Behringer Autoquad - modding HP filter slope

[emrr]

I assume the calculator assumes a zero ohm source impedance.  As actual impedance moves up, so does frequency.  I believe that to be the difference in any accurate calculation versus reality.   

 

[Gertius]

Ok, thanks for your explanation emrr!
I guess I will just aim for those values then when trying to calculate the 3rd order filter as well.

 

[Gertius]

Hey guys,
in case you are interested in following this:

I have just ordered 2 triple pots 47k neg log from the link I have posted above.
(which cost more than the whole box :-D)

With the filter calculator I have determined that I´ll use the existing 909 Ohms resistors (with a third one added) and probably 3x470nF as caps. The calculator gives me a range between 7Hz and 370Hz, which might then be around 15Hz to 750Hz in reality (see emrr´s reply above). I have taken a good look at the existing components at the circuit board, and have high hopes that the triple pot will fit in nicely. At least there is a lot of room around the existing dual pot (see pic attached).
If the pot fits, it should be a relatively easy modification process.

I´ll keep you posted.

 

[Gertius]

The pots arrived yesterday.
2x 47kOhm triple gang neglog.

They are a bit bigger than the old ones (20mm instead of 16mm) and the spindle is also a little bit bigger, but from the looks of it it might just fit. Anyone know where I could get knobs for the spindle of 1/4 inch diameter?

Waiting for the rest of the parts (should arrive tomorrow) and then will start modding.

Cheers!

 

[Gertius]

Ok so yesterday I received the other parts for the mod and started modifying two of the Autoquad´s channels.
It went quite well and was not too much work. See first pic for "after the mod".

But then I measured the new channels vs. the old and it looks a little strange to me. The curves are just a tiny bit steeper than with second order, and the difference doesn´t even kick in until 15dB down or so (depending on fc).
At the very bottom of the curve it does seem to be -18dB/oct though.
I have attached the measurement in two points, 40Hz and 1000Hz, old vs. modded.
Component values on modded channel were 3x 1,8kOhm resistors and 3x 220nF caps.

What do you think? Does it look right to you?

P.S.: the new 20mm pots just barely fit into the device. It is functional, but not pretty. If I were to mod more channels I would look around for 16mm diameter pots. The circuit board on the top just barely fit over the two pots, so that the LED´s do not reach to the front panel holes anymore

 

[Gertius]

And the measurement

 

[emrr]

I suppose the definition of 18dB/oct is what's in question.  I imagine it may be relative to eventual slope, well after the knee.  Here the knee is very shallow, we aren't all that far down 2 octaves below 1K; looks closer to 6dB/oct from defined 3dB point of 1K down to 250.  I had a look at my Intelligate filters, and they look very similar to your 'before' image for both 40 and 1K, with 40 being sharper due to other overall HPF elements of the unit.  In practice is the mod worth it, given the knee shape? 

 

[Gertius]

Hi emrr,
sorry I haven´t checked back here for a while.
Thanks for your reply and assessment of the results!

Yeah, it looks like you are right. It seems to be close to -18dB very far down.

Regarding your question about in practice, I must honestly say that after seeing the measurements results I kind of lost interest in doing intense listening comparisons. I have quickly listened to the new filters on their own and they seemed very close to the "before" version. Right now I have so many other priorities, but if I´ll do a listening comparison still I´ll post back here.

For now at least I have decided not to mod all my other channels. It was a really interesting project though!

Cheers,
Christian

 

[abbey road d enfer]

But as far as I can see it doesn´t match up with the schematics.
I enter the C´s with 150n and 68n and the R´s in one case with 100909 Ohm (result fc = 15.6 Hz) and in the other case with 909 Ohm (result fc = 1733.6 Hz). In my mind the result should say 30Hz and 3000Hz. Can anyone please tell me what I´m doing wrong?

LTSpice sim says 27Hz and 3k. Looks like there's something wrong with the Japanese abacus.

 

[emrr]

I had taken some measurements of an Intelligate filter set, and just today I built up a Sallen-Key /Butterworth filter based on a unity-gain follower, with different values for R1/R2 as is needed.  When I pulled up the Intelligate measurements on top of the new filter measurements, I was struck by how lazy the slope was on the Intelligate, which looked like your Autoquad readings.  I think given how they have implemented the filter, with such a long knee, what you are seeing is correct regarding slope change after mod.  It's jut not apparent until way down the slope. 

 

[abbey road d enfer]

In order to achieve proper Butterworth response, the values need to be altered considerably or the voltage-follower modified to provide gain.
For the LPF with equal value resistors (pots), the values of the caps must be approximately 1, 4 and 0.2.
For example, with 100k pots, the LPF lowest frequency set at 20Hz, the first cap is 80n, the second (the one that goes from junction of R2 and R3 to the opamp's output) is 320n and the third one (from opamp's non-inverting input) to ground is 16n.
The problem is that the HPF with equal value pots does not lend itself to Butterworth response. The only possibility to get a satisfying response is to run the opamp with about 6-7dB gain. Of course that makes the circuit prone to clipping and makes necessary the adjunction of a suitable attenuator to compensate the extra gain.

 

[Gertius]

Hi guys,
haven´t checked back here for a while...
Interesting findings, emrr and abbey road.

Does that make the initial Behringer 2nd order filter an inferior design? Is it worth modding that to get closer to Butterworth response?
Would you care to post the filter response comparison from the Butterworth to the Intelligate, emrr?

It seems, the excerpt from the SSL schematic that I posted on page 1 of this thread uses 3x47K pots and at least 2x220nF caps.
Of course, it is different in that there is another opamp involved. Would be interesting to see that filter response in comparison, too...

Cheers!
Christian

 

[abbey road d enfer]

Hi guys,
haven´t checked back here for a while...
Interesting findings, emrr and abbey road.

Does that make the initial Behringer 2nd order filter an inferior design? Is it worth modding that to get closer to Butterworth response?
Would you care to post the filter response comparison from the Butterworth to the Intelligate, emrr?

It seems, the excerpt from the SSL schematic that I posted on page 1 of this thread uses 3x47K pots and at least 2x220nF caps.
Of course, it is different in that there is another opamp involved. Would be interesting to see that filter response in comparison, too...

Cheers!
Christian

Look at the 2nd schemo (Harrison): the opamp runs with a gain of about 3dB. that gives the correct Q for te filter. They had to compensate at the input; this inverter stage runs with about -3dB gain (3db attenuation).
Look at the SSL schemo, how T12 runs with about 3dB gain also (even adjustable via RV6) and how the signal is attenuated right at the input by the voltage divider RV5/R56.

Now, is the equal value/no gain HPF an inferior design? Maybe yes, on a purely technical point of view, the slope is not as nice as the SSL/Harrison, Maybe no, since it has a slightly better transient response, which some claim to actually hear. Debate is open...

 

[Gertius]

Now this sparked my interest again and I would like to continue tampering around with the HPF.
As far as I understand, in the Autoquad, the Opamp stage following the HPF is configured as a unity gain buffer (see schemo under Behringer in files section).
Is it as simple as changing the config to a non-inverting amp with amplification around 3dB and then adding a passive voltage divider at the input with -3dB to get closer to Butterworth response?

I´m quite happy about the Autoquads LPF, so I don´t really want to mod that.

 

[abbey road d enfer]

As far as I understand, in the Autoquad, the Opamp stage following the HPF is configured as a unity gain buffer (see schemo under Behringer in files section).
Is it as simple as changing the config to a non-inverting amp with amplification around 3dB and then adding a passive voltage divider at the input with -3dB to get closer to Butterworth response?

Yes. The resistor values for the voltage divider must be low enough to provide a low source impedance to the filter even when the frequency pot is at max, and high enough not to overload the preceding stage.

 

[Gertius]

Ok, great! Thanks for the tips.

I sketched out the mod (based upon the 3rd order mod from page one of this thread), just to be sure.
Does this look right to you?
I based the values on the schematics from page one. Just to be on the clear side: the gain of the opamp seems to be around 4dB, the one from the voltage divider -4dB.

The resistor values for the voltage divider must be low enough to provide a low source impedance to the filter even when the frequency pot is at max, and high enough not to overload the preceding stage.

Does a resistor value in the "hundreds" look good for a start (like in the sketch)? Can I just experiment with that and will hear/measure distortion when it´s not quite right yet? Or can I break something here?

Oh, and one last question, can I mod the second order filter in the same way, for experimentation purposes?
I´d like to prepare one 2nd and one 3rd order Butterworth to compare...

 

[abbey road d enfer]

I sketched out the mod (based upon the 3rd order mod from page one of this thread), just to be sure.
Does this look right to you?

For a 3rd-order Butterworth you need to increase the gain to a little over 6dB. The optimization of the single-stage 3rd order Sallen & Key is quite difficult to solve mathematically. Using a simulator is almost mandatory.
I get acceptable results with a gain of 2.2 and the first cap at about 60-70 nF.

The resistor values for the voltage divider must be low enough to provide a low source impedance to the filter even when the frequency pot is at max, and high enough not to overload the preceding stage.

Does a resistor value in the "hundreds" look good for a start (like in the sketch)?

As it is, it presents a load of less than 500r to the output of the 1st opamp; it is too low. Now since you need a different attenuation (2.2) you need to change the values; unfortunately a 6dB pad is the worst case in that respect, presenting the lowest input impedance and the highest output impedance. I think you'll have to compromise, by using a somewhat higher resistance like 1.2k and 1k, which will present about 500r impedance to the filter ( the upper frequency range will be reduced somewhat and the slope less steep), or you may move the attenuator either between the HP and the LP section or at the output of the LP section.

Can I just experiment with that and will hear/measure distortion when it´s not quite right yet?

I recommend you do that anyway.

Or can I break something here?

No.

Oh, and one last question, can I mod the second order filter in the same way, for experimentation purposes?
I´d like to prepare one 2nd and one 3rd order Butterworth to compare...

It is somewhat easier, because you don't need gain, but you need to increase the Q by increasing C13 to 30n, decreasing C12 to 2.5nF and making the added cap 10n.

 

[Gertius]

Ok thanks!

For a 3rd-order Butterworth you need to increase the gain to a little over 6dB. The optimization of the single-stage 3rd order Sallen & Key is quite difficult to solve mathematically. Using a simulator is almost mandatory.
I get acceptable results with a gain of 2.2 and the first cap at about 60-70 nF.

Ok, I´ll rethink it then more into the directon of around 6dB gain and attenuation.
Do you have any tips for a good simulator? Maybe online? I would have access to MATLAB at a friend´s place. The Japanese site posted before seems a little funky, and the scale of the diagrams isn´t really too great.
Ah ok, so the caps need to be changed as well? I thought changing caps was mostly for the LPF? For modding the HPF knee I was planning equal value caps and pots, and just changing the opamp gain+attenuator... did I get that wrong, is that maybe not enough?

I think I´ll start by moving the attenuator then. Maybe then the values are less critical? Should I still start with 1k and 1.2k?

It is somewhat easier, because you don't need gain, but you need to increase the Q by increasing C13 to 30n, decreasing C12 to 2.5nF and making the added cap 10n.

Hm, could it be that there is a misunderstanding here? What I meant to ask was to modify the knee of an unmodded 2nd order HPF (as opposed to the 3rd order one that I attached the diagram of) to get closer to Butterworth response. C12+C13 seems to be in the LPF part of the circuit.

Anyway, thanks a lot already for your patience and all the answers! I´m very confident that I can do the mod once I´ve figured out what to do, but my electronics knowledge is not really advanced enough to design the mod... I think we´re almost there, though 

 

[emrr]

The Japanese site posted before seems a little funky, and the scale of the diagrams isn´t really too great.

The only thing I can add right now is that I've built many unity gain HPF using that site, and the response has been dead on with the calculator result.  It gives more data that any other place I've found online.  I personally have no simulator recommendations. 

Maybe later today I will be able to post the comparison pictures I'd promised. 

 

[abbey road d enfer]

Do you have any tips for a good simulator?

LTSpice.

The Japanese site posted before seems a little funky, and the scale of the diagrams isn´t really too great.

I agree; I've noted that the frequency box does not understand k or M - one has to type 20 000 for 20kHz. failure to do this results in an error message in japanese characters. But the worst is that the calculator does not cater for S&K structure with gain, only unity-gain.

Ah ok, so the caps need to be changed as well?

Yes

I thought changing caps was mostly for the LPF? For modding the HPF knee I was planning equal value caps and pots, and just changing the opamp gain+attenuator... did I get that wrong, is that maybe not enough?

Definitely not enough.

I think I´ll start by moving the attenuator then. Maybe then the values are less critical? Should I still start with 1k and 1.2k?

I would recommend you put it after the two filters; you may have to adjust the values in order to compensate the loading, but it would make the attenuation frequency-independant.

It is somewhat easier, because you don't need gain, but you need to increase the Q by increasing C13 to 30n, decreasing C12 to 2.5nF and making the added cap 10n.

Hm, could it be that there is a misunderstanding here?

Yes, some misunderstanding.

What I meant to ask was to modify the knee of an unmodded 2nd order HPF (as opposed to the 3rd order one that I attached the diagram of) to get closer to Butterworth response. C12+C13 seems to be in the LPF part of the circuit.

OK, so you want a 2nd-order S&K HPF with equal value resistors. then you have to make the gain about 4 db.