PDA

View Full Version : Active Audio Filtering



KG4CGC
12-22-2009, 01:00 AM
Hiya :)
Looking for schematics for AF filtering made from commonly obtainable parts. DSP a plus :p

N2NH
12-22-2009, 08:41 AM
How's this for starters?

http://www.arrl.org/tis/info/dsp.html

http://www.arrl.org/search/newsearch.ph ... cessing%22 (http://www.arrl.org/search/newsearch.php?q=audio+%22digital+signal+processing %22)

http://www.arrl.org/tis/info/pdf/9609x008.pdf

N1LAF
12-22-2009, 08:45 AM
Hiya :)
Looking for schematics for AF filtering made from commonly obtainable parts. DSP a plus :p

Charles, I have to take care of business this morning, but when I get back, I'll help you out here. You can purchase a small prototype board, a few inexpensive op-amps, and various resistors (resistor kit) and non-polarized capacitors between 1uf and 10 uf. We will build some filters!

In the mean time, do an internet search on Sallen-Key low pass and high pass filters. We will start from there, and post the links here so we have a common source to discuss.

X-Rated
12-22-2009, 10:41 AM
Chebyshev or Butterworth?

WØTKX
12-22-2009, 11:10 AM
But, but, we don't need no stinkin' circuits... :snicker:

http://www.io.com/~n5fc/res_spkr.htm

http://www.io.com/~n5fc/res_spkr_sketch_sm.jpg

N4VGB
12-22-2009, 02:12 PM
But, but, we don't need no stinkin' circuits... :snicker:



Reminds me of the old Dixie Cup cw speaker. Years ago some ham accidently discovered that a particular size of the plastic Dixie Cup line of plastic cups had a resonant freq of exactly 750 or 800hz? Outside it's self resonant freq it would hardly respond at all. I built one and it worked like a charm.

It was about like the 100hz bandwidth audio filter in an R-390A, very careful tuning required or your received station went away instantly.

KG4CGC
12-22-2009, 04:59 PM
Chebyshev or Butterworth?
I don't know that it matters yet. I've looked at some stuff online, there are a lot of schematics out of the UK and Asia. Note that values there need some translation, for example: a 4K7 resistor in American terms translates to 4.7K.
Yes, I want to use commonly available parts from RS. Common ICs and transistors. I built a CW filter some years ago from a Vectronics kit but it had 4 set values and sounded hollow. Not good on the ears. Tried some freeware DSP software but I can't get past the excessive oversampling that results in echo echo echo. Using that last night, DSPFIL V1.0 2000 and BR-universal-Filt resulted about the same but the BR program seemed a little better other than having to figure out German to English translations. Yeah, stuff gets stuck in your brain from years of comedy in HS and somehow helps you figure out what foreign words mean LOL.

I may fiddle about with that old Vectronics kit some. I also have an assembled and unassembled amp consisting of a 1458 and a 386 op amps. I believe I can add variables "R's" and switched caps and fiddle about with those results.

KG4CGC
12-22-2009, 05:00 PM
Hiya :)
Looking for schematics for AF filtering made from commonly obtainable parts. DSP a plus :p

Charles, I have to take care of business this morning, but when I get back, I'll help you out here. You can purchase a small prototype board, a few inexpensive op-amps, and various resistors (resistor kit) and non-polarized capacitors between 1uf and 10 uf. We will build some filters!

In the mean time, do an internet search on Sallen-Key low pass and high pass filters. We will start from there, and post the links here so we have a common source to discuss.

Very cool. Stuff that can be inserted in the first IF section would be of interest too :)

N4VGB
12-22-2009, 05:29 PM
Chebyshev or Butterworth?
I don't know that it matters yet.



It's hard to beat a nice simple Butterworth for voice band audio.

KG4CGC
12-22-2009, 05:36 PM
Chebyshev or Butterworth?
I don't know that it matters yet.



It's hard to beat a nice simple Butterworth for voice band audio.
That's certainly something to take into consideration.

N1LAF
12-22-2009, 07:54 PM
Charles, Are you looking at a bandwidth of 20Hz to 5Khz, or 20Hz to 9Khz?

KG4CGC
12-22-2009, 08:32 PM
Anything really. :)
[On edit on purpose] Eventually, I plan to make different parts (sections) modular. Being able to swap out different front ends, pre-amps, detectors, audio etc on the fly.

N1LAF
12-22-2009, 09:25 PM
Try this - I haven't bread boarded it yet, but in theory should work
[EDIT] Gain should be 1 to 11x
R1, R2 double ganged potentiometer.

fc (HP) = 1/(2pi * R1 * C1)
fc (LP) = 1/(2pi * R2 * C3)
R1A = R1B
R2A = R2B
C1 = C2
C3 = C4

KG4CGC
12-22-2009, 09:44 PM
That's your design?
I'll give it a shot this weekend. Double balanced astable vibrator?

KG4CGC
12-22-2009, 10:12 PM
Upon further inspection, it looks like a similar but simplified version of this. Of course my knowledge is limited.

http://i25.photobucket.com/albums/c79/bebop5/autekresearch_com_qf1.png

N1LAF
12-22-2009, 10:19 PM
The circuit I provided are Sallen & Key Low pass and high pass filters. It is arranged to use single voltage supply, not plus and minus voltage supplies.

KG4CGC
12-22-2009, 10:21 PM
Thanks :agree:

N1LAF
04-27-2010, 08:24 PM
We have DSP that can be implemented in software.

Take something like a PIC chip or equivalent. Write a program that reads the analog inputs, convert to double precision floating point, process, and then back to digital and output.

I did some data simulation, and son of a gun, the filtering works.

Beginning steps:

1. y[n] = (b0/a0)*x[n] + (b1/a0)*x[n-1] + (b2/a0)*x[n-2] - (a1/a0)*y[n-1] - (a2/a0)*y[n-2]

where x[n] is present data, x[n-1] is last data, and x[n-2] is the second to last data processed.
y[n] is the output of the equation, y[n-1] is the last result, and y[n-2] is the second to last result.

a0-a2, b0-b2 are constants

This is the basic equations.

2. Setup.

Fs = Sampling frequency
Fo = filter corner/center frequency
Q = similar to quality factor

w = 2*pi*Fo/Fs
wc = cos(w) in radians
ws = sin(w) in radians

qa = ws/(2*Q)

3. Filter coefficients

For low pass filtering:
b0 = (1-wc)/2
b1 = 1 - wc
b2 = (1-wc)/2
a0 = 1 + qa
a1 = -2*wc
a2 = 1 - qa

For high pass filtering:
b0 = (1+wc)/2
b1 = -(1+wc)
b2 = (1+wc)/2
a0 = 1 + qa
a1 = -2*wc
a2 = 1 - qa


When I prove out the bandpass and notch, I'll add it to the collection here. Screen prints to follow soon

N1LAF
04-27-2010, 08:44 PM
Here we have two waveforms added together, 50 Hz and 5000 Hz. Applying a high pass filter with a corner frequency of 1000 Hz results in a filtered, 5000 Hz waveform only.

Second graph following expands the view.

N1LAF
04-27-2010, 08:46 PM
Now we use the low pass option. Same two frequency components, but this time we set the corner frequency to 500 Hz. Note only the 50 Hz signal is recovered.

N1LAF
04-27-2010, 08:49 PM
By adding a moderate amount of noise, most of the noise is filtered out.

N1LAF
04-27-2010, 08:50 PM
Drop the corner frequency, and most of the noise has been filtered out. There is some phase shifting.

N1LAF
04-27-2010, 08:57 PM
LabVIEW Implementation of the Low Pass Filter

KG4CGC
04-27-2010, 10:34 PM
This is cool stuff. :agree:

N4VGB
05-04-2010, 05:25 PM
Barkeep, you do realize you're probably typing on a very good active audio filter???

PCs mostly have very good soundcards these days and if you don't like the basic sliders provide to tailor audio provide on your PC, there are a ton of free shareware programs on the net to process audio in PC soundcards.