Part Number: ONET1151L
Dear Mohammadi,
in a 10.3125 Gbps SFP+ application with ONET1151L I am seeing a data dependent jitter spectrum with two distinct gaussian distributions and an according two lines eye.
When I am increasing deemphasis on the laser output to 6 of 7 I can make the two diracs vanish in the pdf at the cost of one broad jitter spectrum as shown in the attached screenshot.
The layout can also be seen attached - the laser connection is short and all components are 0201 size, PCB layout target is 25 ohm for each single trace.
Neither removing the small snubber network (82 ohm with 0.4pF) nor changing the series termination resistors from 5 to 0 ohm results in a better histogram.
Looking at the individual bits in the time plot it seems single 1 or 0 bits are not reaching the peak value fast enough.
Is there any setting or suggestion beyond output side driver deemphasis to speed up the bits?
Could it be that the 100nF power supply capacitors close to the IC output stages are too small?
Could a 0.4pF capacitor between MOD+and MOD- do any good as your reference schematic suggests?
Is there any other adjustment method to make the output stage more aggressive?
Maybe I should admit that my setup has another SFP+ electrical-optical connection before without a CDR in between - so jitter is higher by default.
However I am not seeing the data dependent split of the spectrum on other SFP+ units ...
Cheers
Philipp
