MSI tv@nywhere (+other cx23881) comments

[mg262]

Dear all,

I have just realised that the CX23881 chips don't seem to offer much theoretical advantage over older chips (unless I'm being stupid). Here's why:

1) The 10-bit colour capture cannot be used by proprietary programs like VDub, etc. as Conexant wont release specs:

http://forum.doom9.org/showthread.php?s=&postid=329074&highlight=conexant+10bit#post329074

2) The remaining advantage, perfect C/Y splitting:

http://www.msi.com.tw/html/products/per/per_htm/8876-2.htm

is irrelevant with a S-Vid input.

So unless you're using the internal tuner or a composite connection....

Thoughts?

Mohan

[Owen]

I have two CX23881 based cards, The MSI TV@nywhere and a Leadtek TV2000 XP Expert and have had many BT8x8 cards. I can assure you the CX cards are noticeably better then any BT8x8 card. Resolution is not much different, but Luma range is improved and color is MUCH better.
I have also tried a FlyVideo 3000 with Phillips chip but I was not impressed by it. I found it to lacked sharpness and color accuracy compared to the CX cards.

If you are looking to purchase a card, I recommend the Leadtek. The drivers for the MSI are bug ridden. I had to use drivers for the Pixelview Xcapture to get it working properly.

Regards,

Owen

[homersapien]

I have a philips 7133 currently but used to have a cx23881 card. There is definitely a difference in picture quality between the two. The 7133/23881 both produce a noticably more detailed picture.

[Arachnotron]

The BT878 is an 8 bits chip, most philips are 9 bits and the CX23881 is 10 bits.

The phillips samples at 27 MHz both PAL/NTSC, the 878 and cx23881 both at 28.6 /35.5 MHz NTSC/PAL.

The CX23881 is 10 bits internally up untill the last step where the pixels are generated. So even if the final pixels are 16 bit yuv 4:2:2, they were interpolated from 10 bits raw data and so likely to be more accurate.

Also, the CX23881 has a 32 Kbit fifo, against only 630 bytes for the bt878. Should make it a lot more stable under heavy system load.

So even with s-video and generating 16 bits yuv pixels, the CX23881 should be the best chip.

On the other hand, driver support is minimal. The BT878 can be tweaked more at the moment.

By the way, the MSI cards use a solid state tuner. I have no idea if this makes any difference to the standard tin can tuners. Anyone?

[mg262]

Right... sounds like I was completely wrong, or at least that the theory isn't reflected in performance.

I have a MSI tv@nywhere, my first card - and I bought it because the chroma/luma separation was advertised to be better than other cards. I get a noisy picture, perhaps because I am actually using a composite connection (only have scart-RGB, composite and coax outputs from TV/VCR), so I thought a RGB-input card might give better results. I have yet to try the on-chip tuner... that may work better. Anyway, thanks for your thoughts, everyone.

[Arachnotron]

The difference in quality between s-video and composite is much larger then the difference in quality between the chips/cards.

s-video on the BT878 will allways look better then composite on the CX23881.

[mg262]

Hmm... in that case I might try another card sometime, or else find out about RBG->s-video converters. (Or I might just persuade the tuner to work and see how that is!) thakns for the info.

[Arachnotron]

The tuners on a TV card are allways composite video.

[mg262]

Yup, but I'm only interested in VHS source (not direct broadcast), and (I *think* that) if I use the RGB (s-vid) output, all that's happening is that the essentially composite videotape source is being sent through the VCR tuner to split and then passed through 3 (2) channels to a card... which may not be worse than passing the source directly to the tuner on the card to be split there? Or is the problem that signal quality degradation occurs in the cable?

I need to go and check up on how VCRs actually store data...

Mohan

[Arachnotron]

The video tape source is not essentially composite, but stores chroma and luma more or less separate. So if you record a composite source (say a TV show) the signal is comb filtered to separate luma and Chroma and stored, and on playback the signal is joined again to composite.

Unfortunatly, allmost no VHS-VCR will output a s-video signal. (this has nothing to do with the difference between the VHS and S-VHS tape format) Instead, chroma and luma are joined again to a composite output.

This is why a normal VHS tape, when played on a S-VHS deck which does have a S-video output looks better when captured then when played on a VHS VCR.
You avoid a second comb filtering step in the TV card, because Y and C stay separate after playback.

The tuner in a VCR has nothing to do with tape playback, and is only used when you are recording a TV signal.

By the way, RGB and S-video are different signals. Most VHS VCR's have only Composite output, and no RGB or S-video. The RGB pins in the SCART are not connected in this case.

[mg262]

I thought S-VHS stored chroma/luma separately but VHS did not? However...

>This is why a normal VHS tape, when played on a S-VHS deck which does have a S-video output looks better when captured then when played on a VHS VCR.

That makes perfect sense.

>By the way, RGB and S-video are different signals. Most VHS VCR's have only Composite output, and no RGB or S-video. The RGB pins in the SCART are not connected in this case.

I know... I meant RGB (or s-video) and 3 (respectively 2); I was being lazy/careless. I have SCART-RGB out of the VCR, but no S-Video; I think this is relatively common in Europe, but it is a pain as the MSI tv@nywhere (and AFAIK other 10-bit cards) have S-Video inputs but no RGB inputs.

Thanks for the info!
Mohan

[DAvenger]

I am thinking of purchasing this card ... I have a question.

Does it use WDM drivers so I can use alternative viewer software (I am not big fan of WinDVR)?

Thanks!

Martin

[Arachnotron]

If you are running w2k or XP, all drivers are WDM.