HDV Canon Vs. HDMI Sony For Chromakey!

[Mike Schell]

I can absolutly assure you that the HDMI output from a "live camera" has never been compressed. MPEG2 compression adds anywhere from 15 to 30 frames of additional latency to the signal. We see this latency when an HDV camera is connected to our HD-Connect SI box, which does HDV -> HD-SDI conversion. There is an undeniable delay of the video and audio signal (about 1 second). However, when we use our HD-Connect MI (HDMI -> HD-SDI) converter there is no perceptible delay in the audio and video.

Furthermore, if the HDMI output from a "live camera" was indeed compressed, then the camcorder would need to perform both a compression and decompression process simultaneously. This would add even more latency and further increase the power consumption (always a concern for battery powered devices).

So, the HDMI output from a live camera has indeed never seen a compression, just like the HD-SDI ouput from the Canon (in live mode). Sony likely has a number of marketing reasons to not advertize or promote this capability.

Mike Schell

[Harm Millaard]

Originally Posted by Mike Schell

So, the HDMI output from a live camera has indeed never seen a compression, just like the HD-SDI ouput from the Canon (in live mode). Sony likely has a number of marketing reasons to not advertize or promote this capability.

Mike Schell

If you are correct (I'm still waiting for confirmation by Sony) then the marketing reasons of Sony seem like BMW introducing a new 2.0 liter 4 cylinder car, which underneath has a 4.2 liter V8 but does not want the public to know. Does not make a lot of sense, does it?

From another perspective, how do you explain that Sony confirmed that their single chip range of cameras, the HC3, UX1 and SR1 output compressed HDMI signals @ 25 Mbps? That there is no way to switch from a 1.485 Gbps stream to a 25 Mbps stream? This means that you can only use HDMI live and not afterwards. How do you explain the fact that the HDMI connection is TV compatible, but the TV HDMI input is limited to 25 Mbps and the camera has no way of switching between the two different signal streams?

How do you explain, from Sony's perspective, the price difference between the XDCAM-HD 330 and 350 due to HD-SDI (and some other differences), the relatively high price of the Canon H1 with HD-SDI, the price differential between the Canon A1 and G1 with HD-SDI and a long list of possible other examples?

If you are correct, Sony is definitely embarked on a route that can be designated as a price fighter. That is a new one for me. Also, for the first time in their existence, they act mute in promotional terms.

Sorry, I just don't buy it without confirmation from Sony.

Mike, please do not consider this a (personal) attack. Just expressing my doubts.

It might be worthwhile to check whether a HD-SDI signal from a camera (H1, G1, 350 or similar) converted to HDMI can be directly displayed on TV during live recording, and repeat that for HDMI. I assume (hopefully I'm wrong) HDMI is no problem, but the HD-SDI data rate is just too much for the TV and chokes on it.

Just in the lucky case that I find a 4.2 liter V8 under the hood of the car sold as a 2.0 liter 4 cylinder, that nobody told me about, I consider the start of 2007 very, very good.

[Barry Green]

Not sure why you think this is so breakthrough of a feature -- every HD camera on the market outputs uncompressed high-def through at least analog component outputs; the XHG1 and XLH1 do so through genuine HD-SDI, and the tiny little Sony HC3 and the AVC-HD cameras from Sony and Panasonic all output uncompressed digital HD through their HDMI ports.

It's a great feature, sure, but it's not like it's a) hidden, or b) unique to the V1U. The HC3 brought uncompressed HDMI output to the market almost a year ago.

[Barry Green]

Originally Posted by Harm Millaard

From another perspective, how do you explain that Sony confirmed that their single chip range of cameras, the HC3, UX1 and SR1 output compressed HDMI signals @ 25 Mbps?

I would chalk that up to a misinformed individual giving you erroneous information. HDMI is always uncompressed.

The only way they could be outputting the 25mbps signal would be if they uncompressed it to full uncompressed before transmitting, and you'd know that real quickly on an HDV camcorder because you'd see a 1/2-second lag between action in front of the lens and when that action shows up on the monitor. HDV has to buffer 1/2 second of footage before it can compress a GOP, so HDV output is always delayed by a minimum of 1/2 second (shorter GOPs can reduce that time, but Sony always uses 1/2-second GOPs).

So -- if someone were to report that the HDMI monitoring on their Sony camcorder was lagging 1/2 second behind reality, then yes that would be an indication that the HDMI output was perhaps being compressed to HDV, and then uncompressed and output over HDMI. But AFAIK there is no lag, which would in and of itself be proof positive that the HDMI output has not undergone 25mbps HDV compression.

[Harm Millaard]

Originally Posted by Barry Green

Not sure why you think this is so breakthrough of a feature -- every HD camera on the market outputs uncompressed high-def through at least analog component outputs; the XHG1 and XLH1 do so through genuine HD-SDI, and the tiny little Sony HC3 and the AVC-HD cameras from Sony and Panasonic all output uncompressed digital HD through their HDMI ports.

It's a great feature, sure, but it's not like it's a) hidden, or b) unique to the V1U. The HC3 brought uncompressed HDMI output to the market almost a year ago.

and this is exactly what Sony denied. According to them it is a compressed signal being output for their single chip camera line.

[Mike Schell]

Hi Harm-
This is also not a personal attack, but I don't think you have a clear understanding of HDMI. May I suggest that you visit hdmi.org and also download a copy of our white paper: "HDMI in HDV and AVCHD camcorders".

Firstoff, HDMI only supports uncompressed video, not 25 Mbps compressed video. Camcorders and decks use an MPEG2 decoder chip to decompress the (25 Mbps) HDV or AVCHD video/audio before it is output on the HDMI port (this assumes playback from a tape, DVD or hard drive). In the case of live capture, the video stream out the HDMI port has never been compressed.

Our HDMI white paper, available from the Convergent Design web site, shows the video/audio dataflow through the camcorder/deck. This diagram should help everyone better understand how the video and audio are processed.

I can't comment on the marketing reasons behind adding HDMI output to camcorders and decks. I do know it works and produces excellent quality video/audio.

Yes, 2007 will indeed be a very very good year.

Mike Schell
Convergent Design

[Ali Husain]

lots of good responses to the initial line of questions. i'll just add this...

Originally Posted by Mike Schell

I can't comment on the marketing reasons behind adding HDMI output to camcorders and decks. I do know it works and produces excellent quality video/audio.

probably a good guess that the hdmi connectors are so that people can conveniently connect their cameras to their hdtv's to playback footage, not so that people can connect their cameras to raw video acquisition systems. that would be a nice side-effect which has only recently become cost-effective (cineform, blackmagic, fast laptops, harddrives, etc.).

[Barry Green]

Originally Posted by Harm Millaard

and this is exactly what Sony denied. According to them it is a compressed signal being output for their single chip camera line.

I would suggest asking someone else then, as apparently the person who answered your question was misinformed.

[Thanasis Grigoropoulos]

As usual, a thread full of useful replies!

Some more questions though that I am really hoping for an answer to:

1) If you want to capture live using the HDMI but without using RAID, which is the best way to go, for bluescreen work? Is there a way to keep full 4.2.2 without using DVCPROHD and thus scaling down the image? And if there is such a way, what are it's potential drawbacks?

2) How about this Gefen adaptor? Feed component and sound, export HDMI. Does it work? What are the loses? Or what could be the loses? (I mean, technically, is there a lot of ways to convert component to HDMI?)

3) Thomas: You mention in your post that "if you have a progressive frame sitting inside of a 1080i stream then you end up with chroma jaggies because the chroma samples may not line up exactly with the luma samples because they alternate every other line." From what I gathered you are discussing the V1's recording of progressive signal on tape. How about the signal of say A1's 24f from component? From what I understand, A1's component signal is always interlaced - therefore another "progressive frame sitting inside of a 1080i stream." Would that mean that this signal would suffer from the same problems (in reference always to bluescreen work). Given Mike's point that component's signal is exactly the same as SDI, I guess bluescreening with H1 -> 24F -> SDI would give the same results (and problems, IF any).

4) Finaly, one thing I really don't understand: How come the sensors of these cameras capture 4.2.2 and not 4.4.4? I really cannot understand what these sensors "see"! I thought sensors "see" all the colour spectrum and AFTER they capture that, then their signal is processed and encoded to whatever codec is to be used. Andromeda plug in for the DVX claims a full 4.4.4 capture out of the sensor. How come these HD sensors, made so many years after DVX, capture only 4.2.2? Please excuse my ignorance! I just cannot understand how this works!

Thanks a lot,

Thanasis

[Thomas Smet]

Originally Posted by Thanasis Grigoropoulos

3) Thomas: You mention in your post that "if you have a progressive frame sitting inside of a 1080i stream then you end up with chroma jaggies because the chroma samples may not line up exactly with the luma samples because they alternate every other line." From what I gathered you are discussing the V1's recording of progressive signal on tape. How about the signal of say A1's 24f from component? From what I understand, A1's component signal is always interlaced - therefore another "progressive frame sitting inside of a 1080i stream." Would that mean that this signal would suffer from the same problems (in reference always to bluescreen work). Given Mike's point that component's signal is exactly the same as SDI, I guess bluescreening with H1 -> 24F -> SDI would give the same results (and problems, IF any).

Thanks a lot,

Thanasis

This is a good question. One thing to remember however is that video coming out of the component, SDI, and HDMI ports assumes uncompressed video. If 24F is on the tape it will come out the ports upconverted directly to 4:2:2. It may not look any better but the chroma samples will not have to alternate because the output isn't 4:2:0 but 4:2:2. It's like how DVCPRO HD will be fine as interlaced footage because it is 4:2:2 by nature. A progressive 4:2:0 should decode to 4:2:2 better then a Interlaced 4:2:0 would. Not to mention progressive 4:2:0 will always have less artifacts and give an overall cleaner compression. My opinion to your question is that the camera would have to decode the tape signal first to an uncompressed siganl and then send it out as an uncompressed 4:2:2 stream even if those 4:2:2 pixels only have 4:2:0 chroma samples the fields will still show proper samples per field. At least this is the way I hope it works. I know with component the chroma usually gets smoothed so you kind of now end up with a blurred 4:2:2 anyways. I'm not sure if the same upsampling works with SDI and HDMI however.

[Derek Green]

Hey everyone thanks for contributing to this humble thread. Some of the points you guys have brought up raised a few questions for me.

Originally Posted by Nick Jushchyshyn

My personal impression (I have no experience with any of these cameras) is that a G1 or H1 with SDI capture would be your best bet. A1 next, due to my instinct (could be wrong) that 3CCDs fed by the A1 lens would perform better than a hand-cam using a single CMOS.

Thanks Nick for your opinion. This is exactly what I was looking for. And I can't say I don't agree. The G1 or H1 both produce great images. Too bad they're both overpriced IMO. $3400 difference right now from the A1--just for a jackpack is crazy. But they know if you need that option, you'll pay.

A soft, over sharpened image through HDMI is generally not going to be as useful for keying as a sharp clean image compressed to 4:2:0.

Again, thanks.

Also, if you control your subject and screen selection to use either a dark subject (Dark hair, dark clothes, dark skin) in front of a green screen ... or use a bright subject (blond hair, fair skin, white clothing) in front of a blue screen .... you can basically avoid color compression based artifacts in your composits altogether and the 4:2:2 vs 4:2:0 concern becomes a non-issue.

I haven't come across this tip before. I understand what you're getting at from a generalist point of view, but I'm lost on why this would work beneficially from a technical standpoint.

Originally Posted by Peter Ferling

I can understand the difference in cost between a G1 (or used H1) and a Sony HC3. Including a system capable of working with 4:2:2 10bit. But after all that work of setting up your scene and the human effort involved, why make the most important tool in your pipeline the weakest link? Can't you just rent what you need?

Thanks for bringing this up Peter. It's something I have thought about. Cost vs Performance vs Effort so to speak. Is the camera really the most important link though? Especially in lower budget projects. I know we're all video guys and want the best image so that is our mentality (at least mine) but when it comes to creating an artistic project there are soooooo many factors. For example, would Lord of the Rings have been as cool if the hobbits had been wearing plastic Halloween costumes and using wooden swords? Probably a good thing they spent some money in that area too. There's so many potential things that can take the audience out of the moment including, like you point out, capture quality. It really is a balancing act. I know I would want my characters dressed as authentic as possible. Yes, renting is always a possibility. I always keep all my options open until I feel I've gathered all the relevant data.

Originally Posted by Mike Schell

I can absolutely assure you that the "live" output from the Sony camcorders has never seen any compression whatsoever. The uncompressed video is 1920x1080i 4:2:2, while the audio is 48Khz stereo, 16-bit uncompressed. Additionally, it is possible, using our converter box to get 720p or 480i out of the Sony camcorder (or M25U deck) in either live mode or while playing back from a tape.

Thanks for chiming in with specs Mike. I love this stuff. Ok, in regards to what you said about the 1920x1080i out through HDMI, can you clarify if you are talking about the V1U or all Sony HDMI enabled cams? Because it's early stages for HDMI capture here, I've been reading a lot of misinformation from blogs/filmaker/videosites across the net and maybe we can cement this right here, right now. Now, from what I've gathered, you could only get 1920x1080 HDMI output from the new HDR-SR1 and HDR-UX1 because their imaging processing chips are newly designed for AVHCD and they are different from the HDV image processing chips from the V1U and HC3. The reason for this new design is because the AVHCD standard supports up to full 1920x1080 where as the HDV standard only supports up to 1440x1080. Even though all 4 cameras have the same kind of CMOS that captures 1920x1080 at the start both the V1U and HC3 have image processors that, by design, limit HDMI out to 1440x1080 because the resolution is being converted before it hits the HDMI. Hence the reason you see Sony's flashy new black & gold FULL HD1080 banner plastered all over the SR1/UX1 but not on the HC3/V1U.

The HDMI ouput will be exactly the same image as the 8-bit over SDI. The The same data stream would be used to drive either type of output as well as the analog component encoder.

Now I could be reading this wrong but I gather you're basically saying the HDMI image output should be the same as the SDI and component because they're being driven by the same data stream?

The diagram at the sony V1U site http://bssc.sel.sony.com/Broadcastan...U/devices.html
seems to contradict this as it shows the component out AFTER the image has been converted to 1440x1080. Wouldn't this contradict what you're saying about component data stream being the same as HDMI? Unfortunately this diagram doesn't label where HDMI out is in the chain.

The bottom line HDMI question here seems to be, as I know this is still a confusing subject for a lot of people on the net, what do the following resolution size do the following cameras output through HDMI?
Sony V1U
Sony HC3
Sony SR1
Sony UX1

Originally Posted by Thomas Smet

If you will be able to capture the live output from the camera then I'm not yet sure which camera will be better.

Yep, hard to say for keying anyways. My guess it'll come down to personal taste.

Originally Posted by Harm Millaard

I posed this question to Sony, asking whether the signal out of the V1 through HDMI was a compressed 25 Mbps signal or a 1.485 Gbps signal and they will investigate further, but the initial reaction was, since it is a TV compatible signal, that it is only 25 Mbps and thus compressed.

Harm, pardon my ignorance...I do all my viewing through a monitor and don't own a TV and have never heard of a TV being limited by data rate? Is this true, TVs can only handle certain data rates? I assume you're talking HD TVs only, is there an industry standard data rate?

If you are correct (I'm still waiting for confirmation by Sony) then the marketing reasons of Sony seem like BMW introducing a new 2.0 liter 4 cylinder car, which underneath has a 4.2 liter V8 but does not want the public to know. Does not make a lot of sense, does it?

Alot of the deciscions a company like Sony does doesn't make sense to the average joe like you and me. Did you see their PS3 commercials? LOL. Marketing people are not consumers, they don't get it. Companies manufacture what THEY want, and their maketing tells US why we should want it. Sometimes we do, sometimes we don't, and most of the time WE end up wishing for something that doesn't exist. Seriously, it's a mentality block most companies have. They won't promote anything they don't have total control over. They want to control then entire package inhouse (think Beta/Bluray/UMD) and make money on the entire process chain. To them it's dollars lost to some other company selling the HDMI capture cards, not money gained on selling more video cameras. Sony doesn't have a consumer HDMI capture product so it's not a focus for them that's all. Of course what they don't understand from a street level perspective is they would sell a ton more cameras if they pushed it.

Originally Posted by Thanasis Grigoropoulos

1) If you want to capture live using the HDMI but without using RAID, which is the best way to go, for bluescreen work?

Shouldn't the SheerVideo Codec allow you to get away with maybe an inexpensive 2 disk RAID?

[Nick Jushchyshyn]

Originally Posted by Derek Green

re:dark subject greenscreen/bright subject bluescreen negates 4:2:2: vs 4:2:0.
I haven't come across this tip before. I understand what you're getting at from a generalist point of view, but I'm lost on why this would work beneficially from a technical standpoint.

The answer is in the 4. Luma is sampled at full resolution with both compression schemes. Most of the luma signal is translated into the Green RGB channel, resulting in sharp detail in this channel vs blockier data in Red & Blue due to color compression. So, if you select your background and subject colors to maximize luma contrast, a sharp image is passed through to the green channel.

The Green channel should be virtually white for a green screen, so using a dark subject provides lots of high-detail contrast to work with, resulting in a better starting point even for color difference and vector based keyers like Keylight and Primatte.

If you put a bright subject in front of a green screen, though, there's little-to-no contrast in the green channel to work with. Your color keying operations end up being based on the blocky data in Red & Blue. This is why it's helpful to use a bluescreen for bright subjects. The green channel for a blue screen is almost black, so the bright subject provides lots of high-detail contrast in the green channel, again helping your keyer independant of color compression.

You'll still be battling artifacts from the lens, sensor, A->D converter noise, and post processing like in-camera sharpening (which is why my instinct favors the idea of using the A1 vs a consumer targeted hand-cam) but by setting up your scene this way, color compression itself becomes less of a factor in calculating your compositing keys.

[Mike Schell]

Originally Posted by Derek Green

Hey everyone thanks for contributing to this humble thread. Some of the points you guys have brought up raised a few questions for me.

Thanks for chiming in with specs Mike. I love this stuff. Ok, in regards to what you said about the 1920x1080i out through HDMI, can you clarify if you are talking about the V1U or all Sony HDMI enabled cams? Because it's early stages for HDMI capture here, I've been reading a lot of misinformation from blogs/filmaker/videosites across the net and maybe we can cement this right here, right now. Now, from what I've gathered, you could only get 1920x1080 HDMI output from the new HDR-SR1 and HDR-UX1 because their imaging processing chips are newly designed for AVHCD and they are different from the HDV image processing chips from the V1U and HC3. The reason for this new design is because the AVHCD standard supports up to full 1920x1080 where as the HDV standard only supports up to 1440x1080. Even though all 4 cameras have the same kind of CMOS that captures 1920x1080 at the start both the V1U and HC3 have image processors that, by design, limit HDMI out to 1440x1080 because the resolution is being converted before it hits the HDMI. Hence the reason you see Sony's flashy new black & gold FULL HD1080 banner plastered all over the SR1/UX1 but not on the HC3/V1U.


Now I could be reading this wrong but I gather you're basically saying the HDMI image output should be the same as the SDI and component because they're being driven by the same data stream?

The diagram at the sony V1U site http://bssc.sel.sony.com/Broadcastan...U/devices.html
seems to contradict this as it shows the component out AFTER the image has been converted to 1440x1080. Wouldn't this contradict what you're saying about component data stream being the same as HDMI? Unfortunately this diagram doesn't label where HDMI out is in the chain.

The bottom line HDMI question here seems to be, as I know this is still a confusing subject for a lot of people on the net, what do the following resolution size do the following cameras output through HDMI?
Sony V1U
Sony HC3
Sony SR1
Sony UX1

The native HDMI output from all these cameras is 1920x1080i YCbCr 4:2:2 with 8-bits of resolution (same as HD-SDI). It is possible to force the output to RGB 4:4:4 but this does not any additonal color resolution.

The compressed HDV/AVCHD video is 1440 x 1080, but is automatically resized to 1920x1080 before being output to the HDMI port or to the analog component encoder. HDMI does not support a 1440x1080 video format.

Using our HD-Connect MI box we can get either 1080i, 720p or 480i/576i out of the HC3 camcorder. While we have not tested these modes on the other camcorders, we are fairly certain they all perform in a similar manner. Note that the HC3 does not list 720p as an output option, but we have found a technique to coax this output from the camcorder.

Ultimately, I think the analog connections will disappear completely and we'll move to serial digital formats like HDMI and HD/SD-SDI. They include both the video and audio and offer no degredation of the signal (no A/D and D/A conversions). They also can be less expensive to implement. Connections are also simplified (only one cable).

I am working on an update to our HDMI white paper with a compatibility chart for playback of Sony, Canon and JVC tapes in the variious camcorder and decks. Bottom line, HDMI offers far more playback options compared to 1394. The Sony M25 deck, for example, can playback JVC 720p30 tapes through HDMI, but not through 1394.

[Ken Hodson]

Originally Posted by Derek Green

Shouldn't the SheerVideo Codec allow you to get away with maybe an inexpensive 2 disk RAID?

I have never heard of anyone doing live capture using SheerVideo codec. It would still need a fair amount of drive speed and space. ProspectHD is a proven codec for this application. It can do 10-bit although only 8-bit is needed. Its downside is high price and a fast PC for capture. I think it can capture to a single 7200rpm drive, but who wouldn't want to use a cheap big RAID 0, for overall speed.

[Derek Green]

Originally Posted by Nick Jushchyshyn

The answer is in the 4. Luma is sampled at full resolution with both compression schemes. Most of the luma signal is translated into the Green RGB channel, resulting in sharp detail in this channel vs blockier data in Red & Blue due to color compression. So, if you select your background and subject colors to maximize luma contrast, a sharp image is passed through to the green channel.

The Green channel should be virtually white for a green screen, so using a dark subject provides lots of high-detail contrast to work with, resulting in a better starting point even for color difference and vector based keyers like Keylight and Primatte.

If you put a bright subject in front of a green screen, though, there's little-to-no contrast in the green channel to work with. Your color keying operations end up being based on the blocky data in Red & Blue. This is why it's helpful to use a bluescreen for bright subjects. The green channel for a blue screen is almost black, so the bright subject provides lots of high-detail contrast in the green channel, again helping your keyer independant of color compression.

You'll still be battling artifacts from the lens, sensor, A->D converter noise, and post processing like in-camera sharpening (which is why my instinct favors the idea of using the A1 vs a consumer targeted hand-cam) but by setting up your scene this way, color compression itself becomes less of a factor in calculating your compositing keys.

This is a great explanation Nick. Appreciate it.