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

[Cliff Totten]

For me, I'm shooting in a rec709-ish profile that is "pre-graded" with fairly high contrast and color saturatiuon, than 60Mbp/s concerns me less. And make no mistake, the fact that all (I think) Sony cameras have it along with 100Mbp/s is great. (60mbp/s needs to be there)

However, if shooting in flat de-saturated colors to help protect RGB channels form clipping and grading is necessary? I really feel strongly that Sony is doing the right thing in trying to support higher PXW-X70 bit rates.

As a general rule of thumb for almost all compression schemes...higher bit rates provide less and less quality loss. We know that many compression types allocate bandwidth to the most vital and visible areas if an image...mid tones. Shadows and the darkest areas often get less bits because of this. If you want to lift the shadows of an image, you will typically find more macro blocking, large block rounding and banding in those tones.

So for me to declare that less compression is a good thing for image quality and grading in post. (in general). I don't see how anybody can say that is nutts.

Yes,..if Sony succeeds in getting the X70 100Mbp/s....YES!,..I will still use 60Mbp/s on occasion for certain applications!

I just got the new RX10-II and I'm loving the SLOG-2 on it so far. This little monster has 100mbp/s and I NEVER would want to try and grade/normalize SLOG-2 from a 60Mbp/s codec. That is too much bending and stretching for 8bit 60Mbp/s. Can we at least all agree with that? H.264 is great but it DOES have it's limitations.

[Noa Put]

Is the slog2 from the rx10II also not 4:2:0 8bit?

[Cliff Totten]

The RX10-II is 8 bit, 4:2:0. Like all Sony 4k/UHD cameras today, it can shoot in 60mbp/s and 100mbp/s. (Whoops,...with the only exception being the PXW-X70)

Almost forgot there.

CT :-)

[Noa Put]

Yes, but you said you never would grade/normalize SLOG-2 from a 60Mbp/s codec, do you think 100mbs would make such an enourmous difference?

[Cliff Totten]

I'll let you know when I try on the RX10-II. My SLOG-2 experience so far only comes from the A7s and ProRes from the Shogun. I typically use the 600Mbp/s flavor of Prores. (is that right?...yeah, it's somewhere around the 600Mbp/s area)

That is only 8bit as well. I have to admit though, ProRes is tough as nails, even in 8bit.

I have only done a few minutes of SLOG-2 footage and that was just to see how much dynamic range I could see in the RX10-II's sensor

Will let you know!

60Mbp/s average bit rate = 7.5 megaBYTEs ber second.

100Mbp/s average bit rate = 12.5 megaBYTES per second.

That's an average of 5 additional megaBYTES per second across 30 frames using this higher bit rate.

Can anybody actually say with a straight face that all that extra image data (40Mbp/s additional information) is "useless"...or "worthless"??

The Sony XAVC engineers and marketing departments certainly wont say that!!! ;-)

[Noa Put]

Can anybody actually say with a straight face that all that extra image data (40Mbp/s additional information) is "useless"...or "worthless"??

If you can show what that 40mbs means in the real world, like showing actual footage that would support your statement, then it would not be necessary to talk numbers again. Also If that slog2 from the rx10 would be 10bit 4:2:2 this would help a lot more when you want to grade your footage then a higher bitrate only.

[Cliff Totten]

OK. So to clear all this up for everybody, I'll setup a nasty stress test for both 60Mbp/s and 100Mbp/s. I guess we can all inspect the results. (whatever that may be)

I can only do this with the h.264 XAVC-S .MP4 wrapper. (AX100 or RX10-II) I'd love to try it with the .MXF wrapper but the X70 wont do 100Mbp/s and I dont have any other .mxf 4k camera.

Will post something this weekend. I think it will be a nice test to possibly put this all to rest. One way or the other.

Who knows? Maybe the "60Mbp/s is already the best in 4:2:0, 8bit long GOP and no more is needed" crowd is right!

Let's break them and see what happens!

CT ;-)

[David Heath]

Originally Posted by Cliff Totten

Can anybody actually say with a straight face that all that extra image data (40Mbp/s additional information) is "useless"...or "worthless"??

But is it extra information - or just extra bitrate?

Think of a word document and using a zip file to compress it. You'd expect the file size to decrease substantially wouldn't you?

So does that mean that it's therefore losing a lot of information in the zip process? The answer's no. In this case none at all. The file size is smaller but no information has been lost. Unzip the file and reopen and the document will be identical to the original.

It's the difference between lossless and lossy compression that's significant, and as I've been trying to say all along, don't draw too many conclusions from headline numbers.

For the record, my bet on such as this is that the 100Mbs sample is likely to be better - but not such a night and day difference as might be first expected. The question then is whether any difference is worth the extra overhead.

[Cliff Totten]

Well, ZIP works totally different. If you zip a book of text, you will get huge gains. What is rebuilt on the unZip is identical to the original text. 0% loss. Text and other types of data compress very well.

Try to Zip a .jpg or an .mp4 file. What happens? Almost nothing. The only way to compress "video" is to throw away image data to make it smaller. When it's uncompressed, it doesn't equal the same as the source. Very very different than zip.

Yup,..numbers alone don't tell all the story. 60Mbp/s on the X70 is not bad, I know this.

I'm actually curious to see a real world stress test between the two. Let's see how far we can push each one in a really complex moving scene.

I'm doing a side by side test between the PXW-X70 "XDCAM" at 60Mbp/s .MXF against the AX100 "HandyCam" at 100Mbp/s .MP4.

I have no evidence of this but I strongly believe they use the exact same encoder chip. It's the same h.264 codec inside two different wrappers.

It will be a good fight. Let's see how they fair this weekend.

[David Heath]

Originally Posted by Cliff Totten

Try to Zip a .jpg or an .mp4 file. What happens? Almost nothing.

Yes, true - a .jpg or an .mp4 file is already compressed.

Originally Posted by Cliff Totten

The only way to compress "video" is to throw away image data to make it smaller. When it's uncompressed, it doesn't equal the same as the source. Very very different than zip.

Not true. Not if you start with the uncompressed original.

You'll be able to compress *to a certain extent* without any loss at all (same as zipping) and then the uncompressed signal will be exactly the same as source. You're taking advantages of redundancies in the signal. Preserving all the information, but in a smaller file size. And if you take advantage of similarities between frames it works even better - which is why interframe compression will give better results than intra for a given bitrate.

Practically, it's still likely to end up too large a file to be practical. So to compress further then yes, you have to discard information. The trick of a good coder is to lose bits which are least important, and it's quite surprising how far you can go before it becomes noticeable.

But it's an exponential graph. No loss at all initially, then very little, even for quite big bitrate reductions, then eventually even a small decrease in bitrate will make a big difference. The trick is to choose the most sensible point on the curve. It'd be lovely to do as zip files do - as much as possible but remain lossless - but impractical.

[Cliff Totten]

OK,..just did a super quick...VERY quick stress test. Both files are the exact same zoom/crop.

These are two frame grabs. One set is 60Mbps and the other set is 100Mbps. Each set has 1 color and 1 Luma channel view only.

Let's see how good our eyes are. Can you tell which set comes from the 60Mbps frame grab and which set comes from the 100Mbps?

Let's all toss it around and speculate. In a while, I will reveal the answer.

Yes, yes, yes...I know that is quick and unscientific. A more accurate test will happen this Saturday. This one wont be frame grabs, it will be motion video, side by side. Cheap consumer Handycam 100Mbp/s vs. Professional XDCAM, 60Mbp/s.

For now with have this to toss around....

Hmmm...who has the best eyes here? This one is not easy. Jody, Noah, David, Ron, Craig and anybody else, what do y'all think?

CT ;-)

[Ron Evans]

My guess is one on the left is AX100.

Ron Evans

[Mark Watson]

I don't see any difference, imagined or otherwise. Pretty much what I found with my AX100 60vs100Mbps testing.

Mark

[Cliff Totten]

Fair enough. I can understand that....

Did you inspect the color removed images?

[Craig Seeman]

JPEG compressed files not the best way to compare. PNG or TIFF might have been better.