ML605 5Gbps noisy eye

[jochenkueppers]

Hi there,

I've got a problem with the GTX Transceivers of the LX240T on my ML605 Board.

On the pictures you see two different eye diagrams. One is the output of an Anritsu MP1800A and the other is the output of the ML605 with IBERT Design. Both are 7bit PRBS Signals. The output of the ML605 seems to be very noisy doesn't it? I think there must be some kind of Jitter, but where does it come from? Im sure I've had better eye patterns of the ML605 before. I'm still using the reference design from the usb stick... Can somebody help me?

[austin]

j,

You seem to have both noise in time (jitter), and noise in amplitude. In both cases, the eye is open, and there should be no bit errors at the receiver (if the link is low enough loss, and has the proper signal integrity -- impedance, no mis-matches, etc.).

The eye at the transmitter is a sanity check: if it is "bad" there, then you can not possibly get very far.

In this case, not knowing your measurement set-up, I can't say it is "bad." But I can say it doesn't looks "as good" as the other.

Given the amplitude noise present, I suspect that the reference design is running, and the other IOs on the chip are toggling, which causes the jitter, and the amplitude noise.

I would go ahead and see if the link has any bit errors (it probably does not).

[jochenkueppers]

Hi Austin,

Both, ML605 (J26/J27) and Anritsu MP1800A are connected the same way by a matched SMA cable to the Tektronix DSA8200 Digital Serial Analyzer (with differential probe). So there should be no big influence by the link.

Yes you're right i just took the reference IBERT Design from the USB Stick. So today I've customized a new design with the Core Generator with only one active GTX Transceiver. Unfortunately I got nearly the same result (see attachment)...

Is there any other reason for such a noisy signal? I think the eye pattern of a Virtex6 GTX Transceiver should be better, shouldn't it? I am confident with the opening but the jitter unsettles me...

Thank you for your help!

Jochen

[mcgett]

Activity in the CLB fabric or with other transceivers will not cause any degradation of the GTX. If it did it would be a major design flaw.

The ML605 is an evaluation board with multiple tradeoffs that can affect the performance. The ML623 is a higher end board for transceivers. The Anritsu MP1800 is high end test equipment which is a class onto itself. Expecting the ML605 to be as good as the MP1800 is unrealistic.

The quality of the TX eye will be dependent on the correct reference clock usage, PLL multiplier, TX swing and pre-emphasis. In addition the measurement configuration will also need to be setup correctly, this includes matched coax cables, AC coupling, 50 ohm termination to GND, and triggering with the same clock source that is used with the GTX.