View Full Version : Blackmagic Intensity Pro capture 4:2:2 (1080x1920) over component?


Pages : 1 [2]

Jim Andrada
September 29th, 2007, 07:01 PM
Steve,

It would sure be nice if they had given some idea of the improvement due to hardware upgrades vs the driver issue.

Also more than a little troubling that they never fixed the bug in the driver. Makes me wonder how good they'll be at fixing the (inevitable) bugs in the drivers for the newer cards.

However, since I'm still in the "should I get it" phase, it sounds like more good than bad for me. Sorry about the studio card owners, though.

Oh well, welcome to technology obsolescence, where you can't fix product problems before the product is replaced by a new one.

And sincere thanks for posting this. Wonder why they never gave the same response to me on the same issue. You must have gotten to them on a good day!

Steven Thomas
September 29th, 2007, 07:24 PM
Wonder why they never gave the same response to me on the same issue.

I hear you. I saw your message and wondered the same thing. It makes me wonder which response was correct.

I'm trying to decide if I want to give the Intensity Pro a try.
It'a benefits may be great in the studio, but a pain to lug around a large frame computer.

Also, I would not be surprised if a "reasonably priced" future component/HDMI to FLASH drive capture is available in the near future.

I'm looking forward to shooting at least 4:2:2. I'm getting tired of blocky REDS using 4:2:0.

Jim Andrada
September 29th, 2007, 09:14 PM
I think I saw that Cineform (and maybe others) takes 4,2,0 to 4,2,2. Also It's my understanding that playback from tape is up-converted to 4,2,2 in camera (over SDI?)

Anyhow, just wondered if you had tried something like Cineform and whether it made an improvement over 4,2,0. I have to admit that I'm a bit puzzled over how you can get 4,2,2 out of 4,2,0 material. Probably just my ignorance!

I guess I'll soon get a chance to find out - I just ordered an HD110 so I'll soon be "enjoying" all the "benefits" of HDV. Like worying about codecs and best capture methodology and workflow, etc.

But I sure think I'll like the images.

By the way, I was just thinking it would be interesting to do a capture over firewire and one over component via a card, then do the same captures of a playback from tape and compare the results to see if capturing the video stream directly really does look better than capturing a playback.

After I get the camera I'll run the experiments over firewire, and if I decide to get the card I'll re-run the same experiments.

Glenn Chan
September 29th, 2007, 09:45 PM
There are different ways of upconverting 4:2:0 to 4:2:2.

MPEG-2 calls for interstitial chroma vertically. Cineform decodes it incorrectly, which causes the chroma center to get shifted 0.5 pixels; and it arguably looks better (than box reconstruction) since it sort of blurs the chroma up/down.

I believe the majority of NLEs use box reconstruction... on upsampling it's the same thing as nearest neighbour resizing in Photoshop.

Steven Thomas
September 29th, 2007, 10:13 PM
Yes, I have Cineform NEO HDV. NEO HDV converts MPEG's 4:2:0 chroma to the 4:2:2 chroma space. Although you can not get what's not there in the first place, the compressed CineForm Intermediate and its 4:2:2 color space allows for higher quality when working in post, especially with multiple renderings.

Capturing from firewire via tape verses capturing directly from the camera in realtime will yield the exact same results. Firewire out of the HD110 will always be HDV 4:2:0. For that matter this stands true with all HDV cameras.

Component out of the HD100 is 10bit 4:2:2.
Ideally, when shooting with the HD110 component out into the Intensity Pro should work great. Of course you're tethered to your computer during your shoot.

So I'm not only wondering on the Intensity Pro's component quality, but also if the component in on the Intensity Pro offers 10 bit?

Glenn Chan
September 30th, 2007, 12:23 AM
Capturing from firewire via tape verses capturing directly from the camera in realtime will yield the exact same results.

???

That may not be the case since the signal will likely go down a different image processing path.

If the image has already gone down to tape and you're playing it back, then you might see very subtle differences. There are different ways of converting 4:2:0-->4:2:2.

If you are comparing against live capture via HD-SDI out (or analog component), then I'd expect a much bigger difference.

Jim Andrada
September 30th, 2007, 12:49 AM
Steve

I may be wrong on this as on so many things, but anlog is analog - the bits come later.

Or in the case of coming from the camera, the bits come before and after.

I know on the Blackmagic web page they talk about 10 bit a lot with respect to the Decklink but I didn't see anything re the Intensity.

Bill Ravens
September 30th, 2007, 08:14 AM
I've tried to find info re: the A-D sampling depth for the Intensity Pro Card. Nowhere, on any of the BMD sites, do they say what it is. I've written to BMD tech support for the answer and the responded with a very weasly response, basically providing no info. This makes me suspect that the Intensity (Pro) card is less than 10 bit.

Steven Thomas
September 30th, 2007, 08:59 AM
???

That may not be the case since the signal will likely go down a different image processing path.

If the image has already gone down to tape and you're playing it back, then you might see very subtle differences. There are different ways of converting 4:2:0-->4:2:2.

If you are comparing against live capture via HD-SDI out (or analog component), then I'd expect a much bigger difference.

Glenn,
I was not aware the HD100 gives you anything but HDV via firewire.
I thought the same data being saved internally to the HD100's tape is the same data available during live capture via firewire.

Steven Thomas
September 30th, 2007, 09:07 AM
Jim,
You are correct, the camera's CCD is an anlog device. The A/D Converter turns each pixel's value into a digital value by measuring the amount of charge at each photosite and converting that measurement to binary form.

I also looked for the bit depth for component (and HDMI) on BM's web site with no luck.

Rolf Seitz
September 30th, 2007, 02:43 PM
I've tried to find info re: the A-D sampling depth for the Intensity Pro Card. Nowhere, on any of the BMD sites, do they say what it is. I've written to BMD tech support for the answer and the responded with a very weasly response, basically providing no info. This makes me suspect that the Intensity (Pro) card is less than 10 bit.

Sampling is 8Bit for the Intensity/Intensity Pro, read that somewhere here in a other thread.

Bill Ravens
September 30th, 2007, 04:05 PM
If, in fact, it is 8-bit, then this is the weak link in the chain. The whole object is to bring in 10-bit, isn't it? It does no good to injest 10 bit from the HD100 thru an 8 bit throttle.

Steven Thomas
September 30th, 2007, 04:08 PM
Hmmm.
I have not seen anything on it's bit depth.

It would be nice to know where this info is located?

The only info I read is from there web site about playback:
" Play out through the HDMI video output using 8 and 10 bit uncompressed AVI and QuickTime™ files, HDV, Motion JPEG and DV files, as well as 32 bit TARGA and BMP image sequences, which can all be played back on the same Premiere Pro timeline without rendering."

Even this info does not tell us if it's playing back all 10 bits from the 10 bit video file.....

Update:
Looking closer at the Intensity Pro manual online, I found this info on page 20:
http://www.blackmagic-design.com/downloads/manuals/IntensityProMan-en.pdf

"When using Disk Speed Test, you need to account for disk seeking, so it’s best to add a healthy
margin to the results. If a disk array tests at 32 frames per second HD 1080, it doesn’t mean you can
do 29.97 HD capture and playback, as the margin is too tight. When checking the results, look at the
8 bit YUV 4:2:2 video sections, as Intensity Pro uses the 8 bit uncompressed video format."

Bad news, it looks like the Intensity Pro only support 8 bit.
I wonder if this partially explains the color shift? Glenn?

Daniel Browning
September 30th, 2007, 04:26 PM
I have not seen anything on it's bit depth.

The output is 8-bit, there is no question about that: the 1X bus can't handle 10-bit 4:2:2. There has been a little speculation about 10-bit sampling to 8-bit out, but no one has clarified that for sure.


The whole object is to bring in 10-bit, isn't it?


There are many more benefits: live capture and less compression (4:2:2 YUV beats HDV any day of the week and twice on Sundays).

Glenn Chan
September 30th, 2007, 04:56 PM
The color shift looks like improper handling of 601 versus 709 luma coefficients, and some issue with analog calibration (the calibration of the analog signal coming off the camera and into the card).

Bill Ravens
September 30th, 2007, 06:10 PM
There are many more benefits: live capture and less compression (4:2:2 YUV beats HDV any day of the week and twice on Sundays).

Daniel...
I'm afraid I'll take issue with this statement. While the benefits may be realizable if one is chroma keying, otherwise I challenge you to see the difference.

Daniel Browning
September 30th, 2007, 06:22 PM
Daniel...
I'm afraid I'll take issue with this statement. While the benefits may be realizable if one is chroma keying, otherwise I challenge you to see the difference.

Try comparing highly saturated colors; there was a recent thread about it. Furthermore, a host of color-related filters benefit greatly from 4:2:2. I'm sure you agree that 4:2:0 must be upsampled before *any* grading is done to it, but I would take that further and say that native 4:2:2 is visibly superior.

Robert Ducon
October 1st, 2007, 03:04 AM
Daniel, I'm working on a production that I shot in ProRes 422 HQ at full raster 1080P.. and yes, it's beating all the HDV b-roll I shot. It's sharper, and cleaner when being played back and being viewed on a still frame on my computer LCDs.

HOWEVER.

I'd bet it'd be hard for me to even tell my own footage apart being played on even a high-end LCD TV.

It was worth capturing to RAID at 4:2:2 though, especially since making a claim like this will mean I'll have eventually prove it, right? ;)

Btw, I captured the digital HDMI signal at 1080P 4:2:2, rather than an analog component signal (which I'd say had less resolution than HDV). When talking about analog YUV vs. digital HDV.. tough call. The jury's out on that one - I tested analog and decided it wasn't for me and when full digital.