Data Rate for UNCOMPRESSED video

[redwave]

hi everyone,
I was looking into recording uncompressed via HDMI, bellow is chart for anyone interested in UNCOMPRESSED video (for special effects etc).

The storage and data rates for uncompressed video are listed below.

525 NTSC uncompressed;
8 bit @ 720 x 486 @ 29.97fps = 20 MB per/sec, or 70 GB per/hr.
10 bit @ 720 x 486 @ 29.97fps = 27 MB per/sec, or 94 GB per/hr.

625 PAL uncompressed;
8 bit @ 720 x 576 @ 25fps = 20 MB per/sec, or 70 GB per/hr.
10 bit @ 720 x 576 @ 25fps = 26 MB per/sec, or 93 GB per/hr.

720p HDTV uncompressed;
8 bit @ 1280 x 720 @ 59.94field = 105 MB per/sec, or 370 GB per/hr.
10 bit @ 1280 x 720 @ 59.94field = 140 MB per/sec, or 494 GB per/hr.

1080i and 1080p HDTV uncompressed;
8 bit @ 1920 x 1080 @ 24fps = 95 MB per/sec, or 334 GB per/hr.
10 bit @ 1920 x 1080 @ 24fps = 127 MB per/sec, or 445 GB per/hr.

8 bit @ 1920 x 1080 @ 25fps = 99 MB per/sec, or 348 GB per/hr.
10 bit @ 1920 x 1080 @ 25fps = 132 MB per/sec, or 463 GB per/hr.

8 bit @ 1920 x 1080 @ 29.97fps = 119 MB per/sec, or 417 GB per/hr.
10 bit @ 1920 x 1080 @ 29.97fps = 158 MB per/sec, or 556 GB per/hr.

1080i and 1080p HDTV RGB (4:4:4) uncompressed;
10 bit @ 1280 x 720p @ 60fps = 211 per/sec, or 742 GB per/hr.
10 bit @ 1920 x 1080 @ 24PsF = 190 per/sec, or 667 GB per/hr.
10 bit @ 1920 x 1080 @ 50i = 198 per/sec, or 695 GB per/hr.
10 bit @ 1920 x 1080 @ 60i = 237 per/sec, or 834 GB per/hr.


ps. Anyone tried MICROCOSM codec (http://www.digitalanarchy.com/micro/micro_main.html) for vfx?

cheers

[jmorton]

Thanks. :hv20-smilie03:

[BitrateBabe]

Thanks Redwave - are these bitrates for 4:2:2 or 4:2:0? I'm assuming that it's not for 4:4:4.

[elantric]

Microcosm seems to be going free
http://anarchyjim.digitalanarchy.com/
http://mograph.net/board/index.php?showtopic=16666&mode=threaded

http://www.redgiantsoftware.com/featured-news/da-faq/

[kmlo1984]

Hi, just wonder anyone succesfully use other codecs?
so far i can only use Blackmagic own capture software and cineform HDlink
the virtualdub works occasionally and then suddently becomes black.

and only cineform HDlink allows u to capture in cineform format, however it's not for MOS ( no sound capture).

BTW i also see something funny too while in virtualdub capture.
the uncompress capture is very low in cpu usage , as low as 5-10% on my duo core 2675. so be sure u buy a fast HD or build a raid 0.

in MJpeg, some dark scene can make cpu works at 100%.

it;s too bad i can't tell from cineform HDlink about cpu usage and most important - is there any drop fram??

anyone has a solution??

thanks
K.M.Lo

[lordtangent]

I've only been successful capturing to Blackmagic codecs with MediaExpress and Cineform with Cineforms tool "HDLink" . I had the same issue with HDLink, it seemd to suck up 100% CPU all the time. So I wasn't sure if I was getting dropped frames or not. Even though when I manually checked my test captures I didn't notice any dropped frames, I'm still not sure I would trust Cineform for live capture until the tool was a little more advanced. I'd at least want to have a little statistics window or something that would report dropped frames.

My solution to date has been to simple capture to the uncompressed BlackMagic codec and run my custom processing scripts to remove the 3:2 pulldown and re-transcode to another lossless format like HUFFYUV or LAgarith. (I usually use Lagarith these days). My custom scripts keep everything in 4:2:2 YCbCr all the way though and I get about 1/3rd realtime on the conversion. REally there is no loss at all other than losing the 3:2 pulldown and packing the data to make it a little more compact.

If you were in the mood to spend money, you could also use Cineform of Sheer Video as the target codec in a similar scheme. Cineform is lossy and would give you better compression and Sheer is lossless but reatime, unlike Lagarith, which at HD res is pretty pokey, even on a pretty hot new machine.

[tatsuo]

Hi lordtangent,

Hope you get this message, I was just wondering on your experience capturing uncompressed from the hv20 and if you think it is really worth it vs hdv. I am looking into getting into this kind of setup soon and mainly want to take the plunge for the better colour space for keying, colour correction etc, also wanting to get out of hdv because it can sometimes look a little funky on movement. Just wanting to know your experience of uncompressed vs hdv on things like greenscreen and grading, can you push the image farther etc, also is there any resolution increases etc. Basically I am looking to get a denser base image if that makes sense. Thanks for your time

[kmlo1984]

Hi
HDV is good as u r happy with 1440x1080 and 4:2:0, i personally think nothing wrong with it. just a format good enough for certain level.
also there's a mepg2 compression artifact issue.

if u r doing chroma key and this is another issue.
1920x1080P 8bit 4:2:2 - this is much nicer by technical point, u might not notice the difference but when u key u will know HDV is not ideal.

but as of now - theres only1 way to use BM intensity card on laptop. so it's not ideal for location shooting with uncompress /cineform capture.

i think many still not getting the point - how to force intensity card to let u use the codec u prefer instead of uncompress/mjpeg or in HDlink - cineform

in mac osx - the quicktime capture can let u use mp4 or h264
but thats all, i cant find any other means to choose codec

anyone try FCP??

KM LO

[funnychicken11]

sorry if i seem naive, but 4:2:0 seems fine at a resolution of 1440 X 1080, because the only thing that im going to be doing with the footage, and--correct me if im wrong--you guys will be doing with the footage is compressing to DVD, where it won't matter.:hv20-smilie50:

[kmlo1984]

i think in most cases, HDV is good enough as really u dont see much different if u just aim and shoot and edit. and if u r going to downres to DVD or SD.

uncompress capture at 1920x1080 is only good when u r doing chroma key or heavy post that requires u to have highest color resampling and no GOP /artifact issue.

BBC is not selecting HDV nor even EX1 for their HD programs ( www.reel-show.tv) so there's still issues to be sorted out.

to me is rather the easy way with minimal resources or hard way with a pc/mac and raid 0 to capture live from camera in uncompress.

i dont think people will see the differences in ur daughter birthday party footage on youtube/ipod to even DVD, from HD to DV it might look the same.
and if u work smartly to output all HDV to another codec/format ( cineform/uncompressed) and carefully upress it. many professional might not even know that is originally from HDV.

after all it';s about content. if the format itself is pervasive enough to deliver the content is OK.

KM Lo

[lordtangent]

Hi lordtangent,

Hope you get this message, I was just wondering on your experience capturing uncompressed from the hv20 and if you think it is really worth it vs hdv. I am looking into getting into this kind of setup soon and mainly want to take the plunge for the better colour space for keying, colour correction etc, also wanting to get out of hdv because it can sometimes look a little funky on movement. Just wanting to know your experience of uncompressed vs hdv on things like greenscreen and grading, can you push the image farther etc, also is there any resolution increases etc. Basically I am looking to get a denser base image if that makes sense. Thanks for your time

Wither or not it's worth it really boils down to your priorities, budget and schedule.

I think it is worth it. My personal choice is to go with Intensity capture when it's convenient because there really is a visible quality boost over HDV. (Skipping the MPEG compression and INTERLACED 4:2:0 chroma subsampling has pretty much the exact impact on the quality as you would expect) But... when I absolutely must shoot HDV due to the constraints imposed by the need to use the Intensity card, a full size machine with a RAID-0 and an HDMI cable, I don't lose sleep over it. HDV is still pretty freaking good. It's almost magically good. It seems to defy reason how good it is most of the time.

I did some early tests of HDV vs direct capture and I was impressed by the improvement HDMI capture gave. But I admit my tests were not super comprehensive or scientific in any way. I'd like to do a little more testing but lately my priorities have been more focused on just finishing all the projects I've started and getting more writing done.

I still have not done extensive green screen from Intensity captured material, though I do have a bunch that I shot that's still waiting to be done. My preliminary tests are very encouraging though. Really, it's the best material I've ever keyed from. I have HDV for the same material, but I can't imagine why I would bother capturing it from tape since I already have the HDMI captured stuff. To date that has been the biggest impediment I've had in regards to doing a really comprehensive comparison. I mean, if you already have the HDMI captured stuff on disk, you don't really feel like going though the added hassle of capturing the HDV. And all my really heaviest shooting sessions are real projects, not tests. So I'm more excited to just finish the project than do tests.

My main advice if you want to do direct capture is to make sure you consider how you plan to handle all the data it generates. One full day of shooting (single cam cinema style) is going to generate 600+ gigs depending on your shooting habits. For a multi-day shoot those levels of data are gonna add up really fast. That's one place where Cineform capture could really make a big difference. If I was going to try to do multi-day shoots with direct capture I would seriously consider Cineform since it can also remove pull-down at capture time. (The separate step I do to remove pulldown WILL run pretty reliably over night. But what if it doesn't? That means the 24p in 60i media will be piling up for every night I don't get it done, taking up a lot of space. Cineform at it's highest quality gives you about 5:1 compression which is head and shoulders above HDV or even HDCAM SR but still provides decent compression. Way better compression than any lossless codec but at the same time with no real impact on image quality.)

Anyway, my point is think about all the details to help you decide if HDMI capture is right for you and how you actually intend to use it. There is no free lunch and the extra quality you get comes at a cost in terms of a little added complexity and a LOT of extra storage requirements. Some people might consider the added requirements disproportionate to the gain in quality. Others (like my self) think the opposite. But there is a definite gain in quality and almost no one disputes that.

[lordtangent]

sorry if i seem naive, but 4:2:0 seems fine at a resolution of 1440 X 1080, because the only thing that im going to be doing with the footage, and--correct me if im wrong--you guys will be doing with the footage is compressing to DVD, where it won't matter.:hv20-smilie50:

If you only plan to go to DVD this is 100% correct.

But what if you want to go to some HD output or even film?

Direct capture give you a boost in that case, making the output from the humble little HV20 look like that of a more expensive camera.

[Rumpelgeist]

[COLOR="Orange"]hi everyone,
I was looking into recording uncompressed via HDMI, bellow is chart for anyone interested in UNCOMPRESSED video (for special effects etc).
Instead of listing different combinations of frame rates, frame sizes and color resolutions you could simply provide just one formula:

data rate = data / time. Isn't is simpler?

Also, your numbers are off. For example, where did this come from:

> 525 NTSC uncompressed: 8 bit @ 720 x 486 @ 29.97fps = 20 MB per/sec, or 70 GB per/hr.
> 1080i and 1080p HDTV uncompressed: 8 bit @ 1920 x 1080 @ 29.97fps = 119 MB per/sec, or 417 GB per/hr.

How come it is only 8 bits? What color depth did you assume? Say, it is 4:1:1 for DV, then color depth will be 12 bits/pixel. If you mean true uncompressed video, it is 24 bits/pixel assuming 8 bits per color (HDMI 1.0 - HDMI 1.2) or up to 48 bits/pixel (HDMI 1.3).

Therefore, the number for 60 Hz SD video will look like this:

8 * 8 * 8 * 720 * 480 * 29.97 ≈ 248 Mb/s ≈ 31 MB/s

For 60 Hz 1080i video it will be:

8 * 8 * 8 * 1920 * 1080 * 29.97 ≈ 1.5 Gb/s ≈ 186 MB/s

> 1080i and 1080p HDTV RGB (4:4:4) uncompressed
> 10 bit @ 1280 x 720p @ 60fps = 211 per/sec, or 742 GB per/hr.

You just wrote that it is uncompressed. You even specified color depth 4:4:4. How come it is only 10 bits/pixel? It should be at least 24 bits. What is "per"? By the way, it is 720p, not 1080p. Here is what it should look like:

24 bits * (1280 * 720) pixels * 60 frames/s ≈ 1.3 Gb/s ≈ 166 MB/s

After calculating using your formula, I got 553 Mb/s. Now I found what per is, it equals to 2.6 Mb. In future I would prefere you used bits or bytes instead of pers.

Now when we are done with second grade math, what was the point of listing these rates?

P.S. I went back and recalculated your equations, even with your assumption of 8 bit/pixel color depth the results do not match the formula, for example:

525 NTSC uncompressed: 8 bit @ 720 x 486 @ 29.97fps = 20 MB per/sec, or 70 GB per/hr.

If you calculate using this formula you will get about 10.4 MB/s, not 20 MB/s.

[lordtangent]

You just wrote that it is uncompressed. You even specified color depth 4:4:4. How come it is only 10 bits/pixel? It should be at least 24 bits.

This is another area where colloquial phraseology often gets us in trouble. People tend to use labels like "24 bit" and "8 bit" interchangeably. When they mean more precisely: "24 bits per pixel" and "8 bits per component". Of course, that still doesn't mean anything without more info. Are they talking about RGB? YCC? 4:2:2? 4:4:4? 4:2:0?, Heck, they could even be talking about XYZ color space, which almost nobody seems to actually understand.

This makes your second grade math into a 6th grade "word problem". And a lot of people are not well enough versed in the particulars of graphics/video formats to do the math from scratch since they don't even know how many bits per pixel most formats actually are. That's why companies print those tables of data rates for particular formats. (The table at the top of the thread looks like it was copied and pasted right from the Black Magic website)

[Rumpelgeist]

Found the IntensityManual.pdf (http://www.blackmagic-design.com/downloads/manuals/IntensityManual.pdf) on Black Magic website. It reads a little bit different:

Uncompressed Video
Standard Definition 4:2:2
720x486/29.97fps (NTSC) ==> 21.5 MB/s or 75.5 GB/hr

The formula does not mention 8 bits, because it is about broadcast SD video with 4:2:2 color subsampling as required by Rec.601, which effectively uses 16 bits/pixel, not 8 bits. Now the formula works:

16 bits * 720 x 486 * 29.97 fps ≈ 168 Mb/s ≈ 21 MB/s

Wait, now I see what redwave meant by "8 bit @ ...". This is "8-bit YUV 4:2:2 @ ...", which translates to 16 bits/pixel. Bingo! I went overboard as usual, and as usual I look like an idiot now. Redwave, please excuse me for being so harsh :)

While I understand now what Black Magic means by uncompressed video, I think this is small lie calling 4:2:2 uncompressed. This is akin saying that DV has 5:1 compression rate. Right, after dropping 3/4 of color information!

For me as a computer person uncompressed means 8 bits per component without any further lossy data compression. I understand that if SD video is recorded and transmitted as 4:2:2, there is no way Black Magic can injest it as 4:4:4. But even HD video is injested as 4:2:2 which is not exactly uncompressed to me. This does not seem to be HDMI limitation, as HDMI seems to be able to transfer 4:4:4 video. Is this Black Magic limitation? Or NLE limitation? Or camcorders' limitation?

Another interesting question: what looks better, 10-bit YUV 4:2:2 or 8-bit YUV 4:4:4 ?

[mattias]

10 bits doesn't look better at all, it's more than most displays can show and usually more than our eyes can see, but it's much easier to grade, key, scale, whatever. and unless you're keying i'd be extremely surprised if you saw a difference between 4:2:2 and 4:4:4. as for hdv, the artifacts from the dct compression are much worse than from the color subsampling. uncompressed 4:2:0 looks a lot better than heavily compressed 4:4:4. color subsampling may sound like a bad idea but just do some tests and you'll see that it's not bad while extremely useful. you cut bandwidth requirements by one third at almost no cost in image quality.

/matt

[Halsu]

you cut bandwidth requirements by one third at almost no cost in image quality.

...Unless you have brightly colored objects in the scene. Say, red flowers on a black background...

But yep, in most situations subsampling is OK.

[mattias]

Absolutely, but even then in most cases. Unless your flowers are computer generated the saturation isn't usually that high and the luma will hold the edge fine. On the computer screen it often looks worse than it is because most codecs use nearest neighbor scaling. /m

[jjanes]

Direct capture give you a boost in that case, making the output from the humble little HV20 look like that of a more expensive camera.

Let us see. Post two short clips of the same material: one shot HDMI uncompressed and one shot HDV 24p.

Thanks.

JJ :hv20-smilie03:

[mattias]

i agree. while i do recongize the technical difference i think the notion that the image will look better is verging on the insane. post some clips by all means. not lab tests but actual footage.

here's some mandatory reading btw. it doesn't talk about 4:2:0 but the principle is the same:

http://www.kenstone.net/fcp_homepage/chroma_investigation_nattress.html

/matt

[lordtangent]

Then consider me virtually insane. (That's actually a much better opinion than most of people who actually know me have.)

mattias, I know you are a smart guy and I totally hear where you are coming from. I am also well aware of the technical differences between the various sub-sampling schemes. For me the biggest reason for wanting to work from Intensity captured stuff vs. HDV is the avoid the interlaced 4:2:0 and mpeg compression. Interlaced 4:2:0 is the devils own sub-sampling scheme. The article you link doesn't mention INTERLACED 4:2:0. You should check it out. Truly evil.

http://en.wikipedia.org/wiki/4:2:0#4:2:0


Though... I agree one would be hard pressed to see the difference with the naked eye on most "natural" scenes. But we are not always talking about human vision or "natural" scenes. What about pulling mattes? Re-Compression? Tracking? Pushing a grade hard? Removing 3:2 pulldown without spreading chroma across 8 scan lines in some crazy way? (Side effect of interlaced 4:2:0) I realize a made a bold statement. But the above are some of the reasons I made that comment.

jjanes, I'll show you mine when you show me yours...

Seriously, the issue for me is time. I have more time right now than usual but I'm still focused on things other than doing tests to "prove" stuff to strangers. Especially when my in-house tests have provided me with satisfactory answers for my own use. I've posted some uncompressed clips here in the past if you are interested in searching for them. They are still being hosted by the site.

[mattias]

Then consider me virtually insane. [...] Interlaced 4:2:0 is the devils own sub-sampling scheme. The article you link doesn't mention INTERLACED 4:2:0.
i know, but it still doesn't look that bad. it's pretty much impossible to see a difference and to put my previous statement in perspective i'm talking about people who think a worklight or an lp micro is lighting, that's the insane part. when somebody asks what they need to make a feature there's always somebody answering "an intensity card", which isn't just verging on the insane.


What about pulling mattes? Re-Compression? Tracking? Pushing a grade hard?
definitely. uncompressed is better than hdv, but that's a no-brainer in my book. the question is whether there's actually something wrong with hdv that typically needs to be fixed, and also whether people really understand what color subsampling does to an image. sometimes i get the feeling that some people think the reason the hv doesn't produce as good looking images as the red one is that it uses subsampling. they also think that subsampling removes color and sometimes that it magically changes the latitude of the sensor. i can't think of a better word than insane.

/matt

[lordtangent]

sometimes i get the feeling that some people think the reason the hv doesn't produce as good looking images as the red one is that it uses subsampling. they also think that subsampling removes color and sometimes that it magically changes the latitude of the sensor. i can't think of a better word than insane.

/matt

Well, yes, insane is the right word for that kind of thinking. Irrational at the very best. Some of it might be coming from the whole "A little knowledge is a dangerous thing" phenomena. Or the "more is always better" school of thought.

Yet, it's amazing how even some people who should know better have those attitudes.

I agree about other factors being more important to perceived quality than chroma sub-sampling and bit depth.

IMHO, As far as the perceived photographic quality of an image goes, you get a lot more mileage out of real lighting and exposure control than you will from almost any technical improvement in video quality. The differences in dynamic range between devices is pretty much meaningless until you master those skills anyway. And chroma fidelity usually only really matters in cases like pulling mattes or FX type stuff, where the COMPUTER needs the fidelity to get the best results. (Funny thing, that describes a lot of my projects...)

We are totally talking about the same thing here. I should have qualified my statements to limit them to the specific technical areas I mentioned previously so newbies don't get the wrong idea about what I'm getting at.

So, for the newbs: The number one thing you can do to improve your images is to use good lighting and exposure. The number one thing you can do to improve your movie overall is make sure you get good sound! Stuff like the Intensity should be pretty low on your list. I have/use an Intensity myself but it's because of my technical requirements. I always light things carefully regardless of what I'm shooting be it HDV, direct capture on and Intensity, P2, HDCAM or even film. What makes the image good is the lighting. (And by that I mean "Movie Lighting"... not just having enough light)