Jump to content


Photo

16-bit solar with Flea3

  • Please log in to reply
17 replies to this topic

#1 bunyon

bunyon

    Gemini

  • *****
  • Posts: 3325
  • Joined: 23 Oct 2010
  • Loc: Winston-Salem, NC

Posted 30 January 2013 - 10:06 PM

I had a chance yesterday to shoot some images through my PST with the Flea3. I once again set it to 16-bit capture. A commenter in the solar forum said he didn't bother shooting a separate image for prom and disk, that with the 16-bit capture he could get both in one go. I haven't been able to do that with 8 bit captures but it did work well with 16 bit. I was going to shoot 8 bit but clouds rolled in.

Compare the one linked with the original which was made by combining 8 panels - 4 for disk, 4 for proms. The linked image is 4 panels - prom and disk from same image in each.

Posted Image

#2 MvZ

MvZ

    Surveyor 1

  • *****
  • Posts: 1583
  • Joined: 03 Apr 2007
  • Loc: The Netherlands

Posted 31 January 2013 - 06:14 AM

Very nice image Paul! Still a couple of merging artifacts visible (the coarse cross), but not much you can do about that if the seeing isn't the same for each panel.

Sorry if this gets old to anyone, or if I sound like a **** by repeating this, but I again like to point out some 8 vs more than 8-bit stuff. Not interested? Than skip the following: ;)

Even 8-bit images can capture both prominence and disk at that same time. Usually, 16-bit (or 12 or 10) doesn't add anything and you get the exact same quality with 8 vs. higher bits. ( Unless you use wrong camera settings: you didn't watch the histogram closely, no clipping and neutral gamma! OR the gain value was extremely low). If you can capture more than 8 bits with the same fps as 8-bits, and you have the disk space (or you just throw the recordings away or the money to buy it), then of course you CAN use more than 8-bit without problems.

But to avoid people from spending money when it isn't necessary, or even getting worse results: for planetary and even h-alpha solar recordings (and it some cases even white light solar recordings at extremely low gain values), when using the correct camera settings, 8 bit recordings are 99% of the time just as good as 12-16-bit recordings. They capture the same information in both the dark and bright areas of the image.

If imaging in more than 8 bits hurts your FPS, the final stack is usually actually WORSE than an 8-bit recording.

#3 bunyon

bunyon

    Gemini

  • *****
  • Posts: 3325
  • Joined: 23 Oct 2010
  • Loc: Winston-Salem, NC

Posted 31 January 2013 - 06:43 AM

It's true that I'm still very new at H-alpha imaging so not having previously been able to do this with 8 bit images may well simply be me getting better.

I also agree in general with the idea that if it hurts frame rate (it does) that it will hurt the image. I wonder if it holds with solar imaging. For instance, my stack here was 150 frames. 150 out of 2000 or 150 out of 4000 seems to not make much difference?

And, yes, the merging artifacts bug me. But they're far better than my first dozen or so attempts in which it looked like I had a painted a big X on the sun.

#4 MvZ

MvZ

    Surveyor 1

  • *****
  • Posts: 1583
  • Joined: 03 Apr 2007
  • Loc: The Netherlands

Posted 31 January 2013 - 07:12 AM

> 150 out of 2000 or 150 out of 4000 seems to not make much difference?

If the frames are of the same quality, than of course it does not make any difference at all.

However, if you capture twice the amount of frames, this means you'll also capture twice the amount of good frames. So the average quality of the 150 best frames is higher in the 4000 recording compared to the 2000 recording. For 'normal' seeing conditions, this probably means a (small) increase in sharpness for the surface features, and thus also decrease the merging artifacts a bit. You could also choose to stack more frames (say 300/2000, the same quality percentage) and adjust your processing a bit. This will probably lower the noise in your stack, especially for the prominences this could make a difference.

You can never have too many frames to play around with.

#5 ValeryD

ValeryD

    Vendor (Aries)

  • *****
  • Vendors
  • Posts: 1147
  • Joined: 26 Nov 2005
  • Loc: Kherson, Ukraine.

Posted 31 January 2013 - 05:07 PM

Very nice image Paul! Still a couple of merging artifacts visible (the coarse cross), but not much you can do about that if the seeing isn't the same for each panel.

Sorry if this gets old to anyone, or if I sound like a **** by repeating this, but I again like to point out some 8 vs more than 8-bit stuff. Not interested? Than skip the following: ;)

Even 8-bit images can capture both prominence and disk at that same time. Usually, 16-bit (or 12 or 10) doesn't add anything and you get the exact same quality with 8 vs. higher bits. ( Unless you use wrong camera settings: you didn't watch the histogram closely, no clipping and neutral gamma! OR the gain value was extremely low). If you can capture more than 8 bits with the same fps as 8-bits, and you have the disk space (or you just throw the recordings away or the money to buy it), then of course you CAN use more than 8-bit without problems.

But to avoid people from spending money when it isn't necessary, or even getting worse results: for planetary and even h-alpha solar recordings (and it some cases even white light solar recordings at extremely low gain values), when using the correct camera settings, 8 bit recordings are 99% of the time just as good as 12-16-bit recordings. They capture the same information in both the dark and bright areas of the image.

If imaging in more than 8 bits hurts your FPS, the final stack is usually actually WORSE than an 8-bit recording.


My results tell me quite opposite. The final images I obtained with my Flea 3 with 16 bit (12bit at the ADC) are consistently better than the images I took with 8 bit.

8 bits I use only for central areas of the disk. And if one want to have best results with his not so cheap solar equipment, a large hard disk and intermediate SSD disk are the must have. After the final pictures are finished all the movies can be erased and disks are ready to take more movies again. I see no any problem with this.


Valery.

#6 MvZ

MvZ

    Surveyor 1

  • *****
  • Posts: 1583
  • Joined: 03 Apr 2007
  • Loc: The Netherlands

Posted 31 January 2013 - 05:25 PM

Valery, you should provide details. Now there is no way of telling if 1) Physics is wrong, 2) your camera is wrong, or 3) you missed the part where I told about camera settings.

edit: 4) processing is wrong (median stacking, software clipping your 8-bit avi recordings,etc).

#7 DesertRat

DesertRat

    Fly Me to the Moon

  • *****
  • Posts: 5228
  • Joined: 18 Jun 2006
  • Loc: Valley of the Sun

Posted 31 January 2013 - 08:19 PM

Revisiting 8 vs 12 bit again I think what Emil states is correct. The relationship is in favor for 16 bit capture if the number of frames is quite low. How low depends on the well depth and read noise of the ccd. The actual number of 'good' bits in a 618 chip camera like the Flea3 is probably somewhere between 9 and 10.

In nearly perfect seeing (!) you can get really nice planetary images with a true astrocamera summing just a handful of frames. I know a guy that does just that, but his seeing is a lot better than mine.

Edit: Sorry Paul, should have mentioned - nice shot! I'd like to get back to halpha solar imaging someday myself.

Glenn

#8 DonAtPointGrey

DonAtPointGrey

    Vendor - Point Grey

  • -----
  • Vendors
  • Posts: 44
  • Joined: 09 Mar 2012

Posted 01 February 2013 - 12:50 PM


Not sure if this comment is applicable but I'll make it anyway. When reviewing this issue a couple of months ago I completely convinced myself that for planetary imaging 16-bits was of no us because the gain settings typically used are very high. For example, most Point Grey cameras have a maximum gain setting of 24 (which is the equivalent of a 4 bit shift - which consequently throws the top 4-bits of the dynamic range away - which is not a problem for planetary because the images are so faint, there is no information in the top 4 bits anyway). Given that the signal-to-noise ratio of our current cameras that are used in this application provide a maximum of 11-bits of signal, when you throw away 4 of those bits it is clear that an 8-bit image will give you all the resolution you can expect.

However, can someone let me know what kinds of gains are typical on solar imaging? I have no idea how much is lost through the filter etc but am hoping that in solar imaging you are using 0db gain and a lower-than-max shutter. In which case there may be an argument that 16-bit imaging will gain you something.

Just a thought - I'm sure Emil has this figured out already :)

Don

#9 MvZ

MvZ

    Surveyor 1

  • *****
  • Posts: 1583
  • Joined: 03 Apr 2007
  • Loc: The Netherlands

Posted 01 February 2013 - 01:45 PM

Yes, it depends on gain a lot. I'm not entirely sure what the typical gain settings are, for white light imaging they are very low (too low is often not recommended to get good image quality, some cameras start with 3.6dB instead of 0.). H-alpha requires much more gain. Unless you use high exposure times (which usually isn't a good idea)

I'm interested in a 16 bit H alpha recordings for which is claimed they give better results than 8 bit recordings.

P.s. some software support real-time noise measurement. That can be used to quantify the noise in practice. If a pixel changes randomly by a couple of values on an 8 bit scale even though its true brightness shouldnt change, then 8 bit is enough.

Sorry for the typing errors, Im in the train working on my phone...

#10 bunyon

bunyon

    Gemini

  • *****
  • Posts: 3325
  • Joined: 23 Oct 2010
  • Loc: Winston-Salem, NC

Posted 01 February 2013 - 01:49 PM

Don, I use "zero" gain for Ha captures. That is, the slider is all the way to the left (160 on the dial). The exposure was 1.27ms, so well below max.

I am complete agreement, empirically and theoretically, that 16 bit recordings with planets doesn't help. I get the theory that it doesn't help with the sun, but it isn't what I observe.

Thanks.

#11 mikewirths

mikewirths

    Vendor-Baja Dark Skies

  • *****
  • Vendors
  • Posts: 749
  • Joined: 16 Jun 2008
  • Loc: San Pedro Martir Mexico

Posted 01 February 2013 - 01:59 PM

My experience using the Lumenera Infinity 2-2 (using Sony 274 chip) on the moon with my 18" Dob is that 12 bit images are slightly better than 8. I use almost no gain (except right on the terminator) and get about the same framerate with 12 bit as with 8 bit (11-12 fps using 1600X1200 roi) The difference in the final image is that with 12 bit tiff stacks brightly lit crater edges have a much smoother transition to black shadow area than with 8 bit where they appear "stepped". Wnen I get my grasshopper express I will be experimenting to see if the experience is the same.

cheers

Mike

#12 MvZ

MvZ

    Surveyor 1

  • *****
  • Posts: 1583
  • Joined: 03 Apr 2007
  • Loc: The Netherlands

Posted 01 February 2013 - 02:15 PM

Ok, (close to) zero gain values it is. That clears things up a bit.

Small quote from myself before the battery dies:
>8-bit no use .. "( Unless .[...] the gain value was extremely low)"

P.s. that also means if you do use 8 bits, you MUST apply enough gain in order to avoid quantization effects giving a huge loss of detail!

#13 ValeryD

ValeryD

    Vendor (Aries)

  • *****
  • Vendors
  • Posts: 1147
  • Joined: 26 Nov 2005
  • Loc: Kherson, Ukraine.

Posted 01 February 2013 - 04:19 PM

Valery, you should provide details. Now there is no way of telling if 1) Physics is wrong, 2) your camera is wrong,


The dynamic range with 16 bit (really 12 bit expanded to 16) is wider than 8 bit one. This is a s clear as a day. The brightness variations at the disk and prominences are much higher than in the center of the disk.
I need to differentiate the very dim parts of prominences and in the same time I need not overexposed disk details - I need a wide dynamic range, significantly wider than 256 levels.

And I am not intendent to discuss the subject which is clear and the results I obtain consistently prove this physics.

The guys who do use CCD cameras with true 16 bits obtain results whcih are even better than mine, significantly better, because they have 16 bit instead of my 12 bit.

My camera works fine and I have correct settings. I have had a discussion in length with PGR engineer at he help me out to make my camera work with 16 bit at the same speed as with 8 bit.

Valery.

#14 ValeryD

ValeryD

    Vendor (Aries)

  • *****
  • Vendors
  • Posts: 1147
  • Joined: 26 Nov 2005
  • Loc: Kherson, Ukraine.

Posted 01 February 2013 - 04:29 PM

However, can someone let me know what kinds of gains are typical on solar imaging? I have no idea how much is lost through the filter etc but am hoping that in solar imaging you are using 0db gain and a lower-than-max shutter. In which case there may be an argument that 16-bit imaging will gain you something.

Just a thought - I'm sure Emil has this figured out already :)

Don


Very often I don't use any gain for the sun imaging. With 150mm F/5 aperture and 0.7A band wide I have enough light for imaging without any gain. 150mm provide much better S/N in photon noise vs, say, 40mm, 50mm, 60mm, 80mm or 90mm solar telescopes wihch are often used by most solar imagers. F/5 give enough energy density at each pixel - so I don't need to use gain.

Also, I do as long exposure as the seeing permits. I better summarize more frames that I will loss dynamical range.


Valery.

#15 MvZ

MvZ

    Surveyor 1

  • *****
  • Posts: 1583
  • Joined: 03 Apr 2007
  • Loc: The Netherlands

Posted 04 February 2013 - 10:00 AM

Valery, it is clear now that you used the extremely low gain settings, that explains everything. This means you get close to what your camera can actually do (10.something bits). But without mentioning that you use low gain settings and then claiming 16-bit is needed is making the discussion less clear than it should be.

It might also have been a good idea to quote the entire sentence, instead of picking out just the first two items where I wrote why you should provide details

Now there is no way of telling if 1) Physics is wrong, 2) your camera is wrong, or 3) you missed the part where I told about camera settings.



The text in bold refers to my first post here:

...OR the gain value was extremely low


>And I am not intendent to discuss the subject which is clear and the results I obtain consistently prove this physics.

But 1) details matter a lot! and 2) we are discussing now, are we not ;)

#16 ValeryD

ValeryD

    Vendor (Aries)

  • *****
  • Vendors
  • Posts: 1147
  • Joined: 26 Nov 2005
  • Loc: Kherson, Ukraine.

Posted 05 February 2013 - 06:39 PM

I will told you more, Emil. I plan to buy even better camera soon after they will appear on the market and will be of very low noise and at least 12-14bit (real). This will makes my images better and I hope to have 0.35" resolution for details on sun disk and register as narrow filaments as 0.17" or even 0.12". For this I will build a 14" solar Ha scope. In this case I surely will need 12bit mimimum.

Valery.

#17 bunyon

bunyon

    Gemini

  • *****
  • Posts: 3325
  • Joined: 23 Oct 2010
  • Loc: Winston-Salem, NC

Posted 07 February 2013 - 08:32 AM

Well, it isn't a huge difference in disk detail. Below are two crops from an 8 bit and a 16 bit image. 8 bit is 150/3500 frames and 16 bit is 150/1980 frames. Both were processed identically and recorded within the same minute.

One thing I can say, though, is the effect on dynamic range. In the 8 bit images of the limb, I can't pull out the prominences well at all. However, in the 16 bit image, I can. For that reason, I will probably continue to use 16 bit on these images.

Not sure the order these will go but the first is a link to a full disk 16 bit image on Feb. 6. The other is a crop, 8 bit and 16 bit. Do you see a difference? I'll reveal later which is which.

Posted Image

Attached Files



#18 DesertRat

DesertRat

    Fly Me to the Moon

  • *****
  • Posts: 5228
  • Joined: 18 Jun 2006
  • Loc: Valley of the Sun

Posted 10 February 2013 - 01:54 PM

In the cropped image the left side appears to have a somewhat lower noise level. But it is not much of a difference. I think I would judge this too close to call really. It would be interesting to see what if any difference was realized by stacking maybe just 10 frames.

Glenn






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics