Back when I designed that filter, the coils I used were fairly
expensive, so I used the varying impedance to allow using the same coils
for all the frequencies. Plus large value capacitors were not easy to
acquire which made the higher impedance nicer using smaller capacitors.
Today its possible to buy large polyester film capacitors for about the
price of the transformers and smaller value inductors are as cheap as 18
cents each or as expensive as $1.75 each from Mouser and JW Miller so
reuse is not so critical. So I have reverse engineered my filters and
designed a new set without transformers with a more useful set of cut
off frequencies, but they do take capacitors 10 to 47 mfd which are in
the class of motor run capacitors or switching snubbers. Tantalum or
aluminum electrolytic capacitors won't work without imposing a DC bias
and that can make the capacitors into a noise source which we are trying
to eliminate from the speaker or headphones.
I have chosen cut off frequencies of 600, 1200 (for CW) 1800, 2400, and
3KHz (for SSB, the last for HIFI SSB.) I haven't built them yet so I
haven't tested. These are designed for 8 ohms impedance. When driving
higher impedance headphones they need to be terminated with a suitable
resistor to load them to 8 ohms or else an impedance matching transformer.
The filter circuit is the same as my speaker filter mentioned earlier
without the transformer and switching the whole filter instead of the
inputs and outputs and the capacitors.
Three inductors in series, the outer ones are the smaller, the middle is
1.5 times the outer inductor.
E.g. for 600 Hz low pass, input and output inductances 4.1 mh, center
6.15 mh. Capacitors from the inductor junctions to ground (2 caps) 47 mfd.
1200 Hz, inductances 2 and 3 mh, Cs 24 mfd.
1800 Hz, inductances 1.35 and 2 mh, Cs 15 mfd.
2400 Hz, inductances 1 and 1.5 mh, Cs 12 mfd.
3000 Hz, inductances .82 and 1.22 mh, Cs 9.5 or 10 mfd.
Going from the available inductors in the JW Miller pages of the Mouser
catalog, I'd use 4700 and 1500 microhenries in series to make 6 mh. 3300
and 680 microhenries in series to make 4 mh. 2500 and 470 in series to
make 3 mh. 1000 + 1000 to make 2 mh. 1000 + 330 to make 1.35, 1000 and
1500 are stock values. 1000 and 220 to make 1.22 mh, and 820
microhenries is a stock value.
For 9.5 mfd, I'd use 10. For 12 I'd use 12 or 10 + 2 in parallel. For 15
I'd use a 15 or a 10 and a 4.7 in parallel. For 24 I'd use 24 or two 12s
or two 10s and a 4.7 in parallel. For 47, I'd consider a 47 (it is in
the mouser catalog, but I'm not sure its in stock, might be easier to
get from W.W. Grainger or a motor repair shop) or 4 x 12 or 5 x 10 in
parallel.
To recap the circuit, say for 2400 Hz lowpass: input to a 1 mh inductor,
then 12 mfd to ground. From the junction of the inductor and capacitor a
1.5 mh inductor. From the free end of that inductor a 12 mfd to ground.
From that coil-capacitor junction a 1 mh to the output pin.
I'll write it up more and maybe build some to test the frequency
variations from rounding off the component values and using 10%
tolerance parts.
73, Jerry, K0CQ
On 8/12/2010 4:44 PM, John Graves wrote:
Jerry,
That is a very interesting article. 10 years ago you could have bought
most of the components at Radio Shack. I just had a quick look at the j.
W. Miller line....Winding toroids seem to be the way to go now. Now to
find that switch you used. Only commenting. What you have said makes
sense to me (a network security person) My Dad was the mostly self
taught audio guy (Thanks to WWII).
John / WA1JG
Dr. Gerald N. Johnson wrote:
One solution I use for really noise free audio is posted at:
http://www.geraldj.networkiowa.com/papers/speakerfilter.pdf This
filter has selectable cut off frequencies and being passive and built
for speaker impedance there is no high frequency noise introduced
after it by 100 KHz bandwidth audio output ICs.
73, Jerry, K0CQ
|