Then we basically agree. I only wanted some clarity on the key question that gets tossed around the forums without anyone actually reaching a conclusion. It's not a question of personal opinions, but of the best explanations as to why analog sound has the characteristics that it does.
The first thing is variance. If you take an analog synth with 6 voices, each with two oscillators then you have 12 oscillators. Each of these oscillators is made up of certain components, as an example lets take a resistor. Resistors are graded at tolerance levels, ranging from 0.05% to 20%. So the absolute closest match we could hope for is 0.05%, each oscillator is going to be using slightly different resistors. The same applies for lots of other components we use as well. So each oscillator is going to be different to the others in various ways as various components are basically "different", so it will sound slightly different.
Now in a digital synth it is pretty easy to knock up some basic oscillator code, now if I want 12 oscillators I run the oscillator code 12 times, once for each oscillator and every oscillator sounds the same. Aliasing nowadays also should not be a problem there are solutions for this including FM and sync.
If you play a 4 note chord on the two machines above the analog one is going to sound fuller and a bit more alive as the oscillators are not perfect while the digital ones are.
Nowadays though digital synths don't tend to take this naive approach to oscillators, they build in the variance to each oscillator so it acts differently to the others, on a digital synth like this then the oscillators sound analog.
The same sort of thing needs to be done for things like VCAs, you need to model a VCA and its distortion not just multiply samples like digital synths used to do.
The next major issue is filters, filters use feedback in the analog world this is pretty instant but in the digital would it isn't, it is running at the sample rate say 44.1Khz, so a feedback is going to take around 22.7 microseconds. Although you are not going to hear this directly it messes around with the phase response and frequency response of the filter. In general this means that the cutoff frequency is not accurate and also the resonance and frequency effect each other rather than being largely independent as in analog.
You can minimise this in digital systems by oversampling to reduce the feedback time but of course this uses more processing cycle, nowadays filters are being used (say in Monark/Diva etc) that are called Zero Delay filters that remove this feedback problem.
So if we compare an analog synth to a naive digital synth (all oscs the same, non modelled VCAs, feedback delay filters) then the analog synth is going to sound nicer, it has movement from its variance and it's filter is acting linearly across the frequency range.
Now if we compare it to a digital synth with variance, VCA modelling and zero delay filters you are going to find it very hard to tell the difference.
There is no point comparing analog synths to bad digital synths, you need to compare them to good digital synths.