Reply by glen herrmannsfeldt February 22, 20122012-02-22
Jerry Avins <jya@ieee.org> wrote:

(snip, I wrote)

>> I haven't thought of this for a while, but I believe gigabit >> ethernet is also phase locked. The two ends agree on who will >> supply the clock that is used.
> I can see a frequency lock, but phase will vary with distance > along the line.
For 10baseT, all transmission is asynchronous. There is a preamble, and it is usual to use that to synchronize a PLL in the receiver. 100baseTX uses continuous signaling, where once the link is up, bits are continuously sent. Both ends have a crystal for the transmitter, the receiver locks onto that. I believe for 1000baseT that only one end uses its crystal. It is again continuous signaling. Note also that with different twist pitch the delay among the four pairs will be different. The receiver on each pair has to adjust for the signal on that pair. But okay, one end supplies the reference clock, the other end lock, likely with a PLL, to that clock. The link only has two ends (there is nothing like the tapped coax cable used for original ethernet). Once the link is up, there will be a fixed phase difference between the two ends, unless the length of the wire changes. (or its dielectric constant). Now, if you ask in the special relativity sense, someone going by on a rocket will notice a different phase between the two ends, (and also a different clock rate). If you power off the system, then on again with the same wiring, the relative phase will likely be the same. Close enough to phase locked for me. (Most likely with actual PLLs.) -- glen
Reply by Jerry Avins February 22, 20122012-02-22
On 2/21/2012 8:53 PM, glen herrmannsfeldt wrote:
> Vladimir Vassilevsky<nospam@nowhere.com> wrote: > > (snip) >> There is a class of zero IF transceivers which use output stage as input >> mixer at the same time in full duplex operation. Both sides of the link >> are phase locked to one another. This approach is simple and it works at >> low power levels. > > I haven't thought of this for a while, but I believe gigabit > ethernet is also phase locked. The two ends agree on who will > supply the clock that is used.
I can see a frequency lock, but phase will vary with distance along the line. Jerry -- Engineering is the art of making what you want from things you can get. &#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;
Reply by Eric Jacobsen February 21, 20122012-02-21
On Tue, 21 Feb 2012 18:28:33 -0600, "steveu"
<steveu@n_o_s_p_a_m.coppice.org> wrote:

>>On Tue, 21 Feb 2012 10:54:08 -0600, "steveu" >><steveu@n_o_s_p_a_m.coppice.org> wrote: >> >>>>On Tue, 21 Feb 2012 06:44:50 -0800 (PST), maury <maury001@core.com> >>>>wrote: >>>> >>>>>On Feb 21, 7:06=A0am, "steveu" <steveu@n_o_s_p_a_m.coppice.org> wrote: >>>>>> >On Feb 19, 8:18=3DA0am, "Fender123" >>><joeborg123@n_o_s_p_a_m.excite.com> >>>>>> >wrote: >>>>>> >> Reading up a 1988 edition of Lee & Messerschmitt's "Digital >>>>>> Communication=3D >>>>>> >", >>>>>> >> I came across this gem (Ch18 pg 643): >>>>>> >>>>>> >> "In principle, echo cancellation could be used to share any >medium, >>>su= >>>>>ch >>>>>> =3D >>>>>> >as >>>>>> >> a radio channel, for the two directions... would be of great >>>practical >>>>>> >> interest... but unfortunately is impractical in today's >>>technology..." >>>>>> >>>>>> >> And the punch-line: "However, we cannot rule it out for the >>>future". >>>>>> >>>>>> >> Just how far are we in 2012? Any takes on when and how this might >>>>>> become >>>>>> >> reality? Can MIMO be considered a more practically realisable >form >>>of >>>>>> >> this? >>>>>> >>>>>> >> Brian. >>>>>> >>>>>> >Clearly echo cancellation can be used to facilate full duplex >>>>>> >communications. It's done every day over the land-based telephone >>>>>> >network, and was being done in 1988. The problem with radio was the >>>>>> >scaling of the frequency. I.e, how to run the technology of 1988 at >>>>>> >radio frequency (HF and higher) speeds. That, too, seems to have >been >>>>>> >largely solved with A/Ds running in the GHz. >>>>>> >>>>>> There is a huge difference between echo cancelling a relatively >stable >>>PS= >>>>>TN >>>>>> channel and an endlessly changing radio channel. >>>>>> >>>>>> Steve- Hide quoted text - >>>>>> >>>>>> - Show quoted text - >>>>> >>>>>Changing channels is not new to echo cancellation. It's called >>>>>reconvergence. Any "good" acoustic echo canceller made to operate in a >>>>>closed area (room) must contend with changing characteristics (the >>>>>room impulse response changes as people move around). Echo cancellers >>>>>for DSL must also contend with changing channels. >>>>> >>>>>The question was can it be done. >>>> >>>>I've been struggling with what is meant by "echo canceller" in this >>>>context. I'm getting a sense of deja vu as I think this has been >>>>discussed before in comp.dsp and I had the same confusion. >>>> >>>>Just cancelling the channel echos/reflections is routinely done with >>>>an equalizer, and I know you know that. So I don't think that's it. >>>> >>>>If you mean cancelling the transmit signal so that the same channel >>>>can be used in both directions, that can be done, too, and is done in >>>>some modern applications. I recall the previous time this came up it >>>>involved a discussion of telephone modem hybrids, which did that job >>>>for that application, so that the digital "echo canceller" only had to >>>>deal with equalizing the reflections in the channel. >>> >>>Telephone hybrids do little more than keep the returned signal below >>>howling. The main job of a digital echo canceller on a telephone line is >to >>>cancel the 4 echoes from the 4 poor quality hybrids you typically have >>>between 2 communicating parties. In fact, for really long international >>>echoes, especially for modems, they insert 4 short time domain >cancellers >>>where the 4 echoes occur. >> >>Yes, the hybrid cancels the primary transmitted power, the "echo >>canceller" then equalizes the channel distortion due to reflections >>from impedance mismatches in the far hybrids. > >Saying a hybrid cancels the primary transmit power is like saying that >someone shouting in my ear is cancelled if I just put a bit of thin sponge >between us. :-)
Nevertheless, that's what a hybrid does.
>Calling a line canceller an equalizer is a strange way of looking at it. An >equalizer unjumbles a number of distorted versions of the same signal. A >canceller removes an independent signal that's screwing up the one of >interest.
Echos are reflections, which is essentially what a channel equalizer corrects. You seem to be talking about interference, which I think isnt' within the same context of the current thread.
>>>>Is there another meaning that I'm missing? > >Steve >
Eric Jacobsen Anchor Hill Communications www.anchorhill.com
Reply by glen herrmannsfeldt February 21, 20122012-02-21
Vladimir Vassilevsky <nospam@nowhere.com> wrote:

(snip)
> There is a class of zero IF transceivers which use output stage as input > mixer at the same time in full duplex operation. Both sides of the link > are phase locked to one another. This approach is simple and it works at > low power levels.
I haven't thought of this for a while, but I believe gigabit ethernet is also phase locked. The two ends agree on who will supply the clock that is used. -- glen
Reply by glen herrmannsfeldt February 21, 20122012-02-21
Jerry Avins <jya@ieee.org> wrote:

(snip on full duplex communication)

>> It is also used by gigabit ethernet, which sends in both directions >> on all four pairs in a Cat 5 cable at the same time.
> Thomas Edison was issued a patent for full-duplex telegraph > signaling in 1892. http://www.google.com/patents/US480567
It does get harder at higher frequencies, especially meeting the RFI requirements. The attenuation is pretty high at gigabit frequencies, so it has to see a much smaller signal coming in on the same pair that it is sending on. Fairly small impedance discontinuities along the way, such as connectors, will also generate reflections. -- glen
Reply by steveu February 21, 20122012-02-21
>On Tue, 21 Feb 2012 10:54:08 -0600, "steveu" ><steveu@n_o_s_p_a_m.coppice.org> wrote: > >>>On Tue, 21 Feb 2012 06:44:50 -0800 (PST), maury <maury001@core.com> >>>wrote: >>> >>>>On Feb 21, 7:06=A0am, "steveu" <steveu@n_o_s_p_a_m.coppice.org> wrote: >>>>> >On Feb 19, 8:18=3DA0am, "Fender123" >><joeborg123@n_o_s_p_a_m.excite.com> >>>>> >wrote: >>>>> >> Reading up a 1988 edition of Lee & Messerschmitt's "Digital >>>>> Communication=3D >>>>> >", >>>>> >> I came across this gem (Ch18 pg 643): >>>>> >>>>> >> "In principle, echo cancellation could be used to share any
medium,
>>su= >>>>ch >>>>> =3D >>>>> >as >>>>> >> a radio channel, for the two directions... would be of great >>practical >>>>> >> interest... but unfortunately is impractical in today's >>technology..." >>>>> >>>>> >> And the punch-line: "However, we cannot rule it out for the >>future". >>>>> >>>>> >> Just how far are we in 2012? Any takes on when and how this might >>>>> become >>>>> >> reality? Can MIMO be considered a more practically realisable
form
>>of >>>>> >> this? >>>>> >>>>> >> Brian. >>>>> >>>>> >Clearly echo cancellation can be used to facilate full duplex >>>>> >communications. It's done every day over the land-based telephone >>>>> >network, and was being done in 1988. The problem with radio was the >>>>> >scaling of the frequency. I.e, how to run the technology of 1988 at >>>>> >radio frequency (HF and higher) speeds. That, too, seems to have
been
>>>>> >largely solved with A/Ds running in the GHz. >>>>> >>>>> There is a huge difference between echo cancelling a relatively
stable
>>PS= >>>>TN >>>>> channel and an endlessly changing radio channel. >>>>> >>>>> Steve- Hide quoted text - >>>>> >>>>> - Show quoted text - >>>> >>>>Changing channels is not new to echo cancellation. It's called >>>>reconvergence. Any "good" acoustic echo canceller made to operate in a >>>>closed area (room) must contend with changing characteristics (the >>>>room impulse response changes as people move around). Echo cancellers >>>>for DSL must also contend with changing channels. >>>> >>>>The question was can it be done. >>> >>>I've been struggling with what is meant by "echo canceller" in this >>>context. I'm getting a sense of deja vu as I think this has been >>>discussed before in comp.dsp and I had the same confusion. >>> >>>Just cancelling the channel echos/reflections is routinely done with >>>an equalizer, and I know you know that. So I don't think that's it. >>> >>>If you mean cancelling the transmit signal so that the same channel >>>can be used in both directions, that can be done, too, and is done in >>>some modern applications. I recall the previous time this came up it >>>involved a discussion of telephone modem hybrids, which did that job >>>for that application, so that the digital "echo canceller" only had to >>>deal with equalizing the reflections in the channel. >> >>Telephone hybrids do little more than keep the returned signal below >>howling. The main job of a digital echo canceller on a telephone line is
to
>>cancel the 4 echoes from the 4 poor quality hybrids you typically have >>between 2 communicating parties. In fact, for really long international >>echoes, especially for modems, they insert 4 short time domain
cancellers
>>where the 4 echoes occur. > >Yes, the hybrid cancels the primary transmitted power, the "echo >canceller" then equalizes the channel distortion due to reflections >from impedance mismatches in the far hybrids.
Saying a hybrid cancels the primary transmit power is like saying that someone shouting in my ear is cancelled if I just put a bit of thin sponge between us. :-) Calling a line canceller an equalizer is a strange way of looking at it. An equalizer unjumbles a number of distorted versions of the same signal. A canceller removes an independent signal that's screwing up the one of interest.
>>>Is there another meaning that I'm missing?
Steve
Reply by Eric Jacobsen February 21, 20122012-02-21
On Tue, 21 Feb 2012 14:56:53 -0600, "Fender123"
<joeborg123@n_o_s_p_a_m.excite.com> wrote:

>Hi all >I asked my question in the context of sharing an RF channel in both >directions at the same time. > >That is: without the usual FDD, TDD, CDD etc. Maybe we can call it DDD, >direction division duplex ;-) > >While the issues are 'well understood', I have to agree with SteveU that >going from wireline to wireless, the problem is in a 'whole different >class' from a processing point of view.
As long as one understands the constraints posed by the application, it may not be that much different from wireline. This has been done in satellite communications for some time: http://www.comtechefdata.com/technologies/doubletalk.asp
>One can safely assume that the massive processing is doable, see cellular >soft-modems form ICERA/Cognovo etc. The basics are there. > >With respect to wireline, would need faster converging algorithms and some >retraining bursts, but I suppose these are doable too. > >I also agree with MarkusN that the main problem is the AFE. With >traditional thinking, TX/RX on the same frequency is not practical. Even >with a very good diplexer, I would imagine the ADC resolution needed would >be challenging, unless the main echo component can be cancelled by analog >electronics to limit the dynamic range needed.
Again, depending on the system constraints for the particular application, it can range from manageable to not affordable.
>I'm surprised no one commented on MIMO. It's not quite the same thing, but >in my mind it's related. It also creates 'pipes' on the same frequency, >it's just that they all point one way. > >Brian.
I don't consider MIMO comparable in this context. In MIMO the channel reuse is in the same direction and the Tx cancelling problem does not exist in the receiver. Eric Jacobsen Anchor Hill Communications www.anchorhill.com
Reply by Eric Jacobsen February 21, 20122012-02-21
On Tue, 21 Feb 2012 10:54:08 -0600, "steveu"
<steveu@n_o_s_p_a_m.coppice.org> wrote:

>>On Tue, 21 Feb 2012 06:44:50 -0800 (PST), maury <maury001@core.com> >>wrote: >> >>>On Feb 21, 7:06=A0am, "steveu" <steveu@n_o_s_p_a_m.coppice.org> wrote: >>>> >On Feb 19, 8:18=3DA0am, "Fender123" ><joeborg123@n_o_s_p_a_m.excite.com> >>>> >wrote: >>>> >> Reading up a 1988 edition of Lee & Messerschmitt's "Digital >>>> Communication=3D >>>> >", >>>> >> I came across this gem (Ch18 pg 643): >>>> >>>> >> "In principle, echo cancellation could be used to share any medium, >su= >>>ch >>>> =3D >>>> >as >>>> >> a radio channel, for the two directions... would be of great >practical >>>> >> interest... but unfortunately is impractical in today's >technology..." >>>> >>>> >> And the punch-line: "However, we cannot rule it out for the >future". >>>> >>>> >> Just how far are we in 2012? Any takes on when and how this might >>>> become >>>> >> reality? Can MIMO be considered a more practically realisable form >of >>>> >> this? >>>> >>>> >> Brian. >>>> >>>> >Clearly echo cancellation can be used to facilate full duplex >>>> >communications. It's done every day over the land-based telephone >>>> >network, and was being done in 1988. The problem with radio was the >>>> >scaling of the frequency. I.e, how to run the technology of 1988 at >>>> >radio frequency (HF and higher) speeds. That, too, seems to have been >>>> >largely solved with A/Ds running in the GHz. >>>> >>>> There is a huge difference between echo cancelling a relatively stable >PS= >>>TN >>>> channel and an endlessly changing radio channel. >>>> >>>> Steve- Hide quoted text - >>>> >>>> - Show quoted text - >>> >>>Changing channels is not new to echo cancellation. It's called >>>reconvergence. Any "good" acoustic echo canceller made to operate in a >>>closed area (room) must contend with changing characteristics (the >>>room impulse response changes as people move around). Echo cancellers >>>for DSL must also contend with changing channels. >>> >>>The question was can it be done. >> >>I've been struggling with what is meant by "echo canceller" in this >>context. I'm getting a sense of deja vu as I think this has been >>discussed before in comp.dsp and I had the same confusion. >> >>Just cancelling the channel echos/reflections is routinely done with >>an equalizer, and I know you know that. So I don't think that's it. >> >>If you mean cancelling the transmit signal so that the same channel >>can be used in both directions, that can be done, too, and is done in >>some modern applications. I recall the previous time this came up it >>involved a discussion of telephone modem hybrids, which did that job >>for that application, so that the digital "echo canceller" only had to >>deal with equalizing the reflections in the channel. > >Telephone hybrids do little more than keep the returned signal below >howling. The main job of a digital echo canceller on a telephone line is to >cancel the 4 echoes from the 4 poor quality hybrids you typically have >between 2 communicating parties. In fact, for really long international >echoes, especially for modems, they insert 4 short time domain cancellers >where the 4 echoes occur.
Yes, the hybrid cancels the primary transmitted power, the "echo canceller" then equalizes the channel distortion due to reflections from impedance mismatches in the far hybrids.
>>Is there another meaning that I'm missing? >> >> >>Eric Jacobsen > >Steve > >>Anchor Hill Communications >>www.anchorhill.com >>
Eric Jacobsen Anchor Hill Communications www.anchorhill.com
Reply by Vladimir Vassilevsky February 21, 20122012-02-21

Fender123 wrote:

> With > traditional thinking, TX/RX on the same frequency is not practical.
There is a class of zero IF transceivers which use output stage as input mixer at the same time in full duplex operation. Both sides of the link are phase locked to one another. This approach is simple and it works at low power levels. VLV
Reply by Fender123 February 21, 20122012-02-21
Hi all
I asked my question in the context of sharing an RF channel in both
directions at the same time. 

That is: without the usual FDD, TDD, CDD etc. Maybe we can call it DDD,
direction division duplex ;-)

While the issues are 'well understood', I have to agree with SteveU that
going from wireline to wireless, the problem is in a 'whole different
class' from a processing point of view. 

One can safely assume that the massive processing is doable, see cellular
soft-modems form ICERA/Cognovo etc. The basics are there.

With respect to wireline, would need faster converging algorithms and some
retraining bursts, but I suppose these are doable too.

I also agree with MarkusN that the main problem is the AFE. With
traditional thinking, TX/RX on the same frequency is not practical. Even
with a very good diplexer, I would imagine the ADC resolution needed would
be challenging, unless the main echo component can be cancelled by analog
electronics to limit the dynamic range needed.

I'm surprised no one commented on MIMO. It's not quite the same thing, but
in my mind it's related. It also creates 'pipes' on the same frequency,
it's just that they all point one way.

Brian.