Saturday, March 2, 2013

Polysix - Resonance-Controlled VCA

On the Korg Polysix Yahoo Group, there was a question posed about the purpose of the resonance-controlled "VCF" that was on the KLM-368 PCB.  After some consideration of the schematic and after reviewing the datasheet for the LM13600 IC, it's my opinion that this part of the circuit is a voltage controlled amplifier (VCA), not a voltage controlled filter (VCF).

My initial hypothesis was that this part of the circuit was used to decrease the amplitude of the audio signal when the resonance was increased.  My thinking was that, because the audio signal usually gets very large once the filter starts to self-resonate, this circuit was used to lower the volume and keep that self-resonating signal under control.  It turns out, though, that this hypothesis was totally wrong.  It turns out that this circuit increases the volume of the signal when the resonance is increased.  Let's look at the data to see why...

Measuring the voltage across R152 to see how the LM13600 is being controlled.
First, I wanted to see how the resonance knob affected the control signal to IC20 (an LM13600 OTA) that is at the heart of this resonance-controlled VCA.  The LM13600 is a current-controlled device, so I measured the voltage across R152, which is the last element before the control current goes into the LM13600.  From this voltage value, I can compute the current simply by dividing the voltage value by R152's resistance (10K).

Excerpt from the Polysix Schematic for the Resonance-Controlled VCA on KLM-368 Prior to the Effects Section.
My procedure was to dial in a particular resonance setting on the resonance knob and to manually record the voltage at R152 using my digital multi-meter (DMM).  I repeated the measurement for different settings on the resonance knob.  The results are summarized in the first few columns of the table below.  Note that for settings between 0-1, the control current is unchanged.  Then, by turning the resonance knob from 1-3, the control current changes rapidly.  Above a setting of 3 (all the way to 10...I checked), there is no change in the control to the LM13600.  This surprised me!

To dig a little deeper, I went back and recorded the audio signal going into this circuit and coming out of this circuit.  The easiest place to grab the input signal is from TP1 on KLM-366.  The easiest place to grab the output is TP4 on KLM-368.  My procedure was to set the resonance knob at the desired setting, to play a note on the keyboard (C1), and to record the audio with my M-Audio recorder.  I repeated the recording at several resonance settings.  I then brought the recordings to the PC and analyzed them in Cool Edit Pro V1.2, where you could immediately see what was happening...

Raw Audio Waveform Recorded With Difference Resonance Settings. 
As you can see, at TP1 (the input to this circuit) we see that the amplitude of the signal decreases as the resonance is increased.   This means that, if we were listening to the audio at TP1, it would get quite a bit quieter when we turned up the resonance.  Compare this to TP4.  At TP4, the amplitude barely changes in response to the resonance knob.  That's what this circuit must be must be compensating for this drop in volume.  With this resonance-controlled VCA, the user can twiddle with the resonance knob without experiencing a big drop in volume.  Cool!

I'm not satisfied, yet, though.  Let's further quantify things.  With these signals in Cool Edit Pro, I can use the "Analyze" feature to assess the RMS value of each of these blocks of audio.  This is one way to estimate how loud we would perceive each block of audio.  By comparing RMS values, you can guess how much louder or quieter each bit of audio would be.  These values are summarized in the middle columns of the table.  As could can see, at TP1, the volume would drop by about 10 dB when you turn up the resonance.  That's a lot!  By contrast, at TP4, the volume only drops by a couple of dB.  That's pretty good compensation!

Ideally, I'd be able to relate the amount of compensation (gain) back to the performance parameters of the LM13600 IC that is doing all of the work.  To try to understand how the LM13600 is working, I already computed the control currents, which should control the amount of gain of the LM13600.  Using my measured RMS values, can can compute the (relative) amount of gain actually produced by the LM13600 in response to the resonance knob.  As you can see in the last column of the table, the effect of the resonance knob is to increase the signal gain by 6 dB once you get to a knob setting of "3".  Is this what we should be getting?

Measured Control Current to IC20 at Difference Settings of the Resonance Knob.  Also shown is the Measured Gain from IC20 Relative to the Gain at a Resonance Setting of Zero.
Sadly, the analysis of the LM13600 proved too challenging for me.  The datasheet has a couple different equations for the gain of the LM13600, but these equations assume a certain configuration (or values) of the circuit elements around the LM13600.  The Polysix circuit doesn't quite match any of these examples.  I tried to use a PSpice model of the LM13600 (via WinSpice) without any luck...the results were garbage.  So, sadly, I cannot confirm if this is the way that it is supposed to work.  Given that we've already seen that the circuit seems to nicely compensate for the resonance-induced change in volume, I'd say that the circuit is working well.  If it's working well, it's probably working as intended.

So, what's the take-away from this?  First, that the designers of the Polysix appear to have added this circuit so that you can play with the resonance knob without also getting changes in volume.  That's great.  Second, in my own notions of by-passing the whole KLM-368 board (to eliminate the added noise of the effects unit and to avoid the changes in frequency response due all the other circuit elements), I will lose this nice functionality.  That gives me pause.

What do you see in this circuit?  Have I missed any other aspects of its behavior?


  1. Same post resonance VCA in the Monopoly but with a simple transistor and not OTA. I was thinking about improving the Monopoly VCA because the double transistor VCAs make a lot of noise.
    Instead of 2 VCAs I think it is possible to use only one in the audio path and adjust the control signal according to the resonance setting.
    Candidates for the new VCA are LM13700 and SSM2164.

    1. Hi Covariance, definitely interested in more details on any mods that you make to the Mono/Poly VCA. Drop me a link if you post any details!

  2. I definetaly implemented this VCA curcuit + added a mixer stage for the resonance volume compensation.