FS5 - R U shooting 25/30p or 50/60p?

[Piotr Wozniacki]

Interesting discussion. Let ask you a question unrelated to compression, data rates and other thinks influencing PQ differences between shooting 25p vs 50p, but having to do with shutter speeds/angles, and available amount of light differences between the two framerates.

Suppose I'm shooting 25p, and - due to poor light - pick 1/25 shutter instead of 1/50 (which is the same as shooting with 360 deg rather than the "right" 180). While not quite "by the book", this can give you 1 stop of additional light so you don't need to use as much gain... Now suppose I'm in 50p mode; on FS7 at least I can - for the purpose described above - go from 180 deg (1/100th) shutter to 360 (1/50 the), and gain the same full stop of sensitivity; but I can also go as low as 1/25th! Now, it terms of angle, what is this shutter speed really: 720 deg?!!!

I wonder what your answers will be...

[Noa Put]

If I am at a controlled shoot I use the "proper" shutter but if I am at a wedding I shoot at 1/50 of a shutter with 50p all the time, I honestly don't see the difference + I gain some sensitivity. 1/25th of a shutter I only would use in extreme conditions when I absolutely need it but else just stay clear from it.

[Ron Evans]

Shutter angle is hang over from film cameras. Video cameras do not have a mechanical shutter. They sample the sensor so it is time not shutter angle that the camera processing uses. The net of the discussion is motion blur for a particular frame rate. How much do you want the image motion to blur. If you want it pin sharp then use a fast shutter speed. Problems comes with fast shutter speed and slow frame rate in that the judder is emphasized. Hence the practice of 180 shutter for film. You can still mess it up by moving the camera though !!! I shoot all my indoor stuff at 1/60 at 60P.

Video camera terms are a mix of still and film cameras. They use shutter speed ( from still cameras ) and gain rather than ISO ( film speed from both still and film cameras ). So how long do you let the sensor collect light and then how much do you amplify that reading.

Ron Evans

[Dave Sperling]

Personally I'll normally shoot with the equivalent of a 180 degree shutter when shooting progressive. Since I'm normally shooting at 30p that means my shutter is at 1/60th. I find the image a little 'smeary' if I keep the shuter off. Obviously this is not much of an issue if there's no movement in the frame, so if extra exposure is needed for a static landscape at dusk/night going with a slower shutter can give more exposure. For sit-down interviews where there's minimal movement it's also not so much of an issue. But when shooting performances or events we always use the shutter, and normally at twice the frame rate.
When shooting slow motion I really like the crispness you get from a faster shutter speed, so when shooting at 120fps I'll make sure the shutter is at 1/240 or faster.

On the long-op discussion, my feeling is that if you are recording twice as many frames (and twice as many I-Frames) into the same bit rate there must be at least a small quality loss -- which would be greater with more camera movement combined with fine detail, or less with static images. Whether you will see the difference is another question... I was always very happy with the 35 megabit (HQ) image quality from my EX1. But if I wanted to do green screen or extreme motion I'd record to a NanoFlash at 100 megabit long-gop. Yes, that's a lot of extra data for a very slight (or sometimes imperceptible) improvement in image -- but in certain cases it was worth it... The main question is how good the 60p compression is to begin with, since slight improvement in image quality may not be detectable.

One more compression story, going a few years back -- I always thought the SxS images recorded to the F3 were really nice, but then I did a test comparison with s-log between the cards and an external uncompressed 4:4:4 recorder for a theatrical feature film. We viewed the results in a color correction suite (sitting front row looking at the image on a 16'x9' screen, projected with a 4k projector with color correction through an Assimilate system). The quality difference was far more than any of us had expected. So if you're thinking about what's appropriate to use, always think in terms of where the final result is going to be seen.

[Doug Jensen]

+1
Well said, Dave.

The only thing I disagree with slightly is the 180 degree shutter. I find it too slow, thus resulting in too much motion blur for my taste. For 24P I prefer 1/60th, and for 30P I prefer 1/80 if I have a camera that can do it.

[Nigel Davey]

OK, in light of Sony confirming no loss of quality in 50P, I've been thinking about shooting in it all the time. But here are some of my thoughts over a workflow that I'd welcome feedback on.

When editing in Premiere CC and using a FS5 XAVC HD 50P clip to automatically create a sequence, Premiere always generates a 'AVCHD 1080P square pixel' sequence with (unsurprisingly) 50 frames per second. If I do the same with 25P footage Premiere generates a 'Arri Cinema' sequence with 25 frames per second. After doing various tests relating to render and export times the bottom line is 50P is more CPU intensive than 25P, by about (again no surprise) 50%. Initially I wondered if the type of sequence (i.e. AVCHD vs Arri) had any bearing on results re the sequence codecs, but I don't think it does since you can go into the sequence settings and turn an Arri 25fps sequence into a 50fps and all of my test results will then pretty much match their AVCHD equivalents.

So my conclusion here is 50P requires approx. twice as much CPU time as 25P regardless of the sequence codec (at least between Premiere's auto choices for XAVC HD).

So then I started to wonder if Premiere can retain access to all of the 50P frames on a 25P sequence in much the same way you can scale up a 4K clip (and use the resolution above 1080) on a 1080 sequence in Premiere? Alternatively does it throw away every other frame on a 25P sequence and you can't get it back without generating a 50P sequence?

If the former is true you would be able to edit most of your 50P footage on a 25P sequence (avoiding heavy CPU usage) and then for those odd clips where you want to slomo them (to 50%) Premiere would still have access to all 50 frames of the 'original' clip and thus can stretch them on your 25P sequence producing smooth 50% playback.

Unfortunately my edit PC is away for repair and so I can't test this until it comes back. But maybe someone else can?

[Noa Put]

If the former is true you would be able to edit most of your 50P footage on a 25P sequence (avoiding heavy CPU usage) and then for those odd clips where you want to slomo them (to 50%) Premiere would still have access to all 50 frames of the 'original' clip and thus can stretch them on your 25P sequence producing smooth 50% playback.

I always edit 50p footage in a 25p timeline in Edius and slowing down the footage 50% gives smooth motion, this should also apply for Premiere.

[Dave Sperling]

Originally Posted by Nigel Davey

This is how Sony responded:

"That way of explaining would be correct in an Intra codec. However, since you record in XAVC Long GOP, there is no such "double quality" in a 25p frame compared to a 50p frame"

Originally Posted by Nigel Davey

OK, in light of Sony confirming no loss of quality in 50P, I've been thinking about shooting in it all the time.

This may just be semantics, but I'm not sure that the Sony statement says 'No Quality Loss' -- I would interpret it to mean 'minimal quality loss for most shots.'

I think the 'double quality' reference really goes back to arguments about the original Panasonic Varicam 100mbps recording system (which here in the States recorded at 60 fps intra) -- where you were essentially throwing away 50% of your frames at 30p, or 60% of your frames at 24p -- and thus your effective usable compression at 24p was actually 40mbps, not 100mbps.

I'd be careful about inferring that you aren't losing any quality at 50p or 60p -- since as mentioned elsewhere in the thread Long-GOP is most efficient when there is minimal change between frames, but the flip side of this statement is that when there is significant change between frames the Long-GOP is not as efficient.

As a side note -- one of my clients has a bunch of XDcam disc cameras which they have standardized for many years at 25mbps (SP) compression. (It's a well known syndicated tv show that airs both in the US and internationally, and they generate more than 250 hours of programs a year). For them the quality difference between the 25mbps Long-GOP and higher bit rates was considered minimal when compared to their savings in discs and data storage. That was their decision, based on their needs and perceptions. Meanwhile, for any of my other clients, 35mbps (XDcam 4:2:0 HQ) would be the absolute minimum data rate they would accept, with many requiring 4:2:2.

So -- I'd still recommend doing your own tests to determine whether the higher frame rate is better for you -- based on what you'll be shooting and where it's being seen... I just can't construe the Sony statement as meaning 'Absolutely no potential quality loss' at the lower effective bit rate.

[Ron Evans]

Originally Posted by Dave Sperling

....

I'd be careful about inferring that you aren't losing any quality at 50p or 60p -- since as mentioned elsewhere in the thread Long-GOP is most efficient when there is minimal change between frames, but the flip side of this statement is that when there is significant change between frames the Long-GOP is not as efficient.

....

FS5 is all LongGOP so there are less incremental data needs when recording 60P than at 30P. But there are twice as many data points. At 100Mbps I doubt there is much difference but suspect that 60P is still less and consequently a better choice. This is just an argument for LongGOP it definitely does not apply for iFrame codecs where there is twice the difference. If you follow your argument it is 30P that will have the quality loss ?

Ron Evans

[Dave Sperling]

Originally Posted by Ron Evans

FS5 is all LongGOP so there are less incremental data needs when recording 60P than at 30P. But there are twice as many data points. At 100Mbps I doubt there is much difference but suspect that 60P is still less and consequently a better choice. This is just an argument for LongGOP it definitely does not apply for iFrame codecs where there is twice the difference. If you follow your argument it is 30P that will have the quality loss ?
Ron Evans

I'm going on the assumption that the length (in frames) of each long-gop sequence will be the same no matter what the frame rate -- hence when shooting at 60p you will be recording twice as many 'GOPs' as at 30p in the same length of time. But recording bit rates will be based not on number of frames but on time. My guess is that the codec would be pushed harder when shooting motion-intensive subjects with plenty of change every frame, no matter what the frame rate (sports action, waves breaking, sparklers, corn flakes pouring, pans of large groups doing calisthenics, etc) at 60p than at 30p if recording at the same bit rate -- since there would be significant changes in every frame, and you are still utilizing less data per GOP sequence at 60p than at 30p, since you have twice as many GOP sequences. So no... I would expect the 30p to look marginally better, depending on subject...

Thus my suggestion to test with the type of footage you will be shooting.

[Ron Evans]

I guess my point is that at the rates we are taking about the motion changes at 60P is a lot less than at 30P so although there will be more of them each change is a lot less load on the encoder than the changes at 30P for the fast action you mention where there will be also judder involved in any camera movement at 30P. If you were talking about lower data rates there may be a difference but at 50Mbps I doubt anyone will see a difference. Clearly there is a data rate at which the encoder just cannot manage to encode. However we are talking about 50Mbps not the lower limit. The data rate differences at AVCHD for 30P and 60p may give some indication of these lower limits. 24Mbps for 30P( or 60i same sampling rate as 60P but half frames ) and 28Mbps for 60P. So to make up the full frames( 60i to 60P ) requires another 4Mbps for AVCHD. Both 24Mbps and 28Mbps are max levels too not the average. XAVC-S is more efficient and at 50Mbps.

Ron Evans

[Tom Van den Berghe]

Originally Posted by Noa Put

I export my videos in 25p as well to vimeo and the noise you see in my videos on vimeo are there in the native files, it's because I"m not shooting with a c100 or a7sII. :)



I shoot at 50fps at 1/50 shutter so the same amount of light comes in but with extra benefits, there is no light loss in this case and again, show me one shot in above video where shooting at 50fps and 1/50shutter would somehow be worse then 25fps and 1/50shutter because there are plenty of shots like that.




I never change framerates on the same camera at a wedding unless I absolutely have to, it only makes my job more difficult for no reason or benefit. If I would forget to change the framerate I might regret it later on if I want to slow footage down and it was shot at 25fps.

so shooting 25P 1/50 or 50P 1/50 has the same amount of light hitting the sensor?

[Ron Evans]

Yes of course. The shutter is staying open for 1/50 of a second for each time the camera records a frame. It is how long the shutter is open governs how much light falls on the sensor not how often. Just with 50P you have the option to use it at 50P for smooth motion or on a 25P time line for slower frame rate motion.

[Tom Van den Berghe]

I thought (wrong) that 25P or 50P also affected the light fall on the sensor. Thx for the answer Ron!