Panning 'mud', reasons why and firmware update
- Thread starter commanderspike
- Start date
anthonybsd
Active member
This is all quite perplexing. 17 megabits = ~ 2 megabytes per second, i.e any cheapo class 4 and above card should be able to more than keep up with this camera. Something seems fishy here, folks.
John Caballero
Veteran
You have to get the fastest card you can get. That goes without question. This cameras are very demanding. You are going to compromise your recordings in order to save a few dollars? It makes absolutely no sense. You can get away with it for the stills but not video. If after you get the fastest SD cards available problems persist thats another matter, but you gave it the chance to capture to the best quality environment. Kholi already said that he saw improvements with faster cards and I have no doubt that they are indeed a primary part of the equation for better capturing.
Last edited:
admactanium
Well-known member
When I get mine later this week I'll run some tests and see. I bought a 16Gb Extreme III 30Mbs card and I have a number of other off-brand cards to test it with.
Barry_Green
Moderator
I'm running some tests on a loaner unit right now. Sharply focused on a computer monitor with plenty of fine detail on the screen, then panning slowly, and quickly, tilting up and down, and rotating. I am using 1/250 shutter speed to hide the natural motion blur and let us more clearly see what's happening in the frame. I also used a cheapo card (class 4) and a SanDisk Extreme III Class 6 card to see if that makes any difference.
First observation -- noise reduction seems to make a difference. With the noise reduction at -2, the image can go pretty much totally to hell, but what you can see is basically still sharp amongst the compression artifacts.
But, if you put noise reduction at +2, then the image gets blurry as the compression falls apart. It still gets nasty, but it's a different nasty. Smudgy and blurry everywhere.
I'm also thinking there's a big effect going on of people not properly removing pulldown. I pulled the footage into Vegas to extract the still frames, and I couldn't find any interlace artifacts (presumably impossible on 1080/24p footage); I brought the footage into Premiere Pro CS4 and - OH YEAH - there's the interlace artifacts. Which tells me that CS4 is showing the footage more accurately and, yeah, I'm not gonna bother with Vegas for AVCHD anymore.
But if you didn't remove the pulldown properly, as happened on my first run through Vegas, you can see massive "smudging" happening on vertical tilts, but it's not really in the footage! It's an effect of the image getting de-interlaced for some reason, so you're seeing both fields at that point. If you strip out the pulldown properly (as I'm doing, I'm only checking the three "whole" frames and ignoring the split frames) then that vertical smudge is gone. Sure, the picture still goes to hell in a dragster-powered handbasket, but at least it's not like what's been reported.
Note: these files that I'm viewing were converted by Cineform Neo Scene and had 24p pulldown removed. I also am going to compare them from the original frames, manually scanning for pulldown-free frames to find comparable representative frames.
In summary: card quality appears to have nothing to do with it. If you've got a class 4 card, formatted in the camera, you're getting the footage as the camera makes it. You can't blame this stuff on a sub-standard card. If your card was too slow to support the mode, the camera would warn you and wouldn't let you record.
The artifacts pretty much look like HDV artifacts all over the place, and reminds me once again why I generally *hate* long-GoP codecs. Grrr. The image degradation is on par with the worst I've seen from HDV in the worst cases. But I suspect HDV might have done better under the given test circumstances. This implementation of AVCHD is not nearly as robust as the implementation in the HMC150, that's for sure. Screw long-GoP, I'm shooting AVC-Intra from now on... but -- okay, okay, long-GoP has its place, I know. Bah. The good news here is that nobody's going to be still-framing your video, so we're making a mountain out of a medium-sized mole hill, but -- if you're planning on doing some shakycam footage and then compositing in some special effects... yeah, get an HPX170 for those shots. Put on the Letus if you have to to match the DOF.
Okay, so enough with pixel-peeping; I'm now putting the card back in the camera and playing the footage back on a 67" 1080p-native DLP TV. And there, it looks a hundred times better. The glitched frames are there, and I can spot 'em, but -- I'm me, right? I know what I'm looking for. I had someone else watch it and said "do you notice anything about this footage?" Answer: "nope. Looks fine. Annoying, but fine." (keep in mind it's basically horrific shakycam footage, flinging the camera around like a spastic dweeb with fire ants in his pants, making "Cloverfield" look like it was shot on a Steadicam).
Tomorrow I'll try a wide-angle grass 'n' trees shot.
First observation -- noise reduction seems to make a difference. With the noise reduction at -2, the image can go pretty much totally to hell, but what you can see is basically still sharp amongst the compression artifacts.
But, if you put noise reduction at +2, then the image gets blurry as the compression falls apart. It still gets nasty, but it's a different nasty. Smudgy and blurry everywhere.
I'm also thinking there's a big effect going on of people not properly removing pulldown. I pulled the footage into Vegas to extract the still frames, and I couldn't find any interlace artifacts (presumably impossible on 1080/24p footage); I brought the footage into Premiere Pro CS4 and - OH YEAH - there's the interlace artifacts. Which tells me that CS4 is showing the footage more accurately and, yeah, I'm not gonna bother with Vegas for AVCHD anymore.
But if you didn't remove the pulldown properly, as happened on my first run through Vegas, you can see massive "smudging" happening on vertical tilts, but it's not really in the footage! It's an effect of the image getting de-interlaced for some reason, so you're seeing both fields at that point. If you strip out the pulldown properly (as I'm doing, I'm only checking the three "whole" frames and ignoring the split frames) then that vertical smudge is gone. Sure, the picture still goes to hell in a dragster-powered handbasket, but at least it's not like what's been reported.
Note: these files that I'm viewing were converted by Cineform Neo Scene and had 24p pulldown removed. I also am going to compare them from the original frames, manually scanning for pulldown-free frames to find comparable representative frames.
In summary: card quality appears to have nothing to do with it. If you've got a class 4 card, formatted in the camera, you're getting the footage as the camera makes it. You can't blame this stuff on a sub-standard card. If your card was too slow to support the mode, the camera would warn you and wouldn't let you record.
The artifacts pretty much look like HDV artifacts all over the place, and reminds me once again why I generally *hate* long-GoP codecs. Grrr. The image degradation is on par with the worst I've seen from HDV in the worst cases. But I suspect HDV might have done better under the given test circumstances. This implementation of AVCHD is not nearly as robust as the implementation in the HMC150, that's for sure. Screw long-GoP, I'm shooting AVC-Intra from now on... but -- okay, okay, long-GoP has its place, I know. Bah. The good news here is that nobody's going to be still-framing your video, so we're making a mountain out of a medium-sized mole hill, but -- if you're planning on doing some shakycam footage and then compositing in some special effects... yeah, get an HPX170 for those shots. Put on the Letus if you have to to match the DOF.
Okay, so enough with pixel-peeping; I'm now putting the card back in the camera and playing the footage back on a 67" 1080p-native DLP TV. And there, it looks a hundred times better. The glitched frames are there, and I can spot 'em, but -- I'm me, right? I know what I'm looking for. I had someone else watch it and said "do you notice anything about this footage?" Answer: "nope. Looks fine. Annoying, but fine." (keep in mind it's basically horrific shakycam footage, flinging the camera around like a spastic dweeb with fire ants in his pants, making "Cloverfield" look like it was shot on a Steadicam).
Tomorrow I'll try a wide-angle grass 'n' trees shot.
Nice test, Barry.
So commanderspikes HAIL artifacts footage is NORMAL!? say it ain't so!
John Caballero
Veteran
Kholi said he saw some improvement with the faster card. So Barry, your tests prove his statements as incorrect?
Also, the question is how good, or bad, does it look when you do a normal smooth moving pan? Are you finding horrible artifacts and stuff in the footage? Because thats basically the question. Can we get something workable out of panning as a whole from this camera? We already know that if we do pans as if we were possesed by the panning demons it won't work! LOL.
Also, the question is how good, or bad, does it look when you do a normal smooth moving pan? Are you finding horrible artifacts and stuff in the footage? Because thats basically the question. Can we get something workable out of panning as a whole from this camera? We already know that if we do pans as if we were possesed by the panning demons it won't work! LOL.
Last edited:
Boz
Well-known member
Thanks for doing the tests, Barry. Your findings are sad news indeed; and what I feared all along. So, what you're saying is right now is that my lowly HV20 outperforms the GH1 in compression detail when panning. That's pretty pathetic. Which leads me back to my original point, how in the world did Panasonic find this to be acceptable? I sure wish there was a Panasonic rep on this board to see these things and relay news back to the mothership (and address such issues).
That is PRETTY much my summation right there. If you're not looking for it, it looks MUCH better on an actual viewing source in your living room. Thanks for quantifying that, Barry.
Also, I dunno maybe I got my settings right or just haven't noticed? But if it's not the cards then okay, one less thing to worry about being the culprit outside of it being HDV and its worse. LoL
ydgmdlu
Veteran
I don't see how a faster card could possibly improve video performance. It simply does not make sense on a technical level. The codec writes at a maximum rate of 17 Mbps, which is slightly more than 2 MB/s. A lot of the footage posted so far doesn't even use all of that bandwidth. Any card with an actual sustained write speed of over 2 MB/s should be enough. At that point, the card is no longer the bottleneck for performance; the codec becomes the bottleneck. A faster card won't make the codec write at a higher data rate. If that were true, then that would necessarily mean that the codec would be encoding with a higher bitrate.
The worst thing that one could say about the cheap cards is that maybe their write speeds are highly erratic to the point of falling below the codec's requirements. But even if that were true (and it hasn't been proven so), then you should say that one should buy the best quality card within one's budget, not the fastest. After all, manufacturer's spec listings are often misleading.
Last edited:
Actually, it's possible if el-Cheapo SDHC cards specs are not what they're being reported as. Earlier in the thread I stated that SDHC cards online were being sold as SanDisk Extreme III's, but were infact not really that.
So if El-Cheapo's write-rate was more like 15mb/s instead of a proposed 20mb/s then there's a bottleneck someone may not have ever expected. Also, check over in the 5D camp and ask them what horror stories they have to share about CF Cards.
I also know that Read/Write specs for Solid State media in larger capacities are not always what they're advertised as. So it could have played a very small role, or could have been a possibility.
Barry tested, directly, a Sandisk EXtreme III and a Cheap SDHC card and ruled that out. Great. Less for everyone else to test.
But do keep in mind specs are not always as they're told, so it really was a possibility.
Nope, I can't disagree with you here. THere's always the possibility that people are buying these on Ebay, where the poorest of quality will live.
So I don't disagree with you entirely, just want to point out that there's always a chance, that's all. And when you've got high hopes you gotta try everything. Nothin' wrong with process of elimination.
ydgmdlu
Veteran
Well, there are plenty of reasons to avoid buying cheap flash memory on eBay, whether in the form of memory cards or USB drives. The biggest one is that there are a lot of shady sellers, mostly based out of Hong Kong and China, who sell memory at falsely stated capacities. I'd say that that is arguably a bigger issue than card perfomance.
John Caballero
Veteran
So we are back to square one. What is exactly causing the bad break ups? I would just like to know how bad is it if you pan slowly and deliberately? Also do they show up on tracking shots?
And that's the thing, man. A lot of people will initially flock to Ebay because of prices not realizing what they're getting themselves into. It's just like those adapters for Nikon to EOS or whatever: waste your money going with cheap ones that don't let you do what you're supposed to.
Anyway, yeah.
commanderspike
Carbonite Member
I don't think we're back at square one just yet. I still think it's the card.
I don't think it's the lens. Why would a optical part cause digital break down?
I don't think it's post processing issues or computer related. Yes, I have noticed interlacing artefacts, such as a section of red on an umbrella completely floating interlaced in the sky whilst panning.
But what put the post processing theory to bed for me was when I looked very closely at the GH1 screen during playback direct from the El Cheapo card, of footage I'd just shot, and you see EXACTLY the same break down of the image as I and others have posted on this thread.
There are a lot of fake Sandisk Extremes out there. I made sure mine was genuine from a high street store - paid a little more but can't take any chances.
Heading out now to ram as much detail as I can through the codec and see if the breakage occurs with the Sandisk, versus my El Cheapo card.
Highly erratic write performance combined with a bug in the GH1 which causes the bit rate to spike when a pan begins could be the reason. A quality card won't fix the bug but it will hide the effects.
I don't think it's the lens. Why would a optical part cause digital break down?
I don't think it's post processing issues or computer related. Yes, I have noticed interlacing artefacts, such as a section of red on an umbrella completely floating interlaced in the sky whilst panning.
But what put the post processing theory to bed for me was when I looked very closely at the GH1 screen during playback direct from the El Cheapo card, of footage I'd just shot, and you see EXACTLY the same break down of the image as I and others have posted on this thread.
There are a lot of fake Sandisk Extremes out there. I made sure mine was genuine from a high street store - paid a little more but can't take any chances.
Heading out now to ram as much detail as I can through the codec and see if the breakage occurs with the Sandisk, versus my El Cheapo card.
Highly erratic write performance combined with a bug in the GH1 which causes the bit rate to spike when a pan begins could be the reason. A quality card won't fix the bug but it will hide the effects.
paulgandersman
Well-known member
noticed this post and I was thinking about buying one of these:
http://cgi.ebay.com/Nikon-lens-to-P...5|66:2|65:12|39:1|240:1318|301:1|293:1|294:50
would you not recommend it? if not what is a good product that does the same thing?
thanks!