Help with cleaning up noisy 10 bps communication

Hello, I am a software engineer but am new to DSP, except for college a
long time ago.  I have been reading The Scientist and Engineer's Guide to
Digital Signal Processing, Steven W. Smith.  I am trying to decode
communications over a long cable to a well motor, that is digital at 10
bps, but may have ringing and noise and various frequencies.  For example,
without the motor running, there is ringing at about 10 Hz on the signal
transitions, that makes reception difficult.

I will try to clean up the signal some with a hardware filter, but this may
not be enough.  I read that windowed-sinc filters are very effective and
can have a very sharp cutoff, but are not good when information encoded in
the time domain, because of overshoot and ringing.  So what would be a
better filter?

I have seen a hardware Bessel filter utilized to clean up such a signal. 
Any ideas on how to build one of these in software?

I am using a PC to write my filter so computing power is not a problem.

Thanks for any ideas
Ron

	 

_____________________________		
Posted through www.DSPRelated.com

On Tue, 12 Aug 2014 09:56:07 -0500, BasicPoke wrote:

> Hello, I am a software engineer but am new to DSP, except for college a
> long time ago.  I have been reading The Scientist and Engineer's Guide
> to Digital Signal Processing, Steven W. Smith.  I am trying to decode
> communications over a long cable to a well motor, that is digital at 10
> bps, but may have ringing and noise and various frequencies.  For
> example, without the motor running, there is ringing at about 10 Hz on
> the signal transitions, that makes reception difficult.
> 
> I will try to clean up the signal some with a hardware filter, but this
> may not be enough.  I read that windowed-sinc filters are very effective
> and can have a very sharp cutoff, but are not good when information
> encoded in the time domain, because of overshoot and ringing.  So what
> would be a better filter?
> 
> I have seen a hardware Bessel filter utilized to clean up such a signal.
> Any ideas on how to build one of these in software?
> 
> I am using a PC to write my filter so computing power is not a problem.

If you had control over the source, a spread-spectrum modulation scheme 
may work better.

If you have a continuous-time prototype filter, and you have the transfer 
function, then there are approximate transforms from the s domain to the 
z domain.  Look in the index of your book for "bilinear transformation" 
-- if it's there (or in the table of contents) that should put you smack 
in the middle of the relevant section.  Then you just need to have a 
Bessel transfer function in hand, and you can start experimenting.

Your statement "*without the motor running*, there is ringing at about 10 
Hz on the signal transitions" is ominous -- this implies to me at least 
that the characteristics of the line changes with the motor on vs. the 
motor off.  You may find that you need two filters, and to switch.  It 
should be your fervent hope that this is not the case.

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

long cable = transmission line,  is it terminated in its charaterisitc impedance?

what happens if it isn't?


"ringing"  at the data rate  google intersymbol interference isi, and adaptive equalization 


Mark

Tim Wescott wrote:
> On Tue, 12 Aug 2014 09:56:07 -0500, BasicPoke wrote:
>
>> Hello, I am a software engineer but am new to DSP, except for college a
>> long time ago.  I have been reading The Scientist and Engineer's Guide
>> to Digital Signal Processing, Steven W. Smith.  I am trying to decode
>> communications over a long cable to a well motor, that is digital at 10
>> bps, but may have ringing and noise and various frequencies.  For
>> example, without the motor running, there is ringing at about 10 Hz on
>> the signal transitions, that makes reception difficult.
>>
>> I will try to clean up the signal some with a hardware filter, but this
>> may not be enough.  I read that windowed-sinc filters are very effective
>> and can have a very sharp cutoff, but are not good when information
>> encoded in the time domain, because of overshoot and ringing.  So what
>> would be a better filter?
>>
>> I have seen a hardware Bessel filter utilized to clean up such a signal.
>> Any ideas on how to build one of these in software?
>>
>> I am using a PC to write my filter so computing power is not a problem.
>
> If you had control over the source, a spread-spectrum modulation scheme
> may work better.
>
> If you have a continuous-time prototype filter, and you have the transfer
> function, then there are approximate transforms from the s domain to the
> z domain.  Look in the index of your book for "bilinear transformation"
> -- if it's there (or in the table of contents) that should put you smack
> in the middle of the relevant section.  Then you just need to have a
> Bessel transfer function in hand, and you can start experimenting.
>
> Your statement "*without the motor running*, there is ringing at about 10
> Hz on the signal transitions" is ominous -- this implies to me at least
> that the characteristics of the line changes with the motor on vs. the
> motor off.  You may find that you need two filters, and to switch.  It
> should be your fervent hope that this is not the case.
>

But does he actually have a hardware problem to solve first.
His description shouts "voltage source with possible ground loop 
problems".
If he has physical access to both ends, should a pair of 4-20mA 
transducers installed?
OK, 40 yrs ago I worked for a transducer manufacturer and before 
that computers frequently used current loop driven KSR33's ;/


hth

On Tue, 12 Aug 2014 10:50:48 -0500, Richard Owlett wrote:

> Tim Wescott wrote:
>> On Tue, 12 Aug 2014 09:56:07 -0500, BasicPoke wrote:
>>
>>> Hello, I am a software engineer but am new to DSP, except for college
>>> a long time ago.  I have been reading The Scientist and Engineer's
>>> Guide to Digital Signal Processing, Steven W. Smith.  I am trying to
>>> decode communications over a long cable to a well motor, that is
>>> digital at 10 bps, but may have ringing and noise and various
>>> frequencies.  For example, without the motor running, there is ringing
>>> at about 10 Hz on the signal transitions, that makes reception
>>> difficult.
>>>
>>> I will try to clean up the signal some with a hardware filter, but
>>> this may not be enough.  I read that windowed-sinc filters are very
>>> effective and can have a very sharp cutoff, but are not good when
>>> information encoded in the time domain, because of overshoot and
>>> ringing.  So what would be a better filter?
>>>
>>> I have seen a hardware Bessel filter utilized to clean up such a
>>> signal.
>>> Any ideas on how to build one of these in software?
>>>
>>> I am using a PC to write my filter so computing power is not a
>>> problem.
>>
>> If you had control over the source, a spread-spectrum modulation scheme
>> may work better.
>>
>> If you have a continuous-time prototype filter, and you have the
>> transfer function, then there are approximate transforms from the s
>> domain to the z domain.  Look in the index of your book for "bilinear
>> transformation" -- if it's there (or in the table of contents) that
>> should put you smack in the middle of the relevant section.  Then you
>> just need to have a Bessel transfer function in hand, and you can start
>> experimenting.
>>
>> Your statement "*without the motor running*, there is ringing at about
>> 10 Hz on the signal transitions" is ominous -- this implies to me at
>> least that the characteristics of the line changes with the motor on
>> vs. the motor off.  You may find that you need two filters, and to
>> switch.  It should be your fervent hope that this is not the case.
>>
>>
> But does he actually have a hardware problem to solve first. His
> description shouts "voltage source with possible ground loop problems".
> If he has physical access to both ends, should a pair of 4-20mA
> transducers installed?
> OK, 40 yrs ago I worked for a transducer manufacturer and before that
> computers frequently used current loop driven KSR33's ;/

Given that he called out a baud rate I'm assuming that he's adopting 
existing hardware.  But yes, implementing elaborate signal processing 
solutions to basic hardware problems aren't always the best system-level 
decisions.

Ron: some more information would be appreciated, here!

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

On Tue, 12 Aug 2014 09:56:07 -0500, "BasicPoke" <101269@dsprelated>
wrote:

>Hello, I am a software engineer but am new to DSP, except for college a
>long time ago.  I have been reading The Scientist and Engineer's Guide to
>Digital Signal Processing, Steven W. Smith.  I am trying to decode
>communications over a long cable to a well motor, that is digital at 10
>bps, but may have ringing and noise and various frequencies.  For example,
>without the motor running, there is ringing at about 10 Hz on the signal
>transitions, that makes reception difficult.
>
>I will try to clean up the signal some with a hardware filter, but this may
>not be enough.  I read that windowed-sinc filters are very effective and
>can have a very sharp cutoff, but are not good when information encoded in
>the time domain, because of overshoot and ringing.  So what would be a
>better filter?
>
>I have seen a hardware Bessel filter utilized to clean up such a signal. 
>Any ideas on how to build one of these in software?
>
>I am using a PC to write my filter so computing power is not a problem.
>
>Thanks for any ideas
>Ron

As others have mentioned, first try to make certain that the line is
properly terminated (on both ends if it is bi-directional).  Also,
take a look at the line when there is no data transmission,
preferrably with the motor on and off, so see if there is interference
energy that could be potentially tracked down and reduced.

Once all of that is done as well as can be, the best performance will
be with a "matched" filter at the receiver, and the matching has to do
with how the signal is modulated and transmitted.   Do you have
control over that or are you just trying to interface to an existing
system?

For example, if the modulator transmits rectangular pulses and those
are pretty much unfiltered when they make it on to the line, then the
"matched" filter in the demodulator is an integrate-and-dump filter.
On the other hand, if there is pulse shaping done in the transmitter,
then the receiver needs to have a filter that "matches" whatever is
used in the Tx.

If terminating and interference mitigation doesn't fix the ringing,
you may need an equalizer (maybe even an adaptive one) in the receiver
to get the best performance.   Passive equalizers can be made and
tuned by hand, but it's a bit of a pain and requires some skill.

You haven't mentioned what sort of resources you have to work with or
what your system constraints might be.


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

On Tue, 12 Aug 2014 16:32:02 +0000, Eric Jacobsen wrote:

> On Tue, 12 Aug 2014 09:56:07 -0500, "BasicPoke" <101269@dsprelated>
> wrote:
> 
>>Hello, I am a software engineer but am new to DSP, except for college a
>>long time ago.  I have been reading The Scientist and Engineer's Guide
>>to Digital Signal Processing, Steven W. Smith.  I am trying to decode
>>communications over a long cable to a well motor, that is digital at 10
>>bps, but may have ringing and noise and various frequencies.  For
>>example, without the motor running, there is ringing at about 10 Hz on
>>the signal transitions, that makes reception difficult.

<< snip >>

> If terminating and interference mitigation doesn't fix the ringing, you
> may need an equalizer (maybe even an adaptive one) in the receiver to
> get the best performance.   Passive equalizers can be made and tuned by
> hand, but it's a bit of a pain and requires some skill.
> 
> You haven't mentioned what sort of resources you have to work with or
> what your system constraints might be.

Fixed-tune equalizers have the distinct weakness that they only work for 
the particular circumstance that they are tuned for -- so if the 
characteristics change (say, when the motor comes on) then they break, and 
if the characteristics differ from unit to unit (say one pump on a 100 
foot wire vs. another on a 10 mile wire) then they need to be fiddled with 
at installation time.

Adaptive equalizers need time to tune, and can do some oddball stuff to 
the unwary.

You definitely want to start by seeing if you can rearrange the physical 
world so that an equalizer is unnecessary.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Those are some great ideas, thanks guys.  I should have known about the
termination with characteristic impedance, but did not occur to me.  I will
have to check this out, as I'm not the hardware guy on this.  I need to
figure out how much detail I can go into without getting into trouble.

We have control of the hardware on both ends of the cable, but generally
not control over the cable and motor.  Not sure how the characteristics
change with the motor on, but I know there is lots of noise at various
frequencies.
Ron	 

_____________________________		
Posted through www.DSPRelated.com

On Tuesday, August 12, 2014 12:48:56 PM UTC-4, BasicPoke wrote:
> Those are some great ideas, thanks guys.  I should have known about the
> 
> termination with characteristic impedance, but did not occur to me.  I will
> 
> have to check this out, as I'm not the hardware guy on this.  I need to
> 
> figure out how much detail I can go into without getting into trouble.
> 
> 
> 
> We have control of the hardware on both ends of the cable, but generally
> 
> not control over the cable and motor.  Not sure how the characteristics
> 
> change with the motor on, but I know there is lots of noise at various
> 
> frequencies.
> 
> Ron	 
> 
> 


So this brings up a question I have tried to raise here beofre.
And nothing personal to Ron, this is a generic issue.

It seems clear Ron's task is in  a professional setting, i.e.  his empoyer is paying him to figure this out.  I'd like to hear a discussion about "giving away" this kind of information given the job market we are in.  If Ron were not able to come here and get this kind of help for free, Ron's employer might have to hire one of us. 

This issue was brought up often the the alt.hvac group where the professional hvac guys would,  shall we say,  respond in a very impolite fashion, to "homeowners" looking for free help.  And after thinking about it, I think they may have a point. 

I'd like to hear other opinions about giving away knowledge of one's craft. 
I like information to be free, but I also like to be paid for my expertiese.  How do we balance these?

Again, sorry Ron, I used your particular question as an example, nothing personal.

Mark



 

On Tue, 12 Aug 2014 10:11:27 -0700 (PDT), makolber@yahoo.com wrote:

>On Tuesday, August 12, 2014 12:48:56 PM UTC-4, BasicPoke wrote:
>> Those are some great ideas, thanks guys.  I should have known about the
>>=20
>> termination with characteristic impedance, but did not occur to me.  I wi=
>ll
>>=20
>> have to check this out, as I'm not the hardware guy on this.  I need to
>>=20
>> figure out how much detail I can go into without getting into trouble.
>>=20
>>=20
>>=20
>> We have control of the hardware on both ends of the cable, but generally
>>=20
>> not control over the cable and motor.  Not sure how the characteristics
>>=20
>> change with the motor on, but I know there is lots of noise at various
>>=20
>> frequencies.
>>=20
>> Ron	=20
>>=20
>>=20
>
>
>So this brings up a question I have tried to raise here beofre.
>And nothing personal to Ron, this is a generic issue.
>
>It seems clear Ron's task is in  a professional setting, i.e.  his empoyer =
>is paying him to figure this out.  I'd like to hear a discussion about "giv=
>ing away" this kind of information given the job market we are in.  If Ron =
>were not able to come here and get this kind of help for free, Ron's employ=
>er might have to hire one of us.=20
>
>This issue was brought up often the the alt.hvac group where the profession=
>al hvac guys would,  shall we say,  respond in a very impolite fashion, to =
>"homeowners" looking for free help.  And after thinking about it, I think t=
>hey may have a point.=20
>
>I'd like to hear other opinions about giving away knowledge of one's craft.=
>=20
>I like information to be free, but I also like to be paid for my expertiese=
>.  How do we balance these?
>
>Again, sorry Ron, I used your particular question as an example, nothing pe=
>rsonal.
>
>Mark

I answered this for myself a long time ago, so I'll share my point of
view FWIW.

Everybody here can share information as they see fit to whomever they
see fit.   So, basically, it's everybody's individual personal choice
to decide what they will or will not share or how to respond to an
inquiry.

There have been many, many times when I've refrained from going into
detail on a response for a number of different reasons.   Clearly
nobody should improperly share proprietary information (e.g., "my
employer spent years and a lot of $$ figuring out this trade secret
that solved your problem, so I can't share it here"), and, as you
mention, sometimes there is "professional" information or information
that you just don't want to give away for whatever reason.   Those
reasons can be professional or personal or a total whim, it doesn't
matter, it's your choice to share or not.    Sometimes there are the
cases where you might feel like you're hurting your own business or
employer or industry by giving things out too easily, and I think
that's legitimate.

I don't however, think it's legitimate to use it as a stick to beat
people with.   If I don't want to share something, but somebody else
does, that's fine, that's their choice.   I don't share the view that
comp.dsp, or other parts of usenet, or most internet forums exist
solely for professionals or vendors or contractors to promote
themselves and solicit business (and there have been some here who
clearly thought that was the case, said as much, and treated people
horribly).

So if somebody here, or an hvac pro, wants to berate people just for
asking questions, I think that's really, really poor form.   We've all
been noobs at some point, and it's usually possible to help somebody
or send them in the right direction, or at least help them not make a
bad mistake, without violating any disclosure issues or hurting your
own business.   Sometimes it's possible to go into great detail on a
method or solution, too, so often there's no barrier to doing that.

I think it still boils down to a personal choice, though.   For the
most part, I think the sort of people who hang out on comp.dsp are
already the sort of people that are willing to help at least a certain
amount, or it wouldn't exist as it does.

I've asked questions here before that had to do with solving
work-related problems, and I've helped others on the same.   I don't
see an issue with it within obvious limits (e.g., "do this job for me"
doesn't fly), and I think this is the basic fundamental purpose of
usenet and many other internet forums.  

I think it'll be a very sad day if it ever gets to where people are
never willing to help each other to some degree unless money changes
hands.   I hope it never gets to that point.




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