FPGA Central

Has anyone had any sucess in using RocketIO from Virtez 2 or Virtex 4
for Serial ATA applications? From the information collected on various
news groups and web sites, my understanding is SATA OOB cannot be done
using RocketIO in Virtex 2s. Did any one try Virtex 4s?

What is the comment from FPGA gurus out there?

- sg

it can be done with v2pro, there are some tricks though

I have tested this as a working solution with real sata chips on memec VP20 board, sata cable to SMA and OOB sequence all passed ok

Antti

Hi Antti,

Can you share those tricks with the group?

-sg


Antti Lukats wrote:
> it can be done with v2pro, there are some tricks though
>
> I have tested this as a working solution with real sata chips on
memec VP20 board, sata cable to SMA and OOB sequence all passed ok
>
> Antti

"sg" <[email protected]> schrieb im Newsbeitrag
news:[email protected] oups.com...
> Hi Antti,
>
> Can you share those tricks with the group?

I think I have done that sooome time ago, but I had to reformat the HD on
openchip server so the files are not accessible.

basically there are 2 possibilities
1) small external circuitry, requires 1 regular IO pin from FPGA this is
Xilinx solution, please contact Xilinx for details.
2) my solution, OK to use for development at least, no external hardware,
tested with real V2Pro and real SATA chip

so my trick is that use 2 MGT RX from one and TX from the other. and using
POWERDOWN to turn of the transmitter as the powerdon on-off delay are not
equal the SATA OOB transmit timing has the pre-adjusted to get the required
chirp timing in the output stream. In the receive section there a specially
developed for MGT squelch detector, thats a bit magic too as the MGT
receiver when not seing proper inpout (eg during SATA OOB) has weird repeat
noise pattern (repeating over 4 bits) this has to cancelled in order to
properly and reliably detect the incomign OOB

detailed enough?

Antti

Antti,
I am interested in your solution to the problem of using SATA on
V2Pro. I still don't understand how the receive takes place. Did you
use an external ckt to detect the OOB ? If not, how did you achieve
the cancellation of the repeated noise patterns ?
It would be great help if you could share the files on the
openchip server again.
Thanks
Ajith

"Antti Lukats" <[email protected]> wrote in message news:<cucjbe$1c8$01$[email protected]>...
> "sg" <[email protected]> schrieb im Newsbeitrag
> news:[email protected] oups.com...
> > Hi Antti,
> >
> > Can you share those tricks with the group?
>
> I think I have done that sooome time ago, but I had to reformat the HD on
> openchip server so the files are not accessible.
>
> basically there are 2 possibilities
> 1) small external circuitry, requires 1 regular IO pin from FPGA this is
> Xilinx solution, please contact Xilinx for details.
> 2) my solution, OK to use for development at least, no external hardware,
> tested with real V2Pro and real SATA chip
>
> so my trick is that use 2 MGT RX from one and TX from the other. and using
> POWERDOWN to turn of the transmitter as the powerdon on-off delay are not
> equal the SATA OOB transmit timing has the pre-adjusted to get the required
> chirp timing in the output stream. In the receive section there a specially
> developed for MGT squelch detector, thats a bit magic too as the MGT
> receiver when not seing proper inpout (eg during SATA OOB) has weird repeat
> noise pattern (repeating over 4 bits) this has to cancelled in order to
> properly and reliably detect the incomign OOB
>
> detailed enough?
>
> Antti

"Ajith" <[email protected]> schrieb im Newsbeitrag
news:[email protected] om...
> Antti,
> I am interested in your solution to the problem of using SATA on
> V2Pro. I still don't understand how the receive takes place. Did you
> use an external ckt to detect the OOB ? If not, how did you achieve
> the cancellation of the repeated noise patterns ?

OOB detect is sure possible without external circuitry
basically COMMADETECT should assert on the chirps
but we did see actual capture of OOB chirps where the
COMMADETECT was not always asserted ,
the digital squelch solution was still working

its the basic problem of detecting silence (eg open input
of MGT) from non silence.

open MGT input is not silence and not constant on the
MGT RXDATA pins ! so some decision logic is required

our work was internal was proof of concept mostly, succesful
though to show that the MGT OOB is doable and does
work with real SATA silicon

> It would be great help if you could share the files on the
> openchip server again.

those files have actually never been publicly available, published was only
chipscope capture around the OOB sequence,

Antti