mmontag wrote:
When we talk about FM synth chips, it's my understanding that there is a fixed pool of operators that can be reconfigured in various ways. i.e., the YMF262 has 36 operators that can be configured in various 2-op and 4-op modes.
Actually there’s only one operator which is used 36 times per sample

. But I guess that’s a technicality

.
Other than that I think 6 operators is really not as great as it seems, as the number of modulators put in series increases it becomes very hard to do sensible sound design. More than 4 is rarely used I think, most of the 6-op patches have at least two carriers and can thus also be played with a 4-op or 2-op FM chip by combining channels. Such channel layering is not very practical for a performance instrument, so that’s why 6-op is better for the DX7. But for chips 4-op is just fine.
Because you can’t control the frequency of the operators independently, but rather specify a frequency per channel, the 2-op and 4-op gives you more flexibility than a 6-op chip would, which binds 6 operators to a single channel. You can assign and use the operators more efficiently on the 2-op and 4-op variants. Had the OPNA been 6-op, it would’ve had only 4 channels rather than 6, definitely a bad trade-off.
The bigger problem with the YMF262 OPL3 is not that it’s got “only” 4 operators, but that the selection of four 4-op algorithms (configurations) is small and also not great, when compared to the OPM / OPN series which were natively 4-op and offered 8 much more varied algorithms. The strength of the OPL3 on the other hand is the massive amount of channels, and the ability to select 7 more waveforms in addition to sines, giving it a whole different dimension of sound design.