USB 3.0 Camera - a hint of the future?

[Karel Bata]

USB3 has been talked about here before, but I think that most folk see it as a way of getting quicker backups to external drives. Actually it offers a whole lot more.

Here's an interesting development. Sony prototypes a USB3 camera that records HD 1080P60 direct to hard drive: http://bit.ly/USB3Camera

Now imagine a RED on the end of that cable, and a Mac on the other. That's goodbye to CF cards and the Red-Drive... (well, on a lot of shoots anyway)

From their site:

"We have no plans to release this particular prototype camera as a product. We built it solely as a technology demonstration to showcase the benefits of SuperSpeed USB (USB 3.0).

"USB 3.0 will be a key digital interface in the imaging and vision industries in the years to come. By introducing the world's first USB 3.0 camera, we are demonstrating that Point Grey is at the forefront of innovation in imaging (which just happens to be our corporate tag line) and that we are committed to the advancement of USB 3.0 technology. We also hope to raise awareness within the computer and machine vision industries of the performance benefits of SuperSpeed USB."

But a USB connection doesn't just transfer data one way. The first thing I thought when I first encountered the RED's menus was "Isn't there some way to access all this from a remote workstation?" and adding a USB3 connection to allow a live feed allows for exactly that. All those menus in an easy to use interface? And without having to physically go anywhere near the camera? Brilliant!

It would also be a boon for multi-camera shoots too - those with fixed cameras, or tied to a tripod or dolly. You could fed all those USB3 connections back to one central point where the recording itself takes place.

And there's more! What about being able to adjust camera speed in shot from the Mac? Or any other such parameters? Even the focus...

Some way off, but bound to happen I'm certain.

Anyone here able to say what the limit on image size is if it's fed live down a USB 3.0 connection? (I can't be bothered to look the data up)

...

Interesting to see too how quickly other manufacturers get in on this too. It certainly heralds the death of the DSLR for video.

And it's also great news for indies looking to get into 3D on the cheap! You can put two of those cameras together with the proper i/o of 2.5" (something you can't do with DSLR). In fact it looks like you can get them even closer. In fact, you can put them at any distance you want! And goodbye to those cumbersome mirror rigs! (which means you gain a stop)

...

Thoughts?

[Jeff Kilgroe]

Anyone here able to say what the limit on image size is if it's fed live down a USB 3.0 connection? (I can't be bothered to look the data up)

USB3 isn't anything magical. Other interface standards are continuously evolving too. USB3 is technically a bit faster than SATAII, but nowhere near as fast as 6Gbps "SATAIII" it also has higher latency and for direct streaming data to be sustained at a high rate, SAS or SATA are still very much superior.

USB3 is 4Gbps with the ability to do 4.8Gbps in burst or using the "Super-Speed" bus setting, which imposes some restrictions. The marketing calls this mode "5Gbps", which it is not. If we cut out the overhead imposed by the communication protocol and data packing, we actually have a real-world maximum of about 3.2Gbps in standard mode or about 3.8Gbps in "Super-Speed".

In my own experience with recent experimentation on USB3 devices, I'm getting much better performance by connecting HDDs to an eSATA port than by connecting to USB3. Of course, there's an extra interface in the middle within the drive enclosure, going the USB3 route... But at this time, I don't see an advantage to it. I do expect this to improve over the coming months and more USB3 interface options, devices and chipsets enter the market. But don't expect miracles of any sort...

Firewire3200 has lower latency and overhead and is about on par for speed with SATAII. It's 3.2Gbps before overhead and operates around 2.8Gbps for raw data throughput after that overhead is considered. FW3200 seems to be getting ignored -- just as FW800 was by most of the market.

We have other standards coming down the pipe in the form of C.Fast -- the next iteration of Compact FLASH cards. They're roughly the same size, but based on SATA and 6Gbps interface is allowed in the spec.

All this aside, the interface isn't so much important as the devices it is connecting. For example, if we look at CF/C.Fast media, it's great that C.Fast is hitting the market... However, there are no CF cards out there that can perform fast enough to utilize all that the ATA-based CF standard gives us.

As for a camera like Epic/Scarlet, I expect them to base their interface off of whatever media they are writing to. So, the interface will probably be contained within the media module[s] installed on the camera and they can be changed and upgraded. For now, it makes the most sense to stay with SATA or similar as it still is the most efficient.

[Dan Hudgins]

Recording live streaming high resolution video to a computer is not as easy and reliable as on might think, at least if you are recording the full true RAW data.

I had a drive speed test program I wrote up here a while back and some kind RED USERs helped alpha test it for me. When I get time I planned to revise it and post it for "beta test" on my web site.

Recording to a SATA II HD I have not seen anything close to even 1280x720x12bit@24fps to work with an un-buffered camera, so far.

It seemed possable to record 2592x1102x12bitx24fps into two SSD in Raid 0.

I have talked to other people who make or have tried such cameras, and a fast computer is needed to avoid data glitches from the OS, it can be better to use hardware desigined for real time "on-the-fly" recording rather than a random PC or notebook.

If the camera has a FIFO, then things get much better since the FIFO can make up for the delays in the PC which can be over a second when recording to SATA II HD disks, I don't know what takes so long, but without a buffer, Raid SSD seem to be a better choice, and putting Raid SSD into a notebook may not be that easy.

As far as the path into the computer, some kind of buss card may work better than a standard port because even recording 150MB/s non-stop can be hard on a PC.

What can work better is a "shoot-then-dump" camera that records to its own FLASH or DRAM for a minute or more per shot, then dumps that over highspeed connection to a notebook, that way no mater how slow the PC and storage are you don't get glitches or dropped frames.

[Michael Romano]

I always sort of thought that the CF cards and Drives were a forward thought to get away from SRW decks and Dual Link HD-SDI cables littering the floor. Its a lot easier when I just have a single HD-SDI cable for DIT to get a peak and wireless SD for everyone else. Sometimes the cable ideas sound great but never forget how they play out on set when the company is wanting to move fast.

[Karel Bata]

Thanks for the tech update - I am truly humbled by the technical know-how of some of the folk here!

But also puzzled by Point Grey (aka Sony) plugging it as the future when they presumably know better. Personally, I would have thought the future with such a connection would be in some fibre optic link. It wouldn't have limits on it for length, and could be very thin, though maybe armoured.

Michael, I wasn't suggesting that everyone go over to a wired connection! It would be added as an option, and if it did appear I think a lot of people would go for it. As to any inconvenience regarding when the camera is on the move - I imagined a simple connection that could be unplugged at any time. Would be great for cranes.

[Karel Bata]

Who was it who last week said "Personally, I would have thought the future with such a connection would be in some fibre optic link" ?

Oh, it was me.

And now it's here:
"Telecast Fiber Systems (Worcester, USA) today introduced the CopperHead 3400, an innovative new fiber optic video transceiver specifically designed for high-end 3D cinema production."

http://www.stereoscopynews.com/index...=891&Itemid=34

[Michael Hastings]

The hint of the future isn't the interface, it's that point grey's supplier has already announced an off the shelf 4096x3072 5.5 micron APSC+ (4:3 anamorphic anyone) sensor that does 300fps@10 bit and 180fps at 12bit and has an HDR mode with all kinds of on-chip tech to make integration of various modes easier. I don't think pricing has been announced but I suspect less than $500 (please correct if anyone knows), since they are going toward much higher volume markets. No reason PG, SI, Phantom, Panavision or others can't upgrade their existing technology using it and use something like cineform to compress it.

That's why I'm champing at the bit to get Epic sooner rather than later.

[Jeff Kilgroe]

Announcing a product and delivering are two different things, as all of us here should really understand by now. If they can deliver such a sensor, which they may as well, then great. It won't be long before many others can offer those specs or even better anyway.

Sensor technology is improving at an exponential rate. Companies like RED must continue to move forward with new designs and technology. And the sensor is only a piece of the entire package. PG could release the most amazing sensor ever, but if camera makers can't or won't pick it up to build an amazing camera round it, or won't do it in a timeframe that brings it out as a competitive offering, then it's just another cool tech gadget that could've been. We se lots of those all the time...

It's only going to be a few more years before our iPhones have a 1080p native display and the built-in camera shoots 2K/1080p video without shutter skewy jell-o and also offers 18 megapixel HDR stills.

[Michael Hastings]

Announcing a product and delivering are two different things, as all of us here should really understand by now. If they can deliver such a sensor, which they may as well, then great. It won't be long before many others can offer those specs or even better anyway.

....
.

Pt grey is just an integrater, the sensor company Is a substantial player with existing products and leading edge tech - 3or4 yo startup whose principals sold their previous sensor company to one of the biggest.

Fab is one of the most well established. So I think the sensor is for real, and probably deliverable on schedule. Integration into a camera is another story.

[Dan Kanes]

The control aspect can be done with a simple serial interface. Controls don't require any more bandwidth than USB 2.0 has.

As far as capture - I guess the fact that USB 3.0 has bandwidth at all is what it allows it to be an effective capture interface.

Redmote (and what I can see as some enchancements of that concept) will be the future of epic remote control.

Capture and control are two different things, and the power of solid state recording is that it opens the doors for a variety of workflows.

Now do that without wires ;)