Balanced High Pass Filter

[earthsled]

Lack of info there. What other changes have you done? You must have changed the caps...

No, the caps are consistent. The circuit here is the same as HPF Test Setup v3 which uses 1.5uF caps. I made both these graphs using the 10k:10k in parallel for L1 (about 6.7H). The only thing that has changed is the sweep generator's output level.

The orange shows the frequency response with an input signal of -20dBu (-33dBFS)
The purple shows the frequency response with an input signal of -40dBu (-53dBFS)

(same graph attached)

 

[earthsled]

Here is the same test as above using the shielded 7k:10k in proper series for L1 (about 12.3H). The caps are 1.5uF in this test too.

The blue shows the frequency response with an input signal of -20dBu (-33dBFS)
The green shows the frequency response with an input signal of -40dBu (-53dBFS)

Which level is more accurate for testing?

Thanks!

 

[abbey road d enfer]

Here is the same test as above using the shielded 7k:10k in proper series for L1 (about 12.3H). The caps are 1.5uF in this test too.

The blue shows the frequency response with an input signal of -20dBu (-33dBFS)
The green shows the frequency response with an input signal of -40dBu (-53dBFS)

Which level is more accurate for testing?

Thanks!

I think the differences are due to two factors:
The inductance decreases with level - that accounts for the fact that the transition zone is less sharp - and the noise contribution, which tends to artificially increase the lower levels, making the slope apparently less steep.
Ideally, this measurement should be done with a sweepable filter, in order to eliminate noise.
In practice, whichever of these measurements is good enough.

 

[earthsled]

Ideally, this measurement should be done with a sweepable filter, in order to eliminate noise.

I was assuming the software I've been using did the filtering in the digital domain, but perhaps not. -- I'll send this question to the developer.

In practice, whichever of these measurements is good enough.

Okay. I initially was using -20dBu to represent a "hot" mic. Would -40dBu be a good estimate for a "nominal" output, or is this too low to be realistic?

----------------------------------------------------------------------------------------------------

The 10k:10k transformer I'm using for L1 measures a bit higher than the ideal value, but it's probably as close as I'll find using off-the-shelf cheap-o models. I find that increasing the capacitance values to around 2uF helps to compensate and brings the cutoff frequency down to the neighborhood of 90 to 95Hz.

Another thing I need work out is the loading of my preamp's input transformer. Attached is a schematic showing the HPF feeding two possible input transformers, each followed by a DI input. The Cinemag model uses a Zobel on the secondary, while the Lundahl does not. For simplicity, the 1M was kept in the circuit to serve as loading for the DI, but now I'm concerned this fixed 1M load will alter the filter's response.

From what I understand, a load of 100k on the secondary would reflect a more ideal 2k at the primary. Should I return to a setup that will switch the 1M into the circuit only when the DI is in use?

Thanks!

 

[abbey road d enfer]

I initially was using -20dBu to represent a "hot" mic. Would -40dBu be a good estimate for a "nominal" output, or is this too low to be realistic?

Both measurements are realistic. The one made at the lower level would be characteristic of most "normal" applications such as vocals, acoustic guitar, piano, and so on. The other, at -20 would be typical of drums and electric guitar.

Another thing I need work out is the loading of my preamp's input transformer. Attached is a schematic showing the HPF feeding two possible input transformers, each followed by a DI input. The Cinemag model uses a Zobel on the secondary, while the Lundahl does not. For simplicity, the 1M was kept in the circuit to serve as loading for the DI, but now I'm concerned this fixed 1M load will alter the filter's response.

Theory and simulation indicate that the unloaded version would have a significant hump at the turnover frequency. You need to experiment; you may well find that you need to add a resistor in parallels with the inductor to tame this resonance.

From what I understand, a load of 100k on the secondary would reflect a more ideal 2k at the primary. Should I return to a setup that will switch the 1M into the circuit only when the DI is in use?

Thanks!
[/quote]

 

[earthsled]

Theory and simulation indicate that the unloaded version would have a significant hump at the turnover frequency. You need to experiment; you may well find that you need to add a resistor in parallels with the inductor to tame this resonance.

I'm a little confused here. When you refer to the "unloaded" version, are you talking about the schematic without the Zobel? Also, are you suggesting it would be better to change the loading within the filter rather than at the secondary of the mic pre's input transformer? I was under the impression that 2k would be the ideal impedance to have at the primary of the input transformer. Is this not so?

 

[abbey road d enfer]

I'm a little confused here. When you refer to the "unloaded" version, are you talking about the schematic without the Zobel?

The one with the Lundahl, where the secondary load is 1Meg, that reflects as 20k at the primary.

Also, are you suggesting it would be better to change the loading within the filter rather than at the secondary of the mic pre's input transformer?

You'll have to experiment, but quite often, these xfmrs are happy to be lightly loaded at the secondary and microphones also are happy with a lighter load. In terms of LF response, there should be no great difference between 2k at the primary or 100k at the secondary, but when the filter is out, there is a difference of response from the microphone between loaded or unloaded. Before solid-state, many mic manufacturers specified a mic pre based on an unloaded transformer. That applies to dynamic or ribbon mics, and somewhat but to a lesser degree to transformer-balanced microphones. The inductive nature of these mics likes to see an inductive load.
By not loading the sec, you allow the possibility of giving an inductive impedance when the filter is out. When the filter is in, you need to load the mic. That is why, as has been mentioned earlier, the design of a passive filter right at the input involves many aspects and compromises.

 

[earthsled]

The one with the Lundahl, where the secondary load is 1Meg, that reflects as 20k at the primary.

Okay, this is clearer now. Is the version with the Cinemag less of a concern because it has a similar reflected impedance compared to my most recent experiments with the HPF?

You'll have to experiment, but quite often, these xfmrs are happy to be lightly loaded at the secondary and microphones also are happy with a lighter load.

What is the best way to experiment with these variables? Can I continue to breadboard components and run sweeps, or will this require building out the preamp and testing with different mics?

Is there any harm in maintaining the same load / Zobel for both transformers? What about simply loading the secondary of the Lundahl with a 100k resistor?

That is why, as has been mentioned earlier, the design of a passive filter right at the input involves many aspects and compromises.

...and a certain degree of pulling-out-one's-own-hair!

 

[abbey road d enfer]

The one with the Lundahl, where the secondary load is 1Meg, that reflects as 20k at the primary.

Okay, this is clearer now. Is the version with the Cinemag less of a concern because it has a similar reflected impedance compared to my most recent experiments with the HPF?

Yes.

You'll have to experiment, but quite often, these xfmrs are happy to be lightly loaded at the secondary and microphones also are happy with a lighter load.

What is the best way to experiment with these variables? Can I continue to breadboard components and run sweeps, or will this require building out the preamp and testing with different mics?

You'll need to assess this aurally. The variations in frequency response are tiny but the subjective effect can be significant. It depends very much on the actual microphone.

Is there any harm in maintaining the same load / Zobel for both transformers? What about simply loading the secondary of the Lundahl with a 100k resistor?

That is quite possible for the load resistor, because it defines the input impedance. The Zobel must be tailored to the actual xfmr.

 

[earthsled]

Is there any harm in maintaining the same load / Zobel for both transformers? What about simply loading the secondary of the Lundahl with a 100k resistor?

That is quite possible for the load resistor, because it defines the input impedance. The Zobel must be tailored to the actual xfmr.

I've modified the circuit using the Lundahl transformer to include a 100k load resistor (attached). I'm almost afraid to ask, but are there any caveats to doing this?

-------

When I explained all of the headaches involved with designing a HPF to go between a mic and a preamp to a friend, he suggested to have the filter at the output of the preamp rather than the input. The more I think about this, the more logical it seems. The output impedance of my preamp circuit is known at 200 ohms, and the input of the next stage will either be high impedance (about 10k for modern line-level gear) or medium impedance (about 600 ohms for vintage gear).

Are we correct in thinking that having the HPF post-preamp would help to eliminate some of the impedance variables?

Thanks!

 

[abbey road d enfer]

I've modified the circuit using the Lundahl transformer to include a 100k load resistor (attached). I'm almost afraid to ask, but are there any caveats to doing this?

As I said earlier, there is no problem, but you loose the possibility to "unload" the mic; in some cases, unloading the mic gives somewhat better results.

Are we correct in thinking that having the HPF post-preamp would help to eliminate some of the impedance variables?

I think I already said earlier in this thread that using an active filter gives a perfectly predictible performance and allows steeper slopes that can hardly be achieved in passive.
Advocates of the filter right at the input claim that it avoids VLF saturating the xfmr and clipping the electronics; unfortunately, it is not always that simple, because the increased source impedance induces THD problems.
For example, your HPF presents a source impedance at least 10 times higher than nominal between 50Hz and 200Hz, with a peak of 100 times at the corner frequency. This is a cause of significant THD (variable with level and nature of magnetic core).
Contrary, an active filter, located after the 1st preamp stage behaves very linear over a wide frequency and level range. But if very high levels of VLF are present, it may saturate the core and clip the active stage. In practice, this not a real concern, except for special applications like recording in industrial environments, but I don't think that's your main interest.

 

[earthsled]

Out of curiosity, I reverse-engineered the roll-off filter from a Sennheiser MD421 (schematic attached). The inductance values are approximate - measured with my LCR meter. I expect there are compromises with a passive circuit such as this, but I thought it would be good to bring to this discussion for what it's worth.

Cheers!

 

[abbey road d enfer]

Out of curiosity, I reverse-engineered the roll-off filter from a Sennheiser MD421 (schematic attached). The inductance values are approximate - measured with my LCR meter. I expect there are compromises with a passive circuit such as this, but I thought it would be good to bring to this discussion for what it's worth.

Cheers!

These work well because they are shunt-filters, so the impedance the mic pre sees is low, which is good for the input xfmr behaviour. At higher frequencies, the mic is not unduly loaded. Care must be taken for the inductors not to saturate, though.

 

[earthsled]

Care must be taken for the inductors not to saturate, though.

What parameters should I keep in mind to avoid the saturation point?

 

[abbey road d enfer]

Care must be taken for the inductors not to saturate, though.

What parameters should I keep in mind to avoid the saturation point?

Exactly the same as a mic input xfmr, which saturates above a certain level of induction. The specs could be relaxed:
in a mic input xfmr, the inductance must be quite high, in order not to swamp the mic's impedance, but here, you want this to happen - let's say you want the HPF at 100Hz, for a nominal 150r microphone, the inductance computes at 240mH. For a mic input primary, you want minimum 2-5H.
Also, for a mic input xfmr, the expected max level is about 0dBu, witha mic, it is safe to assume the max level would be about -20dBu.
A ferrite-core would probably be right for this application, but one has to sort out the many different variants of ferrite materials and cores.

You could use air-core inductors (they don't saturate) but they would be quite bulky and you would probably need to make a humbucking arrangement.

 

[earthsled]

let's say you want the HPF at 100Hz, for a nominal 150r microphone, the inductance computes at 240mH

I'd like to know more about the calculations. Are you using the equation L = R / ( 2*∏*ƒ ) for your estimate?

R1 through R4 in the filter would be in parallel with the nominal 150r mic, so the overall resistance would be reduced with these fixed values, correct?

A ferrite-core would probably be right for this application, but one has to sort out the many different variants of ferrite materials and cores.

There seem to be radial lead "RF chokes" available at online distributers. These have ferrite cores, but how does one differentiate between materials?


Thanks very much for your help!

 

[abbey road d enfer]

let's say you want the HPF at 100Hz, for a nominal 150r microphone, the inductance computes at 240mH

I'd like to know more about the calculations. Are you using the equation L = R / ( 2*∏*ƒ ) for your estimate? 

Yes. 

R1 through R4 in the filter would be in parallel with the nominal 150r mic, so the overall resistance would be reduced with these fixed values, correct?

They are in series and define the amplitude and slope of attenuation.

A ferrite-core would probably be right for this application, but one has to sort out the many different variants of ferrite materials and cores.

There seem to be radial lead "RF chokes" available at online distributers. These have ferrite cores, but how does one differentiate between materials? 

None of these RF chokes is suitable for this application. Their value is from a few uH to about 10mH. You need something like 100-500mH here.
Check this
http://www.audiomaintenance.com/downloads/carnhill_design_guide.pdf
VTB9044 or 54 seem to be about right

 

[earthsled]

They are in series and define the amplitude and slope of attenuation.

Okay. Let me try to calculate the cutoff frequency of the "S" position of the MD421's filter...

L = 45.6 mH
R = 150r[mic] + 47.5r[R1]

ƒ = (150+47.5) / ( 2*∏*0.0456)

ƒ = 689 Hz

The frequency seems about right.

None of these RF chokes is suitable for this application. Their value is from a few uH to about 10mH.

I believe the MD421 has the cutoff frequency of 500Hz with a very gradual slope. Looks like less than 6dB per octave to me (graph attached). Maybe this is not the most desirable response, but it seems to do the job of LF reduction okay.

Considering the inductance values found in the original filter, I was thinking standard values of 47mH and 82mH would be close enough for substitutions.

Mouser seems to have RF chokes in stock that extend up to the 100 to 150mH range. Here's my search for an 82mH coil: http://www.mouser.com/Passive-Components/Inductors/Fixed-Inductors/_/N-wpczZscv7?P=1z0wpweZ1z0z7l5&FS=True

Are these chokes suitable for the application?

Don't get me wrong, Carnhill iron is very nice, but with exchange rates and international shipping, it gets pricey for me.

Thanks!

 

[bruce0]

I got several lovely little shielded inductors from Nebraska Surplus, I can send some pic's .  2 in particular one is 110mH looks mumetal shielded, the other is 47mH encasulated, both are smallish, might work in a mic but maybe not a pencil mic, they would fill the tube.

 

[abbey road d enfer]

Mouser seems to have RF chokes in stock that extend up to the 100 to 150mH range. Here's my search for an 82mH coil: http://www.mouser.com/Passive-Components/Inductors/Fixed-Inductors/_/N-wpczZscv7?P=1z0wpweZ1z0z7l5&FS=True

Are these chokes suitable for the application?

Some have too high DCR; in your application, you're looking at DCR >40 ohms.

EDIT: you're looking at DCR <40 ohms