Various things

Hi there,

Just to keep up date with a few things.

I've been testing the pitch accuracy of the FM7 and
the relative pitch was better than a hundredth of a cent
for a major chord. I thought it was probably theoretically
perfect (up to the limit of normal computer floating poitn
work) and Michael Kurtz said that indeed it probably is
within the limits of human hearing anyway - programmed
to a little better than 24 bit precision.

Another bit of news, another Native Instruments soft synth will
receive the full range of midi tunign standard sysexes when
it is released like the FM7, the Pro-53
http://www.harmony-central.com/Newp/2002/Pro-53.html

Maybe this is the start of a new trend?

Another thing, Michael Kurtz tells me that
Sysexes work with VST plug ins in Cubase 5.1
but not in CubaseSX, and Native Instruments will be
pushing Steinberg to do somethign about it - would
probably do no harm to contact them too if you use
CubaseSX with VST plug ins and want them to
have microtuning capabilities.

I've also been testing the speed of midi relaying
and was surprised to find that Midi Yoke Junction
outperforms a hardware loopback - connecting the
midi out of my soundcard round to the midi in and
then to the midi in of another program.
It's ten to fifty times faster. Midi Yoke can
send a bulk tuning dump round the loopback
in on average a third of a millisecond on my machine,
(which means that most of them go all the way round
the loop in a single Windows tick as the multimedia
timer only has a resolution of one tick per millisecond).
When it goes via the hardware loopback it takes on average
180 milliseconds.

For note on /off pairs the difference isn't quite so dramatic
but Midi Yoke takes about the same time to relay a note on /off
pair as a bulk tuning dump - just a little faster.

Hubi's loopback is between the two, and about as fast as
Midi Yoke but has a cut off at 64 notes for a fast cluster
of note ons - discards the rest.

I also found in the case of the physical loopback, some of the bulk
tuning dumps got lost if you sent a them at great speed one
after another through the physical cable - of 20 bulk dumps
sent in a single batch, 4 didn't make it round the loop
- on my machine anyway.

Next upload of FTS will have an option to test the speed
of your loopback, and that's what I used.

Here are the details:

Midi Yoke: (10000 events)

Note on /off pair: 0.27 milliseconds
SNR Sysex: 0.3 milliseconds
Bulk tuning dump: 0.38 milliseconds

Max delays for note on / SNR sysex pairs:
SNR sysex 8 milliseconds. Note on: 6 milliseconds.

(note that the 8 millisecond delay here is very rare
- in a 1000 events test I got a max delay of 3 milliseconds)

Physical cable: (1000 events, 20 for the bulk tuning dump)

Note on /off pair: 1.916 milliseconds
SNR Sysex: 2.81485 milliseconds
- 44 out of 1000 didn't make it round the loop
Bulk tuning dump: 181.375 milliseconds
- 4 out of 20 didn't make it round the loop

Max delays for note on / SNR sysex pairs - 100 of each:
SNR sysex 8 milliseconds. Note on: 6 milliseconds.

Hubi's loopback: (100 events, 20 for the bulk tuning dump, )

Note on /off pair: 0.21875 milliseconds
- first 64 out of 100 make it round the loop
SNR Sysex: 0.32 milliseconds
Bulk tuning dump: 3.55 milliseconds

Max delays for note on / SNR sysex pairs:
SNR sysex 3 milliseconds. Note on: 1 millisecond.
- again, first 64 out of the 1000 made it round the loop

For the pitch measurements:

440 (desired) -> 439.9214 (wave counting), 440.0848 (FFT)

550 (desired) -> 549.9039 (wave counting), 549.7696 (FFT)

The wave counting method is more accurate, so it makes
sense to use those.

Using those measurements, the ratio is
386.3205 cents
- should be
386.3137 cents
so relative pitch error is less than a hundredth of a cent.

Absolute pitch is 0.309289 cents flat relative to A = 440,
but absolute pitch is harder to measure and especially to
check for any systematic errors so I'm not sure about that
and as far as the measurements are concerned, and would
probably need to do more tests to find out what is happening.
It could well be perfect too (assuming they target A = 440
that is)

Will be releasing FTS 1.09 with the next upload, prob.
later today. Will still continue to update it once
released and particularly there are sure to be
some gaps in the help, and I'm hoping that
I can get a few reviews from the download sites
as I did last time, and the reviewers comments
may be helpful, and could be good to have while
I'm finishing the help.

Then I'll be working on the Virtual flowers program for
a while, of course not abandoning FTS or anything,
but maybe work on it somewhat less for a while, also
spend some time enjoying using it and maybe have
a bit of a composing break as I did last year
I think about this time.

Robert