I don’t believe that. One osc can be enough for a good pad.
So, I bounced some comparisons down to get to the bottom of things.
In each clip below it’s an A minor chord playing one triangle oscillator. First one is the A4, then a generated Triangle wave, and finally the Jupiter 8 sample from Omnisphere. I did my absolute best to normalise levels using a meter.
First clip is with all filters disengaged. On the A4 that is F2 on the default setting and F1 fully open and resonance turned off. I know res 25 is supposed to be flattest, but to my ears you get more body with no res at all.
I looked at the spectrum for this and, contrary to what I thought, there is approx. the same low mid content for all of them. The A4 is buzzier and the JP8 is more saw-like, but in terms of “body”, there isn’t much difference to my ears.
Next I tried to test the filter. On ther A4 I set F1 to 64, again with no res. I used Omni’s most bog standard LP12 filter to avoid any enhancements, with no resonance, and again I did my best with the spectrum analyser to match cutoff values. I admit this is really not easy but I think you’ll agree that, tonally, they are close enough.
Frequency-wise this time there was slightly more thickness on the JP8 in the low mids, but again it’s a small difference.
The most notable thing is that the level of the A4 completely fell off a cliff. It will be more apparent if you loop the clip because it starts with the A4. Again, I haven’t changed anything here except the Filter cutoff.
This is kind of what I’ve been getting at from the start, but based on this unscientific test, the (triangle) oscillator isn’t at fault, it’s Filter 1. I don’t know enough about this stuff in truth but a simple LP12 filter set to 50% cutoff should not behave like this.