On Tue, 04 Jun 2024 13:53:07 -0400, Mike S. <
Mike_S@nowhere.com>
wrote:
On Mon, 03 Jun 2024 19:24:49 -0400, Rin Stowleigh
<rstowleigh@x-nospam-x.com> wrote:
>
Yes, it can give you that "90% accurate" return to nostalgia as you
hear / remember sounds that are familiar and relatable... and that can
be quite cool.
>
I'd say that Roland MT-32 Munt emulation is at about 90% accurate at
this point. I may switch to emulation if my real MT-32 ever fails
which it likely will considering its age.
>
I know Munt isn't 100% accurate as I can hear (don't hear?) dropped
notes and wrong sounding instruments occasionally when Munt is used.
Looking this part up on Wikipedia "Given the MT-32 was intended to be
a relatively low-cost prosumer product, many corners were cut in the
design of its DAC output. For example, the circuitry needed to
properly calibrate the DACs was omitted, resulting in distortion of
the analog signal", I have to assume that emulation for hardware with
a crappy DAC is probably closer to the original than most software is
compared to high end audio gear.
Softsynths in pro-grade studios are a funny thing. Sometimes they can
really wow you with the sound, until you hear them side by side with
the real thing -- the hardware they're trying to emulate. And then
it's like the software can be 90-95% there, but its that last 2%-5%
that makes all the difference in the world in terms of inspiration
(which is hugely important in writing or playing music, but not so
much for just listening to music in-game for example).
Also just the signal path alone often makes a difference. One synth I
own (but admittedly don't use often) is a Roland System-8. It is a
virtual analog that you can data dump emulations of many great
sounding vintage synths (Jupiter 8, Jupiter 4, Juno series, JX3P etc)
into the hardware, effectively swapping out the synth engine. But
unsurprisingly some of the engines (called "plug-outs") sound
substantially better when run on the hardware, I suppose because the
signal is coming out of the DAC and then into the pre-amps of a
high-end audio interface (that alone can make a big difference). But
with some of them I actually hear subtle tuning differences in the
oscillators that just sound better on the hardware than when they are
run as "plug in" versions in the DAW, and the signal path doesn't
really impact tuning per se. So I'm not sure if that's because the
synth is taking advantage of more available processing power on the
synth than the computer* or if that's just an intentional strategic
decision by Roland.
* not implying here that the quad DSP system in the synth is really
faster, per se, than modern CPUs... it's just that they are dedicated
to audio and not trying to run a generic operating system that's
juggling a gazillion other things.
As far as dropped notes, wrong sounding instruments etc. I have no
experience with Munt and never owned an MT-32 so it's hard to
speculate what the issues could be there. Dropped notes can be that
polyphony limit is exceeded (the MT-32 advertised 32 notes but I
believe that was a maximum polyphony, which is typicaly of some Roland
products.. more complex patches can reduce polyphony count). It
could also be that the developer(s) of Munt just implemented their own
method of voice stealing to make it more CPU efficient or to overcome
other problems.