Sony X70 4K - Lowest bit rate in the industry!

[Ron Evans]

Can you imagine the reaction 10 years ago if someone said those images were from cameras cost around $2000 ?

Ron Evans

[Craig Seeman]

Jody, Cliff doesn't believe eyeballs. That's why I mention a comparison using MediaInfo. That will provide the "numbers" Cliff loves. My own hunch is there might be things other than the data rate that are different which may show why a lower data rate file can be better than a higher data rate file.

[Cliff Totten]

Originally Posted by Craig Seeman

Might I suggest someone do a short recording using the AX100 and x70 in 4k and open and compare the files in MediaInfo. While it won't tell the whole story it will tell you what OTHER differences there are besides bit rate. Things like Entropy and GOP Structure can result in a lower bit rate file looking better than a higher bitrate file.

Good idea. Will do that in the next day or so!

Edit: Had the files and exported the metadata form MediaInfo as text. I included the AX100@100Mbp/s too

[Cliff Totten]

Originally Posted by David Heath

And - like with a lot of things(!) - I don't think the answer is simple.....

Firstly, I think you mean "Will file #2 require 4 times the bit rate to ......", don't you? You're quadrupling the number of pixels, so "common sense" would seem to indicate you'd need 4 times the bitrate?

The answer is generally taken to be "no", but it depends a lot on what variables change. You're changing the image resolution - but are you changing the size of the viewed image as well? If your original image is viewed in a frame 8"x8", the answer will vary depending on whether file#2 is also viewed as 8"x8" - or scales up proportionally to 16"x16".

That's important when you think of the difference between 4:2:0 and 4:2:2 chroma sampling. In such a case you are far more analogous to "fixing the size of frame" - as the 1920x1080 luminance aspect is unchanging. And it then follows that although there is a doubling of chroma samples, it shouldn't necessarily be thought of as needing a doubling of bitrate for equivalence - it should be less.

Yeah,..., total brain fart on my part. Good catch.

I'm thinking of the general scalability of h.264. (either 4:2:0 or 4:2:2) Let me put it a better way than my original bone-headed explanation:

If you took 15Mbp/s, h.264, 8bit, 4:2:0 at High profile with CABAC at whatever long GOP structure you want at 1920x1080 frame size. The would give you a certain quality level.

Wait!,...do we agree that 15Mbp/s for the above specs for 1920x1080 are not too good? (I hope so...it's actually way below AVCHD's h.264, High profile, CABAC, 24Mbp/s)

Now, take this video, multipy it's bitrate and frame size times 4 and you get 60Mbp/s UHD. NOW!,...take your 70inch UHD TV, get a piece of cardboard and cover the bottom half of your screen as well as the top left 1/4. This leaves you watching a single 1/4 size or ""1080" quadrant of your 4k video. When you watch this, will you see video that equals the same "15Mbp/s" of quality or artifacts that you see in a normal 1080 video??

The short question: Is h.264 "proportionally" scalable in quality as you increase it's frame size and bit rate?

Or,..does it GAIN or LOSE bit rate effeciency as you increase or decrease it's frame size with the bitrate?

I do not know the answer. I only suspect that because it operates in fixed block calculations that you MUST increase the bitrate TOGETHER in a direct "linear" or "proportional" fashion with a proposed frame size increase. Both factors must increase "evenly" together??

So, UHD h.264 @ 60Mbp/s = FOUR 15Mbp/s "1080" quadrants in your UHD TV???

Again, yes, yes, yes this is speculation on my part. No doubt. But I have always wondered h.264's scalability and efficiency for years.

Anybody know? Anybody care?.. or maybe it's just too geeky of me to think about. lol

[Cliff Totten]

Originally Posted by Craig Seeman

Jody, Cliff doesn't believe eyeballs. That's why I mention a comparison using MediaInfo. That will provide the "numbers" Cliff loves. My own hunch is there might be things other than the data rate that are different which may show why a lower data rate file can be better than a higher data rate file.

My eyeballs tell me that my X70's 4k looks good today. I think Sony will make it look better in the 100Mbp/s codec update they are working on. (hopefully they succeed)

I wonder why Sony is even trying this?

Official Sony statement = *We are looking to support a higher bit-rate recording mode than 60 Mbps for 3840x2160 XAVC-L in the future.

What is Sony thinking here?? I don't think Sony understands yet that nobody want's it and it's really not needed? ;-)

Are we really arguing against the reasons for something that Sony is trying to do?



On a completely different note: Just took my new RX10-II for a 30 min spin. The new stacked 1 inch-type sensor has NO rolling shutter in 4k!!...it's GONE! Woah! SLOG-2 looks fantastic, highlight handling is a HUGE improvement....dynamic range is EASILY way better than previous 1 inch-type models I have used (RX10, AX100 and X70) with SLOG-2 turned on. (rec 709 highlights seems the same as previous 1inch-type) Way to go Sony...you fixed it!!

The high speed frame rates are spectacular for a cheap $1,300 camera.

This thing is a little MONSTER!

[Christopher Young]

Originally Posted by David Heath

Most importantly, you don't need twice the bitrate to code 4:2:2 chroma as you do 4:2:0.

David ~

Agreed and I believe that is why Sony settled on 50-mbit for its XDCam MPEG-2 422 Profile@High Level as opposed to its 35-mbit 420 version MPEG-2 Main Profile@High Level.

Originally Posted by David Heath

Secondly, Alister assumes the same compression ratio for luminance and chrominance, and for MPEG2 this is not necessarily true.

This I believe to be correct for MPEG-2 420 in the transmission chain but this is at a considerably lower bit rate than typical camera compression rates. I believe Alistair to be correct with regards to Sony's proprietary camera 422@HL and MPEG-2 Main Profile@High Level implementations as they use the same compression algorithm for both luminance and chroma. The same goes for Canon's and Convergent Design's use of these codec as they was licensed from Sony. In Alistair's mathematical examples I think he was keeping it reasonably simple to understand for those who want an overview of compression schemes relating to cameras but without drowning them.

With regards to artifacts. Even with the newer DBS transmission systems based on DVB-S2 we are now seeing 420 HD bit rates between 8~19Mbps with a mean average of 12Mbps for both MPEG-2 and MPEG-4 AVC transmissions so it it hardly surprising we are seeing banding artifacts. With many broadcasters now using rate shaping to fit three 1080i channels on a 31.6Mbps DBS transponder I think we will see more of these compression banding artifacts. Without laboring the point I think to compare low bit rate transmission compression artifacts with some of the later compression schemes used in more recent camera designs is probably not the best example to put forward.

Chris Young
CYV Productions
Sydney

[David Heath]

Originally Posted by Christopher Young

This I believe to be correct for MPEG-2 420 in the transmission chain but this is at a considerably lower bit rate than typical camera compression rates. I believe Alistair to be correct with regards to Sony's proprietary camera 422@HL and MPEG-2 Main Profile@High Level implementations as they use the same compression algorithm for both luminance and chroma. The same goes for Canon's and Convergent Design's use of these codec as they was licensed from Sony.

I confess I don't know whether the compression ratios in XDCAM for luminance and chrominance are the same or not. They may use the same algorithms but does that automatically mean the ratios will be the same? The point I was trying to make is that it's dangerous to assume they are - they need not be for MPEG2, and in various implementations they are not.

Originally Posted by Christopher Young

In Alistair's mathematical examples I think he was keeping it reasonably simple to understand for those who want an overview of compression schemes relating to cameras but without drowning them.

But "keeping it simple" shouldn't mean saying anything incorrect, and I'm afraid here it does. It "proves" that the basic compression ratios between XDCAM EX and 422 are the same (the only difference between them is the 4:2:0/4:2:2 aspect) and regardless of the truth about differing luminance/chrominance compression ratios, there is still the fact that doubling the number of chrominance samples won't mean a doubling of bitrate. And that's before we even start to consider any other factors that may differ between them.

It's similar to the comparisons made some time ago between XDCAM EX when it first came out and HDV. Same basic codec (MPEG2) and XDCAM EX not that much higher bandwidth...... True as far as it goes, but totally failing to take into account less quantifiable factors, of which I believe the dynamic bitrate allocation between Intra and difference frames of XDCAM was one major fact.

Originally Posted by Christopher Young

Without laboring the point I think to compare low bit rate transmission compression artifacts with some of the later compression schemes used in more recent camera designs is probably not the best example to put forward.

I was trying to illustrate a different point - that there is no need whatsoever for luminance and chrominance compression within a codec to have the same ratio, and in some cases they don't. Hence taking the combined number of samples per frame and dividing by bitrate doesn't (by itself) prove very much.

[David Heath]

Originally Posted by Cliff Totten

If you took 15Mbp/s, h.264, 8bit, 4:2:0 at High profile with CABAC at whatever long GOP structure you want at 1920x1080 frame size. The would give you a certain quality level.

Wait!,...do we agree that 15Mbp/s for the above specs for 1920x1080 are not too good? (I hope so...it's actually way below AVCHD's h.264, High profile, CABAC, 24Mbp/s)

I'll say again, beware of simple headline numbers.

As far as AVC-HD goes, then I understood that the 24Mbs figure (for 720p and 1080p/25, 1080i/25) is a peak value, not an average value. I seem to remember when the first Panasonic AVC-HD cameras came out that the average value I was seeing reported was far more around the 17-19Mbs mark?

And that was for early implementations, not using anything like all the tricks that AVC-HD is capable of. So my first question to you is this 60Mbs XAVC implementation you are unhappy about an AVERAGE of 60Mbs or a PEAK of 60Mbs? Because without being absolutely crystal clear about that, this entire debate becomes meaningless!

Secondly, for AVC-HD we may be talking about interlace (1080i/25) - with what you're referring to (XAVC for 4K) it's progressive. It has to be - 4K is always progressive. And progressive lends itself better to compression than interlace.

Thirdly, XAVC is a more advanced codec than AVC-HD. It's more recent - so requires a more recent decoder version - but that allows it to be more complex than earlier H264 incarnations (like AVC-HD).

Take all that together and it all becomes feasible that a quadrant of such a 60Mbs QFHD image *MAY* indeed be better than an AVC-HD single image. Frankly, there are so many variables that I wouldn't like to predict. All I will really say is DON'T rely on "simple" headline numbers.

(And note also that as Chris says, digital broadcasting of 1080i/25 is indeed using bitrates averaging around 12Mbs, and I can't remember seeing any noticeable artifacting on BBC One HD at home for a long time. OK, they've got statmuxing with other channels in the multiplex to help get it so low, along with expensive broadcast encoders, but the implementation is unlikely to be able to be as sophisticated as XAVC (to maintain backwards compatibility with all existing decoders), and it's for an interlaced system.)

[Wacharapong Chiowanich]

I think the thing about XAVC either S or L's published 60Mbps being the peak or average sustainable bit rate is the key. I myself still have had no idea that being one or the other after all these times!

[Ron Evans]

I have seen the following.
AX1 QFHD 60P 150Mbps shown as 155Mbps in Catalyst Browse and 149 to 152 in VLC player tools
AX1 QFHD 30P 60Mbps shown as 59 in Catalyst browse and 59 in VLC
AX1 QFHD 30P 100Mbps shown as 99 in Catalyst Browse and VLC player

AX100 QFHD 30P 60 and 100 show the same as the AX1

AX100 HD nominally 50 Mbps show as 50 to 53 in VLC player and 53 in Catalyst browse.

Same clips used of course to compare in Catalyst browse and VLC player. Not sure if Catalyst Browse just gives the maximum data rate as one can watch the data rate as the clip plays in VLC player.

Not sure if these mean the 60 and 100 are maximum or not but the 50 for HD and the 150 for 60P QFHD are not the maximum as I have observed more in several clips. I have not shot XAVC-S HD in the AX1 to try but will do some time. Expect it is just like the AX100.

Ron Evans

[Wacharapong Chiowanich]

Looks like we could say that at least for Sony cameras' published XAVC bit rates the numbers actually mean "around xxxMbps". They mean neither peak nor average but it seems safe to refer to for the sake of arguments since the variances, as Ron has found out, are pretty small.

The definition aside, my experience shooting with my AX100 at both 60Mbps and 100Mbps for the past 2-3 months has yet to reveal any clear proof the higher bit rate files have any noticeable IQ improvement in moving (either the camera itself moving or shooting at moving subjects) shots. In static shots or shots with little movement, they look exactly identical.

[Ron Evans]

I am not a big fan of slow frame rates being particularly sensitive to judder. I can tell the difference on a very slow pan at 30P on the AX100 and the AX1 between 60Mbps and 100Mbps but it is almost insignificant. Noticeable if the pan is across grass or leaves. Also very little difference between the two cameras. For relatively still frame there is no difference that I can see between 60 and 100 Mbps maybe as one would expect. 30P frame rate motion cadence is more dominant. For me there is a big improvement in moving to 60P at 150Mbps.

Ron Evans

[Cliff Totten]

In the side by side test I posted a pager earlier in this thread, there are only "mild" difference between 60 and 100Mbp/s on the AX100. However, all those test scenes were on a tripod pointed at palm trees and a very gentle breeze. So obviously, both bit rates were relatively close. Neither one was too stressed.

I must admit that I'm a little surprised at how many people had a strong reaction to this topic. I figured we'd all jump on the band wagon with our pitch forks and shout for 100Mbp/s. I guess it didn't really happen that way and some disagreed completely with the general concept that less compression is a good thing.

I also had a really difficult time explaining to people that I think 60Mbp/s looks "OK" (they think I hate it...I dont) but that I strongly agree with Sony on their attempt to add higher bit rates to the X70. I think that with the rest of Sony's models and the rest of the industry doing 100Mbp/s, the X70 was left all alone at 60Mbp/s.

I think Sony is attempting to do the right thing with the X70. (I guess some see it differently)

I guess we can all agree to disagree. That is what makes this forum so much fun.

CT;-)

[David Heath]

Originally Posted by Cliff Totten

I must admit that I'm a little surprised at how many people had a strong reaction to this topic. I figured we'd all jump on the band wagon with our pitch forks and shout for 100Mbp/s. I guess it didn't really happen that way and some disagreed completely with the general concept that less compression is a good thing.

It depends which way you look at it. Higher compression rates may indeed be seen as a "good thing" in respect that they mean lower file sizes, and lower data rates - which taken together can mean smaller memory card sizes, and speed performance is less critical - so much lower per GB costs as well. But yes, go too far and quality may suffer. And the question becomes where you draw the line. That's the question constantly confronting designers and manufacturers. Compress too high and people complain about the quality. Compress too little and people complain about media costs.

It's easy enough to simply say "less compression is a good thing" - but higher bitrates come at a cost. If it makes a noticeable difference, well, fair enough, but if any difference is marginal then you have to strike a cost/benefit deal.

For my own part, I was really just trying to get over the general concept that drawing strong conclusions on headline numbers is dangerous. The more I find out about the whole subject, the more I realise I don't know - but I do now realise that simplistic assumptions can be seriously wrong. Which is why I haven't given any direct opinion to your fundamental question - I don't know the answer.

In the very first post you said:

Originally Posted by Cliff Totten

Considering that the rest of the industry and all other Sony cameras shoot at 100 Mbp/s or higher. .......... Literally ALL Panasonic, JVC and Canon 4K cameras shoot at 100Mbp/s or higher. (because realistically, every manufacturer knows it's very necessary)

But I put it to you - what if investigations showed that the 60Mbs was an average, and the 100Mbs of the other cameras was their peak value? Indeed, what if it turned out that for that "100Mbs" codec the AVERAGE rate was only (say) 50Mbs? Would you still be pushing for it?

Or what if the 60Mbs footage had a different GOP structure - more difference frames between I frames?

Yes, those scenarios are unlikely to be true - I'm certainly not saying they are - but..... it just shows how headline number "reasoning" COULD lead to very wrong conclusions. That's the point.

And same with comparing one quadrant of the 4K image with a 1920x1080 image. Factor in that XAVC bitrates are average, AVC-HD are peak, then add in the fact that 4K is progressive (so will compress better), and a modern XAVC encoder is almost certain to be more efficient than an early AVC-HD one, and the conclusions become very different. (And there may well be other factors I'm not accounting for, which may work either way.)

What you're asking for (100Mbs instead of 60Mbs, all else equal) may well give an improvement - but if it's only very slight, then is it worth it? That's why we're not picking up the pitchforks and marching...... :-)

[Greg Boston]

Originally Posted by Cliff Totten

II guess we can all agree to disagree. That is what makes this forum so much fun.

Indeed, as long as it's done with tact and respect. Good discussions about the merits or downsides of a subject are often enlightening to both sides of the table. It's just sad to see how despicably it often plays out on social media or as comments under news stories.

-gb-