Thomas, I hear your points and I understand the differences between Pro HD-SDI and HDMI. I agree, I'll wait till there is some proof either way, or make sure a store I purchase from would have a return Policy and then test myself. It looks like a great cam regardless - home movies always have their place. I already have a Decklink HD card with component and a 4drive RAID in my Mac Pro, so I could test that as well. If every cam does 4:2:2 component out, that's great - then I have high hopes for this one as well. Exciting!

Thanks for your detailed explanation above. Though I haven't stayed up with the all latest bandwidth requirements I was pretty sure uncompressed was still out of the realm of most normal systems/storage. I'm also aware of the various codecs to some extent and their advantages/disadvantages in relation to HDV as I work a lot with the HVX200 and DVCPro HD codecs as well. Much of the talk earlier centered around whether the HV20 outputs 4:4:4 uncompressed or 4:2:2 uncompressed which is reality isn't as critical in my mind as the codec ingested to/edited with. I think it's safe to say that working with uncompressed in either format will be a challenge.

I deliver all my work as Animation codec set at full quality (uncompressed) with 48khz 16 bit stereo sound, I have done it this way for years.

This is then usually dumped to digi-beta (not by me).

To 'proof' (and view) a project I am working on, I render to PhotoJpeg with the quality set to 99 (100 pushes the file into 4:4:4 and it size leaps up) - a 1920*1080p @ 25fps play back perfectly off my 7200/FW800 La Cie.

I agree with pretty much all that is said here, although we might want to make a distinction between live action footage/Editing and FX/Motion graphic work, in the case of the latter uncompressed might often be desirable.

In FX/Motion graphics files are often much shorter and worked on uncompressed and then rendered off only to view - if all looks good I then render a uncompressed version for the client - which will be put onto digi-beta.

I totally agree with you there. I am a visual effects artist and I usually work with uncompressed SD. If you need uncompressed the beauty of HDMI and Intensity is that we can get uncompressed. With that said however I know a lot of sweet keying projects that were done with DV footage and the jpeg codec blows DV way out of the water. I wouldn't be surprised at all to see an indy pull off an amazing project with just using the jpeg codec.

I have done some tests of the jpeg codec over the weekend and it really is amazing. I rendered some stuff from 3D Studio Max to create an uncompressed 1920x1080 segment and converted it to jpeg and compared the two by switching the top layer off and on in After Effects and they were very close. There was only a tiny bit of noise in sharp edges which was pretty hard to see when viewed at a 1:1 ratio. When zoomed in 400% percent you could see it a little bit more but for the file being 1:16th the filesize of the uncompressed version this was amazing. Of course I will have to test this out on some real world footage to see how it holds up.

Just to be sure when you say 'Jpeg' you do mean photojpeg don't you (and not MotionJpeg) ?

The nature of Jpeg compression means it is very reliant on the content you give it, but I would agree it is often visually lossless.

_______________


Instead of switching between two layers in After Effects to reveal any compression artifacting try this:

Stick the original into the timeline and the compressed version on the layer above.
Then apply the 'invert' plug-in to the compressed (upper) layer and set the opacity of this layer to 50%.

At this stage if both layers are identical you should be looking at a perfectly flat 50% grey image - if not any differences introduced by compression to the upper layer's file will show up as small aberrations.

To really get super forensic - pre-comp these two layers and apply a levels plug-in to the comp - then crush the black and white until the aberrations show up clearly.

**I have kept this simple - of course there are many other ways of doing it

I'd be interested in the quality of the Blackmagic JPEG codec. They have the uncompressed codecs available for download but I don't think they have the JPEG codec available, am I wrong?

The Blackmagic JPEG codec captures at 12MB/sec and I tested QT Photo JPEG on an HV10 clip resized to 1920x1080 at 95% quality and it came in around 12MB/sec. Of course at 95% quality the image is outstanding.

Just go to the Blackmagic website and download the drivers for Intensity. It installs the drivers for the card but it should not hurt anything. With the Install you will get all the codecs for the PC including the AVI uncompressed codecs the jpeg AVI codec and the uncompressed Quicktime codecs.

The AVI jpeg codec is a variable bitrate codec which is why it is hard to pinpoint the exact level of compression used for each frame. It changes for how complex each scene is. For example bluescreen footage should compress very small because most of the screen is a solid color with very little detail. The more noise and detail in a shot the larger the file will be. No matter how complex however the video should always fit on a single drive.

I did another test yesterday. This time it was SD but it way live component captured to uncompressed and converted to the jpeg codec. Visually I could not see any difference at all. I mean no artifacts at all. Doing the difference adjustment to compare the two only showed a slight amount of added noise but it wasn't anything any human eye on the planet would ever be able to see. Especially when the footage is playing back at a normal speed. The footage was shot in front of a green screen and it keyed just as well as the uncompressed version. Clearly I need to so more testing but so far this codec seems to hold up very well for compositing. Just to note I shot this greenscreen footage very bad on purpose. It was very low light with a lot of gain and only one light causing shadows and uneven light spread across the screen. With well lit and shot blue/green screen footage I cannot see this codec holding anybody back in getting high quality results.

Thanks for the tip Thomas. I contacted Blackmagic and that's what they told me also, they said datarates can range from 12-14 MB/sec for the JPEG codec. I was also able to see a sample and it looked very good to me.

I've been able to get excellent keys by taking interlaced HDV, magic bulleting and deartifacting it, and finally resampling to 960x540. I have found this resolution affords me an excellent image and very fast processing. So I'm thinking capturing Intensity video with the JPEG codec would be more than enough for my purposes. I think for the users of the HV20 it'd be a good solution. If I could afford an XH-A1 then I could afford a handful of drives to capture uncompressed, but for now that's out of the question.

The next issue would be image quality loss in the pulldown removal process. Would the video have to be re-encoded with another JPEG pass when removing the pull down? For FX shots I would remove the pull-down and export to Tiff sequences but if I wouldn't edit an entire video with uncompressed footage.

The JPEG compression is also done by the CPU and not on the Intensity card according to Blackmagic, which I asked them about.

The data rate is close to 266MB/s. The new version of PCI-E is to be around 533MB/s for an single lane (let's hope they replace USB 2.0 with it (there is an external device version)).

HDMI 4:4:4 is not impossible on cameras, Version 1.3 supports it, but it seems the cameras are an in between version. What we really need is an single chip mode (complimentary and RGB/Bayer etc) that would save three times over 4:4:4 (I wonder if an differential 4:4:4 could be possible). The data would then come in an smaller native format before compression. I knew an guy once, that was trying to specify an HDSDI bayer format.

Wayne I think the card needs that much because it does in and out at the same time. Therefore it would need to be able to push 240 MB/S just for 4:2:2 due to a stream in and out of the card if you want to monitor what you are capturing.

While PCI Express may increase the bandwidth in the future that doesn't help those who already have systems built. The Intensity card is clearly aimed at a lower end market of those who already have decent systems and want to add in a card. If you want 4:4:4 you can move up to the Declink Studio card. I'm not sure if the HDMI will do 4:4:4 but the component will. Even if the HDMI did 4:4:4 the jpeg codec will not. The only codec that could be used right now for compression would be photojpeg at 100%.

I do agree with you that it sure would be nice to have a bayer based video format kind of like how the Cineform version works for the Silicon Imaging camera. By recording the bayer pattern we would get data equal to 4:2:2 but with video that would be easier to convert to 4:4:4 with the correct bayer processing. In order for this to work however a camera would have to be designed to push a bayer pattern out of the DSP which no video camera will do so this is a little pointless right now.

I really like what this Blackmagic Intensity and new HDMI cams bring to the market. I just wish it could be incorporated into a laptop. I am a big fan of HDVrack type monitoring, is there anything like this available for a solution like this?

Thomas, I know that they have separate path for each direction, is it for sure that the bandwidth is split between the two paths? You could use component for monitoring.

HDMI 1.3 does support 4:4:4. 1.2x had extensions, but I am not sure wherever that included 4:4:4.

I read Jpeg supports 4:4:4 above something like 80%, doesn't it. You can use whatever codec you want with it.

I think there will be cameras out there already that could reconfigure their outputs to be grey-scale bayer, or some other pattern. But it is really an industry suggestion, because it would make sense from the production side, and even on the distribution display side. If they could use something like HDMI as the basis for an general I/O interface, it could handle most things and new devices. I used to advocate HDMI as an possible target for USB 3.0 (10.2GB/s versus 480Mb/s, but with more host independent I/O management like Firewire and USB GO (but preferably not, Firewire was better)). But now I realise that the external desktop bus version of PCI-Express (finally coming) is probably meant to be this, but need to be plug and play like firewire and preferably much faster like modern HDMI speeds).

In coincidence, information on the external PCI-E cabling has just come.

http://www.theinquirer.net/default.aspx?article=37528

Looks a bit more than would be cheap for USB3.0, but still a start in the smallest version.

100% supports 4:4:4

Below 99% supports 4:2:2

That's right, below 80% is 4:2:0? What's the compression ratio at 4:4:4 again?

below 75% is 4:2:0. Photojpeg quicktime files are usually captured as 75% because that is the starting point for 4:2:2 color. Anywhere between 75% and 99% is 4:2:2. This only applies to mac users however. The AVI codec cannot be adjusted and it adjusts itself based on the footage.

On another note some of you may have noticed that Cineform just announced support for the Intensity card. This means those of you using Premiere and Cineform now have another codec to use on the PC. Of course the Cineform codec and software costs lots of money but if you have it that is great. Those of you already using Cineform to edit HDV will have files that are the exact same size on your hard drive but much higher quality if captured live over HDMI. The AVI files captured live will be the same size and work exactly the same as video captured as HDV. This means your level of workflow will be exactly the same. The only bad part is that right now the Intensity card will not work for live HDMI output from Premiere with the Cineform files but I guess that should change in the near future. So if HDTV live output is important you may want to stick with the jpeg codec for now until the drivers are updated.

Wes, how are you "deartifacting" the HDV footage? If with Magic Bullet, how?

Do we know if the signal coming out of the hv20 is 10bit or 8bit?

Having problems - and I think it's a simple answer as I don't mean to hijack the thread further than it already has been.

Just wondering what preset to use to transcode my HDV project to a JPEG-based one.

I just exported an HDV clip I had in FCP to a QT movie and chose the preset "Blackmagic HDTV 1080i 59.94 - JPEG" and there was no option for the percentage of compression. There was no 90%, 75% or 100%. And, it looked like crap in the end - macroblocks.

Now, should I have just chosen "Photo - JPEG" or "Motion JPEG B" or "JPEG 2000"?

I want great 4:2:2 quality with little or no perceived digital grain or artifacting (no gradient banding) - and space savings of course when compared to uncompressed. I have a Decklink HD card, so if there is an advantage to using a Blackmagic Codec, I will.

So, I've captured full HDMI from the HV20..

It's 4:2:2, as suspected.

It's 8bit, as suspected.

My workflow is now ProRes 422 which works well with the high-output of the HV20.

Case closed!

Give us a looky look!! If you don't mind.

Svyatoslav Pylypchuk's thread will show the same thing and it's ready to go!

http://dvinfo.net/conf/showthread.php?t=104431

I'll have something ready in a month or so lol. Mr. Pylypchuk's footage is tack sharp!

Robert,

IS that 1440x1080 or 1920x1080?

The signal that comes out of the HDMI port when looking at any content off the HV20 is 1920x1080. However, I don't know if the image is internally processed at 1440x1080 or 1920x1080. Frankly, it looks just great.

When I captured LIVE video via HDMI to the computer, when viewing on playback, the chroma looked very good.

Looking Svyatoslav's footage, I see the same thing with the red's as in my footage - small horizontal lines. While 4:2:2 is bound to have this (as it's not 1:1 luma to chroma mapping), because I don't have footage from any higher-end 4:2:2 camera's to view, I'm not sure if this is 100% normal for 4:2:2 or if this is particular effect is due to the camera internally processing the image at 1440x1080.

Someone else care to prove if the cam is processing it's true 1920x1080 from the chip?

It captures 1440 tall "video" pixels, which when corrected for a computer's square pixels equals, wait for it, exactly 1920.

Looking for proof here..

The HV20's sensor as Canon states, is and uses 1920 pixels by 1080 pixels, square pixels.

Without referring to HDV at all, and rather, the way that the video signal is internally processed, then if anyone can provide proof that the sensor is seeing/using only 1440 rectangular pixels, please do. Likewise with the 1920 square pixel internal processing. Perhaps seeing that the chroma is 'stretched' is proof enough.

Personally I believe this is where their claim for "True" 1920x1080 comes into play. Why would it be otherwise?

In the end it does not matter...just look at the magnificent picture it produces.

In additon to my experience shooting charts, I'm pretty sure that I've read definitively that the signal is 1440 (remember, it's a debayered CMOS reconstruction, so there isn't a 1-to-1 relationship between sensor pixels and RGB signal pixels.)

If I'm confusing cameras, I apologize, but I think this is the case.

Maybe, maybe, but HV20 is very good:
http://www.dvinfo.net/conf/showthread.php?t=104431

I agree, it looks incredible. My point, exactly, actually. I think people get way too hung-up on this 1920 vs 1440 concern. It's important to remember that until very recently NO camera recorded 1920. Its also important to understand the difference between tall video pixels and square computer pixels, because the difference between 1440 video and 1920 computer is even smaller than the raw number suggests.

This thread isn't about the merits of square-to-rectangle pixels. It's about whether anyone knows/can prove either way. If no one is able to prove, it's a non-issue. If someone can prove, then it's proven and then known, and then a non-issue. Just seeking knowledge, trying to avoid debating if it's a debatable issue or not.

If Svyatoslav could shoot a chart, we could see if HDMI uncompressed resolved any more actual horizontal resolution.

My understanding is....

its 1920 - non compressed straight from the HDMI port (pre-tape)
its 1440 - compressed from the firewire port (post tape)
and its back to 1920 - un-compressed from the HDMI port ( post tape )

So the camera will un-compress the 1440 tape footage out of the HDMI
port at 1920...

so if your workflow is 1920, just stay with the HDMI port ( pre or post tape )

if your workflow is 1440, just stay with the firewire port ( pre or post tape )