How to model a "complex" audio filtering device?

[mr_monster]

I hope this is the right forum for this type of question.

There is a device which is a kind of band-pass filter used with various musical instruments (analog) which I would like to copy. Just building an identical circuit is not good enough since the different components in the circuit have unique imperfections which contribute to the overall response of the device. Myself and other have built many prototypes and invested a lot of work but so far the results are not good enough. Since I am not much of an analog guy I thought to go for a more "brute force" approach - replicate the response of the entire device (looked at as a black box). The question is how to start, what is the method I should use?

I think the general idea is to see freq. and gain response to various input signals, come up with some formulas and / or graphs that describe various behaviors and replicate that using a DSP. This must be very methodic so I would appreciate any insight to the work process as required. I have access to scopes and a signal analyzer.

 

[FvM]

You should tell the name of the device, edaboard members might already know it.

The term filter usually means a 1. linear and 2. time-invariant system. Is this true for the device of interest?

 

[BradtheRad]

1.

The device that comes to mind which contains a bandpass filter, is a wah-wah pedal.
You cause its center frequency to sweep up and down the spectrum.
Is this what you are making?

To make it sound right I believe you need to apply automatic gain control to the signal, due to different instruments having weak areas in their frequency profile.

2.

Do you use digital sound processing software? A popular free program is Audacity.

You can examine waveforms in detail.
It has a spectrographic display.
It has a large assortment of effects which you can apply to a small time span, or the entire file.
The equalization curve allows you to fine-tune individual frequency bands.

 

[mr_monster]

The device is a wah pedal (from whatever reason I thought no one knows what it is), but now that you do it might be helpful!

I have a very "special" wah pedal. It is dated back to 1967 I believe and it has a truly remarkable sound, unlike any other wah pedal I have heard and trust me that I played with a lot of them. I have tried building a similar device using modern components and from reasons I am not entirely sure about they do not sound the same. The main suspects are: the transistors which are different from the same model transistors being made today (BC10x) and I could see that on a curve tracer as well. Not that I didn't try to find other devices that would behave the same but even after trying about 20 different new production devices I could not find a good replacement.

The second problematic component is the inductor which is highly not ideal and behaves very differently from other similar inductors. It seems like the inductor clips the signal at certain levels and that it's inductance is somewhat amplitude dependent and some other weird and non ideal effects I don't want to get into. I have been trying to get similar components but with no luck and even had two magnetic engineers from companies who make inductors have a go at it but until now with nothing to right back home about.

So here is the plan - model the response of the "special" wah I have and try to build a chain of filters and dynamic processors in software that will emulate the behavior of the device.
I understand that the 1st step would be to gather data on the device. This means the output for different freq. at the input, different amplitudes at the input and different positions of the pedal. If I want to check 100 freq. at 10 amplitudes at 100 positions of the pedal I need to chart 100,000 graphs which is a lot. Because this should be done in a methodic way and I want to know if I can cut back on the tests and still extract the data I need from them. So I guess my questions is what is the process I should be going for here?

 

[BradtheRad]

I would go with 10/10/10 right off the bat. Or, set the pedal at position #1, apply a sine sweep, examine response on an oscilloscope. Push the pedal to position #2, repeat the above steps. You'll quickly get an idea what the pedal is doing, and that will tell you what range of data you need and how many frequencies to apply, etc.

Rather than observe a number of static images, have you examined the dynamic frequency response in real time in a spectrographic display?

You would pipe its signal into a computer running a suitable program which displays a spectrogram real-time.

There is also the Windows Media Player visualization called 'Firestorm' (under category Bars and Waves). It displays a moving spectrogram in high resolution. You can make out individual frequencies. However it only plays recorded audio, not real time (I think).

 

[FvM]

Wah-wah is a filter involving little complexity, akthough the inductor adds some non-linearity. I believe that it's possible to determine it's parameters with sufficient accuracy.

I don't think that there's something special about the transistors in your device. They are most likely small-signal silicon BJT, probably germanium in case of a very old design, may be JFETs. A classical "Cry Baby" device uses a darlington transistor in the input stage.

Anyway, I agree that the linear frequency response can be measured, the non-linear effects need to be determined with varying inout voltage.e

 

[mr_monster]

I know you would think that the wah is a simple circuit thus the device is easy to reproduce, that is true if you are dealing with a device which is closer to being ideal like the schematic. Once you have a device with parts which are far from being ideal you start to get a complex system which is hard to reproduce. The transistors are very standard BJT BC10x type, however you would be able to see that they do not distort the same way as modern devices of the same type even if biased perfectly the same. These non-linearities are something that needs to be accounted for because it is very noticeable. But I agree that the effect of the highly non-ideal inductor is much greater than anything else. Overall it makes the filter to be tuned to different freq. for different input amplitudes. Sounds crazy huh? But it does that in addition to clipping the signal sometimes. This is not your ordinary off the shelf wah pedal, something was clearly very wrong when they did this unit but somehow it works magically. I used to believe that these devices are easy to reproduce, until I got my hand on this one.