Originally Posted by blue_z
A nit, but it's not downsampling
but rather downscaling
. There's a big difference between sampling and scaling.
Ditto on the possible bad box or cable.
Thanks for the suggestion about the S-Video cable, I'll try another one, but I doubt it will make much difference, since I did my side by side tests by moveing the source end of the same cable back and forth between the DVD player and the CM7000 (so that the S-Video cable, and port on the TV remained exactly the same).
Sorry if my comments seemed a bit negative. I just don’t like to see a basically good design like the CM7000 hurt by unnecessary silly mistakes, like putting tiny fonts that need 400 or 500 lines of resolution to read, on a box required to support connections through an NTSC off-the-air tuner that can handle (at most) about 330 lines of resolution.
One of the things that bugs me most about the guide text being so small and the S-Video being sub-par on sharpness is that these problems interact in a very annoying way.
I can actually use my Trinitron's sharpness
setting to punch-up the overall quality of the CM7000 image quite a bit, so the sharpness with actual on-the-air video looks much better (quite nice actually), but this causes some slight ringing artifacts, which I could normally live with, but because the channel guide text is so damn small, even minor ringing causes the guide to becomes totally illegible and useless (which I can't live with), so this work-around is not a practical solution.
On the plus side, the Channel Master CM7000's ST chipset did do significantly better in my side by side testing with the Tivax T8 on things like dynamic multipath and hiding blocky low-bit-rate MPEG artifacts and mosquito noise (deblocking and deringing seem to be better).
Multipath is almost always a concern with ATSC's 8VSB modulation, and the other artifacts mentioned can get pretty bad when broadcasters try to push their bit budget too far with sub-channels. So, I don't want to trash the CM7000 too much, because in a lot of ways it does deliver where it counts.
In fact, my friend was impressed enough by these factors when I showed him my CM7000, that he will probably also get a CM7000 dispite the other issues.
So, I will be able to do a side by side with his box when it arrives, to see if my CM7000 is sub-par.
The Tivax STB-T8 also does a pretty reasonable job considering what these boxes have to deal with when decoding 8VSB signals in a residential setting (though not quite as good as the CM7000).
As to down-sampling vs. down-scaling . . .
Not to nit pick back, but I was trying to be fairly precise. You see, from what I am seeing, dignifying what my Channel Master seems to be doing as 'down-scaling' may be, I think, perhaps a bit over generous.
A decent bicubic resize from an HD source would be enormously sharper, and even the simplest of respectable "down-scaling" algorithms (a simple bilinear resize) should be very sharp when you have a good quality 1920x1080 source to start with.
But I don't think that's what is going on . . .
The cheap dirty approach is to just down-sample the video to the lower resolution using a resizing algorithm called 'nearest neighbor
', which as you probably already know, just down-samples (or down-scales if you prefer) by grabbing some pixels and skipping others.
As a test I took a good sharp HD capture, then used Virtual Dub to downsize it to 704x480 using Virtual Dub's built in cheap dirty 'nearest neighbor' resize.
The result from the nearest neighbor type resize was a nice sharp image, but one with very noticeable 'jaggies' from the non-integer ratio down-sampling.
No problem though; just apply a mild gaussian or box blur function (for our S-Video analog video, just use a simple low pass filter on the converted signal) and BINGO - no more jaggies (but now, the video sharpness is nowhere near as sharp as it should be).
Virtual dub also has bilinear, bicubic, and lanczos resize, so, just for the sake of comparison, I tried downsizing the HD video with these more sophisticated true re-scaling
algorithms. Every one gave very clean resizes, without jaggies or other artifacts, while maintaining very sharp output.
As I said above, the ST chipset in the Channel Master CM7000 does seem to do an above average job in dealing with dynamic multipath, and MPEG decoding artifacts are suppressed better than most of the other CECB boxes, but from what I have seen so far, it looks like the 'down-scaling' you refer to may be nothing but simple nearest neighbor down-sampling, followed by mild lowpass filtering
I could be wrong though – Maybe the CM7000 does use a more sophisticated rescaling algorithm. If so, it is either very poorly implemented, or, as has been suggested, my CM7000 is just not quite up to par.
As I said above, I will recheck with another sample of the CM7000 when my friends box arrives.
To Rammitinski, who commented that my expectations were probably too high to start with, I would only say that I agree 100%. As an engineer who was involved with the RF engineering of some of the very first high capacity digital radio links in this country, including some providing broadcast quality STL (studio to transmitter) video links, I have little patients for half baked hardware designs, and even less for bone headed software.
The manufactures are not allowed by the CECB rules to output an HDMI, DVI, or Split Component signal, but anything you can squirt out of a composit or S-Video jack is FINE, so it would have been nice if they had done a little better job in maximizing this option.
I just wish they had sweated some of the details on these boxes a little more before pushing them out the door.