Correspondence with Dominico De Seta

 This appendix reproduces correspondence with a Mr. Dominico De Seta of Fondatzione Ugo Bordoni. Mr De Seta was one of the authors of [14]. As has been mentioned previously, there is at the time of writing a contradiction in the standards in that it has been stated that the model has been used to derive the wander masks for individual clock types yet in the same document ([4]), it has been stated that the model parameters have been changed so that the masks are satisfied. When the modelling was originally carried out with parameters specified in [4] and [5] and were found to produce wander that lay above the masks, it was unclear whether the correct approach would be to propose new masks that satisfied the model or to change the model parameters so that the masks were satisfied. It was the correspondence reproduced here that was responsible for dictating the strategy that was followed.

Dear Mr De Seta,
You will be pleased to know that I now have simulations of one SEC and SSU running and the SSU falls below the prescribed wander mask. However I was wondering if I could possibly ask you a couple of more questions.

In your latest contribution to part 4 annex A you derived the expression

$$\tau_c=0.3\sqrt{1+\frac{1}{B^2}}$$

For an SEC of bandwidth 1Hz this is about 0.42s which is as I obtain by my simulation. Given this value I do not understand how a 1-10Hz SEC can possibly satisfy the mask in part 4 which assumes that this value is 25s. I am also very confused by the working document you submitted to the Vienna meeting, entitled ``On the choice of SSU/SEC model parameters in annexes C of DE/TM-3017-4 and DE/TM-3017-5''. In this document you propose that the maximum flicker phase gain satisfying part 5 should be 2977 then use a value of 5953 for your simulation on the last page. Your simulation depicted in fig. 4 of the document, (like mine) also does not satisfy the wander mask that it purports to define as the flicker phase behaviour of your model looks like it begins at about 0.42s as predicted by your expression for $\tau_c$.

This leads me on to another question. I was of the impression that the motivation was that the model was being used to derive the wander masks in locked mode. The standards certainly seem to imply this: ``The model used to derive these numbers is described in (informative) Annex A''. However the model has been modified so it obeys the masks that it itself has supposedly derived. I am unsure if I have missed something but to me this does not seem to make sense. If the model is modified to obey the masks, what is it modelling? Or what is the value of the masks if the model deriving them has been changed?

I would be very interested in your insight on the above matters. If it would be possible for you to email or fax me I would be most grateful.

With Regards,

Mark Ivens

From deseta@fub.it Mon Aug  5 10:36:19 1996
Mime-Version: 1.0
Date: Mon, 5 Aug 1996 11:40:56 +0200
To: Mark Ivens <eemt10@eleceng.ucl.ac.uk>
From: deseta@fub.it (Domenico De Seta)
Subject: Re: choice of f0=5microHz

>Dear Mr. De Seta,
> I was wondering if I could ask you another quick question. Why was the
>particular value of f0=5microHz chosen? As always I would be very grateful
>for any help you could give.
>You will be pleased to know that my simulation is going well and I have
>nearly implemented the noise model apart from a problem I have with filter
>A.
>Regards
>Mark Ivens


Dr Mark:

back to my office after being away for two weeks I found both your e-mail
and fax: I understand your doubts, which I hope to solve in a moment.

First, as far as the value of 5 microHz is concerned my argument is that
such a value is well below the lower frequency limit in the  MTIE and TDEV
masks for wander generation: so, even though flicker noise should
theoretically be modelled up to zero frequency (!) the flickering filter
given in Annex C yields a practical good approximation whose results cannot
be distinguished from those obtained using a lower limit for f0.


As to your comments on  parameter values for modelling noise according to
TDEV mask in part 5 (that is for a SEC clock), the problem with such a mask
is that it cannot precisely be satisfied (i.e. following its shape) by the
model assumed in Annex C if bandwidths in the range of 1 to 10 Hz have to
be used: this is clearly shown in fig. 4 of my contribution to the Vienna
meeting, where trying to meet the long term TDEV  value of 6.4 ns (in the
FPM dominating region) makes it impossible to stay below 3.2 ns at
observation interval less than 25 seconds (this could be achievable with a
much smaller bandwidth, in the order of a few mHz). Therefore, the only way
to manage with both the given mask and the assumed model is to set the GI
parameter at a value less than 2977 such that the 3.2 ns limit holds also
in the long term region: in this case the bandwidth can be set  at values
even higher than 10 Hz without affecting the capability of the model to
"meet" the TDEV requirements.

This leads me to give you an answer to the last questions concerning the
entire meaning of our modelling: the TDEV or MTIE masks in the standards
are the result of long discussions  among network operators, manufacturers,
researchers and possibly other actors: hence, they take into account many
different, sometimes conflicting, exigencies. The agreed common
understanding, though, is that it shouldn't be required of
telecommunication equipment more than is necessary to achieve the desired
overall quality of the network. This means that often requirements can be
fulfilled with different choices in the implementation of equipment, with
different margins in different operating conditions. In such a situation it
can be very difficult to devise models for representing a multiplicity of
implementations, unless some degree of mismatch between model behaviour and
actual equipment behaviour is allowed. This seems to be the case for the
specification of SEC performance.

I hope to have been helpful: let me know how your work is progressing. Best
wishes.


Domenico

Domenico de Seta
Q21/13 Rapporteur

Fondazione Ugo Bordoni, Roma, Italy
Phone: +396 5480 3311
Fax:   +396 5480 4403