352 vs 720 resolution at 2700 Kb/s

[davideck]

Last night, I did an A/B comparison of two 20 minute captures of the same source segment, one at 352x480, the other at 720x480. Both bitrates were 2700 Kb/s. I was capturing a Beta III recording through a TBC-3000 to a PVR-250, trying to fit 3.5 hours onto one DVD. I burned them to separate DVD-Rs and compared them simultaneously using two DVD players and a PIP TV, swapping between the captures every several seconds or so. Then I exchanged the DVD-Rs in the DVD players and did the A/B comparison again. IMHO, the 720x480 capture looks better. The images were sharper, and more detail was preserved. Any areas of blockiness were more pronounced at 720, but this seems directly related to the extra sharpness obtained at that reolution. I think it would be very hard to tell the difference if you weren't comparing them side by side. Would anyone care to clue me in on the advantages of capturing at 352?

[lordsmurf]

No.

An observance of 720 being "sharper" is DIRECTLY related to the hardware, software and source. You cannot surpass the resolution of the source (maybe if you use a detailer, but that's a whole different conversation). Most sources are at, under or around 352x480. Almost nothing aside from DV is available to the consumer in that higher res.

Macroblocks are an artifact created from lack of bitrate. 2700k is a SERIOUS lack of bitrate for 720x480, and can even be a bit to low for 352x480.

Offhand, I have no idea what the source res is for a Beta III tape. I'm not familiar with that format. It may be high res.

Your DV box only captures 720x480, DV is that res. If you saw softening of the footage at encode, blame the encoder.

If it was the Hauppauge PVR doing MPEG encoding, those boxes are slightly soft. Same for MainConcept (unoptimized) and TMPGEnc.

Double digitizing the signal (first to DV) and then to the PVR-250 was probably very damaging too. You have a DV box ... transfer it.

Hope that clears some things up for you.

[babyboo]

I have done my own comparison with VHS and analog cable at 352 vs 720 and quite honestly I do not see any difference when the source was vhs between the 2 resolutions. I do, however, notice a somewhat sharper, clearer and more detailed picture when the source is analog cable. What is the source resolution for analog cable? Is it higher than 352 (I realize a resolution of 352 is NOT the same as 352 lines of broadcast resolution) I am just wondering approximately how much detail you are losing if you cap analog cable at 352?

[lordsmurf]

It's in the low 400's at best. (440x480, max possible in theory ... often nowhere close in practice) ... 352x480 is going to be pretty close

[babyboo]

Thanks. I figured it must be slightly higher since 720 seems better than 352

[Cornucopia]

Beta III is, if I'm understanding his reference correctly, Betamax LongPlay (LP) speed which give ~3.5 hours to a tape.

While Betamax is somewhat of a better signal than VHS, you're still starting with an LP recorded signal. My guess is that the nearest equivalent resolution you can resolve this signal to is ~ 300x300. (Off-tape blurriness reduces 480 lines to visual perception of ~230 in VHS, I expect more for Betamax.

BTW, analog signals don't have a "resolution" as defined the way digital signals do. They have a bandwidth and an equivalent resolution can be approximated with TVL charts, etc. I'm really getting tired of people throw around numbers for analog signals (Betamax, VHS, LD, Betacam, SVHS, Analog cable and broadcast, etc.), when there is so much variability.

[davideck]

Yes, Beta III is the LP version of Betamax; 4.5 hours with an L750 tape. It has more detail than VHS SP at roughly the same noise level.

I am wondering under what conditions a 352 capture would yield better results than a 720 capture at the same bitrate. From the replies so far, it seems like 720 is the better choice, regardless of source detail.

[lordsmurf]

BTW, analog signals don't have a "resolution" as defined the way digital signals do. .

That's a given, but they have approximate equivalents.

I am wondering under what conditions a 352 capture would yield better results than a 720 capture at the same bitrate. From the replies so far, it seems like 720 is the better choice, regardless of source detail.

No. To upsize that much is to create new data, which in turn creates new artifacts. One of the most predominant ones is linear errors of sorts.

[Wilbert]

Yes, Beta III is the LP version of Betamax; 4.5 hours with an L750 tape. It has more detail than VHS SP at roughly the same noise level.

Beta 3 and VHS have the same bandwidth (afaik).

[davideck]

Wilbert - I believe you. What I should have claimed is that the BIII recordings that I have made play back with more detail than the VHS SP recordings that I have made. Probably depends on the particular VCRs used.

Nevertheless, it seems to me that for a given source and bitrate, if the 720 capture looks sharper than the 352 capture, then detail is being lost at 352.

[The_Doman]

Nevertheless, it seems to me that for a given source and bitrate, if the 720 capture looks sharper than the 352 capture, then detail is being lost at 352.

That's just my experience too. 720/704 looks so much better for me, even on VHS sources. (Using PAL with a D8 camera er and a Pinnacle 848 card)

I always go for the better resolution. You really can't define a ANALOG (cable/VHS) source in a fixed digital resolution. Specially with VHS sources there can be a lot of difference in quality depending on the VCR being used.

And about that bitrate of 2700, depending on the source material you can do some nice things with Procoder and a 2passVBR. I often make recordings (tennis) of over 3 hours in full resolution that way.

[Nelson37]

Damn I really get tired of this one.

These comparisons of 352 versus 720 captures, and the sources involved, (surprised no charts or graphs yet) are ignoring the most important aspect. YOU ARE ASSUMING A PERFECT CAPTURE!!!!! If you actually have hardware that does this, skip the rest of this post.

It has been widely discussed on this board, lots of sample pictures, factual info on exact procedures followed by hardware, etc., etc. The simple fact is that almost all capture hardware caps at a fixed res, usually in the vicinity of 640 x 480. When a lower resolution is specified, the card applies internal resizing algorithms, and often various smoothing filters, which result in a LOWER QUALITY VIDEO. Detail is not lost, it is captured and then destroyed.

It doesn't make a damn bit of difference what the res of your source is, other than a lower quality source somewhat "masks" the quality loss.

I am constantly amazed at the way many supposed "experts" continually ignore these facts, which have been repeatedly demonstrated both by simple visual comparison, and by technical info which clearly shows what the true issues are.

I saw my first post on this over two years ago, and ran my own visual tests confirming the problem. Hardware technical reference info was posted over a year ago. People bitched at me for decrying this reliance on theory over observation, I don't care what the charts and graphs say. I determine what is best by actually looking at it, A-B comparisons, with confirmation from others. 352 caps are worse than 720 caps, PERIOD.

BTW, whether 2700 is sufficient bitrate depends entirely on the source, for lo-motion, this could be acceptable.

[Cornucopia]

Yeah, that's one reason I always cap at 720x480 (I have the drive space) and resize in software (Vdub, AVISynth, TMPGEnc, AE, etc) if needed. And I've got great cap boards! (Canopus DVStorm2, AVIDs)

Scott

[lordsmurf]

352 caps are worse than 720 caps, PERIOD.

You destroy your own argument with this ascertion.

[Nelson37]

Well, that pantload of information certainly helps clarify the situation.

Perhaps I should have specified "with my capture hardware", and added that same is true with most, but perhaps not all, capture hardware. Though I already had stated that possible exception earlier, and just assumed that anybody with the brains God gave a walnut could make the necessary extrapolation.

If you have some information which contradicts the known, previously posted in great detail, hardware reference data which specifies the problems capping lower res video, which is common to virtually all cards, then I and probably many others would be interested in seeing it.

[mpack]

Well, I would certainly like to see this issue cleared up. I've read comments similar to these from Lordsmurf before, and while it didn't really register at the time - I found it puzzling when I thought about it later.

First: I'm not sure how one can talk with certainty about the horizontal resolution of analog video. Vertical resolution is of course determined by the number of lines, but horizontal? I freely confess ignorance on this subject - I can think of iron oxide grain size etc, but otherwise I don't know what limits horizontal resolution on a purely analog VHS recording system.

Second: even if some chart says that the effective horizontal resolution on the VHS tape is 352 pixels, that IMHO still implies that you should capture at 720 pixels minimum - ie. at the Nyquist rate, otherwise the signal will be aliased, and the error distributed as noise over the residual lower frequencies. If you claim that you can capture at 352 pixels per line and not lose definition then that would imply that VHS only has an effective horizontal resolution of 176 pixels, which I find hard to believe.

Like I said, I wish some knowledgeable person would clarify...

[BrainStorm69]

Well, that pantload of information certainly helps clarify the situation.

I think what LS is driving at is that you cannot make a blanket statement like "352 caps are worse than 720 caps, PERIOD" and be correct in all cases.

For instance, if you are trying to capture a source with lots of fast motion, and fit it into a given file size, you may not be able to use a high enough bitrate to keep the 720 cap from macroblocking badly, whereas if you had capped at 352, the bitrate you can use for the given file size would be enough to prevent the occurrence of macroblocks. In that case, the 352 cap will look better in the end analysis, even if for the sake of argument, you have lost some detail capping at that resolution.

[Nelson37]

Mpack - reference to #1, analog doesn't really have a resolution, but we can speak of equivalents. Important factor here is that resolution "overkill" on capture does NOT involve a resize operation, it is simply using more sampling points than may be theoretically necessary.
It is generally accepted that VHS "equivalent" resolution is somewhere below 352x480. That's where all the charts and graphs come in.

#2. I've seen this discussed, more charts and graphs again.

The factor that overrides all this concern about source res lies in the nature of how most capture hardware works. They use a FIXED resolution. That's it, done deal. They use a resizing algorithm to arrive at designated res, this algorithm being in virtually all circumstances notably inferior to post-capture resizing. Some cards also apply filtering, and the same inferiority applies.

The observed visual defects caused by these processes vary by card, and by specified resolution. The "trigger point" resolution where they become really objectionable is usually somewhere below 480x480, and almost always above 352x480. That is SPECIFIED resolution, the actual or apparent resolution of the source does not matter.

These hardware specs were well discussed and documented some time ago. I haven't searched for it, but it's there for all to see.

For instance, the actual fixed capture res for ATI cards was either at, or very close to, 688 x 448, which fascinatingly is the exact res you get when specifying a 720x480, cropped, DVD MPG-2 capture. Amazing, ain't it?

[Nelson37]

Brainstorm, I'll concede the point, but I would think it is a given that 720 res requires more bitrate than a 352 cap.

In fact, no I don't concede this. You're invalidatng the 720 cap for a reason outside the principles of the discussion. Since those wishing a 352 solution are using that as the targeted goal, it is obvious that a 720 cap would require a resize to reach that goal. Therefore, a re-encode is necessary rendering capture bitrate irrelevant.

You might as well argue that 352 is better if the cable is unplugged when doing the 720. Or that you really need an 80 x 60 file for Web broadcast.
Or that the filesize must fit on a floppy disk.

OK, restatement - Assuming necessary changes in bitrate are made, and you are testing both on the same source, resulting filesizes fulfill your needs, and best quality is your goal, with virtually all known capture hardware and specifically ATI AIW cards, then 720 is better than 352 for capturing purposes, PERIOD. How's that?

[lordsmurf]

ATI's clock in at about 704-710x486 ... FYI.

[BrainStorm69]

In fact, no I don't concede this. You're invalidatng the 720 cap for a reason outside the principles of the discussion. Since those wishing a 352 solution are using that as the targeted goal, it is obvious that a 720 cap would require a resize to reach that goal. Therefore, a re-encode is necessary rendering capture bitrate irrelevant.

I could have sworn this thread started with a question about 720 vs 352 at the given bitrate of 2700 kbs.

[Nelson37]

Ah, crap, I forgot.

He did, however, note that the 720 cap in fact looked better.

The actual question was what advantages there would be to a 352 cap. Accepting the given of lesser bitrate usage, I still think the disadvantage of hardware-induced degradation outweighs this.

It is the overlooking of this degradation, and continual (incorrect) citation of source resolution as the important factor, that I am arguing against.

[lordsmurf]

Ah, crap, I forgot.

He did, however, note that the 720 cap in fact looked better.

The actual question was what advantages there would be to a 352 cap. Accepting the given of lesser bitrate usage, I still think the disadvantage of hardware-induced degradation outweighs this.

It is the overlooking of this degradation, and continual (incorrect) citation of source resolution as the important factor, that I am arguing against.

Huh? A bit too much rhetoric, hard to understand what you're saying.

[FulciLives]

This thread has gotten very "messy" so let us get back to the basics.

A cap of 720 width is almost always going to look better in terms of sharpness and detail than one done at 352 width.

You could argue all day that the source maybe doesn't have enough resolution for 720 width to look better but at the end of the day it will. End of story.

So issue ONE is the resolution in terms of SHARPNESS and DETAIL of the image.

Issue TWO is the MPEG compression.

The benefit of 352 width over 720 width is that you will get much less MPEG artifacts (such as blocking and mosquitoe noise etc.) when using a low bitrate. I am sure we can all agree that a video bitrate of 2700kbps is a low bitrate for a width of 720

So in this particular situation (2700kbps video bitrate) I cannot imagine for the life of me that the MPEG artifacts in the 720 width capture are worth the extra sharpness or detail of the image. In this case I would rather have the ever-so-slightly softer look of the 352 width capture which would have MUCH less MPEG artifacts.

- John "FulciLives" Coleman

[lordsmurf]

Let me be brief and frank...
to expand on what Fulci said...

More pixels needs more bitrate. That easy.

A 352x480 cap at 3500 will look equal to
a 704x480 cap at 7000

Easy as that. Simple, huh?

2700 is already low for 352x480
it will be unbearable and full of blocks at 720x480
5400, on the other hand, would be more acceptable

While 720 is higher res, and thus sharper, it only works out that way if the source was that sharp too. VHS isn't very sharp. Unless you buy an image enhancer (like the Vidicraft Detailer or SignVideo DR-1000), that's not going to change. There is no magic in the capture card.

On the other hand, film in a studio is high res.

That's why people at home use 352x480 and commercial releases are 720x480. Stick to your source, and you'll be fine.

The only exception is probably half or more of the video cards out there suck. They cannot accurately capture, and most of them look better at higher res. This is why buying good hardware and software is of utmost importance.

There are some bitrate graphs halfway down this page:
http://www.digitalfaq.com/capture/avivsmpeg.htm

[vhelp]

I think everyone has good points.

But, I believe that it's all about *sample* rates and how much detail it
returns back (after the cap/hardware induces it's own level of algorithems
and things, aka 'taintedness') ...
.
And, the greater the resolution (max the given cap card was destined/set
to) the better the results.

Course, now that given the above, resizing is another deal all together.
.
Given a good *sampling* of a capture (assuming the disccusion is w/
respect to VHS sources) the recommneded (mine at least) for VHS sources
to capture in (bare in mind the cap cards limits) the resolution should
always be "highest" your card or driver will allow. That means, if your
card can do 720 pixels, then set it as such. But, if your card does a
little "funny jimmy", then you have to employ trickery, such as caping
at odd sizes to get to the same goal. What ever that is. I'm saying
here, that capture cards have a funny way of yielding final results. And
we can all say, that its dependant upon the card and its feature (or
limits)
.
If your capture card does a good job at *sampling* at a resolution
setting of 720 pixels, BUT that your end goal is for another
product (and for other pobably obvious reasons) 352 x 480 res, then
your chances are, that you'll get better results if you HAD capped
at 720 vs. 352 res/pixels (or whatever)

Sampling ...

I beleive that the secret (and misunderstanding) w/ respect to resolution
is the *missing link* .. and that link is called sampling

For instance, when we capture something from a tv source (actaully, any
source really) we are not capturing resolution. The device is taking
samples and incprorating them (however the machnics/ciruitry is done)
and applying them to this timeline (resoution of dots/pixels) within
a container or two-demenion plain, and resulting in what we see (or
perceive as resolution) oh my
.
I've ben plauged by this resolution theory for time ends, when talking
specially with VHS sources. But this applies to everything that we
capture - videowise. we are not capturing a resoution. WE ARE actually
sampling dots or 1's and 0's and the machincs of things (inside the
device) is puting each 1's and 0's in their respective places. that
mean, all the LUMA and CHROMA and what ever else there is to *fill*
this timeline (or container) with.
.
so, when we talk about resolution, think of it, in terms of containers
hold dots.. or drops of pee-pee
.
Then, w/respect to resolution..
* the better the sample circuitry the better the results
* the greater the resolution (for the samples of pee-pee) the greater
.. the detail (or smell)

When your capture source is VHS, the same is applying. Or shoud I say,
when you capture a TV source the same applies (w/ respect to sampling)

Some cards do a better job a sample the source. Others' do a better
job at interpolating the data. Others do a better job at... etc.
It all depends on the card and it's features and abilities (or limits)

Do you recall those newer cards that are 10bit (vs. 8bit) samplers ??
That's another hint to you.
.
It's *sampling* the detail (or color information) but at 10bit
units vs. 8bit units. There is obvisiouly going to be better results
w/ a 10bit sampler than an 8bit.
.
But, this sampling, takes a certain container of color infomation
and samples out of the millions of collors (or given pallete) and
sample at a *greater resolution* (demention) of colors and picks
out those that are closest in detail of the original source's.

Now, on the speaking of 352 vs 720 (w/ respect to MPEG and Bitrates) ...

I agree w/ fulci and LS on this. However..

No one is wrong here. Think about it for a moment.

* 352x480 cap at 3500 will look equal to
* 704x480 cap at 7000

This is true. However, it's only theory, because as we all know..,
and based off of experience, that its the encoder that dictates this
theory to be true or false. What I mean, is given the above, will
the encoder in use, fullfill the theory above 100% ..and the anser is
yes/no
.
It all depends on the encoders engine.. or algorithem and how it
utilizes bitrates; modes; and thechniques in its encoding process
and final MPEG quality/detail (if you're measureing fast scenes as
a final guage)

There is nothing wrong w/ finalizing projects to 352 x 480 res.
(after considering the above)
.
But do bare in mind, that today is DVD. And Todays' tvs are getting
bigger and better, and will require better/greater details from
our louzy processed MPEGs (our captures/encodes)
No matter how good your source 352 is, and into these tvs of today (and
tomorrow) chances are, you'll be back here looking for resolutions
(I mean, answers) to improved details from these sources.

But, I think the argument is w/ respect to 352 vs. 720 AND the bitrates
used (as posted'er above started on)
.
I think that the answer should have ben better responded to w/
repsect to Methods of Encoding from w/ Bitrates and formats from
these two resolutions. It wasn't about capturing resolution (at least
I don't think so) It was about Encoding Technques for either res

Cheers,
-vhelp

[bunyip]

Well I do all my TV & VHS captures at 352x576 (PAL) because in the time it takes to render a single pass of 720x576 I can have 2 passes of 352x576 completed, fully authored and burn't to DVD. And playing it back on an analogue TV there is no visual difference, so why bother with Full D1 sizes. :P

[Cornucopia]

I pretty much agree with what you said, but to clarify:

* 352x480 cap at 3500 will look equal to
* 704x480 cap at 7000

352x480 cap at 3500 with exhibit (roughly) the same amount of artifacting as
704x480 cap at 7000

Resolution would still be higher with 704.

So at same bitrate (~2700), example of "Artifacts = K * Resolution / Bitrate" and "Quality = Resolution / Artifacts" so e.g. 352/200artifacts = 704/400artifacts.

"K" is determined by the efficiency of the filter/resizer/encoder.

Scott

[vhelp]

oops.. I applogies.

I was capturing (of all things) cut/pasting the data
and didn't revise it.

352x480
.. vs..
704x480

I ment that I agree to an extent on the two w/ respect
to bitrate, etc.

[johns0]

Trying encoding a football game at 720x480 at 2700kbps they you will see the flies following the football players in fast action scenes.