hierarchy of digital video

[Jason Maeda]

can someone list for me the different types of digital video mediums, in order of compression/quality?

(example: HD, digibeta, dvd, etc...ending with the worst/most compressed)

jk :ph34r:

[David Mullen ASC]

Well, the problem is how do you list Digital Betacam, which is a standard def video format with less compression, against HDCAM and DVCPRO-HD, which has more information, more resolution, but more compression?

[Filip Plesha]

You can make two lists.

One for SD and one for HD.

[David Mullen ASC]

It would probably be:

 

Standard def (digital):

D1

Digital Betacam

DVCPRO-50

IMX

DV25 (DVCAM, DVCPRO, Mini-DV)

DVD?

 

Standard def (analog):

Beta-SP

3/4" U-matic

Hi-8mm

VHS

 

Hi-Def:

HDCAM-SR

HD-D6 VooDoo

HD-D5

HDCAM

DVCPRO-100 (HD)

D-VHS

HDV

 

Don't where to stick DVD, since it can use different compression rates. It's not really a mastering format but a distribution format.

 

And how do you compare HDCAM (150 mgb/sec) to DVCPRO-HD (100 mgb/sec)? The first has more data but less color information (3:1:1 versus 4:2:2). Compression is high for both.

 

And what about HDCAM-SR, which is 4:2:2 and 4:4:4 switchable? And how does that compare to HD-D6, which cannot record 4:4:4 in real time but at half speed?

[Filip Plesha]

Where does sony's digital 8 go?

I think it is about the same quality as mini DV,

the only difference is that mini DV cameras get better CCD's

[Alvin Pingol]

Where does sony's digital 8  go?

I think it is about the same quality as mini DV,

the only difference is that mini DV cameras get better CCD's

Digital8 records the exact same 25Mbps signal as MiniDV. The only difference is in the tape.

 

However, as you said, most Digital8 cameras are very low-end consumer.

[Guest Ultra Definition]

The way I see it; modified David Mullen's table; overall quality, not compression is compared; similar quality is grouped together:

 

 

Standard def (digital):

 

Digital Betacam

IMX, quality close to Digital Beta

DVCPRO-50, quality close to IMX

 

DVCPRO progressive

 

DV25 (DVCAM, DVCPRO, Mini-DV), Digital 8

 

Micro MV

 

DVD

 

 

Standard def (analog):

 

Beta-SP

 

3/4" U-matic SP

 

Hi-8mm, Super VHS

 

VHS, 8 mm, Beta

 

 

Hi-Def:

 

HDCAM-SR uncompressed, Viper

 

HDCAM SR recorder

HD-D5, quality close to HDCAM SR recorder

 

HDCAM

 

DVCPRO-100 (HD)

 

D-VHS, HDV, Blu-Ray DVD (as of now only available in Japan)

 

 

When we mix all these formats together:

 

HDCAM-SR uncompressed, Viper

 

HDCAM SR recorder

HD-D5, quality close to HDCAM SR recorder

 

HDCAM

 

DVCPRO-100 (HD)

 

Digital Betacam

IMX, quality close to Digital Beta

DVCPRO-50, quality close to IMX

D-VHS, HDV, Blu-Ray DVD (as of now only available in Japan)

 

DVCPRO progressive

 

DV25 (DVCAM, DVCPRO, Mini-DV), Digital 8, Beta-SP

 

Micro MV

 

DVD

 

3/4" U-matic SP

 

Hi-8mm, Super VHS

 

VHS, 8 mm, Beta

[Marty Hamrick]

So where would Beta SX fit?

marty

[Phil Rhodes]

Hi,

 

Beta SX is a pretty horrible format designed to deliver 4:2:2 images in a data space that's actually smaller DVCAM's 25mb/s (although SX uses beta-form-factor cassettes.) It uses I-frame MPEG-2 compression at (controversially) some horrible ratio between 6 and 10 to one.

 

This was clearly done to satisfy broadcasters' common but somewhat naive technical requirement that all material be 4:2:2. The result is a format which records less data than DVCAM but locks you into unremarkable HAD-based cameras which must use bulky, expensive, power-hungry Beta-scaled VTRs, and requires you to use horribly expensive Sony equipment throughout your post chain. Basically it's a format designed to satisfy a requirement that made little sense in the first place, and as a result is a total botch-up.

 

Sky News in the UK shoot SX - and then downconvert (well, upconvert really!) to 25-megabit DVSD firewire to cut on laptops in the field, then re-convert it to SX for interchange and storage. They will claim, however, that DVCAM is not a professional format because it's "not 4:2:2." The mind boggles.

 

In any ranking, Beta SX sits about equal with DVCAM and good Beta SP on raw image quality, but in terms of actual worth and usefulness, is in a fight with Pixelvision for last place.

 

Phil

[Tenolian Bell]

Phil

 

Why do you think Sony is going down this MPEG route with future formats? I don't know any one who understands it, everyone's scratching thier heads and saying mpeg is a horrible codec for shooting on. Digital cable in this country has gone mpeg and they promote it crystal clear when in reality it looks worse than if you recieved the channel with a regular antena. What is Sony saying that justifies mpeg?

[Guest Ultra Definition]

Everyone is going MPEG. MPEG2 is more efficient than the Sony developed DV based codec that is used in HDCAM. DVCPROHD, DVCPRO50, DV, DVCPRO, DVCAM. If you'd apply good MPEG2 processing to these formats, you'd end up with a significantly better image overall. MPEG4 is more efficient than MPEG2. There are new even more efficient codeces than MPEG4 but today's technology to make these is too primitive to make them small enough, inexpensive enough, with small enough power consumption.

[Guest Ultra Definition]

HDTV has too many MPEG2 compression artifacts because the bit stream is too narrow. The standard HD bit stream is on the order of 100 Mbps. HDTV is normally 19 Mbps, or less.

[Phil Rhodes]

Hi,

 

No idea what the official Sony line is. However, it's important to differentiate things like digital cable and a shooting format - clearly one is inferior to the other in terms of absolute technical standard, but they're very different animals in other ways.

 

Cable is a stream - its content is known in advance (at least fifteen frames in advance for most DVB streams, and often up to two seconds in reality) and the bitrate may vary. Neither of these things are true of tape based formats, which is why they invariable (Oh, okay, almost invariably) use I-frames only - the frame to frame difference management of MPEG-2 is completely overlooked for these formats.

 

However the kicker is this - cable MPEG won't be edited or manipulated in other ways, such as colour correction, at any downstream stage. This knowledge empowers much higher levels of compression, because the originating device knows that its compression decisions won't be made more obvious by later actions it can't predict, particularly contrast increases.

 

Neither of these things intrinsically make I-frame MPEG a bad choice for origination, though. After all, it supports 4:2:2 colour space, which DVSD does not. However, trying to make any codec perform at such a low bitrate (20mbps for SX) is simply going to make for sucky pictures. It's 4:2:2, sure, but it's so heavily crunched that this is almost meaningless. Well-upsampled DVSD, particularly PAL DVCAM or miniDV, has more information in it.

 

Good reasons for using MPEG in any situation include its scalability - an MPEG device can be built which will handle hi-def, and it will more or less intrinsically handle lower resolution images. MPEG decoders also intrinsically handle lower revisions; an MPEG-2 decoder can handle MPEG-1 and decoders aren't really supposed to care about bitrate either, including variable bitrate within a stream, to within certain sensible limitations. MPEG decoders aren't supposed to care about GOP size or I-frame-only sequences. The format is flexible. MPEG also includes well-compressed audio options. Interchange is also a bonus; MPEG is implemented in many systems and the intrinsic scalability and interopability of the format is robust. One particular gain here is that future MPEG codecs are intended to continue to retain compatibility with earlier formats, which mitigates concerns over archive format obsolescence to a certain degree.

 

The problem is that bugger-all of these bonuses relate to cameras. MPEG is clearly designed to be a distribution format - videoCD, DVDs, Internet streaming, video phones, and the like. There's no tremenous contra-indication to using it in acquisition - frankly, there's very little technically to choose between motion-JPEG, DVSD and I-frame MPEG, other than that they're all scalable except DVSD and JPEG is somewhat inferior to both. Sony even managed to scale DVSD for its HDCAM format. Why did they go to MPEG? Well, probably because it was there.

 

MPEG-4 is a somewhat different matter. It is conspicuously clever and is certainly worthy of the Cunning Plan Medal.

 

Phil

[Guest Ultra Definition]

Phil,

 

I don't hink that there is any reason to use I frames only. The uncompressed stream is not that large. You can have plenty of buffer memory and processing power to create quite large GOP's. It would not make any sense to use I frames only.

 

No matter what, MPEG definitely appears to be the current and future trend. XDCAM is MPEG2, CineAlta SR is newer and is MPEG4. HDV and Blu-Ray is as old as XDCAM and is using MPEG2. All these are acquisition formats.

 

Joe

[Phil Rhodes]

Hi,

 

> I don't hink that there is any reason to use I frames only.

 

It's hugely easier to compress. It's easier to transport, and it's easier to cut. The image fidelity is higher, and I choose he word "fidelity" over "quality" in this context.

 

> The uncompressed stream is not that large

 

Well, you'd probably need thirty or forty megs of RAM to contain a single PAL 8-bit 602 GOP, assuming fifteen frames, and that's at least thirty or forty rather audacious dual-port framebuffers which would have a fairly extreme cost and power consumption implication. Don't forget that modern cameras are engineered to be micropower devices in much the same way as mobile phone handsets, for stamina.

 

But that's not the issue.

 

> You can have plenty of buffer memory and processing power to create quite

> large GOP's.

 

Actually you can't. GOP-based compression is orders of magnitude more mathematically complex than just the I-frames. The realtime MPEG-2 encoders you get in things like set-top DVD recorders and computer video handling boards are absolutely horrible; the image quality is not suitable for broadcast. Decent-quality MPEG-2 IBP encoding is the realm of very upscale realtime encoders by the likes of Hitachi and Mitsubishi, which work in realtime but cost half a million, and very fast desktop computers which may take ten to fifteen times realtime to come up with a similar result. Bear in mind that a modern desktop processor - just the processor - comfortably pulls 60 watts, which is twice the entire power consumption of the average ENG style video camera. It DOES make sense to use I-frame only, since the compression load is similar to M-JPEG or DVSD. Where d'you think all that juice goes in a DVW-790 or a HDW-F900? The beta-sized decks are hungry, but there's things going on in there which would have been considered supercomputing a decade ago.

 

Phil

[Guest Ultra Definition]

Why would they use MPEG4 then; it would not make any sense. And the MPEG4 processor is a stand alone item and it is just now coming to the market so latest microtechnology is used. Plus even the cheap JVC HDV camera is using GOPs in their MPEG2 codec; not only I frames. I guess we'll have to ask Sony about the GOP size, if only I frames are used, etc.

[Jarin Blaschke]

It's sad that one can't be a photographer anymore without also being an electronics tech.

[Mitch Gross]

You keep saying that everyone is going MPEG these days when it is simply not true. Consumer formats and distribution formats are going MEG, and Sony is going MPEG. But others are not and they'll be going very far.

 

The challenge in aquisition isn't compression. Compression sucks and no one wants it if they can possibly avoid it. Any serious shooting format is try to limit compression as much as possible. Dalsa uses a scalable format for their Origin camera that uses Mathematically Lossless compression. That means that you can expand the signal back out for manipulation without ANY signal degredation or artifacts. Not "any noticeable" loss, NO loss. Kinetta will have zero compression--that's absolutely no compression of the signal. Same goes for some of the solutions coming out for the Viper.

 

BTW, do you even know of the Viper? It's been out there for a while but you never seem to mention it. It uses a Dalsa chip. Elsewhere you mentioned Sony & Panasonic having some sort of strangehold on HD chip technology, which is stunningly untrue and never has been true. Which corporate newsletter did you read that one from?

[Guest Ultra Definition]

Comon Mitch, I don't need this. Of course I know Dalsa. But do you know that it is only a prototype as of today; it is not a production camera. Of course I know Viper. Can you name a major motion picture that was shot with it? And Viper is not using Dalsa technology. Both Thomson (Viper) and Dalsa are using Philips technology after Philips get trid of a particular small division. Of course uncompressed is better and of course Sony knows what it is doing when it is using MPEG2 and MPEG4. This is the real world.

[Mitch Gross]

Of course I know Viper. Can you name a major motion picture that was shot with it? And Viper is not using Dalsa technology. Both Thomson (Viper) and Dalsa are using Philips technology after Philips get trid of a particular small division.

Michael Mann is currently shooting a feature with the Viper. It took two years from when Sony introduced the f900 until the first feature film shot on it was released (Jackpot, photographed by David Mullen). I never said the Viper was using Dalsa technology. I said that Dalsa makes the sensor chip that the Viper uses, which is true. This was to point out that Sony & Panasonic do not have any sort of stranglehold on chip sensor technology, which they do not and have not. The division of Dalsa that makes the chip did used to be owned by Philips, which if you know this it makes me curious as to why you made the statement implying that Sony & Panasonic did have this power.

[Guest Ultra Definition]

Sorry, I use the term camera for a camcorder. Sony and Panasonic make HD camcorders. HD cameras are made also by Thomson, Hitachi and Ikegami. I haven't seen anyone else's HD cameras. Panavision HD camcorders are basically Sony CineAlta. I'm sure that there will be more players in this field in the uture. I've always been with you on that one.

[Phil Rhodes]

Hi,

 

Companies such as Ikegami and I believe Hitachi also make hi-def studio cameras.

 

Phil

[Guest Ultra Definition]

that's what I said

[Mitch Gross]

Well now that you've edited your post to reflect what Phil said...

 

If you or anyone else would care to do some Google searching I'm sure you'll find a number of specialty companies who make HD cameras for many specialized industrial purposes. There are some high speed cameras that go up to 15,000 frames per second (that's 15,000p in video parlance) in HD resolutions.

[Phil Rhodes]

Hi,

 

I've often wondered about using those instrumentation cameras for dramatic production. I've been talking to a production company recently about a hi-def feature where they want to use high frame rates, and there was a company promoting these specialised cameras at a recent trade show. The type I saw shot 1024x1024-pixel 8-bit images, which I'd consider marginally suitable for hi def work, particularly cut in with Varicam material. Anyone used them?

 

Phil