100% Quality Factor MJpeg Settings (very low compression - class 10 required!)

[modulr]

Same static scene no camera movement no change in lighting.

Unfortunately JPEGSnoop doesn't work well with the movie files but I helped myself by cutting out a few frames near the end with QuickTime Pro and saving this as a self contained movie without any recompression.
We haven't looked any further than the first frame so far and I suppose the camera jumps to Q4 very soon with such high bitrates.
Another real life scene showed a "degradation" over time from 97 to 92 quality with the old settings from this morning.

We have to test this "degradation" with real life scenes. I would love to see that it's minor.

I followed suit and can confirm similar results... the plot thickens! haha.

My initial instinct after the [400,300,250,188] settings was that we'd want less spread between Q1 and Q4, because the codec would be much more likely to utilize those Q values when possible. I shot some complex scenery today, but can't discern much difference between the 600 settings and the 400 settings.. So it could very well be riding on Q4. That being said, the footage remains pretty jaw dropping. I honestly just haven't seen detail like this before.

I think an interesting test will be to go from complex to simple, and see if there is a bump in rate.. def more experiments to be done. Gonna try Pappas chart too.

@abolit - I'm on a mac, use quicktime, and the .mov have always worked fine.

@coryshubert - The shots I had taken were with the Kit 14-140, and a Nikkor 50mm f1.4... Beyond that, nothing special was done.. just pushed the button and let it record.

 

[mpgxsvcd]

Thanks for the chart Pappas. However, how realistic is this test case? Is there a real world scenario that would ever come close to this?

The settings are working great for me so far. I even have been using the 4:2:2 settings with them.

 

[sammysammy]

i was going to say the same thing mpgxsvd..i could be wrong, but that typs of shot will never happen, at least with me..but its a good measure stick, but i think its over kill to say if my settings don't pass it ,than the hack is not working right, im not saying anyone has said that, but just saying..

 

[modulr]

Just tried the Pappas chart with the [400,300,250,188] as well... and got 10s at 103mbps. That chart is hardcore. It ultimately only tests the Q4 value. You can continue to lower Q4 it until it works. I find pappas star chart much much closer to complex real life scenery in terms of how it taxes the codec.. which is definitely much more practical.

 

[GraemeH]

Definitely good to hear the settings are working so well for everyone! And really good to know 422 is an reliable option. BTW, awesome shot Martin! not in a million years did I expect the camera to be able to output footage at this level of detail.

Even better news, I'm testing some new numbers that are yeilding even better results for complex scenes. The complex scene being some little bamboo bushes outside my house... lots of tiny leaves and patterns.

[/COLOR]

(first post)
Hi Modulr / Martin
Just Bought the GF1 over the weekend on the back of seeing your footage on vimeo and the threads here- Just to say thanks for these settings- i'm blown away, i dived in with the 384,330,300,276 -24,24,24,24. then found this thread.

Iv'e been using your 400,300,250,188-4,4,4,4- 720P-422 with the 20mm lens and a Sandisk 8GB class 10 card. iv'e shot off about 20 clips- no failures. The image is amazing . will try the 14-40mm lens tomorrow. also tempted to try your new numbers.

a fair few clips reached 105Mbit/s most around 90 and some around 60-70- they all looked stunning though

Again many thanks guys
Graeme

 

[abolit]

Modulr,
Thanks for a quick response!
Did you increase the audio bitratings as well? I did and I guess that's why I couldn't play the clip with quicktime.

I'm on a mac, use quicktime, and the .mov have always worked fine.

 

[abolit]

I got it figured! For some reason quicktime wouldn't play when sound card is turned off. Go figure!

Modulr,
Thanks for a quick response!
Did you increase the audio bitratings as well? I did and I guess that's why I couldn't play the clip with quicktime.

I'm on a mac, use quicktime, and the .mov have always worked fine.

 

[Martin Koch]

Martin,
I have a new codec test chart put together

Thanks for the chart. It's a great texture and I'm sure it will give a lifetime insurance that settings that survive this chart will work in every possible situation. Unfortunately this also means we must lower Q4 more than maybe reasonable and high detailed scenes will loose much more quality. So far I found my high detailed scene sufficient to ensure that real world scenes work.

My problem at the moment is that with high detailed scenes and these high quality settings the camera too quickly falls back on the Q4 quality so the quality or mbps of the first frame we are seeing in QuickTime are not held throughout a clip.

 

[PappasArts]

Thanks for the chart Pappas. However, how realistic is this test case? Is there a real world scenario that would ever come close to this?

The settings are working great for me so far. I even have been using the 4:2:2 settings with them.

Well this is an ongoing project to see our boundries ( AVCHD and MJPEG R&D ). We can't hang out in the safe waters cause we feel comfortable - until we know the extremes. In a funny way we are re-manufacturing the software parameters in this camera, and like manufactures we need to see those limits. That's how we learn.

I designed this test chart to push it. If you can get it to last the chart, or at least for a short while, you can bet the real world might not be as harsh. Also charts are consistent- not trees in our back yards or the hairy dog next door ( as well the bright levels are always exact ) When doing R&D we need testing tools like the chart to map it out. If all settings are the same every time however results are changing with every little adjustment we make in Ptool- we know were making progress one way or another based on how well the camera is performing.

MIKOS • MLPappas
http://TWITTER.COM/PAPPASARTS
http://PAPPASARTS.WORDPRESS.COM
http://MIKOSarts.wordpress.com
http://MIKOSarts.wordpress.com/page-2-of-blog
http://MIKOSarts.COM

 

[Martin Koch]

(Modulr) Did you increase the audio bitratings as well? I did and I guess that's why I couldn't play the clip with quicktime.

I was under the impression that the audio settings in Ptool only affect the AVCHD part and never tried it. Can I increase the MJPEG audio sample rate from 16kHz to lets say 44 or 48? Does this work?

 

[modulr]

@graemaeh - nice! Definitely good it's working out!

@abolit - I have yet to mess with the audio stuff... did it work with up'd audio quality settings?

@Poppas - true!

@Martin - I performed some additional tests. I shot the star chart for 10s, then moved to black for 10s, then back to the star chart. If the codec moved to Q4, but never returned to Q1 levels when reintroducing the image I'd be pretty concerned. That being said it seemed to demonstrate it's variable nature... The returning image was initially back to Q1 quality, then slid. This would suggest the camera is definitely dynamically selecting Q1 thru Q4. If a complex scene is really taxing the codec constantly (like a static shot of these horrid bushes out side my door.. lol, star chart, or Poppas new chart, etc), it's simply going to ride Q4. The more I think about this, it makes sense. It might explain why much of the footage with these settings still look superb.. because when it can, the camera seems ok to use the higher Q settings.


Regarding the last round of settings [600,330,260,188]...

After realizing this, the 600 settings are definitely overkill. As Martin observed the initial tables look spectacular, but the camera just doesn't spend much time at Q1=600 even when the scene is dead simple which makes the setting not particularly useful.

I'm back to thinking about the Q settings along these lines... (sorry if this seems repetitive, I'm just writing this out so it makes sense in my head... lol).

Q1 = Simple scenes. This includes a lot of night / low light content because it usually compresses easier, possibly some dof content since the background is usually very soft, and likely anything that includes solids and/or a lack of detailed textures.

Q2 = (in between)

Q3 = (in between)

Q4 = Complex scenes. Think highly patterned scenery with lots of small details and loads of textural shifts. Heavy foliage is a great example of this. The most complex scenes will evidently ride this value.


At this point I'm sticking with the original [400,300,250,188]. It seems to cover that range from simple to complex very well, and has been working superb. I can shoot [Q1-4]=400 in night/low light and simple settings... so I'm thinking it makes a great Q1 value. Ultimately, I can't think of any way to get the footage any better other than faster cards.

<tangent>We shoot with 7D/5D's at my job quite a bit. I can't wait to sneak this one in on a shoot. hahah! </tangent>

 

[mimirsan]

I found one little issue with the new super sharp settings 400,300,250,188

Highly detailed brickwork display strobing effect with small globs of blue blotches (see centre brickwork near door and lower middle cant upload video right now).
Almost like a moire effect. Cany What could be causing this?
Taken with 14-45mm panny lense GH1 same thing with 20mm both tried reduced sharpness also (Tried on GF1...same lense no strobe(!))

 

[philiplipetz]

I found one little issue with the new super sharp settings 400,300,250,188

Highly detailed brickwork display strobing effect with small globs of blue blotches (see centre brickwork near door and lower middle cant upload video right now).
Almost like a moire effect. Cany What could be causing this?
Taken with 14-45mm panny lense GH1 same thing with 20mm both tried reduced sharpness also (Tried on GF1...same lense no strobe(!))

same hack settings on GF1?

 

[crunchy]

Almost like a moire effect.

In my opinion it IS moire effect.

 

[mimirsan]

same hack settings on GF1?

Same settings looks great! Both look fantastic actually...but im guessing with more detail brings the moire demon! :evil:

 

[mpgxsvcd]

Well this is an ongoing project to see our boundries ( AVCHD and MJPEG R&D ). We can't hang out in the safe waters cause we feel comfortable - until we know the extremes. In a funny way we are re-manufacturing the software parameters in this camera, and like manufactures we need to see those limits. That's how we learn.

I designed this test chart to push it. If you can get it to last the chart, or at least for a short while, you can bet the real world might not be as harsh. Also charts are consistent- not trees in our back yards or the hairy dog next door ( as well the bright levels are always exact ) When doing R&D we need testing tools like the chart to map it out. If all settings are the same every time however results are changing with every little adjustment we make in Ptool- we know were making progress one way or another based on how well the camera is performing.

Pappas

Pappas, I didn't mean to critique your chart. I appreciate everything you have done. I just wanted to know if there was a real world scenario that comes close to this. From other's posts it sounds like there isn't anything that comes close to this in the real world.

I agree we should definitely test the firmware to its limits. However, we should also try to achieve something that is stable and reliable for each of our own uses.

Right now, for me, these 100% quality settings and lpowell's "35Mbps MJPEG In-Camera Playback Patch" are perfect.

I shot some fireworks on the 4th with the 100% quality settings. Here is the video on youtube. Unfortunately, youtube's compression kind of destroys it. I will try to post it on vimeo where you can download the original as well.

http://www.youtube.com/watch?v=_hcdbmoqwPE

 

[Martin Koch]

Just for fun a comparison to see how much the MJPEG values have been altered. I also think that depending on scene complexity there's an interpolation between the points going on because otherwise we would only see four fixed tables. But I may be wrong.

@mimirsan That is moire. I have a clip where I see the same colored rainbow moire on a roof in the distance as if it where shot with the 5DMarkII. The GH1 can do this also. Fortunately very seldom.

 

[abolit]

Modulr,
I got a chance to increase Audio sampling rate along with your mjpeg setting but after testing Papas chart and camera's failure I got back to factory default(Audio) . It's gonna be my next step. I don't want anything to "suck" resources to get the best video.

 

[Lpowell]

Pappas, I didn't mean to critique your chart. I appreciate everything you have done. I just wanted to know if there was a real world scenario that comes close to this.

Pappas' test charts were very helpful in developing my 35Mbps MJPEG In-Camera Playback patch: http://dvxuser.com/V6/showthread.php?t=215624

Though the test charts are artificial, they provide a repeatable high-detail pattern that can be used to compare the performance of subtle changes to a patch. In my experience, the demands they place on the codec are not unrealistic, as I have been able to construct real-life scenes that required even more intensive bitrates than the test charts.

For a thrilling exercise in patience, I'd suggest setting your patched camera on a tripod in front of a Pappas test chart, and watch the seconds tick off till you reach the 2GB file limit. :thumbsup:

 

[mpgxsvcd]

Pappas' test charts were very helpful in developing my 35Mbps MJPEG In-Camera Playback patch: http://dvxuser.com/V6/showthread.php?t=215624

Though the test charts are artificial, they provide a repeatable high-detail pattern that can be used to compare the performance of subtle changes to a patch. In my experience, the demands they place on the codec are not unrealistic, as I have been able to construct real-life scenes that required even more intensive bitrates than the test charts.

For a thrilling exercise in patience, I'd suggest setting your patched camera on a tripod in front of a Pappas test chart, and watch the seconds tick off till you reach the 2GB file limit. :thumbsup:

Both of your guys work is very much appreciated. There is no question on that. However, I just think the new chart is overkill for the average user. Most people can use the settings you guys have come up with for 100% of their shooting needs.

I don't see why those average users should step their quality down in order to have the ability to shoot a test pattern that they will never shoot again.

This is just my opinion but I think our time would be better spent trying to optimize 1080p MJPEG. So far I haven't seen any settings that are stable for even the star chart. I would also like to see some proof as to whether it is really scaled 720p or not. Just my opinion though.

I have tested the 1080p settings extensively but I still can't find anything stable. Got any ideas?