EX3 event coverage advice

[Eric Gulbransen]

I shot some footage a couple of weeks ago at a motorcycle race. Put a few minutes together as a sample for the promoter. Got hired to shoot the next event. Here's my problem:

I shot everything in 60p, 1/120th shutter. Slo mo looks fantastic. But the 60p footage plays like crap in a 24p timeline. Too juddery. Not only the lack of motion blur necessarily, from the 1/120th shutter instead of 1/48th, but more FCP-6 creating jumps in frames. Don't know the technical term but when I advance through the timeline one frame at a time sometimes FCP skips a frame. Seems too inconsistent to be math. Also some of the 60p footage played back super fast once it was captured. Pretty random how it occurred to some clips but not others.

How would you guys shoot an event like this with the EX3? Dedicate some shooting for slo mo, other for 24p? Seems an obvious solution but it would sure be nice to have the option to overcrank if a wreck just so happened to occur - which is why I shot 60p all day long last time. Don't want to deal with the judders again though.

Any suggestions would be appreciated.

[David Hutchinson]

Why are you putting it on a 24p timeline? You'd be much better off using 30/60 surely?

[Craig Seeman]

As someone who's faced similar decision (for me it was shooting tennis) you need to grasp some technical concepts and then make some aesthetic decisions.

Slow Mo - Shoot in either 24p or 30p time base and overcrank to 60p for slow mo. The recoded data rate is much higher shooting slow mo than simply shooting 60p. There's some math involved but that's the basic fact you need to know 24/60 or 30/60 are much higher record data rates that 60 native.

Targeted use - what's the likely final destination. If Broadcast/Cable TV than shooting 720p60 (or 1080i60) are the standard. For Blu-ray 24p can work in addition. For web or other playback it may be as you prefer or device limited spec.

Temporal resolution - 720p60 natively will give you best temporal resolution (otherwise it's 1080i60). These are generally the best options for shooting fast action sports.

Back to slo mo - If you can't shoot overcrank you're faced with some workarounds and alternatives, each with some sacrifices. If you shoot 60p for 60p time base then you may need to use Optical Flow to make 60p look as if shot at a higher frame rate and slowed. Optical Flow calculates/creates new frames (and sometimes gets it wrong). You can also convert to another time base (properly done eliminates the judder you're seeing). You can use Cinema Tools for that.

Aesthetics - it's hard to say which is "best" as each has a cost/benefit. You have to know where your going before your start. You may have to shoot some tests to make some aesthetic and technical decisions before hand.

Your options - use a 60p time base and use optical flow to slow the shots you want and deal with risks of creating new frames. Convert video to a lower time base 24 or 30 which will give you room to play video at normal speed at that time base and use the 60p in the above time base for slow mo. Going to lower time base looses temporal resolution.

IMPORTANT TO GRASP - shooting 60p is NOT the same as shooting 24/60 or 30/60 for slow mo

Math - XDCAM playback targets 35mbps (VBR so some variation). 24/60 records at about 87.5mbps so it'll play at about 35mbps. 30/60 records at about 70mbps for 35mbps playback. 60p native is recording at 35mbps so slowing that will give you a LOWER mbps.

[Eric Gulbransen]

O my god. Back to school. Thanks loads Craig. I will have to dissect your post one sentence at a time.

One of my challenges is this project will ideally end up in various formats - web being the first, which is why, David, I chose 24p for the first attempt. Thinking file size mainly.

All I know at this point is, holy mother I don't know anything.. Thanks for the help guys.

Here is the sample showing my struggles so far - the real time footage looks horrible but I love the slo mo.

Supermoto USA Infineon 2009, By GoGo

[Sean Adair]

ignore - glitched my edit - it's all in the following post

[Sean Adair]

I'll chime in to also endorse avoiding a 24p timeline for this type of project. Simply inappropriate motion look for one, and if slomo is part of the game, achieve it another way.
An "optical flow" or interpolated slo-motion would be ideal. Truth is, this temporal slowdown will be much less noticeable from a 60fps base than slowing down 24fps footage (more real frames per second).
I suspect your secondary delivery will be looked at much more critically for quality, and I'd focus on making that the best it can be.
You could even use slomo import of 60p down to 24p, conforming 60p footage to 24p playback in post for true overcrank - (as opposed to "in camera" overcrank, thanks Craig) as a comfortable to the eye temporal look. It will come back in with the familiar 3:2 pulldown and will most probably be seen on playback with a 60 frame base (or 60 field CRT). For the web, export it at 30fps (throwing out 1/2 the frames) and damn the torpedoes. It will still have a little judder (shutter to playback relationship) but it's an even division, and won;t be an irregular effect.

Or compromise at 30p with a 1/60 shutter. This will look reasonably natural in the temporal aspect. It will keep the frame rate to the web. Slomo will have to be extrapolated (optical flow in AE or motion) to look decent. I'm assuming that you don't have the luxury of CHOOSING when to overcrank, but want to take those delicious action highlights out of full coverage - when flesh, metal and asphalt do the dance of bump and grind. PS Eric - are you still IN the racing game?

Craig - I'm intrigued with the report that overcranked footage is actually captured at this relative to playback speed. IOW 35mps PLAYBACK rate as opposed to 35mps RECORD.
I first wondered about this in your discussion of SDHC card performance with an adapter.
How have you confirmed this? For instance, is an overcranked 60/24 clip of 10 secs the exact same size as a native (24/24) clip of 10 seconds?

For the record, I'm here poaching EX-cam HD info since I've jumped on the new JVC bandwagon which uses the same codec (although in a QT "shell" nyeh, nyeh!. As you probably know, we default to SDHC (class 6 supposedly mandated, although 'm also interested in testing sandisk extreme class 4/2!). SXS is possible with an op[tional JVC adapter (which I opted out of, at $500 more and no published advantage for my workflows). We can also do overcranking, and I will no doubt try this comparison myself next play break.

[Craig Seeman]

Originally Posted by Sean Adair

Craig - I'm intrigued with the report that overcranked footage is actually captured at this relative to playback speed. IOW 35mps PLAYBACK rate as opposed to 35mps RECORD.
I first wondered about this in your discussion of SDHC card performance with an adapter.
How have you confirmed this? For instance, is an overcranked 60/24 clip of 10 secs the exact same size as a native (24/24) clip of 10 seconds?

For the record, I'm here poaching EX-cam HD info since I've jumped on the new JVC bandwagon which uses the same codec (although in a QT "shell" nyeh, nyeh!. As you probably know, we default to SDHC (class 6 supposedly mandated, although 'm also interested in testing sandisk extreme class 4/2!). SXS is possible with an op[tional JVC adapter (which I opted out of, at $500 more and no published advantage for my workflows). We can also do overcranking, and I will no doubt try this comparison myself next play break.

Hi Sean,

There are several ways to test and verify the data rate change on overcrank. Note how fast it fills up the card. Much faster with overcrank because it's a much higher data rate record. Check data rate on playback of overcranked clip. They play at about 35mbps. If this were not the case, overcranked files would look horrible since the data rate playback would be very very low.

I'm using Sandisk 32GB Ultra II Class2.

Curious if you can actually get 720p24/48 on SDHC on JVC HM EX. Generally that's not the case with Sony EX although Hoodman claims their expensive SDHC can (I've heard otherwise though).

JVC HM also has 2.9 second cache record (NTSC) but Sony has none (that may change though). I actually think buffer management is related to Sony's handling of overcrank and cache record (or lack of) and SDHC/USB.

[Tim Polster]

If you need to record audio in-camera at the same time you can rule out any overcranking.

I put some 720p60 material on a 24p timeline in Edius and the results looked pretty good. So much that I see 720p60 as the best aquisition format for multiple delivery formats.

720p60 slowed down even on a 60p timeline still looks really good. A lot better than SD interlaced!

[Sean Adair]

Actually Craig, I've been happy with 60p footage conformed down to 24fps that was shot at 19mbs! Here looking at individual still frames isn't always great, but the result was still good. Think of it this way - those 35mps HQ frames at 60p don't degrade when played back at 24 - you just see each one for a bit longer. We're talking 720p here, and this is not going to be as much compression crunching as 1080p in any case.

The 24/60 or 24/48 slomo features with instant feedback are a nice feature, but there is much more flexibility when conforming 60p footage in post. Not only having the audio, but the ability to CHOOSE which footage will be used at real time and which will be slowed is often essential.

But the increased data rate on slomo record is fascinating. I'm surprised that it works this way. As I understand, even the varicam records 60p at a standard data rate - no matter what the playback speed is.

I don't have quality media yet. I'm experimenting with a 4gb PQ micro SDHC card in an adapter. It works, and I did a little in camera overcrank with it, but will wait for my transcend cards before doing more serious tests. Certainly the camera is marketed, and user reports appear to confirm full functionality with SDHC. I opted not to get the optional SxS adapter since m4p wouldn't be part of my workflow.

JVC specs class 6 very specifically, and it was very interesting to find information here suggesting that Class 2 sandisk might actually outperform some class 6 cards.

Perhaps a "buffer" helps SDHC performance on the JVC HMs as you mention? The cache recording is definitely a useful feature for some things. There would be a logical connection between these.

[Craig Seeman]

Originally Posted by Sean Adair

As I understand, even the varicam records 60p at a standard data rate - no matter what the playback speed is.

DVCProHD works under different rules. Data rate distributed equally to each frame rather than by second. That's why if you use P2 cards you'd find 720p24n (n = native as opposed to in 29.97) will allow you to record more on the card than 1080p30.

With XDCAM, 720p24 or 720p60 or 1080p30 all are 35mbps VBR.

Originally Posted by Sean Adair

those 35mps HQ frames at 60p don't degrade when played back at 24

They do actually. Some see it and some don't and it depends on the content. They don't "degrade" per se. There's fundamentally less data per frame. One may or may not see it but it's happening.

[Sean Adair]

Update. I was curious, so I took the com around for a quick spin (onto that very cheap 4gb PQ sdhc, which is rated class 6).
It records 24/48 and 24/60 onto this card without complaint. The footage plays back smoothly full size in quicktime player (after dragged to the hard drive).

Comparison of file sizes shows that on this camera, all clips were RECORDED at 35mps.
The 24/48 overcrank file size is exactly 1/2 the size of a standard clip of the same playback length. The 24/60 overcrank is smaller for it's playback length.

Interesting eh? This is exactly what I predicted here. The same compression and identical results to conforming playback rate of 60p in cinema tools. The convenience and playback are sweet advantages, and the range of speeds available are greater including undercranking (60,48,40,30,24,15,12,10) but not as flexible in some situations. I would do the undercranking effect previously by using a slow shutter speed, then speeding up footage the exact amount required to remove duplicate frames. (eg double material shot a 1/12 shutter). However post work often has workflow advantages, especially if you aren't certain which footage will be used at realtime, and which slowed down, or if a motion ramp is desired.

[Sean Adair]

Originally Posted by Craig Seeman

DVCProHD works under different rules. Data rate distributed equally to each frame rather than by second. That's why if you use P2 cards you'd find 720p24n (n = native as opposed to in 29.97) will allow you to record more on the card than 1080p30.

With XDCAM, 720p24 or 720p60 or 1080p30 all are 35mbps VBR.

They do actually. Some see it and some don't and it depends on the content. They don't "degrade" per se. There's fundamentally less data per frame. One may or may not see it but it's happening.

I'm familiar with the HVX-200, PN mode & in camera slomo on that camera. Also HDV1 interestingly is 19mbs at 30/60. But true 24p takes less space because the pulldown frames are duplicated as proxies when you record direct to drive.

My point is that clearly there is not "less data per frame" between 35mbps capture rate of 60p footage AND 60p footage played back at 24fps. Certainly less data (higher compression ) is not desirable and introduces the likelihood of degrading quality. We are still in a highly compressed format.

If the Sony EX cameras actually ramp up CAPTURE compression rates during in camera slomo to well over double the standard 35mbs rate that all other formats use, this is an incredible feat. I'm not doubting you - just surprised and curious how it is accomplished.

[Craig Seeman]

Basically than JVC HM EX does not do overcranking the same way the Sony EX does then.
In EX the recorded data rate is much higher during overcrank. 720p24/60 plays at the same data rate as 720p24 and 720p60. 720p24/60 will fill up a 16GB card in under 23 minutes compared to 57 minutes running at any "normal" frame rate.

EX overcrank assures that a frame of 720p24 and 720p24/60 has the same date per frame. Apparently JVC does NOT do that.

Fundamentally if you're recording 720p24/60 at the 35mbps compared to 720p24 35mbps, the overcrank will have LESS data per frame on JVC unfortunately.

mbps - mega bits per second (time not frame based).

Sony 720p24/60 = records 87.5 mbps to play back at 35mbps and it's easy to verify just by timing how fast the recording fills up a 16GB card.

If JVC is recording 720p24/48 at 35mbps, it is playing back at 17.5mbps.

[Sean Adair]

Hey Craig, we've chatted before, and I think this exchange is missing the direct contact!
Also - we're heisting poor Eric's thread.
The cameras do approach this differently then, The JVC RECORDS all HQ footage (under & over) at 35mps. The time available on the card doesn't change when put in overcrank. As you mention, effective playback rate is proportionately less. It isn't something I personally find a critical deficiency, given the robustness I see in the 720/60p codec (and of course with diminishing effect at 48/40/30 frames recording). It's a huge advantage over extrapolated frames without the temporal information. It's conforming frames live for instant review and saves steps in post.

Still, I have to go back to your comment "There's fundamentally less data per frame" & "If this were not the case, overcranked files would look horrible since the data rate playback would be very very low."
The comparison I made was with 35mps 60p footage, conformed (true overcranking) to a 24fps timeline. When these frames are on a 24p timeline they retain the same amount of data per frame. I thought this was clear before - Yes this is less data than a "native" 24p recording. However the HQ 720/60p footage is not bad at all. 19mbs 60p 720 hdv could show noticeable stress. I can now contemplate shooting projects with 60p as a base which I didn't feel comfortable with before. I suspect compression artifacts will be less than 1080p/30 .

In this situation, Eric may just want slo-mo of a dramatic moment that takes place during his race. It's not predictable when that moment will be, and he needs realtime coverage of everything. A common scenario for better or worse. 24p native footage would have precious little of the important temporal information for even the best extrapolation. 60p gave him the opportunity to choose what was slowed down. It's the uneven conversion (doesn't divide evenly) of 60p to 24p, combined with 24p being "unnatural" for this high motion that I think is bugging him. I still think 30p for the web, and 60fps for a final delivery is best - even if it means 60 interlaced fields (DVD or 1080i)

Part of the reality of shooting is also that as frame rate increases, the eye detects less detail or flaws in an individual frame. Both math and single frame analysis are limited in drawing conclusions as to the final motion clip. I'm coming from HDV, which may be pooh poohed, but some fine 19mbs 720p footage has gone far and stands up to scrutiny.

Finally, It's not that I doubted you! It's a great thing that the overcranking on the Sony preserves the 24p data rate, rather than using the data rate of the capture frames. How about when you undercrank on the Sony? Do you use less data in 24/10 or 24/12 modes? (keeping the 24p PLAYBACK data rate). It just strikes me, that since the hardware can sustain this type of data rate recording, that it would be useable for an even higher bit rate realtime mode. I think we've both learned a distinction between these cameras that wasn't obvious before.

[Craig Seeman]

OK I think I understand the perspective.

Sony EX records overcrank at higher data rate to target 35mbps playback
JVC HM is recording at 35mbps even in overcrank.
The previous JVC HD series was recording at 19mbps (although with 6 frame GOP instead of 15).

JVC HM at 720p60 (not overcrank) will hold up better getting slowed to 720p24/60 in post.
JVC HD at 720p60 would not hold up as well.

If JVC HM p60 can hold up getting reflagged to 24/60 then of course Sony EX also recording at p60 should hold up getting reflagged 24/60.

While Sony EX 720p24/60 is recording at extremely high data rate (87.5mbps) actually standard p60 at 35mbps should be sufficient to hold up being reflagged to p24/60.

Thus Sony EX 720p60 flagged as 24p/60 might be a lower data rate than 720p24/60 it will provide greater flexibility at no apparent loss since this is what you'd see with JVC HM and that holds up.

BTW I may be able to test this very soon. I did a shoot for a client tonight (tennis tournament) that wants some shoots entirely slow mo. Today he decided to shoot normal speed. I shot 720p60 just in case he changes his mind (as clients are want to do). BTW the way, one reason why he wanted normal speed is that shooting slow mo doesn't record the audio. Hence that's one of the flexible options of shooting 720p60.

BTW I'll probably test undercrank soon to see what that's doing regarding data rate.