I love my job

I sat down today to get in a little bit of work over the weekend on a 
communications project I'm doing for a customer.  The channel is not only 
highly dispersive, but I'm stuck with a legacy modulation method that 
does not cough up a non-correlated mix of symbols, and at any rate the 
system as a whole is used for feedback where you just can't throw away a 
bunch of time waiting for acquisition.  Finally, the whole thing has a 
host of complications that I can't go into because of non-disclosure, but 
basically it's an oddball, unique, critical communication task for which 
I'm having to invent some theory to fit the problem.

So this morning put in my first-cut adaptive equalizer, thinking "gosh, I 
hope this isn't so mangled that I can't get something useful out of the 
results".

What I got was REALLY GOOD EQUALIZATION!  I mean really good!

On my desk right now is a stack of three books: A.P. Clark's "Equalizers 
for Digital Modems", closed; on top of that is Astrom's "Adaptive 
Control", open, and on top of that is Simon's "Optimal State Estimation".

"Fun" is applying control systems knowledge to communications, and having 
it work profitably.

-- 
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com

On Sat, 18 May 2013 15:22:31 -0500, Tim Wescott
<tim@seemywebsite.please> wrote:

>I sat down today to get in a little bit of work over the weekend on a 
>communications project I'm doing for a customer.  The channel is not only 
>highly dispersive, but I'm stuck with a legacy modulation method that 
>does not cough up a non-correlated mix of symbols, and at any rate the 
>system as a whole is used for feedback where you just can't throw away a 
>bunch of time waiting for acquisition.  Finally, the whole thing has a 
>host of complications that I can't go into because of non-disclosure, but 
>basically it's an oddball, unique, critical communication task for which 
>I'm having to invent some theory to fit the problem.
>
>So this morning put in my first-cut adaptive equalizer, thinking "gosh, I 
>hope this isn't so mangled that I can't get something useful out of the 
>results".
>
>What I got was REALLY GOOD EQUALIZATION!  I mean really good!
>
>On my desk right now is a stack of three books: A.P. Clark's "Equalizers 
>for Digital Modems", closed; on top of that is Astrom's "Adaptive 
>Control", open, and on top of that is Simon's "Optimal State Estimation".
>
>"Fun" is applying control systems knowledge to communications, and having 
>it work profitably.
>
>-- 
>Tim Wescott
>Control system and signal processing consulting
>www.wescottdesign.com

It's always fun when something comes together like that.

There is a lot of technology overlap in controls and comm.


Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

On 5/18/2013 3:22 PM, Tim Wescott wrote:> I sat down today to get in a 
little bit of work over the weekend on a
 > communications project I'm doing for a customer.  The channel is not only
 > highly dispersive, but I'm stuck with a legacy modulation method that
 > does not cough up a non-correlated mix of symbols, and at any rate the
 > system as a whole is used for feedback where you just can't throw away a
 > bunch of time waiting for acquisition.  Finally, the whole thing has a
 > host of complications that I can't go into because of non-disclosure, but
 > basically it's an oddball, unique, critical communication task for which
 > I'm having to invent some theory to fit the problem.

Toughest challenges arise when they want to adapt legacy system for 
something that it was not originally intended for; while keeping full 
compatibility. This is good chance for expert to earn his bucks.

 > So this morning put in my first-cut adaptive equalizer, thinking "gosh, I
 > hope this isn't so mangled that I can't get something useful out of the
 > results".
 >
 > What I got was REALLY GOOD EQUALIZATION!  I mean really good!
 >
 > On my desk right now is a stack of three books: A.P. Clark's "Equalizers
 > for Digital Modems", closed; on top of that is Astrom's "Adaptive
 > Control", open, and on top of that is Simon's "Optimal State Estimation".

There isn't need for many books; mostly the books are retelling each 
other; and the rest you could figure out yourself.

 > "Fun" is applying control systems knowledge to communications, and having
 > it work profitably.

DSP is about multiply and accumulate. Once you got main idea, you could 
do anyting from image processing to error correction codes.


Vladimir Vassilevsky
DSP and Mixed Signal Designs
www.abvolt.com







On 5/18/2013 3:22 PM, Tim Wescott wrote:
> I sat down today to get in a little bit of work over the weekend on a
> communications project I'm doing for a customer.  The channel is not only
> highly dispersive, but I'm stuck with a legacy modulation method that
> does not cough up a non-correlated mix of symbols, and at any rate the
> system as a whole is used for feedback where you just can't throw away a
> bunch of time waiting for acquisition.  Finally, the whole thing has a
> host of complications that I can't go into because of non-disclosure, but
> basically it's an oddball, unique, critical communication task for which
> I'm having to invent some theory to fit the problem.
>
> So this morning put in my first-cut adaptive equalizer, thinking "gosh, I
> hope this isn't so mangled that I can't get something useful out of the
> results".
>
> What I got was REALLY GOOD EQUALIZATION!  I mean really good!
>
> On my desk right now is a stack of three books: A.P. Clark's "Equalizers
> for Digital Modems", closed; on top of that is Astrom's "Adaptive
> Control", open, and on top of that is Simon's "Optimal State Estimation".
>
> "Fun" is applying control systems knowledge to communications, and having
> it work profitably.
>

I love it when a plan comes together.

I had some downtime at work and while I've made dozens of modulators, I've never made a demod.  So I was working a model of a QAM-16 demod.  The first thing I did was work on the Gardner loop and when I got it to lock it was a wonderful thing. When I was working on the the decision-directed carrier synchronization I referred to a lot of textbooks but didn't find a whole lot of helpful info.  The best one was by a guy named Rice.  I eventually got it to work (in the model). 

An equalizer would be fun to work on.  Maybe I'll play with that in my next downtime phase.
Kevin

On Saturday, May 18, 2013 1:22:31 PM UTC-7, Tim Wescott wrote:
> So this morning put in my first-cut adaptive equalizer, thinking "gosh, I 
> hope this isn't so mangled that I can't get something useful out of the 
> results".
> 
> What I got was REALLY GOOD EQUALIZATION!  I mean really good!
...
> 
> "Fun" is applying control systems knowledge to communications, and having 
> it work profitably.

Adaptive filters are fun and map to control system design nicely. After all we are designing a state-space model of the system we are interested. If you model make the state a set of coefficients, a transversal adaptive filter solution pops up nicely.

How are you dealing with synchronization in this system? In my experience most of the challenge is in timing error estimation if one is dealing with baud-rate sampling. I hope you can afford to oversample nicely.

On Mon, 20 May 2013 08:59:59 -0700 (PDT), Kevin Neilson
<kevin.neilson@xilinx.com> wrote:

>I love it when a plan comes together.
>
>I had some downtime at work and while I've made dozens of modulators, I've =
>never made a demod.  So I was working a model of a QAM-16 demod.  The first=
> thing I did was work on the Gardner loop and when I got it to lock it was =
>a wonderful thing. When I was working on the the decision-directed carrier =
>synchronization I referred to a lot of textbooks but didn't find a whole lo=
>t of helpful info.  The best one was by a guy named Rice.  I eventually got=
> it to work (in the model).=20
>
>An equalizer would be fun to work on.  Maybe I'll play with that in my next=
> downtime phase.
>Kevin

The modulator you worked on for us about ten years ago is still in
service.   ;)


Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

On Mon, 20 May 2013 14:17:47 -0700, muzaffer.kal wrote:

> On Saturday, May 18, 2013 1:22:31 PM UTC-7, Tim Wescott wrote:
>> So this morning put in my first-cut adaptive equalizer, thinking "gosh,
>> I hope this isn't so mangled that I can't get something useful out of
>> the results".
>> 
>> What I got was REALLY GOOD EQUALIZATION!  I mean really good!
> ...
>> 
>> "Fun" is applying control systems knowledge to communications, and
>> having it work profitably.
> 
> Adaptive filters are fun and map to control system design nicely. After
> all we are designing a state-space model of the system we are
> interested. If you model make the state a set of coefficients, a
> transversal adaptive filter solution pops up nicely.
> 
> How are you dealing with synchronization in this system? In my
> experience most of the challenge is in timing error estimation if one is
> dealing with baud-rate sampling. I hope you can afford to oversample
> nicely.

The existing modulation format makes synchronization pretty easy -- 
basically there's pulses separated by lots of 0.

-- 
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com