UPGRADE YOUR BROWSER

We have detected your current browser version is not the latest one. Xilinx.com uses the latest web technologies to bring you the best online experience possible. Please upgrade to a Xilinx.com supported browser:Chrome, Firefox, Internet Explorer 11, Safari. Thank you!

cancel
Showing results for 
Search instead for 
Did you mean: 
Observer ganeshtk
Observer
443 Views
Registered: ‎12-06-2016

GTX Near End PMA does not work with strange behaviour

Jump to solution

Hallo,

(I have already followed other related threads, but were not helpful here)

I have a GTX Transceiver in my SATA Host design on Zynq 7100 device, which I want to test in Near End PMA loopback mode with internal PRBS Pattern Generator/Checker.

So, I first start loopback mode by setting LOOPBACK[2:0] Port to '010' and then start GTX TRPRBS Generator and RXPRBS Checker at the same time by setting for example '001' (PRBS-7, I also tried other patterns) to TXPRBSSEL[2:0] and to  RXPRBSSEL[2:0]. Then I reset GTRX channel by toggeling GTRXRESET pin. To verify the result I read RX_PRBS_ERR_CNT through DRP interface and which is always 65535 (I frequently reset the count) and this I consider as the loopback mode does not work.

I have following setting for GTX Channel.

Equalizer is set to DFE

TX and RX gearbox - False

TX8B10BEN - 1

RX8B10BEN - 0

TXBUFEN - True

TX_XCLK_SEL - TXOUT

RX_BUF_EN - False

RX_XCLK_SEL -  RXUSR

With ILA, I see RXDATA path shows some pattern as follows

0x0F3CF0C3 0x0FFC0F03 0xF0FFF00 0x0F3CF0C0 then repeat the same pattern again for couple of times.

Now the strange things are follows

1. When I turn off the pattern generator and checker but keep the near end PMA loopback mode, then also I see above similar patterns

2. When I turn off both pattern and loopback mode, then I see that RXDATA line constantly shows link training patterns D21.5 (which is 1010101010) or D10.2 (which is 0101010101) charachters even if the Analog Front of the GTX transceiver is open.

So my quations are, why does loopback mode with pattern generator not working? am I missing any settings?

an why does GTX Transceiver shows above behaviour?

I appriciate any help in this direction.

 

Thank you in advance

 

1 Solution

Accepted Solutions
Observer ganeshtk
Observer
315 Views
Registered: ‎12-06-2016

Re: GTX Near End PMA does not work with strange behaviour

Jump to solution

Hallo,

the problem is now solved. The reason was obviuos , that the GTX transceiver is configured for SATA Phy and so it uses SATA OOB block on TX path.

To use any loopback mode in such case, I must first turn off  the SATA OOB block. I did it by setting TX/RX PD to 00, TXDETECTRX to 1 and TXELECIDLE to 0 and now it works. But before that I balanced the TX and RX paths by setting 8B10 Encoder and Decoders active.

I think "0x0F3CF0C3 0x0FFC0F03 0xF0FFF00 0x0F3CF0C0 then repeat" are the OOB bursts. And D21.5 / D10.2 are link idle charachters.

@gguastiI tried with LPM, then I see 0xFFFFFFFF instead of D21.5 / D10.2.

Thank you guys for your valuable inputs.

Ganesh

0 Kudos
4 Replies
Moderator
Moderator
418 Views
Registered: ‎07-30-2007

Re: GTX Near End PMA does not work with strange behaviour

Jump to solution

It is odd to have 8B10B on TX and not on RX.  An asymmetric channel could have clocking mismatches on a loopback and loopback may not be viable.  I would suggest analyze the TXOUTCLK and RXOUTCLK frequencies.  I would first start with one of the clock patterns to verify the loopback is running.  You also need to be checking that the T/RXRESETDONE is completeing.  You should start this with the example design which often delivers a working simulation at least on symmetric designs.  It will be much faster and easier to check this out in simulation.

Roy


----------------------------------------------------------------------------
Don't forget to reply, kudo, and accept as solution
----------------------------------------------------------------------------


0 Kudos
Observer ganeshtk
Observer
404 Views
Registered: ‎12-06-2016

Re: GTX Near End PMA does not work with strange behaviour

Jump to solution

Hello Roy,

thank you so much for the reply.

I will examine the T/RXOUT and T/RUSR/2 clocks. Also check if T/RXRESETDONE is completeing.

Where should I look for example design? Do you mean IBERT? But as long as I know IBERT has no simulation included in example design.

-Ganesh

0 Kudos
Xilinx Employee
Xilinx Employee
363 Views
Registered: ‎11-29-2007

Re: GTX Near End PMA does not work with strange behaviour

Jump to solution
hello
I just want to try to answer to the second "strange" behaviour. I suspect this is due to the combination of no signal present and usage of the DFE equalizer.
Please can you switch to LPM and check if it is still receiving 010101...?
0 Kudos
Observer ganeshtk
Observer
316 Views
Registered: ‎12-06-2016

Re: GTX Near End PMA does not work with strange behaviour

Jump to solution

Hallo,

the problem is now solved. The reason was obviuos , that the GTX transceiver is configured for SATA Phy and so it uses SATA OOB block on TX path.

To use any loopback mode in such case, I must first turn off  the SATA OOB block. I did it by setting TX/RX PD to 00, TXDETECTRX to 1 and TXELECIDLE to 0 and now it works. But before that I balanced the TX and RX paths by setting 8B10 Encoder and Decoders active.

I think "0x0F3CF0C3 0x0FFC0F03 0xF0FFF00 0x0F3CF0C0 then repeat" are the OOB bursts. And D21.5 / D10.2 are link idle charachters.

@gguastiI tried with LPM, then I see 0xFFFFFFFF instead of D21.5 / D10.2.

Thank you guys for your valuable inputs.

Ganesh

0 Kudos