Looking at your results, the voltage measurements say how well the DAC and buffer combination is doing. This should normally be measured into a fixed load, say 100k, but in your case the output buffer has a very low output impedance making the voltage almost independent of load.
Given the constraints of your meter, all you can say is that the circuit is performing correctly. Any errors in the trimming of the circuit are swamped by errors in the meter and can't be measured directly. (They can, within some constraints, be measured indirectly).
The frequency measurements say how well the complete system of DAC, buffer, and oscillator is doing. Again the sources of error in the measurement are the system itself and the meter used to measure it.
The tracking of an analog oscillator is determined by the linearity of the exponential converter and the care taken to correct capacitance losses at higher frequencies. (This is why some oscillators have range or footing switches; they switch in different values of capacitance so that on a given range they have to cover less octaves and are thus less affected by high frequency error). A digital oscillator should track a large number of octaves as it is immune to those effects.
For the measurement, there are two ways to measure frequency (count cycles in a fixed period of time, or time one cycle) which have different error behaviour; the first is accurate at high frequencies but inaccurate at very low ones and the second is the opposite way around. I suspect that your meter is using the first method, as it is both more accurate and quicker for most frequencies of interest; that may explain the non-linearity in the errors towards the lower octaves.
So, lets look at your frequency results expressed as errors in cents. 100 cents is a semitone. 50cents is an easily heard error. 20 to 10 cents is hard to hear (depends a lot on the person and their sense of pitch). Below 10 cents is imperceptible even for people with perfect pitch.
Code:
Freq Ratio Cents Error
16.4 0.031538 -5984.0 15.907
33.4 0.064230 -4752.7 47.290
66.4 0.127692 -3563.1 36.892
130.8 0.251538 -2389.3 10.621
260.9 0.501730 -1194.0 5.9823
520.0 1.000000 1.0000 0
1038 1.99615 1196.6 -3.3325
2069 3.97884 2390.8 -9.1798
4120 7.92307 3583.2 -16.727
8194.5 15.7586 4773.6 -26.313
In general excellent, there is a bit of non-linearity on the lower octaves. That might be due to measurement inaccuracy and there are also less digits of precision in the lowest measurements.
As there are ten measurements, we can look at the overal linearity and guess (with more confidence than we could from a single measurement) that the gain of the op-amp buffer is very slightly less than it should be.
Looking at a
linear regression of the the frequencies (in cents relative to the 0V frequency) from the best seven octaves only (to avoid what seems to be low frequency measurement error, and dropping the uppermost octave as well) I get a slope of 1192.5 when it should be 1200. So a slight trim to increase the gain might give even better tracking. This error probably is due to the tolerance of the resistors used to set the gain.