Some q's about blackmagic capturing.

[Dagnanigans]

I'm still trying to figure out how to maximize the quality of the camera. I have a laptop at home at the moment and doubt there is much room for adding a card in there but more to the point, say I got a desktop, would I have to drag that to every shoot with me in order to capture? How does that work? Is there a storage device that you would feed the HDMI output to? And on an unrelated note, I heard that there was teh potential of someone rewriting the manual control capabilities. Any headway there?

[Terfyn]

Buy more miniDV tapes. Download back at base. The miniDV tapes are your storage device. Shooting video and capture are mutually exclusive events.

[RHKFilm]

You can capture via HDMI either on set (for higher quality) or during ingest back at your editorial suite. The former is a hassle, but sometimes worth it. The latter (capturing at your edit suite) does not make much sense unless you have an unused raid array sitting around. The only reason you would need to capture via your HDMI port is to avoid editing in HDV. Editing with HDV works just fine on current computer systems and it is therefore not a requirement to capture to a less compressed (or uncompressed) format.

Contrary to popular belief, the visual quality will not be any better UNLESS your editorial suite's HDV codec is inferior to your cameras codec. I have seen this in the past but am very happy with Sony Vegas' HDV codec. You can either decode the digital data on the miniDV tape during ingest or inside your computer when you edit. Given the same codec (which we admittedly don't have), the visual quality would be exactly the same, bit-for-bit.

The most sensible reason to capture via the HDMI port during ingest is to remove the pull-down. The blackmagic intensity card will not do this and I am only aware of one program that does and it is made by Cineform. I do not like editing with the Cineform codec, so ingesting with it does not seem sensible to me. I have yet to find a codec (other than uncompressed) that looks great and it highly responsive during editing. Any suggestions are welcome.

[Flyingsheep]

I've seen two frame grabs in comparison with each other, one being from the blackmagic intensity card (uncompressed) and another from a mini-DV tape with HDV compression. To be honest, the ONLY difference was the compression... and it wasn't that big of a difference. The colors were visually identical, and even the HDV frame grab being upscaled from 1440x1080 to 1920x1080 had no visible difference in resolution. I looked very closely and I couldn't even see the interpolation, unless some of it just happened to look like the HDV compression. My opinion is that it's not worth the hassle of always having a computer with your camera just so that you can get uncompressed video that will end up being compressed later anyway.

[Dagnanigans]

My boss (I'm a VFX artist) confirmed what I already thought. The compression happens as its being written onto the DV tape. Capturing from and HDMI cable afterwords is still going to have the compression. I want uncompressed footage because the compression does play a significant factor when it comes to keying. Have a look at the different color channels and you can see the artifacting quite clearly. So now I'm going to have to get a ton of extra gear to bring with me to any shoot. HV20 is sounding less amazing at the moment.

[Ian-T]

..... HV20 is sounding less amazing at the moment.???Why is that? Because you have to bring a ton of gear to your shoot to use HDMI? Buy a much more expensive camera...you would be faced witht the same issues. It is amazing really because f what it ofers at this price range. One thing I did not see anyone mention here in regards to HDMI out is the motion artifacts of HDV. Though it is more of a non-issue to me the benefits of HDMI is that you don't get that at all with the HDMI ('live") capture. The fact that this cam offers this "and" 24P tells me that it's amazing. Plus...the picture is incredible. I think that has something to do with the fact that its sensors "native" resolution is higher than 2k. I wish someone could apply a Hydra to this cam.

But i do get your point.....

[Frank]

Have a look at the different color channels and you can see the artifacting quite clearly.

What are you using to look at this and what color artifacting are you seeing? I ask because when you look at MPEG-2 video, you have to take the codec into account (Of course, motion artifacting is to be expected with MPEG-2 compression).

[Dagnanigans]

It's true that I didn't follow the process myself from start to finish with the footage I'm referring to, that again was my boss. But he guranteed me that what I was looking at was straight from the capture and then saved out in a lossless sgi format. The atrifacting looks similar to the kind you would find in a jpg image. My comment on it being less amazing is partly about the frustrating lack of aperature control and that annoying auto-gain function. That's why in my first post I was inquering about some programming hack. I agree it is awesome for the price, just hope that we can get a fix for some of these gripes.

[Ian-T]

I can dig what you're saying especially if you are used to a lot of manual controls from other cams. I try to keep things in perspective though and remind myself that this is only a consumer cam..a dam good one however. Hopefully the HV30 addresses these issues. Until then... there's no complaining from this neck-a-da-woods.

[Flyzilla]

The compression happens as its being written onto the DV tape. Capturing from and HDMI cable afterwords is still going to have the compression.

I had a discussion with Black Magic Designs about this a week ago, and I thought that you were wrong here. Still, I doubted myself, so I called them twice today and both techs confirmed my understanding of this camera and the process of capturing via HDMI.

It is true that the camera compresses to HDV when writing to tape, but the camera also uncompressed that HDV (using the same alogorithm that it used to compress to HDV in the first place) when sending the footage back out the HDMI port. So, at the HDMI port on the camera, the footage is Raw and uncompressed. It is at a full 1920 at this point. The camera is the best device to uncompress the HDV because it is using the same algorithm.

No more editing in HDV. This also works the same with AVCHD, so I can now edit in Premiere CS3 which doesn't support AVCHD.

Also, I won't need to build a computer to haul with me to the beach just so that I can film my kids in Full HD.

Black Magic did tell me that they are getting the "Do I need to bring computer equipment on site?" question a lot. No you don't have to. But if you want to, the Intensity card will allow you to do so.

I. personally, am not going to edit in Raw mode, but I am going to use Black Magic's MotionJPEG format (or maybe CineForm) during the capture to recompress the footage into something just a little bit smaller for the editing process. What this does it the following:

1. Gets the footage out of HDV 1440 format and into a Full HD 1920 format.

2. Keep me from having to build some crazy 8-disk RAID array just to edit. (I will, however, be using a 2 disk RAID 0 array.)

3. Output a great looking final render.

I don't really feel qualified to tell anyone that they are wrong about this kind of stuff, it's just that I am getting comflicting information.

I would love if the thread continued because I am going to be buying everything within the next couple of weeks, and I want to be sure that my workflow is going to work.

Any thoughts? -Flyzilla-

[jmorton]

"...It is true that the camera compresses to HDV when writing to tape, but the camera also uncompressed that HDV (using the same alogorithm that it used to compress to HDV in the first place) when sending the footage back out the HDMI port...."

Why don't you ask your contacts at Blackmagic to make it official and put out a press release addressing the issues and conclusions and implications you are making regarding the HV20.

I need to keep what is fact separate from rumor, conjecture, and assertions from a story twice told.

JM :)

[Dagnanigans]

if what the blackmagic people are saying is true then that is awesome. I would like to see this as fact before I spend the money though.

[benr]

So, at the HDMI port on the camera, the footage is Raw and uncompressed. It is at a full 1920 at this point. The camera is the best device to uncompress the HDV because it is using the same algorithm.

I'm not sure what the BlackMagic guys have told you exactly, but your statement above is a bit misleading. If you're capturing direct from the camera via HDMI, with no tape involved, then yes, it's uncompressed (I'd avoid the word raw, because that means many other things). If you've recorded to tape, and you're playing the tape out over HDMI, then it's a very different story.

There are two types of compression in the world: lossless and lossy. A true lossless codec reduces filesize (compared to uncompressed), but allows you to recreate the exact image data of the original. SheerVideo is a lossless codec, but still produces very large files (about half the size of uncompressed).

All other (lossy) codecs throw information away and degrade the image. Some, like Apple's new ProRes422, or the CineForm codecs - although lossy - can be described as "visually lossless" - they have been designed to hide the loss of information from all but a severe technical anaysis. HDV, however, crams an HD image onto a MiniDV tape. To do this, it uses MPEG2 compression at 25Mb/sec (megabits, not megabytes). MPEG2 is very much a lossy codec - you can tell just by looking at the numbers:

Uncompressed 8bit Yuv 1080i29.97 = 124 MB/sec
HDV 1080i = 3.12 MB/sec

These are megabytes, and yes, that's 124 vs 3. You're not going to get your uncompressed original back from that. (Reality check: HDV footage still looks pretty damn good for what it costs, and you have to pay a lot to get truly uncompressed capture to portable media.) There may be some minor advantage to going back via the same hardware codec to recreate the full raster 1920x1080 frame, but it's going to be marginal.

None of this is to say that going out via HDMI off tape is necessarily a bad idea, particularly if you have fast enough hardware and drives to capture uncompressed (or to a visually lossless codec) on the fly. On the other hand, going HDV over firewire, and re-compressing from the HDV footage at a later stage is going to produces very similar results.

The simple conclusion is: you're not getting any better image quality as a starting point by playing out HDV over HDMI instead of Firewire.

If you plan to do normal cuts-only editing, and your NLE supprts it, then you may as well stick with the HDV footage. If you're going to do major colour-work, filtering, compositing or multi layered editing (and you have the hardware and disk speeds to support it) then think about capturing or converting your HDV footage into a higher quality codec either on the fly via HDMI or by re-compressing from the captured HDV. It won't give you a better quality starting point, but it will hold up better over generations of work.

[RobPhoboS]

It won't give you a better quality starting point, but it will hold up better over generations of work.

Exactly what I wanted to hear after shelling out £150.
:hv20-smilie01:

Good info guys.

[Flyzilla]

Uncompressed 8bit Yuv 1080i29.97 = 124 MB/sec
HDV 1080i = 3.12 MB/sec

These are megabytes, and yes, that's 124 vs 3. You're not going to get your uncompressed original back from that. (Reality check: HDV footage still looks pretty damn good for what it costs,

It won't give you a better quality starting point, but it will hold up better over generations of work.


Yes, benr, you make perfect sense to me. That's why i kept asking Black Magic Designs "How can you get the "full HD" information back after it has been compressed to tape?" Even during subsequent calls to Black Magic, they stick to their answer. I think that I, being a noob, am just speaking a different language that they.

I posted the above message a while back when I was doing my research to buy a new camera. Since that time, I purchased a Canon HG10 and a Black Magic Intensity Card.

When I capture from the HDD through the Intensity card, I end up with a 1920 x 1080 (1.0) MJPEG image that looks, as you say, pretty damn good.

What the Intensity card does for me is gets my footage out of HDV or AVCHD and into a better CODEC. This is much better to edit with.

I am also playing with a trial version of Cineform and I love it, but I cannot afford $500 to purchase it at this time.

-Mike-

[David Newman]

I am also playing with a trial version of Cineform and I love it, but I cannot afford $500 to purchase it at this time.

-Mike-

Are you aware that there discount offered only on this site you can get CineForm NEO HDV for only $199, or $200 off NEO HD. See details here http://www.hv20.com/showthread.php?t=3992

[RobPhoboS]

I can't get my Intensity (on the PC) to capture anything :(
Hope they get back to me.

[joe12south]

It is true that the camera compresses to HDV when writing to tape, but the camera also uncompressed that HDV (using the same alogorithm that it used to compress to HDV in the first place) when sending the footage back out the HDMI port. So, at the HDMI port on the camera, the footage is Raw and uncompressed. It is at a full 1920 at this point. The camera is the best device to uncompress the HDV because it is using the same algorithm.
Sorry, but this is pretty specious advice. Once it's been recorded to tape, HDV has done its damage. The differences in decoders are not great enough to materially impact workflow decisions.

[Duke]

Here's the truth:

It is a hastle to drag your computer to a shoot, and it can't be just any computer. It must have (at least) a fast two drive Raid 0 array, or a three drive Raid 0 (striped) array. It helps if you have a separate drive for the programs and page file too.

However, uncompressed video directly from the HDMI is true 1920 x 1080. HDV is only 1440 x 1080 with a 1.3:1 pixel aspect. (The pixel is rectangular, not square.) This means uncompressed has 25% more image date.

Uncompressed direct video also has color space of 4.2.2. Regular HDV is 4.2.0 color space. Uncompressed has twice the color information.

You can't see any of this on a little screen, period. Blow it up 40 feet wide and 20 feet tall and you'll notice the difference. HDV color is only 8 bit, like a gif picture, you'll notice banding in things like a sky or sunset.

Doing it is only worth it if you're really doing a transfer to film. If you are, it's much better. If not stick to HDV. If necessary do static shots by direct capture and moving shots HDV.

Duke

[Mrpc9886]

"Uncompressed direct video also has color space of 4.2.2. "

I thought that was 4:4:4? 4:2:2 is DVCPRO or somesuch. Still compressed.

[Duke]

"Uncompressed direct video also has color space of 4.2.2. "

I thought that was 4:4:4? 4:2:2 is DVCPRO or somesuch. Still compressed.

Yes, 4.2.2 is still compressed. The Blackmagic intensity card does have a setting for 4.4.4, but I haven't tried it yet because that much info won't fit within 8 or 10 bits.

According to the drive test program that comes with the BM card my Raid 0 array won't handle that much data.

Right now I have two 7200 rpm 500 gig drives striped. Those will handle 4.2.0 or 4.2.2 (8-10 bit) I don't know if you'd have to run 3 drives, or 4 because it
would be 16 bit, it may very well be that I'd need 4 identical drives striped together since 16 bit is twice of 8 bit.

Because I haven't tried it I don't know if:
a) the HV20 really outputs to 4.4.4 or only 4.2.2 (I didn't want to claim something I didn't know for sure):hv20-smilie32:
and
b) if the BM Intensity can actually capture that much data.:hv20-smilie51:
and
c) if I can capture it if any of my programs will actually edit it.

Most of the time on the computer you must balance memory, drive speed and CPU speed. I'm running a quad core 2.4Ghz overclocked to 2.8Ghz with 4 Gigs of high speed ram. It might work if I install two more drives and reformat so all four work together and keep the 5th drive for programs.

If some one REALLY knows if the BM Intensity will handle 4.4.4 I'd like to know. It might convince me to get 2 more drives even though I couldn't back them up. Very risky that way could drive me to drink. :hv20-smilie71:

Duke

[Erik Bien]

Here's the truth:

It is a hastle to drag your computer to a shoot, and it can't be just any computer. It must have (at least) a fast two drive Raid 0 array, or a three drive Raid 0 (striped) array.

Or pester David Newman and CineForm to build their proposed Compact Flash Recorder (http://www.hv20.com/showthread.php?t=3672). :hv20-smilie84: Yeah, baby!

[SalaTar]

intensity card can only do 8 bit 4.2.2 (pcie 1x speed limitation )
"HDMI Video Sampling 4:2:2.
HDMI Color Precision 4:2:2.
HDMI Color Space YUV 4:2:2. "



Hv20 Hdmi is 4.2.2 8 bit

Kinda like they were ment for each other

[Duke]

intensity card can only do 8 bit 4.2.2 (pcie 1x speed limitation )
"HDMI Video Sampling 4:2:2.
HDMI Color Precision 4:2:2.
HDMI Color Space YUV 4:2:2. "

Hv20 Hdmi is 4.2.2 8 bit
Kinda like they were ment for each other

That all makes sense then, and I don't need to buy more drives! :hv20-smilie03:
And, yes the direct capture is very noticable in certain situations, but not so much others.

Also, you only need a reasonably fast cpu to work with it (thought they help on effects), but you do need fast drives.

Duke

[lordtangent]

You can't see any of this on a little screen, period. Blow it up 40 feet wide and 20 feet tall and you'll notice the difference. HDV color is only 8 bit, like a gif picture, you'll notice banding in things like a sky or sunset.

Wrong. You were doing just fine until you got to this one. HDV color is 8 bit the way JPEG is 8 bit. It's 8 bits per COMPONENT. (not pallet mapped 8 bit, like GIF which is only 256 colors out of 16.7 million TOTAL)

Of course, the fact that is in YCbCr instead of RGB and 4:2:2 sub sampled hurts it a bit. rec.709 YCbCr does not use the code space very efficiently so a lot of precision is lost. But it's nothing like the loss of pallet mapped ( "GIF-like" ) 8 bit.

[Tiberius13]

Flyzilla - It sounds like you have the answer I seek.

All I really seek is if 16:9 aspect ratio video is created by the Canon HV20/HG10. Clearly, the video is stored in 4:3 ratio - but is it squished/squeezed 16:9 video (it would appear squeezed if it was?)... so... if you get it back to 1920x1080.... does it appear stretched?)

You say that using the BlackMagic card and outputting the video from the hard drive of the HG10, you get 1920x1080 video. Does this video look absolutely bang on when it comes to aspect ratio? (ie. nothing is stretched or squished) If so - how does the video compare to the 1440x1080 version of the exact same video (that you would get if you used USB to put the video on a computer). If it has exactly the same picture - that implies that the 1440x1080 video is squeezed... or... the 1920x1080 video is stretched. One resolution has to be "not as it should be"... either stretched or squeezed?

I hope my question makes sense?

Thanks!


Yes, benr, you make perfect sense to me. That's why i kept asking Black Magic Designs "How can you get the "full HD" information back after it has been compressed to tape?" Even during subsequent calls to Black Magic, they stick to their answer. I think that I, being a noob, am just speaking a different language that they.

I posted the above message a while back when I was doing my research to buy a new camera. Since that time, I purchased a Canon HG10 and a Black Magic Intensity Card.

When I capture from the HDD through the Intensity card, I end up with a 1920 x 1080 (1.0) MJPEG image that looks, as you say, pretty damn good.

What the Intensity card does for me is gets my footage out of HDV or AVCHD and into a better CODEC. This is much better to edit with.

I am also playing with a trial version of Cineform and I love it, but I cannot afford $500 to purchase it at this time.

-Mike-

[lordtangent]

The Sensor of the cam is 16:9

The image is squeezed to record to tape.

Assuming the filter the cam uses for doing the squeeze is good, not too much actual quality is going to be lost on the squeeze. It's only a single chip camera and "1920x1080" photo-sites is not the same thing as "1920x1080" effective resolution. This is an issue of Bayer pattern filters (what is used for color reproduction on single chip cameras) and the simple fact of the Nyquist–Shannon sampling theorem. The MTF possible with a bayer filter sensor is just not equal to teh number of photo sites it has.

(The Panavision Genesis also outputs 1920x1080 and no one would argue about which camera makes a better picture. But they get a better MTF by massively oversampling the image with their large single chip )

Now... the loss due to the HDV mpeg compression is another story. There is definitely loss there.

The only argument FOR Intensity capture from stuff already on tape is that maybe (maybe) The codec in the camera would do a better job of decoding the video (that it made in the first place) than a software codec on your system can. Other than that there is no other area where might get any better quality by capturing from tape via the Intensity.

LIVE Intensity capture (straight from the HDMI without going to tape first) is definitely better. You bypass the motion compression and 4:2:0 color decimation of HDV. But then you have to shoot tethered. How long is your HDMI cable? Obviously, sometimes tape is much more convenient.

[Duke]

LordTangent,

You're right. I didn't mean to say that it was exactly like a GIF. Merely that it had limitations to 8 bits. Each 8 bit channel (like the red channel) is like an 8 bit gray scale gif.

In summary:

Full uncompressed is 4.4.4
Direct capture with BM Intensity is 4.2.2
HDV is 4.2.0
DV is 4.1.1

Not only do you have less resolution in regular DV, but more problems with colors compared to a direct capture. It does as good as the camera can do, and it's a bit better than tape.

Duke

[Duke]

Digital video is represented in either RGB or in YCbCr (often called YUV) colorspace. Both the RGB and the YCbCr colorspace contain three variables, also known as components (or channels.)

RGB's are red, green and blue.

YCbCr's are Y = luma (or black and white or lightness)
and Cb = Blue minus 'black and white',
and Cr = Red minus 'black and white'.

Subsampling (or compression of color) is a reference to keeping less detail for the color components in digital Video. It works because the human eye is more drawn to image and brightness than color.

Standard DV video (4:1:1) is keeping color for every 4th pixel on every line.

HD video and MPEG (4:2:0) is keeping color for every other pixel on the odd lines, and no color for the pixels on the even lines.

Direct video capture (4:2:2) colorspace describes keeping full image information, but color information (CbCr) for only every other pixel, on every line.

The problem comes when you blow up a video sized image and project it on a 40’x20’ movie screen. That means the movie screen is 480” wide.

If you do that with standard DV video, each pixel is then .66” square. If the color is on every fourth pixel you have a 2” rectangle with no color and a .66” square with color. Standard DV is very blocky and pixilated when projected to theater size.

With HD video, each pixel is .33” x .25” rectangular with color when projected, then a .33” x .25” pixel without color. The entire next line down has no color at all, then repeat with color. You heard that right. You will have a series of stripes 1/4” high x 20 feet long with no color unless you use a program that interpolates the color. The only place a program can get that information is already missing half the information you desire and when it does get it the info doesn’t exactly match the image info.

With direct capture video, each pixel is a .25” square pixel with color, followed by a .25” square without color, and every line has this checkerboard of color info. From 50’ away you don’t really notice a ¼” square when it’s surrounded by color.

Those same pixel sizes apply to image quality. The direct capture pixel is simply a smaller square.

In short, if you’re doing a POV (point of view) shot of a skateboarder going fast shoot with HD video and you won’t notice even when projected on the big screen. Do a fairly static shot (especially with difficult light or color) and you’ll really notice the difference of a direct capture when projected on the big screen.

I think I'm done on that subject forever.
Duke

[lordtangent]

I never though about the actual size of the pixels on a big screen. Thanks for doing the math on that. Very interesting when you look at it that way!

One thing you glossed over is how evil 4:2:0 is with interlaced sources. It's a real mess actually, because in order to get efficiency the codec has to treat each FIELD as a single image before applying the 4:2:0 sub-sampling scheme. That means the chroma samples, instead of being spread over 2x2 pixel blocks actually end up being spread over 4x2 pixel blocks. I'm not sure how badly it effects the 24p of the HV20. But it IS encoded in that 60i, so it might get munched by that interlaced 4:2:0 scheme. It's a shame they don't just store the 24p frames progressively.

You can read about interlaced 4:2:0 here (Good illustrations to visualize the problem): http://en.wikipedia.org/wiki/Chroma_subsampling#4:2:0

Yeah, 4:2:2 capture should really help spare the chroma from what I've started to call the "HDV meat grinder", especailly 24p material.

[Frank]

Duke,
That was a very interesting and informative primer. I just have one reservation about the conclusion you draw that video will look blocky on a big screen:
My videos look good on a big screen HDTV. I don't think most people in a movie theater sit close enough to make the screen look bigger than that TV. So I would think the image would be comparable.
That said, I'm sure everything you said about the damage color subsampling does is accurate.

[Duke]

I just have one reservation about the conclusion you draw that video will look blocky on a big screen:

My videos look good on a big screen HDTV. I don't think most people in a movie theater sit close enough to make the screen look bigger than that TV. So I would think the image would be comparable.


You're right, from that distance it doesn't appear to be huge blocks. I did a comparison test here:

http://hv20.com/showthread.php?p=32886#post32886

It just depends on what you want to do with it. Also, you can always degrade an image later in post (blur, etc) if that is your intent, but you can't make it better if it wasn't the best to start with.

Duke

[jmorton]

You're right, from that distance it doesn't appear to be huge blocks. I did a comparison test here:

http://hv20.com/showthread.php?p=32886#post32886

It just depends on what you want to do with it. Also, you can always degrade an image later in post (blur, etc) if that is your intent, but you can't make it better if it wasn't the best to start with.

Duke

How about this variation of the HV20: simply incorporate another mini tape drive or two or three or... However many to achieve the desired output quality.

How many mini tapes would be needed to record uncompressed video?

Then include an output setting that when capturing puts all the information together in a single captured clip. Make it so you could record in this better quality with more than one mini tape or in HDV using one mini tape in a single drive.

Add more manual controls and I'd definitely buy such an HD variant camcorder.

Should I send this suggestion off to Canon right now? I can't hold my breath that long. Maybe Santa will give me a BM Intensity card for Christmas just for trying to be good all year!

J "High Hopes" M :hv20-smilie03:

[lordtangent]

Cineform is working on a protable HDMI capture station that will do just what you are dreaming of here.

I don't think more DV tapes to get up to the required data rate would ever be practical. Uncompressed 1080i is like a firehose ( 119 Mega BYTEs per/sec ) compared to the waterweenie of HDV (25 Mega BITS/sec). Even a hard drive based solution would not be practical without some form of compression. The Cineform box I mention above will be using Cineforms wavelete codec to get the data rate down to something manageable. It's a good codec (and supports the 4:2:2 of HDMI) , so there will be much less loss than HDV.