FS5: Re: PXW-FS5 Presentation by Alister Chapman

E

ectobuilder

Veteran
Re: PXW-FS5 Presentation by Alister Chapman

A very interesting, in-depth, technical and experiential presentation on the PXW-FS5 by Alister Chapman.

I really enjoyed this presentation!
https://www.youtube.com/watch?v=U-HrXmRqpME

On the technical side he dives into the 8-bit vs 10-bit debate and things of that nature. Very eye opening.

EDIT: I love the fact that you can manually "push" the auto focus into a different plane WHILE recording! He demonstrates it using multi-face tracking. And in this case each plane is the face. 30:59
 
Last edited:
i watched that last week and wow was it informative. learned a lot about the FS5 but also learned a lot about things that related to all video cameras. a must watch IMO.

david
 
Interesting watch.

One always need to be able to translate with Alistair.. or I do.

For example 8-10bit - he says 'no banding' then 'not so good for slog'.. which to me translates as.. 'lots of banding'

So which is it? 'no banding' or 'lots of banding'

I assume anyone is going to shoot Slog - he doesnt. Why assume that? Because in all but the most controlled situations you need latitude.

Interesting there is a kind of LUT, I wonder if it works in all modes..(HFR) in which case a big step from the FS7

My overriding question with the camera is 'will it make pretty HD' in a small practical package?

Does this video answer that question? not really!

S
 
I think the 8 bit is to do with the amount you can grade.. rather than just banding .. i.e. 16 million shades V,s 1 billion.. and so to shoot Slog in 8 bit isnt advised .. ?
 
Well banding is the test of grading.

Grading a subtle tone (sky, smooth wall) etc shows up the trouble that you may not see on a more textural portion of the image.

Ali says that 8bit is problematic due to typically being mixed with compressed codecs.. (rather than just compromised by design).. I think it is just problematic in general for anything but nailing perfection in camera.

As you say a load less colours!

S
 
Sure banding can also occur in grading.. it just depends how much you need to ,or want to...pull or push your footage .. and that you can get away with alot more in 10 bit than 8 bit.. as simple as that I thought anyway.. Im still "telling" prod companies that want to shoot log that XAVC is better.. they check with their post..(because they dont know what Im talking about) and yes they then agree XAVC is better.. but even this year alone I dont have to do that much anymore.. as XAVC is getting more established.. I mean why would you shoot 8 bit when you can shoot 10 bit Log anyway.. I suppose would be the question..
 
I was really just commenting on his 'wooly' presentation.

To me it is clear.. 8bit you must shoot 'in camera' (and can make small corrections if required) and 10bit you are starting to explore the ability to properly move footage in post (with 12bit you are really motoring) and shooting Slog implies a big move in post and therefore, to my head, requires 10bit.

With a flat image you need 'in between' tonal values to be able to even normalise it.

I come from stills where one would never shoot a 'baked codec' professionally - it is just too inflexible.

To me 10bitSlog is therefore the minimum quality that can really be described as professionally usable.

S
 
The seems to be very little difference between 10 bit and 8 bit output of the FS7. The damage is done by the XAVC codec to the subtle tones which is the biggest limitation to grading because large areas of the image have been posterised by the compression.

The irony is that the FS100 AVCHD codec that allows much more contrast to be put into the skies than FS7 XAVC because there's less chunky posterisation.

I Sony allotted more bits to the areas of subtle tone e.g. skies I doubt many would need RAW and you could grade like a demon.

Sony are very very savvy.
 
"To me 10bitSlog is therefore the minimum quality that can really be described as professionally usable."

For Log yes I agree sir.. .. other types of shoots XDCAM 422 HD is still a pretty good codec.. as long, as you say . your exposure is pretty much on the nail.. and its not going to be "improved" by demented dir or editors with a new bit of software.. or a hatred of camera persons..

Personally I quite like Big Al,s delivery ..
 
Upon the arrival of the Sony F3, Alister has always stood by not using 8-bit to record S-Log because you need to bring the gamma back to normal (de-log) the image, and after that possibly cc or grade the image and so overall there just aren't a lot of bits to work with. He does say however that if you do shoot S-Log in 8-bit you need to get the exposure bang on because there is less latitude to work with.

If you're going to try shooting S-Log2 or S-Log3 in 8-bit my suggestion would be to use a grey card and WFM or learn how to calibrate your camera to a light meter and you won't have to fuss with any grey cards.

The FS5 is somewhat similar to the PMW-F3 in the sense that the F3 was also only 8-bit internal and featured four CineGammas who's middle grey points were mapped on different points of the curve so you could allocate data differently on the curve if desired. This can have some benefits later in post, even if only 8-bit. So with the FS5 I would probably opt to shoot using the cinegammas and forget S-Log all together.

Here's an image from the F3 I shot using one of the cinegammas, 8-bit, 420.

8092661728_a0765892ae_c.jpg
 
morgan_moore said:
Interesting watch.

One always need to be able to translate with Alistair.. or I do.

For example 8-10bit - he says 'no banding' then 'not so good for slog'.. which to me translates as.. 'lots of banding'

So which is it? 'no banding' or 'lots of banding'

My overriding question with the camera is 'will it make pretty HD' in a small practical package?

Does this video answer that question? not really!
Click to expand...

He specifically talks about his REASON behind the banding. He said it is incorrect to say that 8-bit is the direct cause. He points to the fact that ALL 8-bit codecs are found in consumer level products. And this is the reason.

He also said that 10-bit codecs are NEVER in consumer products. So it is the codec used that causes 8-bit.

According to Alistair, 8-bit has a legacy with consumer grade level that can't be shaken. So 8-bit can avoid banding if only the manufacturers gave it better compression.

I believe he was specifically talking about banding straight from the camera. Yes later on he talks about the behaviours of different gamma / log curves and talks about how 8-bit on certain parts of certain curves allow only 2 shades of information which affects processes like grading. But I believe banding was a different issue he was trying to address.


morgan_moore said:
My overriding question with the camera is 'will it make pretty HD' in a small practical package?

Does this video answer that question? not really!
Click to expand...

He seemed to really focus on the 4K aspect instead of HD. My impression is that the HD out of the PXW-FS5 is so perfect that if he talked about the HD he wouldn't be able to get into the discussions of do's and don't's (i.e. the do's and don't's of gamma and log curves).
 
Exactly my critique - his REASON - is wrong (but may be a contributory factor)
8bit cannot represent enough tonal levels for a flat codec to work when graded due to simple maths that 256 levels is not enough reference points to contain the data needed to delog an image - one needs over sampling - 1000 levels or more. Even uncompressed 8 bit will band if you look hard enough.

As for the perfect hd. Well the compression is twice as high as the fs7 - if that comes at no visual cost if be very interested to know how. The hd on the fs7 is far from perfect - put it up against a red raw file or sr444 from the f5, or even hd made from down sampling fs7 4k, and you will see that. Good yes. Perfect no. Half the data rate from the fs7 and one may or may not drop below the bar that I call pretty.
 
Last edited:
morgan_moore said:
Exactly my critique.
8bit cannot represent enough tonal levels for a flat codec to work when graded due to simple maths that 256 levels is not enough reference points to contain the data needed to delog an image - one needs over sampling - 1000 levels or more. Even uncompressed 8 bit will band if you look hard enough.

As for the perfect hd. Well the compression is twice as high as the fs7 - if that comes at no visual cost if be very interested to know how. The hd on the fs7 is far from perfect - put it up against a red raw file or sr444 from the f5, or even hd made from down sampling fs7 4k, and you will see that. Good yes. Perfect no. Half the data rate from the fs7 and one may or may not drop below the bar that I call pretty.
Click to expand...

What HD from the FS7? From what I know, the FS7 can shoot a variety of codecs
and bitrates in HD......enough that a statement like
'the HD on the FS7 is far from perfect'
is almost confusing to me. What form of HD and what bitrate?
Im sure none of them are perfect, but some are closer than others.

As for the FS5, I think it meets the 'magic, BBC approved' bitrate in HD right?
 
It would never occur to me to shoot below the highest quality settings. So the best setting 10bit xavc.

50mbs hd may be bbc approved (for non log) - but that doesn't make it pretty.

i can tell you the 50mbs setting in the fs7 is far from pretty. I've shot a 60 min bbc doc on that setting so might know.
 
Is the Fs7 XDCAM 422 HD different from the f5 is some way..? Ive got totally acceptable images from XDCAM 422 HD.. on the F5.. surprised that the Fs7 would be a far lesser image..? Maybe Sony keep it all a secret .. I wonder if anyone knows.. that would effect buying it as a B cam.. ?
 
Donny 123 said:
Is the Fs7 XDCAM 422 HD different from the f5 is some way..? Ive got totally acceptable images from XDCAM 422 HD.. on the F5.. surprised that the Fs7 would be a far lesser image..? Maybe Sony keep it all a secret .. I wonder if anyone knows.. that would effect buying it as a B cam.. ?
Click to expand...

I guess it is the same - maybe we have different 'bars' of 'acceptability' - I prefer codecs that will take a deep grade or save my ass if I cock up on set.

I don't go on client or corporate specs just my own perception of 'pretty' which is entirely aribitary and I know it when I see it. Red, Alexa, SR444 et al. are pretty, FS7 4k down to HD is pretty, FS7 HD is OK, FS7 HFR is marginal, the FS5 is an unknown. Pretty is not numbers based - I would clearly acknowledge that the C300i is quite pretty - way prettier than the FS700 onboard for example.

Alistair always opens his seminars with some pretty footage.. shot on his F5 in raw mode.

The HD from the FS5 is different from the HD on the FS7 in being long gop as opposed to i frame and half the bit rate (afaik).

S
 
Last edited:
Sure XAVC i is always better.. but as a freelancer I cant always have the final say.. except in Slog/proper grade mode,where i would also push hard for XAVC.. or have it in writing that it was under duress if i ever had to shoot XDCAM Slog.. so far not happened..!!

But I have to say for many shoots I shoot XDCAM ,Cine EI and burn in a user LUT and it looks really good.. to my eye anyway..
 
morgan_moore said:
I guess it is the same - maybe we have different 'bars' of 'acceptability' - I prefer codecs that will take a deep grade or save my ass if I cock up on set.

I don't go on client or corporate specs just my own perception of 'pretty' which is entirely aribitary and I know it when I see it. Red, Alexa, SR444 et al. are pretty, FS7 4k down to HD is pretty, FS7 HD is OK, FS7 HFR is marginal, the FS5 is an unknown. Pretty is not numbers based - I would clearly acknowledge that the C300i is quite pretty - way prettier than the FS700 onboard for example.

Alistair always opens his seminars with some pretty footage.. shot on his F5 in raw mode.

The HD from the FS5 is different from the HD on the FS7 in being long gop as opposed to i frame and half the bit rate (afaik).

S
Click to expand...

These things are always more complicated than pure numbers, I totally
agree with you there. But that also extends to the codecs.......long gop.....
sure in this case it's half the bit rate of I frame but does that automatically
mean it's worse? Sometimes we get so caught up in the numbers on these
things. Remember DV? Took 13 gigs to store an hour of it, BECAUSE it was
an I frame codec. And AVCHD from say the C100 or FS700? It could store
about that same hour with that same 13 gigs.....but in 1080HD and much
better quality than DV. Long gop isn't always bad....or maybe I should say it this way.
Everything else being equal, I frame is probably going to be better
quality than long gop.....but with technology advancing so fast, all
other things are rarely equal. Could a newer long gop codec at half
the bitrate look as good as an older all I frame codec? I mean isn't this
'Better video in a smaller package' almost have to be a given if we ever want
4k to work for broadcasters? I mean I don't think it's here yet either....but
on the other hand, I'm not going to be surprised when codec advancements again
make higher resolution (4k) video fit in the same space as 1080hd. Or
alternatively get the same quality HD video at half the bitrate.
 
Ask the guys shooting DSLR's with flat picture profiles if their 8-bit consumer camera's are yielding acceptable results for professional work. I think their responses may be shocking.
 
Also worth noting that compression techniques have moved on immeasurably in the last few years and so simple data rate comparisons between different codecs don't have much relevance.

likewise raw dng of the fs7 over XVAC-i is huge by comparison with little in the way of increased image quality.
 
You must log in or register to reply here.
Back
Top