Author Topic: Bit rate investigation  (Read 74464 times)

1%

  • Developer
  • Hero Member
  • *****
  • Posts: 5936
  • 600D/6D/50D/EOSM/7D
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #325 on: September 22, 2012, 11:39:11 PM »
The only thing that is supposed to make it red is going over 4gb and since its counting bit rate wrong it probably thinks thats where it is when its actually like 1 or 2 gb in. After that BR is completely wrong and quality lowering function thinks you're doing 50mbps when you're actually at 150 and stuff like that. I'll have to see whats actually in MVR_BYTES_WRITTEN.

1%

  • Developer
  • Hero Member
  • *****
  • Posts: 5936
  • 600D/6D/50D/EOSM/7D
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #326 on: September 23, 2012, 01:24:45 AM »
Try this:

https://bitbucket.org/OtherOnePercent/tragic-lantern/downloads/autoexec.bin.BetterBRI


Made BR not update so fast but buffer still does. I think the problem was a little conceptual.. don't know why it took 3 cans of "steel reserve" for it to dawn on me.  We were getting megabits / 10ms, lol. Buffer save works with new peaking but indicator will be off from slow redraws. Also has full screen magic zoom but doesn't work in crop mode. There is some danger of stoppages in the first second while BR is calculated. Maybe through sheer luck its working better for me from cursory testing.

JasonATL

  • Freshman
  • **
  • Posts: 54
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #327 on: September 23, 2012, 02:20:23 PM »
Thanks, 1%! I'll give it a shot later today, I hope, and report back.

JasonATL

  • Freshman
  • **
  • Posts: 54
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #328 on: September 23, 2012, 11:59:12 PM »
I haven't shot anything challenging with this new build yet. But, I wanted to give folks a heads up in case they try this beta (any version, probably). I had left the audio on accidentally when engaging CBR on this. I got an Err70. Recovered fine, but Err70 several times without writing a movie file. ML log was written. So, be sure you have disabled audio in the Canon menu before testing this! Otherwise, this version is working - again, haven't shot anything that would tax the bitrate yet.

Also, I edited the video that I shot last week. It isn't "wow look at this picture quality" material. It was me handholding my 600D with my Zeiss 50mm 1.4 and chasing my daughter's classmates around a horse stable, trying to get exposure on the fly (weighing overexposure against seeing into shadows - literally). Also, the color grade is not one that necessarily wows (I was going for a look). But, I shot it all with 1%'s bitrate enhancement and I will say that, unlike working with normal footage, never did the grading or sharpness get to a point that I thought constituted a breakdown. I was able to push sharpening far further than normal.

1%

  • Developer
  • Hero Member
  • *****
  • Posts: 5936
  • 600D/6D/50D/EOSM/7D
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #329 on: September 24, 2012, 01:22:08 AM »
Did you leave record separate wav on + canon audio. Thats how I can reproduce it. It initializes audio stuff before hand so you have to pick 1 or the other.

I didn't user proof this one but I took out the 68% stuff since BR measuring is ok. Won't increase quality if BR is already at 115Mb/s. Can make this customisable or increase it if its too low. Also drops faster if your already above 100.

https://bitbucket.org/OtherOnePercent/tragic-lantern/downloads/autoexec.bin.600D.MoreStopProof

Kabuto1138

  • Freshman
  • **
  • Posts: 87
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #330 on: September 24, 2012, 02:47:16 AM »
Looking pretty good!!

JasonATL, have you noticed any banding on the raw footage?

thanks!!

JasonATL

  • Freshman
  • **
  • Posts: 54
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #331 on: September 24, 2012, 02:04:44 PM »
Thanks again, 1%. And thank you for saying, "user proof" rather than "idiot proof"! The latter might have been more accurate when referring to me, I think.  ;)

We're supposed to have some wind here today, so perhaps I'll be able to test out with some blowing leaves again, since I have experienced how this particular feature has behaved in this circumstance.

I haven't noticed any in the raw footage. Do you ask because you see it in the edited footage or shots? If so, perhaps I'm just not seeing it.

Kabuto1138

  • Freshman
  • **
  • Posts: 87
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #332 on: September 24, 2012, 07:23:25 PM »
Hey Jason, I didn't see any banding on the footage.  I just wanted to know because just THAT improvement... is AMAZING... no frigging banding!!

Thanks!!

Can't wait to try this eventually in my 5d Mark II!!

1%

  • Developer
  • Hero Member
  • *****
  • Posts: 5936
  • 600D/6D/50D/EOSM/7D
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #333 on: September 24, 2012, 11:18:13 PM »
Only time I got banding was when editing a completely under-exposed iso 5k test clip shot in super neutral. Between SN and marvell's advanced... best footage ever. No profile/neutral doing to much color crushing for me.

Hehe, anyone can make the audio mistake. Thats why FPS override audio (from ML main) turns off canon audio itself to make sure. I don't have time to wait for that and leave canon audio off 100% of the time since its easier turning it on/off from audio menu and this way it doesn't fight with canon's functions. I haven't checked sync accuracy very thoroughly but they start at the same time so should be good.

JasonATL

  • Freshman
  • **
  • Posts: 54
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #334 on: September 25, 2012, 12:20:49 AM »
1% - Still working with this latest build and tried GOP=3. It looks to be absolutely awesome. The buffer save is clearly working more effectively now. I gave it quite a workout for about 15 minutes. I'm trying to look at the footage now to see if I can see the difference. My first pass through, I couldn't see when the save function kicks in - so this is great! No more "breathing". I tried to overflow the buffer many times and haven't been able to yet. The recording didn't stop on me in my few tries.

Also, I assume you're seeing this: peak bitrate at 250 Mbps on the i-frames! Craziness!

1%

  • Developer
  • Hero Member
  • *****
  • Posts: 5936
  • 600D/6D/50D/EOSM/7D
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #335 on: September 25, 2012, 05:11:11 AM »
If only that 250 was consistent.

unity2k

  • Global Moderator
  • New to the forum
  • *****
  • Posts: 23
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #336 on: September 25, 2012, 05:32:08 AM »
Guys, for some of us following this, you are doing some intriguing, amazing things on the 600D. I look on with envy. Looking at JasonATL's comparison shots between GOP3 and Canon FW, specifically the leaves that are crisp and clean on the GOP3 while the Canon FW shots are full of jpeg artifacts and mushy edges. These developments appear to be a giant leap forward.

What's missing from this dialogue is, what's it going to take to give more of us the opportunity to start working with this version of ML? I'm using a couple of 550D's and am more than willing to potentially sacrifice the older one in order to give GOP3 a run.

I'd imagine I could get the code and give it a try to port it to my camera, but what about a compiled version we can try out?

1%

  • Developer
  • Hero Member
  • *****
  • Posts: 5936
  • 600D/6D/50D/EOSM/7D
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #337 on: September 25, 2012, 05:56:03 AM »
I can try to turn off assertions and and just change the gop on 550D but you'll only get ~4 minutes of video since there is no limit hack. Dunno why other 550D people haven't tried it. All you have to do is ifdef/ifndef out the cache hack parts. Its working on 1100D too. Its not so much that your camera would break, just that the hack wouldn't produce the greatest results.

Its not just the gop, its the gop combined with direct quality control (technically above qscale limits). That part, I have no idea how it works on other firmwares.

unity2k

  • Global Moderator
  • New to the forum
  • *****
  • Posts: 23
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #338 on: September 25, 2012, 06:00:05 AM »
Thanks 1% for the quick answer. As others have already said, I too do not care about the 4 minute limit. When filming nature elements very rarely do I need a segment of that length. Plus, I'm just as interested in doing some green screen testing.

So, when and where can I download it from you?

You've made my night!

Andy600

  • Contributor
  • Hero Member
  • *****
  • Posts: 1863
  • Have you tried turning it off and on again?
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #339 on: September 25, 2012, 12:16:19 PM »
Hi 1%

I've been trying out your latest build and I'm getting some pretty wild BR fluctuations especially when setting GOP greater than 1. BR is  dropping as low as 2000 and seems to happen when I fast pan or move the camera quickly. It's very noticeable on footage.

I've also achieved sustained BR at around 80-90mbit ALL-I but can't remember the settings doh :( Can the movie logging function be modified to save BR/Encoder settings?

In your tests what have you found to be the best settings for sustaining a higher than FW BR? I'd be very happy with sustaining 80mbit but stopping it from dropping much below that. It's nice to know it can peak above 200mbit+ without stopping but in the end the footage will only be as good as the lowest sustained BR.


Colorist working with Davinci Resolve, Baselight, Nuke, After Effects & Premier Pro. Occasional Sunday afternoon DOP. Developer of Cinelog-C Colorspace Management and LUTs - www.cinelogdcp.com

1%

  • Developer
  • Hero Member
  • *****
  • Posts: 5936
  • 600D/6D/50D/EOSM/7D
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #340 on: September 25, 2012, 04:15:08 PM »
Haven't been getting that but I have to check it out. Shouldn't be dropping that low unless you cover the lens or something. Logging would kind of kill it as would have to write to the card too.

Pan around with debug and see what quality is when you get that drop. Did a quick before work test. I did get a really low BR drop but that was when lens passed my door frame (black, no detail). I didn't see quality move for me when I pan, its all 87. But BR did fluctuate as scene changed. Lots of blur and rolling shutter... are you shooting 30p?

Just spun around at 30P and only get drops when hitting things like heavy motion blur scene/empty sky/overexposure. I'm not seeing Q move for these, just camera compresses the scene. Gop3. Post a frame grab and BR graph.

Quote
BR at around 80-90mbit ALL-I

Low ISO like 100 or 200.

Theres not that much to set at this point, just gop to 1 or 3 and the slice takes care of itself.

This is yesterdays merges, still have to do todays:

https://bitbucket.org/OtherOnePercent/tragic-lantern/downloads/autoexec.bin.MostRecent1

Andy600

  • Contributor
  • Hero Member
  • *****
  • Posts: 1863
  • Have you tried turning it off and on again?
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #341 on: September 25, 2012, 05:42:53 PM »
Post a frame grab and BR graph

I would but I've since reformatted the card. I think it might possibly have been a read/write error on my card because one of the .mov files was corrupt on playback. If it happens again I'll be sure to get a framegrab and the BR graph. Gonna try your latest binary. Thanks as ever! :)
Colorist working with Davinci Resolve, Baselight, Nuke, After Effects & Premier Pro. Occasional Sunday afternoon DOP. Developer of Cinelog-C Colorspace Management and LUTs - www.cinelogdcp.com

1%

  • Developer
  • Hero Member
  • *****
  • Posts: 5936
  • 600D/6D/50D/EOSM/7D
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #342 on: September 25, 2012, 05:51:14 PM »
Yea, all this writing has got to be killing the cards. Oh well, thats what they're for. My patriot card seems broken in at this point. Lowest quality is like 129 which is like qscale +17 or so... 2k bit rates come from lower q's that we shouldn't be going to in this implementation. I made some of the lowest quality vids ever at QP 51.

Andy600

  • Contributor
  • Hero Member
  • *****
  • Posts: 1863
  • Have you tried turning it off and on again?
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #343 on: September 25, 2012, 06:40:42 PM »
I'm using a Patriot card too and it's not bad for the money but it's starting to fall apart... literally ;D

Your new Bin seems good :)
Colorist working with Davinci Resolve, Baselight, Nuke, After Effects & Premier Pro. Occasional Sunday afternoon DOP. Developer of Cinelog-C Colorspace Management and LUTs - www.cinelogdcp.com

1%

  • Developer
  • Hero Member
  • *****
  • Posts: 5936
  • 600D/6D/50D/EOSM/7D
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #344 on: September 25, 2012, 07:34:28 PM »
I got 120mb average over 5minutes on iso 80.   :o

Andy600

  • Contributor
  • Hero Member
  • *****
  • Posts: 1863
  • Have you tried turning it off and on again?
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #345 on: September 25, 2012, 09:00:28 PM »
I got 120mb average over 5minutes on iso 80.   :o

Not bad ;)
Colorist working with Davinci Resolve, Baselight, Nuke, After Effects & Premier Pro. Occasional Sunday afternoon DOP. Developer of Cinelog-C Colorspace Management and LUTs - www.cinelogdcp.com

1%

  • Developer
  • Hero Member
  • *****
  • Posts: 5936
  • 600D/6D/50D/EOSM/7D
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #346 on: September 25, 2012, 09:05:25 PM »
Drove around too when I went to get smokes and only got one stop (a first record) even with trees. BR was a little loopy but I have to see if its visible. I did get 1 frame at 16248kbps somewhere in the begining but haven't watched the footage yet. Also a 234Mb/s peak (in about 25 frames all over 210).

heheh, actually can't play any clips back on my development laptop at all unless really small. even a 400mb clip slows to a crawl.

Andy600

  • Contributor
  • Hero Member
  • *****
  • Posts: 1863
  • Have you tried turning it off and on again?
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #347 on: September 25, 2012, 09:09:12 PM »
On a Patriot card?  :o

My card seems to fall over at about 150mbit (previous builds). I'll try and push it harder tomorrow. Doing some night shots atm to tweak another PS and testing high BR/High ISO for noise performance.
Colorist working with Davinci Resolve, Baselight, Nuke, After Effects & Premier Pro. Occasional Sunday afternoon DOP. Developer of Cinelog-C Colorspace Management and LUTs - www.cinelogdcp.com

1%

  • Developer
  • Hero Member
  • *****
  • Posts: 5936
  • 600D/6D/50D/EOSM/7D
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #348 on: September 25, 2012, 09:12:11 PM »
Yea, patriot 64gb UHS-1. It won't constantly write above 150 but sometimes it will burst, usually that would produce a stop before but now it just drops quality until card can catch up.

unity2k

  • Global Moderator
  • New to the forum
  • *****
  • Posts: 23
Re: Bit rate investigation - altering ratio between P and I frame sizes
« Reply #349 on: September 26, 2012, 12:20:31 AM »
Took me a moment to figure out what settings were affecting what, and I am yet to start making notes as to results as I'd like to know my way around the settings, but, after some false starts, I've had success on my 550D.

I've looked back at some very detailed videos shot in the Grand Canyon on this camera, they are fairly consistent with bitrates between 45,000kbps and 50,000kbps, checked a bunch of others recorded with Canon FW, same bitrates. But now I have this new video I shot outside of some trees with a good wind kicking them around; bitrate reading is 139,793kbps. The 34 second 1920x1080 24fps video is 566MB in size - WOW!

Subsequent tests I have seen consistent bitrates of 166,817kbps with the buffer hanging around 30%. All of these tests were performed on a Pretec 64GB SDXC1 which claims to be a Class 16! Tried using the same settings on a SanDisk 64GB SDXC1 Class 10 and the only consistent numbers I get are at about 65,000kbps, better than Canon FW but not as amazing as compared to what I can record on the Pretec.

I've had some Error 70's, but only on the SanDisk card. Removing the battery cleared the issue. Also had each card become corrupted requiring reformatting.

Anyone care to explain the Encoder Options, or point me to somewhere I can understand what they do?

I do know that if I can have everything set to 3.0x while recording to the Pretec card. While on the SanDisk if I set P factor through D2 factor at 2.5x and GOP0,1,and 4 at 1.0x, with GOP3 factor = 2.5x, I get the results stated above. I haven't changed JP Slice.

And maybe I'm misunderstanding something, but how do I record all I-frames?

Amazing work guys. As soon as I can I'll setup both my 550D's to record the same scene, one with Canon FW and the other with this bitrate magic, I'll post what I see.