[vdr] Re: Increasing SDRAM on DVB cards

[Oliver Endriss <o.endriss@gmx.de>]

Stefaan Coddé wrote:
> Oliver Endriss wrote:
> >The AV7110 is a very complex device with many built-in
> >functions/peripherals. Each of these reserves a part of the address
> >space. There are no external CS signals pins for the built-in
> >functions...
>
> The AV7110 can control 4MB but only the half is used.
> This half, 2MB is selected with the first chip select signal.
> The second half with the second which is not accessible.

Ack.

> They will not use an adress range internal if it's ment original
> for something else eg 4MB instead of 2MB and if that's the case
> of the free area then you can use a non conventional chip select line
> but that means a firmware change anyhow.

The problem is that you have to deselect the expansion SDRAM when 
*internal* devices are being accessed. For example, the PAL encoder 
uses 0x72xxxxxx (according to the datasheet). How do you ensure that 
the expansion SDRAM will not be activated when the firmware writes to 
the PAL encoder? There is no signal available which could be (ab)used.

Firmware changes are necessary (and possible IMHO).

> The question will be: where to fetch an chip select signal from?
>           what about timing?
>
> What about the select lines of the cam slot or combine two signals
> a clumsy way would be applying a exclusive and function on all
> address pins so you get a additional adress range which can be used
> as chip select signal for the second 2MB sdram.

As I said: This is *not* possible because you don't have enough signals 
to create a CS signal for the memory expansion.

> Example :in this case 8 adress lines
>
> A0 A1 A2 A3 A4 A5 A6 A7 A8
> If they become all high and you perform exclusive and then you end up
> in fact
> with A9 which could be used as chip select for a second ram chip with
> 8 adress
> lines in this case.
> There you have your independent chip select signal.

To avoid confusion:
Let's call your "address lines" "chip select lines": CS0..CS8, CS9
You can derive CS9, if (and only if) you can access *all* other CS 
lines.

If I may adopt your example: The problem is that we don't have CS4..CS8, 
because they are used internally, but we still need CS9.
No chance.

> The firmware must be modified then to think that there's a 4MB chip
> mounted
> cause it has to adress the entire 4MB range and not only the 2MB
> range.

I guess that's the easy part.

> That's the advantage of adressing, each address is 'unique'.
>
> Does this makes sence for y'all?

See above. ;-)

Oliver


-- 
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe vdr" as subject.