Some of the simple folk do still claim to be able to tell the sonic difference between a DCO and a VCO.
https://youtu.be/b9UxnbGszaY?t=56
I feel that I can tell the difference between a DCO and a VCO at a certain early stage during sound sculpting, but not so much from the pure output produced by the oscillator waveform. Just listening to a raw saw tooth for example from the OB6, P6, or Rev2, if I asked 1000 participants in an Internet survey to pick out which one was a DCO, I bet the random guesses would be evenly distributed.
However, take those OSCs and do something with them... like set them all to square and modulate the shape slightly for a little PWM, and very quickly you can start picking up the sound profile of each synth, and start identifying which is the Prophet 6 vs the OB-6 vs the Rev2. I'm still not sure that would indicate which is VCO or DCO if one was not familiar with subtle characteristics of how each synth sounds.
Vintage Roland Junos were DCO based synths, yet still considered some of the warmest sounding analogs of their time. Yet, this is not to say all VCOs or DCOs are created equal, or that DCOs and a purely digital oscillator necessarily have the same characteristics.
Yeah, in an A/B test of a raw DCO vs VCO, you can't tell the difference. There is a very minor effect of harmonic jitter that all VCOs exhibit, but it's extremely subtle. I made a video about it a while back:
https://www.presetpatch.com/articles/VCO-Harmonic-JitterAs I've mentioned, the big difference with DCOs vs VCOs is not the raw analog character of them, but the juxtaposed combination of multiple VCOs that are not tuned to their nominal target. VCO synth oscillators are in tune in a "sweet spot" on the keybed, but progressively get sharper or flatter as you go up and down the keybed. Some modern VCO synths are very well tamed, and this difference in tuning may just be a couple cents over five octaves or so, but classic VCO synths will often have 5-10 cents difference in tuning over that octave range... and some (looking at you MemoryMoog), may exhibit upward of 20+ cents variance over a five octave range.
When you play two or three oscillators per voice, and play multiple voices at one, you're combining potentially dozens of oscillators, each playing slightly out of tune. This juxtaposition of non-nominal tuning results in natural phasing / detuning motion - that is what people think of when they refer to "the VCO character".
So, yeah, that's the objective/scientific difference between DCOs and VCOs that I studied. I wrote up my findings in the Voice Component Modeling (VCM) website:
http://www.voicecomponentmodeling.com/Some key points that should be re-emphasised:
1. From a raw oscillator standpoint, there is virtually no difference in sonic character between a single DCO and VCO oscillator. There is a little bit of harmonic jitter as mentioned above, but its extremely subtle. You can actually model this harmonic jitter in the Rev2 if you really want, and I will occasionally do this on Mono Lead and Bass patches, but it has diminishing returns, and in comparison with the large effect of Voice Modeling, its pretty much not needed.
2. The big difference is the juxtaposition of detuned oscillators, per voice. This detuning of each oscillator is RELATIVELY STABLE... ie: It doesn't wildly swing back and forth over time. Once the synth has warmed up, and you hit a certain voice/oscillator, if you hold it, it will have a tuning offset, but that offset will be stable (minus the harmonic jitter described above) This is the reason why Osc Slop / Osc Drift circuitry delivers a somewhat unrealistic / unnatural sound. These Slop circuits use the motion of multiple LFOs to swing the frequency back and forth of oscillators. For quick stab sounds this works okay, as you don't notice the artificial motion. But for strings, pads, brass and other acoustic ensembles with longer decay/release times, its much better to model the offsets per voice (Individual Voice Modeling / VCM). The analogy I make is that if you have a section of string players in an orchestra, you wouldn't want each player in the orchestra to have another person next to them who is constantly turning the tuning pegs on each players instrument as they play. That would produce an unnatural sound. But having each players instrument and the strings on each instrument have small tuning offsets (a couple cents) adds this natural phasing motion of sound when combined with many others.
With the Rev2, you get the best of both worlds in my opinion... In its natural state, you have accurate tuning per osc / per voice up and down the keybed, and you can pull off more technical precise sound design. But then with a couple minutes of alterations to the gated sequencer and mod matrix, you can model in per voice / per oscillator character , and even get specific voice counts modeled (ie: 6 voice OBX, 8 voice CS80). And not only can you model in voice modeling, but you have surgical control over how loose each oscillator and each voice is. You can model a MemoryMoog on a good day with beautiful phasing/detuning, or you can model a MemoryMoog at a hot summer festival where a voice or two are significantly out of tune, creating weird semi dissonant phasing in certain patterns. (Note: usually you won't want to model the hot summer festival behavior of classic VCOs, but you can if you want)
https://youtu.be/jB9HG3k3vvQ?t=349