It is not the codec, it is the onboard computer!

I can only think of two cases of Panasonic Broadcast ever adding features, and that was the flash band compensation for the HPX300 and the addition of 24p to the HPX2000 (if it was the 2000 I'm thinking of).
Flash band compensation looks like a fix to a well-known bug -- or should I say gotcha? -- to me ;-)
 
It's a statement based on observing all these manufacturers for years, as well as taking Canon's statement of "firmware isn't for adding features, it's for fixing bugs." They reserve new features for new models. That's what they said they do, so if you have a problem with it, you'll have to take it up with them.

Do you *really* want to do this here? We can, if you want...

5D
Aliasing
Costs 3x as much, or no kit lens included, take your pick
Chroma moire
Major rolling shutter issues
Can't take PL-mount lenses
Can't use c-mount lenses
No flip-out LCD or swivel screen
Films in a non-video-standard 30.000 fps, instead of video-standard 29.97
No 24p
No 60p
No autofocus when shooting video
vertical banding
black dot noise (source)

GH1
Mud occurrences (which is the same as Compression artefacts on detail)
Motion artefacts -- what motion artifacts?
Fixed pattern noise



What kind of statement is that?

Well, I respect your experience with pro video cameras, but we're talking about a different market and a different era here with the GH1 & 5D. It's a slightly different situation for the consumer market and I can see why Canon felt the need to add the M mode to EOS movies via a firmware update.

GH1 mud occurrences isn't the same as compression artefacts on detail. One of the things I found whilst shooting without the kit lens in the Tiergarten in Berlin was that with highly detailed scenes and the singer's movement, the detail would smudge around the moving areas of the image and the overall frame detail would dramatically reduce. The bitrate of the GH1's footage is variable on the fly, and hovers around 14mbps in 17mbps mode.

Mud is different in that it represents a complete and dramatic failure of the image for a few frames before the camera rights itself.
 
Last edited:
Wait -- you got that off of canonrumors.com, and presented it as fact?

Bro... seriously?

Yes I was making my point -

"I can see why you think it IS misinformation though, because it doesn't come from Canon, it comes from 'sources' close to Canon or rumour sites, like the following:"

And my point was that although it may or may not be FACT, you can correlate real life experience with the camera to statements you read on the rumours site and make your own mind up which is closest to what you observe, and therefore closest to the truth.

It helps to do this because you build up knowledge about how these things work. I personally think the rumour sites have been proved right ahead of time quite a few times. The first I heard of the GF1 was on 43rumors.com - you can't always dismiss these things out of hand. They have good sources. There is no smoke without fire. Canon may say they don't bin pixels, but that doesn't correlate too well with aliasing does it?

I am positively sure I watched Tim-something or other of Canon USA stand up and actually talk about how downsampling is done on the 5D Mark II but I don't remember exactly what he said in enough detail to be sure. But it doesn't take a undercover mole to workout that the current generation of onboard CPU isn't quite ready for the demands of video. Of course they skip lines and subsample pixels in groups of 6, etc. The current chip is only designed from the ground up to feed a live view display. A true video chip needs to be designed from the ground up with video in mind like the next DIGIC processor.

Panasonic and Canon - it would be good for them to be more open with regards technical information. Okay, no need to give away trade secrets, but a few more engineer interviews with the trustworthy camera press and journalists wouldn't go amiss, and it'd help to showcase their brilliance. The interview with the chief Leica lens design guy on Luminous Landscapes recently was interesting for it's uniqueness, if not terribly well put together. A factory tour of the 1D Mark IV design office? Yes please :)
 
Last edited:
gh1 - amazing technology

gh1 - amazing technology

speculation ? misinformation ? lets try conjecture -

(conjecture |kənˈjek ch ər|
noun
an opinion or conclusion formed on the basis of incomplete information - )


I really appreciate all that I've learned from this forum - thank you everyone (conjecture an all). :beer:
 
And my point was that although it may or may not be FACT, you can correlate real life experience with the camera to statements you read on the rumours site and make your own mind up which is closest to what you observe, and therefore closest to the truth.

It helps to do this because you build up knowledge about how these things work. I personally think the rumour sites have been proved right ahead of time quite a few times. The first I heard of the GF1 was on 43rumors.com - you can't always dismiss these things out of hand. They have good sources. There is no smoke without fire. Canon may say they don't bin pixels, but that doesn't correlate too well with aliasing does it?


You lost me around albuquerque. :huh:
 
GH1 mud occurrences isn't the same as compression artefacts on detail.
I would contend that it's the exact same thing, the only difference is the degree. Same cause, same symptoms. It's the same thing that we could observe on HDV when HDV was pushed too far, it's a long-GoP codec without enough efficiency to handle the changes that are occurring.

Man, I sure wish they would have just extended the MJPG mode to handle 1080/24p and 720/24p and 720/60p. I'm fine with needing a class 10 card, just give me some intraframe goodness without long-GoP complications!
 
"I can see why you think it IS misinformation though, because it doesn't come from Canon, it comes from 'sources' close to Canon or rumour sites, like the following:"
Well, I guess that comes down to "who do you want to believe?" I would take the word of Canon's product manager for all digital imaging products, who has obviously talked to the engineers and asked them this very question and gotten the answer he got, instead of "sources close to Canon" or rumor sites.

I will freely admit that sometimes official channels don't let the whole story be known, and that sometimes stuff leaks out, yes that happens. But the way Tim Smith answered the question, it's too reminiscent of the times when I've asked the engineers at NAB the questions I've asked, and they've answered him. He didn't say "we don't pixel skip. Next question?" He said "we don't pixel-skip, we bin pixels in a group of six", etc. He attempted to give a factual answer, not just cover-up an uncomfortable question.

And binning a color sensor is not an easy task. The aliasing that's being seen is entirely possible due to binning and having an extremely inadequate low-pass filter (which, by virtue of it being a stills camera first and foremost, the low-pass filter for video is inherently going to be way inadequate.) You can't optimize one system for both stills and video performance, one or the other is going to have to give.

And my point was that although it may or may not be FACT, you can correlate real life experience with the camera to statements you read on the rumours site and make your own mind up which is closest to what you observe, and therefore closest to the truth.
Ultimately we all have to choose what we're going to believe, I guess, but I prefer to dig in and find out what's actually going on. I've given this example a million times, but commonly-accepted wisdom isn't very frequently right. Bogart never said "play it again, sam", and the line "If you build it, they will come" was never said in "Field of Dreams", regardless of how many people swear that it was.

Hence why I got a little twisted in the knickers about the whole conjecture/speculation/fact thing...

Canon may say they don't bin pixels, but that doesn't correlate too well with aliasing does it?
Yes, actually, it does. Because in the same video, Tim said that the hardware in the 5D is not designed for video. The video feature in the 5D Mark II was never intended to go any further than webcasts. He admits that the hardware that does the scaling is in no way optimized for full-res HD video, and that in future (more-expensive) products they'll have much better hardware in there. The aliasing is coming primarily from the lousy downconverter. He said that the hardware that's in there now, was adequate to the market that they build the 5DMKII for: photojournalists who are going to post footage to the web. He said that's why there's no 25P in there for the EU users either -- because on the web, it doesn't matter, 30fps is okay. And that's also why it's not 29.97, it's 30.00 -- because for the web, it doesn't matter. Now, don't take this wrong -- I am NOT saying that the 5DMkII is only capable of delivering webcam-caliber video! Obviously it's infinitely better than that. But what I am saying is: the issues that we observe are there, they're caused by inadequate hardware, and it's because the camera was never designed to compete in that arena. The fact that it does is a happy accident.

I am positively sure I watched Tim-something or other of Canon USA stand up and actually talk about how downsampling is done on the 5D Mark II but I don't remember exactly what he said in enough detail to be sure. But it doesn't take a undercover mole to workout that the current generation of onboard CPU isn't quite ready for the demands of video. Of course they skip lines and subsample pixels in groups of 6, etc. The current chip is only designed from the ground up to feed a live view display. A true video chip needs to be designed from the ground up with video in mind like the next DIGIC processor.

A factory tour of the 1D Mark IV design office? Yes please :)
hat would indeed be sweet, because the way Tim Smith was talking in that video, it sounds like Canon has every intention of building a killer VIDEO camera around a sensor as big as the 5D Mark II's. And I would expect the 1D Mark IV to be that product.
 
Last edited:
Flash band compensation looks like a fix to a well-known bug -- or should I say gotcha? -- to me ;-)

This firmware update was purposed to address rolling shutter issues attendant to CMOS.

The firmware provided an attenuation of the problem; that quacks like a feature to me.
 
In what Panasonic devices can we see a better AVCHD codec at work, and in any such device is it handling the detail from the optics that the GH1 is encoding? And is any such device subject to the constraints of size that the GH1 imposes on their engineers, which must affect processor capabillty and heat dissipation issues and power consumption issues?

It seems to me that something akin to miracles are being demanded a lot of the time.

The constant search for mud seems to me to risk other aspects of the picture quality being obtained to be overlooked - for instance, while you (a generic you) are looking for mud in your images, are you sure you're not overlooking focus issues or whatever? Sometimes I've looked at footage posted as samples of mud, and wondered why the main subject is poorly focussed, but nobody is mentioning that - and it's the main subject that the eye is supposed to dwell on.
 
Last edited:
This firmware update was purposed to address rolling shutter issues attendant to CMOS.
That's what I meant. "rolling shutter issues" == gotchas.

A "gotcha" is:
- a feature or function which works as advertised, but not as expected;
- a strange quirk, but not exactly a bug.
 
just a thought

just a thought

im kind of new here, so please bare with me, as I may not have all the knowledge that a lot of you seem to have.

I have been shooting with the GH1 for a couple weeks now, and have made some discoveries that perhaps have been mentioned here, though I have yet to read about here. its concerning the whole mud issue. it has to do with what codecs and players are installed on the system (computer) that are working with these avchd files.

i purchased a lumix zs3 while waiting for the gh1 to become available as i wanted something smaller than a sony z1u hdv cam to shoot hd with. after shooting a bunch of video with this little zs3, i played it back only to be a bit shocked that i was noticing mud even on very steady shots. so i also downloaded some clips that others shot with the gh1 to see if the same issues persisted - and sure enough, i saw it with these too. i played the clips in vlc, both the newer 1.0 and an older .8.x.x version, i played them in media player 11, i tried importing into premiere cs2 (no luck there), and attempted to play them in a few other generic players as well - sometimes getting no image at all. one thing i did notice, was that the programs that did play the files (vlc, windows media player, and gom player) didnt display them in the same manner. there was mud in all of them - but not the same mud. i tried playing the files on a few other computers with less luck. i installed and uninstalled ffdshow and various versions of it.

finally i installed a program which i hadnt used in a while - nero showtime. it was like magic! the files played and no mud! after installing, went back to some of the other programs such as vlc - still mud. after having some other computer issues unrelated to video work, i decided to reinstall my windows xp pro sp2. after doing this, i was back to windows media player 9 (did not upgrade) - switched it to skin mode to look like media player classic to use less computer resources- and the files played perfectly here, too! (and with better audio sync than nero vision - which seems to have about a one second delay.)

now i wanted to edit these files. got the trial of premiere cs4 on my home laptop (my job hadnt yet updated our cs3 to cs 4 yet) imported the files and yuck - the mud was back - but only until i attempted something that i did not think would work on my 1.5 GHz centrino duo laptop (granted it has 4 gigs (3 usuable) of ram) i switched the preview mode from 'automatic' to 'high' - which with hdv footage would just eat up all the resources making it bog down and too slow to be productive. lo and behold, no mud! and i can scrub though 1080 or 720 footage pretty smoothly - even from an external 5400 rpm drive!

granted it takes a pc more processing power to edit gop footage rather than i-frame footage, the bitrate is technically lower than sd dv footage.

another thing i will note, is that i did manage to lock up premiere a few times editing this footage until i installed the 4.1 update that just came out this summer. now its smooth as can be and there are 720 30p avchd presets amongst many others.

every once in a while, i get some footage that will play in showtime, though not media player - tends to be any footage that is extremely dark - it just stutters for whatever reason. in showtime, there is an option to display information onscreen while playing video - its a readout of the bitrate of the video that is playing. the footage that stutters is displaying a bitrate of 3-8 Mbits/sec. cant figure this out - cause the footage the reads an average of 12-14 Mbits/ sec plays really smooth.

anyways, i do still have ffdshow installed, though if i have it enabled as the decoder for h.264 footage, i get mud regardless of what program i play it in. with the install of premiere cs4 came the main concept h.264 codec, which is what media player says it is using when i look at the properties for any of this footage. i also export using this codec in premiere.

so my thought is this
- perhaps the camera is doing a fine job recording the images and turning it into data. maybe the computers we view them on are simply having trouble turning this data back into images. due either to codec/software issues or hardware issues (might add that i have turned off any unnecessary processes in windows such as any graphical enhancements, animations, etc - looks more like win 95, than xp to free up processing power)

a few more things before i hit the sack - we just got some new panasonic bluray players at my job so inserted the card into the sd slot on it - played some test footage perfectly on our 1920x1080 samsung 50" monitors

also, i have been shooting almost entirely in manual and only sometimes in ap or shutter priority, which probably is less demanding of the cams cpus. full auto plus auto focus, in theory, could cause the camera's encoding to suffer - though as said - its all speculation. oh and have not been able to get avchd footage to play on any single-core computers for more than a second or two.

as for why 24p wrapped up into 30i? makes mixing the footage a lot simpler with other footage. combined 720 60p and 1080 24p into my test footage that i posted a few days ago and its hard to distinguish what each shot was shot with. a few years ago shot some stuff with a dvx 100a or b in what i believe was true 24p and had tons of issues mixing with footage shot at 30p and 60i.

anyways, good luck to all and look forward to learning more and making the most with what we have :)
 
im kind of new here, so please bare with me, as I may not have all the knowledge that a lot of you seem to have.

I have been shooting with the GH1 for a couple weeks now, and have made some discoveries that perhaps have been mentioned here, though I have yet to read about here. its concerning the whole mud issue. it has to do with what codecs and players are installed on the system (computer) that are working with these avchd files.

i purchased a lumix zs3 while waiting for the gh1 to become available as i wanted something smaller than a sony z1u hdv cam to shoot hd with. after shooting a bunch of video with this little zs3, i played it back only to be a bit shocked that i was noticing mud even on very steady shots. so i also downloaded some clips that others shot with the gh1 to see if the same issues persisted - and sure enough, i saw it with these too. i played the clips in vlc, both the newer 1.0 and an older .8.x.x version, i played them in media player 11, i tried importing into premiere cs2 (no luck there), and attempted to play them in a few other generic players as well - sometimes getting no image at all. one thing i did notice, was that the programs that did play the files (vlc, windows media player, and gom player) didnt display them in the same manner. there was mud in all of them - but not the same mud. i tried playing the files on a few other computers with less luck. i installed and uninstalled ffdshow and various versions of it.

finally i installed a program which i hadnt used in a while - nero showtime. it was like magic! the files played and no mud! after installing, went back to some of the other programs such as vlc - still mud. after having some other computer issues unrelated to video work, i decided to reinstall my windows xp pro sp2. after doing this, i was back to windows media player 9 (did not upgrade) - switched it to skin mode to look like media player classic to use less computer resources- and the files played perfectly here, too! (and with better audio sync than nero vision - which seems to have about a one second delay.)

now i wanted to edit these files. got the trial of premiere cs4 on my home laptop (my job hadnt yet updated our cs3 to cs 4 yet) imported the files and yuck - the mud was back - but only until i attempted something that i did not think would work on my 1.5 GHz centrino duo laptop (granted it has 4 gigs (3 usuable) of ram) i switched the preview mode from 'automatic' to 'high' - which with hdv footage would just eat up all the resources making it bog down and too slow to be productive. lo and behold, no mud! and i can scrub though 1080 or 720 footage pretty smoothly - even from an external 5400 rpm drive!

granted it takes a pc more processing power to edit gop footage rather than i-frame footage, the bitrate is technically lower than sd dv footage.

another thing i will note, is that i did manage to lock up premiere a few times editing this footage until i installed the 4.1 update that just came out this summer. now its smooth as can be and there are 720 30p avchd presets amongst many others.

every once in a while, i get some footage that will play in showtime, though not media player - tends to be any footage that is extremely dark - it just stutters for whatever reason. in showtime, there is an option to display information onscreen while playing video - its a readout of the bitrate of the video that is playing. the footage that stutters is displaying a bitrate of 3-8 Mbits/sec. cant figure this out - cause the footage the reads an average of 12-14 Mbits/ sec plays really smooth.

anyways, i do still have ffdshow installed, though if i have it enabled as the decoder for h.264 footage, i get mud regardless of what program i play it in. with the install of premiere cs4 came the main concept h.264 codec, which is what media player says it is using when i look at the properties for any of this footage. i also export using this codec in premiere.

so my thought is this
- perhaps the camera is doing a fine job recording the images and turning it into data. maybe the computers we view them on are simply having trouble turning this data back into images. due either to codec/software issues or hardware issues (might add that i have turned off any unnecessary processes in windows such as any graphical enhancements, animations, etc - looks more like win 95, than xp to free up processing power)

a few more things before i hit the sack - we just got some new panasonic bluray players at my job so inserted the card into the sd slot on it - played some test footage perfectly on our 1920x1080 samsung 50" monitors

also, i have been shooting almost entirely in manual and only sometimes in ap or shutter priority, which probably is less demanding of the cams cpus. full auto plus auto focus, in theory, could cause the camera's encoding to suffer - though as said - its all speculation. oh and have not been able to get avchd footage to play on any single-core computers for more than a second or two.

as for why 24p wrapped up into 30i? makes mixing the footage a lot simpler with other footage. combined 720 60p and 1080 24p into my test footage that i posted a few days ago and its hard to distinguish what each shot was shot with. a few years ago shot some stuff with a dvx 100a or b in what i believe was true 24p and had tons of issues mixing with footage shot at 30p and 60i.

anyways, good luck to all and look forward to learning more and making the most with what we have :)

Pretty sure what your seeing isn't mud but codec issues with your playing device.
 
Now I want another speculation about 24p-over-60i. Why it was done, why not native 24p. Presently I believe that Panasonic wanted to artificially separate pro digital cinema and amateur camera markets, but maybe there is a better explanation, for example "most NLEs including Panasonic's own video editor do not support native 24p". It this true? Was it the reason? Any takers?
The GH1 is a consumer product created by Panasonic's consumer division. The higher-ups didn't need to tell the consumer division not to trample on the sales of the broadcast division. The consumer division knows what it has to do to gear a product toward the consumer market. There doesn't need to be any ulterior motive. It's the same reason why consumer camcorders have fewer manual features; and when they do have such features, the features are buried within the menus.

The GH1's designers most likely decided to embed 24p within 60i because that's just the way that it's done for consumer camcorders. That's the way that Canon's HV-series does it; even the prosumer Sony HDV camcorders do it that way. Now the reason why pulldown is standard practice for consumer camcorders is presumably because the designers expect consumers to connect their camcorders directly to their HDTVs. Not all HDTVs accept a non-interlaced signal. It's likely the same reason why the GH1 doesn't do live video-out while recording: It's because they expect that most consumers have no use for such a feature. Why is that so hard to understand?
 
Pretty sure what your seeing isn't mud but codec issues with your playing device.
I'd say there are two phenomena that tend to get called 'mud' - first, the severe problem seen with whip pans where, if you freeze a frame, you see something like an impressionist painting. Second, the problem which seems to generally affect AVCHD, namely foliage ceasing to be textured but becoming a kind of green blotch - and that can be seen without any movement being involved. But part of the mud problem is getting agreement on what the problem actually looks like.
 
> The GH1's designers most likely decided to embed 24p within 60i
> because that's just the way that it's done for consumer camcorders.

You call this an explanation?

> That's the way that Canon's HV-series does it

Not the HV40.

> even the prosumer Sony HDV camcorders do it that way

Sony has always dragged its feet implementing progressive. Sony had been an avid proponent of interlaced scanning because they had the technology from long ago. Even with pro models Sony was two years late after Canon implementing native progressive in HDV cams.

> Now the reason why pulldown is standard practice for consumer camcorders
> is presumably because the designers expect consumers to connect their camcorders
> directly to their HDTVs. Not all HDTVs accept a non-interlaced signal.

What about DVD players adding pulldown in realtime to 24p-mastered DVDs? What about Sony and Canon cameras adding pulldown in realtime when feeding 24p through HD-SDI or component?

> Why is that so hard to understand?

It would be easier to understand if Panasonic had not produced consumer AVCHD models with native 24p in the past.
 
after shooting a bunch of video with this little zs3, i played it back only to be a bit shocked that i was noticing mud even on very steady shots.

Well, I'm new here, too. But it's my understanding that the AVCHD-Lite codec used in the ZS3 is different from the (full?) AVCHD codec used in the GH1.

Mark
 
Commanderspike said:
GH1 mud occurrences isn't the same as compression artefacts on detail.

I would contend that it's the exact same thing, the only difference is the degree. Same cause, same symptoms. It's the same thing that we could observe on HDV when HDV was pushed too far, it's a long-GoP codec without enough efficiency to handle the changes that are occurring.

Man, I sure wish they would have just extended the MJPG mode to handle 1080/24p and 720/24p and 720/60p. I'm fine with needing a class 10 card, just give me some intraframe goodness without long-GoP complications!

Nope, I am talking about two separate things here, honestly. The mud is a temporary break-down of the image, and compression artefacts are a constant reduction of detail if the camera can't handle the amount of detail needed in the scene. It's variable bit rate and it's a low bit rate at that.

Mud is separate to that, and probably caused by the lack of B-frames due to the hardware not being powerful enough.
 
In what Panasonic devices can we see a better AVCHD codec at work, and in any such device is it handling the detail from the optics that the GH1 is encoding?
The closest to that would be the HMC40, which uses three 1920x1080 chips and has the full PH mode AVCHD encoder. Or, you could use the HMR10 which is a standalone AVCHD "deck", and feed it SDI input from something that generates tons of detail. I think a great test for the robustness of AVCHD would be to pipe the output of an HPX3000/HPX3700 or F23/F35 through SDI to the HMR10, and see what the codec can do with that signal.
 
Back
Top