cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
NaN_user
Visitor
Visitor
569 Views
Registered: ‎01-28-2021

Worse GTH eye margin using CPLL instead of QPLL?

Jump to solution

FPGA: XCKU060
XCVR type: GTH
Line rate: 10.3125 Gbps

I have performed some IBERT eye scans changing only the QPLL to the CPLL in IBERT IP, and I have the impression the horizontal aperture (UI axis) is narrower with the CPLL.
Is this to be expected or some random factor made the aperture narrower?

Screenshots of both eye diagrams are attached.

quad_224_x1y1_tx_x1y2_rx_qpll.png
quad_224_x1y1_tx_x1y2_rx_cpll.png
0 Kudos
1 Solution

Accepted Solutions
roym
Moderator
Moderator
320 Views
Registered: ‎07-30-2007

You won't have problems with a channel that has that Eye Scan.  I don't think any more modification is justified.  If you're using a Termination of "programmable 800mV" you shouldn't change it and nothing in the Scan suggests any termination problem.




----------------------------------------------------------------------------
Don't forget to reply, kudo, and accept as solution
Be sure to visit the Resources post periodically to keep up with the latest
https://forums.xilinx.com/t5/Serial-Transceivers/Serial-Transceiver-Forum-Guidelines-and-Useful-Resources/td-p/1173590
----------------------------------------------------------------------------


View solution in original post

5 Replies
roym
Moderator
Moderator
549 Views
Registered: ‎07-30-2007

It is not too unusual. QPLL is supposed to be a little better.  You need to set the horizontal and vertical increments lower, probably down to 2 or 1 to get a good view of this.




----------------------------------------------------------------------------
Don't forget to reply, kudo, and accept as solution
Be sure to visit the Resources post periodically to keep up with the latest
https://forums.xilinx.com/t5/Serial-Transceivers/Serial-Transceiver-Forum-Guidelines-and-Useful-Resources/td-p/1173590
----------------------------------------------------------------------------


0 Kudos
NaN_user
Visitor
Visitor
539 Views
Registered: ‎01-28-2021

Hi roym,

In the meantime I have doubled the horizontal the resolution and run two scans for each mode (CPLL/QPLL). The results follow.
QPLL
ch3_quad_224_x1y1_tx_x1y2_rx_qpll.pngch3_quad_224_x1y1_tx_x1y2_rx_qpll_1.png

CPLL

ch3_quad_224_x1y1_tx_x1y2_rx_cpll_1.pngch3_quad_224_x1y1_tx_x1y2_rx_cpll_2.png

My only worry is if I should expect a BER greater than 10^-15 with the CPLL results. I have seen that an amplitude of 45 codes and a horizontal margin of +-0.15 UI should be enough, of course supposing this is stable.
Would you suggest some changes before running the 10^-15 BER test?
I do need two different line rates in this quad, so I think my only option is the CPLL, right?

0 Kudos
roym
Moderator
Moderator
384 Views
Registered: ‎07-30-2007

It looks like your cpll graph is a little marginal and might be improved with maybe 3dB of of pre-emphasis on TX side.  You still need better resolution to prove 10^-15 but it looks like it will qualify.




----------------------------------------------------------------------------
Don't forget to reply, kudo, and accept as solution
Be sure to visit the Resources post periodically to keep up with the latest
https://forums.xilinx.com/t5/Serial-Transceivers/Serial-Transceiver-Forum-Guidelines-and-Useful-Resources/td-p/1173590
----------------------------------------------------------------------------


0 Kudos
NaN_user
Visitor
Visitor
350 Views
Registered: ‎01-28-2021

Hi @roym ,
I have increased the resolution of the eye scan (2x2 increments):
default_txdriver.png

I applied 3.10dB to the TX Pre-Cursor, but it only reduced the vertical margin:

txprecursor_3db.png

Maybe a TX pre-cursor, TX post-cursor sweep is needed? Different values of RX term might help?

EDIT:
Okay, I got much better results by deactivating DFE:
nodfe.png

I noticed the eye scan is actually faster with DFE deactivated? I got 40 minutes instead of a 70 minutes eye scan using the same BER and increment settings.

roym
Moderator
Moderator
321 Views
Registered: ‎07-30-2007

You won't have problems with a channel that has that Eye Scan.  I don't think any more modification is justified.  If you're using a Termination of "programmable 800mV" you shouldn't change it and nothing in the Scan suggests any termination problem.




----------------------------------------------------------------------------
Don't forget to reply, kudo, and accept as solution
Be sure to visit the Resources post periodically to keep up with the latest
https://forums.xilinx.com/t5/Serial-Transceivers/Serial-Transceiver-Forum-Guidelines-and-Useful-Resources/td-p/1173590
----------------------------------------------------------------------------


View solution in original post