As we all know, the FX1 / Z1 records to tape in the MPEG 2 (compressed) format.
I was at a video exhibition in London today and I asked the people demonstrating the Z1 if it was possible to stream raw, uncompressed footage from the Z1 (to reduce artefacts, etc). They said it was NOT possible, but I just searched the net and someone says that it can be done with the “BlackMagic DeckLink”. I went to their site (http://www.blackmagic-design.com/site/decklinkhd.htm) but I can’t work out for sure if it is possible.
Does anyone know for certain if it can be done? Also, does anyone know if such devices are available for the PC?

Thanks

yes it works

The only problem is that you'd need to lug your PC around I guess. =P

As stated the system is possible but you need to consider a couple of things.
Will the system be for capture only or do you want to edit on it as well and next, how portable does it need to be?

Either way if you’re looking to use a HD capture card from Blackmagic it must slot into a 64bit/100Mhz or 133Mhz PCI-X slot.
These are mostly dual G5, Xeon or Opteron boards.
You could also use a Kona2 capture card for the G5.
To begin you’ll need either an AJA HD 10A or a Blackmagic Multibridge to convert the ‘live’ HD component out to HD SDI.
You must capture the component live as if you don’t it will have gone through mpeg2 compression.

Next if you want to use this system for capture only at say 10bit 4:2:2 you’ll need either a scsi or sata array capable of capturing 500GB at 135MB/sec for 2 hours (say).
It’s also fair to say here that nobody outside Sony Japan knows for sure with the ‘effective’ bit depth and colour space of the FX1 component out is.
Sure it won’t have gone thru the mpeg2 encoder but it could well be ‘effectively’ anywhere between 8bit 4:2:0 and 10 bit 4:2:2, probably!
If it’s going to be your edit system as well you’d probably be best using something similiar to what Blackmagic suggests on their site.
The cheapest capture only xeon system I’ve seen that ‘should’ do it is a Dell 1420 for $450.
If this system is a xeon capture only you’ll only need one cpu as you’ll be using DMA. Ask BM for other G5, Opteron capture only specs.
For the drives you’re key areas are i/o bandwidth and backup and of course $$.
I’d look at trying 4 x 300Mb Maxtors Maxlines.
Find a sata controller that let’s you do Raid 0 and raid 1 on the same physical drives.
Make the first (outer partition) Raid 0, say about 1/3 of the drive and the inner partition Raid 1.
Backup as you go!
Or you could go for a 8 port sata controller and have an additional external raid 0 partition to shuttle back and forth with your main NLE system.
Needless to say you have a lot of storage options!
Personally I’d like to see how those scsi 2.5” 10K 250GB Hitachi sonatas go. I heard they handle shock of 500Gs!!
Another thing I’d like to investigate is using 8 sata drives, 4 internal and 4 external. See if you could create an 8 disk, raid 1 partition over the fastest (outer) section of the drives, and then use the 4 external drives as a copy to be used with your main NLE. Not sure if it’ll work but worth investigating
Remember you’re capturing at 134MB/sec so when you’re transferring data, Firewire or 1Gig Ethernet it’ll take 2.5 times the length of the footage to transfer, if that makes sense.
Last is batteries, that’s if you want it portable. Well if you really want it portable I’d be looking at some of those industrial SFF single xeon boards but that’s another thing!
You’ll need Lithium-Ion battery arrays capable of 250W for 2 hours. These have a good power to weight ratio.
I suspect the batteries, battery controller and atx converter will cost around 1500.

You do realize no one is doing this stuff at such ridiculous cost. A portable 10bit uncompressed 4:2:2 SRW1 costs about $95,000 and we’re talking about $5~10K for a portable recorder.
That’s progress!

Dave Farland

Wow, David, that's a pretty comprehensive answer!

If you have seen the documentary on the DVD of the movie "Russian Ark" you will know what recording a 90 minute uncompressed HD movie in a single take requires... About eight people were dragging about 400-500 lbs of equipment around 2 miles of the Hermitage museum in St Petersburg with temperatures of -20C outside!

However, while I appreciate that if you were going to spend two hours recording in any one session, you would need to drag about a LOT of kit, my idea would be to have half the kit (i.e. half the disk drives, half the batteries) and only record uncompressed when you are sure that you are likely to use the footage that you are shooting. Usually, with video, you are not to worried about using a lot of tape as it is re-usable and cheap, but treating the uncompressed footage as if it was on film would put you in an efficient shooting mindset...

They say that Hitchcock used half the footage he shot in the final movie (an efficiency of 50%). People shooting on video probably have an efficiency of 5% or less! So, if you are prepared to shoot test footage and rehersals on tape and save hard-drive space for the "real thing", then a portable uncompressed setup could be viable.

As to price, I think that (based on your specs) a one hour uncompressed, portable storage system could be made for as "little" as $2000. Whether this is good value for money depends on how much you value the difference in quality!

Considered as an investment, such a system, when used in multiple productions (and maybe hired out), could be not only viable, but valuable!

Thanks for your input!

While the topic is raised, I'm curious:

You can take the live component out on an FX1/Z1 and sample (in a low noise environment) some form of uncompressed HD video on a high end consumer PC, hack the camera and equip it with a lens with less chromatic aberation... It strikes me that for less than $10k you could work up a 1440x1080i uncompressed video HD cam that is reasonably portable.

What is the cheapest "real" system (i.e., one that provided uncompressed 1440x1080i or better) available from a manufacturer?

-Steve

Leitch VelocityHD: Boardset/software is approx $10,000
The Media 100 HD: Boardset/software/break-out box is $8,000
HD Fury: Boardset/software is approx $12,000
Kona2: Boardset is $2,500*
Decklink HD Pro is $1,500*
* requires Decklink Multibridge ($2,000) for analog video/audio capture. The multibridge is nice because it'll imbed your audio into the HDSDI signal.

These boards will capture uncompressed 10bit 4:2:2.
If you want ‘real’ HD turnkey version of these boards/software add $8,000~10,000 to the price. Nothing portable here.

Dave Farland

Here is close to the cheapest Mac system you would need:

1x 1.8Ghz G5 w/ 1GB ram $1750
1x Decklink HD I/O card $500
1x AJA HD10A $1500
1x 8 Channel SATA RAID controller $300
6x 300GB SATA drives $1200
Adapters to mount drives inside G5 $600
1x Final Cut Pro HD $999

Total $6849

All you need for this application is a single link HD-SDI card and the Decklink Hd is the least expensive and works just fine.

This system would give you 1.8TB of dsik space to capture to.

Dustin, I dig your list. Do you have this equipment now?

I've got the G5 and FCP HD. But, the other stuff is now on my wish list. Can we squeeze by with Firewire 800 drives? If so, we're all looking at about $2500 after a Mac/FCP purchase.

Chris,

I have a dual 2.5Ghz G5 instead of the 1.8Ghz G5, but the 1.8 does just fine.

I also have a Decklink Multibridge instead of the AJA HD10A. THe Multibridge cost about $500 more, but it does a lot more. Well it will do a lot more once Decklink gets it working correctly.

I also have a JVC 19" HD studio monitor ($2500) for editing and color correction. You can live without this, but it is very nice to watch HD on a good HD monitor and it is required for proper color correction. I convert the HD-SDI signal from the Decklink HD card to analogue YUV using the Multibridge for viewing on the monitor. The Multibridge can convert analogue to digital from the camera while at the same time converting digital to analogue for viewing on the monitor. It is very nice!

The only thing I don't have right now is the 1.8TB of internal disks. Without that I can't do 10bit uncompressed HD.

With firewire drives (25-80MB/s) you can do 1080i DVCPRO HD (14MB/s) just fine.

Since we are shooting 8bit MPEG-2 compressed HDV, it seems that 8bit DVCPRO HD compression would be good enough. I need to do some real test of capturing HDV to DVCPRO HD and 10bit uncompressed to see how much of a difference there is.

10bit would be great, but I have to justify another $2000+ in harddrives and I don't know that the quality difference will be noticed by the audience.

That's really interesting...so, basically we have everything. I've got the dual 2.0 gz machine, so we just need external firewire drives. Even then, we only need 400 drives right? The last I remember, G5's run Firewire 800 drives like crap. I'm not sure if that issue has been resolved or not. Know about that one?

Chris,

Firewire 400 is just fine.

The Firewire 800 problem was fixed on the 2.5Ghz G5, but is still not at fast as Firewire 800 on the G4. As far as I know all G5s before the 2.5Ghz have terrible onboard Firewire Performance. The simple fix is to buy a firewire 800 add-on card.

All you need is:

Decklink HD $500
AJA HD10A $1500 or Decklink Multibridge $2000

And you can capture DVCPRO HD to Firewire HD from any analogue HD source.

DVCPRO HD is about 50GB per hour versus 12GB per hour for DV25, so you will more disk space than with DV25, but a single drive is fast enough.

Actually a single 400GB SATA drive as a second drive inside your G5 should be able to handle about 7+ hours of DVCPRO HD footage.

Thanks for the info Dustin. I expect DVCPRO HD will give very good results
I suggest the Multibridge over the AJA because it'll inbed the analogue audio from the FX1/Z1 into the HDSDI signal .

Or you can use the Decklink HD pro capture card and a digital source.
Or, I'm not sure about FCP but can it capture timecode and even audio from the firewire input or audio from an independant sound card at the same time you're capturing the video from the decklink card?
If you've sync'd from the camera I guess you'd just use the audio from the HDV tape. There are other solutions i.e. lanc but I'd hope these would be cheapest.

Dave Farland

David,

I recommend the Multibridge over the AJA HD10A also.

As for audio, in FCP you can either capture:

1. embeded audio from the Multibridge

2. DV audio from the downconverted HDV footage

3. Audio though an external sound card. I have a MOTU 896HD.

FCP is very flexible about where the audio, video, and deck control come from.

Once the Multibridge has been replaced and is correctly working I will do some more tests to see which way is best for me.

The Multibridge is set to -20db so all of our captured audio is about 16db below where we want it. I am looking into the best solution to work with this. Right now I have my Sound Devices 302 field mixer between the camera and the Multibridge and that works.

The audio in the downconverted DV footage is just right, so maybe using the DV audio will be the best solution.

As mentioned in the Barefeats website:
http://www.barefeats.com/hard35.html,
the internal SW RAID 0 in a Mac can deliver a sustained 114MBytes/s with only two SATA drives and between 184 - 232 MB/sec with 4 drives and a PCI controler.

This is great information, you guys rock!

There is of course no chance that the component outputs are 'compressed', as analog outputs can't be -- but the question is has the output video been through the compressor & back out again -- is it a video output of decoded compressed video.

The simple answer seems to be no -- the component output from a live signal has not been compressed & decompressed ... however, it has been chroma down sampled so it is NOT pristine, 'uncompressed' in the sense that it is full quality.

Is there a gain if the component output is recorded using a higher quality recorder? Yes, in that there is no mpeg compression-decompression cycle and so no mpeg artifacting. But not as good as one could wish, and certainly not at the standard of 10 bit, 4:2:2 chroma.

GB

Don’t want to have arguments of what exactly what ITU standard or effective bit depth/colour space the live component out is until Sony Japan gets back to me with their engineering specification on the output.

Suffice to say capturing the component should remove mpeg artifacts, provide more robust editing/compositing which will ‘make’ the solution for a few people.

Is it 4:2:2 or 4:2:0….dunno.
Currently I’m looking at lithium-ion battery packs to power the capture system for a couple of hours. Pre-built battery arrays seem to range from $1000~1500. 100watts of battery weighs the same as a 3.5” disk unit (.6kg)probably need 6 plus controller boards for 2 hours shooting.
http://www.ocean-server.com/battery_packs.html.
Here is one supplier, Jon Crowell who knows his stuff when it comes to powering pc’s with batteries.

Thank you,
Dave Farland

Can the Z1 or FX 1 output from the component jacks a SD signal?

If so can it do it while shooting HD or does the camera have to be in DV mode?

The equipment costs wouldn't need to be as high then for storage. Many of us will still be doing SD projects for some time yet. You may also be able to get by with a fairly portable capture system.

You could use a high end laptop with Liquid Edition 6 pro. The LE 6 pro break out box connects to a USB2 port and captures from component uncompressed video. You would then only need to lug around a laptop, external G-raid, and the small LE6 bob.

Very good idea Thomas but unfortunately you can't down-rez the component output to SD while recording in HDV. Sure you can on play back though.
Blackmagic sells it’s Decklink SD ($295) card which can down rez HD to SD but they recommend a HD platform http://www.blackmagic-design.com/site/65support.htm to support it. Not sure the minimum BM spec for a HD to SD capture only system as BM talk about their Decklink SD/ HD host system replaying HD footage in SD which wouldn’t be required here.

Thank you,
David Farland

Little more on the FX1 component out specification...

I asked Sony Japan the following question:-
What is the “ITU” standard that describes this component video output?

Sony replied:-
"The ITU standard is IT-R BT.709 and the factory says there is no recommended setting value of Digital Standard". (Sony Japan)

I'm waiting on more info......

Thanks you,
David Farland

A very interesting thread here...

And a little amusing too. Already we are getting used to the notion of employing a focus-puller on video shoots when the final product is meant for theatrical projection, and now we might even bring back the equivalent of the 'video-assist' playback operator!

So I took a look at what equipment playback operators currently use on film shoots.
This is the Omega deck (http://www.ffv.com/product_specs/Omega%20Cut%20Sheet.pdf) (notice the removable hard drives)
The Digiclam (http://www.lemac.com.au/sales/DIGICLAM/DIGICLAM.html)
The Digi Combo (http://www.y-vamp.com/digicombo.html)
and a software version, Cinelog (http://www.videoassist.com/software.htm) which can be used with proprietary hardware (http://www.videoassist.com/hardware.htm)

So what you're aiming at is something not too dissimilar to one of these systems, right? But with a much higher throughput of data to allow the recording of component HD video.

You could also go a step further and build in some of the features that are on these systems such as mixing images (for critical line-ups) or variable speed playback.

Of course not everyone who owns an HDV camera would want one, nor - I think - would many HDV owners who'd have a use for it want it on every shoot. But it opens up the possibly of building a dedicated system (like some 35mm video assist operators already do) which can then be rented out to shoots.

In the UK this looks tempting...

As to this question of a high shooting ratio eating up disk space: well, again you go back to something like traditional film practices. You record everything, including rehearsals, log the 'selected takes' (the old continuity girl's job (I liked her - let's have her back!)) and then delete everything you don't want, making room for the next set up. Meanwhile - if you choose - you can also record everything using the camera's own tape transport to give you a back-up if you have a catastrophic failure on your hard drive recorder.

The director has to think in terms of a limited amount of memory - like a limited amount of film stock.

At the end of the day all the hard drives (which are removable) are taken to the editor.

Batteries: as long as it is understood that a mains supply is required either at studio / location, or via a genny, this is not an issue. But a proper reccy may be needed at each location by a competent electrician (big hotels for instance will exhibit a voltage drop and spikes when all the ovens go on mid morning). Some kind of mains supply cleaner or buffer (not my strong point) would be needed.

Just realised (oh silly me!) that of course the natural person to push the record button on this thing would be the sound recordist - so you don't need an extra member of crew after all.

(The poor Video Playback Op was always the first to go...)

Would the chroma on the Z1 really be reduced right away? It seems like it would like many SD cameras but then I was thinking about how the camera would do it based on different output formats.

If you shoot HD you are getting 4:2:0
Now if you shoot DV or DVCAM you get 4:1:1

How can the camera switch the entire chroma output of the block based on what format you are using?

To me it seems if the chroma was reduced it was only be down to 4:2:2. This would be an easy point for the camera to get to either 4:1:1 or 4:2:0. If the block only did 4:2:0 where would the extra 1 come from for 4:1:1. The same is true if the block outputs only 4:1:1. The camera would have no way of bumping up to a 2 for 4:2:0.

I then thought well maybe during SD shooting the 0 turns to one since only 1 field is used. Well this would work for 480p mode but not regular everyday NTSC video. The two fields for 480i video would have to come from the same alternating fields from the HD block or else the fields would be interpolated. This means that unless the DV from the Z1 is only 4:1:0 the block must output 4:2:2.

I then thought about well maybe they kill the chroma near the component output. Well this would mean the signal would have to be run two different methods of chroma reduction based on the format. This seems like it would be hard and rather complex in a camera this small.

I could be wrong about all of this. Just my boring and rather useless thoughts on the subject.

4:1:1 and 4:2:0 are both the same 'amount' of chroma data -- one quarter the luma resolution -- just differently sited. 4:2:0 describes a chroma 'block' that is two pixels wide and two pixels tall; 4:1:1 describes a chroma rectangle that is four pixels wide and one pixel tall. This is a simplified description, but the salient fact is that both are derived quite easily from the same source chips, as they are -- but for a math trick -- the same.

IIRC PAL DV is already 4:2:0 instead of 4:1:1 -- this was in recognition of the 'extra' vertical resolution of PAL.

I believe the chroma downsampling is done at the imaging block before being sent to the encoder.

GB

I'll chime in here and say that anyone who'd buy an uncompressed card prior to NAB 2005 (between now and then) is acting out of impulse more likely than sensibility. You'll see several wonderful solutions at NAB. I'm testing one now....but can't say more. But it is sweet and stunning!

So you are saying there is a software code hardwired into the camera that will change the way the chroma is output based on what format you are selecting? It then does this before hitting the encoder block? Wow if that is true SONY put a lot of effort into dumping down the live video from the camera as to not compete with the higher end gear. I know how the two chroma formats work, I just didn't think SONY would go to all that trouble to make the chroma dump thing complex enough to switch based on the format chosen.

No, I'm saying that the chroma downsampling is the same for both as far as the imaging block is concerned, and whether the downsampled chroma gets encoded as 4:2:0 or 4:1:1 makes no difference to the imaging block ... Either way the encoder gets a chroma downsampled source, how it chooses to manage that downsampled chroma becomes a codec choice.

GB

Dustin, David et al:

With Decklink HD + Decklink Multibridge... have you had a chance to compare the quality of the video captured off Sony's Component out vs. MPEG (firewire/tape)?

Does the improvement in quality warrant a $2,600 extra hardware cost?

Most importantly, have you seen an improvement in green screen keying? (Any chance of screen grabs...)

OK, seem slike no-one has implemented this yet.

So I decided to bite the bullet and build a uncompressed ingest system.

(Thanks to everyone and especially David Farland for inspiration and info!)

Results and MPEG vs Uncompressed HD comparison photos are posted on my site: www.PrimeHD.com

I'm now looking for a Varicam to check if our system works with it, in practice. If you have a Varicam in NY or Northern NJ area, please let me know (my web site has a contact form...)