Conexant TV card test program!

@agafaster

To be honest, I'm not sure.... but then I'm not sure to which problem you're referring either!

All I know is that on the A1 I can get a good picture using a Cx2388x chipset card, but on my Bt8x8 it raises an interrupt saying it's trying to execute a bad opcode. This sounds to me like it's trying to run from a bad address. However, I've not had time to look into it properly.

Now I've got overlay support I think I don't need to worry about the suspiciously inefficient PCI bus on the A1, as I've got a full-scale PAL picture (768x576) running with no artifacting. Overlay support among other things halves the bandwidth I need as it's using YUV422 (2 bytes) instead of RGB565 (4 bytes).

DMAing isn't too much of a problem at the moment because I've split the bitmap into different chunks, one per line, but it would still fail on large lines (1600x1200 resolution would probably fail, for instance).

The main problems:
1) Cx2388x chipset raises interrupts which never trigger the associated routine (works on Bt8x8)
2) Bt8x8 executes the wrong area of memory. I may know how to get round this but if so it'll only work up to resolutions of about 576 lines until I get the new SDK.
3) Bt8x8 raises too many interrupts when the signal is poor and freezes the hardware solid. Don't know if this is AOS or the A1 at fault.

Bug 1 is a mystery to me.
Bug 2 I think I may be able to get round.
Bug 3 may mean I just can't use interrupts on the Bt8x8, which may slow things down immediately. The next rev of AOS however may allow me to DMA straight into the Overlay, so it may cancel it out.

From what I've seen so far, though, I'd have to recommend the Cx2388x chipset over the Bt8x8. It just seems much more solid. Unfortunately, of course, most people have Bt8x8 chipsets!

@Spirantho

interesting stuff.

the bitmap bit interests me - why would one want a bitmap bigger than PAL (768x576) ? could you not assemble the output and then scale as required, or pump the output to a PAL res screen (say 800x600) ?

I did notice that the BT would lock up AOS when the signal wasnt good, but I was able to work (proper paid work that is!) on me laptop, while watching the Ashes Test series some days on my A1 for long periods. C4 was pretty good in me room, 5 and Two would cause problems sometimes - but then, i was using a loop aerial nicked from me kids' portable - however it would also sometimes lock when displaying RF output from a ZX-Spectrum too.
Is there nothing in the Dev notes for the BT8x8 to limit this? I guess it could just be a bug that Connexant ironed out when developing the Cx2388x. (Brooktree were bought out by Connexant a few years back).

Good luck, and I'll watch with interest !
I think I'll try my current TV-card in my kids' Peecee to compare notes....

@agafaster

You want a bigger bitmap if you've got a window on a large screen, which is going to happen more and more as monitors get higher resolutions. I can assemble the output and scale, but it's slower than I'd like, and rules out proper Overlay support.

Interesting what you said about watching the test match. I didn't even think I'd released a version with the tuner activated! It's something I want to add fairly soon to the Cx2388x card.

As for it being a bug that's not in the Cx chipset - I don't know, as it's not raising any interrupts! It's very possible if I work out how to make those interrupts trigger, then it'll still crash!

@Spirantho

Quote:

Hmm... can you send me a minimal test program with source code and binary that exhibits the problem?

@Rogue

I'll try and write a quick test case when I get chance but it may be a while. I'll email it to you when I've knocked it up.

Thanks!

@Spirantho

Quote:

Ok, understood - just a PITA whichever way I suppose !

-edit-
although could one not get the graphics card to scale the 'overlay' bitmap somehow ?

Quote:

you sent it to me, muppet ! it was really good - worked a charm on Update 2-3, then stopped working for some reason. I tried recompiling - but I had to find a bt878.h from somewhere. as I recall, it struggles around the point of generating P97 bitmaps, but I could be wrong !!

Quote:

I guess that in itself could be a bug - still such is the life of the driver writer eh!