Forums

Digital Modulation Scheme

Started by Charvee December 3, 2007
Hi! 

I want to ask what is the best digital modulation scheme used for high
speed telephone modems?.

Thanks.

Charvee


On Dec 3, 4:01 pm, "Charvee" <char...@kfffoundation.com> wrote:
> Hi! > > I want to ask what is the best digital modulation scheme used for high > speed telephone modems?. > > Thanks. > > Charvee
Hello Charvee, That is a pretty all encompassing question. If you are talking about modems over an analog phone line, V.34 was pretty good - it got up to 33.8kBit/sec. It used QAM with TCM (trellis coded modulation). But then somebody (I think at U.S. Robotics) realized that the bandwidth from the CO to the end user had a bandwidth much greater than the simple 3.7kHz normally associated with analog phone lines, so by letting the modem in the home train to the actual PCM levels from the CO, a 56kbit/sec modem could be realized. Yes the link is 64kbit/sec but with robbed bit signalling (T1) all that survived was 56kbit/sec. And yes with this scheme the uplink speed is lower. Similar lines could be used for T1 and DSL which each use forms of line coding. Hopefully you can narrow the focus of your question. Clay
Clay <physics@bellsouth.net> writes:
> [...] > But then somebody (I think at U.S. Robotics) realized that the > bandwidth from the CO to the end user had a bandwidth much greater > than the simple 3.7kHz normally associated with analog phone lines, so > by letting the modem in the home train to the actual PCM levels from > the CO, a 56kbit/sec modem could be realized.
I think heard somewhere a long while back (10 years?) that it was Robert Lucky who conceived of this. DSP history, anyone? -- % Randy Yates % "She tells me that she likes me very much, %% Fuquay-Varina, NC % but when I try to touch, she makes it %%% 919-577-9882 % all too clear." %%%% <yates@ieee.org> % 'Yours Truly, 2095', *Time*, ELO http://www.digitalsignallabs.com
Clay wrote:

(snip)

> That is a pretty all encompassing question. If you are talking about > modems over an analog phone line, V.34 was pretty good - it got up to > 33.8kBit/sec. It used QAM with TCM (trellis coded modulation). But > then somebody (I think at U.S. Robotics) realized that the bandwidth > from the CO to the end user had a bandwidth much greater than the > simple 3.7kHz normally associated with analog phone lines, so by > letting the modem in the home train to the actual PCM levels from the > CO, a 56kbit/sec modem could be realized.
It is more than that. The demodulation has to be synchronous so it can detect the actual levels at the appropriate time. It also requires only one D/A conversion and no A/D conversion in the path. That is, the ISP has a digital connection to the CO.
> Yes the link is 64kbit/sec > but with robbed bit signalling (T1) all that survived was 56kbit/sec. > And yes with this scheme the uplink speed is lower. Similar lines > could be used for T1 and DSL which each use forms of line coding. > Hopefully you can narrow the focus of your question.
Well, that and that with the mu-law or A-law encoding the higher code points are at a very high level. They could easily cause crosstalk. -- glen
glen herrmannsfeldt wrote:
> Clay wrote: > > (snip) > >> That is a pretty all encompassing question. If you are talking about >> modems over an analog phone line, V.34 was pretty good - it got up to >> 33.8kBit/sec. It used QAM with TCM (trellis coded modulation). But >> then somebody (I think at U.S. Robotics) realized that the bandwidth >> from the CO to the end user had a bandwidth much greater than the >> simple 3.7kHz normally associated with analog phone lines, so by >> letting the modem in the home train to the actual PCM levels from the >> CO, a 56kbit/sec modem could be realized. > > It is more than that. The demodulation has to be synchronous so it > can detect the actual levels at the appropriate time. It also > requires only one D/A conversion and no A/D conversion in the > path. That is, the ISP has a digital connection to the CO. > >> Yes the link is 64kbit/sec >> but with robbed bit signalling (T1) all that survived was 56kbit/sec. >> And yes with this scheme the uplink speed is lower. Similar lines >> could be used for T1 and DSL which each use forms of line coding. >> Hopefully you can narrow the focus of your question. > > Well, that and that with the mu-law or A-law encoding the higher > code points are at a very high level. They could easily cause crosstalk.
There are a surprising number of gain adjustments taking place in the digital network. Its surprising so many adjustments occur, and its surprising they so freely accept the quality loss associated with each adjustment. I understand these gain controls caused the early developers of 56k modems considerable trouble. Steve
Clay wrote:
> On Dec 3, 4:01 pm, "Charvee" <char...@kfffoundation.com> wrote: >> Hi! >> >> I want to ask what is the best digital modulation scheme used for high >> speed telephone modems?. >> >> Thanks. >> >> Charvee > > Hello Charvee, > > That is a pretty all encompassing question. If you are talking about > modems over an analog phone line, V.34 was pretty good - it got up to > 33.8kBit/sec. It used QAM with TCM (trellis coded modulation). But
Simply using QAM and TCM only got to V.32bis at 14,400bps. It required several additional tricks to get to V.34 at 28,800bps and then V.34bis at 33,600bps. Rounding the constellation, to avoid high signal level sticky out corners was a key part of that. The ulaw/alaw codecs in the exchange impose hard limits on the dynamic range of the signal.
> then somebody (I think at U.S. Robotics) realized that the bandwidth > from the CO to the end user had a bandwidth much greater than the > simple 3.7kHz normally associated with analog phone lines, so by > letting the modem in the home train to the actual PCM levels from the > CO, a 56kbit/sec modem could be realized. Yes the link is 64kbit/sec > but with robbed bit signalling (T1) all that survived was 56kbit/sec. > And yes with this scheme the uplink speed is lower. Similar lines > could be used for T1 and DSL which each use forms of line coding. > Hopefully you can narrow the focus of your question.
If you look at the V.90 spec, the speed limits appear to be based not on what the line is capable of, but on what the regulations allow for maximum power on the line. They defined higher rates than 56k, but even E1 systems don't use those because of power level issues. Steve