Current controlled oscillator, LM13700 transconductance amplifier

Theory

There is a neat device for building a voltage, or rather current, controlled oscillator. It's a transconductance amplifier called LM13700. It has a positive and negative voltage input, but a current output. It also has a current bias input. The output current is proportional to the differential input times the current bias input. The LM13700 has a lot of applications such as voltage controlled -amplifier, -resistor and -filter and multiplier.
 In the data sheet, page 16, there is a simple VCO, which works very similar to the VCO i built in an earlier blog post.
I did my own schematic of the VCO with pin numbers of the device:

Voltage (or rather current-) controlled oscillator taken from the LM13700 datasheet.

I want my oscillator to play between C2, 65.4064Hz, and C6, 1046.50Hz. The frequency of the VCO is: f_OSC=I_C2/(4C_fI_AR_A). The current for the lowest key, V_C = 1V, was measured to 64μA, and current for the highest, V_C = 5V, was 1.02mA. That means we have 4C_f I_A R_A = I_C2f_OSC, with the two values I_C2f_OSC = 64μA·65.4064Hz =979nC and I_C2f_OSC =  1.02·mA1046.50Hz = 975nC, which has a mean of 977nC. Our target will therefore be 4C_fI_AR_A = 977nC.
 We also have that the output amplitudes of the oscillator is A=I_AR_A, so that's a good place to start deciding our component values. If we go by the data sheet and set R_A= 4.7kΩ (closest E12 resistor to  5.1kΩ) and we want 1V amplitude, we get I_A=1V/4.7kΩ = 213μA. With a 39kΩ resistor connected to positive rail we get  I_A=231μA, close enough. We now only have to decide the value of C_f.
C_{f =} I_C2f_OSC /(4I_AR_A)=977nC/(4·231μA·4.7kΩ) = 225nF. I found two 120nF caps in my junk box, so I'll use them in parallel to get 240nF.

Circuit build

The LM13700 is hid under the decoupling cap. The black alligator clip is the current source.

The two waveforms at V_C = 3V. 

We can see that the waveforms have a nice shape, but the amplitude is twice of what I calculated. The frequency is a bit on the lower side. I'll hook the VCO up to the log converter and the MIDI2VC, and see how parameters can be calibrated. It would be interesting to figure out why the VCO differs in function from the calculations, but right now I'm more eager to test out the MIDI2VC...