Greetings:
Last night I decided to test TiMidity against Fluidsynth. The results
were interesting.
Here's the test scenario :
run Sequencer Plus Gold in dosemu
route MIDI output to ALSA virmidi module
start TiMidity with 'timidity -iA -B2,8 -Os -c timidity-sf2.cfg
-EFreverb=0 -EFchorus=0'
start JACK (qjackctl) then start qsynth
connect virmidi to either TiMidity or Fluidsynth (qsynth) via
kaconnect
Both synths used the same soundfont (FluidR3). The output from qsynth
was subjectively more to my liking, but I had a persistent problem with
noise, as though periodically the synth output would "fall off the
rails", resulting in severe aliasing. It eventually corrects itself but
until I resolve this problem I would not use qsynth for a production
run. OTOH TiMidity just works fine. The only problem I have with it is
that I'm now rather spoiled by what I take to be the better filters of
Fluidsynth. TiMidity's output was comparatively harsh, but of course I
can modify it with outboard gear (or a LADSPA plugin ?) if necessary.
I tested the synths separately and in unison. No radical timing errors
occurred and intonation was fine. I'll test Fluidsynth without the GUI
today to determine whether the qsynth GUI contributes to the audio
breakup. Meanwhile I welcome any comments or suggestions regarding this
test.
Oh yeh: I'm running a PlanetCCRMA kernel 2.4.22 patched for
low-latency, ALSA 0.9.6, TiMidity 2.11.3, and qsynth 0.0.2 (fluidsynth
1.0.3). JACK version is 0.91.1 and qjackctl is version 0.1.3.
Best regards,
== dp
