VOGONS


First post, by dnewhous

User metadata
Rank Member
Rank
Member

Now that the reverse engineering has been done, what exactly is LA synthesis? I read something on the web from a musician who reversed engineered a D-10 (just to figure out what it did) and said the LA synthesis doesn't work like any of the publications he had read said it did. And it bothers my intellect that I just don't get it either. I understand the principles of FM synthesis and wavetable synthesis but I can't form a coherent explanation of LA synthesis in my head.

Reply 1 of 4, by canadacow

User metadata
Rank Member
Rank
Member

LA Synthesis is half analogue synthesis (subtractive synthesis using sawtooth and square waves) and half PCM synthesis. Basically, the MT-32 has the capacity to play 32 sound elements (Roland called them partials, though this is confusing terminology as partial means something else in DSP as well). Each element could be a PCM sample or an analogue waveform. Up to four elements could make up an instrument (though you were free to use from one to four elements) Each element had its own volume and pitch envelope, and for the analogue parts, a filter envelope. In addtion to the envelopes, each element also had a low frequency pitch osscilator. By adjusting the element parameters and the mixing methods, one could patch together instruments that sounded better than either the pure PCM synthesis or analogue synthesis of the day.

Comparing LA synthesis to modern wavetable synthesis we find several notable differences. First, the MT-32 can't rely completely on its PCM samples as they are incomplete snapshots of instruments. Wavetable synthesis normally has multiple frequency samples of an instrument's attack, loop and decay.

Comparing the MT-32 to FM synthesis, we find FM synthesis rather primitive compared to LA synth. FM synthesis is considered additive synthesis where waveforms (operators) are added together to make a complex, rich sound. In the MT-32's case, the opposite occurs. The MT-32 takes a signal rich sound (square or sawtooth) and applies a lowpass filter to extract the frequencies it needs.

I hope this clarifies it for you. If you're curious about another aspect of LA synthesis just ask and I'll elaborate on what I've learned from this project.

Reply 4 of 4, by alex_mauer

User metadata
Rank Newbie
Rank
Newbie

im not sure how MUNT emulates LA synthesis... but i dont think LA uses real subtractive to generate its sound - otherwise you could filter the PCM - from several sources i have determined that LA must be additive synthesis w/ a subtractive interface

ftp://ftp.roland.co.uk/productsupport/Manuals/D-110_OM.pdf

on page 91 of the roland d-110 manual there are illustrations that lead me to believe the waveforms are generated w/ additive harmonics... not raw waveforms run thru filters - the filter effects and pulse widths could be generated w/ math... not actual filters - thats my guess

http://homepage.mac.com/synth_seal/html/mt32b.html
http://homepage.mac.com/synth_seal/images/lasynth.gif

here's a diagram and explaination of how the filter is not real... nor is the PWM - the way the low pass filter could work... the higher frequency harmonics would decrease in volume... and the resonance would be an emphasis/volume boost at the "cut off point" - no self oscillation here - the PWM may be generated by a formula that changes the "symmetry" of the waveform... or maybe a shift in phase of certain harmonics - PWM works on saw as well as square

i dont have any concrete proof or statement by roland on how LA synthesis works... but i do think my theories are atleast similar to what is correct