Media Size

[AndreasOberg]

How does the media size compare to DSMC2 and the compression quality.
I know that you can film with around 800MB/s. In Helium I'm usually at 8k 60fps that is using 240MB/s if I remember correctly (or was it 270MB?)

What is the data rate at 60fps 8k for example?

Also how is the quality vs Helium. Anyone that has studied in detail?
I did careful studies of the compression levels in DSMC2 and you could clearly see the difference between 1:3, 5, 8, 12 etc.

We often film in the wilderness and I store data on SSD disks so if the data is double or even tripped, that will result in many more harddrives.

Cheers,
Andreas

 

[Phil Holland]

Several points I'll make Andreas.

V-Raptor uses a different encoding scheme, so the difference between the REDCODE RAW compression ratios aren't as "noticeable" as the major gabs previously. You can see a difference between HQ, MQ, and LQ. But it's nowhere near the difference between 3:1, 8:1, and 12:1 for instance.

You've quoted the maximum data rate, which you can hit particularly when filming high frame rates. Like 8K 120fps LQ or 8K 96fps MQ.

At speed you for 24fps you are looking at:
8K HQ 365MB/s
8K MQ 298MB/s
8K LQ 170MB/s

Traditional feature and series compression ratio for DSMC2 is around 5:1, which hit about 260MB/s. So basically you have a data rate that ranges from higher, to slightly higher, to lower than DSMC2 in that range.

Versus Helium. Won't get too brutally honest here, but if you understand the difference between Helium and Monstro, V-Raptor is better than that even. Cleaner and more colorful shadows, more DR, etc.

The biggest leap for you will be the FOV difference between S35 and VV between Helium and V-Raptor. If you are shooting long glass in nature it will be wider on V-Raptor if it covers VV.

Data rate calculators and format comparison tools are up on phfxtools.com if you want to poke around.

 

[AndreasOberg]

Hiya Phil, always a pleasure. I poked around on your excellent site. Used it several times over the years! for MQ at 60fps it says that the data rate would be 744MB. I do not think you can shot in HQ in 60fps, right?
DSMC2 says 250MB for 13:1 which I normally shot with in Helium.
So that is about 3x larger if I picked the right quality mode. But then 13:1 is pretty brutal compression so MQ is probably better, or? Just have not seen any visual comparisons.

Yea, I have used Monstro, remembered it was about 1.7 stops better than Helium and created a more 3d image.
However one problem for me with Monstro is we mostly use our CN20x50 and it covers about 5.5k-6k on Monstro (at 1500mm it covers 8k though which is cool).
So not so happy about stepping down to 5.5k.

However I think what we will do a bit later is replace our second camera that does not use CN20x50. We usually have 2 cameras. Could then use it for macro, Canon 2x600mm lens, gimbal and for diving etc where the lower noise and slomo would be super useful. Raptor would still be useful at 1500mm though since then you have crazy little light so we always fight with noise.

I do have a feeling that if we get the Raptor we will use it as our a-cam and just accept the lower resolution. At lower resolution we also get more slomo which is always very welcome.
Best,
Andreas

 

[Phil Holland]

60fps

Yep. That's pretty much one of the points of V-Raptor is increasing the quality at higher frame rates, at least how I think of it. 8K 60fps can be done at HQ, but you can also set it to MQ or LQ if you desire.

Data at 60fps looks like this, and this is approximate.

V-Ratpor @ 60fps 8K 17:9
HQ = near 850-900MBs
MQ = Around 740MB/s
LQ = Around 423MB/s

Monstro's highest data rate is 12:1 at 8K 17:9 and is 270 MB/s.

The new demosaic RED just implemented improves all cameras at those higher compression levels, but V-Raptor more or less is designed to "look better at any frame rate" comparatively. That improvement comes at the cost of higher data rates. This has been a topic of discussion as some have been shooting well beyond 12:1 even in the past and others have been looking for increased quality. Everybody here knows which side of sword I die by especially in comparison to other systems.

I'd take a close look at V-Raptor at 5K and 6K with those long lenses. Just did a project going between 5K, 6K, and 8K for a variety of reasons and it all looked really good. 5K 17:9 is truly about 3-perf S35 on V-Raptor and you get basically 2X the frame rates there versus anything DSMC2 at increased quality as well. 5K crop on Helium is a bit severe for some too. V-Raptor at 6K, though a bit larger, is about the same format size of Helium at 8K.

 

[AndreasOberg]

Yep. That's pretty much one of the points of V-Raptor is increasing the quality at higher frame rates, at least how I think of it. 8K 60fps can be done at HQ, but you can also set it to MQ or LQ if you desire.

Data at 60fps looks like this, and this is approximate.

V-Ratpor @ 60fps 8K 17:9
HQ = near 850-900MBs
MQ = Around 740MB/s
LQ = Around 423MB/s

Monstro's highest data rate is 12:1 at 8K 17:9 and is 108MB/s.

The new demosaic RED just implemented improves all cameras at those higher compression levels, but V-Raptor more or less is designed to "look better at any frame rate" comparatively. That improvement comes at the cost of higher data rates. This has been a topic of discussion as some have been shooting well beyond 12:1 even in the past and others have been looking for increased quality. Everybody here knows which side of sword I die by especially in comparison to other systems.

I'd take a close look at V-Raptor at 5K and 6K with those long lenses. Just did a project going between 5K, 6K, and 8K for a variety of reasons and it all looked really good. 5K 17:9 is truly about 3-perf S35 on V-Raptor and you get basically 2X the frame rates there versus anything DSMC2 at increased quality as well. 5K crop on Helium is a bit severe for some too. V-Raptor at 6K, though a bit larger, is about the same format size of Helium at 8K.

If the quality is quite a bit better than say Monstro with the same data, then I'm all for it. We will just have to get a lot more harddrives in the field. I think that you are right that 5k Raptor vs Helium 8k would be interesting. You would prob miss a bit of detail, but you would have far less noise of corse and the higher frame rates. We often film with the 1.5x extender on, and then it covers full Vista Vision, but the F value is a very dark f1/14 or so so even in sunlight the light is not always enough.
Currently I use mostly Zeiss Otus 28, 55 and 85 with gimbal and other shots that do not require longer lenses and they cover well of course. May in the future we will get the Zeiss Cine lenses as well, they seem great.

 

[AndreasOberg]

V-Ratpor @ 60fps 8K 17:9
HQ = near 850-900MBs
MQ = Around 740MB/s
LQ = Around 423MB/s

Monstro's highest data rate is 12:1 at 8K 17:9 and is 108MB/s.

Do not think Monstros data rate is correct. Should it not be 271.05 MB? (that is what your webpage says and what I remember as well. I think you are using 24fps at 12:1 maybe
But is not the new compression less efficient as well?

 

[Phil Holland]

Ah yes, corrected my post. Good catch.

The new compression. Interesting word efficient. Efficiency-wise it's allowing the camera to cost less while doing more and using less power and generating less heat.

It's a different encoding scheme. Discrete Cosign versus Wavelet. There should be subtle improvements in micro contrast with the new encoding, but DCT will never go to places like 20:1 as it will fall apart. And honestly most with WCT shouldn't have been filming beyond 12:1 maybe to 15:1 at most.

RED needs to compete with the highest end cameras on an image quality level and what's going on now is they are putting people within a smaller box of compression ratios that all yield high quality results. And for DPs who didn't like the semi-penalty in a dip in IQ for higher frame rates, that's virtually been eliminated now. And the file sizes are still much smaller than they should be IMO for what we're getting.

My unpopular, though based and informed opinion. Acquisition codecs, particularly RAW ones, will increase in size while delivery codecs will be smaller over time. I've already completed a full round of testing on two next gen streaming codecs and they both result in data savings from my testing in the range of 25-40% at equivalent 4K and 8K resolutions and frame rates. This will be apart of a several stage upgrade to streaming quality in general. But first you'll likely see increased datarates tied to H.265 and VP1 first while all this goes down.

Cameras are still increasing in quality in a variety of different directions and the square peg into square hole logic is data rates will go up as part of that.

Keeping in mind on the film side we were doing everything 16-bit uncompressed when digitizing film first and everything else went from there. Early digital was confined greatly by the cost and technology available during each specific era's technology base. Similar to where we are now, but things have moved forward a bit compared to 5-10 years ago.

I do wonder if we'll ever see RED do something wild like H.265 log encoding as an in camera format for some types of work, but nobody would use that for cinema/feature application. Not by a mile. Streaming too, but I could see reality series using it as they do already.

 

[Brian Timmons]

Ah yes, corrected my post. Good catch.

The new compression. Interesting word efficient. Efficiency-wise it's allowing the camera to cost less while doing more and using less power and generating less heat.

It's a different encoding scheme. Discrete Cosign versus Wavelet. There should be subtle improvements in micro contrast with the new encoding, but DCT will never go to places like 20:1 as it will fall apart. And honestly most with WCT shouldn't have been filming beyond 12:1 maybe to 15:1 at most.

RED needs to compete with the highest end cameras on an image quality level and what's going on now is they are putting people within a smaller box of compression ratios that all yield high quality results. And for DPs who didn't like the semi-penalty in a dip in IQ for higher frame rates, that's virtually been eliminated now. And the file sizes are still much smaller than they should be IMO for what we're getting.

My unpopular, though based and informed opinion. Acquisition codecs, particularly RAW ones, will increase in size while delivery codecs will be smaller over time. I've already completed a full round of testing on two next gen streaming codecs and they both result in data savings from my testing in the range of 25-40% at equivalent 4K and 8K resolutions and frame rates. This will be apart of a several stage upgrade to streaming quality in general. But first you'll likely see increased datarates tied to H.265 and VP1 first while all this goes down.

Cameras are still increasing in quality in a variety of different directions and the square peg into square hole logic is data rates will go up as part of that.

Keeping in mind on the film side we were doing everything 16-bit uncompressed when digitizing film first and everything else went from there. Early digital was confined greatly by the cost and technology available during each specific era's technology base. Similar to where we are now, but things have moved forward a bit compared to 5-10 years ago.

I do wonder if we'll ever see RED do something wild like H.265 log encoding as an in camera format for some types of work, but nobody would use that for cinema/feature application. Not by a mile. Streaming too, but I could see reality series using it as they do already.

Nice post.
The interesting thing with Wavelet and DCT is how they have contrasting sweet spots and pain points.

I have a feeling RED saw an eventual dead end or limitation with Wavelet that they thought would be a handicap down the line particularly with higher resolutions and frame rates.
Perhaps from their point of view, higher processing would be akin to fighting against the river where embracing higher data rates would be akin to flowing with it.

As pointed out, it does make sense that RED would put a greater emphasis on tuning REDCODE for higher IQ particularly if that's the needs of the customer base.

I still think a 4th more compressed DCT flavor would help the current storage pain point some are having particularly on documentary shoots.

Brian Timmons
BRITIM/MEDIA

 

[Mark Andersen]

I still think from a PR perspective calling the three choices low compression, medium compression and high compression is better that low, med and high quality. I don’t like telling a client ‘we are going to shoot in Low Quality today for our slow motion shots.

 

[Phil Holland]

I still think from a PR perspective calling the three choices low compression, medium compression and high compression is better that low, med and high quality. I don’t like telling a client ‘we are going to shoot in Low Quality today for our slow motion shots.

The other reality, in DSMC3's case, is "it doesn't matter" what you are shooting in. There is a difference in the finite details, subtle textural stuff too, and yes I am an HQ roller, but you would be pretty hard pressed to A/B/C them shot to shot to identify the difference.

I think that's the whole point of this new direction really.

 

[Steve Phillipps]

Yea, I have used Monstro, remembered it was about 1.7 stops better than Helium and created a more 3d image.
However one problem for me with Monstro is we mostly use our CN20x50 and it covers about 5.5k-6k on Monstro (at 1500mm it covers 8k though which is cool).
So not so happy about stepping down to 5.5k.

Hi Andreas, long time no speak. I have the CN20 on my Raptor so give me a shout with any specific questions.
From my testing it just covers 6k 16:9 (a little soft in corners) and 8k 16:9 with the 1.5x engaged. To my mind it makes sense to use it even if it does tail off a little in corners as you often tend to crop in a little for composition an / or stabilisation with wildlife stuff anyway...so it would be better to start from a higher resolution and allow for a 10% crop in rather than lowering the shooting resolution. Be interested in other thoughts though!

Steve