Source code for process control...

filia&sofia wrote:

If you don't know whether you need a simple PID loop or an adaptive loop
you're probably not ready to start downloading code and randomly
applying it.

What are you tuning?  For any controller to work there has to be a
defined input/output relationship that you can affect to control it, and
that you can sense to use for feedback.  If this I/O relationship is
reasonably straightforward and does not change then you don't need
adaptive control; if this I/O relationship is reasonably linear then a
fairly simple loop will do -- you may not even need a full PID loop.

If you already know that you have something that's fit for PID control
you can see my article at
http://www.wescottdesign.com/articles/Sampling/pidwophd.html .

I wouldn't recommend that you try applying an adaptive control solution
unless you really understand it thoroughly.  In fact, I would recommend
that you investigate every alternative that you can to adaptive control
-- it's a powerful method, but it's not a magic bullet that you can
casually apply and have everything turn out right.

--

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

Posting from Google?  See http://cfaj.freeshell.org/google/

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

I am trying to control audio source. So the input would be audio
stream, and output would be, again, audio stream, which is tuned to
make constant audio power level. At this point, it seems that PID-
control is proper choice for the task, but since the task is to
control audio stream on a computer, I need an PID-algorithm to do
this. The  article was good and it gave me good insight of what my
program should do and how it should be done.

filia&sofia wrote:

If you want to keep the volume flat, look up volume compressors. They
have been around for a long time. (You will need to become familiar with
attack and release times ans be able to set them properly.)

If that's not what you want, and PID is really appropriate, then I don't
understand what you ask for.

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

You do not need a PID algorithm to do this.  A standard PID loop is not
frequency-sensitive.

If you're just trying to feed an audio source into a computer, what you need
is called an Automatic Gain Control (AGC).  A compressor (on it's own) won't
do the job properly because it doesn't expand the quiet bits.

Tri-band AGC's (as used by radio & tv stations worldwide) are available
off-the-shelf from several professional audio suppliers including
www.aphex.com

Google is your friend.

Cameron:-)
 

[snipped response]

Apologies to all - I was on the phone and hit send too fast!...

What I should have said was that a standard PID loop is not
frequency-sensitive *enough for the application*.   An AGC *does* contain a
feedback loop - that's how it works.

Cameron:-)

Cameron Dorrough wrote:

   ...


But ir usually *doesn't* have an integrator and only rarely an
differentiator.

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

On Thu, 24 May 2007 22:16:00 -0400, Jerry Avins wrote:


The only AGC circuit I ever implemented in a radio had an integrator -- it
was for a data link application that was feeding an 8-bit ADC into the
processor.

--
Tim Wescott
Control systems and communications consulting
http://www.wescottdesign.com

Need to learn how to apply control theory in your embedded system?
"Applied Control Theory for Embedded Systems" by Tim Wescott
Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html

Tim Wescott wrote:

Interesting! The closest thing to an integrator in the ones I used --
and used to see -- was the lowpass filter.

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

On Fri, 25 May 2007 23:50:32 -0400, Jerry Avins wrote:


I did it for my Master's thesis.  Afterward, a friend of my thesis adviser
mentioned that an integrating AGC doesn't sound good in an AM system
because it just amplifies noise, but that it isn't a bad way to go in a
digital system.

--
Tim Wescott
Control systems and communications consulting
http://www.wescottdesign.com

Need to learn how to apply control theory in your embedded system?
"Applied Control Theory for Embedded Systems" by Tim Wescott
Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html

Tim Wescott wrote:

   ...


History as I remember it: AGC in receivers (mostly AM receivers) was
done at IF by varying the bias on so-called "variable-mu"  or "remote
cutoff" tubes. The distortion products created by the bias-dependent
transfer characteristic were primarily at harmonics of the IF and
filtered out by the selective coupling between stages. AGC is used in
receivers to keep the carrier level at the detector constant, not the
demodulated volume. Better receivers had an AGC threshold, below which
no gain reduction was applied. That was called "delayed AGC".

Relatively distortion-free voltage-controlled gain is harder to get at
audio. Devices that did that were called compressors to distinguish them
from the 6SK7-style variable-mu approach. At least one early compressor
for broadcast studios modulated the signal to a low RF, varied the RF
gain, and restored the signal to audio with a synchronous detector. In
my lexicon at least, "volume compression" and "audio AGC" are different
names for the same collection of techniques.

Naturally, perfect audio AGC acting instantaneously would remove all of
the signal. To avoid that, some sort of low-pass filtering is always
used. Separately adjustable attack and release times are common, as well
as gain suppression when the signal becomes too small to be useful.

Please expand my picture by telling me how an integrator fits into it.

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

Cameron Dorrough wrote:

   ...


That depends on the implementation. With enough gain before volume
compression takes place, the output can be made quite constant.

   ...

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

Ahh...not really.  What you end up with is an overall increase in background
noise level.

A compressor reduces the volume of the peaks (reduces dynamic range).  An
AGC makes loud bits softer and soft bits louder, without affecting the
dynamic range (or the noise floor if set up properly).

Cameron:-)

Cameron Dorrough wrote:

The compressor built into the microphone amplifier for the Mercury space
capsule had a 2 dB output change for a 60 dB input change*. The
headphone amplifier had only a 40 dB dynamic range, AGC in the receiver
taking care of the rest.

Whenever gain to makes softer bits louder, noise is made louder as well.
How could it be otherwise?

Jerry
____________________________
* The spec was 3 dB. We accomplished it with diodes, resistors, and
capacitors. Distortion never reached 1% THD even at the highest input
levels.
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

Hi, Jerry.  Using a compressor and turning up the volume increases the
background noise level continuously - start, end and at all times during
transmission.

An AGC will not increase background noise level during loud bits (it will
probably decrease it) and will only increase noise level during quiet bits
(volume increase) and then not much at all ifg properly set up, then once
the input level falls below a set threshold, it will turn off the output all
together.

Cameron:-)

Cameron Dorrough wrote:

   ...


How so? When the compressor cuts the gain back, everything through the
amplifier is reduced. We probably mean different things. As I see it, a
compressor is an amplifier with AGC. Here's what we did:

Microphone (5 ohm, 5 millivolts shouting). Matching transformer for 500
ohm load. (50 millivolts at output.) Current-controlled attenuator at
the input to the first base, followed by 80 dB of gain to the
transformer-coupled output. An extra winding on the output primary went
below ground before the collector hit the rail, pulling current through
the attenuator. 5 microvolts was enough to provide -3 dB audio to
modulate the transmitters. 5 millivolts gave only 2 dB more. I managed
to make the amplifier stable on a circuit board 1.4 x 6 inches using all
discrete components (including both transformers). Whew!


We suggested a squelch, but NASA said no because it would have been a
frill that could go wrong. With everything doubly redundant, there were
two noise-canceling throat (contact) microphones, each with its own
AGCed amplifier chain. Background noise wasn't an issue. When the
astronaut wasn't talking, it didn't matter what he wasn't saying.

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

[snipped stuff]

Yes, I think we do mean different things.

The "compressor" part of a "modern professional audio compressor" (likes of
Aphex, Beringer, and numerous others out there) only reduces gain by a set
amount (say 3db) above a set threshold (say 0dB) _leaving the rest of the
the audio stream untouched_.

In this way, the peaks are gently limited and the overall "volume" (as you
hear it) can be increased without risk of distortion by clipping.  There is
no AGC involved (not in a true "compressor" anyway).

It is common to find such things as a "Tri-band
AGC/Compressor/Expander/Limiter" used for radio & TV broadcast that splits
the incoming audio stream into high, mid and low frequencies, adjust the
level of each to what you want, squashes the peaks, lifts the low spots,
blocks any spikes and pops and then recombines it all for transmission...
and from the sound of it would have been ideal for the Space Program. :-)

Cameron:-)

Cameron Dorrough wrote:

Maybe so. I wish car ratios had a way to reduce the dynamic range of a
received signal. When soft passages of, say, Schubert are loud enough to
be heard despite road noise, crescendos get to be unpleasantly loud.

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