How to map frequency response???

[Bill Mountain]

What if I have an effect that I like but don't have a schematic and like the way is sounds in a specific setting, what tools would I need to map the frequency response?

[Jaicen_solo]

Get yourself a class 1 or 2 sound level meter and run a 1/3rd octave frequency analysis on it.

Not sure what you'll do with that to be fair though. The frequency response is likely to be dynamic though, depending on a number of factors such as input gain etc.

I would just find someone with a schematic.

[Mark Hammer]

In olden days, it would be done via a sweep generator and spectrum analyser.  I doubt many here would consider that a fair representation of guitar signal, though.  I suspect that there would probably be something in the software area that could do that while you feed a suitable guitar sample into the circuit.  Hell, I have a spectrum analyzer cartridge for my old mid-80's Radio Shack CoCo that would do that so there has to be somethng out there.

[Bill Mountain]

It's a popular pedal and I have seen a schematic of a clone of it.  I won't divulge what the pedal is because I wanted this thread to be about technique of mapping the response.  I have modeled the clone when I tried to replicate the frequency of a certain eq setting (pedal is played clean) I couldn't get the same sound.  So either the clone is off, there is something I don't understand, or this particular pedal has some "mojo".  My goal is to get a picture of the frequency response that I can attempt to replicate with some filters that I can switch in and out when I want that particular sound.

My goal is also to better understand my tone preferences and a deeper understanding of how frequencies effect my tone.  There is no magic.  This pedal just has a normal eq with some preset filters and that I have a pretty good idea of what they are and how they are set but I would love to do my own tests.  My buddy used to work at a speaker engineering company so I'm sure he could also help me out with it.  Maybe he could even get one of his old work buddies to do it with/for us.  I juts wanted to understand what was involved.

[Jaicen_solo]

Right, I see where you're going with it, but I'm not sure you'll achieve what you want just mapping the freq response. Can you not tweak till it sounds right?

[Bill Mountain]

Right, I see where you're going with it, but I'm not sure you'll achieve what you want just mapping the freq response. Can you not tweak till it sounds right?

I've tried.  Something is missing when I breadboard it?!!!??

[teemuk]

Get yourself a class 1 or 2 sound level meter and run a 1/3rd octave frequency analysis on it.

Oscilloscope or a true RMS multimeter work much better and allow tracking true output of the effect much more accurately. One can compare input vs. output voltage at certain frequency intervals. The difference between the signals is the response but the quantity of voltage must be converted to dB to turn the data into more sensible format. Oh, as this kind of test setup is what it is, somekind of a decent signal generator is needed as well.

It's a popular pedal and I have seen a schematic of a clone of it.

If it's a distortion pedal then frequency response alone doesn't tell everything. You are interested in frequency responses both before and after distorting stages and naturally this requires higher number of tests where samples of output signal are taken from aforementioned places.

If you really want to learn about effects of EQ then make yourself a test setup: graphic EQ (with decent number of bands and preferably with output gain control) running into diode clipping stage feeding another graphic EQ. Have fun tweaking the knobs for hours marvelling how puzzlingly large selection of different tones you have at your hands. When done, cascade that thing to yet another diode clipper (you can experiment with symmetric and asymmetric setups in both) and third graphic EQ. Again have fun tweaking the knobs. This will teach you more about tonal effects of EQ than graphs ever will.

[Bill Mountain]

Get yourself a class 1 or 2 sound level meter and run a 1/3rd octave frequency analysis on it.

Oscilloscope or a true RMS multimeter work much better and allow tracking true output of the effect much more accurately. One can compare input vs. output voltage at certain frequency intervals. The difference between the signals is the response but the quantity of voltage must be converted to dB to turn the data into more sensible format. Oh, as this kind of test setup is what it is, somekind of a decent signal generator is needed as well.

It's a popular pedal and I have seen a schematic of a clone of it.

If it's a distortion pedal then frequency response alone doesn't tell everything. You are interested in frequency responses both before and after distorting stages and naturally this requires higher number of tests where samples of output signal are taken from aforementioned places.

If you really want to learn about effects of EQ then make yourself a test setup: graphic EQ (with decent number of bands and preferably with output gain control) running into diode clipping stage feeding another graphic EQ. Have fun tweaking the knobs for hours marvelling how puzzlingly large selection of different tones you have at your hands. When done, cascade that thing to yet another diode clipper (you can experiment with symmetric and asymmetric setups in both) and third graphic EQ. Again have fun tweaking the knobs. This will teach you more about tonal effects of EQ than graphs ever will.

This is some great advice but I'm actually talking about a clean eq.  No dirt.

[markeebee]

Convolution?

http://designingsound.org/2012/12/recording-impulse-responses/

[Bill Mountain]

I guess another question would be...how do big manufactures do it when they want to copy the frequency response of vintage amps and such.  They must run some sort of tests beyond reading schematics.  What do modelling software companies do?

[Bill Mountain]

Convolution?

http://designingsound.org/2012/12/recording-impulse-responses/

This might work.  I could use a recording interface to generate the sine sweep and record the responses and I'm sure there is software that can translate it all for me.  I knew about convolution reverbs but never thought it could be applied like this.  Interesting.

[markeebee]

Obviously, the really difficult bit will be creating hardware to replicate the impulse response.  There are more questions than answers, as the song says.

[tubegeek]

What do modelling software companies do?

It depends.

For a purely EQ-based analysis, the signal generator/level measurement suggestion above is the way to go. But - as you've learned - that's not always the really important thing about the way the effect changes the signal.

My understanding (which is somewhat shallow here) is that transfer characteristics of active devices are modeled mathematically one by one and then the complete circuit is modeled using a composite of the individual models. In this way, deviations from ideal device behavior can be incorporated into the models and accounted for.

I guess it's kind of like SPICE modeling but with fancier device models?

[waltk]

I made/used this...

[Bill Mountain]

So there's this...

http://www.waves.com/plugins/q-clone

It's a vintage eq modelling and capture software.

[R.G.]

There are sound-card apps (even free, I think) that are intended to do just this for speakers. They may have to have mike inputs. Can't remember

[PRR]

As R.G. says: there's an app for that.

You may need a pad between soundcard out and pedal in. You may need a buffer between pedal out and soundcard in. You should first sweep just pad+buffer to prove they are "flat" (or flat-enuff).
___________________________________

Sound level meter requires sound. As I understand your question, you want the pedal alone, and sound means going through a speaker. Speakers are VERY un-flat. In principle you can compare the jaggy curve from the good pedal against the jaggy curve from the bad pedal. In practice the peaks/dips in a speaker are so deep, and change with the slightest movement of speaker, mike, experimenter, or cat, that this may be futile.
___________________________________

"Sweep generator and spectrum analyser" is a fairly specialized technique and the boxes are not household items. It is how we used to do speakers and rooms, although it can be done on pure electric signals.
___________________________________

THE way to do it, before computers, is:

Lots of note-paper.

Get graph paper (ask your grandpaw). Work in dB versus log-freq, so you want lin/log scales. If all else fails, find old Stereophile articles and copy the graphs, plot your pedal with a different color pen. (I suppose you can do it in eXcel, but I hate that garbage.)

Get a decent audio signal generator. Must be a respectable Sine. Function generators with sine-approximation are good-enuff for most audio as long as there is no strong TIZZ from the 3rd harmonic. Classics are Hewlett-Packard 200AB, and the HeathKit transistor sig-gens.

Get a decent AC/Audio voltmeter. A tube VTVM is a good tool because it is dead-flat FAR above the audio range. However sensitivity is limited. Digital MMs can be adequate or awful. Many fall-off by 400Hz, especially "true RMS" products; worse, the fall-off may vary with signal level which is real counter-productive when measuring level versus frequency. The ideal tool is a Boonton, or more likely an H-P ACVM: high sensitivity and a dB-calibrated meter.

In any case you "calibrate" a short straight wire, to know how un-flat your sig-gen and meter are. A H-P 200AB will rise/droop a dB at the top of each range, make a note. (The trim is tedious, because fixing the un-flat screws the freq calibration, so you go back and forth until you go nuts.) It is neat that it is continuously variable, so you can find narrow dips; however any freq-change means waiting a few seconds for the bobble to settle. The Heath transistor jobs are switched, which isn't such a problem as you might think (it was my main tool when I did freq-resp checks for a living). The function-generator is usually continuous-freq and also zero-bobble, so may be easier for non-precision work (and all musical pedal freq-response is non-precision compared to hi-fi and broadcast checks).

Signal must pass through the pedal *below* distortion. (Response of a distorty channel is a complicated thing of dubious meaning; here the Spectrum Analyzer may be of use; however things distort different on one-tone, multi-tone, and many-tone signals, so hand-plotting is probably futile.)

To start, feed 1KHz and observe the output. Bring it up to distortion then way-way down from that point. Note the meter reading. Switch to 100Hz, then to 10KHz, note the readings. If they are WAY different from 1KHz (shouldn't-be in studio/hi-fi, but likely in some pedals and tape/phono-EQ) then figure out how you are going to cover the spread while staying far above noise, well below distortion, and stay on your meter scale.

Now go to 20Hz, make note. 40Hz, make note. 80Hz, make note. 160Hz, make note. 320Hz, make note. There's less than a dozen Octaves in the audio band, this does not take long. ("Most" guitar is 80Hz-8KHz, so you can do just 8 readings... but if you are not sure, plot it all and a bit more. One preamp's glitch was only uncovered with a sweep below 5Hz; it boomed like the Sea of Japan and the backwash upset the audio clarity.)

If these don't plot smooth, go to the in-between frequencies (but narrow dips are rare in simple audio (Fender ToneStack the exception; narrow dip near 700Hz)).

Repeat with other pedal.

[FiveseveN]

Why not use a (white/pink) noise source and integrate the FFT over 2-10 seconds? Or even use infinite peak hold (works for sweeps too)?

[Bill Mountain]

These are all great ideas.  I called up my engineer buddy who use to do design work at a home/car speaker manufacturer.  He has some precision mics he uses for his own home stereo builds but he said we really need to use an AP machine.  He suggested we get in touch with his old coworkers and see if they could run it for me.

I don't need to capture amp/cabinet sound because if I do model this eq I will just be playing it through my speakers anyway.

I have a little Focusrite recording interface so I think I'm going to try some of the app versions.

That Waves Q-Clone stuff I showed is pretty cool here's a video of it in action (with an effect pedal no less):

http://www.youtube.com/watch?v=yEFrvj_VQR8

[tubegeek]

so you go back and forth until you go nuts.

Is it perhaps possible to stop just prior to that point or am I asking the wrong guy?