Data Rates When Overcranking

[Sumfun]

Does anyone know if the AF100 increases the recording data rate when overcranking? If so, what is the maximum data rate, as this will affect which SD cards will work with the camera. I would assume that the worst case is when shooting 1080P24 in PH mode with VFR engaged and overcranking to 60P.

For comparison, the EX1/EX3, which also use a VBR long GOP codec, increase the recording bit rate from 35 mbps to 87.5 mbps in case of 24P/60P overcrank. The data rate increases by 2.5x because it is recording 2.5x more frames per second, thus allowing the amount of data per frame to remain constant. I hope the AF100 does something similar, because the alternative would be to degrade the picture. And who wants that?

Jan or Barry?

[Cristi Olariu]

As far as I know, there isn't any major increase of bitrate in undercrank or overcrank (I don't remember which one is used for slowmotion) on EX cameras because the under/overcranck is working in 720p mode automatically. I guess this is the case with every other sub-10k camera.

[Barry_Green]

The AF100 data rate stays the same. But it doesn't "degrade" the picture, it stays about the same quality.

You'd think there'd be a significant difference, but there isn't. You can see the exact same thing in the other AVCCAM/HMC cameras, when shooting 720/24p vs. 720/60p. The data rate stays the same, and the picture quality stays about the same.

How can this be? Shouldn't more frames per second mean a big drop in quality? Actually, no, not necessarily, because the codec doesn't encode each frame. It only encodes the amount of change between frames. And when you're shooting 24fps, there's much more change between frames, than when you're shooting 60fps. At 60fps, the changes between frames are minimal, whereas at 24fps obviously the amount of change could be 2.5x as much. So if you're tracking a moving object that's moving at 120 pixels per second, does it really matter if you encode 24 frames of five pixels' change, or 60 frames of 2 pixels' change? With a highly efficient codec like AVC, it doesn't seem to make much difference.

I'm sure there are extreme circumstances that might show the difference, but in general use it sure didn't seem to make any difference. I shot lots of 1080/60p and 1080/36p and the individual frames looked just like the 1080/24p frames. It should be noted that I didn't dig deep for pixel-peeping purposes, as pixel-peeping is kind of pointless on an unreleased/unfinished product. I'll dig around more in the final version.

On the EX, it does indeed increase the data rate, but that only brings the efficiency of the codec up to where the AF100 starts anyway. H.264 is 2 to 2.5 times more efficient than MPEG-2, so you'd need 2.5x as much bandwidth to get equivalent image quality anyway. The EX cameras are using MPEG-2, which is nowhere near as efficient as AVC, so cramming 2.5x as many frames may have had a detrimental effect on the picture, and running 2.5x as much bandwidth would definitely help. You can actually test that on the EX cameras; put one in 720/60p mode, and shoot a sequence, then put it in 720/24p mode and S&Q it to 60fps. The first test will be 720/60fps at 35mbps, and the second test will be 720/60fps at 87.5mbps. How much of a difference is there in the image? On MPEG-2, I'll wager it's quite a bit cleaner, but I don't know as I never ran that test.

[hunter richards]

If the overcrank footage plays back as ~24mbs avchd, than I guess when your shooting 60p you could be running at a max of 60mbs strait to the cards (but it becomes 24mbs on playback)

Barry- have you checked your overcrank mts files to see what the data rate is?

[Sumfun]

Thanks for the explanation Barry. If you have time, it may be an interesting experiment to do a side by side comparison of 1080P30 and 1080P60 pictures while the camera is in motion.

Also, do you know if the overcranked data is fed to the SDI output? I'm not sure if 1080P60 is supported.

[Barry_Green]

If the overcrank footage plays back as ~24mbs avchd, than I guess when your shooting 60p you could be running at a max of 60mbs strait to the cards (but it becomes 24mbs on playback)
Not sure what you're getting at. 24mbps is 24 megabits per second. That doesn't change on playback.

The # of frames per second changes, but the stored data files are at a fixed rate.

Remember AVCHD isn't like DVCPRO-HD. It doesn't allocate a set amount of data per frame, and then if you shoot less frames it takes less space or if you shoot more frames it takes more space. That's not how it works. It's a long-GoP codec that compresses an entire group of frames simultaneously, and operates at a consistent bitrate.


Barry- have you checked your overcrank mts files to see what the data rate is?
Yes, all 21-24.

[Barry_Green]

If you have time, it may be an interesting experiment to do a side by side comparison of 1080P30 and 1080P60 pictures while the camera is in motion.
I don't have any time, but I also don't have the camera, so ... neither prerequisite will work out, at least right now! :)

But yes, I agree that that would be a very interesting test to run. Should be exactly the same, percentage-wise, as running an HMC40 at 720/30p vs. 720/60p.


Also, do you know if the overcranked data is fed to the SDI output? I'm not sure if 1080P60 is supported.
That's the thing I don't know how it does what it does. 1080/60p is most definitely not fed to the SDI output. So my guess is it probably either embeds the frames into 1080/60i, or just shows 30fps on the output. But I don't know.

[hunter richards]

Not sure what you're getting at. 24mbps is 24 megabits per second. That doesn't change on playback.

The # of frames per second changes, but the stored data files are at a fixed rate.

Remember AVCHD isn't like DVCPRO-HD. It doesn't allocate a set amount of data per frame, and then if you shoot less frames it takes less space or if you shoot more frames it takes more space. That's not how it works. It's a long-GoP codec that compresses an entire group of frames simultaneously, and operates at a consistent bitrate.

Gotcha, I was confused because 1080/60p avchd isnt a timebase standard on this camera and is only for overcrank with playback at 24 or 30. That theoretically if the camera records 1080/24pN at 24mbs- that just maybe if you played back your 60p for 60p (instead of 24p as it does it camera)- you might have 60 frames at 60mbs which would be the same compression rate per frame as not overcranking. I thought thats what the OP was asking- but im sure you know better than I do about this stuff. Thanks.

[Barry_Green]

When shooting the overcrank stuff, you are shooting 60 frames per second -- but the data rate is still 24 megabits per second. So when you play it back, it'll play back at 24fps, which means that your original 24mbps will be spread out across 2.5 seconds. If you speed the footage up so you're playing back 60-for-60, then you'll be getting the 60fps you imaged, playing back at 60fps, but still at 24mbps.

[Sumfun]

But yes, I agree that that would be a very interesting test to run. Should be exactly the same, percentage-wise, as running an HMC40 at 720/30p vs. 720/60p.



Yes, but 720p60 is supported by many cameras and has about the same amount of data as 1080i60. On the other hand, 1080p60 has twice the data and is only supported by a few cameras. So it would be interesting to see if the picture quality still holds up.

[mcsmooth]

Barry, it sounds like you are saying it can only record at a bitrate of 24mbps regardless of VFR. If that were the case, the psudo-60p files should show up with a bitrate under 10mbps. On the other hand wouldn't that mean the 12fps files would end up with 48mbps?

The way I saw it, it looked as if all progressive frame rates, (including VFR) always ended up with a 24mbps playback file. This would ensure compatibility by holding the files to the specs of the original format. Anything else would be a variation of the PH mode. This is the first time we are seing VFRs with this codec, and I haven't seen any files so please correct me if anyone has raw files.

Assuming the above, the RECORDING bit rate would be 2.5 times (< 60mbps) as high per second than that of standard 24p. The resulting file is normal 24p that PLAYS back at 1x (< 24mbps) which appears as slow motion. Play it back at 2.5x in post and you have a 60mbps codec. All frame rates inside a 24p container should have the exact same efficiency on playback. This would also mean that 30p over 24p would be slightly more efficient than standard 30p.

Again, all of this is assuming all VFRs end up with the same bit rate file. The main hurdles to make that happen are that the encoder and SD writes have to be capable of 2.5x speeds. If we see mention of class 10 cards being needed for 60p over 24p, this will probably be the best in-camera implementation of 1080p60 out there. If the bit rate is stretched... it will still be a cool feature, just not as robust.

[Faust]

When shooting the overcrank stuff, you are shooting 60 frames per second -- but the data rate is still 24 megabits per second. So when you play it back, it'll play back at 24fps, which means that your original 24mbps will be spread out across 2.5 seconds. If you speed the footage up so you're playing back 60-for-60, then you'll be getting the 60fps you imaged, playing back at 60fps, but still at 24mbps.I think Barry made a clean explanation.
Spreading 24 mbps out across 2.5 seconds generates a 9.6 mbps playback rate for 60:24p slowmotion.
Loss of playback quality will be the price to pay for slowing down 60fps originally coded @24mbps.

As mcsmooth wrote, VFR@1080p is a very cool and unique feature, yet not as robust as a desirable 1080/60p with a higher bitrate.