Is all analogue video sampled at 4:2:2?

[DNICE_ONE]

Hi!

The manual of my DataVideo TBC-3000 says that it has a 4:2:2 sampling rate. And since it can only work with analogue signal, does it mean that all analogue video signals use this sampling rate or is the TBC-3000 doing a colourspace conversion internally? And if analogue signal uses 4:2:2 then which of the following will result in better quality video?

1. Analogue(4:2:2)-->DV(4:1:1)-->MPEG2(4:2:0)

2. Analogue(4:2:2)-->MPEG2(4:2:0)

[lordsmurf]

Most traditional sources are 4:2:2, yes.
VHS, broadcast, S-VHS, 8mm, etc

Then you have some like Betacam (NOT BETAMAX!) that are 3:2:2

FYI, in the UK, PAL DV is 4:2:0 ..... and 4:1:1 is only NTSC DV.
Those are, of course, consumer DV25. The DV50 is 4:2:2

[davideck]

Analog formats do not have a specific sampling structure associated with them. The DataVideo TBCs sample at 4:2:2 to preserve analog bandwidth.
They tend to soften the image a slight bit, however.

Theoretically, it is best to colorspace convert as little as possible.
I would not expect that an intermediate conversion to 4:1:1 would provide any improvement.

[DNICE_ONE]

Most traditional sources are 4:2:2, yes.
VHS, broadcast, S-VHS, 8mm, etc

Then you have some like Betacam (NOT BETAMAX!) that are 3:2:2

FYI, in the UK, PAL DV is 4:2:0 ..... and 4:1:1 is only NTSC DV.
Those are, of course, consumer DV25. The DV50 is 4:2:2

So, if PAL DV is sampled at 4:2:0 then theoretically only one colourspace conversion takes place (4:2:2 to 4:2:0) when the analogue source is first captured to DV AVI and then encoded to MPEG2 right?

LS can you please elaborate on this. Thanks everyone for your valuable feedback.

[Cornucopia]

"Sampling" has to do with the conversion from analog to digital--taking discreet SAMPLES at definite points in time. Up until it is sampled (digitized), it is continuous. That is, THERE IS NO SUCH THING AS SAMPLES OR PIXELS IN ANALOG. Those are digital terms.

Analog has bandwidth constraints. This "corresponds" to a certain number of pixels if sampled at a standard, fixed/constant sample rate. The ITU-601 spec uses 13.5MHz, IIRC.

Using this for the conversion factor, VHS can be said to have an evivalent usable resolution of 352x480, SVHS 480x480, Betacam at 720x480.

In the same way, colorspace isn't really sampled either in the analog domain. The "4:" etc refers to the sampling ratio between the luminance and chrominance components. For expediency, let's say that all full-color-carrier formats (1"C, BetacamSP, etc) are the equivalent of 4:4:4 at best (but usually equivalent to 4:2:2), whereas color-under-subcarrier formats (VHS, SVHS, 8mm, Hi8, 3/4" umatic)--because of bandwidth restrictions and comb filter error--are equivalent to 4:2:2 at best, but usually more like 4:1:1 (as the color is there in every line of every field, unlike 4:2:0).

Rule1-it's always best to convert as little as possible
Rule2-it's always best to convert down as opposed to converting up

PAL DV is 4:2:0. This sampling happens during the digitizing. It converts the ~4:4:4 colorspace of analog to 4:2:2 (while internally sampling) and saves as 4:2:0. Re-encoding to MPEG doesn't change this at all (also 4:2:0).

Guess what? Your Datavideo TBC is a digital device! That's why it says something about sampling.
It does this:
AnalogIn-->Sampling/Digitizing-->TBC processing-->D-to-A-->Analog Out

Looks like analog on both ends, but isn't internally.

So it's doing:
4:4:4 (analog) --> 4:2:2 (internally digital) --> 4:4:4 (analog again)
Of course, once lost, those chrominace samples aren't regained, so even though it's analog it really has only 4:2:2 of color gamut at best.

Scott

[lordsmurf]

Nice explanation, Cornucopia.
Very neat overview.

[DNICE_ONE]

Guess what? Your Datavideo TBC is a digital device! That's why it says something about sampling.
It does this:
AnalogIn-->Sampling/Digitizing-->TBC processing-->D-to-A-->Analog Out

Looks like analog on both ends, but isn't internally.

So it's doing:
4:4:4 (analog) --> 4:2:2 (internally digital) --> 4:4:4 (analog again)
Of course, once lost, those chrominace samples aren't regained, so even though it's analog it really has only 4:2:2 of color gamut at best.

Thanks for a very elaborate feedback cornucopia.

So if the source I am trying to capture is a very good quality VHS then will it affect the quality of the captured video if I don't pass it through the TBC and capture directly from the source?

It may save the signal from being downsampled and then upsampled again by the TBC and losing all those chrominance samples.

[lordsmurf]

The TBC won't cause any real loss because the carrier signal is basically just 4:2:2 data anyway. The TBC resamples won't really lose anything. Using a TBC is completely safe, and a TBC helps an image 99%+ of the time, not harm it.

Is that what you're asking?

[DNICE_ONE]

The TBC won't cause any real loss because the carrier signal is basically just 4:2:2 data anyway. The TBC resamples won't really lose anything. Using a TBC is completely safe, and a TBC helps an image 99%+ of the time, not harm it.

Is that what you're asking?

Yes! that was my question and thanks for clarifying that.

Can you please tell me which method for capturing results in the minimum colourspace conversions? And what colourspace does ATI MMC capture in using HuffYUV with ATI AIW?

[vhelp]

@ dnice_one

Basically, what the specs of a given device is saying is..

That it will give you 4:2:2 sampled data, which is about the
max you would get from a tipicle source.., be it from your
cable/antenna; vcr; tv tuner; laserdisc player; dvd player; etc.

All of these sources above have 4:2:0 format, and are stored
that way onto the medium, be it through cable/satellite/antenna;
or vcr; or laserdisc; or dvd player;

When you connect a given device to it, your device is basically
starting from scratch (w/in limits, of course) and is pulling
from this pile of data bits.. and re-piecing them back together
again. During this process, it performs a number of internal
mesagues and things like conversion, etc.

In the end, what you (your card) will end up with, is an image
that is in a sample format of YUV 4:2:2 and this is the max you
will receive in that given card (or device) and if you are
knowledged about this, and many other facets of video, you will
know what to do next. (I'll tell you) ...

You would capture *RAW* this YUV 4:2:2 video with your capture
card's codec that provides this format of capture. That is,
you don't want to use a codec that captures this 422 and then
compresses it, like huffy does, and then say you are capturing
4:2:2 raw data.. becuase, IMHO, you are not. You want to capture
RAW, period.

If you want to capture the maximum that your capture card will
provide, then you want to capture in the RAW, this YUV 4:2:2
[Or, YUV422, as I refer at times] then you want to be sure that
your given capture card comes equiped with the proper RAW codec
for this format of capture.. but expect to prepare yourself with
a large HDD for this. Because for every one minute of capture,
your HDD consumes 1.25G space.. or approx 75G per hour.. or 150G
per two hour.. etc. This is why many users resort to another
codec that provides similar 4:2:2 sampling, but w/ compression
added in to the chain of events.. and comprising on whatever
quality loss may result. There are many that believe huffy is
lossless. That is also a debatelable topic.

Now, as far as a proper codec for this RAW yuv422, I know that
with ATI, it comes with a codec that allows this. I don't know
about other capture card brands, however.

By default, ATI 's setup is with the ATIVYUY.DLL codec. This is
a RAW YUV422 codec.. and uses the VYUY formating for the 422 sample
layout. But, from my experience with installing such an ATI type
card and MMC, it will usually set you up in MPEG real-time capture
mode. So, be watchful of this. Don't assume you are going to
start capturing RAW yuv422

There is also M$ 's MSYUV.DLL codec, which is 422 sampling as well
and can use the VYUY or YVYU layout.. still 422 sampling.

And, also HUFFY, is 422 but compresses it w/ a huffman algorithem
or something.. making the final filesize much smaller.

These all do the same thing, [capture to 422] but the layout of the
pixels are different, is all.

i420 uses msh263.drv codec. It is an 4:2:0 sample codec.
Personally, I would not use it, if your goal is as much quality as
possible.

y41p uses the btwc32.drv codec. It is an 4:1:1 sample codec,
similar to DV which is 4:1:1 also, [but compressed].

There are many more, but this should be enough.

But, put it to you this way, if you want to be sure you are
capturing yuv422 in the RAW, do a 1 minute capture, and verify
the filesize is +/- 1.25G. If it is smaller, then you are prob
using a codec that is incorporating a compression scheme.

Getting back to your TBC device ...

So, it is giving you a 4:2:2 sample format of video.

The next step in your process, is to capture *this* 4:2:2 sample
data. That is why I commented above on the codecs.. to get you
started in the RAW.. if that is what you are after.

Now, you have to know one more thing. Does your capture card
give you 4:2:2 sampled video. Most, if not all capture cards do.
And, now that you know what you cap card is capable of, you can
react by using the proper codec.

Your next step is to know which is which, (codec) and use it.
If you don't select a given codec in your capture's app setup,
then you may most likely end up capturing uncompressed RGB data
which is even larger than 4:2:2 sample format data. This is
usually the default when you first configure or install a given
capture application.

To Recap:

** your TBC device is telling (per spec) is serving your 422 samples
** you learned that all (consumer limited) sources output 4:2:0 sample format.
** determine if your capture card serves your 422 sample video
** educate yourself about your cap cards codecs, and know which is 422 RAW sampling
** determine how your capture cards app sets you up w/ what avail codec(s)
** educate yourself on what codecs incorporate compression (stay way, if pref)
** capture in 422 sample format, RAW for maximum reproduction quality.
** .
** After here, your next steps ...

-vhelp 3580

[davideck]

Getting back to your TBC device ...

So, it is giving you a 4:2:2 sample format of video.

The TBC is analog in / analog out.
Internally it is 4:2:2, which hopefully makes it sufficiently transparent so as not to reflect its sampling structure on its analog output.

Particularly with respect to S-Video in/out (to eliminate any composite to component to composite artifacts), as long as bandwidth requirements are met, the internal sampling structure of the TBC should not matter to any subsequent operations.

[DNICE_ONE]

Thanks Vhelp! That was some fantastic peice of info. Surely, now I have enough knowledge to decide what to do. Now I can buy a capture device according to what I want based on all the info I have received in this forum. I has been really helpfull being a member of this forum.

Thanks all of you guys for helping me decide. Soon I will do my first capture and let you know about my experience and the results.