Piotr

I zoomed in 400x on the line in the left hand corner and this is the image I'm getting. I think its obvious which is which so obviously the HDSDI Output is giving better quality is this the difference your seeing?

You'll need to download it and zoom in

Mark, there is no need to look that deep into these pictures - just take a look at the foreground red areas, and specifically compare:

- macro-blocking inside them
- edge definition between them, and the neighbouring whites
- the thin horizontal line (of red on white background)

I hope you can see it now :)

No sorry cant see a difference on this screen. However I have to now agree there is quite a difference Obviously not one I can apreciate on my limited resolution. But I really should invest in a decent HD monitor. I'd be interested in seeing more tests done still and am now interested much more in the Nanoflash. I know its cheap compared to whats gone before. Just wish it was a bit cheaper still!

Let's not forget that, while the HD-SDI output is "uncompressed", when it gets recorded by (nano)Flash, it's being mpeg compressed. Just not as much as in the native EX1 scenario.

Exactly - and, while before the 420 vs 422 hype started, I was aiming at the 50Mbps, 422 format as my "sweet" nanoFlash setting (with direct HD422 MXF support in Vegas), I'm afraid that with more goodies coming from relaxed compression than colour sampling, we might need the 100Mbps to really get what we're after... With the I-frame only, 160Mbps for "special purposes" only.

As to the uncompressed (only available on the bigger XDR box), I personally think I don't need it - with all the storage requirements being an overkill, considering my real editing scenarios.

Sorry Bill, I wasn't clear. I was actually referring to the uncompressed samples captured directly to a Blackmagic or Kona card when I made that statement, though I understand you were referring to the lesser compression that will be offered via Nanoflash. In either case, the macroblocking will be minimized.

-gb-

Yes, that image alone easliy shows 4:2:0 vs 4:2:2 difference.

Hi Bob-
The live HD-SDI (or HDMI, for that matter) output is derived from an internal video processing chip, which gets it's input from the image sensor A/D converter. The output of the video processing chip is simultaneously fed to the analog encoder (analog component output) and the HD-SDI (or HDMI) driver circuit. So, the HD-SDI/HDMI output has not seen any analog processing except for the original A/D conversion off the CMOS/CCD sensor(s).

Also, just to clarify another point of confusion, HD-SDI is always, always 4:2:2 10-bit, full-raster (1920x1080 or 1280x720). No other formats are permitted. Yes, the video stream may have originated in another format (such as HDV which is 1440x1080 4:2:0 8-bit), but it absolutely has to be scaled, upsampled or appended with 2 lower (zero) bits (to go from 8-bit to 10-bit) accordingly to meet the HD-SDI specifications. There are no exceptions permitted!

Hi Bill-
Just wanted to point out that the Sony MPEG2 module used in XDR/nano actually has two hardware CODEC chips for 4:2:2 processing (same chip as used in the EX1/EX3, but two instead of one). I suspect that Sony sends the luminance data to one CODEC and the chrominance data to the second chip. You therefore, should have considerable more processing power to further eliminate motion artifacts.

Many people examined the high-motion airplane footage we played this past week at IBC. No one noticed any blocky or dropped frames whatsoever. We showed some very high-motion scenes with water and smoke. All captured at 50Mbps 4:2:2.

So, the higher bit rate should help, but the additional MPEG2 processing power should also minimize motion artifacts.

wow! great info. thanx, Mike.

Thanks Mike...
Man, the XDR/Nano are really going to offer a lot for the XDCAM users.
I'm looking forward to owning one.

I haven't had time to get back here and join in the discussion, but I've taken that original side by side image I posted, kept it at 100% native rez and made a layered Photoshop file out of it. You can now just toggle the top layer on and off to compare, while zooming your view to any degree of magnification you want.

http://ftp.datausa.com/imageshoppe/o...5mb_layers.psd

I also mention an error, on that original image I put text saying it was 600X. I meant 6X or 600% enlargement. Sorry for the confusion.

In addition, in hopes of making this even more clear, I did another test, this one photographing solid chroma shapes with vibrant colors side by side, and captured SXS and 8bit 4:2:2 uncompressed and made a layered Photoshop file out it, same as above, as well as a 400% enlarged side by side .png for easy web viewing.

http://ftp.datausa.com/imageshoppe/o...b_layers_2.png

http://ftp.datausa.com/imageshoppe/o...b_layers_2.psd

The results are very striking, as it also shows some of the subtle textural differences between the two color spaces as well as the broad edge differences. It also should make clear that the HD-SDI out isn't simply chroma-smoothed 4:2:0 passed off as 4:2:2.

Finally, I want to point out a few of the reasons I'm personally excited about the NanoFlash.

First, it helps keep the EX1 and the work I do in green screen competitive with other camera systems, such as full blown F900 and RED packages. I know one client of mine interested in the NanoFlash as a record alternative to HDCAM for his F900 for green screen shoots. Currently, the only way he can get this high of quality is to record out the F900 "live" into either a single channel of an HDCAM SR deck, or into an NLE system.

Second, the NanoFlash gives you the recording quality of two generations down the line in Sony prosumer camera systems with today's existing camera. How so? As Mike mentioned, they're using Sonys MPEG module which is already over designed for the needs of either the EX1 or any other Sony mid-range system. It will be the cornerstone of several generations of cameras. I'm sure the follow-up to the EX1/EX3 will be 50 Mb/sec 4:2:2, and the generation after that might be 100/160 Mb/sec with I-frame recording ability. You get that TODAY. Not three years from now.

Third, for client playback and special presentations in D-Cinema theatre, the NanoFlash will allow you to walk in and play back something from a device the size of a pack of smokes that blows away not only Bluray, but an HDCAM deck.

Cool beans, I say.

Regards,

Jim Arthurs

I've said this before and I'll say it again - what motion artifacts?? I've been using the EX for lots of fast moving aircraft filming and I have not seen ANY. I also see no more loss of definition in fast pans than I do with footage filmed on our HDCam790.

Jim - excellent demonstration! That should settle this once and for all. And you make some very good points about the Nano-Flash as well. Couple them with it needing memory a fraction of the price of SxS or P2, and outperforming recording wise all the SxS/P2 systems currently on the market and it really makes you think.

It does leave me wondering how much of the difference is due to colour space, and how much due to the compression. I'm assuming the examples are with an interlace signal signal, it would be interesting to see the difference in progressive mode.

It would also be interesting to see the comparison you give with the XDR 50 and 100Mbs modes as well.

Actually, this latest test was my first time out with the new 23.976 PsF HD-SDI mode... my EX1 just got back from service with new firmware and this is a new feature. Nice not to have the 3:2 padding eating up precious uncompressed space!

I haven't tested the 50 Mb/sec mode, but from the limited samples I've seen, I feel it is probably the same level of compression as current EX1 35Mb/sec, with the difference in data rate to handle the increased color space. The 100 Mb/sec is where you get your big compression benefits.

Regards,

Jim ARthurs

Jim, once again your information has been EXTREMELY valuable.
You have always been a great asset to these technical video conferences.

If these examples do not bring closure to this, nothing short of branding 4:2:2 on the side of the camera and having the president bless it will. LOL

This example is a PRIME 4:2:2 example.
http://ftp.datausa.com/imageshoppe/o...b_layers_2.png


Not to mention using SDI to capture the cleanest possible from the EX1 / EX3.

Cineform is also working on its own hardware recorder that will have HD-SDI option.

That one captures directly into Cineform Prospect HD format on CF card.

Here:

CineForm - Frequently Asked Questions

Thanks Steven! Of course if companies like Sony and Panasonic got all their folks on the same page, there would be lots less conflicting information floating around. It's hard to fault any individual who is told one thing when something else is being said by another from the same company.

Alex, here's a real world example of the difference in a key between 4:2:2 and 4:2:0. This is a bluescreen foreground shot by the EX1, keyed over a grey background in Digital Fusion. On the left is my uncompressed 4:2:2 capture, on the right is unfiltered 4:2:0. Notice the lack of definition between the shirt and background and the cross-hatch chroma stepping on the shoulder.

http://ftp.datausa.com/imageshoppe/o...p_vrsXDCAM.jpg

In this next example, I've used some color space conversions and a blur to blur just the chroma and not the luma... just enough to blend out the cross-hatching in the chroma. Notice that while smooth now, the fine distinction between the shoulder and the background is gone.

http://ftp.datausa.com/imageshoppe/o...CAMblurred.jpg

I like CineForm... any idea how far along this product is?

Regards,

Jim A.

That's a good question shared by many.
One thing for sure, with Cineform codec many will be very interested in this product.

With all these SDI cameras now on the market and more becoming available, I'm hoping they release an SDI verison first.

Jim - your stills inspired me to look at my footage again, now at 4x and 8x magnification, frame by frame.

Yes, now I see that there's a clear difference in quality, HD-SDI being noticeably better for color fidelity, and even shapes of objects (I have a small red area surrounded by blue; in SxS footage its shape is slightly wrong because blue did bleed into red area. HD-SDI one shows correct shape and colors, as well as significantly less artifacts. Also greens overall are slightly better with HD-SDI. I did not notice much diff with blue channel, surprisingly to me.)

To me, this shows that EX1 in fact outputs different quality (better) signal over HD-SDI than being recorded to SxS card.

When I'm looking at the image at 100%, I can still see some difference in the problematic areas.

So I'm now convinced that EX1 in fact does output higher quality video over HD-SDI. Thanks for insisting, and for illustrating your point (although I have not seen such pronounced stepping in my own greenscreen images, thankfully. I suspect this is because your model wears green shirt that might have some blue color in its fabric, which results in the keyer eating into it at lower sampling. Just a thought.)

So I guess now we are down to practical ways of capturing that HD-SDI signal, especially in the field. (I do have a PC built with Hd-SDI capture card for in-studio work, and it is big and not very well suited to be hauled around...)

Since nanoFlash does not seem to offer uncompressed capture like you did in your illustration, how much of a quality difference can we expect in SxS vs nanoFlash?

(I'm not worrying about Cineform's upcoming recorder, btw - all my current footage is in Propsect HD codec, and I'm very happy with it. I just have to convert from MP4 - SxS card - or use HDlink for in-studio capture. If the video was recorded in Prospect HD in the field - all the better. Cineform says that they are working on the recorder... not fast enough, if you ask me! :)

......

Ditto!

Yeah... I must say that these examples from Jim not only show the even more pronounced difference between SxS compression and HDxSDI. They are also an eye-opener on the quality of SxS chroma keying being much lower than I expected!

Like Alex, I've never seen such a pronounced stepping in my (limited number of) keying.

Alex, I don't think the (nano)Flash (XDR)'s 50 or 100Mbps compression will "spoil", or negate, the 4:2:2 advantages over 4:2:0 - the difference vs fully uncompressed should only influence motion-related phenomena. OK, perhaps some more tendency for macroblocking than in the uncompressed - but edge definition, I expect to be equally good.

Here is the chart that shows the quality of compression vs non compression listed on
the CD site...

http://www.convergent-design.com/dow...ty%20Chart.jpg

You should see little if any difference between the non compressed footage from the HDSDI port ingested with the Sony Codec at 100 mbit ( Nano/XDR ) and the actual non compressed footage.

Looking at the chart, it only goes out to 50 mbps for the Sony PDW-700... you can put the
EX1/EX3 a little further out on the graph at 35 mbps and you will be able to see how the Nano/XDR will handle the non compressed footage at 100 mbps... it is approaching the HDCAM SR performance which is 440 Mbps at 1920 4:2:2 10 bit.... :-)

I'd have to say Sony has fine tuned this CODEC quite well....

While we are talking about the output quality of the camera I'd like to ask another question that some of you can answer please....

I know that no matter how good any footage is going into the encoder for DVD/Blu ray
authoring, you can actually damage the quality of the footage with the compression going to the disk during the write process....

I know that the compression is dependent on how much footage you plan on putting on the disk.. but I'd sure like to know at what rate can you put the footage on the disk before
you begin to introduce compression artifacts.... and what to use, CBR or VBR ....

It would be a shame to pull the footage from the HDSDI port and then just mess it all up
during the authoring process....

I can't address specifics of DCT compression(what's used on mpeg for DVD), but, I can say that the artifacting introduced in DCT is proportional to the frequency. What this means, in practical terms, is that the compression algorithm is stressed when there are very fine details in the image, such as foliage, waves on water, etc.

So, if the in-camera compression introduces small detail artifacts, the artifacts are greatly amplified in the compression process out to DVD/Bluray. Not only are the artifacts made more noticeable, but, the amount of disk space needed for a highly detailed image frame is higher than for an image frame with less detail. If you're up against a disk storage limit, you will then have to increase the compression ratio to get it to fit on the disk. End result is you get hit twice on quality because of the noise introduced on the in-camera capture process.

What I don't understand is how nanoFlash (or other) will be able to feed a compact flash card (or even 4 of them at the same time) with a 100mbps data rate... If it would have been so simple, why would Sony use SxS ?
I think there is some kind of mistery here...

Who knows what Sony's reasons really are?

My opinion? Fast enough CF technology wasn't ready when the EX was being planned for. Sony wanted a card format that would work with mobile devices drectly at max speed. There is a progression from PCMCIA/Cardbus to ExpressCard as well.

The interface is more than just a "memory card", making additional options possible (PHU-60K, maybe even eSATA later on). And I also believe Sony wanted a "Professional" solution for which they could control the media quality (although by using the USB interface for the PHU-60K that seems to be out the window).

Anyway just my 2c.

George/

SxS has other advantages. SxS is very fast so it has plenty of head room for future development. It is very difficult to corrupt the file structure on an SxS card. Pull one out of the camera mid-shot or remove the power from the camera mid shot and the most you will loose is the last 4 seconds of the shot.