Forums

Adaptive equalization and synchronization

Started by Tim Wescott May 30, 2013
How does one maintain synchronization when one is using adaptive 
equalization?

I'm currently working on a highly dispersive, varying channel, that needs 
adaptive equalization to work well.

But I've come to realize that one thing that adaptive equalization adapts 
for is the exact timing of the signal.  This means that you can't really 
used the output of the equalizer as a phase reference, unless you have 
some knowledge of what the equalizer is doing to the phase of the 
signal.  So you either need to be able to figure out the equalizer's 
effect, or you need some side channel for synchronization.  

I'm currently able to dodge this problem because right now I'm making 
improvements on a legacy signaling scheme, and the cruder reception 
methods for this scheme are good enough for synchronization, if not for 
the best reception.  But that won't last.

Our goal is to increase data rates, and that's going to make any new 
signaling schemes dependent on adaptation.  At turn-on we can have a 
training/synchronizing sequence, but that's just to spin things up 
quickly: after that things run long enough that the adaptation constants 
change, and extant clock frequency errors will cause timing errors 
greater than a symbol time.  Thus we cannot just set the phase at the 
beginning and forget it: we must maintain synchronization.  Being able to 
do so without putting special features in the signaling scheme would be 
very cool.

Any guidance?  Is there a way to look at the equalizer and discern what 
its doing to the signal phase?  Are there other techniques that work well?

If this is a complicated topic (and it sure sounds like it could be) I'll 
happily entertain book recommendations, particularly if the theory 
therein has actual practical usage beyond just making a name for the 
author and getting him grant money.

-- 
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?

Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com
On Thu, 30 May 2013 11:58:18 -0500, Tim Wescott <tim@seemywebsite.com>
wrote:

>How does one maintain synchronization when one is using adaptive >equalization? > >I'm currently working on a highly dispersive, varying channel, that needs >adaptive equalization to work well. > >But I've come to realize that one thing that adaptive equalization adapts >for is the exact timing of the signal. This means that you can't really >used the output of the equalizer as a phase reference, unless you have >some knowledge of what the equalizer is doing to the phase of the >signal. So you either need to be able to figure out the equalizer's >effect, or you need some side channel for synchronization. > >I'm currently able to dodge this problem because right now I'm making >improvements on a legacy signaling scheme, and the cruder reception >methods for this scheme are good enough for synchronization, if not for >the best reception. But that won't last. > >Our goal is to increase data rates, and that's going to make any new >signaling schemes dependent on adaptation. At turn-on we can have a >training/synchronizing sequence, but that's just to spin things up >quickly: after that things run long enough that the adaptation constants >change, and extant clock frequency errors will cause timing errors >greater than a symbol time. Thus we cannot just set the phase at the >beginning and forget it: we must maintain synchronization. Being able to >do so without putting special features in the signaling scheme would be >very cool. > >Any guidance? Is there a way to look at the equalizer and discern what >its doing to the signal phase? Are there other techniques that work well? > >If this is a complicated topic (and it sure sounds like it could be) I'll >happily entertain book recommendations, particularly if the theory >therein has actual practical usage beyond just making a name for the >author and getting him grant money. > >-- >My liberal friends think I'm a conservative kook. >My conservative friends think I'm a liberal kook. >Why am I not happy that they have found common ground? > >Tim Wescott, Communications, Control, Circuits & Software >http://www.wescottdesign.com
Usually symbol synchronization is done prior to the EQ for the reasons you indicate. This does mean that sometimes judicious selection of an appropriate TED must be made so that it is well-behaved in the expected conditions. Mostly it just needs to be stable/smooth in dynamic channels so that it doesn't jump around. Even if it has some timing error in distortion, as you've discovered the EQ can deal with it, as long as it's stable. There can be similar issues with phase-lock, too, that can be handled other ways. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
On Thu, 30 May 2013 17:12:59 +0000, Eric Jacobsen wrote:

> On Thu, 30 May 2013 11:58:18 -0500, Tim Wescott <tim@seemywebsite.com> > wrote: > >>How does one maintain synchronization when one is using adaptive >>equalization? >> >>I'm currently working on a highly dispersive, varying channel, that >>needs adaptive equalization to work well. >> >>But I've come to realize that one thing that adaptive equalization >>adapts for is the exact timing of the signal. This means that you can't >>really used the output of the equalizer as a phase reference, unless you >>have some knowledge of what the equalizer is doing to the phase of the >>signal. So you either need to be able to figure out the equalizer's >>effect, or you need some side channel for synchronization. >> >>I'm currently able to dodge this problem because right now I'm making >>improvements on a legacy signaling scheme, and the cruder reception >>methods for this scheme are good enough for synchronization, if not for >>the best reception. But that won't last. >> >>Our goal is to increase data rates, and that's going to make any new >>signaling schemes dependent on adaptation. At turn-on we can have a >>training/synchronizing sequence, but that's just to spin things up >>quickly: after that things run long enough that the adaptation constants >>change, and extant clock frequency errors will cause timing errors >>greater than a symbol time. Thus we cannot just set the phase at the >>beginning and forget it: we must maintain synchronization. Being able >>to do so without putting special features in the signaling scheme would >>be very cool. >> >>Any guidance? Is there a way to look at the equalizer and discern what >>its doing to the signal phase? Are there other techniques that work >>well? >> >>If this is a complicated topic (and it sure sounds like it could be) >>I'll happily entertain book recommendations, particularly if the theory >>therein has actual practical usage beyond just making a name for the >>author and getting him grant money. >> >>-- >>My liberal friends think I'm a conservative kook. My conservative >>friends think I'm a liberal kook. Why am I not happy that they have >>found common ground? >> >>Tim Wescott, Communications, Control, Circuits & Software >>http://www.wescottdesign.com > > Usually symbol synchronization is done prior to the EQ for the reasons > you indicate. This does mean that sometimes judicious selection of an > appropriate TED must be made so that it is well-behaved in the expected > conditions. Mostly it just needs to be stable/smooth in dynamic > channels so that it doesn't jump around. Even if it has some timing > error in distortion, as you've discovered the EQ can deal with it, as > long as it's stable. > > There can be similar issues with phase-lock, too, that can be handled > other ways.
Thanks Eric. Reading your post made the light dawn in one of those ways that makes one feel simultaneously profoundly smart and profoundly stupid, for figuring out my solution but having it be so obvious that I should really have gotten it earlier. I realized that if worse comes to worst I can synchronize on the training sequence, then free-run for the first 50 or so symbols, then fix the resulting equalizer setting to use for continued synchronization. At the same time that I'll allow the equalizer to continue to evolve for data decoding. The channel changes enough that I need to keep the equalizer adapting -- but it doesn't change so much from end to end that I can't use the initial capture for synchronization. (And it may be that there's some generic, fixed equalization or family thereof that'll be good enough for synchronization -- I need more field data from the customer to know that, however). Now I have a really dumb question, though -- what's "TED"? -- My liberal friends think I'm a conservative kook. My conservative friends think I'm a liberal kook. Why am I not happy that they have found common ground? Tim Wescott, Communications, Control, Circuits & Software http://www.wescottdesign.com
On 5/30/2013 11:58 AM, Tim Wescott wrote:
> How does one maintain synchronization when one is using adaptive > equalization? > > I'm currently working on a highly dispersive, varying channel, that needs > adaptive equalization to work well. > > But I've come to realize that one thing that adaptive equalization adapts > for is the exact timing of the signal. This means that you can't really > used the output of the equalizer as a phase reference, unless you have > some knowledge of what the equalizer is doing to the phase of the > signal. So you either need to be able to figure out the equalizer's > effect, or you need some side channel for synchronization.
Optimal: Make Kalman-like filter observing current decision. Derive equalizer update, symbol sync and carrier sync updates altogether. Quazi optimal: Derive symbol sync and carrier sync updates from equalizer coefficient dynamics. Dumb: Put equalizer outside of timing/carrier loop. I.e. perform synchronization before EQ. Also dumb: Put equalizer inside loop. Make the dynamics of sync and carrier much faster then equalizer update. Vladimir Vassilevsky DSP and Mixed Signal Designs www.abvolt.com
On Thu, 30 May 2013 12:35:27 -0500, Tim Wescott <tim@seemywebsite.com>
wrote:

>On Thu, 30 May 2013 17:12:59 +0000, Eric Jacobsen wrote: > >> On Thu, 30 May 2013 11:58:18 -0500, Tim Wescott <tim@seemywebsite.com> >> wrote: >> >>>How does one maintain synchronization when one is using adaptive >>>equalization? >>> >>>I'm currently working on a highly dispersive, varying channel, that >>>needs adaptive equalization to work well. >>> >>>But I've come to realize that one thing that adaptive equalization >>>adapts for is the exact timing of the signal. This means that you can't >>>really used the output of the equalizer as a phase reference, unless you >>>have some knowledge of what the equalizer is doing to the phase of the >>>signal. So you either need to be able to figure out the equalizer's >>>effect, or you need some side channel for synchronization. >>> >>>I'm currently able to dodge this problem because right now I'm making >>>improvements on a legacy signaling scheme, and the cruder reception >>>methods for this scheme are good enough for synchronization, if not for >>>the best reception. But that won't last. >>> >>>Our goal is to increase data rates, and that's going to make any new >>>signaling schemes dependent on adaptation. At turn-on we can have a >>>training/synchronizing sequence, but that's just to spin things up >>>quickly: after that things run long enough that the adaptation constants >>>change, and extant clock frequency errors will cause timing errors >>>greater than a symbol time. Thus we cannot just set the phase at the >>>beginning and forget it: we must maintain synchronization. Being able >>>to do so without putting special features in the signaling scheme would >>>be very cool. >>> >>>Any guidance? Is there a way to look at the equalizer and discern what >>>its doing to the signal phase? Are there other techniques that work >>>well? >>> >>>If this is a complicated topic (and it sure sounds like it could be) >>>I'll happily entertain book recommendations, particularly if the theory >>>therein has actual practical usage beyond just making a name for the >>>author and getting him grant money. >>> >>>-- >>>My liberal friends think I'm a conservative kook. My conservative >>>friends think I'm a liberal kook. Why am I not happy that they have >>>found common ground? >>> >>>Tim Wescott, Communications, Control, Circuits & Software >>>http://www.wescottdesign.com >> >> Usually symbol synchronization is done prior to the EQ for the reasons >> you indicate. This does mean that sometimes judicious selection of an >> appropriate TED must be made so that it is well-behaved in the expected >> conditions. Mostly it just needs to be stable/smooth in dynamic >> channels so that it doesn't jump around. Even if it has some timing >> error in distortion, as you've discovered the EQ can deal with it, as >> long as it's stable. >> >> There can be similar issues with phase-lock, too, that can be handled >> other ways. > >Thanks Eric. Reading your post made the light dawn in one of those ways >that makes one feel simultaneously profoundly smart and profoundly >stupid, for figuring out my solution but having it be so obvious that I >should really have gotten it earlier. > >I realized that if worse comes to worst I can synchronize on the training >sequence, then free-run for the first 50 or so symbols, then fix the >resulting equalizer setting to use for continued synchronization. At the >same time that I'll allow the equalizer to continue to evolve for data >decoding. The channel changes enough that I need to keep the equalizer >adapting -- but it doesn't change so much from end to end that I can't >use the initial capture for synchronization.
The timing may drift, too, depending on the system, so it's sometimes difficult to juggle multiplexing the EQ training and timing synchronization when you do it that way. It can certainly be made to work in many systems, though.
>(And it may be that there's some generic, fixed equalization or family >thereof that'll be good enough for synchronization -- I need more field >data from the customer to know that, however). > >Now I have a really dumb question, though -- what's "TED"?
Timing Error Detector. The timing phase detector for the synchronization loop.
>-- >My liberal friends think I'm a conservative kook. >My conservative friends think I'm a liberal kook. >Why am I not happy that they have found common ground? > >Tim Wescott, Communications, Control, Circuits & Software >http://www.wescottdesign.com
Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
>I realized that if worse comes to worst I can synchronize on the training
>sequence, then free-run for the first 50 or so symbols, then fix the >resulting equalizer setting to use for continued synchronization. At the
>same time that I'll allow the equalizer to continue to evolve for data >decoding. The channel changes enough that I need to keep the equalizer >adapting -- but it doesn't change so much from end to end that I can't >use the initial capture for synchronization.
IMHO periodical initial capture + continous equalizer adaptation with free running symbol timing is not bad way.
>(And it may be that there's some generic, fixed equalization or family >thereof that'll be good enough for synchronization -- I need more field >data from the customer to know that, however). > >Now I have a really dumb question, though -- what's "TED"?
See paper Solving the Interaction Problem of Timing Synchronization and Equalization Dirk Daecke and Heinrich Schenk http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=4497274&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D4497274
>>I realized that if worse comes to worst I can synchronize on the
training
> >>sequence, then free-run for the first 50 or so symbols, then fix the >>resulting equalizer setting to use for continued synchronization. At
the
> >>same time that I'll allow the equalizer to continue to evolve for data >>decoding. The channel changes enough that I need to keep the equalizer >>adapting -- but it doesn't change so much from end to end that I can't >>use the initial capture for synchronization. > >IMHO periodical initial capture + continous equalizer adaptation with
free
>running symbol timing is not bad way. > > >>(And it may be that there's some generic, fixed equalization or family >>thereof that'll be good enough for synchronization -- I need more field >>data from the customer to know that, however). >> >>Now I have a really dumb question, though -- what's "TED"? > >See paper > >Solving the Interaction Problem of Timing Synchronization and
Equalization
>Dirk Daecke and Heinrich Schenk > >http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=4497274&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D4497274 >
Only a fractionally spaced equalizer will resolve timing phase. If you use a T spaced equalizer you had better have good symbol timing recovery prior to the equalizer. Carrier reference can be resolved after equalization unless the frequency offset is large in which case you can split the loop and recover frequency prior to the equalizer and phase/remaining frequency after the equalizer. -Doug
On 5/31/2013 10:22 AM, DougB wrote:

> Only a fractionally spaced equalizer will resolve timing phase. If you use > a T spaced equalizer you had better have good symbol timing recovery prior > to the equalizer. Carrier reference can be resolved after equalization > unless the frequency offset is large in which case you can split the loop > and recover frequency prior to the equalizer and phase/remaining frequency > after the equalizer.
Whether equalizer is fractional or not, all information about sync, carrier and ISI comes from error vector. You just have to separate it into rotation, sync and linear distortion parts. Vladimir Vassilevsky DSP and Mixed Signal Designs www.abvolt.com
On Fri, 31 May 2013 10:22:43 -0500, "DougB" <60916@dsprelated> wrote:

>>>I realized that if worse comes to worst I can synchronize on the >training >> >>>sequence, then free-run for the first 50 or so symbols, then fix the >>>resulting equalizer setting to use for continued synchronization. At >the >> >>>same time that I'll allow the equalizer to continue to evolve for data >>>decoding. The channel changes enough that I need to keep the equalizer >>>adapting -- but it doesn't change so much from end to end that I can't >>>use the initial capture for synchronization. >> >>IMHO periodical initial capture + continous equalizer adaptation with >free >>running symbol timing is not bad way. >> >> >>>(And it may be that there's some generic, fixed equalization or family >>>thereof that'll be good enough for synchronization -- I need more field >>>data from the customer to know that, however). >>> >>>Now I have a really dumb question, though -- what's "TED"? >> >>See paper >> >>Solving the Interaction Problem of Timing Synchronization and >Equalization >>Dirk Daecke and Heinrich Schenk >> >>http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=4497274&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D4497274 >> > >Only a fractionally spaced equalizer will resolve timing phase. If you use >a T spaced equalizer you had better have good symbol timing recovery prior >to the equalizer. Carrier reference can be resolved after equalization >unless the frequency offset is large in which case you can split the loop >and recover frequency prior to the equalizer and phase/remaining frequency >after the equalizer. >-Doug
One of the disadvantages of using the EQ to correct timing is that it consumes degrees of freedom usually intended for combating the channel distortion. Those degrees of freedom cost complexity, and usually the complexity of the EQ is much greater than the complexity of a decent timing recovery system. So, in my experience, the systems that try to use the EQ for synchronization or to aid in synchronization wind up with a heavier complexity load than alternative methods. Separating the two tasks can be advantageous for the performance of both the EQ and the timing recovery as well as complexity. As always, YMMV, every system is a little different. I think Tim's system may be bursty, which can allow additional simplifications like he's already suggested. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
>>Only a fractionally spaced equalizer will resolve timing phase. If you
use
>>a T spaced equalizer you had better have good symbol timing recovery
prior
>>to the equalizer. Carrier reference can be resolved after equalization >>unless the frequency offset is large in which case you can split the
loop
>>and recover frequency prior to the equalizer and phase/remaining
frequency
>>after the equalizer. >>-Doug > > >One of the disadvantages of using the EQ to correct timing is that it >consumes degrees of freedom usually intended for combating the channel >distortion. Those degrees of freedom cost complexity, and usually >the complexity of the EQ is much greater than the complexity of a >decent timing recovery system. So, in my experience, the systems >that try to use the EQ for synchronization or to aid in >synchronization wind up with a heavier complexity load than >alternative methods. Separating the two tasks can be advantageous for >the performance of both the EQ and the timing recovery as well as >complexity. > > > >Eric Jacobsen >Anchor Hill Communications >http://www.anchorhill.com >
That's true, but if the channel highly distorted, there is no TED that is going to give you the optimum timing phase - that can only be gotten from a fractionally spaced equalizer. Still use timing recovery as normal, but let the FSE optimize the timing phase - it can synthesize the proper fractional delay. -Doug