[SI-LIST] Re: Fibre channel interconnect margins

  • From: Paul Levin <levinpa@xxxxxxxxxxxxx>
  • To: Chris.Cheng@xxxxxxxxxxxx
  • Date: Wed, 05 Jul 2006 13:50:06 -0700

Dear Chris,

You've asked this question before and not received what you consider a 
satisfactory answer. Clearly it has to be better than 1e-12; almost as 
clearly, it has to be worse than 1e-19 (1 error after thirty years at 1 
Gbps if my math serves me correctly.)

Have you ever considered that (in my humble opinion) no company in its 
right mind would ever reveal what their number was since its competitors 
would then try to use that information against it?

Regards,

Paul Levin
Xyratex
_____________________

Chris Cheng wrote:
> Steve,
> I start off by asking the same question before, "does your company ship =
> products with 10e-12 error rate ?" and it seems it comes down to the =
> following answers :
> 
> a) "Its a fact of life, I'll try to keep it to 10e-x13,14,15 or better."
> 
> Well, I still want to know what that number is ? And when you have a =
> customer that has a peta byte of installation, how does that translate =
> into actual errors per second in your FCAL ?=20
> 
> b) It may happen in some part of the link but overall the link can =
> tolerate it
> 
> In particular,
> 
>>If over any finite period of time the RJ causes the frequency=20
>>as seen by the PLL to change then the PLL will move the VCO, thus=20
>>creating a limit to the max observed RJ.  If the RJ is distributed=20
>>so that the frequency does not have to change then the 'single=20
>>incremental interval' effect will apply.
>> Have we not then got a jitter distribution that is Gaussian in=20
>>form but with limits to the maximum deviations?
> 
> 
> I don't know about FCAL jitter tolerance spec but let's take the OC-xx =
> spec as a reference, what happen when the jitter spectrum passes ft ? Is =
> the jitter still bounded ?
> To take it out to a non-ideal world, how does one trade off jitter =
> tolerance (for example, by Rj and Dj) and jitter accumulation (for =
> example, by PLL supply noise) ? You only have one choice of loop =
> dynamics in your PLL and you can not help one without hurting the other. =
> And I would bet jitter accumulation from PLLVDD dwarfs the concern over =
> jitter tolerance. Just look at how elaborate those PLLVDD distribute =
> schemes we have (isolation traces, filters etc). Can you still say your =
> Rj is unconditionally bounded by such PLL's ?
> 
> c) 10e-12 happens only in methodology, you are really running a BERT =
> much much lower, possibly beyond the life of the system or the component =
> or reasonable measurement technique
> 
> Which, I think, is essentially what you are saying in the last sentence. =
> Do we really claim a 10e-12 simply because we got an erroneous Rj from =
> an instrument in a Dj dominated system, multiply the sigma to 14 or 15 =
> times per the BERT spec and scare ourselves to such error rate ?=20
> And before the instrument guys jump on me again, let me be clear that I =
> don't think the instrument is at fault but rather the problem lies on =
> the application (Dj dominate system measurement) that force the =
> instrument to report back a number that it cranked out based on some =
> fixed assumptions not suited for such situation.
> 
> For those who claim a), my hats off to you because you are a braver man =
> than me to publicly say that.
> 
> For those who claim b), I sure would like to know what kind of PLL you =
> have
> 
> And my money will be on c).
> 
> -----Original Message-----
> From: si-list-bounce@xxxxxxxxxxxxx
> [mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of steve weir
> Sent: Tuesday, July 04, 2006 3:40 PM
> To: David Instone
> Cc: si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Re: Fibre channel interconnect margins
> 
> 
> David, I did not regard it as an attack just an opinion that is=20
> different and worth discussing.
> 
> The basis of our disagreement appears to be in the definition of=20
> bound.  I look at things from the standpoint of electrical=20
> noise.  Time interval in a timing circuit is the result of the=20
> magnitude of some electrical quantity, and is always causal, each=20
> event defining a new interval follows the previous.  This means that=20
> noise effectively multiplies the interval by some factor   1/oo <=3D K=20
> <=3D oo.  Jitter is still unbounded, but every incremental interval has=20
> a positive duration.
> 
> So far we have been talking about noise in the oscillator itself.
> 
> Now, let's see what a PLL does to this quagmire.  If noise hammers=20
> the VCO then the PLL feedback loop applies gain to divide the effect=20
> of the noise.  If we still believe in infinity, infinity divided by=20
> anything is still infinity.  In practice will the oscillator stop for=20
> an unlimited time?  It will only when it fails.  On the other end,=20
> two successive pulses can occur essentially on top of each other.
> 
> A receiver PLL will take a finite amount of time to realign within a=20
> fixed amount of phase to the jittered stream for the case of the=20
> oscillator event, and will take a different, much longer amount of=20
> time to align to the short term frequency offset that noise in the=20
> PLL error amplifier causes.  The phase error between the source=20
> stream and the recovered clock in the latter case generally follows a=20
> classic 2nd order step response.  The golden PLL is a PLL with=20
> specific frequency response and damping.  Even if we have a PLL that=20
> uses N=3D1, the PLL only starts correcting after a timing error is=20
> already apparent.  For a timing error of sufficient magnitude data=20
> moves outside the timing window, a data recovery error is guaranteed,=20
> and no PLL is going to prevent that.  A nasty little problem that=20
> gets into systems is power supply noise coupled into the VCO and/or=20
> error amplifier by one means or another.  For systems with high Ns it=20
> can get really ugly.
> 
> On a slightly different tack, for a PLL using a PFD, the unit=20
> interval is that at the phase comparator input which is VCO/N or=20
> Fref.  Noise whacking the error amplifier will push the VCO off=20
> frequency until new information arrives to get it back.  If the noise=20
> jumps the VCO up it can take up to VCO/N cycles before we start=20
> correcting.  If noise slows the VCO down, it will take at least one=20
> cycle of Fref to get it back.
> 
> So, I think the only place that we are having any semantic trouble is=20
> on the notion of unbounded noise.  While we likely will never see=20
> such a thing, the math really does tell us that an interval can go=20
> virtually to zero 1/oo, or last forever.  I think the important part=20
> of this concept is that it says that random noise ( jitter ) will=20
> create data errors sooner or later.  And I think doubt about that is=20
> where the discussion began.  The tough issue is finding the actual=20
> random jitter.  The value is often way overestimated because=20
> deterministic jittter that we have difficulty correlating gets=20
> incorrectly classified as RJ.  People turn the crank on the math and=20
> conclude that their links are 10E-12 or 10E-14 when they are really=20
> more like 10E-20 from an RJ standpoint.
> 
> Regards,
> 
> 
> Steve.
> At 06:35 AM 7/4/2006, David Instone wrote:
> 
>>Steve,
>> Firstly, my initial response was in support of Alan's posting not=20
>>an attack on your reply to him.  Your definition follows that of FC=20
>>and other serial standards. FC defines random jitter in FC-PI-3 as
>>
>>>jitter, random (RJ): Jitter that is characterized by a Gaussian=20
>>>distribution. Random jitter is
>>>defined to be the peak-to-peak value for a BER of 10-12, taken to=20
>>>be approximately 14 times
>>>the standard deviation of the Gaussian distribution.
>>
>>
>> So lets look at it both ways
>>
>>>That means that any single incremental interval can never have=20
>>>jitter of more than -(1UI-epsilon).
>>
>>If that jitter is all Gaussian then hasn't it been truncated, or do=20
>>we have to say that it's not RJ because it's bounded?
>>
>>
>>
>>>If on the other hand we want to integrate phase compared to some=20
>>>distant fixed timing reference, then a stream can theoretically=20
>>>precess total time interval error by an unbounded amount.
>>
>>FC measures jitter against a timing reference derived from a golden=20
>>PLL.   If over any finite period of time the RJ causes the frequency=20
>>as seen by the PLL to change then the PLL will move the VCO, thus=20
>>creating a limit to the max observed RJ.  If the RJ is distributed=20
>>so that the frequency does not have to change then the 'single=20
>>incremental interval' effect will apply.
>> Have we not then got a jitter distribution that is Gaussian in=20
>>form but with limits to the maximum deviations?
>>
>>Regards
>>Dave
>>
>>steve weir wrote:
>>
>>>David,
>>>
>>>I would just like to make certain that we are talking along the=20
>>>same lines here.  The operation of the oscillator, no matter what=20
>>>its construction is causal.  So the closest that any two events can=20
>>>occur is epsilon.  That means that any single incremental interval=20
>>>can never have jitter of more than -(1UI-epsilon).
>>>
>>>If on the other hand we want to integrate phase compared to some=20
>>>distant fixed timing reference, then a stream can theoretically=20
>>>precess total time interval error by an unbounded amount.
>>>
>>>Regards,
>>>
>>>
>>>Steve.
>>>At 03:10 AM 7/4/2006, David Instone wrote:
>>>
>>>>Steve,
>>>>  I didn't disallow an infinite time between events.  I allow for=20
>>>>the time between events to be between 0 and infinity, but not =
> 
> negative.
> 
>>>>Thus if I'm measuring the time between edges and my reference I=20
>>>>can measure an infinite time between my reference and a following=20
>>>>edge but never more than 1 UI between the last edge and my reference.
>>>>That last edge could of course be from a edge that should have=20
>>>>occurred an infinite amount of time in the future, but from the=20
>>>>point of view of the measurement it's only 1 UI early.
>>>>Regards
>>>>Dave
>>>>
>>>>
>>>>steve weir wrote:
>>>>
>>>>>David, I disagree.  It does not change causality.  It changes the=20
>>>>>incremental delay between two events.  Imagine for a moment that=20
>>>>>we have a simple relaxation oscillator as the basis of our=20
>>>>>VCO.  In the presence of an infinitely large noise pulse, which=20
>>>>>is the limit for random noise, it takes an infinite amount of=20
>>>>>time for the ramp to reach the threshold.  The next cycle will=20
>>>>>not begin untilt he current cycle completes.  It may sound like=20
>>>>>something from Douglas Adams, but it really is mathematically and=20
>>>>>physically sound.
>>>>>
>>>>>Regards,
>>>>>
>>>>>Steve.
>>>>>At 01:50 AM 7/4/2006, David Instone wrote:
>>>>>
>>>>>>Because it makes for a nice simple clean definition.  However, I=20
>>>>>>believe you have to take the real world into consideration.
>>>>>>Allowing the RJ to be really unbounded means that each edge in a=20
>>>>>>bit stream could be advanced or delayed by an infinite amount.
>>>>>>Taken to extremes this means that  the order of  edges  could be =
> 
> reversed.
> 
>>>>>>This is obviously absurd, the measured time between edges can=20
>>>>>>reduce until it is zero, it cannot go negative.  The time=20
>>>>>>between edges can of course go to +ve infinity, but that isn't a=20
>>>>>>bit error, the system has failed or been switched off.
>>>>>>steve weir wrote:
>>>>>>
>>>>>>>RJ really is unbounded by definition.
>>>>>>>
>>>>>>>Steve.
>>>>>>>At 09:46 AM 7/3/2006, Steven Kan wrote:
>>>>>>>
>>>>>>>
>>>>>>>>>Date: Fri, 30 Jun 2006 21:48:56 -0700
>>>>>>>>>From: Alan.Hiltonnickel@xxxxxxx
>>>>>>>>>Subject: [SI-LIST] Re: Fibre channel interconnect margins
>>>>>>>>>
>>>>>>>>>In fact, I think that companies DO ship products that toss a =
> 
> random
> 
>>>>>>>>>error approximately every 10e-xx or so. Why? Because the =
> 
> statistical
> 
>>>>>>>>>theory behind those errors is that random/Gaussian noise is, by
>>>>>>>>>definition, unbounded - errors are a fact of life, even if the =
> 
> error
> 
>>>>>>>>>rate is very low.
>>>>>>>>
>>>>>>>>I suppose we're way off in the weeds, here, but is the noise =
> 
> actually
> 
>>>>>>>>unbounded? Or does it just behave in a Gaussian-like manner=20
>>>>>>>>within the realm
>>>>>>>>of times/rates that matter for shipping product? I suppose if=20
>>>>>>>>I sat in my
>>>>>>>>chair for long enough, a truly unbounded system might cause a=20
>>>>>>>>gold bar to
>>>>>>>>pop into existence on my desk, but my empirical GBR (gold-bar =
> 
> rate) is
> 
>>>>>>>>currently 0.
>>>>>>>>
>>>>>>>>------------------------------------------------------------------=
> 
> 
>>
> 
> ------------------------------------------------------------------
> To unsubscribe from si-list:
> si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
> 
> or to administer your membership from a web page, go to:
> //www.freelists.org/webpage/si-list
> 
> For help:
> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
> 
> List FAQ wiki page is located at:
>                 http://si-list.org/wiki/wiki.pl?Si-List_FAQ
> 
> List technical documents are available at:
>                 http://www.si-list.org
> 
> List archives are viewable at:     
>               //www.freelists.org/archives/si-list
> or at our remote archives:
>               http://groups.yahoo.com/group/si-list/messages
> Old (prior to June 6, 2001) list archives are viewable at:
>               http://www.qsl.net/wb6tpu
>   
> 
> 
> 

------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field

or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list

For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field

List FAQ wiki page is located at:
                http://si-list.org/wiki/wiki.pl?Si-List_FAQ

List technical documents are available at:
                http://www.si-list.org

List archives are viewable at:     
                //www.freelists.org/archives/si-list
or at our remote archives:
                http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
                http://www.qsl.net/wb6tpu
  

Other related posts: