From: Nial Stewart on
I've just noticed and read the thread you started just before Christmas
about your Fifo problem.


You said...

> problem fixed!
> solution and explanation in the next Brain issue
> (I will post short post also after the issue release)

Any sign of the new Brain?

BTW, was the read indicated by a single clock width pulse that was always
de-asserted after each read, or could this have been active for a number
of 62.5mhz cycles?


I'd have changed the design so the internal read flag was inverted at
each read, then the number of reads couldn't have been mistaken by the
higher speed clock domain.

Looking forward to the solution..


Nial



From: Antti on
On Mar 2, 4:17 pm, "Nial Stewart"
<nial*REMOVE_TH...(a)nialstewartdevelopments.co.uk> wrote:
> I've just noticed and read the thread you started just before Christmas
> about your Fifo problem.
>
> You said...
>
> > problem fixed!
> > solution and explanation in the next Brain issue
> > (I will post short post also after the issue release)
>
> Any sign of the new Brain?
>
> BTW, was the read indicated by a single clock width pulse that was always
> de-asserted after each read, or could this have been active for a number
> of 62.5mhz cycles?
>
> I'd have changed the design so the internal read flag was inverted at
> each read, then the number of reads couldn't have been mistaken by the
> higher speed clock domain.
>
> Looking forward to the solution..
>
> Nial

sign :)!

just returned from embedded still needing some breathe time


FIFO proble solution:
this was SO STUPID, the reason was that the engineer who made the
"read enable" signal from the PPC user logic
for some mystical reason did move this read enable from the PPC clock
domain to the MGT clock domain that
was in sync with the master clock on the master board, so in the slave
board all was fine, except fifo read enable
was in the clock domain of the master (syncronised to the MGT
recovered clock)

this also explains why the problem was not visible in chipscope:
chipscope did latch the signal in 1 F/F
and no bad disaster happened, what I also witnessed, but in the FIFO
itself the clock enable was controlling
multiple F/F that occasionally did then go nuts what I also witnessed.

so there was no clock issue, no FIFO issue, but.. well things happen.
I was not the only person who did look
at the code, but nobody did see any issue, well we all mostly looked
at the problems with the FIFO and CLOCK
signals...

Antti