Amazing discovery in capture resolutions (IMHO)

[trevlac]

P.S.
So when I capture at 368x480 should I resize that to 352x480 to get a proper aspect ratio?

Full cap size * 1/2 because it is 1/2 D1, then crop/pad to reach 352.

Aver: 688/2 = 344 pad to 352

BTwincap: 712 / 2 = 356 then crop to 352.

And as I recall, the picture is shifted on the Btwincap 2 pixels to the right. On the aver 1 to the left. I'd guess to 'fix' the Btwincap at 368, you'd crop 1 from the left and 3 from the right (since it's 1/2 size). Anyway, this is a bit anal because a TV picture is probably not centered anyways. PS: The vertical is probably also shifted a few pixels.

Personally, I cap at 368 and crop to 352 not worrying about the aspect. But it's nice to know how it works

[SatStorm]

Hmmm....
I test a lot today. And I can't conclude on something...

my kworld card has exactly those numbers as the example of the link trevlac provide. I have virtually the same card and drivers with the test of that guy!

But, I can't find the best framesize to capture with PAL....
On my eyes, the best picture I have is 702 x 576, but again, I'm not sure...

[FulciLives]

P.S.
So when I capture at 368x480 should I resize that to 352x480 to get a proper aspect ratio?

Full cap size * 1/2 because it is 1/2 D1, then crop/pad to reach 352.

Aver: 688/2 = 344 pad to 352

BTwincap: 712 / 2 = 356 then crop to 352.

And as I recall, the picture is shifted on the Btwincap 2 pixels to the right. On the aver 1 to the left. I'd guess to 'fix' the Btwincap at 368, you'd crop 1 from the left and 3 from the right (since it's 1/2 size). Anyway, this is a bit anal because a TV picture is probably not centered anyways. PS: The vertical is probably also shifted a few pixels.

Personally, I cap at 368 and crop to 352 not worrying about the aspect. But it's nice to know how it works

I'm not sure I follow the math here.

I'm now using the BTwincap driver and I have confirmed that 712x480 capture padded to 720x480 matches the original although there is that slight shift I agree it really isn't worth worrying about.

However ...

If I want to capture at 712x480 and resize to Half D1 wouldn't it be more "proper" to first pad the 712x480 to 720x480 then resize that to 352x480 as in this AviSynth script:

LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec3.dll")
avisource("D:\capture.avi")
AddBorders(4,0,4,0)
LanczosResize(352,480)

Here I am taking a capture that is 712x480 and adding 4 pixels of black to each side to get 720x480 then resizing that to 352x480

That makes more sense to me than resizing 712x480 to 356x480 then cropping to 352x480

However I am more concerned with capturing at 352x480

As this thread indicates ... capturing with the BTwincap driver at 352x480 creates a slightly soft picture as opposed ot capturing at 368x480

So I want to capture at 368x480 right?

So NOW how do you get that to 352x480 and keep the aspect ratio?

I can't imagine that just cropping 8 off each side to reach 352x480 would keep the aspect ratio.

I'm not sure (without testing) if simply resizing 368x480 to 352x480 is also proper.

My theory (which I guess I will have to test) is that you will need to resize 368x480 to 348x480 then add 2 pixels of black on either side to get up to 352x480

This can be done like this in AviSynth:

LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec3.dll")
avisource("D:\capture.avi")
LanczosResize(348,480)
AddBorders(2,0,2,0)

Does this make sense to anyone? Trevlac?

- John "FulciLives" Coleman

[FulciLives]

Hmmm....
I test a lot today. And I can't conclude on something...

my kworld card has exactly those numbers as the example of the link trevlac provide. I have virtually the same card and drivers with the test of that guy!

But, I can't find the best framesize to capture with PAL....
On my eyes, the best picture I have is 702 x 576, but again, I'm not sure...

Do what I did ...

1.) RIP a DVD to the HDD then load it into VirtualDub (I used DVD2AVI to create a D2V project file then loaded that into VirtualDub with a simple 2 line AviSynth script).

2.) CAPTURE the DVD using various resolutions and compare a screen grab to that of the original ripped DVD to see what matches up.

In my situation ... changing to the BTwincap driver ... a capture at 712x480 padded to 720x480 exactly matched the original aspect ratio wise except the image was every-so-slightly shifted to the right and slightly higher but nothing to really worry about.

I thought I read that it is best to capture PAL at 768x576 then resize to 720x576 and if that does work then I would think 768x576 would resize just as well to 352x576 but maybe Trevlac can comment on this.

- John "FulciLives" Coleman

[SatStorm]

I don't wish to capture that high... I mean, I don't have the HD space for such captures....
Seems like I gonna end using this card only for PAL60 captures....

Meanwhile I'm testing parallel mainconcept's 1.4.1 realtime capture tool. What an amazing job it does... 352 x 576 with 1000min/2700average/5000max has no blocks. Really excellent picture! Amazing! Makes programs like power VCR and winDV3 look really bad... But this is a subject for other topic.

I'm open for capture suggestions, from those who know best all those maths... For PAL and a BT8x8 card, which framesize seems better to capture?

[vhelp]

I would go with Fulci on this one. Capture at 712 (if his or your card dos
a better job at that resolution) and PAD 4 pixels left/right side each,
for a perfect 720.

By doing this (assuming the filter does not taint w/ sharpening etc) is that
the AR does not get toutched, since you did not resize (aka, distort) the
picture. But, if you resize, using a filter that has other tainting such as
sharpening etc, you are risking distorting the AR to some degree, if not,
your source quality. I don't mind using .AVS scrits, and do so to a limited
degree. I use vdub because I know the resising filter and how to minipulate
to some degree, my source. That's not to say that you all don't know this
either, but I'm saying that I have more experience and dealings w/ vdub
and it's resizing filter to an advantage point in my many projects :P

I do understand AR to a degree, now that I've learned more about it and it's
intricit issues. So my end goals are alwasy to factor AR w/in my projects,
especially if they are from a capture nature.

Trevlac is correct, but he didn't factor (or realize) that not only does or
can tv be missleading or off a few pixels, but that you also have to factor
in the studios or broadcasters decisions to:

* jimmy the source
* alter the source
* use other providers; stations; broadcasters etc source, and then add in their
...own flavor of jimmy/alterations to the source.. ..
.
.
Then, it's up to our devices that bring us the tv signature/picture (ie, cable
box; satellite receiver; capture card's tuner; etc) of which may have its own
set of issues; settings limits etc.. and,
.
.
Then finally, to our capture cards/devices. And, then finally, again, it's up
to our:

* capture card and it's hardware features,
* and driver, and then finally,
* the software app that does the capturing (via cap card driver specs etc)
.
.
Then, its finally up to our abilities/skills to edit (wisely) the brought in
source for a final end result. All this, and hopefully no tainting of the
AR - - well, most unlikely, but possible.
.
.
TV is airing it one way, out capture card is bringing it inside our pc in yet
another way (hence this thread and doom9's) and we're still figuring out which
is the best method. I'm pretty at all this fun stuff, but the only
reason why I continue doing so, is that it's interesting and also my hobby :P
So, lets continue the research.

Oh, yes.. as I was saying..
I would find that capping at 712 and padding to 720, then resizing to 352 would
be the most effective way of maintaining quality.
However, my ATI-TV Wonder card seems to do it's best at 640 x 480, but there
is no smooth way of approaching either 352 or 720 So, in my case (and if I
were to use my ATW) I would just capture at 720. However, if I were to cap
at 704, I would resort to 704 and pad 8 pixels on each side for 720.

I do have other cads to test maximum resolutions with. For instance, my
Osprey-210 is a good analog card. I like it's output very much, and is worthy
of some resolution testing - this process yet to be determined.

-vhelp

[SatStorm]

OK, here are my results without the aspect ratio issue (which isn't a big deal for me)

With the Kworld card, my overall best results was 702 x 576, expand to 704 x 576, resized to 352 x 576 using Lanczos. Also, 368 x 576 > 344 x 576 >expand 352 x 576 look about the same and has a better aspect ratio than the direct capture to 352 x 576.
368 x 576 cropped to 352 x 576 or resized with the "resizer" trevlac provided me (thanks trevlac for this) also gives excellent results. But for PAL tests I need more opinions, I don't have time to test it further, and I'm sure there must be other framesizes for PAL with better results!
Capturing PAL60 (which is nothing more than NTSC covered), is exactly like Trevlac results: 368 x 480 cropping to 352 x 480 is the best results!

With my Hauppauge win TV primio FM, things are more straight: Capture 704x576, resize 352 x 576 is the best. Capturing 368x576, resize 344x576 , expand 352 x 576 is sharper than capturing direct 352 x 576 (slightly...). I didn't run PAL60 tests with this card...

With my Nvidia Asus 7700 card, all those framesizes are bullshit (sorry...) and capturing direct to 352 x 576 gives me an excellent picture (far better any hauppauge or kworld capture). Unfortunatelly, no PAL60 capture with this card at all (it is possible only with w98SE, with old vfm drivers, plenty lipsynch issues, frame drops, etc...). Capturing to 704 x 576 is amazing, overkill for VHS of course...

Conclusion: I gonna use that cheap kworld card only for my rare PAL60 captures from NTSC VHS tapes. Also, I hope for better drivers in the future!

[ronnylov]

..........
I'm not sure I follow the math here.

I'm now using the BTwincap driver and I have confirmed that 712x480 capture padded to 720x480 matches the original although there is that slight shift I agree it really isn't worth worrying about.

However ...

If I want to capture at 712x480 and resize to Half D1 wouldn't it be more "proper" to first pad the 712x480 to 720x480 then resize that to 352x480 as in this AviSynth script:

LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec3.dll")
avisource("D:\capture.avi")
AddBorders(4,0,4,0)
LanczosResize(352,480)

Here I am taking a capture that is 712x480 and adding 4 pixels of black to each side to get 720x480 then resizing that to 352x480

That makes more sense to me than resizing 712x480 to 356x480 then cropping to 352x480
.....
Does this make sense to anyone? Trevlac?

- John "FulciLives" Coleman

Would it not be more correct to capture 712x480 and crop to 704x480 and resize to 352x480? Remember that 352 is not half of 720, it is half of 704.

This can be done in avisynth:
avisource("D:\capture.avi")
Crop(4,0,704,480)
LanczosResize(352,480)

[trevlac]

Would it not be more correct to capture 712x480 and crop to 704x480 and resize to 352x480? Remember that 352 is not half of 720, it is half of 704.

Bingo!

You could also resize to 356 and crop to 352. Same result. Point is that 352 is 1/2 of 704 not 1/2 of 720.

[FulciLives]

Would it not be more correct to capture 712x480 and crop to 704x480 and resize to 352x480? Remember that 352 is not half of 720, it is half of 704.

Bingo!

You could also resize to 356 and crop to 352. Same result. Point is that 352 is 1/2 of 704 not 1/2 of 720.

I guess the question is this then ... how does a stand alone DVD player resize 352x480?

Does it resize it to 720x480 or 704x480?

Also how does a 704x480 DVD (I know the Panny stand alone DVD recorders use this resolution) display on a TV vs a DVD with a 720x480 resolution. Understanding this first is key to understanding I think how the 352x480 gets resized ... or at least how it will look.

- John "FulciLives" Coleman

[lordsmurf]

Would it not be more correct to capture 712x480 and crop to 704x480 and resize to 352x480? Remember that 352 is not half of 720, it is half of 704.

Bingo!

You could also resize to 356 and crop to 352. Same result. Point is that 352 is 1/2 of 704 not 1/2 of 720.

I guess the question is this then ... how does a stand alone DVD player resize 352x480?

Does it resize it to 720x480 or 704x480?

Also how does a 704x480 DVD (I know the Panny stand alone DVD recorders use this resolution) display on a TV vs a DVD with a 720x480 resolution. Understanding this first is key to understanding I think how the 352x480 gets resized ... or at least how it will look.

- John "FulciLives" Coleman

You're where I am. This is one of my current thigns to find out here too.

[ronnylov]

The 704x480 and 720x480 DVD resolutions both have the same pixel aspect ratio. This mean that you can crop 720 width to 704 width and it will still look the same on TV at playback. At least in theory. Maybe some DVD players does not do this sizing correctly at playback (and gives a difference between 704 and 720 width)?

I think you may find this link interesting:
http://www.iki.fi/znark/video/conversion/#conversion_table

The table say that the pixel aspect ratio of 352x480 DVD resolution is 144/79. The pixel aspect ratio of both 720x480 and 704x480 is 72/79.

To convert between 352x480 to 720x480 you need to change the pixel aspect ratio from 144/79 to 72/79. That would mean twice as many pixels horisontally, but then you get 704x480 resolution, so you need to pad pixels to make it 720x480 and still have the same pixel aspect ratio. And I have already explained how to resize from 720x480 to 352x480, first crop to 704x480 and then resize to half horisontal resolution.

I do almost never create DVD with 720 pixels width because 704 pixels width will look exactly the same on the TV and there are less pixels to encode so I'll save a tiny amount of bitrate. The extra pixels from 704 to 720 is hidden in the overscan area of my TV so why encode it? I live in a PAL country so my actual picture area is 702x576 at both 720x576 resolution and 704x576 resolution but I can understand if some of you NTSC people want to encode your full 711 pixels width and then use 720 pixels width to avoid cropping instead of 704 when encoding to DVD format.

[ronnylov]

Would it not be more correct to capture 712x480 and crop to 704x480 and resize to 352x480? Remember that 352 is not half of 720, it is half of 704.

Bingo!

You could also resize to 356 and crop to 352. Same result. Point is that 352 is 1/2 of 704 not 1/2 of 720.

Yes, but I beleive cropping first and resizing after would be a little bit faster because the resizer get less pixels to work with. But maybe the cropped edges would look better if resizing first and cropping after? I'm not sure which method is best.

[trevlac]

Yes, but I beleive cropping first and resizing after would be a little bit faster because the resizer get less pixels to work with. But maybe the cropped edges would look better if resizing first and cropping after? I'm not sure which method is best.

You know your stuff...

Crop 1st is probably faster because there are less pixels to filter. However, a problem with any resize filter is that it is based upon an infinately wide picture (signal). Because in practice they are not, the edge pixels are filtered (averaged) with unknown pixels that are off of the edge. An implementation probably pads with grey pixels.

I don't think you'd notice the speed increase or the filter edge problem if you watch on a TV. BTW: black borders cause the same filter edge problem.

I too am not sure which is the 'best' method.

[trevlac]

Also how does a 704x480 DVD ... display on a TV vs a DVD with a 720x480 resolution. Understanding this first is key to understanding I think how the 352x480 gets resized ... or at least how it will look.

Here is my understanding. You have to think in terms of what an analog TV does. That is, it has a beam that takes a (1 dimensional) signal and paints it back and forth across the screen creating a 2 d image. This signal contains a picture portion and other info so the TV knows when a line ends and when an entire frame ends. That way it can start on the next line or at the top of the screen again.

ITU 601 spec says the digital active picture portion of one scan line is 53.333 microseconds long. This number equates to 720 pixels. You may have heard of the sample rate of 13.5MHz. Well, if you sample a 53.333 microsecond line at the rate of 13.5 samples per microsecond, you get 53.333 * 13.5 = 720 samples.

For NTSC tv, the spec says the active picture is 52.666 microseconds. This is about 711 pixels at a 13.5 sample rate.

So, when a DVD player encounters a 704 or 720 dvd, it must convert to the analog standard. Well, I don't really know, but it makes sense that they just start creating the entire TV line; for 720 they layout the 53.333 signal and overlay the edges with sync stuff to get a real 711 width. For 704 they layout a 52.148 signal and pad the edges.

For 352 I'd guess they work at a 6.75MHz rate (13.5/2=6.75). For svcd I'd say they use 9MHz (13.5 * 2/3 = 9MHz). That is why these are 1/2 D1 and 2/3 D1.

Hope this helps.

[SatStorm]

I did more tests today....

On the device settings, there is an option: It is called "Video Proc Amp" on Virtualvcr and it is also present in virtualdub on Video source > device settings.

This has Brightness, Contrast, Hue, Saturation, white balance, Gamma and Backlight comp adjustments, also one called "sharpness".
By default, this option is set on "50" from a possible of 81. Also, there is a tick there, "auto", which is enable.

When you first visit this option, only few of them can be tweaked. But if you untick the "auto" near to "sharpness", then this became programable.

This is possible on my bt8x8 cards, but not possible on my nvidia one.

When I set this "sharpness" to 81, the difference between 368 x 576 and 352 x 576, simply disapears!
So, what I conclude, is that this setting, 368 x 576, is actually a bug of the driver: It overides this "sharpness" built in filter of the cheap, and set it to the most possible value!
Same story at 702 x 576 vs 704 x 576, which I reported before that looked better. No: If i set the sharpness to the full value (81), there is no difference between 702x576 and 704x576 on this matter...

Can someone test this on NTSC?

For me now, it is back to old good 352 x 576. Strange through: If I didn't read this tread, I would never discover this setting and the use of it!

[lordsmurf]

So is our final finding on this driver and card-type showing that it artificially sharpens the image? Filters are applied regardless of the desire of the user?

If so, that sucks.

[fmctm1sw]

So is our final finding on this driver and card-type showing that it artificially sharpens the image? Filters are applied regardless of the desire of the user?

If so, that sucks.

I guess so. I've started capturing at 712x480 and cropping off 4 pixels each side. Then I use avisynth's horizontalreduceby2() to get my 352x480. I guess I'm keeping the aspect ratio close that way...

[BrainStorm69]

So is our final finding on this driver and card-type showing that it artificially sharpens the image? Filters are applied regardless of the desire of the user?

If so, that sucks.

I believe you can disable the sharpness by unchecking the box next to the setting in VirtualVCR, although I'm not positive about this. I haven't really done any testing to confirm.

Also, for those of you that are capturing at 712 and cropping, have you tried cropping 5 pixels from the left and 3 from the right to shift the image back to the left? Seems to work for me.

[fmctm1sw]

I did more tests today....

On the device settings, there is an option: It is called "Video Proc Amp" on Virtualvcr and it is also present in virtualdub on Video source > device settings.

This has Brightness, Contrast, Hue, Saturation, white balance, Gamma and Backlight comp adjustments, also one called "sharpness".
By default, this option is set on "50" from a possible of 81. Also, there is a tick there, "auto", which is enable.

When you first visit this option, only few of them can be tweaked. But if you untick the "auto" near to "sharpness", then this became programable.

This is possible on my bt8x8 cards, but not possible on my nvidia one.

When I set this "sharpness" to 81, the difference between 368 x 576 and 352 x 576, simply disapears!
So, what I conclude, is that this setting, 368 x 576, is actually a bug of the driver: It overides this "sharpness" built in filter of the cheap, and set it to the most possible value!
Same story at 702 x 576 vs 704 x 576, which I reported before that looked better. No: If i set the sharpness to the full value (81), there is no difference between 702x576 and 704x576 on this matter...

Can someone test this on NTSC?

For me now, it is back to old good 352 x 576. Strange through: If I didn't read this tread, I would never discover this setting and the use of it!

I never thought of that... I'll look at it as soon as my 3 Stooges are done encoding.....

[SatStorm]

Well, what it seems...

Those third party drivers we are all using, are universal ones.
The bt8xx family, has variations. Differences, too hardware-like to understand with my minimal english skills. (Don't mention that I don't have time to read and learn all those stuff...)
So, what I guess, is that by default, the universal drivers are "optimised" for the older, least capable hardware, for compatibility reasons.
But this is not fair for later hardware, so the authors of those drivers, implement an "auto" fuction, who try to guess the sharpness abilities of any chip. Well, it seems that this "auto" fuction, don't work so well and we didn't know about it. Maybe on older drivers, it worked and now is broken. Who knows...
Now: What trevlac objervate, was that when he used non standard framesizes to capture, calculated and based on the maths, it has better picture. Doing the tests the way he show us, this is true. But why?
I don't believe it is the maths only, for 2 reasons:
1. I'm PAL and when I used those ideas for PAL I also had some wierd results. But the PAL maths are not the same with the NTSC maths. Why this is possible? Something else here...
2. Why using the same drivers on 2 different bt8xx cards, I have different results? I mean, there is not only the results between 352 vs 368, but also 368 from Hauppauge vs 368 from Kworld.
So, I was sure from the start that it is an adjustment, a driver issue which cause all this! But which one?

Also, is that factor "X" of the bt8xx cards, which make it harder to conclute on something while testing. That those cards don't capture boarders, if the boarders are detected (who knows how...). I'll explain:
All bt8xx cards, with the universal 3rd party tweak drivers we are all using, somehow don't capture overscan boarders, so the result is somehow resized always. Doing the maths as trevlac provided, we can calculate (detect) and use some specific framesizes, which don't do any kind of resizing while capturing. So, we don't distort the picture that way and since no built in chip "resizing" is done, the results look better (expecially on lines, shapes, etc. Cartoons is an excellent example).

And here is the accident which reveal us this "bug" of the drivers (I name it bug, because this is my opinion on this, I might be wrong...)
It seems that when we use those specific framesizes, the chip overides the "default" settings of the drivers, about sharpness. So, a bt8xx card shows 100% of the sharpness it is capable off

It is a silly bug if you ask me...

So, when we use those framesizes provided by the objervation of Trevlac (and others) , two things happen at the same and we have better picture:
1. The picture is not resized by the built-in resizer of the chip (or the bult-in resizer of the drivers)
2. The sharpness is set automatic to the highest possible value.

That also explains why I have different results on my 2 Bt8xx cards (Hauppauge win TV primio FM (older) and Kword RF Pro (new). They are capable for different amount of sharpness (or it has to do with those 8bit, 9bit and 10bit. Older bt8xx did8bit only, new ones do much better...).

Also, it seems that if you manual set the sharpness filter to a value higher that the default (50 of the possible 81), on the older bt8xx cards it sharps artificially from a point and beyond. On the cards with the newer bt8xx chips, this point maybe simply don't exist and the 81 value is the default value for those cards..

Of course, some testing is needed. Expecially from you, the NTSC people which you affected by this issue much more than me, the PAL user.

When I set sharpness to full value (81) the only minus I have when I capture 352 x 576 compared to 368x576, is the built in resizing. 352 x 576 is a result of resizing from who knows which framesize (maths are needed to discover the true here, I'm open in suggestions), while with 368 x 576 no resizing is made at all. The difference of course is unoticable on a final mpeg 2 @ 352 x 576 (IMHO), but more opinions are needed on this

[fmctm1sw]

Damn SatStorm.. Just when I though I had this all figured out...

I did a screen cap of 352x480 with a sharpness of 49 and 50. Only a one click difference. Here's what happened:



And this is 368x480 at sharpness 49 and then 50....

[SatStorm]

Yeap, exactly, you confirm it...

Just compare 352 x 480 sharpen at 51 with 368 x 480 on auto....
About (if not...) the same picture.
Set now the sharpen filter to the highest possible value (81) and prepare for great suprises!

[Northstar]

Just a small question about this sharpening issue.

As you said "This is possible on my bt8x8 cards, but not possible on my nvidia one"

Its the same with my Winfast, the option to set this sharpness is ghosted out. What i tried was to set this sharpness manualy, in the vcr option file, i turned it upto 81 and saved the file, i gave it a quick try to see if the picture looked any different, havent done any testing on it other than a few minutes, i wonder if setting it manualy has any effect as i wasnt sure i could tell a difference, i thought i could...

Maybe someone out there with the ghosted out sharpness option would like to try it this way, see if you can see any difference doing it manualy, i'd sure like to know, maybe you could try it this way with your nvidia card SatStorm.

Looking at fmctm1sw pictures, there is quite a difference in quality, those are much sharper.

Cheers.

[FulciLives]

Well here is my test.

I captured the THX DVD of ARMY OF DARKNESS a total of 3 times. Each time at 352x480 but changed the sharpness from 49 then 50 then 81

Loaded the capture into VirtualDubMod and pasted into Adobe PhotoDeluxe and saved as JPG at quality 8 on a scale of 10

Original capture done with PICVideo MJPEG on the 19 quality setting.

First up the UNIVERSAL logo at the start of the film:

352x480 with sharpness set to 49


352x480 with sharpness set to 50


352x480 with sharpness set to 81


Here is a close up shot of ASH holding the BOOK OF THE DEAD:

352x480 with sharpness set to 49


352x480 with sharpness set to 50


352x480 with sharpness set to 81


There is definately a HUGE difference between the soft looking 49 sharpness setting and the MUCH SHARPER 50 sharpness setting.

The 81 sharpness setting seems to be too high because there is some color distortion ... sort of like a moire pattern effect.

Ironically the 50 setting is the default so I'm not sure what the whole 368x480 thing was. I mean a 352x480 capture defaults to 50 so why would it ever be too soft?

This is using the BTwincap driver.

When using the BTwincap driver I get the follow as DEFAULT values:

BRIGHTNESS = 750
CONTRACT = 100
HUE = 0
SATURATION = 100
SHARPNESS = 50
GAMMA = 140
WHITE BALANCE = 0
BACKLIGHT COMP = 1

I would assume that anyone using the BTwincap drivers would have the same default values?

- John "FulciLives" Coleman

P.S.
All pictures have been posted. This post is finished

[BrainStorm69]

When using the BTwincap driver I get the follow as DEFAULT values:

BRIGHTNESS = 750
CONTRACT = 100
HUE = 0
SATURATION = 100
SHARPNESS = 50
GAMMA = 140
WHITE BALANCE = 0
BACKLIGHT COMP = 1

I would assume that anyone using the BTwincap drivers would have the same default values?

Those are the default values for me in VirtualVCR.

[FulciLives]

When using the BTwincap driver I get the follow as DEFAULT values:

BRIGHTNESS = 750
CONTRACT = 100
HUE = 0
SATURATION = 100
SHARPNESS = 50
GAMMA = 140
WHITE BALANCE = 0
BACKLIGHT COMP = 1

I would assume that anyone using the BTwincap drivers would have the same default values?

Those are the default values for me in VirtualVCR.

Seems to make for a slightly dark capture though. Also the color seems a bit muted. That's easy enough I suppose to fix with SATURATION but I'm not sure what combo of BRIGHTNESS/CONTRAST is best.

Has anyone else noticed the slight darkness and lack of color?

- John "FulciLives" Coleman

[SatStorm]

No, it wasn't the default setting!!!!

The bar was on 50 BUT the "auto" at the right was ticked. So, it was on the auto mode, NOT on the 50 value!!!

When you unticked and make this bar programable, ONLY THEN the manual sharpness default value activates!

What 368 x 576 do? Overides the "auto" fuction and because of a driver bug the default sharpness activates to 50 (the default value in your case).
On my Hauppauge, bt878 the value is also between 50 - 56. On my Kworld with bt878a this value is much higher. I think it is 65 or something, maybe I gonna test more next weekend.
IMHO, each bt8xx based card, has a different sharpness default value. Of course this needs testing!

One interest thing in case we wish to have the best possible aspect:
Grabb at 344 x 576/480 and expand to 352 x 576/480. With the sharpen filter activated manually, the picture is much better that way and with the correct aspect!

[fmctm1sw]

One interest thing in case we wish to have the best possible aspect:
Grabb at 344 x 576/480 and expand to 352 x 576/480. With the sharpen filter activated manually, the picture is much better that way and with the correct aspect!

Achieving a good level of sharpness (not oversharp), keeping the aspect ratio correct, and keeping the filesize down. That's the 3 things I am looking for now. Looks like more weekend testing....

[trevlac]

I havn't been following.

But I posted on 1/19 what the chip specs say and what the code does with regard to sharpness "peaking" settings.

It looks to me that the code works like this:

(Sharpness value - 18 ) / 16 & 3

For this anything above 65 gives 3
65-50 gives 2
49-34 gives 1
<34 gives 0

1 = Max full resolution peaking
2 = Min Cif resolution peaking
3 = Max Cif resolution peaking
0 = Min full resolution peaking

The only way NOT to use the filter is to check the AUTO box and use a frame size > 360.


--------------------
As you continue to test, I'd use the links I pointed out before and the spec doc. You can read the registers directly and look to see what the spec says is happening.

If you don't like that, you can set some registers directly using BT8x8 tweeker which comes with virtual dub. Using it, you may be able to cap at 352 and turn off the vertical filter that normally messes with the resolution. However, i'm not sure you can use it for timed caps.

Guessing what the driver is doing is just guessing. You may have it correct, but the story seems to be a bit convoluted.

If anyone wants to know more about how to read the spec or how to read the registers on a BT878 chip, email me trevlac (at) trevlac.us.

--------------------------

This does seem like it's been great fun!