inductors for audio frequency LC filters?

[ssbothwell KJ6RSG]

what kind of filters do you guys use at audio frequencies?

i want to build a 10kHz brickwall low pass filter. i came up with a design using RFSim99 which i am attaching below. it calls for 12mH and 22mH inductors which seem really huge to me.

can i use power inductors such as this: http://www.mouser.com/Search/ProductDetail.aspx?R=07MFG-123J-50virtualkey60130000virtualkey434-02-123J

or do i need to wind toroids? what material would you use for mH size inductors?

[Steve - K4HX]

Use toroids for that sort of inductance.

I'd go with an active filter and skip the inductors.

[W4NEQ]

+1 on skip the inductors.  Used to be the surplus 88 mhy torroids were used for this sort of thing, but with opamps it's easier.  Download the TI "FilterPro" from their site, and inside of 30 minutes you'll have several alternatives worked ...

[W7TFO]

Active and/or RC networks are much better at audio frequencies due to the inherent phase shift of LC nets.

73DG

[W1DAN]

Hi:

A passive L/C filter will work fine. Toroid-wound inductors are best here. More info on design in old handbooks as this was used in the tube days.

Today, I use a Maxim MAX294 switched capacitor filter. Brick wall and I can adjust the corner frequency with a timing capacitor adjustment.

http://www.maxim-ic.com/datasheet/index.mvp/id/1443

A main tradeoff is S/N, but is not an issue for me as I use this filter after my processing. After the filter I should use a phase-equalizing network to correct for the tilt, but use a diode to clip the negative peaks that overshoot instead. This replaces the analog op-amp based FDNR circuit. See 8.70 here:

http://www.analog.com/library/analogDialogue/archives/43-09/EDCh%208%20filter.pdf

Dan

[WBear2GCR]

Skip the standard cascaded "pi" filters.
Go directly to elliptical or Cauer filters - active or LCR...

Fewer elements, much much steeper cutoff slope.

Afaik, the phase shifts will be virtually identical for active and passive filters of the same "alignment".

The size of the inductors will depend on the current and voltage being passed - line level stuff is pretty small. UTC used to make fixed inductors for 600 ohm studio level stuff - but today you can buffer the in and out of the filter with opamps or buffer chips, or discrete parts and make the impedance high, and so make the parts smaller in value and lower in power as well!

I like 'em.


                    _-_-bear

[W9BHI]

Go to Frequency Devices.com
They have everything in the way of filters.
Even DSP.

W9BHI

[DMOD]

Steve Cloutier has some suggested circuits in the ClassE section:

http://www.classeradio.com/easy_e_pwm_rev_c_right_half.pdf

http://www.classeradio.com/easy_e_pwm_alternate_filters.pdf

Phil - AC0OB

[ssbothwell KJ6RSG]

thanks for all the advice. i think i will go with an active filter on this. ive never made anything with op-amps and this seems like a great chance to learn about them.

bear thanks for the tip on elliptical and cauer filters. i'll look them up.

[DMOD]

yes active filters are really cool.

If you do decide to go passive, I have various inductors designed specifically for passive audio.

Phil - AC0OB

[ssbothwell KJ6RSG]

hi guys,

this message is aimed at dan and anyone else who has used those max29* ICs.

i received some max294,293,and 295 filter IC samples from maxim today. i have a couple questions.

the example circuit in the datasheet is setup for an external clock. they say that in order to use an internal clock you place a capacitor on the clock input pin. do they mean a capacitor from ground to the clock input or do they mean something else?

also they give the formula for the capacitor as: Fosc = 10^5 / 3Cosc

would i be correct to say that i want a 20,000pF cap for a 5khz cutoff?

lastly, i am currently operating the device with no external clock input and no capacitor. the lower half of the output waveform is severely clipped. is this to be expected?

edit: i solved the clipping problem with a 100nF cap on the audio input. also think i misread the equation for clock capacitor. with a 10pF cap i am getting a cutoff around 10kHz. i think i am going to make the device switchable between 5kHz and 10kHz cutoffs. are ceramic disc caps suitable for this circuit or should i be using some other type?

also, should i buffer the audio input?

[ssbothwell KJ6RSG]

i built a prototype filter using a TL071 as a butterworth highpass at 50cycles -> 10kc butterworth using the max294 uncommitted opamp -> max294 10kc elliptical filter.

it seem to work pretty well. the final unit is going to be switchable between 3.5kc, 5kc, 7.5kc and 10kc using a rotary switch to cycle through switching capacitors. i included the 10kc butterworth for anti-aliasing purposes as recommended by the datasheet.  should i make that filter switch along with the main elliptical filter or would it be fine to keep it as a static 10kc filter?

also, how can i measure phase shift? once i know if there are any phase issues would an all-pass filter be the solution?

[Opcom]

On the inductors thing, some people like to use them for filters but high values like 0.5H to 5H can be hard to find. I have a nice little collection of them and most have come from the estates of electronics men (may they rest in peace) where I 'cleaned out' the unwanted electronic parts after it had been picked though by collectors and flippers. The majority are of military industrial quality. Little JS'd 3, 4, and 5 pole filters performed pretty well in tests as did a phase rotator which was a crude temporary copy of the Kahn unit.

[KM1H]

Ive never been satisfied with external active filters from HB, MFJ and Timewave, the audio is very uncomfortable and about the only thing I find useful is the notch filter for hetrodynes.

What about the various switching supply toroids? I see several in the 25H and higher category that hold promise even for low level modulators.

[W4NEQ]

... also, how can i measure phase shift? once i know if there are any phase issues would an all-pass filter be the solution?

To measure phase shift, use a dual channel Oscope, with first channel tied to input of filter (also sync) second channel looking at output.  Using a sinewave audio tone, tune the audio frequency through the passband and note the shift.  You will have shift, and a lot of it near the filter knees.   Does this matter?    Anytime you EQ audio you are creating phase shift. 

An all-pass filter is designed to "rotate" the peaks of midband and  or high frequencies opposite to the lower frequencies to lessen their addition and minimize asymmetry.  Improperly employed, it may have the opposite effect.   Sometimes useful, other times not.  It is not a magic cure-all.

Chris

[ssbothwell KJ6RSG]

... also, how can i measure phase shift? once i know if there are any phase issues would an all-pass filter be the solution?

To measure phase shift, use a dual channel Oscope, with first channel tied to input of filter (also sync) second channel looking at output.  Using a sinewave audio tone, tune the audio frequency through the passband and note the shift.  You will have shift, and a lot of it near the filter knees.   Does this matter?    Anytime you EQ audio you are creating phase shift. 

An all-pass filter is designed to "rotate" the peaks of midband and  or high frequencies opposite to the lower frequencies to lessen their addition and minimize asymmetry.  Improperly employed, it may have the opposite effect.   Sometimes useful, other times not.  It is not a magic cure-all.

Chris

i figured that it would be something simple like that for checking the phase. up to 1kc the phase shift is about 180 degrees but then they start to go back into phase. at ~5.5kc it is in phase and then it goes out of phase again as it approaches 10kc.

i guess an all pass filter wouldn't be the solution in this case. it might fix the phase in some ranges but it would cause more problems in other ranges. are there any other techniques for phase correction that might be suitable here?

also, i just noticed something strange in the waveform coming out of the max294. there are tiny 'steps' in the wave. the output from the earlier filter stages do not have these steps. i am attaching a zoomed in photo from my scope.

what is going on and is this a significant problem?

[W4NEQ]

The steps are the remnants of the switching function and are normal.  I'm not familiar with that chip, but typically they are on the order of 100 x the highest audio frequency so generally won't pass through subsequent audio stages.  If you don't like it, then add a simple RC LP to the output.

[ssbothwell KJ6RSG]

oh i see. i added a 25kc low pass and it cleaned up those steps perfectly. this project is working out real nice.

i think the next step is to build a clipping detection circuit of some kind. i've seen some fairly simple designs online. does this circuit look like a good starting point? http://www.circuitstoday.com/audio-clipping-indicator-circuit

[W4NEQ]

The traditional approach in analog audio is to build in a certain amount of "headroom" throughout your audio path to prevent clipping.  This headroom figure ranges from say, 10 to 20 dB or more if you can afford it, depending upon the variability of the audio level.  With unprocessed voice, when actively monitored with a VU meter, 16 to 20 dB works well.  If you choose the more conservative 20 dB figure, your audio circuits will clip at 10 times the normal peak voltage level.  With reasonable care monitoring the VU meter, 16 dB works fine for ham stuff.

The amount headroom is usually limited by your total dynamic range, traded against noise floor; if your SCF can deliver 70 dB of dynamic range, which I think might be typical, and you allocate 20 dB to headroom, your normal signal to noise ratio would be 50 dB.

You decide this when setting your normal operating point and calibrating your VU meter.  With opamps, the power supply voltage(s) usually determine your clip point.

Chris

[ssbothwell KJ6RSG]

hi everyone.

i ran into another weird roadblock. i etched a circuit board and assembled the device in an enclosure.

the circuit as i designed it goes like this: buffer amp -> 50cycle highpass filter -> max293 uncommitted opamp butterworth LPF -> max 293 switched capacitor filter

i dont know if the buffer amp was necessary but i only have dual op-amps (tl082) so i figured i might as well throw it in there.

it works or does not work at random intervals but usually it is does not work. turning off and on, or adjusting the amplitude on the input signal sometimes gets it working but is it very unpredictable. i dont think the problem is with the MAX293 because the brickwall cutoff works fine.

when it is not functioning correctly it clips the waveform as you can see in the attached photo. as it approaches the LPF cutoff point the waveform looks correct but is attenuated. the max293 still seems to operate correctly and the massively attenuate past the cutoff point.

when it is operating correctly there is no distortion and very little attenuation up until the max293 cutoff point.

it seems like the problem is probably with the TL082 highpass and buffer but i have no idea. i am using a batch of op-amps i bought on ebay for really cheap. i've tried a couple of them and they all seem to perform the same.

[ssbothwell KJ6RSG]

i just discovered something. i was checking various voltages with my DMM and noticed that when i short the buffer amp's non-inverting input to ground temporarily then the circuit operates correctly. if i turn off and on the input signal generator then the problem returns until i once again short the non-inverting input. this is extremely consistent.

what is going on here?

[K7MCG]

Use a good scope to verify that your op amps are not oscillating....far above audio.
Did you remember to bypass the op amp power leads to ground, as close to the op amp as possible, using capacitors that do not have an internal resonance frequency within the opamp's gain bandwidth product ?

(I learned these lessons the hard way, long ago, with Fairchild 709 op amps.)
73,
Chuck K7MCG

[ssbothwell KJ6RSG]

i am using 10uF and 0.1uF capacitors across V+ and ground. they are not physically as close as possible to the op-amp. however, i tried holding various sized capacitors directly across the V+ and ground pins of the op-amp IC but it did not change anything. shorting V+ to the inverting input also seems to correct the problem.

i have a pretty nice scope (HP54600B) and a 10x probe.  i'm not noticing any oscillation from the op-amp.

also, i have no idea if this is a problem but my circuit does not have a voltage divider biasing the buffer amp input. i dont know if buffer amps are supposed to have bias when used with single ended power supplies.

[ssbothwell KJ6RSG]

i just read this document: http://www.kennethkuhn.com/students/ee431/text/op_amp_practical_applications.pdf

i tried a couple of the suggestions and putting 100k resistance in series with the inverting input seemed to solve the problem.

i guess that means there was internal parasitic capacitive feedback. the article mentions this is usually a problem when the inverting op-amp input is far away from the source. the trace on my circuit board is only about 1cm from the op-amp pin to the rca jack and i was using a 6inch rca cable to connect to my signal generator.

oh well, at least this solved the problem.

[W1DAN]

Hi:

I have been away from this thread for a while. Your circuit looks very nice and etching a board is a step ahead of me!

Sometimes op-amps lock up if the input pins do not have a reference to ground or a supply (i.e. they are truly floating and you have high gain). If I build an op-amp circuit with balanced plus and minus rails, then a resistor from the positive or negative inputs to ground usually helps to center the inputs to ground.

On the SCF I built (Thanks to Stu-AB2EZ for the idea), I put it after my audio compression and limiting, and gave a few db of headroom as checked by a scope. This availed the most signal to noise. I skipped the all-pass filter to correct for the phase shift and just used a diode to clip the negative-going spikes that result from the phase shift. As you see, this is not a perfect form of an elliptical LPF, but sure is easier that designing and building multi-stage FDNR's. For transmitting work, I feel the elliptical filter is better than the gentler forms. It provides more energy up to the cutoff, at the expense of very fast phase shifts near the corner frequency.

Thanks for posting your progress!

73
Dan