Balanced High Pass Filter

[MicDaddy]

Would a first order low pass, using a single inductor to shunt the lows into ground, be a viable option?  Forgive me if this is obvious, the conversation is quite a bit above my head already.

 

[earthsled]

Would a first order low pass, using a single inductor to shunt the lows into ground, be a viable option?  Forgive me if this is obvious, the conversation is quite a bit above my head already.

Yes. This should give you a single-pole filter with -6dB per octave.

This is Butterworth, an excellent filter when you have no spectific goal.

Has anyone used a Butterworth filter calculator like this one : http://www.pronine.ca/buthf2.htm ? I'm attempting to find the ideal values for a 2-component HPF with a cutoff of 100Hz. I'm confused about the "characteristic impedance" value. If the filter is to be placed between a 200R source and a 2K load, what should the characteristic impedance of the filter be?

Thanks!

 

[PRR]

> a first order low pass, using a single inductor to shunt the lows into ground

If you are shunting lows to ground, that's a high-pass.

To specify a 1-pole filter you also need an R some place.

However, why a coil? In audio, caps are generally much cheaper.

An R-C low-pass is a series resistor and a shunt capacitor. Assuming your source has low impedance and can drive 1K, and your load is much over 1K: add 1K in series and then a cap to ground. I happen to recall that against 1K, a 0.01uFd will cut above 17KHz, so 0.1u gives 1.7KHz and 1U gives 170Hz, so you want 1.7uFd. (Or higher impedance so you don't need such large caps.)

> I'm confused about the "characteristic impedance" value.

In LOOONG-line work the source and load impedances are made equal. 50 or 75 ohms for coax, 100-600 for twinlead.

> to be placed between a 200R source and a 2K load

Obviously power is not precious (as it can be in radio work). To use that calc, add 1,800 ohms to the source and proceed as a 2K characteristic impedance circuit.

It defaults to RF-size units, you will want to change to uFd and H.

For 100Hz I get numbers near 1uFd and several H. The several-H parts will look like output transformers and you are unlikely to find these values in the choke-store. If there is DC power anywhere near, an active filter will be cheaper and less work.

 

[abbey road d enfer]

The calculator there is for matching source and load impedance; doesn't apply to your case. You should look for a calculator for loudspeaker crossovers, which is based on zero source impedance, or use the formula Lc=Z/(2pi.F) and Cc=1/(2pi.Z) and for Butterworth
Cb=1.414Cc and Lb=Lc/1.414
The actual frequency response with non-zero source impedance will be somewhat overdamped, but in a much less significant way than any tolerance or standardized value .

 

[earthsled]

Thanks for your help! Surprisingly, I'm getting different results from each of the suggested methods using a cutoff frequency of 100Hz with a load (or characteristic) impedance of 2K ohms.

Results from this calculator: http://www.pronine.ca/buthf2.htm
C = 0.56uF
L = 2.25H

Results from this calculator: http://www.apicsllc.com/apics/Misc/filter2.html#second
C = 0.56uF
L = 4.5H

Results from Abbey's equations:
C = 112uF  ???
L = 2.25H

So, for some reason, using the calculator designed for speaker crossovers yields an inductance value that is double compared to the rest. I'm thinking there might be a frequency variable missing from Abbey's capacitance equation. Perhaps it should be Cc=1/(2pi•Z•F) ? This gives me a capacitance value that is 1.125 uF (which is double compared to the other results).

 

[abbey road d enfer]

I'm thinking there might be a frequency variable missing from Abbey's capacitance equation. Perhaps it should be Cc=1/(2pi•Z•F) ? This gives me a capacitance value that is 1.125 uF (which is double compared to the other results).

Sure, I managed to beat a world record by including 3 typos in one sentence... Sorry about this.
Cc=1/(2pi.F.Z)=>0.8uF
Lc=Z/(2pi.F)=>3.2H
Cb=0.707.Cc=>0.56uF
Lc=1.414Lc=>4.5H
Which is consistent with the loudspeaker crossover calc.
The "pronine" calc is relevant to matched impedances, i.e. the source impedance is equal to the load impedance, which induces a 6dB loss in the pass band. This is valid only for RF circuits, where a filter is inserted in an antenna  or interstage connection.

 

[earthsled]

Very interesting.

The Butterworth filter requires a smaller capacitance and a larger inductance compared to the Shure schematic. Of course, 4.5H makes for an impossibly large component (as PRR suggested).

If I were to begin with the Shure values and gradually increase the inductance while decreasing the capacitance, would the result be a gradually flatter pass band (up to the point of the maximally-flat Butterworth)?

Also -- for a microphone-level filter, would a 75mW transformer (used as an inductor) be sufficient, or would the 200mW model be a better choice? I ask because Mouser has some shielded models with the 75mW rating.

One more thing -- if the filter were to be placed after a phantom power circuit that offered some protection from DC (like this one-- http://www.groupdiy.com/index.php?action=dlattach;topic=46785.0;attach=10774;image) would tantalum capacitors be okay, or would there still be some risk of reverse voltages?

Thanks!

 

[abbey road d enfer]

The Butterworth filter requires a smaller capacitance and a larger inductance compared to the Shure schematic. Of course, 4.5H makes for an impossibly large component (as PRR suggested).

What do you mean "impossibly large"? A typical mic xfmr has a primary inductance of about 5H. Look at the size of Beyer or Neurik xfmrs.

If I were to begin with the Shure values and gradually increase the inductance while decreasing the capacitance, would the result be a gradually flatter pass band (up to the point of the maximally-flat Butterworth)?

Yes.

Also -- for a microphone-level filter, would a 75mW transformer (used as an inductor) be sufficient, or would the 200mW model be a better choice? I ask because Mouser has some shielded models with the 75mW rating.

-20dBu (which is a very hot level for a mic)=> 77mV=> 30uW into 200 ohms

One more thing -- if the filter were to be placed after a phantom power circuit that offered some protection from DC, would tantalum capacitors be okay, or would there still be some risk of reverse voltages?

There is still an enormous risk of reverse voltage, if the phantom voltage is turned off, and you connect a 48V-powered signal to the input. I would never trust tantalum caps in this position.

 

[earthsled]

Recently, I ran some sweeps using the ideal C and L values calculated (above) for a Butterworth high-pass filter at 100Hz.

Interestingly, the response curve changes dramatically when the filter is used ahead of a mic pre input transformer.
The red graph shows the response of the filter alone.
The green graph shows the response of the filter with an input transformer.

I imagine it's the inductance of the input transformer in parallel with the filter that causes this shift -- is this correct?

 

[abbey road d enfer]

I imagine it's the inductance of the input transformer in parallel with the filter that causes this shift -- is this correct?

The shift in turnover frequency could be explained by a transformer with extremely low inductance, which wouldn't be correct for any quality audio performance, but the asymptotic response at -25dB below is not normal; probably noise. I think someting's wrong in your measurement set-up.

 

[earthsled]

Thanks for your reply.

Attached is a schematic of my testing rig. The components are mounted on a solderless breadboard, connected with twisted pairs and shielded cable where possible. The value of L1 ranges between 3H and 7H, but this doesn't seem to make a drastic difference to the frequency response.

Do I need to adjust L1 so that it equals 4.5H when in parallel with XFMR1?

 

[abbey road d enfer]

Thanks for your reply.

Attached is a schematic of my testing rig. The components are mounted on a solderless breadboard, connected with twisted pairs and shielded cable where possible. The value of L1 ranges between 3H and 7H, but this doesn't seem to make a drastic difference to the frequency response.

Do I need to adjust L1 so that it equals 4.5H when in parallel with XFMR1?

That would make the alignment more exact, but it's not so important.
Are you really running the test at -20dBu? The transformer core may tend to saturate at low frequencies, which could explain the weird frequency shift.

 

[earthsled]

Are you really running the test at -20dBu?

I think so. I'm using a DMM with true RMS to measure the signal output. The highest voltage seems to be near 77mV (about 30 to 40mV when measuring the voltage of pin 2 referenced to pin 1). This output voltage corresponds to about -33dBFS in the software.

My main concerns are that neither of the plots appear to be "maximally flat" like the textbook examples Butterworth filters and the cutoff frequency of the "transformer" plot seems to be much higher than the intended 100Hz.

Perhaps my equipment is just not capable of accurate results?

Thanks!

 

[abbey road d enfer]

My main concerns are that neither of the plots appear to be "maximally flat" like the textbook examples Butterworth filters and the cutoff frequency of the "transformer" plot seems to be much higher than the intended 100Hz. The slightly underdamped response of the transformerless circuit is quite normal. With these values, the optimal load is about 3k. The actual load is much higher, so the circuit tends to be resonant.
As to the circuit with transformer, I continue to think that the deviation is due to the fact that its actual inductance varies with the signal level.

Perhaps my equipment is just not capable of accurate results?

As it is, your test setup is perfectly capable of producing reasonably accurate measurements.

 

[earthsled]

Earlier, you mentioned noise might be the culprit so I put on my headphones and listened closely to the HPF circuit. I heard low-level hum when the HPF was in the audio path. Moving the circuit around, it sounded like the coil was acting like a guitar pickup -- transducing lots of stray magnetic fields from around my desktop. Would induced hum cause this irregular frequency response?

 

[abbey road d enfer]

Earlier, you mentioned noise might be the culprit so I put on my headphones and listened closely to the HPF circuit. I heard low-level hum when the HPF was in the audio path. Moving the circuit around, it sounded like the coil was acting like a guitar pickup -- transducing lots of stray magnetic fields from around my desktop. Would induced hum cause this irregular frequency response?

As I mentioned earlier, it may explain why the response doesn't go down to minus infinity at VLF, but certainly not why the turnover frequency is skewed.
Do you have the possibility to visualise the wave form at the xfmr's output? There's several free apps that do that. I recommend Visual Analyser.

 

[earthsled]

Do you have the possibility to visualise the wave form at the xfmr's output?

Yes. Well, sort of... I hooked up my test circuit to an oscilloscope and used a sine wave generator for the input. I increased the level of the generator to the maximum output (about 0.6V p-p on the scope). With the filter, I can see the amplitude decreasing when the frequency is lowered, but I'm not seeing any obvious clipping or flattening of the waveform.

Maybe this is not the best setup for identifying the issue?

There's several free apps that do that. I recommend Visual Analyser.

Visual Analyser looks nice. Is it the waterfall-type view that would be helpful here? Unfortunately, I'm on a Mac, so the software choices are more limited. The app I'm using for frequency response graphs is capable of waterfall plots, but not in the demo mode.

Thanks for your help!

 

[abbey road d enfer]

Do you have the possibility to visualise the wave form at the xfmr's output?

Yes. Well, sort of... I hooked up my test circuit to an oscilloscope and used a sine wave generator for the input. I increased the level of the generator to the maximum output (about 0.6V p-p on the scope). With the filter, I can see the amplitude decreasing when the frequency is lowered, but I'm not seeing any obvious clipping or flattening of the waveform.

Maybe this is not the best setup for identifying the issue?

There's several free apps that do that. I recommend Visual Analyser.

Since you have an o'scope, you don't need much more. I suggest you use the best audio analyser available: your ears. If you listen to the signal, you can tell if there is hum/buzz/hiss or not.

Visual Analyser looks nice. Is it the waterfall-type view that would be helpful here? Unfortunately, I'm on a Mac, so the software choices are more limited. The app I'm using for frequency response graphs is capable of waterfall plots, but not in the demo mode.

Waterfall is good for analyzing resonances and transients. I was just suggesting using your PC (well...Mac) as an o'scope.

 

[earthsled]

Okay. Thanks for clearing that up!

Using an SM57 and the HPF into a mixer and headphones... I do notice an obvious increase in LF noise / hum when I switch the filter on. This noise / hum increases when the circuit is brought near almost anything electronic. The level of noise is such that I question if the circuit is even practical for use in recording.

Maybe I should explore a HPF that doesn't use an inductor?

 

[abbey road d enfer]

Okay. Thanks for clearing that up!

Using an SM57 and the HPF into a mixer and headphones... I do notice an obvious increase in LF noise / hum when I switch the filter on. This noise / hum increases when the circuit is brought near almost anything electronic. The level of noise is such that I question if the circuit is even practical for use in recording.

Maybe I should explore a HPF that doesn't use an inductor?

That's probably because the inductor is not shielded. You can make a shield out of mild steel (cheap) or Mumetal (expensive), or you can use two inductors of half-value arranged in a humbucking configuration - that's probably the most efficient solution.
Now you understand why inductors became out of fashion and were replaced with active RC circuits!