[10GMMF] Further comments on the noise interferer in the TP3 test
I would like to also comment on the noise interfere in the TP3 test. Before getting into Vipul's message, I would like to make a few comments on the original proposal as well as what I think we need to achieve in the TP3 test, because I have been having discussions with others which indicate that people have different assumptions about why we even proposed a sinusoidal interferer and in general what we need to accomplish with the interferer in a TP3 stressed sensitivity test.
First, I'd like to clarify the reason we proposed the sinusoidal interferer for the static stressed RX sensitivity (S-SRS) test in the July presentation. The reason we proposed any noise loading in the S-SRS test (as opposed to just reducing the required OMA level) was based on the idea that it was possible/likely that an EDC implementation would suffer a penalty from the noise allocated in our link budget (0.4 dB for RIN and 0.5 dB for high frequency modal noise) which is higher than the 0.9 dB penalty which comes from normal eye closure calculations. The reason for choosing the sinusoidal interferer was not because we thought it was similar to the RIN or Modal noise, but rather out of a desire to use the same hardware involved in the 802.3ae SRS test. Essentially, this being based on a probably incorrect assumption that the sinusoidal interferer would stress the EDC to the same degree as say a broadband Gaussian noise source.
As I had pointed out many times, this was only a starting point and I fully expected there to be feedback from the EDC makers on the question of whether this was a good stressor.
With all this said and given comments from Vipul and other, let me state my opinion of what we should do and why.
1) I think there should be little argument that RIN is very close to broadband Gaussian noise. A measured noise spectrum from a 10G FP laser is relatively flat out to and beyond 10GHz, with a typical peaking in the range of 7 - 9.5 GHz of 2 - 4 dB (at least this is data I have on some parts of ours).
2) I don't know the spectrum of modal noise, but as Vipul said, we should not confuse the low frequency modal noise from mechanical perturbations (which will essentially be captured in our dynamic testing) with high frequency effects from laser mode partitions etc.
3) It is not important that we precisely simulate the nature of the real noise in the link. It is only important that we use a stressor where we are confident that the penalty it imposes on the EDC will equal or larger (but not much too much larger) than that resulting from the actual noise. Thus, if RIN is broadband and the Modal noise is narrower or less flat, but we feel that a broadband source will be tougher on the EDC than the modal noise, than at worst we are just adding a stress a few 0.1dB too high.
Given this, and the feedback I have gotten that a sinusoidal interferer may stress the EDC much less than broadband noise, I would have to agree with Vipul that a broadband white Gaussian noise source is the right way to go. I have briefly looked into this and found what looks like an inexpensive source.
(see for example the NC1128 in : http://www.noisecom.com/content/Products/Components/NC1000/nc1000.pdf).
(I have not found, however, the equivalent of a sinusoidal generator instrument for noise in the range up to 10 GHz.)
I don't think the calibratation is a terribly big issue. In any case, in the setup and characterization of the test signal, I expect a procedure similar to that in 802.3ae, where the noise is added to the noise which exists in the test source in an amount to achieve a total desired S/N ratio (which we still need to calculate).
Other than whether we can in fact not include the noise loading in the test signal (i.e.make it similar and choose a penlaty by which to lower the required OMA sensitivity), I don't see whiy we would have to make the problem any harder.
Lew
Lew Aronson (lew.aronson@finisar.com)
Finisar Corporation
1308 Moffett Park Drive
Sunnyvale, CA 94089-1133
408-542-4215 (PH)
408-543-0083 (FAX)