Different formats effect on compression at DVinfo.net

Ola Christoffersson · November 22nd, 2007, 06:40 AM

I have been thinking about something that I do not think has been up for discussion. Maybe because it is obvious for everybody but I would be interested to hear your thoughts on this:

As the HQ mode seems to use approximately the same bandwith regardless of recording mode (judging from the "tennis files" and the "street files" posted) should this not mean that some modes will suffer much less from compression artifacts than other?

For example wouldn't 720P50 be twice as hard compression per frame compared to 720P25? This would probably be visible if this is indeed how it works. Maybe someone would care to make a test? This could tell us a lot about how good the codec really is. If 50P is very similar to 25P that would tell us that there is quite a lot of headroom for fast movements etc in 25P-mode.

Also 1080P25 contains 2,25 times as much raw data as 720P25. This should effect compression results, right? So - does this mean that it is better to record 720P if the end result is going to be 720P or is scaling down 1080P equally good?

I'd be interested to hear your thoughts on this and also see some tests if someone feels up for it. I am afraid that I don't have a camera (yet) to do this.

/ola

Mike Marriage · November 22nd, 2007, 07:28 AM

Remember that with higher frame rates, there will be less change between frames when recording the same subject. This should allow the interframe compression to work more effectively so in theory, 720p50 should NOT show double the compression artifacts of 720p25.

Craig Seeman · November 22nd, 2007, 08:25 AM

It's much like people mistakenly believe if they compress video for the web at 15fps or 12fps (instead of 29.97 or 25) they cutting the file size in half. They aren't. The slower the frame rate, the more movement between frames, the harder the codec works. Of course "talking heads" shoot with very little motion would certainly use less data.

Keep in mind we're also talking about a VBR encode here. For that very reason some people have said that an 8GB card may actually hold more than 25 minutes of video at 35mbps VBR. It does depend on what you're shooting and at what frame rate.

Also keep in mind that the playback is 35mbps VBR and the record may gather at higher data rates than that at times. Think of 60p being played back at 24p for slow motion. That 60p has to gather at a higher data rate for it be 35mbps VBR playback. If that weren't the case, 60p would look pretty poor for slow mo use. This is also why there's 720p60 but not p60 for 1080. 1080 would probably have to gather data at such a high rate that there would be another bucket load of technical issues to solve.

Dave Elston · November 22nd, 2007, 08:27 AM

I think this is a good issue that deserves some real-world tests.

While I agree with you Mike that higher frame rates will tend to have lower frame-to-frame differences, another consideration is the fact that at 50p/60p you will likely be using a higher shutter speed than at 24/25/30p, therefore the amount of motion blur decreases pushing up the level of detail per frame (for scenes that are heavy in motion) thus pushing the compression algorithm harder.

I would expect, in general terms, that 720/25p compressed should stand up better than at 50p but the difference may only be noticeable in the most extreme of 'testing' scenarios. That's just my hunch though.

Thomas Smet · November 25th, 2007, 07:56 PM

http://www.dvinfo.net/conf/showthrea...=ultimate+720p

Please check out my thread here on this exact same subject.

November 22nd, 2007, 06:40 AM	#1
Ola Christoffersson Major Player Join Date: Nov 2007 Location: Stockholm, Sweden Posts: 462	Different formats effect on compression I have been thinking about something that I do not think has been up for discussion. Maybe because it is obvious for everybody but I would be interested to hear your thoughts on this: As the HQ mode seems to use approximately the same bandwith regardless of recording mode (judging from the "tennis files" and the "street files" posted) should this not mean that some modes will suffer much less from compression artifacts than other? For example wouldn't 720P50 be twice as hard compression per frame compared to 720P25? This would probably be visible if this is indeed how it works. Maybe someone would care to make a test? This could tell us a lot about how good the codec really is. If 50P is very similar to 25P that would tell us that there is quite a lot of headroom for fast movements etc in 25P-mode. Also 1080P25 contains 2,25 times as much raw data as 720P25. This should effect compression results, right? So - does this mean that it is better to record 720P if the end result is going to be 720P or is scaling down 1080P equally good? I'd be interested to hear your thoughts on this and also see some tests if someone feels up for it. I am afraid that I don't have a camera (yet) to do this. /ola

November 22nd, 2007, 07:28 AM	#2
Mike Marriage Major Player Join Date: Sep 2005 Location: NE of London, England Posts: 788	Remember that with higher frame rates, there will be less change between frames when recording the same subject. This should allow the interframe compression to work more effectively so in theory, 720p50 should NOT show double the compression artifacts of 720p25. __________________ www.mikemarriage.com

November 22nd, 2007, 08:27 AM	#4
Dave Elston Regular Crew Join Date: Oct 2004 Location: UK 50i/25p Land Posts: 103	I think this is a good issue that deserves some real-world tests. While I agree with you Mike that higher frame rates will tend to have lower frame-to-frame differences, another consideration is the fact that at 50p/60p you will likely be using a higher shutter speed than at 24/25/30p, therefore the amount of motion blur decreases pushing up the level of detail per frame (for scenes that are heavy in motion) thus pushing the compression algorithm harder. I would expect, in general terms, that 720/25p compressed should stand up better than at 50p but the difference may only be noticeable in the most extreme of 'testing' scenarios. That's just my hunch though. __________________ Play to Learn, Learn to Earn, Earn to Play... Dave - Broader Pictures

November 22nd, 2007, 08:25 AM	#3
Craig Seeman Inner Circle Join Date: Feb 2004 Location: Brooklyn, NY, USA Posts: 3,841	It's much like people mistakenly believe if they compress video for the web at 15fps or 12fps (instead of 29.97 or 25) they cutting the file size in half. They aren't. The slower the frame rate, the more movement between frames, the harder the codec works. Of course "talking heads" shoot with very little motion would certainly use less data. Keep in mind we're also talking about a VBR encode here. For that very reason some people have said that an 8GB card may actually hold more than 25 minutes of video at 35mbps VBR. It does depend on what you're shooting and at what frame rate. Also keep in mind that the playback is 35mbps VBR and the record may gather at higher data rates than that at times. Think of 60p being played back at 24p for slow motion. That 60p has to gather at a higher data rate for it be 35mbps VBR playback. If that weren't the case, 60p would look pretty poor for slow mo use. This is also why there's 720p60 but not p60 for 1080. 1080 would probably have to gather data at such a high rate that there would be another bucket load of technical issues to solve.

November 25th, 2007, 07:56 PM	#5
Thomas Smet Trustee Join Date: Mar 2004 Location: Milwaukee, WI Posts: 1,719	http://www.dvinfo.net/conf/showthrea...=ultimate+720p Please check out my thread here on this exact same subject.