PDA

View Full Version : DMK21AF04 Artefacts at 60fps



oliver_p
December 17, 2009, 14:48:17
Dear TIS support,

I experienced some ghost image edge artefacts with my DMK21AF04 using it at 60fps and since several DMK21 users reported the same issues (e.g. http://www.theimagingsourceforums.com/showthread.php?t=322135 I conducted some "dry testing" and could confirm that the 60fps mode seems to be the root cause, my test summary can be found here http://www.astro-imaging.de/astro/dmk_artefacts_20091211.html.
Looking at the ICX098BL datasheet I recognized that the specs seem to state a max. progressive scan frame rate of 30fps, 60fps only in monitor mode with two lines read and two lines skipped. Might this be the reason for the edge artefacts? Today I read about a user experiencing the same artefacts with his DMK31 at 30fps and the datasheet of the ICX204AL states a max. progressive scan of 20fps - same issue here?

Thanks for any clarification & best regards,
Oliver

oliver_p
January 6, 2010, 09:19:41
Hello,

ok, it seems that the max horizontal drive frequency of 12.27MHz of the ICX098BL is not enough to drive 60fps without artefacts. This issue might be overcome with true Region of Interest processing inside the camera, not in the capture software like today. Is the FPGA/-Controller used in the DxKs capable of ROI with a different firmware? Are there any plans to update the cameras with ROI? It would be extremely helpfull for planetary imaging and overcome the framerate issues.

Best regards,
Oliver

Stefan Geissler
January 6, 2010, 09:51:49
Oliver,

I am very sorry, but there are no ROIs planned right now. Also the artifacts seem to be created by noise only and are not visible in single image as far as I know.

oliver_p
January 7, 2010, 11:38:47
Stefan,

hm, I don't think the artefacts are caused by noise. They are very present at 60fps and disappear completely at 30fps, all other camera setting identical. They are also present in the raw frame data and a lot of users experience this issue. 12.27MHz horizontal drive frequency is just not sufficient to drive a 60fps progressive scan?

Best regards,
Oliver

Stefan Geissler
January 8, 2010, 09:16:00
Hello Oliver,

Please understand, that we do not publish anything about the firmware in the camera.

I woud like to know, do you see any of these artefacts in the single images?

Emil
January 10, 2010, 19:01:40
You can't really see the artifacts in single images, because the noise in the single images is higher than the signal of the artifact. But that doesn't mean the effect doesn't exist. Averaging out the noise - by stacking multiple images -, shows that the artifact does exist. Just like planetary details can be brought up in a stack of images, that are often impossible to see in the raw frames, those planetary details are just as real as this artifact.

http://www.astrokraai.nl/dump/DMK21AU04.as-Ghost-Artifact---11-pxiels-right.jpg

Edit; an image to further clarify my story. The top row shows raw stacks of images of 1, 10, 100 or 1000 frames, the row below are sharpened versions of those images. At 10 frames stacked, you can cleary see the ghost effect appearing (especially in the sharpened frames, but it's also visible in the 10 frame raw stack if you stretch it a bit).

So altough the effect might be minimal in single raw frames - just like you can't see a lot of cloud structures in single frames - it is clearly there, as can be seen in the larger stacks.

So what exactly do you mean with it being a noise artifact Stefan ??

Planetary imaging is always about imaging with noise; it's not just 'freezing the seeing' (lucky imaging) to get good images, but you basically MUST average out the noise the camera produces. Single shot images are useless if you want to get the most out of your equipment. Normally the noise is pretty much random (and the static noise is averaged out because the planets move about a little bit during imaging). But this ghost artifact noise is always present at 60fps, and it's location is depending on the contents of the frames (more specifically, +-11 pixels to the right of high contrast area's), so averaging out is not possible anymore, hence the effect shows up more clearly in bigger stacks.

oliver_p
January 11, 2010, 08:45:29
Hello Stefan,

I understand that you won't publish any internals regarding your firmware but I refer to the CCD characteristics and specs (horizontal drive frequency) according to the Sony datasheet - nothing to do directly with the firmware. CCDs capable of 640x480 pixel with 60fps progressive scan have about twice the max. horizontal drive frequency of the ICX098 (e.g. ICX424: ~24MHz).
The artefacts are barely visible in the raw frames since they are buried in the noise but you get a hint if you know where to look at in contrast enhanced raw data. Since the artefacts are systematic/static and not random like the noise just like the planetary details we capture they appear if only a few frames are stacked and processed to eliminate random noise just like Emil demonstrated. This is the same approach that makes the planetary details visible in our usual processing workflows and can't be avoided.

Best regards,
Oliver

Emil
January 12, 2010, 15:11:47
Hi Stefan,
I don't want to bother you too much, but just wondered if you already have an answer to my question two posts above (and any other information on the matter?)?

Stefan Geissler
January 12, 2010, 15:25:01
Hi Emil,

If I had an answer, I would have answered already.

I guess, but I am not sure about, because of lack of knowledge about the physical background of CCD technics, that the read out noise of the CCD creates the artifacts. This noise can be reduced by reducing gain.

I guess, it is noise, because the artifacts are not visible in the single frames. You need to stack frames to make them visible. This may means, that the artifact pixels have no consistent brightness, thus they are kind of noise. This happens, if the CCD runs with a very high clock of 60 fps.

As you can see from my answer, there are many "guess" and many "may" used.

Emil
January 12, 2010, 17:23:44
Thanks for answering Stefan (altough there are indeed a lot of guesses and maybes:-) )

But reducing the gain is just not an option for planetary imaging. Well, you could go to about 700 gain for example, but even then the ghost effect is still visible. And because the image is dimmer when gain is lower, after brightening the image back up, you are back at where you started: a pretty strong ghost image effect at 60fps, making the camera pretty much useless for planetary imaging at 60fps.
And that is kinda weird for an (astronomical) camera that is said to work well at 60fps.


If I have time, I'll try to do some more testing this weekend, also at very low gain rates to try and get a better understanding of the effect. I'll report back here (if Oliver is fine with that of course ;-) )

Anyway, please let us know if you have more information. Do you, or any engineers at TIS, know anything about the max horizontal drive frequency of 12.27Mhz of the icx098 ccd's, and that it might be related??

oliver_p
January 13, 2010, 14:17:19
Hi Emil,

sure, why not. All input is helpfull to evaluate the root cause (although I think it's already been found ;)).

Best regards & clear skies,
Oliver

Emil
January 14, 2010, 17:40:37
According to the specification sheet by the manufacturer of the icx098bl ccd (and comfirmed by a big imaging components company in germany) the maximum framerate for these ccd's is 30 fps.

Direct quote from the datasheet of the icx098bl ccd: "Progressive scan allows all pixels signals to be output independently within approximately 1/30 second."

So if you want to try it faster (at 1/60s or 60fps), the pixels can't be read-out independently anymore (and the 'ghost artifact' seems to fit this story well?). The icx098 does support 60fps in monitoring mode, but that means that only 50% of the pixels can be read-out at the same time.

the ICX098BL doesn't support full progressive scan imaging at 60fps, and the DMK21 camera's are running the ccd out of specification when trying to do achieve that framerate.
By the way, the DMK31, and DMK41 are also running their ccd's out of specification at 30fps and 15fps respectively.

Stefan Geissler
January 15, 2010, 10:42:04
I have asked one of our developers about the 60 fps issue.

He told me that we - as a big customer of Sony - have much information beyond the things they provide in the "official" datasheet.

He has no permission to publish this information but he guaranteed that it is possible to get the full resolution in progressive scan mode with 60 fps.

oliver_p
January 15, 2010, 15:03:12
Hello Stefan,

obviously it is possible to operate the CCD out of specs @60fps but at the expense of artefacts in high gain applications - you can't beat physics ;)

Best regards,
Oliver

Stefan Geissler
January 15, 2010, 16:56:00
Oliver,

it seems, you get nothing for free, you always have to pay a price. However in machine vision, where stacking is not used, the 60 fps are very useful, especially when machines run very fast.

Emil
January 15, 2010, 21:49:26
Stefan, for ISS imaging I often use the camera at 60fps. The increase of fps just outweighs the little noise that is hardly visible at those small stacks (I only use a maximum of 10 frames per stack at most).

I'm very much interested in creating a software solutions to this ghost image problem. I think the effect is highly predictable, and that it should be possible to 'fix' it in the raw frames of a 60fps avi. It's just something I'm thinking about at the moment...

Stefan Geissler
January 18, 2010, 10:21:45
Hi,

since the effect is not very predictable, be cause it seems to be created by kind of noise, I have no idea for a software workaround. The effect is barely visible in the single images and it is on right side of bright-dark (and vice versa) changes. It is as faint in the raw images, so I have no idea, how to detect it by software and correct it. The "wrong" pixels are randomly on the single images. Only after stacking them, you can see them.

oliver_p
January 19, 2010, 16:58:49
Stefan,

of course, for applications demanding no high gain settings and no heavy sharpening applied to the image (e.g. machine & production line inspection) the 60fps mode is perfectly fine. Unfortunately the same camera is marketed as an astronomy camera targeted at planetary imaging. We require high to max. gain for a lot of targets like e.g. Saturn. The camera works absolutely fine switched back to 30fps but 60fps would be very nice for e.g. solar and lunar targets and other bright objects like Venus, Jupiter & Mars. Maybe there should be a note in the documentation for astro applications.

Best regards,
Oliver

Stefan Geissler
January 19, 2010, 17:03:01
Oliver,

I will forward your hint regarding a note to my bosses.

oliver_p
January 20, 2010, 12:01:38
Ok, thanks!

Best regards,
Oliver

oliver_p
February 3, 2010, 14:40:28
Hello,

one update: The DxK31 show the same issue at max. FPS 30fps, the artefacts are gone @15fps. Root cause is again the max. horizontal drive frequency of the ICX204.

Best regards,
Oliver

Emil
February 28, 2010, 01:37:38
Oliver (and TIS-team), I don't believe the root cause is the horizontal drive frequency of the CCD, because my DMK21 that is equiped with the ICX618ALA CCD (which has double the horizontal drive frequency of the ICX098BL-E) shows the exact same problems at 60fps. This can be seen in the following image of a (very bad) indoor star-test using the modified DMK21 camera. Again the ghost effect appears +-11 pixels to the right of a bright edge at 60fps, and goes away at 30fps or lower (the fps setting is the only difference between both images)

In infrared light (>807nm):
http://www.astrokraai.nl/dump/icx618ala-e_whaaaaaaaaa_ghost.jpg

And in UV light (350-400 nm):
http://www.astrokraai.nl/dump/ghost_in_UV.gif


So the source is probably something else in the DMK21 electronics that kicks in at 60fps (or perhaps it is related to the CCD, and it's always there at higher FPS? I can't really believe that though, because the ICX618ALA is rated at a full frame speed of 60fps).

Stefan Geissler
March 1, 2010, 10:11:31
Emil

Thank you very much for the images. Unfortunately the exact reason for the artifacts is not known. I am very sorry for this.

jmelquist
July 15, 2010, 08:37:26
My name is Jason Melquist. I opened case # 156977 on Sept 8th 2009 on this exact topic. (see artifact in this image: http://www.flickr.com/photos/jasmel90/3851845554/) I was told it was an artifact of my processing despite providing much evidence to the contrary. I was passed back and forth among reps but I put up enough of a stink that I was reluctantly granted an RMA. The New camera I received exhibited the same behavior, but I gave up and have been imaging at 30fps ever since. I was extremely upset at the time - mostly by The Imaging Source's poor support, and now finding others complaining about the exact same behavior, I'm really mad. This is a big issue which needs to be addressed and rectified. Please keep me up-to-date with any progress or updates with this issue! I can be reached at jmelquist(at)gmail(dot)com.

Mars4ever
November 14, 2010, 17:41:29
Hi all, I own a DMK21 and I've noticed the same issue, so I came on this forum to look for this discussion because I was sure that I couldn't be the only one with this problem!
I quote 100% of the words said by Emil: his analysis is perfect and I confirm that theese artifacts happen ALWAYS and ONLY at 60 fps, so they NEVER appear at 30 fps or less, whatever the others parameters are!
I show you a comparison between two pics of Venus (the only planet where I can use an exp. time <=1/60 s), taken with same exposure (about 1/90) and gain.
30 fps:
http://img535.imageshack.us/img535/6577/2090806043930fps.th.jpg (http://img535.imageshack.us/img535/6577/2090806043930fps.jpg)
60 fps:
http://img197.imageshack.us/img197/2518/2090806044160fps.th.jpg (http://img197.imageshack.us/img197/2518/2090806044160fps.jpg)

patry
November 22, 2010, 14:44:47
Is it possible it could be due to a cable ?
For my part, I've never had it with my DMK21

Here's what I've done at 60fps (more than 54fps really acquired)

http://www.theimagingsourceforums.com/attachment.php?attachmentid=1328&stc=1&d=1290431328

I said that because I've changed early the 1 meter or so cable for a better quality (hope so) 2m USB cable.
I have never had such artefacts from then (here's another image captured recently)

http://www.theimagingsourceforums.com/attachment.php?attachmentid=1329&stc=1&d=1290432464

(Same capture mode, C11, prime focus, DMK21, R,G,B, and IR742 Astronomik filters, captured on a DELL 6400, planes processed by Registax, color reconstruction using Registax). No N/S inversion for the last image.


Marc

Mars4ever
November 22, 2010, 19:22:29
Is it possible it could be due to a cable?
...
I said that because I've changed early the 1 meter or so cable for a better quality (hope so) 2m USB cable.It's very easy to verify this hypothesis: do you still have (or can find by someone else) the original cable? ;-)

patry
November 23, 2010, 15:19:12
Probably yes but I 'm not sure to recognize it. I may test all I have (dispatched against the cable "length and use" at home).

I will try to set up a procedure to test all.


Marc

Mars4ever
November 28, 2010, 16:03:14
It's grey and 1,5 m long.

tea drinker
February 9, 2011, 16:56:58
Heavy sharpening and processing can influence this kind of noise. That's why de-ringing exists in registax, it's not only for imaging source camera's.
As already mentioned USB conditions could be affecting this. Maybe some noise is coming apparent on the PC USB at high rates, or the additional framerate is causing some voltage drop.
My own laptop seems to have 2 higher power ports on one side than on the other, might be worth using another port.

But it could be a real issue in the camera / control software of course.

jmelquist
August 19, 2011, 16:18:43
All, I have been working with the TIS support group and wanted to document my issue here as well - FYI:

Background: I'm using the DMK21AU04.as camera for astronomical use, imaging planets (Jupiter in particular which results in a bright disk against a black background, a very high contrast situation with sharp edges) using a Celestron CGEM800 telescope at high gain levels (full gain), mid gamma levels (~100) using the IC Capture software. The [dark ring artifact] appears when I use the camera in 60fps image-read-rate mode. This is NOT to be confused with the individual frame exposure (which is unrelated), but rather with the image read-rate of 60fps (30fps is the other image read-rate option and does not exhibit the artifact in question.) With these parameters, I record an AVI with around 3000 frames in uncompressed Y800 format. I then use frame-stacking software (Registax v5) to combine the frames into a single stacked image which makes the artifact evident. If the image is sharpened using wavelets, the ring artifact becomes VERY pronounced and ruins the image - a certain amount of wavelet sharpening is needed to bring out planetary detail. This process is currently the most common way among astronomers of doing astronomical/planetary imaging - which this specific camera is marketed for. In my opinion, it will be very difficult (although not impossible with highly-controlled lighting conditions) to reproduce these conditions in the lab.

Until this issue is resolved i'll be using a DFK21au04.as (color version) at 30fps read-rate. The color sensor will is about 1/2 as sensitive as the DMK and i'll obtain only 1/2 as many frames per second, but it will allow me to image for 4x the time since I don't have to switch between L, R, G, and B filters. if others have an 8" aperture scope or higher (thus having plenty of brightness), this makes for nearly a wash in terms of image quality.

jmelquist
September 7, 2011, 23:03:43
Just an update with some recent shots using the DFK21au04...w/o the 60fps artifact:
1414

jmelquist
September 7, 2011, 23:07:33
sorry, forum is not allowing updates... here are the image specs:

Just an update with some recent shots using the DFK21au04...w/o the 60fps artifact (see attached). Taken with C8 (Celestron 8"sct) on CGEM with DFK21au04 thru 2.5x barlow from Minneapolis, MN area in early Sept, 2011.

1414

Stefan Geissler
September 8, 2011, 09:52:51
Hello,

thank you very much for the Jupiter images.

You are right, there is a problem with editing posts in the forum. I am searching for a solution.