On Mon, 7 Feb 2005, Ian Trider wrote:
I did all my tests with a static luminance test pattern (B&W). It does not take into account motion or the digital scaling that must occur (since the actual sampling dotclock of the bt878 is fixed at 4*fsc IIRC.). I'm not sure how the driver sets up the digital antialiasing filters when scaling to small sizes (like 352x480), but it might artifically cut out even more data to prevent aliasing. There's also the perceived resolution difference between continuous analog and sampled (Kell factor), which I'm not super-familiar with and sounds like a "fudge factor."BT878-based cards to not have a comb filter and rely on a notch filter to separate luma and chroma. That means that for any composite input signal (NOT S-vid), they're limited to about 3MHz of luminance resolution. That translates into 352x480 capture. Anything over that is wasted space. Again, if used for s-vid, that is not the case.
I'd hate to disagree, since you clearly know what you are talking about, but I must. Either the theoretical numbers fail to match the practical numbers or some cards suck less than others because I've done some tests myself on this before and my bt878 card (ATI TV Wonder VE) tops out above that (on composite). I haven't done any extensive checking, but there is a difference between captures at 352x480 and 640x480. I'm sure the resolution tops out somewhere in between.
Anyway, certainly the default MythTV setting of 480x480 is perfectly reasonable (and perhaps advisable) under most circumstances.
Now that I look at the datasheets more, however, I see that the biggest difference is the other side of the notch filter. The bt8xx chip uses a big fat notch filter at the color subcarrier (3.58MHz for NTSC). That's what kills the luminance channel at 3MHz. The *other* side of the notch filter starts coming back at about 4Mhz. Some of that would be sent through at capture resolutions higher than 480x480. Normally, there's not much information there, except rainbow swirls on pinstripped B&W ties.... :)
I'm not arguing with your results, I'm just saying that from the datasheet, most luminance information is lost above 3MHz on the bt8xx chips. Capture cards that have a comb filter can better separate Y/C without resorting to the brutal notch.
-Cory
************************************************************************* * Cory Papenfuss * * Electrical Engineering candidate Ph.D. graduate student * * Virginia Polytechnic Institute and State University * *************************************************************************
_______________________________________________ mythtv-users mailing list [email protected] http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
