XAVC S vs XAVC L

[Paul Anderegg]

1080p60 4:2:2 10 bit 50Mbps vs 1080p60 4:2:0 8 bit 50Mbps.........wouldn't the "consumer" format have better picture and motion quality than the "professional" version if you don't do any color correcting or need all that unseen color information?

Just curious.

Paul

[Paul Cronin]

Paul, I am sorry but 8 bit 4.2.0 will not look better then 10 bit 4.2.2.

[Ron Evans]

I think Paul is asking the question of wheather the extra 20% compression required in the XAVC-L will leave the consumer version with a better picture ? If there is not a lot of colour but a lot of motion it may. Especial when final use will be 8bit. Maybe someone with AX100 as well as X70 can test.

Ron Evans

[Paul Cronin]

I did understand the question, but I will back out and let others discuss.

[Christopher Young]

I think Alistair Chapman argued the pros and cons of 8-bit vs 10-bit quite succinctly a while back. I kept a Word doc on his observations for future reference. I tend to think he is spot on on his analysis of the subject. Below is a 100% Quote from that document. The original link I cannot find at the moment. I trust Alistair doesn't mind me reproducing it here.

"Over the years there have been many, often heated debates over the differences between 8 bit and 10 bit codecs. This is my take on the situation, from the acquisition point of view. The first thing to consider is that a 10 bit codec requires a 30% higher bitrate to achieve the same compression ratio as the equivalent 8 bit codec. So recording 10 bit needs bigger files for the same quality. The EBU recently evaluated several different 8 bit and 10 bit acquisition codecs and their conclusion was that for acquisition there was little to be gained by using any of the commonly available 10 bit codecs over 8 bit because of the data overheads.

My experience in post production has been that what limits what you can do with your footage, more than anything else is noise. If you have a noisy image and you start to push and pull it, the noise in the image tends to limit what you can get away with. If you take two recordings, one at a nominal 100Mb/s and another at say 50Mb/s you will be able to do more with the 100Mb/s material because there will be less noise. Encoding and compressing material introduces noise, often in the form of mosquito noise as well as general image blockiness. The more highly compressed the image the more noise and the more blockiness. It’s this noise and blockiness that will limit what you can do with your footage in post production, not whether it is 10 bit over 8 bit. If you have a 100Mb 10 bit HD compressed recording and comparable 100Mb 8 bit recording then you will be able to do more with the 8 bit recording because it will be in effect 30% less compressed which will give a reduction in noise. Now if you have a 100Mb 8bit recording and a 130Mb 10 bit recording things are more evenly matched and possibly the 10 bit recording if it is from a very clean, noise free source will have a very small edge, but in reality all cameras produce some noise and it’s likely to be the camera noise that limits what you can do with the images so the 10 bit codec has little advantage for acquisition, if any.

I often hear people complaining about the codec they are using, siting that they are seeing banding across gradients such a white walls or the sky. Very often this is nothing to do with the codec. Very often it is being caused by the display they are using. Computers seem to be the worst culprits. Often you are taking an 8 bit YUV codec, crudely converting that to 8 bit RGB and then further converting it to 24 bit VGA or DVI which then gets converted back down to 16 bit by the monitor. It’s very often all these conversions between YUV and RGB that cause banding on the monitor and not the fact that you have shot at 8 bit."

It always interested me that Sony's 35-mbit 420 XDCam VBR MXF codec had lower compression than the 50-mbit 422 XDCam CBR MXF implementation, 16:1 as opposed to the 50-mbit's 21:1. It's all a pretty black art really. There are many things that influence the final picture outcome.

Chris Young
CYV Productions
Sydney

[David Heath]

Originally Posted by Christopher Young

It always interested me that Sony's 35-mbit 420 XDCam VBR MXF codec had lower compression than the 50-mbit 422 XDCam CBR MXF implementation, 16:1 as opposed to the 50-mbit's 21:1. It's all a pretty black art really. There are many things that influence the final picture outcome.

Well, yes to the last sentence ( :-) ) but be very careful with the XDCAM 35 versus XDCAM 50Mbs comparison.

I think you're basing what you say on summing values for luminance and chrominance samples and dividing by the datarate to get the compression ratios? I'm not quite sure about your maths, because the 4:2:2 signal has 33% more bits overall than the 4:2:0 signal, but about 40% higher datarate? But there's more to it than that anyway. (A lot more!)

For starters, then whilst the 4:2:2 signal may indeed have twice as many chroma samples as the 4:2:0, it *does not* therefore follow that it needs twice the chroma bitrate to encode to the same quality. That may not seem logical, but think of it this way. As simply as possible, if you've got the basic 4:2:0 chroma signals (one chroma pair for every four luminance sample) then you can make a guess at the missing samples of the 4:2:2 signal by simple interpolation. Hence to get the true values you only need correction figures for the values - not the values themselves. Hence you don't need twice the bitrate to encode the 4:2:2 chroma matrix as the 4:2:0.

The more you go into it, the more complex it gets, and just for example the compression applied to the chroma signals is likely to be higher than that applied to luminance. (And that is often the cause of chroma banding on saturated coloured backgrounds. Banding CAN be caused by insufficient bitdepth, but also by to heavy compression.)

Fundamentally, what you quote from Alister is correct. What Paul Cronin says ("I am sorry but 8 bit 4.2.0 will not look better then 10 bit 4.2.2") may be true *if the bitrate goes up proportionally* but if we are talking about the same bitrate in each case, then Pauls statement does not hold. You may indeed be better off (overall) with a 4:2:0/8 bit and lower compression than 4:2:2/10 bit with higher compression.

And a lot depends on the camera itself. If the noise level of the camera front end is relatively high, all that 10 bit will do anyway is more accurately define the noise! And if the chroma resolution is less than the luminance resolution (as is the case for Bayer sensors), then what's the point of recording 4:2:2 anyway? (We'll assume progressive - interlace brings in a whole load of other factors.)

[Paul Anderegg]

On my old Panasonic SPX800, when you switch from DVCPRO to DVCPRO50, things like red car tail lights at night stay bright red lit like neon cherries, with the 4:2:0 flavor, the bright centers would wash out a bit and loose color. The entire DVCPRO50 codec looked better, but it was running twice the bitrate, and had a 3:1 compression ratio as opposed to 5:1 for standard DVCPRO. That was the only prior experience I had with 4:2:2, and I do not see any of those positives in my new XAVC L codec as opposed to shooting AVCHD. I also do most of my shooting above 15db at night on my X70, and typically a lot of ClearImage zoom as well.

I wish they included an XAVC S option in the X70, I am sure a lot of people would like to have that available especially since most people can't ingest or edit XAVC L at this point in time.

Paul

[Christopher Young]

Paul ~

Just check our my latest post over on the other X70 thread. There is light on the XAVC-L decoding horizon, even to ProRes no less!

http://www.dvinfo.net/forum/sony-pxw...-ax100-39.html

Chris Young
CYV Productions
Sydney

[Christopher Young]

Originally Posted by David Heath

Well, yes to the last sentence ( :-) ) but be very careful with the XDCAM 35 versus XDCAM 50Mbs comparison.

Totally agree with all you say David. For years I have used both codecs and my preference is the 50-mbit MXF and that's why I still shoot on it. It's basically the default delivery standard here. Shoot and post in it and deliver final programs to network on XDCam disc in 50-mbit 422.

There again though we have included many miles of XDCam 1/2" disc camera 1440x1080 420 into those productions and to be 100% honest? I can't see the difference when it goes to air. Mind you we have tweaked the 1/2" sensors to the max from the handling point of view and were getting 460% highlight handling out of them with DCC on. But as for the rest of it it's all in the processing and delivery of the end product really.

I think the reasons for not seeing any differences are explained quite well here re end product delivery. PDF attached.

Chris Young
CYV Productions
Sydney

[David Heath]

Originally Posted by Paul Anderegg

On my old Panasonic SPX800, when you switch from DVCPRO to DVCPRO50, things like red car tail lights at night stay bright red lit like neon cherries, with the 4:2:0 flavor, the bright centers would wash out a bit and loose color. The entire DVCPRO50 codec looked better, but it was running twice the bitrate, and had a 3:1 compression ratio as opposed to 5:1 for standard DVCPRO. That was the only prior experience I had with 4:2:2, ...........

But remember that was comparing 4:2:2 colour encoding with the 4:1:1 of DVCPro - not 4:2:0

And 4:1:1 only has a quarter the horizontal chroma res cf luminance. I'd suggest that what you were seeing was less how good 4:2:2 was, but more how BAD 4:1:1 was! :-)

In the 50Hz world, then of the 25Mbs SD codecs only DVCPro was 4:1:1 - both DVCam and DV were 4:2:0 and the difference between 4:2:0 and 4:2:2 is far less pronounced, which I think is one reason why DVCam was far more successful in PAL territories than DVCPro.

(The whole reason behind 4:1:1 coding was because of interlace and the early days of digital video recording. It was foreseen there would be a lot of analogue-digital-analogue conversion through the chain. Because of interlace, it was seen as far more important to maintain full vertical chroma - even at the expense of savaging the horizontal chroma. As soon as the signal stayed more or less in the digital domain (esp interconnects being SDI rather than analogue) 4:1:1 ceased to make sense, and 4:2:0 was seen as far superior. Fortunately, (in PAL land) then although too late for the DVCPro spec, they were able to define 4:2:0 into the DV/DVCam specs.)

[Aaron Holmes]

Don't forget that the job of a video codec is, among other things, to find similarity between frames. Having a clean image is a huge part of enabling the codec to do that effectively and to produce the most efficient results. We've probably all seen noisy DV go completely to hell when encoded as MPEG-2 at DVD bitrates. With bandwidth-constrained codecs, noise isn't just an unwanted picture element, it steals bandwidth away from the other elements, making them look even lousier.

A 10-bit sample conveys more information than an 8-bit sample, but unless the corresponding photosites are well exposed, that "extra information" is really just noise. From the standpoint of the uncompressed source signal, that noise might represent two bits of wasted data per channel, but on the other side of the codec, whose ability to find similarity between frames is thwarted by that noise, those two bits of useless source information could be far worse than useless.

The short of it is: If you're not lighting the hell out of your shots and micromanaging your exposure, it's hard for me to see how 10 bits would help, and easy to see how it would hurt.

[Christopher Young]

A 10-bit recording if it's from a very clean noise free camera will have a very small edge over an 8 bit recording but in reality all cameras produce some noise and it’s likely to be the camera noise that limits what you can do with the images. A 10 bit codec has little advantage for acquisition unless the camera has an SNR of at least 60dB. Anything under an SNR of 60 means that in a ten bit codec the last one to two bits recorded will probably be mainly noise.

SNR for an image is defined as the ratio of the light detected on the sensor to the sum of the noise in reading that signal. SNR is expressed in units of power or decibels (dB).

SNR (dB) = 20log (Signal e- / Noise e-)

The maximum number of electrons (e-) that a sensor can collect within a photosite (pixel) is the full well capacity. The larger the pixel, the more electrons (e-) a photosite can hold. Hence the quietness of the Sony A7s with its 8 micron photosites. Now look at the Sony S7 it claims a SNR of 57dB (typical) in the Y luminance channel. So let's work out where it lands on the bit-depth to SNR scale.

Assuming the image could be read from the sensor with no other noise, your A/D converter’s resolution can be determined from the table attached.

Therefore a minimum 10-bit A/D would be required in the camera but note that the image you get will not be a true 10-bit image simply because your camera has a 10-bit A/D converter. In fact if your camera could only deliver a 52 dB signal to noise ratio your image pixel depth is not even 9 bits so the last bit of your 10-bit encoded signal is all noise.

Not many cameras can make the 60dB requirement without some kind of NR circuit. The Sony PMW-400 for example is claiming 56dB and only attains 60dB with its noise suppression circuit engaged.

Amongst TV engineers this is well known. Signal to Noise and Signal to Carrier Noise Ratios are some of the important things you learn about because it is important right through the transmission chain to maintain the lowest SNRs possible and the lower it is on the camera chain to start with the better. All things being equal on the receive end a peak signal to r.m.s. noise ratio of 45 db at the set is generally considered adequate to produce a good quality picture.

Chris Young
CYV Productions
Sydney

[Paul Anderegg]

I think my best solution will be to wait for the 4K XAVC S firmware upgrade, then downres to my final HD output. Will be sad to loose 60p, as I do a lot of moving scenes and my footage airs 720p60. I am hoping the 8bit codec as well as the downsampling will allow less noise at the high gain settings and Clear Image zoom I use often.

Paul

[Ron Evans]

You may not have Clear Image zoom with 4K as it works by cropping into the sensor until it reaches 1920x1080. 4K will need all the sensors pixels. The X70 is already downsampling to 1920x1080 so I do think the noise will not improve much either.

Ron Evans