Output Impedance of a Function Generator

[YouAre]

I'm building a quick XR2206 based function generator, mainly to test guitar pedals and possibly amplifiers. The main features would be Frequency, Amplitude, Bias Voltage, and Sine/Triangle. The output of the function generator will likely come from an non-inverting opamp.

My question is whether or not it would be useful to include a 15k-ish series resistance (switchable) on the output of the non-inverting opamp stage to simulate the output impedance of a guitar pickup. I understand that the Geofex article on guitar pickup simulation ( http://www.muzique.com/lab/pickups.htm ) mentions that a resistor alone won't get us the proper frequency response, but I'm assuming that's a non-issue since this is a variable frequency device.

Does anyone have any thoughts on adding this artificial output impedance to this circuit? I guess it boils down to...does a guitar effect circuit's input impedance loading down our voltage source (guitar or function generator) affect the frequency response or does it affect the overall output as well?

Thanks guys!

[R.G.]

My question is whether or not it would be useful to include a 15k-ish series resistance (switchable) on the output of the non-inverting opamp stage to simulate the output impedance of a guitar pickup. I understand that the Geofex article on guitar pickup simulation ( http://www.muzique.com/lab/pickups.htm ) mentions that a resistor alone won't get us the proper frequency response, but I'm assuming that's a non-issue since this is a variable frequency device.

Bad assumption.

Plus you're making the standard mistake: you're forgetting that the loading on the output of the thing is what makes the frequency response an issue, not just what's in side the generator. Guitar pickups look like a resistor of 4K to 18K (single coil to humbucker) in series with a 2H to 4H (again, single coil to humbucker)  in series with the actual voltage generated. If you want your generator to interact with a load the same way a guitar pickup might (depends on guitar controls, too) then you could switchably include a resistor plus an inductor. Then the load plus the impedance will get close to what some guitar might do.

This is only important when driving low impedance input effects, and primarily fuzz face and similar. The reason effects have 1M and greater impedances is that they make the guitar source impedance not give you frequency rolloff, so then it wouldn't matter.

Does anyone have any thoughts on adding this artificial output impedance to this circuit?

It's good if you're going to experiment with fuzz face like pedals. Otherwise, not so much.

I guess it boils down to...does a guitar effect circuit's input impedance loading down our voltage source (guitar or function generator) affect the frequency response or does it affect the overall output as well?

Both.

[PRR]

IMHO, for general testing....

Signal generator should be Solid, low-impedance. Test bench should have ways to modify this. For my needs, a couple clip-leads and a resistor-drawer usually filled any need for specific source impedances. If you are on government salary, you can take the time to build a switch-box with a vast range of possible source impedances.

Though when I was messing with loudspeakers a lot, I added a booster to my Heath sine oscillator. The Heath was nominal 600 Ohms, which is not "low" or "solid" when facing 8 Ohms. I added a power amp chip which would drive 4 Ohms happy, and again clip-leaded whatever 4 or 100 Ohm added resistance I wanted for a specific test.

The XR and friends can have output far above what is needed to test low-level guitar stuff. Aside from a pot, I would want a couple levels of heavy attenuation, perhaps -20dB (10:1) and -40dB (100:1). (That Heath had -10dB to -60dB attenuations.)

[YouAre]

Bad assumption.

Well...sh*t.

Plus you're making the standard mistake: you're forgetting that the loading on the output of the thing is what makes the frequency response an issue, not just what's in side the generator. Guitar pickups look like a resistor of 4K to 18K (single coil to humbucker) in series with a 2H to 4H (again, single coil to humbucker)  in series with the actual voltage generated. If you want your generator to interact with a load the same way a guitar pickup might (depends on guitar controls, too) then you could switchably include a resistor plus an inductor. Then the load plus the impedance will get close to what some guitar might do.

I should've been much more clear about the use and purpose of my generator. One of the uses would be to test the frequency response of certain clipping circuits. I feel like filtering the output of our signal source would introduce a variable that we don't want, hence why I figured that we wanted the impedance without the change in frequency response.

This is only important when driving low impedance input effects, and primarily fuzz face and similar. The reason effects have 1M and greater impedances is that they make the guitar source impedance not give you frequency rolloff, so then it wouldn't matter.

I figured that if I was running this into a standard guitar pedal circuit, I'd be fine as you'd stated. But what if I start running into only a circuit block of a circuit (e.g. a Tone Clipper, inverting op-amp stage, etc)?

It's good if you're going to experiment with fuzz face like pedals. Otherwise, not so much.

Noted.

I guess it boils down to...does a guitar effect circuit's input impedance loading down our voltage source (guitar or function generator) affect the frequency response or does it affect the overall output as well?

Both.

Well that was a dumb question. Again, should've been more clear. This device would be used to test the headroom, frequency response, observe clipping at different frequencies, and so on. As I'd mentioned earlier, that filtering the output of this would not give us the cleanest results. Everything would have the inherent upper mid bump and high end roll off. This is why I wanted to include a non frequency dependent impedance.

IMHO, for general testing....

Signal generator should be Solid, low-impedance. Test bench should have ways to modify this. For my needs, a couple clip-leads and a resistor-drawer usually filled any need for specific source impedances. If you are on government salary, you can take the time to build a switch-box with a vast range of possible source impedances.

I agree to a certain extent. But if we're always testing what a 15k source impedance does to a circuit, wouldn't it be most efficient to include that option as a switch on something you always use?

Though when I was messing with loudspeakers a lot, I added a booster to my Heath sine oscillator. The Heath was nominal 600 Ohms, which is not "low" or "solid" when facing 8 Ohms. I added a power amp chip which would drive 4 Ohms happy, and again clip-leaded whatever 4 or 100 Ohm added resistance I wanted for a specific test.

The XR and friends can have output far above what is needed to test low-level guitar stuff. Aside from a pot, I would want a couple levels of heavy attenuation, perhaps -20dB (10:1) and -40dB (100:1). (That Heath had -10dB to -60dB attenuations.)

You're right, and I had thought about this after posting yesterday actually. Instead of running the output of the XR into a non-inverting opamp, I will hit it into an inverting opamp. That way, I can get attenuation and "boost" on a single dial. I'm not sure what the output impedance of an inverting amplifier is off hand.

Sidenote: the triangle is larger in amplitude than the sine wave output, so I'm compensating by switching the "amplitude" parameter when I switch output types. I'm calibrating to match as best as possible of course.


Thanks again guys.

[duck_arse]

youare - I have an xr2206 based sweeper, I can circuit-you-up with it, which I only recently started hack-modding to add spot f. I've since discovered the sine/tri signal difference, a problem I still need to fixxe. I long ago added a V/10 switch, to make the output pot usable at sig levels. now I find the original circuit, w/ volume control for level, is not the app-note method. more fixxe needed.

[ashcat_lt]

Step back a minute.  The 15K by itself will interact with cable capacitance to roll of some of the highs, but unless the cable is REALLY long it won't be much in the guitar range.  Much more important is the inductance of the pickup, which works with the in-Z of the load to kill quite a lot of what might be properly called treble.  You keep saying impedance and then only talking about resistance.

But!  What exactly are testing here?  You want to stick a sine or triangle into some circuit (or snippet thereof) and see how sinish or triangular it is on the other end, and maybe how much bigger or smaller it might be.  You're not testing the frequency response of the guitar plugged into it, you're not even really getting a clear picture of the frequency response of your circuit.  You're seeing (mostly) the amplitude transfer curve of your circuit.

What happens if you send a sine wave through a high-pass filter to simulate the pickup response?  Nothing, if its frequency is in the pass band.  If it's above cutoff, the sine wave gets smaller.  Non-linear circuits will react differently to smaller inputs, but will this tell you much about your circuit specifically?  I tend to think not.

What happens if you run a triangle through a low-pass filter?  It starts to round off.  The closer the fundamental is to the cutoff, the more the thing looks like a sine wave.  As fundamental passes cutoff, the whole thing gets smaller.  How can this tell you anything useful about what your circuit is actually doing to the triangle?  I like triangles for testing non-linearities because the will just show you the transfer curve and how it deviates from the linear function, but if you round it off before you even get there...

Now, fuzz face pretty much depends on destroying all of the highs and most of the mids coming out of the guitar, but this doesn't really much affect its transfer curve.

To put this whole thing another way - by turning the Amplitude knob and Frequency knob on your signal generator you can get the effect of any damn filter you want!

[YouAre]

youare - I have an xr2206 based sweeper, I can circuit-you-up with it, which I only recently started hack-modding to add spot f. I've since discovered the sine/tri signal difference, a problem I still need to fixxe. I long ago added a V/10 switch, to make the output pot usable at sig levels. now I find the original circuit, w/ volume control for level, is not the app-note method. more fixxe needed.

The sine triangle difference isn't an issue. Instead of using a SPST to switch between sine or triangle, use a DPDT to simultaneously switch the amplitude resistor. Works for me.

The output volume control in my setup is independent of the amplitude resistor. This simplifies the setup greatly.

Step back a minute.  The 15K by itself will interact with cable capacitance to roll of some of the highs, but unless the cable is REALLY long it won't be much in the guitar range.  Much more important is the inductance of the pickup, which works with the in-Z of the load to kill quite a lot of what might be properly called treble.  You keep saying impedance and then only talking about resistance.

Won't be using a guitar cable. I have a very short shielded cable ( 10") running from terminal blocks, so I doubt capacitance will be a huge issue. You're absolutely right, I am referring to resistance because I'm trying to remove the frequency dependence for reasons detailed below.

But!  What exactly are testing here?  You want to stick a sine or triangle into some circuit (or snippet thereof) and see how sinish or triangular it is on the other end, and maybe how much bigger or smaller it might be.  You're not testing the frequency response of the guitar plugged into it, you're not even really getting a clear picture of the frequency response of your circuit.  You're seeing (mostly) the amplitude transfer curve of your circuit.

I'm mainly testing headroom/clipping characteristics. I also wanted to play around with frequency dependent clipping. Something like that would show up more dramatically on a scope than watching the amplitude vary.

What happens if you send a sine wave through a high-pass filter to simulate the pickup response?  Nothing, if its frequency is in the pass band.  If it's above cutoff, the sine wave gets smaller.  Non-linear circuits will react differently to smaller inputs, but will this tell you much about your circuit specifically?  I tend to think not.

What happens if you run a triangle through a low-pass filter?  It starts to round off.  The closer the fundamental is to the cutoff, the more the thing looks like a sine wave.  As fundamental passes cutoff, the whole thing gets smaller.  How can this tell you anything useful about what your circuit is actually doing to the triangle?  I like triangles for testing non-linearities because the will just show you the transfer curve and how it deviates from the linear function, but if you round it off before you even get there...

Did not know this! I thought the triangle would get smaller in amplitude too! That's huge, thank you.

I still think that the sinewave would be good for slightly closer look at what a clipping circuit would do to our guitar signal.

Now, fuzz face pretty much depends on destroying all of the highs and most of the mids coming out of the guitar, but this doesn't really much affect its transfer curve.

To put this whole thing another way - by turning the Amplitude knob and Frequency knob on your signal generator you can get the effect of any damn filter you want!

Luuuckily, I don't think I'll be touching a fuzz face for a little while.


Thank you both!