each time I've looked at the Canons, they do some real mucking with the image to get it so that the thermal signal doesn't boost the actual background - which makes dark subtracting a pain. It also makes estimating the true dark current a pain. In addition, you're talking noise on the dark current.

Personally, if I could figure out how to undo what Canon does to the image to put it back to being purely linear, I'd be a lot happier with them.

Craig

#4 Renae Gage

Renae Gage

    Apollo

  • *****
  • Posts: 1120
  • Joined: 13 Dec 2005

Posted 18 June 2012 - 11:47 AM

Great review. I'm not sure I understood the part about the odd/even line scans. Are you saying that it takes 2 minutes to get a 1 minute exposure because it is actually a composite of 2 images?

#5 rmollise

rmollise

    Hubble

  • *****
  • Posts: 15704
  • Joined: 06 Jul 2007

Posted 19 June 2012 - 09:36 AM

Hi Craig:

Another DSLR vs. CCD head to head would be great...

#6 Renae Gage

Renae Gage

    Apollo

  • *****
  • Posts: 1120
  • Joined: 13 Dec 2005

Posted 20 June 2012 - 01:12 AM

Hi Craig:

Another DSLR vs. CCD head to head would be great...


Well, ok...just this once. Send me the cameras.

#7 MichaelAK

MichaelAK

    Sputnik

  • -----
  • Posts: 27
  • Joined: 11 Jun 2012
  • Loc: Las Vegas, NV (suburb)

Posted 21 June 2012 - 05:02 PM

Craig,

Definitely an article I will be coming back to. Thanks.

Meanwhile, regarding:

Since these are not meant for planetary imaging, the fact that, for very short exposures, the odd and even lines come from different exposures some time apart is not a significant handicap.


1. I wasn't clear as to why these are not meant for planetary imaging ; and

2. What technology would be recommended for planetary imaging?


Michael

#8 Craig

Craig

    Vendor (Stark Labs)

  • *****
  • Vendors
  • Posts: 768
  • Joined: 16 Sep 2003
  • Loc: Irvine, CA

Posted 22 June 2012 - 12:29 AM

Great review. I'm not sure I understood the part about the odd/even line scans. Are you saying that it takes 2 minutes to get a 1 minute exposure because it is actually a composite of 2 images?


No. This only applies for very short exposures. Let's say it takes 1s to read out one set of lines. If you were to take a 1s exposure and read the odd and then the even lines, you'd have 1s of odd and 2s of even exposures. Lame eh?

Instead, it takes 1s and reads the odds and then 1s and reads the evens. So here, yes, it does double the exposure duration. But, let's now make this a minute exposure (and BTW, I'm just making up these times). Sure, one set of lines gets a bit longer exposure than the other, but it's now a small fraction of the total duration rather than 2x the total duration. So, for anything beyond a short threshold gets a normal exposure.

This is the basic issue though with interlaced chips. StarlightXpress has another trick to handle it (they dump the charge of the even lines during the exposure - just at the time delay you'd expect from a readout) to make these quasi-progressive. But, the upshot is that there are several solutions to the issue and that the only time you really run into the problem is when you're looking at short exposures.

Craig

#9 Craig

Craig

    Vendor (Stark Labs)

  • *****
  • Vendors
  • Posts: 768
  • Joined: 16 Sep 2003
  • Loc: Irvine, CA

Posted 22 June 2012 - 12:33 AM

1. I wasn't clear as to why these are not meant for planetary imaging ; and

2. What technology would be recommended for planetary imaging?


Planetary imaging exploits the "lucky seeing" principle. Every so often you're going to get a clear patch of sky to shoot through (or a clear patch of sky you partly shoot through). The trick is to find those patches. Since we can't predict them well, we grab a whole mess of frames and sort things out after the fact.

This means we need a lot of frames - a lot of frames before the target rotates, changes lighting, etc. So, we need high frame rates and want 10-100 frames per second. Big chips like these with 16-bit ADCs geared for low noise (faster readouts mean higher noise) run at fractions of a FPS. This is also why I tell people not to use Nebulosity for planetary work. The single 16-bit FITS files I use mean 1 FPS at max and that's just not how you're going to get killer lunar / planetary stuff (or at least you have a deuce of a time exploiting the lucky seeing idea).

There are fast streaming cams out there to fit every budget from basic webcams through the TIS cams and to the high end jobs. But, the chips are far smaller, the ADCs run at far higher clock rates, and as a result, they can stream a heck of a lot faster.

Craig

#10 Craig

Craig

    Vendor (Stark Labs)

  • *****
  • Vendors
  • Posts: 768
  • Joined: 16 Sep 2003
  • Loc: Irvine, CA

Posted 22 June 2012 - 12:35 AM

FWIW, I finally am making some progress into the image processing Canon is doing on the images before they hit the CR2 file. I've suspected they're doing both scaling and shifting of the data but the evidence was more indirect. Now, it's pretty clear. They're doing it for a solid reason and one that makes sense for the target audience of normal photography. But, it does make things a bit tougher for the astro crowd.

Craig

#11 freestar8n

freestar8n

    Soyuz

  • *****
  • Posts: 3988
  • Joined: 12 Oct 2007

Posted 22 June 2012 - 02:20 AM

My impression is that different astro software extracts canon raw files in different ways - and some might be better than others. One time when I messed with this stuff, Maxim seemed to do an ok job - compared to other things anyway. I haven't worked with the canon sdk myself, but I believe there are many ways to extract the "raw" data from the .cr2 files - and it has to be done right. Unless the extracted values are "raw" - the whole calibration process is a song and dance that won't make mathematical sense - and would make it hard to compare ccd with dslr.

It used to be, and still may be true, that it was simply impossible to get raw data from nikon because a median filter was applied that could not be undone - except with a hack. Not sure if that is still true, but if so it would make it hard to compare nikon results to ccd or even to canon.

So - if there is ever an article on dslr vs. ccd, I would put priority on whether or not you can access the raw values - and which processing software will actually do it that way. The only way to compare them may be to try them under the skies under typical usage conditions and compare results - rather than go by measured parameters.

Frank

#12 gdd

gdd

    Surveyor 1

  • -----
  • Posts: 1633
  • Joined: 23 Nov 2005
  • Loc: Lynnwood, WA (N/O Seattle)

Posted 22 June 2012 - 09:28 AM

FWIW, I finally am making some progress into the image processing Canon is doing on the images before they hit the CR2 file. I've suspected they're doing both scaling and shifting of the data but the evidence was more indirect. Now, it's pretty clear. They're doing it for a solid reason and one that makes sense for the target audience of normal photography. But, it does make things a bit tougher for the astro crowd.

Craig


Hi Craig,


Does this mean you can stretch the DSO images much further from a QHY camera than a DSLR? Does this mean I can get the same results from a QHY camera with a fraction of the integration time? Regardless of the advantages, would you still recommend a beginner like me should stick with a DSLR for a while?

Thanks,

Gale

#13 Craig

Craig

    Vendor (Stark Labs)

  • *****
  • Vendors
  • Posts: 768
  • Joined: 16 Sep 2003
  • Loc: Irvine, CA

Posted 02 July 2012 - 06:41 PM

My impression is that different astro software extracts canon raw files in different ways - and some might be better than others. One time when I messed with this stuff, Maxim seemed to do an ok job - compared to other things anyway. I haven't worked with the canon sdk myself, but I believe there are many ways to extract the "raw" data from the .cr2 files - and it has to be done right. Unless the extracted values are "raw" - the whole calibration process is a song and dance that won't make mathematical sense - and would make it hard to compare ccd with dslr.


Hey Frank,

I spent several days analyzing things and it's clear that Canon is both shifting and scaling the data before it hits the CR2 file. I've got about a 12 page report ready to put up here but that conclusion seems undeniable.

Running Canon's or Adobe's recon is really adding a ton. If you run:
"dcraw –D –r 1 1 1 1 -4 –T IMG_1000.CR2"

You're running linearly, just decompressing the lossless JPEG, and applying no white balance correction. (This is, in effect, what I do). You'll find if you do this that the mean signal in a dark frame does not increase smoothly with exposure duration. The dark noise (stdev of the difference b/n darks divided by 2) does and the width of the histogram nicely expands. But, the mean does not (and on some cameras it actually goes down).

The shift component is easy to see. The scale is a bit tougher, but if you calculate the system transfer function at different exposure durations you can see different slopes / gains. I'll submit this to the queue here soon for folks to look at.

One thing to keep in mind is that these effects make great sense for the target use of DSLRs. They do throw a wrench into pre-processing.

Craig

#14 Craig

Craig

    Vendor (Stark Labs)

  • *****
  • Vendors
  • Posts: 768
  • Joined: 16 Sep 2003
  • Loc: Irvine, CA

Posted 02 July 2012 - 06:46 PM

Does this mean you can stretch the DSO images much further from a QHY camera than a DSLR? Does this mean I can get the same results from a QHY camera with a fraction of the integration time? Regardless of the advantages, would you still recommend a beginner like me should stick with a DSLR for a while?



Gale,

There is a lot that goes into any puchasing decision. If you've got a DSLR and are working through the learning curve, enjoying things, etc. by all means stick with it. If you need something to do double-duty as an astro cam and a daytime use cam, you'd be hard-pressed to argue for any dedicated astro CCD. But, I do think that you'll get better SNR and things like dark subtraction will work better with a dedicated, cooled camera.

Craig

#15 SBrian

SBrian

    Vostok 1

  • -----
  • Posts: 144
  • Joined: 04 Sep 2010
  • Loc: New Hampshire USA

Posted 02 July 2012 - 07:11 PM

Craig I would be interested in hearing any feedback from you regarding the QHY10 vs the QHY12, Realestate vs well depth. I'm looking to purchase one or the other and I'm have a tough time choosing. Any feedback would be greatly apriciated.

Thanks
Steve

#16 Craig

Craig

    Vendor (Stark Labs)

  • *****
  • Vendors
  • Posts: 768
  • Joined: 16 Sep 2003
  • Loc: Irvine, CA

Posted 04 July 2012 - 02:27 PM

The 10 and 12 are very similar beasts. The 10, with it's bigger pixels (6 vs. 5.1) has a greater full-well (45k vs. 32k). The real question is whether you're going to be able to make good use of the smaller pixels or not. If we figure where you are at 1"/pixel, we get a decent rule of thumb for the longer-end you should be running at. Sure, under great conditions if you want the best resolution, you can go under 1"/pixel to some effect, but I find this as a nice starting point. The 10's 6u pixels put this at 1250 mm and the 12's smaller pixels put this at 1050 mm. With either, you can, of course go for shorter f/l rigs (I often run at 400 mm on my 7.4u pixels).

But, if you've got a 1m rig and are interested in higher resolution, the 12 is looking good. If you're at 1.5 m or longer, the 12's not looking so hot. You'll have lost SNR and full-well and not gotten much of anything in return. Personally, I'd take the bigger pixels and greater full-well of the 10 unless you're very short. Running a Hyperstar though and the 12's smaller pixels will actually get you more resolution and be likely worth the trade.

Craig

#17 SBrian

SBrian

    Vostok 1

  • -----
  • Posts: 144
  • Joined: 04 Sep 2010
  • Loc: New Hampshire USA

Posted 17 July 2012 - 08:35 PM

Thanks Craig and sorry for the delay in my response.
Craig the scope I am using is a celestron 11HD with a Hyperstar. So the 12 might make good sense. I currently use a canon t3i 18mp but not real happy with the results so far. Although not bad for my first year at Astrophotograph. The next step regarding the cameras seems to be very challenging. CMOS, CCD. I came across this web site adding even more confusion. http://www.centralds...-600d_1100d.htm
Cooled modified CMOS canon modifications to helping remove the major source of noise in the DSLR. Sighhhh

Question for you.

Q: will the deeper wells help prevent star bloating.
Q: will the deeper wells help with color depth.

Thanks

#18 Craig

Craig

    Vendor (Stark Labs)

  • *****
  • Vendors
  • Posts: 768
  • Joined: 16 Sep 2003
  • Loc: Irvine, CA

Posted 20 July 2012 - 04:38 PM

Question for you.

Q: will the deeper wells help prevent star bloating.
Q: will the deeper wells help with color depth.


It may. The trick to figuring this out is to look at your current raw images before any stretching. Are the stars saturated white disks there? If you mouse-over, do you get basically the same value (at the max) for a decent diameter? If so, yes, the deeper wells will help both of these. If not, if only a few pixels perhaps are saturated and the bloating and color loss is coming from your stretching.

It's worth checking that, but overall, yes, deeper wells will give you greater capacity and therefore won't saturate so easily. This lets you preserve the star shape and color better.

Craig

#19 freestar8n

freestar8n

    Soyuz

  • *****
  • Posts: 3988
  • Joined: 12 Oct 2007

Posted 31 July 2012 - 07:40 AM

I spent several days analyzing things and it's clear that Canon is both shifting and scaling the data before it hits the CR2 file. I've got about a 12 page report ready to put up here but that conclusion seems undeniable.

Running Canon's or Adobe's recon is really adding a ton. If you run:
"dcraw –D –r 1 1 1 1 -4 –T IMG_1000.CR2"


Woops - I was on vacation at the time and didn't notice you had replied to my note. And since then you did the other report on canon dslr behavior.

What's not clear to me is if you did anything with the canon sdk, or if you did it all with dcraw. If I understand correctly, dcraw does not use the canon sdk, and since aspects of the cr2 format are proprietary - I think - it seems like dcraw may not have full capability to recover the linear signal, while the sdk could. This would involve properly using metadata in the cr2 that allows recovery of the raw data.

I encountered this stuff years ago with a canon 20d and working with ImagesPlus as described her. Many people were saying that the early canon's had defective detectors and the newer ones were better, but by accessing the pixels differently through the sdk, the result was much more linear. I haven't characterized how linear it is, but it's clearly much improved.

So - are you sure that the cr2 data are corrupted from being linear in a lossy way, or is there perhaps a way to get linear behavior via the sdk?

If the behavior is nonlinear, then everything about flats, bias, darks goes out the window. It may even be better to use the in-camera noise reduction while imaging - even though you are spending half your time not collecting photons. This is heresy in a situation where the camera is well behaved - but if there is proprietary stuff going on, and the in-camera reduction does have access to the raw data, then it might be better in the end.

If dcraw is acting like my top image of m101, and the sdk with bayer basic raw looks like the bottom one, then there could be a big benefit from improved linearity by using the sdk.

Frank

#20 SMigol

SMigol

    Ranger 4

  • -----
  • Posts: 318
  • Joined: 30 Jul 2010
  • Loc: California, USA

Posted 08 November 2012 - 07:24 PM

Was thinking about this article and the even/odd lines readout detail and this thought came to mind:

How does this effect bias for noise reduction?

In the past, I understand that bias frames are the shortest exposure possible for the device. What is the shortest exposure for the 10 and does this exposure exhibit problems with bias frame capture? Is it better to make bias at some other duration?

Thanks for your thoughts on the matter.






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics