Flash XDR

[Steve Phillipps]

Just a crazy thought - If Convergent Designs can make a box that takes an HD-SDI signal from the likes of Canon XL-H1, Sony EX1 etc. and record it at 160 mb/s 4:2:2 onto CF cards, why can't the camera makers just make the cameras do that internally? Would anyone buy a Sony EX3 with a data rate of 160 mb/s 4:2:2 that records to CF, even for a few thousand extra?
Steve

[George Kroonder]

Life is a compromise. Convergent Design actually uses a Sony MPEG codec in the Flash XDR.

Basically for development of a reasonably priced product you go for a "good enough" solution with a few distinguishing qualities.

35Mb/s VBR MPEG2 4:2:0 is definately good enough for most, if not all "prosumers" and even for lots (and lots) of pro's. That you can get 4:2:2 out before the encode/record is gravy.

I believe the EX1 is great as technically there are few compromises, and I don't really regard 4:2:0 as a compromise at all.

George/

[David Heath]

Originally Posted by Steve Phillipps

Would anyone buy a Sony EX3 with a data rate of 160 mb/s 4:2:2 that records to CF, even for a few thousand extra?

Recording 160Mbs does need high spec CF, but the XDR has proved that 50Mbs, (even 100Mbs at a pince) is recordable onto much cheaper cards, also giving 4:2:2.

All of which begs the question of why not use CF for an EX3 type camera, especially for 35-50Mbs data rates? SxS and P2 may have merit for high end cameras, but what advantage do they give for the large extra cost here?

[Dan Keaton]

Dear Steve, George, and David,

I have been working with Convergent-Design on the development of the Flash XDR for many months now.

We have learned quite a bit about the trade offs between various encoding rates as well as the visual differences between 4:2:0 and 4:2:2.

I agree completely that the Sony EX1 does a great job and has few compromises. 35 Mb 4:2:0 is a good choice in many situations.

We are in the process of testing various options. We are sharing the results of our testing in other threads on DVinfo.net.

We do see a visual improvement when going from 4:2:0 to 4:2:2. For example, fine text is more readable.

John Richards has loaded some of our test images into Photoshop.
Then he used the histogram function of Photoshop to compare the images. The histogram of the 50 mb 4:2:2 looks very similar to the uncompressed 4:2:2,while the standard HDV and DVCPro HD do not.

To see his results and conclusions, please go to Post 238 in the following thread and then download his results. Be sure to view all four pages.

http://www.dvinfo.net/conf/showthrea...102312&page=16

These test images are available on the home page of our website:

www.convergent-design.com

Posts 231 and 232 on the above thread provide more information about these images.


Our conclusions so far:

50 Mb 4:2:2 is visually better than DVCPro HD.

4:2:2 is visually better than 4:2:0 in the fine details.

100 Mb 4:2:2 Long GOP is visually very close to uncompressed 4:2:2.

50 Mb 4:2:2 Long GOP provides a high quality image.

Long GOP is an excellent way to get more visual data using a limited data bandwidth. (For example, 100 Mb 4:2:2 Long GOP is visually better than DVCPro HD (non-Long GOP) at 100 Mb.

The normal tradeoff using Long GOP is that a single drop out is a major problem. Using 30 frames a second, and 15 frame Long GOP, one usually loses 1/2 second of images if a dropout occurs.

Now, with the Flash XDR, creating two simultaneous original recordings to Flash XDR, a dropout should be a thing of the past. Even if one Compact Flash cards fails, a failure in the other Compact Flash card should be highly unlikely (if you are using quality Compact Flash cards). I recommend avoiding the counterfeit Compact Flash cards that are widely available.

We have not yet performed comparison testing on our 100 Mb or 160 Mb I-Frame only images. However, we strongly feel that visually the 100 Mb Long GOP is superior.

We estimate that it would take around 300 Mb I-Frame only to visually compete with 100 Mb Long GOP.

I-Frame only, of course, has its advantages. For ENG work, where speed from the camera to a final edited product is of the utmost concern, I-Frame only can be a good choice. So we provide 100 Mb and 160 Mb I-Frame only capabilities.

We feel that one of the benefits of using the Flash XDR is that you get to choose the bit rate and chroma subsampling option (4:2:0 vs 4:2:2) for each situation.

Wikipedia has a good article on chroma subsampling:

http://en.wikipedia.org/wiki/Chroma_subsampling

[Steve Phillipps]

Thanks Dan,
When you say 50 mb/s is better than DVCPro HD, do you mean from HVX or from proper DVCPro ie Varicam, HPX2100/3100?
Steve

[Dan Keaton]

Dear Steve,

We captured a fully uncompressed image from a Canon XL H1, into Final Cut Pro.

Then, for the DVCPro HD comparison, we encoded the file using the DVCPro HD codec in Final Cut Pro.

For an exact explanation of how we created each of our test images, please download the images from our website:

www.convergent-design.com

Each of the download files has a "Read Me" document that describes the processes that we used.

We did not use a DVCPro HD camera, as that would introduce differences in the lenses, sensors, viewing angle, etc. In other words, we used the same exact image so we could compare just the codecs (coder – decoder).

We do feel that Final Cut Pro has a quality implementation of the DVCPro HD codec.

With DVCPro HD being 100 Mb I-Frame only, it does not have the advantage of using the information from previous frames. As such, it requires a higher bit rate to achieve the same quality as a Long GOP codec.

In other words, a 100 Mb Long GOP codec can provide a higher quality image than can a 100 Mb I-Frame only codec.

What we found is that a 50 Mb Long GOP codec is visually better than a 100 Mb I-Frame only codec.

Also, we found that the DVCPro HD 100 Mb codec image is better than standard HDV at 25 Mb.

We are trying to be very clear in the procedures that we are used for testing. We hope the "Read Me" documents provide the information that you need.

We are also open to suggestions and criticism.

[Mike Schell]

Originally Posted by George Kroonder

Life is a compromise. Convergent Design actually uses a Sony MPEG codec in the Flash XDR.

Basically for development of a reasonably priced product you go for a "good enough" solution with a few distinguishing qualities.

35Mb/s VBR MPEG2 4:2:0 is definately good enough for most, if not all "prosumers" and even for lots (and lots) of pro's. That you can get 4:2:2 out before the encode/record is gravy.

I believe the EX1 is great as technically there are few compromises, and I don't really regard 4:2:0 as a compromise at all.

George/

Hi George-
I certainly agree that the EX1 is a excellent camera and the at 35 Mbps 4:2:0 gives great results for many applications.

Flash XDR enables you to significantly improve the video and audio quality without jumping to a $30K+ price range. Some applications, such as film-out, stock footage, green/blue screen, or delivery to National Geographic, BBC, or Discovery Channel demand 4:2:2 high bit-rate quality. Flash XDR enables your $7K camera to perform at levels close to a $30K camera (the lens may be a notch lower, but the recorder is a definitely much better).

So it all depends on the job requirements. However, Flash XDR does offer some compelling advantages in terms of media cost as well as a content delivery device. 32GB Compact Flash cards are $150, while 16GB SxS cards are running $850. So, on a GB basis, CF cards are 1/10 that of SxS. If you buy many SxS cards, you will quickly surpass the cost of Flash XDR.

Also, Flash XDR can be used as a content delivery device for trade shows / client preview or transporting your finished video to a post house for output to tape. As Dan Keaton mentioned, we find the 100 Mbps Long-GOP to be near visually lossless to uncompressed.

Yes, I agree, life is full of compromises. Flash XDR may be an overkill for many applications, but for those jobs that really need the best possible quality, we think XDR is a great value.

[George Kroonder]

Hi Mike, Dan,

I think Steve was merely asking what reasons Sony could have (had), not to implement a recorder like the XDR. With better quality and cheaper media.

Regarding the media; my opinion is that Sony went with SxS as the ExpressCard technology would have a longer projected lifespan, which they could also control better. It would also enable the cards to be used in higer end models, providing a way to 'grow' to bigger/better cams, keeping your investment in media and being able to share cards between cams (standardizing). It was mainly positioned against P2 I believe.

Flash cards are basically simple memory blocks; the SxS are a little more complicated. I don't think you can detect the speed/type of Flash memory, so responisibility for using 'workable' cards lies with the user. That is not a good plan for a mass production "prosumer" camera.
Also I beleve at time of development (couple of years ago) there were no flash cards that were fast enough, so partnering with Sandisk to develop SxS was a good choice at least from a product & projectmanagement standpoint.

As to quality, well product is developed for a market/segment and I believe Sony hit a sweet spot. In the end everything is built to some spec and cost. While you should always try to apeal to as many customers as possible, leaving room for "upselling" to the next model is a good policy. Where to place that "feature cap" is always a delicate decision.

Not having everything is also good for "add on" manufacturers; the XDR adds a "better recorder" to the mix, but they aren't limited to EX1 owners. Many more cams can use this to upgrade their rig. It adds more choice to the mix, and in and on itself that is a good thing.

Both Sony and Convergent Design can build their products "for the masses", which is good for us, the users.

George/

P.S. I know I'm shot-changing the XDR a bit Mike, but only for the sake of topical clarity!

[Steve Phillipps]

George, it's not so much the media I was wondering about it was more the uncompressed recording to any sort of media. If the camera can output through a socket on the back which Convergent Designs can make use of why don't Sony et al just enable it in some form in camera?
Dan, I suppose what I'm really asking is to give an idea if the output will compete with or even out compete the image from a Varicam or 750, as that's what I know BBC etc. are happy with.
Thanks all,
Steve

[Dan Keaton]

Dear George and Steve,

These are not really easy questions, but I will attempt to answer them. Please remember that I can not speak for Sony.

The Sony EX1 is a great camera and produces a great image.

In comparing it to cameras that cost many multiples of its cost, one has to take into consideration the lens and many other factors.

With the much more expensive cameras, one uses a much more expensive lens. Now, this does not mean that the much more expensive lens is better, but in some cases it actually is.

There is an economy of scale with cameras and a manufacturer can put a much better lens than one would normally expect, on a lower cost camera, if there is a wide market for the camera.





Yes, the Sony EX1 can output a very nice HD-SDI signal with 4:2:2 chroma subsampling.

But it can not actual record a 4:2:2 signal in the camera. The Sony EX1 has a wonderful, custom designed, Sony MPEG2 encoder chip in the camera. With only one of these chips, the camera does not have the horsepower to actually encode 4:2:2.

So to answer one question, why didn't Sony enable 4:2:2 encoding within the camera (regardless of the media): it could have been the cost, or space, or power requirements, etc., to put two of these encoding chips in the camera.

It is also a trade off. Remember that 4:2:0 35 Mb Variable Bit Rate MPEG2 encoding looks very nice visually.


Steve, in time we will be able to tell if the Sony EX1, with the Flash XDR, be able compete with the Varicam or 750. Personally, I believe that it may be a close race, but there are differences between CMOS and CCDs and there can be substantial differences between lenses and many other factors.

While we will be testing the Flash XDR with a Sony EX1 soon, the best tests will occur when we get the units in the hands of our customers. Then they can compare it to the wide range of cameras at their disposal. And we hope they will post their results.

George, you made some points concerning the advantages of SxS media, specifically about it communicating its speed with the camera.

Since there are a wide range of Compact Flash memory speeds, we will, in the Flash XDR, qualify a card to ensure that it will meet the recording demands of the selected bit rate. We may do this when a card is inserted, or as a menu option so you can pre-quality your memory cards.

We will also be testing cards at Convergent-Design so that we can recommend qualified cards for various bit rates.

Steve, another question is why don't other manufactures build in tapeless recording. I believe that there is a solid trend for camera manufacturers to do so.

We are seeing more cameras equipped with HD-SDI, which is a good thing. I wonder if we will see new cameras that can do what the Flash XDR can do.

For example, will new cameras have four Compact Flash (or other media) slots?

Will they be able to record both 4:2:0 and 4:2:2, at user selectable maximum variable bit rates?

Will they be able to create two master copies simultaneously?

Will they be able to switch to another media slot when a media card fills up?