DSP sound/audio demos

Well, I think that whatever you do with a laptop computer it will not
change your reputation if you explain it from the "weird side" of
things. People need a "drawer" to fit you in, so you have to give them
some key words they know. If you just fire up your laptop and hack
away before them to produce strange sounds out of their voices they
will still think "what a useless profession - how come someone is
being paid to play with sounds?". So, you have to explain it first in
simple terms before you do that to show them how much fun it can be.

I often have the same problem, and I usually succeed by explaining it
differently:

- The general approach: You could say that DSP is some hybrid science
between physics, information technology, communications engineering
etc. and mention some of the (everyday) equipment that makes use of
DSP, such as cell phones, TV sets, DVD players, ECG machines, modems,
satellites, radio astronomy telescopes, etc. to illustrate what kind
of work you're doing and where it can be found at work.

- The specific approach: You could pick one application and explain it
to them. If you're into audio, take a reverb for example. Tell them
that DSP makes it possible to simulate a room with its acoustic
properties without actually having it available. It's easy to see that
a concert hall might not be available (or affordable) for some
orchestra projects when you need it, so you have to simulate it
digitally.

Or take time stretching: a dialog that was recorded which is 1:30 min
long and needs to fit in a 1:00 slot in a radio ad playlist.

Or take de-noising to clean up noisy data.... or speech recognition...
the list is quite endless...

Cheers
Stephan

Jon Harris wrote:

>>Reflections cause phase-cancellation, which cause notch-filter-type
>>behaviour in a frequency response. This is non-linear, because it cannot
> 
> be
> 
>>compensated by a linear system (ok, perhaps that's my definition). I
>>definitely did not mean harmonic generation (although that is non-linear
>>also).
> 
> 
> Sorry, a notch filter is a linear operation.  

Of course.

> In theory, it can be
> compensated for with a linear filter with an inverse characteristic.

On second thoughts I see that at every point we have a sum of 
reflections (original signal filtered with the reflection 
characteristics of the surface) - this could be modelled with a 
transversal filter with IIR filters instead of simple tap coefficients. 
I guess that's linear. But it still isn't correctable with an eq, 
because every point in the room has a different frequency response.


> Sound contractors do that all the time using an RTA.

Yeah right - I used to do that as well. However, it doesn't result in an 
overall flat frequency response, and I explained why in my previous 
post. Usually your measurement does not correlate very much with your eq 
setting (unless you are standing right before or beneath the louspeakers).

> Sure, you cannot perfectly correct a room for all locations, but you can
> improve the frequency response for a large number of locations, perhaps even
> the majority of seats.

You can try to correct an overall trend (sound is muffeled, flat, washy, 
  etc.) - but the effect of EQ is negligible compared to the acoustic 
change when a room is half or fully or quarter occupied with people. 
It's just not worth the bother.

> 
> 
>>The only justification I see for equalizers in PA systems is for feedback
>>suppression.
> 
> 
> Of course, if the room was perfectly flat, there would be no dominant
> frequency to feed back!  In theory, you would have no feedback until all of
> the sudden, every frequency would feed back.  When you equalize for feedback
> suppression, you care correcting the system frequency response.  When this
> is done right, this also tends to correct the room frequency response (OK,
> also microphone/speaker frequency response).

Feedback correction is just a couple of tight notch filters - this 
hardly evens out any frequency response or change the "timbre" of the 
program material. It just cancels resonant frequencies.

Regards,
Andor

Jerry Avins wrote:
> 
> >
> >>>>All
> >>>>you have to do with an equalizer is compensate the major non-idealities
> >>>>caused mainly by the speakers and the acoustic environment.
> 
> I believe he meant, "All you _can_ (or maybe, ought to) do ...", but he
> can clarify that.

The equalizer can significantly flatten the response of an audio system.
I did that many times and the improvement may be very noticeable.  


> > - it seems that at least some people are of the opinion that equalizers can
> > be used to correct loudspeaker (or room) response.
> 
> Indeed they can. 

You bet.

> No equalizer can flatten the spiky response that shows up if
> you run the response curve slowly enough.

Flattening the major spikes is quite simple task for the parametric EQ.
The notch is more difficult to correct because of the limited power
available, however the notches in the response are less noticeable then
the spikes.

> >>> You cannot equalize a room, because a room does
> >>>not have a well-defined frequency response.

That is correct. However you can equalize the speaker axis emission in
ideal non-echoic room. That makes the speaker sound better in general.  

> >>
> >>If that were germane, one would need different equalization for every
> >>cubic foot of space in the room. The phenomenon is real, but it
> >>matters very little.

I have seen acoustic arrays which were pretending to do that. The idea
is similar to the holographic technology.

> > The only justification I see for equalizers in PA systems is for feedback
> > suppression.

That is done better with LMS feedback cancellation and/or adaptive
notches.

But what really surprises me in all that arguments about FIRs ans IIRs
that very few people understand that there is no direct relation between
the "ringing" and the phase linearity. The passive RC filter can have
very nonlinear phase, however it exibits no ringing.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

Jerry Avins wrote:

> glen herrmannsfeldt wrote:

>> I had a physics lab class once where we measured the impedance, 
>> attenuation, and propagation velocity for different cables.  One was a 
>> coax cable with a spiral wound center conductor which increases its 
>> inductance relative to capacitance, for a velocity of about 0.1c.  It 
>> was designed for use as a delay line.

> An interesting feature of that cable is that if you take the length of 
> the stretched-out inner conductor as the length of the cable, The group 
> velocity comes out not much different from RG59U. You could say that the 
> signal takes so long to come our because it gets dizzy along the way. 
> Did the cable have a blue jacket?

I think the cable is commonly used in physics experiments, such as 
connecting photomultiplier tubes to the associated electronics.  I don't 
remember the color, though.

Yes, it is interesting that the inductance of a coil works out just 
about right so that the propagation velocity scales with wire length.

I also worked some on TWT (traveling wave tube) designs which use a 
similar helical structure to slow down the wave.

If you tried to argue that the signal was following the helix, though, 
how do you explain that the outer conductor is not helical?  Besides, 
everyone knows that signals don't travel in wires, but in the space 
between the wires.

-- glen

glen herrmannsfeldt wrote:
> Jerry Avins wrote:
> 
>> glen herrmannsfeldt wrote:
> 
> 
>>> I had a physics lab class once where we measured the impedance, 
>>> attenuation, and propagation velocity for different cables.  One was 
>>> a coax cable with a spiral wound center conductor which increases its 
>>> inductance relative to capacitance, for a velocity of about 0.1c.  It 
>>> was designed for use as a delay line.
> 
> 
>> An interesting feature of that cable is that if you take the length of 
>> the stretched-out inner conductor as the length of the cable, The 
>> group velocity comes out not much different from RG59U. You could say 
>> that the signal takes so long to come our because it gets dizzy along 
>> the way. Did the cable have a blue jacket?
> 
> 
> I think the cable is commonly used in physics experiments, such as 
> connecting photomultiplier tubes to the associated electronics.  I don't 
> remember the color, though.
> 
> Yes, it is interesting that the inductance of a coil works out just 
> about right so that the propagation velocity scales with wire length.
> 
> I also worked some on TWT (traveling wave tube) designs which use a 
> similar helical structure to slow down the wave.
> 
> If you tried to argue that the signal was following the helix, though, 
> how do you explain that the outer conductor is not helical?  Besides, 
> everyone knows that signals don't travel in wires, but in the space 
> between the wires.
> 
> -- glen

I could argue for the space _around_ the wires, but that's quibbling.
It creates in me a wonder at the fitness of things to know that that
component of the Poynting vector perpendicular to the power-line
conductor heats it, while that component parallel to it is delivered
downstream. Trying to visualize that while contemplating a drooping
catenary is awesome. 8-)

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

"Andor Bariska" <andor@nospam.net> wrote in message
news:3fbb3cfa$1@pfaff2.ethz.ch...
> Jon Harris wrote:
>
> >>Reflections cause phase-cancellation, which cause notch-filter-type
> >>behaviour in a frequency response. This is non-linear, because it cannot
> >
> > be
> >
> >>compensated by a linear system (ok, perhaps that's my definition). I
> >>definitely did not mean harmonic generation (although that is non-linear
> >>also).
> >
> >
> > Sorry, a notch filter is a linear operation.
>
> Of course.
>
> > In theory, it can be
> > compensated for with a linear filter with an inverse characteristic.
>
> On second thoughts I see that at every point we have a sum of
> reflections (original signal filtered with the reflection
> characteristics of the surface) - this could be modelled with a
> transversal filter with IIR filters instead of simple tap coefficients.
> I guess that's linear. But it still isn't correctable with an eq,
> because every point in the room has a different frequency response.

We agree here.

> > Sure, you cannot perfectly correct a room for all locations, but you can
> > improve the frequency response for a large number of locations, perhaps
even
> > the majority of seats.
>
> You can try to correct an overall trend (sound is muffeled, flat, washy,
>   etc.) - but the effect of EQ is negligible compared to the acoustic
> change when a room is half or fully or quarter occupied with people.
> It's just not worth the bother.

No direct experience, you might be right about that.

> >>The only justification I see for equalizers in PA systems is for
feedback
> >>suppression.
> >
> >
> > Of course, if the room was perfectly flat, there would be no dominant
> > frequency to feed back!  In theory, you would have no feedback until all
of
> > the sudden, every frequency would feed back.  When you equalize for
feedback
> > suppression, you care correcting the system frequency response.  When
this
> > is done right, this also tends to correct the room frequency response
(OK,
> > also microphone/speaker frequency response).
>
> Feedback correction is just a couple of tight notch filters - this
> hardly evens out any frequency response or change the "timbre" of the
> program material. It just cancels resonant frequencies.

But you have to ask yourself, why did the feedback eliminator pick those
particular notches to fix the feedback?  The answer is that the system
response has a peak at that frequency!  Hence my assertion that filtering
done my feedback elimination devices does tend to flatten a system response.
Not that it's perfect correction by any means, but it should be at least
moving in the right direction.

Jerry Avins wrote:

> glen herrmannsfeldt wrote:

(snip regarding propagation velocity in coax cables with high inductance 
center conductors)

>> If you tried to argue that the signal was following the helix, though, 
>> how do you explain that the outer conductor is not helical?  Besides, 
>> everyone knows that signals don't travel in wires, but in the space 
>> between the wires.

> I could argue for the space _around_ the wires, but that's quibbling.
> It creates in me a wonder at the fitness of things to know that that
> component of the Poynting vector perpendicular to the power-line
> conductor heats it, while that component parallel to it is delivered
> downstream. Trying to visualize that while contemplating a drooping
> catenary is awesome. 8-)

Depending on the geometry between, around, or something in between.

Between isn't too far off for coax, though.  For twisted pair it is 
mostly between, but some a little farther out.

Pretty often in the ethernet newsgroup it is necessary to explain the 
effect of mispaired cables.  If signals really travel inside a wire then 
mispairing would not have any adverse affect.

Though there was once a system for transmitting signals on a single 
wire, so in that case around would be the right description.

-- glen

glen herrmannsfeldt wrote:

   ...

> Though there was once a system for transmitting signals on a single 
> wire, so in that case around would be the right description.
> 
> -- glen

My father, a sheet-metal worker, made a number of G-string launching
cones for some of my ham friends. I don't think anyone ever came up
with a satisfactory explanation of the wave distribution in the case
of curved wire. (That catenary again!) So a G-string is not exactly 
co-ax with an infinite-diameter outer conductor.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Jerry Avins wrote:
> glen herrmannsfeldt wrote:
> 
>   ...
> 
>> I had a physics lab class once where we measured the impedance, 
>> attenuation, and propagation velocity for different cables.  One was a 
>> coax cable with a spiral wound center conductor which increases its 
>> inductance relative to capacitance, for a velocity of about 0.1c.  It 
>> was designed for use as a delay line.
> 
> 
> An interesting feature of that cable is that if you take the length of 
> the stretched-out inner conductor as the length of the cable, The group 
> velocity comes out not much different from RG59U. You could say that the 
> signal takes so long to come our because it gets dizzy along the way. 
> Did the cable have a blue jacket?
> 
> Jerry

Hmmm if that's generally true as a first order approximation, it might 
explain why something I tried as an ignorant teenager interested in 
Ham radio _apparently_ worked. I needed/wanted a halfwave dipole. Did 
not have space. Used a couple of handy pieces of Al tubing as long as 
possible. Made up rest of length required with wire loosely wrapped 
around .5->.75 inch form. OK, so that was >40 yrs ago ,]
As I recall, it did work.

Richard Owlett wrote:

> Jerry Avins wrote:
> 
>> glen herrmannsfeldt wrote:
>>
>>   ...
>>
>>> I had a physics lab class once where we measured the impedance, 
>>> attenuation, and propagation velocity for different cables.  One was 
>>> a coax cable with a spiral wound center conductor which increases its 
>>> inductance relative to capacitance, for a velocity of about 0.1c.  It 
>>> was designed for use as a delay line.
>>
>>
>>
>> An interesting feature of that cable is that if you take the length of 
>> the stretched-out inner conductor as the length of the cable, The 
>> group velocity comes out not much different from RG59U. You could say 
>> that the signal takes so long to come our because it gets dizzy along 
>> the way. Did the cable have a blue jacket?
>>
>> Jerry
> 
> 
> Hmmm if that's generally true as a first order approximation, it might 
> explain why something I tried as an ignorant teenager interested in Ham 
> radio _apparently_ worked. I needed/wanted a halfwave dipole. Did not 
> have space. Used a couple of handy pieces of Al tubing as long as 
> possible. Made up rest of length required with wire loosely wrapped 
> around .5->.75 inch form. OK, so that was >40 yrs ago ,]
> As I recall, it did work.

Well, a pair of opposed whips is a dipole. An opposed pair of end-loaded
whips is an end-loaded dipole. Why not? It should work!

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������