cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Highlighted
Observer
Observer
450 Views
Registered: ‎11-22-2019

Eye Mask with GT-Debuuger and IBERT

Jump to solution

Hello there,

I am a newbie, working on to plot eyemask for my design with Zynq Ultrascale+.

My requirement is to check link quality and margin for xdma in design, I am following XAPP1295 and added some debug ports into my .dcp file, with debug clk from the design. I have also added following ports::

- GTTXRESET

-GTRXREST

-EYESCANRESET

-EYESCANTRIGGER

-

I am able to scan all four channels using GUI (IBERT), as per xapp1295 to use eye-qualify.tcl 

gives a message in vivado tcl console: link might be closed.

In GUI, I get eye diagram open as nearly 30 % of UI.

Here are the steps I am following:

- source ./insert_gt_dbg/insert_gt_dbg_hwproc.tcl

source ./insert_gt_dbg/igd_eyescan.tcl

following this, I reset GTs as without this scan hangs.

igd_eyescan init cX0Y12 res_sig_prefix=igd_gpio_cX0Y12_GTTXRESET, (to all channels) then I get the eyemask but with 2 or 3 codes showing up (eye is not normal).

is this  method correct? 

Specifications:

LVDS standard, GTH Serial transceivers

Here are the attatchments, please find it.

Thanks in advance

IMG_20200928_212604.png
0 Kudos
1 Solution

Accepted Solutions
Highlighted
Observer
Observer
135 Views
Registered: ‎11-22-2019

Hello,

Even after enabling PMA loopback did not solve the 'link closure' issue.
Here, I have found a possible solution, saying XAPPs won't work for Ultrascale devices. (Although, the xapp document clearly does not mention)
https://forums.xilinx.com/t5/Serial-Transceivers/Eye-Scan-on-GTH-on-VCU108-under-MicroBlaze-Control/m-p/1036472

- Another, matching explanation I got from ug576, RX margin analysis

  ''At high line rates, the received eye measured on the printed circuit
board can appear to be completely closed even though the internal eye after the receiver equalizer is open. " (possibly, this could be the reason of closed link).

- So, I have calculated eye margin using In system Ibert (which was known earlier, but somewhere I was looking for the Eyemask plot), for the channels.

Thanks,

View solution in original post

0 Kudos
7 Replies
Highlighted
Xilinx Employee
Xilinx Employee
357 Views
Registered: ‎10-19-2011

Hi @joychauhan43 ,

so far I do not see what could be wrong. Can you try the 'igd_eyescan run ...' command by itself to see what the eye looks like? Maybe try the matrix option to have a compare with the IBERT GUI.

------------------------------------------------------------------------------
Don't forget to reply, give kudo and accept as solution
------------------------------------------------------------------------------
0 Kudos
Highlighted
Observer
Observer
329 Views
Registered: ‎11-22-2019

Thanks for your reply.

In case of script, eye gets an 'X' value at the center of eye mask. However, running igd_eyescan run for individual channels, I get the plot as 'X' near edges of the eye depending on max_circle_div value.

Why is this 'X' insted of BER value, please help me out.

(Kindly, refer the attached document)

Thanks

IMG_20201006_100544.png
0 Kudos
Highlighted
Observer
Observer
272 Views
Registered: ‎06-16-2020

Hi @joychauhan43,

have a look at XAPP1295 on page 23. There you have a complete description of the eye scan feature.

There it reads: "A setting of –1 [for the prescale value] selects the ARMED branch of the FSM and provides the quickest possible scan with just pass/fail results."
A bit further down (p. 26) it also reads:
"The characters in the plot represent the following:
      • X – fail, measured with ARMED scan (1.0)
      • . – pass, no errors (0.0)
      • 0-9 – represent measured BER with exponent 10–0 to 10–9
      • a-v – represent measured BER with exponent 10–10 to 10–32
      • z – anything else "

You are using the setting "-1" for prescale and that's why you get this result. As far as I understand it, it's just finding the inner borders of the eye but it's not testing for any BER value.

0 Kudos
Highlighted
Observer
Observer
260 Views
Registered: ‎11-22-2019
Hey @friede,
The above figure is generated after runnng this command with prescale, you are right that is the reason, it is giving eye border.
What my point is when I run eye_qualify.tcl from Xilinx, that is using 'ber' to plot the scan (as I have understood); gives a message 'Link might be closed' and the report comes as 2-3 code for the eyemask, So what do I need to check here? Does the hardware side need a check? Actually, I have no idea about it.
Second thing, In normal IBERT GUI, eye diagram is open that can be said to be a good eye, so if the link close when I reset GTs ?
0 Kudos
Highlighted
Observer
Observer
209 Views
Registered: ‎06-16-2020
To be honest I don't have to much knowledge about this either, as I just started working with the GT-Debugger. I've also only read a few parts of the AR#70915 that describes the eye qualification, but never worked with it.
If you have a opening of 0.3UI horizontally and ~52 codes vertically with IBERT you should be fine. If IBERT and GT-Debugger produce different results, I'm not really able to tell what the problem is.
Your first picture shows 0.03UI and 2 codes opening which isn't good. But I would rely on the result of the IBERT.
Maybe try using a near-end PMA loopback as reference. Activate it and then have a look at the results. The eyes normally should be wide open.
Highlighted
Observer
Observer
203 Views
Registered: ‎11-22-2019

@friede 

Thanks for your suggestion, will try it for sure.

Thank you.

0 Kudos
Highlighted
Observer
Observer
136 Views
Registered: ‎11-22-2019

Hello,

Even after enabling PMA loopback did not solve the 'link closure' issue.
Here, I have found a possible solution, saying XAPPs won't work for Ultrascale devices. (Although, the xapp document clearly does not mention)
https://forums.xilinx.com/t5/Serial-Transceivers/Eye-Scan-on-GTH-on-VCU108-under-MicroBlaze-Control/m-p/1036472

- Another, matching explanation I got from ug576, RX margin analysis

  ''At high line rates, the received eye measured on the printed circuit
board can appear to be completely closed even though the internal eye after the receiver equalizer is open. " (possibly, this could be the reason of closed link).

- So, I have calculated eye margin using In system Ibert (which was known earlier, but somewhere I was looking for the Eyemask plot), for the channels.

Thanks,

View solution in original post

0 Kudos