Meanwhile, back at gowgow's original question...
"I've heard that lots of people just use a fairly cheap tv input card and then use dscalar to make the image progressing and then a high end video card (and powerstrip) is used to output the image to the projector in it's native resolution.
What does this card do better than a cheaper tv input card would do? Is it the Faroudja processing is better at making the progressive image than dscalar?"
Those cheap TV input cards do a poor job of converting an input signal into a digital video stream. They vary greatly in quality and should be chosen with care. Some have sought out modified or 'professional' cards, a solution that is no longer 'cheap' (in either sense). Even then, you can't do better than the quality of the bt8x8's video decoder, which frankly leaves a lot to be desired.
The Holo3D uses the Philips SAA7118, an excellent video decoder. The hardware design and board layout was done by a good friend of mine, Chris Coley, who's the best hardware engineer I've ever worked with. He's also a home theater enthusiast like you and I, with a nice CRT projector setup at home.
He did the layout by hand (rare these days), with particular attention that there's no cross polution between analog and digital sections. We also put local power supply regulation on the board, to ensure none of the usual crud on the PCI's power supply makes it into the analog sections (hence the extra four pin 'drive power' connector on the board).
Even setting aside the Faroudja processing for a moment, the quality of the video decoder section on the Holo3D is in a completely different class from any bt8x8 board I know of.
That's important because the later processing stages (be it DCDi, dScaler, or both) greatly benefit from a high quality signal. It's the old 'garbage in - garbage out' principle. It doesn't matter how many MIPS and how fancy an algorithm you apply to a signal if it's already full of noise, chroma decoding errors, etc. For some algorithms (e.g. motion estimation) too much noise will severely compromise the algorithm's performance.
Besides all that, the board has component video inputs, something I don't believe is currently available on any other dScaler-compatible board. And we're planning to support those with SCART-based equipment, too.
"And if the Faroudja processing is better than what dscalar, then why are they boasting that this holo3dgraph card is compatible with dscalar?"
We can start with a couple of reasons:
a) dScaler is software, the Faroudja chip is hardware. DCDi isn't something that's practical to implement on today's PC processors, and since it's done in hardware, you'll get the same performance today as you will 2 years from now. dScaler continues to improve, and who knows what will be possible in future versions, running on the latest processors.
b) dScaler covers much more than deinterlacing - there are a variety of plugins that do image correction/enhancement.
The point here is that you can't lose. You'll have a very high quality input section, much better than any bt8x8 card I know of, with all the right kinds of inputs. Then you have the options of using DCDi in hardware or dScaler's algorithms in software to do de-interlacing, and in either case, you'll have the ability to run dScaler's other image processing plugins on the deinterlaced result.
Another nice aspect is that using hardware to do the deinterlacing reduces the CPU requirements considerably, which means either you can use a less expensive machine, or use the extra processing power you've freed up to do more post-processing.
Also keep in mind that the card has a sizable FPGA that could be used in the future for further pre- or post-processing, and a number of headers on the board that can be used for other purposes like direct links to MPEG decoder cards.
Hopefully that clears up the difference between the Holo3D and 'cheap TV input cards' :)
"Please, somone enlighten me. I might be interested in this card if someone could convince me it provided *dramatic* improvements over what you can get with solutions that are much less expensive."
This is what everyone's been waiting for - the card that lets the HTPC compete with a standalone scaler on the same footing. Arguablly better than a scaler in some respects, since it's possible to tap the large processing power of a contemporary processor to further process the image.
Is the improvement 'dramatic' over less expensive solutions? Obviously my opinions are not unbiased, I'm a part of the team that brought this to you. The people in the early adopter program will be able to post their opinions before long, and it will be shown publically at CEDIA (next week).
This was a labor of love, and I think it shows...
Holo3DGraph Team Member