Color Fringing: What's the Cause? at DVinfo.net

Marc Young · November 2nd, 2003, 09:35 AM

Given the sensitivity of the DVX100 to manual focusing, I created a diy focusing chart in powerpoint, consisting of True Type fonts ranging from 36 - 72 pt in size (0.5 to 1"). The paper was slightly tilted and the camera was focused on it from a distance of around 4 - 5 feet. The firewire output was connected to a panasonic dvd recorder and the image was displayed on a progressive sony 36" tv using component video connections. Please explain the red and cyan overhang on the black lettering. How is this possible on a b&w image? Even if the color is undersampled under 4:1:1 coding, shouldn't it appear gray? Are the CCD's aligned properly or is there distortion in the lens? Anyone else have an eye chart or other pattern to test this on?

I did the same experiment with a cheap JVC single chip camera, and did not see the same thing. Admittedly, the Panasonic looked much smoother (no ringing). This experiment also demonstrated that during panning, 24p is better than 30p because most deinterlacers on consumer tv's are not geared to turn the signal into 60p. They still think they're being fed 60i. Also, 24p advanced is not interpreted correctly by Panasonic's dvd recorder. It displays correctly with a static image, but it twitters slightly with a moving image. Nothing is more merciless than an image with sharp edges, like a whole bunch of letters.

Boyd Ostroff · November 2nd, 2003, 10:31 AM

<<<-- Originally posted by Marc Young : The firewire output was connected to a panasonic dvd recorder and the image was displayed on a progressive sony 36" tv using component video connections.. -->>>

Just curious because this was discussed before... when using the DVD recorder to transcode from D25 to component, isn't the image undergoing MPEG compression first? Maybe this is a factor?

Also (although a bit off topic) which model Panasonic recorder to you have and how do you like it? I've been thinking about getting one myself.

Marc Young · November 2nd, 2003, 11:22 AM

My Panasonic dvd recorder models are the DMR-HS2 and the DMR-E100. Both have firewire inputs. On live display, there is no mpeg-2 encoding (this is only done by the original Philips recorder).

The latest Panasonic model outputs a very good picture, and the mpeg-2 encoding (recording) is near the top of the field (surpassed only by Sony or Pioneer, and the margin, if any, is very slim). It is an alternative to a pc for getting your mini dv camera output to a dvd-r. We all know about the problems with choosing the right codecs and mpeg encoders on a pc. Or standardizing on one nle. That is a huge can of worms. Pick the wrong one and your end product will look worse than any standalone recorder. In fact, I'm starting to reach the conclusion that unless you pick the top mpeg-2 encoders on a pc, it is better to print your nle output to tape, and let the standalone recorder handle the dvd creation.

The JVC dvhs recorders are another alternative for people who want to convert firewire, or record their mini-dv tapes on another format, for viewing on a large screen tv. Another display option is one of the marantz digital linear projectors and one rear-screener (forget the model #) which have firewire inputs. Before you edit any recording, make sure it looks good enough to author on a dvd. That takes a big tv imho. Nothing exposes the circle of confusion like a big screen. Otherwise, just go out and re-shoot.

Andre De Clercq · November 2nd, 2003, 12:53 PM

>On live display, there is no MPEG-2 encoding...
Where did you get this information. As far as I know there is no consumer equipment on the market which outputs component video straight from a DV input.

Andre De Clercq · November 2nd, 2003, 01:15 PM

On the fringing...I would first turn the saturation of the Sony to zero. If you still see "fringing" there is a problem with the set: convergence and/or focus tracking is not OK. (Remember that unsharp edges from yr old cam hide those problems). If you then don't have fringes I would turn up sat again and enter luma only straight from the cam into the Sony Y/C input (special cable). If you still see fringes yr comb or notchfilter in the Sony is not optimal. If OK I would enter Y signals from the Panasonic into the Sony....Also remember, if yr component interconnection is not short or unmatched (in length)you can get fringes.

Stephen van Vuuren · November 2nd, 2003, 01:19 PM

You can purple fringing on the DVX100 from overexposing highlighted edges in cinegamma.

Marc Young · November 3rd, 2003, 10:26 AM

<<<-- Originally posted by Andre De Clercq : >On live display, there is no MPEG-2 encoding...
Where did you get this information. As far as I know there is no consumer equipment on the market which outputs component video straight from a DV input. -->>>

It gets mighty close, meaning that on a static image, it looks reasonably good compared to my PC monitor. Whether consumer dvd recorders save a firewire transmission, or for that matter, an s-video signal, in their field/frame buffer as 4:2:0, 4:1:1, 4:2:2 or 4:4:4 is not known to me since I don't design this equipment, nor do I have the test equipment to verify that the component video is an exact analog representation of the firewire data. It just seems logical that the Panasonic recorder would synthesize the 720x480 frame buffer using two field buffers, without undersampling the chroma more than it is in the incoming signal. After all, when the recorder is not recording, it is merely transcoding the digital signal using simple math. In fact, it is more likely they resample/interpolate the incoming data, so the color resolution is finer than what came in. Unlike Philips, the feed-through image to the tv does not degrade based on recording mode (XP, SP, LP, EP). Because in the Panasonic, you're not recording yet, just watching the feed. Also, even I can see the difference between the firewire signal through the recorder and out on component (progressive) video vs. s-video from the camera directly into the tv, so the E100 must be improving something. Now clearly a consumer tv does not offer 27 MHz bandwidth on non-HD component video, but then again, neither does this camera. It would be interesting to A/B a professional mini-dv tape deck/editing station with component video out vs. the dmr-e100, but I doubt the pro-stuff will do 3:2 pulldown (at least not cheaply). Try doing something as "straightforward" as a frame grab on pro equipment, and you can be sure they'll charge you a $5000 entry price.

Anyhow, since I seem to have gotten off on a tangent, I'll get back to the color fringing anomaly. It may be the Sony TV and its DRC (digital reality creation) or some other signal processing quirk. Possibly even the color matrix. When in doubt, blame the color matrix (quote Joe Kane). The problem appears on s-video as well.

On a pc, the image of the character sheet was recorded using the firewire connection. The avi file was dropped into windvd and the computer screen was zoomed. Any pixels that had a color like magenta, cyan or red were difficult to find (although they do exist to some degree). Not to the extent of the Sony, which looked like it was attempting to smoothly anti-alias the characters. Thus another mystery solved. Thanks to the members who pointed me in this direction.

Shai Levy · November 4th, 2003, 10:58 AM

there was a great article by Adam Willt in www.dv.com

one should register to read.. but it worth it. this article by Adam explains the thing.

my url to the article doesnt work.. but it must be there in their archive

Gints Klimanis · November 4th, 2003, 02:37 PM

Shai, any hints as to the name or location of the article?
I tried searches on dv.com for various keywords: color bleed,
purple fringe,

Also, how are you all saving articles? I've been saving articles
with embedded pictures as Windows .mht files. Generally, this works, and leaves me with a single file instead of the HTML file and a folder of all sorts of stuff. Now, it seems as if Windows Internet Explorer fails to save some articles when flashy ads are present. Any suggestions?

Gints Klimanis · November 4th, 2003, 02:42 PM

Never mind. I guess it was Adam Wilt's article on
Panasonic AG-DVX100 .

http://www.dv.com/features/features_item.jhtml?category=Archive&LookupId=/xml/feature/2003/wilt0203

Shai Levy · November 5th, 2003, 07:49 AM

found it!!!

http://www.dv.com/columns/columns_it...uestid=1317583

and i found now that it refers to the 4:2:0 in PAL, but you can understand something about the color sampling limitations

Andre De Clercq · November 5th, 2003, 01:54 PM

This is another illustration of Wilt's frustration w.r.t. the better 4:2:0 sampling used for PAL, DVD...Has nothing to do with the color fringing problem in this thread. It somewhat suggests that the people who decided several years ago for 4:2:0 didn't know what interlace effects were. There are more people in the world that know what subjective image quality, and its trade-offs are. The only slight advantage with 4:1:1 is(was!) somewhat easier post processing approaches.

November 2nd, 2003, 09:35 AM	#1
Marc Young Regular Crew Join Date: Oct 2003 Location: Honolulu, HI Posts: 112	Color Fringing: What's the Cause? Given the sensitivity of the DVX100 to manual focusing, I created a diy focusing chart in powerpoint, consisting of True Type fonts ranging from 36 - 72 pt in size (0.5 to 1"). The paper was slightly tilted and the camera was focused on it from a distance of around 4 - 5 feet. The firewire output was connected to a panasonic dvd recorder and the image was displayed on a progressive sony 36" tv using component video connections. Please explain the red and cyan overhang on the black lettering. How is this possible on a b&w image? Even if the color is undersampled under 4:1:1 coding, shouldn't it appear gray? Are the CCD's aligned properly or is there distortion in the lens? Anyone else have an eye chart or other pattern to test this on? I did the same experiment with a cheap JVC single chip camera, and did not see the same thing. Admittedly, the Panasonic looked much smoother (no ringing). This experiment also demonstrated that during panning, 24p is better than 30p because most deinterlacers on consumer tv's are not geared to turn the signal into 60p. They still think they're being fed 60i. Also, 24p advanced is not interpreted correctly by Panasonic's dvd recorder. It displays correctly with a static image, but it twitters slightly with a moving image. Nothing is more merciless than an image with sharp edges, like a whole bunch of letters.

November 2nd, 2003, 10:31 AM	#2
Boyd Ostroff Wrangler Join Date: Dec 2002 Location: Mays Landing, NJ Posts: 11,800	DVD recorders <<<-- Originally posted by Marc Young : The firewire output was connected to a panasonic dvd recorder and the image was displayed on a progressive sony 36" tv using component video connections.. -->>> Just curious because this was discussed before... when using the DVD recorder to transcode from D25 to component, isn't the image undergoing MPEG compression first? Maybe this is a factor? Also (although a bit off topic) which model Panasonic recorder to you have and how do you like it? I've been thinking about getting one myself.

November 2nd, 2003, 01:19 PM	#6
Stephen van Vuuren Space Hipster Join Date: Oct 2001 Location: Greensboro, NC Posts: 1,508	You can purple fringing on the DVX100 from overexposing highlighted edges in cinegamma. __________________ stephen v2 www.insaturnsrings.com

November 2nd, 2003, 11:22 AM	#3
Marc Young Regular Crew Join Date: Oct 2003 Location: Honolulu, HI Posts: 112	My Panasonic dvd recorder models are the DMR-HS2 and the DMR-E100. Both have firewire inputs. On live display, there is no mpeg-2 encoding (this is only done by the original Philips recorder). The latest Panasonic model outputs a very good picture, and the mpeg-2 encoding (recording) is near the top of the field (surpassed only by Sony or Pioneer, and the margin, if any, is very slim). It is an alternative to a pc for getting your mini dv camera output to a dvd-r. We all know about the problems with choosing the right codecs and mpeg encoders on a pc. Or standardizing on one nle. That is a huge can of worms. Pick the wrong one and your end product will look worse than any standalone recorder. In fact, I'm starting to reach the conclusion that unless you pick the top mpeg-2 encoders on a pc, it is better to print your nle output to tape, and let the standalone recorder handle the dvd creation. The JVC dvhs recorders are another alternative for people who want to convert firewire, or record their mini-dv tapes on another format, for viewing on a large screen tv. Another display option is one of the marantz digital linear projectors and one rear-screener (forget the model #) which have firewire inputs. Before you edit any recording, make sure it looks good enough to author on a dvd. That takes a big tv imho. Nothing exposes the circle of confusion like a big screen. Otherwise, just go out and re-shoot.

November 2nd, 2003, 12:53 PM	#4
Andre De Clercq Major Player Join Date: Sep 2002 Location: Belgium Posts: 804	>On live display, there is no MPEG-2 encoding... Where did you get this information. As far as I know there is no consumer equipment on the market which outputs component video straight from a DV input.

November 2nd, 2003, 01:15 PM	#5
Andre De Clercq Major Player Join Date: Sep 2002 Location: Belgium Posts: 804	On the fringing...I would first turn the saturation of the Sony to zero. If you still see "fringing" there is a problem with the set: convergence and/or focus tracking is not OK. (Remember that unsharp edges from yr old cam hide those problems). If you then don't have fringes I would turn up sat again and enter luma only straight from the cam into the Sony Y/C input (special cable). If you still see fringes yr comb or notchfilter in the Sony is not optimal. If OK I would enter Y signals from the Panasonic into the Sony....Also remember, if yr component interconnection is not short or unmatched (in length)you can get fringes.

November 3rd, 2003, 10:26 AM	#7
Marc Young Regular Crew Join Date: Oct 2003 Location: Honolulu, HI Posts: 112	<<<-- Originally posted by Andre De Clercq : >On live display, there is no MPEG-2 encoding... Where did you get this information. As far as I know there is no consumer equipment on the market which outputs component video straight from a DV input. -->>> It gets mighty close, meaning that on a static image, it looks reasonably good compared to my PC monitor. Whether consumer dvd recorders save a firewire transmission, or for that matter, an s-video signal, in their field/frame buffer as 4:2:0, 4:1:1, 4:2:2 or 4:4:4 is not known to me since I don't design this equipment, nor do I have the test equipment to verify that the component video is an exact analog representation of the firewire data. It just seems logical that the Panasonic recorder would synthesize the 720x480 frame buffer using two field buffers, without undersampling the chroma more than it is in the incoming signal. After all, when the recorder is not recording, it is merely transcoding the digital signal using simple math. In fact, it is more likely they resample/interpolate the incoming data, so the color resolution is finer than what came in. Unlike Philips, the feed-through image to the tv does not degrade based on recording mode (XP, SP, LP, EP). Because in the Panasonic, you're not recording yet, just watching the feed. Also, even I can see the difference between the firewire signal through the recorder and out on component (progressive) video vs. s-video from the camera directly into the tv, so the E100 must be improving something. Now clearly a consumer tv does not offer 27 MHz bandwidth on non-HD component video, but then again, neither does this camera. It would be interesting to A/B a professional mini-dv tape deck/editing station with component video out vs. the dmr-e100, but I doubt the pro-stuff will do 3:2 pulldown (at least not cheaply). Try doing something as "straightforward" as a frame grab on pro equipment, and you can be sure they'll charge you a $5000 entry price. Anyhow, since I seem to have gotten off on a tangent, I'll get back to the color fringing anomaly. It may be the Sony TV and its DRC (digital reality creation) or some other signal processing quirk. Possibly even the color matrix. When in doubt, blame the color matrix (quote Joe Kane). The problem appears on s-video as well. On a pc, the image of the character sheet was recorded using the firewire connection. The avi file was dropped into windvd and the computer screen was zoomed. Any pixels that had a color like magenta, cyan or red were difficult to find (although they do exist to some degree). Not to the extent of the Sony, which looked like it was attempting to smoothly anti-alias the characters. Thus another mystery solved. Thanks to the members who pointed me in this direction.

November 4th, 2003, 10:58 AM	#8
Shai Levy Regular Crew Join Date: Feb 2003 Location: Berlin, Germany Posts: 31	there was a great article by Adam Willt in www.dv.com one should register to read.. but it worth it. this article by Adam explains the thing. my url to the article doesnt work.. but it must be there in their archive

November 4th, 2003, 02:37 PM	#9
Gints Klimanis Inner Circle Join Date: Jun 2003 Location: San Jose, CA Posts: 2,222	Shai, any hints as to the name or location of the article? I tried searches on dv.com for various keywords: color bleed, purple fringe, Also, how are you all saving articles? I've been saving articles with embedded pictures as Windows .mht files. Generally, this works, and leaves me with a single file instead of the HTML file and a folder of all sorts of stuff. Now, it seems as if Windows Internet Explorer fails to save some articles when flashy ads are present. Any suggestions?

November 4th, 2003, 02:42 PM	#10
Gints Klimanis Inner Circle Join Date: Jun 2003 Location: San Jose, CA Posts: 2,222	Never mind. I guess it was Adam Wilt's article on Panasonic AG-DVX100 . http://www.dv.com/features/features_item.jhtml?category=Archive&LookupId=/xml/feature/2003/wilt0203

November 5th, 2003, 07:49 AM	#11
Shai Levy Regular Crew Join Date: Feb 2003 Location: Berlin, Germany Posts: 31	found it!!! http://www.dv.com/columns/columns_it...uestid=1317583 and i found now that it refers to the 4:2:0 in PAL, but you can understand something about the color sampling limitations

November 5th, 2003, 01:54 PM	#12
Andre De Clercq Major Player Join Date: Sep 2002 Location: Belgium Posts: 804	This is another illustration of Wilt's frustration w.r.t. the better 4:2:0 sampling used for PAL, DVD...Has nothing to do with the color fringing problem in this thread. It somewhat suggests that the people who decided several years ago for 4:2:0 didn't know what interlace effects were. There are more people in the world that know what subjective image quality, and its trade-offs are. The only slight advantage with 4:1:1 is(was!) somewhat easier post processing approaches.