View Full Version : DV compression question
John Jay December 12th, 2002, 09:36 PM My question concerns the signal at the 1394 port on any DV camcorder
Is each field of a frame compressed with the DV compression codec or is the compression done when a whole frame (2 fields) is ready before output at the port?
Rob Lohman December 13th, 2002, 05:21 AM You will get the whole frame at the port. Actually the frame is
devided into several firewire packages, I can get you specifics
if yoy want.
Whether the compression is field or frame based I'm uncertain.
But that doesn't matter since you WILL get the whole frame
(+ audio samples and extra information) at one time.
Hope this helps!
Bill Ravens December 13th, 2002, 08:18 AM hmmm...interesting reply, Rob. Considering that the codec is NOT used in a firewire transfer...the data is merely copied from one medium to another. The only time the codec is employed is during the "write to tape" internal to the camera and during a rendering process inside the computer. I'm not so sure the data is useable, either in field or frame format, at the firewire port. It's, quite simply, a data stream...in packets, yes, but, I don't know if those packets bear any relation to the discrete frame info cantained within.
Andre De Clercq December 13th, 2002, 09:06 AM Like Rob writes, the compressed DV video data stream consists of so called isosynchronious packets which contain compressed whole frame data. There is basically no "field after field" involved. Therefore field order setting in NLE and convertors is important. On the compression level itself DCT uses, depending on the motion in the picture a field based (2:4:8) or frame based (8:8) pixel structure for the DCT transformation, which is only the first step in the DV compression concept.
John Jay December 14th, 2002, 07:40 AM thanks all
the reason I asked is concerning those cams which support frame mode by two line averaging.
If the compresion is at the frame level then the signal should be better (less DV artifacts) since compression is applied to effectively half the signal whereas field compression would have no such advantage from a signal standpoint since it is just copied to the other field
Robert Mann Z. December 15th, 2002, 10:34 AM importing into a computer will compress the signal even more, check out the comparison featuring todays nles and card system, click on my signature link
Jeff Donald December 15th, 2002, 10:54 AM skyonic newyork,
Only the very last image displays. I can't see what you referring to.
Jeff
Robert Mann Z. December 15th, 2002, 11:35 AM in order to preserve the integrety of the images they are bmp @ 1 meg each, however the first image is a collection of all codecs
John Jay December 15th, 2002, 03:40 PM what were the settings used on the avid and main concept codex?
i ask because there are a choice of quality settings
Robert Mann Z. December 15th, 2002, 08:28 PM vegas was set to best...i know of no best settings on avid xpress
Rob Lohman December 16th, 2002, 07:54 AM Skyonic,
Taking footage in a computer will NOT alter its quality at all, since
it will only be a digital copy. It will alter when it is being displayed
(since it is going through a decompressor and color space conversions
and what not). But the main fact were people are loosing quality
is when changing the footage and/or outputting it again (back
to DV or MPEG2 etc.).
Robert Mann Z. December 16th, 2002, 11:18 AM Rob,
>>Taking footage in a computer will NOT alter its quality at all, since it will only be a digital copy. <<
hardly true, you never edit raw dv, i know of NO such app, everything is either an avi or qt wrapper with appropiate codec...the reason it's called dv codec is it's re-compressed into a wrapper from the camera...
even further if you use the matrox codec to bring in your footage your dv signal will be converted into rgb from yuv..
>>But the main fact were people are loosing quality
is when changing the footage and/or outputting it again (back
to DV or MPEG2 etc.)<<
yup... you lose even more at this stage, this actually a heated debate in the industry currently, you might see ads promoting a products lack of gen loss (generation loss). No Gen loss is based on a codecs ability to maintain quality even after 20 renders...
so the myth that dv footage is just moving data back and forth is just that...
John Jay December 16th, 2002, 07:54 PM skyonic
do you know if there are differences in the firmware compression between cams like Sony, Canon, Pana
ie do they use the same algorithms or are there differences in the way they are implemented?
(might explain why video recorded on a Pana plays back better on a Sony)
Robert Mann Z. December 16th, 2002, 10:02 PM jay this is very good question, i only know of the audio sampling differences, i was told a while back that most cams use sonys chip, but i could be wrong...
Jeff Donald December 16th, 2002, 10:11 PM I've had some Panasonic, Sharp and Canon DV cameras apart and I don't recall seeing any Sony chips in the Panasonic or the Sharp. I could be wrong, too. I wasn't looking that close, but I think I would have remembered a Sony chip in a Panasonic.
Jeff
Robert Mann Z. December 16th, 2002, 10:35 PM Jeff your probably correct, i know panasonic actualy sells after market encoder chips, canopus is one of there bigger clients...
i imagine if there are diff. they must be slim considering they are built to spec with little room to wiggle...then again there is a huge diff in how sony and canon sample audio so who knows...
Rob Lohman December 17th, 2002, 06:00 AM Skyonic,
What I meant was that if you transfer your footage to a PC and
would view it, it had no quality degradation. In theory
if you edit the program should just pass along frames without
re-encode (but that might be an interesting thing to test).
Premiere doesn't seem to re-encode when you just edit
(without fades or any other effect except a direct cut). But I meant
that there is no loss in transfer from the camera to your
computer. What you do after that might very well alter the
quality ofcourse!
I think we are saying the same more or less, heh.
Cheers!
Jeff Donald December 17th, 2002, 07:32 AM My understanding is that AVI is not a codec either. It is a file format for Microsoft Video for Windows. It has various codecs available such as Cinepak, Indeo Video and DiviX. So, if I understand you correctly, AVI is a wrapper, like Quicktime.
Jeff
Robert Mann Z. December 17th, 2002, 09:27 AM let me see if i can shed some light on this:
ok your cam sends the raw dv into your computer, your computer must then encode raw dv, this is done on the fly....if your on a mac it encodes it using qt wrapper and qtdv codec, if your using avid on a mac it will be a qt wrapper and a avid dv codec, if your on a pc with a standard fire wire card and no other codecs installed besides preimeres default msvd codec it will be an avi wrapper and a msdv codec...
so the process of adding a wrapper to raw dv is not a problem, the codec is, not all codecs are the same, some use softeners some don't...if the test were conducted with input from a dv source into the computer you would see the same results...this is your first gen
when you render in premiere and you have no effects or transitions it should not do anything to your first gen, however if you apply a transitions, that part will be re-encoded (second gen)...
this is why there are different codecs, this is why some are prized some are not, you can be the judge which is best
:)
Bill Ravens December 17th, 2002, 09:57 AM Sorry to differ with you skyonic, but, when the DV file is "captured" on a PC, there is absolutely NO codec involved. There is only an AVI wrapper put on the bitstream coming from the computer. The codec is applied within the camera..and 99% of the time, the codec is/was written by Sony. The only time the codec is used, again, is when the avi file is read into the NLE software. Within the NLE, the codec is not used until the frame is modified in some way, like a transition or effect. At that point, the codec may or may not me the M$ codec, depending on the software.
Rob Lohman December 17th, 2002, 11:23 AM I agree 100% with you Bill! The only thing that happens when
"capturing" (which is actually a bad term for DV/1394) is that
the raw DV stream is encapsulated into a known file format like
AVI or QuickTime. The data in itself is not altered in any way
(except if you split the audio channels, but even then the audio
samples themselves are left alone)
Robert Mann Z. December 17th, 2002, 11:58 AM bill so you are saying that with avid xpress i can edit raw dv, i think not...avid encodes it's own codec to dv, as does matrox, canopus, pinn...and any other nle...this is simply just how it's done, in our premiere systems we can choose to use main concept or msdv on capture...same goes for canopus (canopus or msdv) on our avid station we can only use avid codecs all other will be re-coded...
>>The data in itself is not altered in any way
(except if you split the audio channels, but even then the audio
samples themselves are left alone)<<
a myth...maybe next month we will add a test to show what each looks like after capture from a dv deck...
Bill Ravens December 17th, 2002, 12:35 PM Skyonic....
All NLE's, including Avid, do NOT impress their own codec on the DV stream, until you load that DV stream into the NLE. Even then, the original DV file is not altered by any codecs, IN ANY WAY, until you re-render and re-save them.
This isn'r rumor, it's fact.
Go read up on the function of a codec at adam wilt's site.
Robert Mann Z. December 17th, 2002, 01:13 PM thats interesting, then how is it that a system like matrox which codec is rgb not native yuv is the only way to get rt effects with there board even before you apply any effects...
changing dv from yuv to rgb is what i call altering... but wait there is more
according to your post a system like canopus would work this way...you capture using your storm system to a 'non altered dv stream' ... import the footage into premiere then it gets tranfered into canopus codec?... or does it get tranfered to canopus codec after you first render?..
sorry it doesn't work that way if you are using a canopus card (for example) and you capture you will be using the canopus codec...for those of you following this and have two different codecs try the follwing capture with one then capture with another...you will see difference, for those that have a storm can see a big difference just by the color space (canopus i s 4:2:2, msdv is 4:1:1 color space) when compared with msdv...
i read adams site where does it say that codecs or for that matter custom codecs don't change anything...
John Jay December 17th, 2002, 01:35 PM skyonic is right on the button
if you are on a PC and you register mainconcept over the default msdv and then grab at the fast settings the vid you get is mush..
game over
Bill Ravens December 17th, 2002, 02:12 PM ahhhh....there is a distinction, and I'm sorry i've overlooked it in my post. HARDWARE based capture cards will use their own codec. It's only the software based capture systems(and one's that use generic IEEE1394 controller cards) that don't rely on a codec to capture. Indeed, the hardware based systems will manipulate the data stream as well as add the avi wrapper.
|
|