I've owned VCS 3s and Moog Modulars, and the early Moog oscillators, drifting away, outdo the EMS for a musically interesting varying tone. I don't mean out-of-control, but as he implies, something that is pleasing to hear.
If you want "more" of that from your LVCOs, consider installing a switch for the oscillator core capacitor, as the EMS (and even the Moog 901b) utilize vintage Mullard mylar capacitors, not polystyrene/polycarbonate.
C3, 2.7nf, is the polystyrene core capacitor, yes?
How to? This applies:
"Ideally, each Oscillator should have the two capacitors on the pcb joined at the current input end, to keep this sensitive point a short path. Fly a screened twisted pair cable wired to the other end of each capacitor to a panel switch such that it switches the larger capacitor in parallel. Ground the screen from a nearby panel control. "
"I've chosen the same noisy 4-transistor exponential converter that EMS used in the VCS3. I'm not overly scientific about it - I just like the sound of it."
Then why change anything at all about it if it sounds good as it is/was?
Adding the less-accurate cap may increase liking the sound of it- it's up to you. I remember Dean Hinton's original EMS mods page mentioned that its lowly vintage power supply had something to do with the "bad" behaviour of the instrument; you may wish to use a less-than-Kosher power supply to help regain the desirable character...I also note that when I sampled a VCS3 all those years ago, that the tip of each sawtooth waveform had a different overshoot, meaning that each cycle was slightly different. I don't know how that will turn out at audio frequencies but it has to be yet another aspect of Difference from the modern. I don't know how to re-create this, or if it is the same on all EMS oscillators. (I posted this at TGS many years ago; someone took my statement incorrectly, imagining that I meant that I could somehow hear each wavecycle at audio frequencies, and differentiate them...certainly not the case, but you can see the difference in a sample. I'm not sure about the cycle-to-cycle homogeny of the LOs, but it's worth considering.)
You may want to make sure that it has the same exponential .32V/Octave response (or so) for deep exponential FM.
Also, part of an interesting drone/its overall interesting character may be Oscillator 3. It doesn't, in the several images I've seen online, have 2C746 matched, single-housing trannies, which aid in heat transfer/stability. The 2N5172 pair which appear to be used in its core instead of the 2C, are sometimes (perhaps always) epoxied together without heat-transfer compound. This may produce more drift than the same oscillator with a 2C746 or other matched, single-housing pair.
"Back to the crow, back to the stone age!" -March Violets. Yes, I'm suggesting that on at least one of your LOs, that you install a matched but heat untreated pair. All of your oscillators don't have to sound like a bomber squadron, but one with a bit more drift will increase interest and perhaps get you closer to what sounds good about the original. Consider two-oscillator patches and use the third oscillator for FX. Sometimes, use all three for the different character. The Minimoog is also set up in this manner, however its Oscillator 3 is likely (haven't looked) of equal stability as the first two.
You can also put the Tempco resistor on a switch, DPDT, to select between it and a resistor of the same measured value but standard carbon comp or metal film. See the above link and considerations per switching core capacitors.
But of course, just ignore me; no one actually likes those old instruments or why they sound and act the way they do.
And to those who leap to fear any time I discuss the facts as I am learning them, get a life. Just get a life.