Low-distortion Audio-range Oscillator

[richiem]

Tune the filter peak to the 1kHz fundamental and let the skirts take care of themselves -- and it does depend on Q. The primary advantage of a bandpass over a lowpass is deeper attenuation at the 2nd H.

[dirkwright]

Originally Posted by richiem

Tune the filter peak to the 1kHz fundamental and let the skirts take care of themselves -- and it does depend on Q. The primary advantage of a bandpass over a lowpass is deeper attenuation at the 2nd H.

What should be the impedance of the filter?

I'm going to need a good sized box to hold all of this stuff!

[scott wurcer]

Originally Posted by Samuel Groner

I'm answering my first Q myself now that I have the preprint in front of me--20 kHz. Would be interested to see the filter schematic (i.e. order and impedance niveau).

2nd order 600 Ohms. Don't remember the values. #16 copper magnet wire and the best film caps we had.

Samuel

[richiem]

@dirkwright -- difficult to say -- it majorly depends on the output capabilities of the device doing the driving. In the case of low-distortion oscillators, I would shoot for a high peak Z, but I haven't done a serious LC bandpass filter in years and just don't remember the practical consequences of various issues. There probably are many sites online that can help with basic design parameters and results.

[dirkwright]

Originally Posted by richiem

@dirkwright -- difficult to say -- it majorly depends on the output capabilities of the device doing the driving. In the case of low-distortion oscillators, I would shoot for a high peak Z, but I haven't done a serious LC bandpass filter in years and just don't remember the practical consequences of various issues. There probably are many sites online that can help with basic design parameters and results.

Wouldn't the filter determine the effective source impedance for the oscillator/filter combination? I'm not sure what kind of load Victor's oscillator can drive, so I don't have a clue as to what impedance to make the filter. I do know that this online calculator I'm using gives me huge inductors for 10kohms characteristic impedance.
http://www.the12volt.com/caraudio/crosscalc2.asp#cc

So, assuming the oscillator can't drive very much, I'd need a buffer between it and the filter if the filter was low impedance.

[richiem]

Well, a bandpass filter will have relatively hi Z at the peak, where the signal level is also highest, and lower Z in the stop band. So using some series R at the output of the oscillator will be a good idea -- I suggest a couple of kohms just to keep the load on the oscillator light, even in the stop band. The resulting insertion loss shouldn't be high at the fundamental, which is the one and only frequency you want.

[dirkwright]

OK, thanks. So, if the filter is 600 ohms, then that would be the source impedance for the oscillator/filter combination, and I can insert a couple of kohms between the oscillator and the filter to keep the oscillator happy. Sounds good. I don't have to make or buy monster inductors!

[richiem]

Well, you do have to worry about the loss between the osc, output and the filter input at the filter's peak frequency. I would design the filter for a higher nominal Z, but as I said, it's been awhile since I did any work with passive LC bandpass filters and I just don't remember the Z characteristics at all. Just see what you can find online as far as designs go. It doesn't take much Q to do a really good job -- I'd shoot for anything over 6 but less than 60....

[dirkwright]

Plot for 18dB/octave, 600 ohms, terminated with 600ohm resistor. Pretty good I think.

[1audio]

That's pretty impressive. What does it spec for parts? And make sure the inductors are real (have realistic DCR). FWIW I use the free version of this Downloads which works really well. Then move the design to LTSpice to tweak.