Hi Guillaume,

 

4600 ppm peak-to-peak (45.3 UI) of sinusoidal modulation at 97 kHz and 1840 ppm pp (6.5 UI) at 270 kHz are OK.

 

I think that it is important that we have these two test points on the sloping side of the Jitter Transfer Function (JTF). The section in between is theoretically a straight line.

 

Some people had understood that we wanted to test at 10 kHz or with 14870 ppm pp at 30 kHz and I think that we have to clarify that these two points exceed the expected frequency deviation range of the receiver VCO, whereas 97 kHz was selected to be the point where the receiver CDR should be able to handle deviation as well as jitter tolerance.

 

Sincerely,

 

Bent Hessen-Schmidt

SyntheSys Research, Inc.

3475-D Edison Way

Menlo Park, CA 94025

(650) 364-1853

www.BERTScope.com

 


From: Guillaume Fortin (Montreal) [mailto:Guillaume_Fortin@pmc-sierra.com]
Sent: Monday, July 21, 2008 12:43 PM
To: Bent Hessen-Schmidt; t10@t10.org
Subject: RE: Receiver JTF tolerance

 

Hi Bent,

 

I agree that the modulation at 97kHz that we will be applying is likely higher than what would be applied in a reasonable SAS-2 system with 30kHz triangular SSC modulation. However, the residual jitter that results from the 45.3UIpp modulation at 97kHz creates a 0.1 UIpp residual jitter after the JTF and thus meets the SAS-2 specification, that is to say that a compliant transmitter could generate it. If the transmitter is allowed to generate this jitter, we must ensure that the receiver can track it.

 

By ensuring that a compliant transmitter never produces more than 0.1UI of residual jitter after the JTF and ensuring that a receiver accepts an input jitter that is at least 0.1UI after filtering by the same JTF, we ensure consistency in our analysis of system robustness. This is what the inverse JTF mask accomplishes.

 

A 4600ppm-pp sine SSC modulation would meet the spec so I disagree with the 1.3x factor for triangular vs sinusoidal. I am not sure that I follow your train of thoughts with the modulation index. It would seem to me that the modulation index (m = df / fm) would actually decrease by a factor of 3.23 at 97kHz vs 30kHz since the modulation frequency increases (30kHz -> 97 kHz) while the df stays constant (2300ppm * 6GHz = 13.8MHz).

 

Regards,

 

Guillaume

 


From: Bent Hessen-Schmidt [mailto:bhessen@SynthesysResearch.com]
Sent: Monday, July 21, 2008 1:16 PM
To: Guillaume Fortin (Montreal); t10@t10.org
Subject: RE: Receiver JTF tolerance

Hi Guillaume,

 

The +/-2300 ppm (4600 ppm peak to peak) deviation may be the same for your proposal, but please realize that the modulation index will be 3.23 times larger at 97 kHz than it is at 30 kHz. Further the sinusoidal component of a 5000 ppm peak-to-peak triangular SSC waveform is merely 3626 ppm at 30 kHz (another factor of 1.3 of difference).

 

Your proposal therefore is about 4.2 x the modulation which can be generated at 97 kHz in SAS system. Instead a factor 2 x, which is very close to your peak (not peak-to-peak) values in ppm as displayed in your graph, will be much more reasonable for receiver testing.

 

Bent

 


From: owner-t10@t10.org [mailto:owner-t10@t10.org] On Behalf Of Guillaume Fortin (Montreal)
Sent: Monday, July 21, 2008 8:29 AM
To: Bent Hessen-Schmidt; t10@t10.org
Subject: RE: Receiver JTF tolerance

 

You are correct Bent: the frequency offset values are peak values.

 

It is true that the frequency offset at 30kHz is too large if we scale the SJ modulation by the inverse-JTF below ~2MHz. This is why my proposal is to stop at 97kHz, which is the point at which the SJ modulation amounts to +/-2300ppm, which is the maximum SSC modulation that has to be tracked according to the SAS-2 spec.

 

I will create figures for the SJ mask that should make things clear.


Regards,

 

Guillaume

 


From: Bent Hessen-Schmidt [mailto:bhessen@SynthesysResearch.com]
Sent: Friday, July 18, 2008 2:40 PM
To: t10@t10.org; Guillaume Fortin (Montreal)
Subject: Receiver JTF tolerance

Hi Guillaume,

 

There appears to be an error going from the equation for frequency offset on page 9 to the figures on page 10 and forward in your presentation document 08-248r0. When the offset plotted is only half of the peak-to-peak offset as the max and min values of Cos(x) are 1 and -1 respectively and the peak-to-peak frequency offset therefore should be double the values plotted. i.e 14870 ppm at 30 kHz for 6 Gb/s.

 

These are extremely large values considering that the peak-to-peak transmitter deviation is 5000 ppm plus/minus frequency tolerance. The transmitter SSC waveform is often a triangular waveform in which the 30 kHz sinusoidal frequency component is merely 3624 ppm peak-to-peak followed by a decreasing amount at each odd harmonic of 30 kHz.   

 

I do not see any source of ppm amplification in our SAS channel. We should certainly make sure that the receivers have some margin (i.e. that the SSC Waveform becomes a small or insignificant portion of our jitter budget). 4x margin may however a too much margin. We therefore suggest that the margin be 2x instead of 4x, this will leave the current ppm frequency offset numbers as is in the graphs while the UI numbers be half of the values of your presentation.

 

Sincerely,

 

Bent Hessen-Schmidt

SyntheSys Research, Inc.

3475-D Edison Way

Menlo Park, CA 94025

(650) 364-1853

www.BERTScope.com