"Elliptical" LFO in Phase 45

[sta63bmx]

I think Bucksears mentioned this with a Phase 100.  In the ELectric Mistress build I did, the LFO sweep was 50/50, really nice and round.  But in the Phase 45 clone I built from the tonepad layout (hooray for easy builds!) the sweep has that elliptical feel to it.  In the phase 45, would a balanced "circular" sweep be accomplished with a sinusoidal LFO or a square-wave 50% duty cycle LFO or...?

I haven't looked at the LFO signal on the scope yet, but I will.  I'm curious about this.  Has anyone experienced this in other builds?  The phasing sound I get from this pedal is pretty strong and I think it's all working properly (has the overloaded input distortion issues other people have mentioned, but it's ok with PAF-style pickups).

[Seljer]

I spent all of yesterday afternoon messing with this and I think it might be of use, simulation of the LFO in the Phase 45

[Transmogrifox]

It could be an interesting LFO that makes a phaser sound "circular" and balanced.  Typically one would think the LFO should be sinusoidal to do this, but the conversion of the sine wave in the "1/(2*pi*R*C)" term changes the sweep to something different.  In theory this should be sounding rather "circular" since it moves through the lower frequencies more slowly, then accelerates through the higher frequencies in the sweep.  Typically synths have an exponential generator for this reason.  

Ultimately, that's what you need to compensate:  Cancel the 1/R term, then apply a 2^X term so the filter spends an equivalent amount of time resonating in each octave.  You could do this by modifiying the LFO, or by changing the control voltage transfer to the FET gates...and the transfer from voltage to resistance on the FETs needs also to be considered.

[Transmogrifox]

You have the 'inverting" and "noninverting" labels backward.  I messed you up on that one.  No wonder you were so confused in the other post. I was being an airhead when I insisted on that.  Just had to see it on a picture for it to register that I had dumbly misinformed you.

[Transmogrifox]

You have the 'inverting" and "noninverting" labels backward.  I messed you up on that one.  No wonder you were so confused in the other post. I was being an airhead when I insisted on that.  Just had to see it on a picture for it to register that I had dumbly misinformed you.

This was with reference to a post by comfortably numb about a Dist+ type clone.  I'm apologizing in advance if you had gotten the misinformation from that post.

[Seljer]

Ah yeah, I see that (for somereason my mind connectrf the - to "non"). The labels were there just so I roughly knew what was what when I was trying to figure out how the thing works in the first place

[sta63bmx]

I will take a picture of that LFO wave later today or tomorrow and compare it to the shape in that simulation.

Are you saying that it's analogous to why a logarithmic pot "sounds" linear?  Like....a triangle LFO for volume might not sound even because it's a linear change up and down?  ANd are you pulling the 1/(2 pi r c) term from RG's technology of phasers article or...? 

In that circuit simulation shown, are R2 and C2 the ones responsible for the charge/discharge time of the waveform?  I could always jack around with those to make something with a different-shaped wave then, I suppose.  This is the fun of building stuff and jacking around with it.  In the lab the scope is annoying because it means something is broken or not working yet and I don't know why so I have to look at it.  At home the scope is fun because it's a tool of exploration.

[Seljer]

from my mucking about with it in simulation:
R5 is the rate control and along with the size C2 are most responsible for the speed of the thing

R2 doesn't really do much (except if its too large it takes a while for the circuit to start up), it lightly affects the frequency
C2 doesn't do much, it lightly rounds down the signal if its a bit larger (but again, if its too large it takes a while for the circuit to start, but it then works like normal, and also some slight change to the frequency), same deal as with R2 overall

smaller values of R2 make the wave more like a regular DC charging/discharging of a capacitor, larger values make it more triangle wave-ish (but too large and then it doesnt work), it also affects the frequency of the thing

and you can get some sawtoothwave-ish type stuff by changing the ratio between R3 and R4

and you can place a low pass filter on the output of the thing (on the left by the postive side of C2) to get it a bit more sinewave-ish by getting rid of some of the harmonics, not sure if that would work in the actual circuit (I was going to build the Phase45 but I found out I bought a 220ohm trimpot instead of a 250kohm one :S)

[sta63bmx]

Awesome info!  I'll look for those parts on my board and see where they're at and then get to fooling around with it.

[Transmogrifox]

I will take a picture of that LFO wave later today or tomorrow and compare it to the shape in that simulation.

Are you saying that it's analogous to why a logarithmic pot "sounds" linear?  Like....a triangle LFO for volume might not sound even because it's a linear change up and down?  ANd are you pulling the 1/(2 pi r c) term from RG's technology of phasers article or...? 

In that circuit simulation shown, are R2 and C2 the ones responsible for the charge/discharge time of the waveform?  I could always jack around with those to make something with a different-shaped wave then, I suppose.  This is the fun of building stuff and jacking around with it.  In the lab the scope is annoying because it means something is broken or not working yet and I don't know why so I have to look at it.  At home the scope is fun because it's a tool of exploration.

Yes, I am saying that it is analogous to why a logarithmic pot "sounds" linear.  I pulled the 1/2piRC thing out of the schematic.  The JFET acts like a variable resistor for small signal levels, and the resistor parallel performs 3 things: sets a maximum resistance, and biases the op amp, as well as biasing the source and drain at the same potential.

The output impedance of the stage preceding is very low, so we disregard it.  Therefore, you're left with an RC high-pass filter feeding the noninverting input of the op amp.  The inverting side is fed with pure, unfiltered signal.  The two add, forming a unity gain output with a linear phase response from 0 to -180 degrees centered at the RC high pass filter's 3dB frequency.  Stack a bunch of these together and add it to the dry signal and you have what's called a phaser.

The 3dB frequency of the high pass filter (neglecting the nonlinearity of the JFET) is 1/(2pi*Req*C), where Req is the equivalent resistance of the JFET in parallel with the bias resistor.

The 3dB frequency changes as "1/X" with increase in Req.  That's just with respect to "Req".

Req does not change linearly with the control voltage, either.  If Rjfet changed linearly with the control voltage (which it does not), you would still have the Rjfet||Rbias effect: 
Rjfet*Rbias
--------------
(Rjfet + Rbias)

So effectively your transfer function from control voltage to center frequency is this:

               R(Vc)+ Rbias
Fc    =  --------------------------
           (2*pi*C)(Rbias*R(Vc))

Where the term, R(Vc) is the function that describes the relationship between the JFET's approximate resistance as a function of the control voltage.

Your best bet to linearize this is not to try to calculate it.  It would be nice to have a scope, so you could see an AC waveform and determine its frequency.  You would want to sweep an AC sine wave on the phaser input, then measure at the first stage output, varying the signal generator frequency until you find the 3dB corner.  Record it in a table along with the corresponding control voltage.  Make 20 or 50 of such measurements, incrementing the control voltage each time and plot it on a graph to view the transfer.

The next step would be to form an analog network that approximates the inverse of this function to pre-process the control voltage before applying it to the gate of the JFET.

Then apply an exponential generator to the control voltage preceding all of this mess so that you have a nice exponential transfer.

Sound complicated?

Try some LFO's until you find one you like.  You can probably tune one in by ear faster than you can do all that garbage I mentioned.