Re: [EFM] Modulation and Digital Duplexing (DD)
Behrooz,
indeed, less ADC bits didn't cut the transmission, but cut the performance
just because the quantization noise grows to high. For short reach (high speeds)
this has much less impact than for long reach, because FEXT is very high and
masks quantization noise. For long reach all changes: FEXT is attenuated by the
loop and quantization noise turns to be the main issue. So, it doesn't surprise
me that people can demonstrate digital duplexing even today.
Regarding orthogonality, I think that both SCM and DMT are "almost
orthogonal". In SCM that's because the leakage, in DMT due to timing issues
(except we assume that all if fully synchronize). I still think the
orthogonality is not really an issue.
Regarding your question - we are definitely going to implement digital
duplexing in SCM at the moment ADC became cheap enough. It is just the same as
for today EC systems (HDSL, ISDN, SHDSL, HDSL2) which use digital EC duplexing.
Vladimir
Behrooz Rezvani wrote:
> Vladimir,
>
> we looked at some of the points you brought up as well as some others and
> evaluated the problem from all angles and we concluded that it is still
> better to do digital duplexing. By the way we are doing digital dulexing
> with a lot lower number of A/D bits and we are getting very high performance
> level. In DMT, digital duplexing does not add anything to the existing cost.
> The DAC or ADC resolution because of it does not increase and no extra
> hardware is required.
>
> Digital Duplexing and Digital filtering have some differences. In digital
> duplexing, the signals are orthogonal whereas in digital filtering they are
> all most orthogonal.
> There are some merits to do it this way.
> 1) the efficient use of frequency, because of orthogonality property of
> FFT/IFFT there is very little loss in guard bands.
> 2) Flexibility, there is no added filtering for Tx and bands can be easily
> divided in any way you choose,
>
> Do you think you digital duplexing makes sence for SCM?
> In any event There are very efficient ways of implementing digital duplexing
> in DMT. This does not mean that QAM needs to do that. I am sure digital
> filtering can be used but I think they have different role
>
> Kind regards
> Behrooz
>
> ----- Original Message -----
> From: "Vladimir Oksman" <voksman@xxxxxxxxxxxx>
> To: "Behrooz Rezvani" <brezvani@xxxxxxxxxx>
> Cc: <aidan@xxxxxxxxxxxx>; "'Wei, Dong'" <wei@xxxxxxxxxxx>;
> <daun@metanoia-technologies.com>; <stds-802-3-efm-copper@ieee.org>;
> <stds-802-3-efm@ieee.org>; "'Zagalsky, Nelson'" <Nelson_Zagalsky@adc.com>
> Sent: Monday, July 08, 2002 8:20 AM
> Subject: Re: [EFM] Modulation and Digital Duplexing (DD)
>
> > Behrooz,
> >
> > the upstream and downstream carriers are always almost orthogonal
> > otherwise one cannot separate them. The grade of this "almost" in the case
> of
> > FDD both MCM and SCM currently use depends on filtering technique. If we
> assume
> > that ADC and DAC has unlimited resolution, there could be always found a
> > suitable bandpass filter to provide small enough crosstalk between
> upstream and
> > downstream. DMT is doing that using FFT and windowing as a filter, SCM
> usually
> > uses digital filters. I don't know exactly which are more complex, but for
> now
> > neither is a problem of digital duplexing we really face.
> >
> > The today problem is lack of resolution of ADC and DAC to deal with
> echo
> > (assuming relevant implementation for the hybrid). Thus all mostly
> depends on
> > the dynamic range of the transmit signal, where SCM due to small number of
> bands
> > usually has 5-6 dB advantage. You can see some calculations for ADC/DAC
> > requirements for DMT in T1E1.4 contributions (Ottawa, 1999).
> >
> > Kind regards,
> >
> > Vladimir
> >
> > Behrooz Rezvani wrote:
> >
> > > Aiden,
> > >
> > > Thanks for touching on the subject of Digital Duplexing and I would like
> to
> > > add few words to it. As you correctly pointed out Digital Duplexing is
> > > independent of modulation and I hope to see the use of this technique
> for
> > > PHY devices to become more readily available.
> > >
> > > On the technical issue I think you covered the "necessary" condition and
> I
> > > would like to discuss the "sufficient" condition in order to have
> digital
> > > duplexing done
> > >
> > > I will focus mostly on DMT base modulation and may be you may want to
> take a
> > > crack at QAM based modulation on this is accomplished.
> > >
> > > In DMT based modulation one has many sub-carriers similar to OFDM based
> > > techniques of 802.11a. The sub-carrier bandwidth is typically 4.135 KHz
> (and
> > > other options also exist) The number of these sub-carriers (or tones)
> can be
> > > as much as 4196. Digital Duplexing (DD)in DMT fundamentally works on
> the
> > > principal of orthogonalizing the downstream transmission to upstream
> > > transmission. Remember that having large dynamic range A/D or D/A is not
> > > sufficient to perform DD. In VDSL-dmt frame there is function that adds
> > > Cyclic Prefix or Suffix to the symbol itself to allow the movement the
> > > downstream waveform wrt upstream waveform. During the timing recovery
> and
> > > training mode symbol boundary is obtained using x-correlation methods.
> Based
> > > on these information then upstream and downstream waveform are
> synchronized
> > > to each other. Cyclic prefix and suffix are used to adjust the waveforms
> in
> > > time domain. (This is somewhat similar to ranging algorithm used in PON
> type
> > > of network where based on the distance of a CPE and CO a time reference
> is
> > > obtained)
> > >
> > > Upon completion of this task then ANY of the tones can transmit in
> upstream
> > > or downstream direction. In other words one can have as many bandplans
> that
> > > one can imagine [combinatory problem: Normally there is no need for
> large
> > > number of bandplans. In VDSL we have 4-bands + an optional band. This is
> > > however designed to address a large population of loops with one shot.
> If
> > > one allows more of local optimization view then one can imagine the need
> for
> > > other band plans or more bands]
> > >
> > > I assume this can be done in QAM based methods, however the number of
> bands
> > > are limited to the number of modulated carriers.
> > >
> > > Best
> > > Behrooz
> > >
> > > ----- Original Message -----
> > > From: "Aidan O'Rourke" <aidan@xxxxxxxxxxxx>
> > > To: "'Behrooz Rezvani'" <brezvani@xxxxxxxxxx>; "'Wei, Dong'"
> > > <wei@xxxxxxxxxxx>; <daun@xxxxxxxxxxxxxxxxxxxxxxxxx>
> > > Cc: <stds-802-3-efm-copper@ieee.org>; <stds-802-3-efm@ieee.org>;
> "'Zagalsky,
> > > Nelson'" <Nelson_Zagalsky@xxxxxxx>
> > > Sent: Tuesday, July 02, 2002 10:49 AM
> > > Subject: RE: [EFM-Copper] RE: [EFM] (forward) progress in EFM copper
> > >
> > > > Behrooz,
> > > >
> > > > In an attempt to provide some clarification I would like to point out
> to
> > > > members of the reflector, that digital duplexing and line coding are
> > > > entirely separate issues.
> > > >
> > > > Digital duplexing is the concept of digitizing a region of spectrum
> and
> > > > employing digital filtering techniques to ensure that out of band
> signal
> > > and
> > > > noise is not applied to the receiving demodulator. Whether or not this
> > > > demodulator employs FFT or QAM demodulation is an entirely separate
> > > matter.
> > > >
> > > > The main requirement for digital duplexing is that you have an A to D
> > > > converter that has enough dynamic range and low enough quantization
> noise
> > > to
> > > > enable digital techniques to be employed. I am not aware of any
> advantage
> > > > that DMT has in its ability to be digitized more readily, or with any
> less
> > > > complexity than single carrier. If you are aware of these advantages I
> > > would
> > > > be very interested to learn more.
> > > >
> > > >
> > > > Best Regards,
> > > >
> > > > Aidan.
> > > >
> > > >
> >
> >