Here's a short compilation of clips I put together last night with the FS700. Slow-Mo is sooooo addictive!

One Summer Evening with a Sony FS700 - YouTube

I've been a slowmotion addict since years; it still has this unreal beauty.
You wouldn't say it's in the evening: it's still very bright :-)

FS700 finally makes it possible to realise ideas I've had for a while.
Too bad I don't know anyone at Sony ;-)
(I case they read this and like to have other slowmo-stuff than skaters, biker, golfers and waves: they can mail me :-p)

At how many fps and in what kind of resolution was this shot?

I also see there is quite some aliasing in the first clip, also when viewed at 1080p, is that a result of the slowmotion mode or has it something to do with youtube? The slowmotion shots do look very smooth, impressive.

The Bikes were shot at about 7pm, so definitely evening, but the sensitivity is such that I had plenty of aperture still to play with.

The first shot was at 480fps. At 480fps the image is 1920x1080 but the vertical resolution is halved to 540. This halving of the resolution leads to the extra aliasing.

The other shots were done at 240fps where the camera is full resolution 1920x1080.

Once I get my own FS700, hopefully in late June, I will be off to shoot some much less common subjects in super slow mo. I'm doing some FS700 workshops at Broadcast Asia if anyone is interested.

Is there much of a delay after each shot while the buffer is writing to the media storage?

It is read out in real time so if you were shooting 240fps and had the camera set to conform to 24fps, each second of captured footage will take 10 seconds to record to card/play out the video ports.

Thats not correct Chris. The buffering is faster than real time. I was shooting at a base frame rate of 23.97 and the buffering appears to happen at 60P ( which presents an interesting possibility with regards to using a higher quality external recorder to record the slow mo, this needs more investigation).

So when shooting at 240fps you get 8 seconds of filming, which then takes 32 seconds to buffer and results in a clip 80 seconds long.

So regardless of the base rate it outputs 60p?

This will be a problem for using external recorders such as the PIX since they don't do 3g.

No, just while it's buffering. In normal modes the Sdi output reflects the shooting frame rate. I have not fully investigated this but that's what appears to be happening.

So when shooting super slow-mo, first you set your base shooting and playback rate which can be 23.98, 25P or 30P then you activate super slow mo ( two presses of the S&Q button). When you hit the record button the camera starts to buffer either the previous or the following 8 seconds (if at 200/240 fps) depending on whether you are using start or end triggering. It's during this buffering period that the camera appears to be operating at 50/60fps.

My suspicion is that the camera is storing the 240fps video in an internal memory buffer and then reads out that memory at 60fps, writing a 60fps AVCHD file to the SD card that is then flagged to playback at 24 or 30fps. This allows the FS700 to write at standard AVCHD speeds and bit rates making it really easy to create the slow motion files.

That is interesting. I was under the impression that after it buffered it would play out at the base rate so you could use an external recorder.

If I am understanding you correctly there won't be any advantage to use an external recorder in the interest of avoiding the internal AVCHD codec.

The buffering happens at 60p so if you have an extnal recorder that can record at 60p there may be some advantage.

I will be interested to find out if frame rates other than 60 will be possible in the production units since the recorder I have doesn't do 60p.

It's going to have to be 50 or 60p. Not many will want to wait 80-90 seconds before being able to shoot another shot. The Gemini and Hyperdeck Shuttle 2 can do it. My full review is online now http://www.xdcam-user.com/2012/05/a-second-look-at-the-sony-nex-fs700/

Unfortunately those aren't the one I already purchased. :(

Hello Alister - always great to get/read your input!
Den Lennie reports they got SloMo recorded via SDI to the Samurai here - about 20% down the web page you will find it...

Sony NEX FS700 on Location - Become a Film Maker Without Quitting Your Job! (http://www.denlennie.com/2012/04/sony-nex-fs700-on-location.html)

What's your take on this please?

Thanks again,
Lonnie

Not quite sure how Den managed that and I'm not sure he is too, as the Samurai on paper can't do 1920x1080 50p or 60p.

I'm still looking in to this. I'll have access to a FS700 on Thursday and I'm taking my Gemini which can do 1080p60 to test it out.

Look forward to your results Alister - Happy Shooting!
Lonnie

Keep us posted Alister. I would enjoy getting to try one out myself.

My suspicion is that the camera is storing the 240fps video in an internal memory buffer and then reads out that memory at 60fps, writing a 60fps AVCHD file to the SD card that is then flagged to playback at 24 or 30fps.

I did a Weisscam 1000fps shoot yesterday and this is the exact same principal that camera works under. It internally records to a RAM buffer then prints out to HD-SDI in real time, so it takes a while until you can take another shot. I have to say filming rugby players at 1000fps under a bevy of 18K HMI's looked absolutely amazing.

I can confirm that you can read the 50/60 fps buffer stream vis 3G Sdi to an external recorder. I was able to test it today recording uncompressed to a Convergent Deign Gemini. So, 240fps uncompressed is possible. Very cool!

Thanks for the follow up, Alister!

Will a single link HD-SDI work at the 240fps or does it need to be 3G?
(the Samurai specs per their website, they list the input as HD/SD and the output as 3G/HD/SD. And based on Den Lennie's report - I'm a little confused...)

Thanks for shedding info and light!
Lonnie

The FS700 writes slow mo from the memory buffer to AVCHD and the SD card at either 50 or 60 fps depending on which country zone the camera is set to. The camera output can be set in the menu to either 25/30p or 50/60p, if set to 50/60p the output is 3G. In either case the buffering happens at 50/60p. So, if the output is set to 25/30p you only get every other frame on the output Sdi at 1.5G. The end result of this is that if you shoot at 200 fps with the camera output at 25p you would end up with an external recording the equivalent of 100 fps with a 180 degree shutter. Still very useful and a nice option. But to get the full benefit of external recording you need to set the FS700 to output 50/60p over 3G SDi and then you must use an external recorder that has 3G SDi and can record at 1080 50/60p which the samurai can not do.

Thanks for the succinct answer and always a pleasure to read and watch your craft, Alister!
Lonnie

The FS700 writes slow mo from the memory buffer to AVCHD and the SD card at either 50 or 60 fps depending on which country zone the camera is set to. The camera output can be set in the menu to either 25/30p or 50/60p, if set to 50/60p the output is 3G. In either case the buffering happens at 50/60p. So, if the output is set to 25/30p you only get every other frame on the output Sdi at 1.5G. The end result of this is that if you shoot at 200 fps with the camera output at 25p you would end up with an external recording the equivalent of 100 fps with a 180 degree shutter. Still very useful and a nice option. But to get the full benefit of external recording you need to set the FS700 to output 50/60p over 3G SDi and then you must use an external recorder that has 3G SDi and can record at 1080 50/60p which the samurai can not do.

This is very, very interesting to know!

God that 240fps is sweet, I thought the S&Q 60p on the FS100 was the beez-neez. This already has me thinking upgrade from my FS100 that I've had for a scant 2 weeks now. Nice...

The FS700 writes slow mo from the memory buffer to AVCHD and the SD card at either 50 or 60 fps depending on which country zone the camera is set to. The camera output can be set in the menu to either 25/30p or 50/60p, if set to 50/60p the output is 3G. In either case the buffering happens at 50/60p. So, if the output is set to 25/30p you only get every other frame on the output Sdi at 1.5G. The end result of this is that if you shoot at 200 fps with the camera output at 25p you would end up with an external recording the equivalent of 100 fps with a 180 degree shutter. Still very useful and a nice option. But to get the full benefit of external recording you need to set the FS700 to output 50/60p over 3G SDi and then you must use an external recorder that has 3G SDi and can record at 1080 50/60p which the samurai can not do.

So if I understand well I can make a burst shoot of 240 fps (ntsc setting), record it as I-frame with my nanoflash set at 1080p30, thus taking 120 fps of the initial 240. With Cinematools flag it as 25p and import it in a 25p timeline. The end result is 120fps brought down to 25p, so almost 5 times slomo with the nanoflash. Correct?

Yes, that is correct. It would be the equivalent of 120fps with a 180 degree shutter or 1/240 shutter.

One thing probably worth considering with overcrank 60fps, 120fps, 240fps footage recorded to AVCHD is that AVCHD has a significant temporal compression component which likely gets more efficient as motion gets slowed down. As you increase framerates, the similarities between frames almost always become significantly greater, which should allow the efficiency of long-GOP AVCHD to increase substantially. I'd be interested in testing this, but my understanding is that GOP codecs do exceedingly well at recording a bunch of very similar frames.

Then again some of this effect may be lessened slightly given the lack of motion blur at such frame rates.

Yes, long GOP codecs become more effective at higher frame rates due to the reduced temporal differences between frames. The FS700 takes full advantage of this. At up to 30fps the bit rate is 24Mb/s, when you go to S&Q mode the bit rate increases to 28Mb/s for 60fps, so only a 4Mb/s bit rate increase but double the number of frames.

isn't the 28Mbps AVCHD mode only for 1080p60 recorded footage? That is how it works on my FS100, if I want to record S&Q it is always records the 60fps footage into 24fps FX mode (24MBps)

You might be right Noah. I'll check again when I have a camera at the end of the week.

Yes, that is correct. It would be the equivalent of 120fps with a 180 degree shutter or 1/240 shutter.

I suppose this means that it would be possible to schoot 240fps with nanoflash and other devices, if Sony would include the option to write the files at a speed of 30fps instead of 60fps? Any chance that this might happen?

Is it by the way confirmed that the output over sdi is before compression?

My guess is that the speed/loop recording is done after video is coded to avchd, to use less memory. If so, that means we could not record slowmo to extern recorder through the SDI port. Only to intern memory. No better quality.
Isn't it also so, if 1080p50 is recorded at 28 mbs, and put on a 25/24p timeline, we use only half the pictures, and the effective mbs would be around the half? Thats 14 mps! The same happens if we stretch the footage to 25/24p.
If used from camera, or from a TV/PC that can play back 50 or 60 frames it can look fine.
But my delivery is mostly broadcast, and DVD projects at 25p. The footage from my nx70 recorded at 1080p50 looks worse than footage recorded at 25p. But I have not measured the bitrate.
Or am I thinking wrong here?

Arnt, Alister Chapman proved already that the FS700 sends the slomo over sdi to external recorders. The Gemini takes full advantage of the 240fps and the nanoflash only half. See his explanation before in this thread. He supposed in other threads that is is before compression, I am looking for confirmation of that.

Yeah, AVCHD compression takes place in the moment the buffer writes to the card.
But the 8bit 4:2:2 signal is active at the SDi and HDMI ports at the same time.

It would take a hell of computing power to apply compression at 240 FPS.

Frank

Here's a controversial sideways thought about that.

1080p50 material at 28 mb - I don't think, because of the way that H.264 and AVCHD work, we can do a straight halving. The extra bitrate can be spent on looking at the (less) difference between frames. There are also 'jumps' in MPEG2 recording, in that the difference between 25 mbit and 35 mbit was quite extraordinary (thinking of XDCAM here).

Best PAL SD comes from 720p50 (or 720p25 if you don't want interlace). I've stuck with this since getting an EX1, and BBC has published guidelines, test results and the like. You'll have to blur 1080p before downscaling to reduce the amount of detail to avoid aliasing. You'd certainly avoid shooting 1080i for SD, as you'd have to de-interlace (thus reducing resolution by 25%) before scaling, and then you'd loose the temporal information OR you'd use fields as frames and find that 540 lines is less than 576, and accept the lumpy scaling UP.

The AVCHD codec implementation in the FS100 and FS700 is really good. Very good indeed. I would pitch it against XDCAM-EX 35mbits and expect it to look wonderful. Only the usual 'Compression Nightmare' suspects will degrade it: scenes with lots of movement throughout the screen - full screen water, a viewfinder full of leaves in the wind, certain types of sports shot.

Various testers have looked at the FS700 SSM and all agree that in order to perform this magic at that money, the SSM modes will have to include a little aliasing just to read off the frame data quick enough. This will happen and can be outputted to a 4:2:2 device like the Gemini, but the sole act of the expediated sensor read will have done the damage. Want more? Want better? Get a Phantom or an Epic. :)

It's going to have to be 50 or 60p. Not many will want to wait 80-90 seconds before being able to shoot another shot. The Gemini and Hyperdeck Shuttle 2 can do it. My full review is online now A Second Look at the Sony NEX-FS700. | XDCAM-USER.COM (http://www.xdcam-user.com/2012/05/a-second-look-at-the-sony-nex-fs700/)


Alister, are you sure the Hyperdeck shuttle 2 can do it? When looking at the specs I only see a normal HD-SDI input, not 3G. Don't you mean the Hyperdeck Studio?

For thought comparison, how much better do you think the FS-700 slow-mo is than using Twixtor with 60p material? Or is it?

Workflow, shooting and image quality?

Thanks for your input.

Without having a fx700 I would say it will be MUCH better, all twixtor slowed down 50 or 60p footage I have seen produces some weird morphing artefacts that you won't get with the fs700.

Twixtor'ed 60p:
Pro: Looks pretty good if you're shooting subjects that are interacting with low/no-detail backgrounds, allows simulation of very high overcranking
Con: Edge artifacting in most normal situations with significant motion, long render times, lots of time and tweaking required for each shot for optimal results which is still always inferior to real overcranked footage

FS700
Pro: No interpolated edge artifacts, no extra hours of plugin tweaking and rendering to get an inferior result, can playback on set to see and evaluate footage
Con: Costs more than twixtor, burst time limits


There is no replacement for actually shooting high framerates.

Here's a controversial sideways thought about that.

1080p50 material at 28 mb - I don't think, because of the way that H.264 and AVCHD work, we can do a straight halving. The extra bitrate can be spent on looking at the (less) difference between frames.
Not controversial at all! And not just to do with H264 and AVC-HD. It's highly likely that with 1080p/25 and 1080p/50 the time difference between I frames is the same (1/2 second?) and hence the same difference between them because of movement.It follows that for 1080p/25 the GOP length is 12, for 1080p/50 it will be double that.

Since I frames need a lot more bits than difference frames, it's then easy to see that doubling the basic frame rate won't need anything like a doubling of bitrate. Practically, there's a lot more to it than that - the content differences between successive difference frames will be less (only 1/50sec apart, rather than 1/25 sec) so the coded size of each can be smaller with 50p than 25p. That will help compensate for the greater number of them.
There are also 'jumps' in MPEG2 recording, in that the difference between 25 mbit and 35 mbit was quite extraordinary (thinking of XDCAM here).
Again, not unique to MPEG2. You need a certain minimum bitrate to get anything worthwhile at all, any increase then gets applied just to reducing deficiencies - and if they weren't that bad in the first place......

It's part and parcel of the same argumentabout the oft (mis)quoted statement about H264 offering half the bitrate of MPG2 for equivalent quality. It's true - but only at quite low bitrates. The H264 "tricks" do improve on areas where low bitrate MPEG2 shows problems. But improve the MPEG2 bitrate and there's less problems for H264 tricks to work on!

Hence H264 has big advantages or broadcast transmission and web video - less point for camera acquisition where bitrate is less of an issue.

Alister, are you sure the Hyperdeck shuttle 2 can do it? When looking at the specs I only see a normal HD-SDI input, not 3G. Don't you mean the Hyperdeck Studio?

You are correct Cees. The Shuttle 2 cannot do it. In BMD's press release they say it can, when in fact it can't.

To be clear, though, it sounds like we could record 240fps and still get 120fps frame decimated footage out to a non-3g recorder like the hyperdeck shuttle 2 or other device? This would be done by setting our recording format to 30p (since 60p requires 3g) and when writing out the buffer every other frame is still played out of hdmi/hd-sdi. Can someone confirm this?

As soon as my camera arrives I will test this idea.

Thanks for your input on Twixtor. Just wondering how the time limit on the FS-700 slo-mo works into a workflow with such a short burst. Making a clear choice between slo-mo or live might be difficult.

I think in most cases the time limit will be far more favorable to the work in post that twixtor requires. More likely you will find you rarely need up to the full limit since things you want to see in slow motion tend to happen quite quickly and you will want to keep down your buffer offload times.

To be clear, though, it sounds like we could record 240fps and still get 120fps frame decimated footage out to a non-3g recorder like the hyperdeck shuttle 2 or other device? This would be done by setting our recording format to 30p (since 60p requires 3g) and when writing out the buffer every other frame is still played out of hdmi/hd-sdi. Can someone confirm this?

I don't have the FS700 (yet), but Alister Chapman confirmed this on the same question when I asked it before. So the answer is yes.

Twixtor'ed 60p:
Pro: Looks pretty good if you're shooting subjects that are interacting with low/no-detail backgrounds, allows simulation of very high overcranking
Con: Edge artifacting in most normal situations with significant motion, long render times, lots of time and tweaking required for each shot for optimal results which is still always inferior to real overcranked footage

FS700
Pro: No interpolated edge artifacts, no extra hours of plugin tweaking and rendering to get an inferior result, can playback on set to see and evaluate footage
Con: Costs more than twixtor, burst time limits


There is no replacement for actually shooting high framerates.

These are not the only reasons to shoot high framerates.
Some organic movements are very hard to interpolate; the software just doesn't know how something like different hairs of a longhaired girl running towards the camera on a windy day move.
I tried this in december: most of the time it looks weird and fake. Only in one shot it was ok.
(And not thanks to Twixtor, but the standard options in AE.)

Or how thrown water splashed as it 'explodes' on a detailed object: it just yields distorted footage. (Which can be funny... lol)

Shooting high framerates wins in my opinion as well.