Reply by steveu August 13, 20092009-08-13
>On Aug 11, 10:15=A0pm, "steveu" <ste...@coppice.org> wrote: >> >> Look at Figure 6.3-4. It is based on a channel of unconstrained
bandwidth=
>, >> and the early late gate approach is based on the symmetry you see there
(=
>as >> the text clearly states), trying to home on the peaks. When the
bandwidth
>> is constrained by nature (e.g. a telephone line) or regulation (e.g.
most
>> radio communications) things don't look anything like that diagram.
You
>> have a signal point on a constellation map that is slewing almost
directl=
>y >> from symbol point to symbol point. For a simple modulation, like QPSK, >> things may still converge. However, even a small QAM constellation has >> little chance of converging in a reliable way. >> >> Steve > >I call this under-appreciation of synchronization issues "job >security" ;-). > >Julius
The secret brotherhood of the bitstream wouldn't let me join. Why shouldn't I give away the precious secrets they'll never put in a book? Steve
Reply by Vladimir Vassilevsky August 12, 20092009-08-12

julius wrote:
> On Aug 11, 10:15 pm, "steveu" <ste...@coppice.org> wrote: > >>Look at Figure 6.3-4. It is based on a channel of unconstrained bandwidth, >>and the early late gate approach is based on the symmetry you see there (as >>the text clearly states), trying to home on the peaks. When the bandwidth >>is constrained by nature (e.g. a telephone line) or regulation (e.g. most >>radio communications) things don't look anything like that diagram. You >>have a signal point on a constellation map that is slewing almost directly >>from symbol point to symbol point. For a simple modulation, like QPSK, >>things may still converge. However, even a small QAM constellation has >>little chance of converging in a reliable way. >> >>Steve > > > I call this under-appreciation of synchronization issues "job > security" ;-).
I haven't seen the direct answer to the simple question in any of the books: For the modulation type of X and SNR of Y, what should be bandwidth of the synch PLL so the implementation losses will not exceed Z ? Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by julius August 12, 20092009-08-12
On Aug 11, 10:15=A0pm, "steveu" <ste...@coppice.org> wrote:
> > Look at Figure 6.3-4. It is based on a channel of unconstrained bandwidth=
,
> and the early late gate approach is based on the symmetry you see there (=
as
> the text clearly states), trying to home on the peaks. When the bandwidth > is constrained by nature (e.g. a telephone line) or regulation (e.g. most > radio communications) things don't look anything like that diagram. You > have a signal point on a constellation map that is slewing almost directl=
y
> from symbol point to symbol point. For a simple modulation, like QPSK, > things may still converge. However, even a small QAM constellation has > little chance of converging in a reliable way. > > Steve
I call this under-appreciation of synchronization issues "job security" ;-). Julius
Reply by steveu August 11, 20092009-08-11
>"steveu" <steveu@coppice.org> writes: > >>>"steveu" <steveu@coppice.org> writes: >>> >>>>>>@BOOK{proakiscomm, >>>>>> title = "{Digital Communications}", >>>>>> author = "John~G.~Proakis", >>>>>> publisher = "McGraw-Hill", >>>>>> edition = "fourth", >>>>>> year = "2001"} >>>>>>@BOOK{couch, >>>>>> title = "{Digital and Analog Communication Systems}", >>>>>> author = "{Leon~W.~Couch}", >>>>>> publisher = "Prentice Hall", >>>>>> edition = "fifth", >>>>>> year = "1993"} >>>>>>@BOOK{sklar, >>>>>> title = "{Digital Communications}", >>>>>> author = "{Bernard~Sklar}", >>>>>> publisher = "Prentice Hall P T R", >>>>>> edition = "second", >>>>>> year = "2001"} >>>>>>@BOOK{leeandmesserschmitt3, >>>>>> title = "{Digital Communication}", >>>>>> author = "{John R. Barry and Edwared A. Lee and David G. >>>>>Messerschmitt}", >>>>>> publisher = "Springer", >>>>>> edition = "third", >>>>>> year = "2004"} >>>>> >>>>>I do not know about the second and the fourth one but the >> first(Proakis) >>>>>and the third(sklar) books are horrible. They just make the subject >> more >>>>>complicated. When I read those two, I thought to forget about >>>>>synchronization topic (both phase and timing). >>>> >>>> They don't make timing recovery complicated. They just gloss over
it,
>> and >>>> don't talk about most of the issues, or most of the interesting >>>> techniques. >>> >>>Then what is section 6.3 (Symbol Timing Estimation) [proakiscomm]
about?
>>>He provides timing loop examples for both decision- and >>>non-decision-directed timing estimation, including the popular >>>"early-late gate" synchronizer, and draws on estimation theory to show >>>where this stuff comes from. >>> >>>Proakis may not provide an *exhaustive* coverage of the topic, but I >>>don't agree that he glosses over it. >> >> The early late gate is too limited to be of much use these days.
Proakis
>> doesn't cover its limitations at all well, or cover methods that do
work
>> with modern massive constellations, and limited bandwidth. In such a
thick
>> book, I thought that was pretty pathetic coverage of one of the key >> topics. > >Steve, > >What is it about the early/late gate that doesn't work well for massive >constellations (I presume you mean something like 256-QAM?)?
Look at Figure 6.3-4. It is based on a channel of unconstrained bandwidth, and the early late gate approach is based on the symmetry you see there (as the text clearly states), trying to home on the peaks. When the bandwidth is constrained by nature (e.g. a telephone line) or regulation (e.g. most radio communications) things don't look anything like that diagram. You have a signal point on a constellation map that is slewing almost directly from symbol point to symbol point. For a simple modulation, like QPSK, things may still converge. However, even a small QAM constellation has little chance of converging in a reliable way. Steve
Reply by Randy Yates August 11, 20092009-08-11
"steveu" <steveu@coppice.org> writes:

>>"steveu" <steveu@coppice.org> writes: >> >>>>>@BOOK{proakiscomm, >>>>> title = "{Digital Communications}", >>>>> author = "John~G.~Proakis", >>>>> publisher = "McGraw-Hill", >>>>> edition = "fourth", >>>>> year = "2001"} >>>>>@BOOK{couch, >>>>> title = "{Digital and Analog Communication Systems}", >>>>> author = "{Leon~W.~Couch}", >>>>> publisher = "Prentice Hall", >>>>> edition = "fifth", >>>>> year = "1993"} >>>>>@BOOK{sklar, >>>>> title = "{Digital Communications}", >>>>> author = "{Bernard~Sklar}", >>>>> publisher = "Prentice Hall P T R", >>>>> edition = "second", >>>>> year = "2001"} >>>>>@BOOK{leeandmesserschmitt3, >>>>> title = "{Digital Communication}", >>>>> author = "{John R. Barry and Edwared A. Lee and David G. >>>>Messerschmitt}", >>>>> publisher = "Springer", >>>>> edition = "third", >>>>> year = "2004"} >>>> >>>>I do not know about the second and the fourth one but the > first(Proakis) >>>>and the third(sklar) books are horrible. They just make the subject > more >>>>complicated. When I read those two, I thought to forget about >>>>synchronization topic (both phase and timing). >>> >>> They don't make timing recovery complicated. They just gloss over it, > and >>> don't talk about most of the issues, or most of the interesting >>> techniques. >> >>Then what is section 6.3 (Symbol Timing Estimation) [proakiscomm] about? >>He provides timing loop examples for both decision- and >>non-decision-directed timing estimation, including the popular >>"early-late gate" synchronizer, and draws on estimation theory to show >>where this stuff comes from. >> >>Proakis may not provide an *exhaustive* coverage of the topic, but I >>don't agree that he glosses over it. > > The early late gate is too limited to be of much use these days. Proakis > doesn't cover its limitations at all well, or cover methods that do work > with modern massive constellations, and limited bandwdith. In such a thick > book, I thought that was pretty pathetic coverage of one of the key > topics.
Steve, What is it about the early/late gate that doesn't work well for massive constellations (I presume you mean something like 256-QAM?)? -- Randy Yates % "How's life on earth? Digital Signal Labs % ... What is it worth?" mailto://yates@ieee.org % 'Mission (A World Record)', http://www.digitalsignallabs.com % *A New World Record*, ELO
Reply by steveu August 11, 20092009-08-11
>"steveu" <steveu@coppice.org> writes: > >>>>@BOOK{proakiscomm, >>>> title = "{Digital Communications}", >>>> author = "John~G.~Proakis", >>>> publisher = "McGraw-Hill", >>>> edition = "fourth", >>>> year = "2001"} >>>>@BOOK{couch, >>>> title = "{Digital and Analog Communication Systems}", >>>> author = "{Leon~W.~Couch}", >>>> publisher = "Prentice Hall", >>>> edition = "fifth", >>>> year = "1993"} >>>>@BOOK{sklar, >>>> title = "{Digital Communications}", >>>> author = "{Bernard~Sklar}", >>>> publisher = "Prentice Hall P T R", >>>> edition = "second", >>>> year = "2001"} >>>>@BOOK{leeandmesserschmitt3, >>>> title = "{Digital Communication}", >>>> author = "{John R. Barry and Edwared A. Lee and David G. >>>Messerschmitt}", >>>> publisher = "Springer", >>>> edition = "third", >>>> year = "2004"} >>> >>>I do not know about the second and the fourth one but the
first(Proakis)
>>>and the third(sklar) books are horrible. They just make the subject
more
>>>complicated. When I read those two, I thought to forget about >>>synchronization topic (both phase and timing). >> >> They don't make timing recovery complicated. They just gloss over it,
and
>> don't talk about most of the issues, or most of the interesting >> techniques. > >Then what is section 6.3 (Symbol Timing Estimation) [proakiscomm] about? >He provides timing loop examples for both decision- and >non-decision-directed timing estimation, including the popular >"early-late gate" synchronizer, and draws on estimation theory to show >where this stuff comes from. > >Proakis may not provide an *exhaustive* coverage of the topic, but I >don't agree that he glosses over it.
The early late gate is too limited to be of much use these days. Proakis doesn't cover its limitations at all well, or cover methods that do work with modern massive constellations, and limited bandwdith. In such a thick book, I thought that was pretty pathetic coverage of one of the key topics. Steve
Reply by Randy Yates August 11, 20092009-08-11
"steveu" <steveu@coppice.org> writes:

>>>@BOOK{proakiscomm, >>> title = "{Digital Communications}", >>> author = "John~G.~Proakis", >>> publisher = "McGraw-Hill", >>> edition = "fourth", >>> year = "2001"} >>>@BOOK{couch, >>> title = "{Digital and Analog Communication Systems}", >>> author = "{Leon~W.~Couch}", >>> publisher = "Prentice Hall", >>> edition = "fifth", >>> year = "1993"} >>>@BOOK{sklar, >>> title = "{Digital Communications}", >>> author = "{Bernard~Sklar}", >>> publisher = "Prentice Hall P T R", >>> edition = "second", >>> year = "2001"} >>>@BOOK{leeandmesserschmitt3, >>> title = "{Digital Communication}", >>> author = "{John R. Barry and Edwared A. Lee and David G. >>Messerschmitt}", >>> publisher = "Springer", >>> edition = "third", >>> year = "2004"} >> >>I do not know about the second and the fourth one but the first(Proakis) >>and the third(sklar) books are horrible. They just make the subject more >>complicated. When I read those two, I thought to forget about >>synchronization topic (both phase and timing). > > They don't make timing recovery complicated. They just gloss over it, and > don't talk about most of the issues, or most of the interesting > techniques.
Then what is section 6.3 (Symbol Timing Estimation) [proakiscomm] about? He provides timing loop examples for both decision- and non-decision-directed timing estimation, including the popular "early-late gate" synchronizer, and draws on estimation theory to show where this stuff comes from. Proakis may not provide an *exhaustive* coverage of the topic, but I don't agree that he glosses over it. -- Randy Yates % "I met someone who looks alot like you, Digital Signal Labs % she does the things you do, mailto://yates@ieee.org % but she is an IBM." http://www.digitalsignallabs.com % 'Yours Truly, 2095', *Time*, ELO
Reply by julius August 11, 20092009-08-11
On Aug 11, 7:01&#2013266080;am, "steveu" <ste...@coppice.org> wrote:
> > Thanks Chintan, Julius and Randy for advising some good books. I have > >them, let me make my question more clear as I have not found any block > >diagram demonstrating my concern. > > > It is clearly mentioned in different books that timing recovery ( > finding > >the correct sampling point i.e. maximum SNR point ) is separately > performed > >on I and Q parts of the received signal. Now my question is that after > >performing timing recovery, whether equalization is done separately on > the > >I and Q branches or we have to combine I&Q output of timing recovery > block > >and perform equalization on complex signal? > > Do they say it *is* separately performed on the I & Q data, or that it > *can be* separately performed on the I & Q data? There are ways to do it > separately, but the most popular approaches - Gardner, cross correlating > the band edges, etc. - work on the complex data, to arrive at a single > timing estimate. Any sane design is going to arrive at a single timing > estimate in the end, so why start with two? > > Steve
Yes, I agree that doing the synch on I and Q separately is stupid. Maybe Ali is confusing the concept of deriving the timing error signal separately from the I and Q channels, with the concept of doing the entire synchronization separately on the I and Q channels. Ali, instead of mentioning these different books as if they were mythical items, why don't you post the references? Julius
Reply by steveu August 11, 20092009-08-11
> Thanks Chintan, Julius and Randy for advising some good books. I have >them, let me make my question more clear as I have not found any block >diagram demonstrating my concern. > > It is clearly mentioned in different books that timing recovery (
finding
>the correct sampling point i.e. maximum SNR point ) is separately
performed
>on I and Q parts of the received signal. Now my question is that after >performing timing recovery, whether equalization is done separately on
the
>I and Q branches or we have to combine I&Q output of timing recovery
block
>and perform equalization on complex signal?
Do they say it *is* separately performed on the I & Q data, or that it *can be* separately performed on the I & Q data? There are ways to do it separately, but the most popular approaches - Gardner, cross correlating the band edges, etc. - work on the complex data, to arrive at a single timing estimate. Any sane design is going to arrive at a single timing estimate in the end, so why start with two? Steve
Reply by julius August 11, 20092009-08-11
On Aug 10, 10:57&#2013266080;pm, "Ali A Nasir" <aliarsha...@hotmail.com> wrote:
> &#2013266080;Thanks Chintan, Julius and Randy for advising some good books. I have > them, let me make my question more clear as I have not found any block > diagram demonstrating my concern. > > &#2013266080;It is clearly mentioned in different books that timing recovery ( finding > the correct sampling point i.e. maximum SNR point ) is separately performed > on I and Q parts of the received signal. Now my question is that after > performing timing recovery, whether equalization is done separately on the > I and Q branches or we have to combine I&Q output of timing recovery block > and perform equalization on complex signal?
For passband-modulated signals you will have to do both timing and phase recovery. Phase offset makes the I and Q parts dependent on each other. Some timing recovery schemes are indifferent to phase offsets, and some are sensitive to it. Equalization should be done AFTER timing and phase synchronization, otherwise you screw up your equalization. If you have a frequency offset, for example, your phase offset will change linearly, and throw off your equalizer. You need to learn to do both equalization and synchronization. Maybe you should take smaller steps and learn them separately, and then figure out how to do both at the same time. Julius