Elektron Musical Instruments

i thought i'd put claims about the stability of the
machinedrum's clock stability to the test.  jump down to the
"findings" section if the details bore you.

environment:

1. MD connected directly to a motu micro lite USB, with no
other devices attached.

2. the micro lite is connected to a mac g4 running os X
panther on a sonnet PCI card.  the keyboard and mouse are on
the motherboard's bus.

3. the mac runs snoize's "MIDI monitor" application.

known issues:

1. MIDI over USB is purportedly not jitter-free.  there is no
published spec for the MIDI interface's jitter that i know
of, so it's something of an unknown quantity here

2. the clock in MIDI monitor may not be consistent, either,
but i hope it is.  at any rate, it times to the millisecond.

actions:

1. fire up MIDI monitor to listen for clock messages from
the MD's port

2. enable clock output on the MD

3. go and pour myself a beer.  samuel adams' imperial
pilsner is surprisingly great, if you like really hoppy
beers.

4. come back in about five minutes, stop the MD clock, and
export the data to a text file.

5. write a shell script that will output the difference
of a timestamp and the timestamp immediately preceding it
{starting with a value of zero to account for line #1} and
cat that to another file.

6. ask herself to make a pretty graph with the numbers in
the new file {i don't have excel}.

findings:

in a sample of 1001 ticks, we see 

797 ticks with 21ms between them
191 ticks with 22ms between them
13 ticks with 20 ms between them

a graph illustrating this can be seen here:

http://syncretism.net/img/drumclock.png

what does this tell us?

i'm not a statistics guy, and don't want to put any thought
into measuring jitter {deviation}, here.  this strikes me as
reasonably accurate and precise over time, and i have no way
of discerning whether the jitter is the product of the MD,
the interface, the system clock, or all three, anyway.  can
the human ear hear this?

i know that there are some tech-types among you.  was there
a better way to do this?  if so, please advise.

-- 
yours,
niall.
.. .  .   .    .     .       .           .             .                 .
aleph null.                             a simple insinuation around silence.
http://syncretism.net
.. .. gpg public key - http://www.aleph-null.net/niall.gpg .. ..

niall,
i'd be curious how this compares to other clocks,
particularly of popular software sequencers, namely
Cubase SX 3  ;0
doug


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com

niall,
i'd be curious how this compares to other clocks,
particularly of popular software sequencers, namely
Cubase SX 3  ;0
doug


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com

Very cool report and graph, too. At which tempo was the MD for the test?

> known issues:
> 
> 1. MIDI over USB is purportedly not jitter-free.  there is no
> published spec for the MIDI interface's jitter that i know
> of, so it's something of an unknown quantity here

Not sure what you mean by jitter in this sense. FWIW the tolerance for 
midi timing depends on the sampling rate of the MIDI wire itself, which 
ends up being about 1%. That is, one device (the sender) can have an 
internal electrical timing that is about 1% off the other (the receiver) 
and you'll still be able to clearly tell what is coming in over the 
wire. (Online there seems to be a good deal of misundertanding about the 
baud rate for MIDI. IIRC it is 32 kilobaud and not 32,000 baud. Or maybe 
I got that backwards heh.) I guess maybe you're talking about things 
getting bumped over in little increments to suit the timing of USB? We 
could get all mathematical, but imho it can be considered negligible here.

> 2. the clock in MIDI monitor may not be consistent, either,
> but i hope it is.  at any rate, it times to the millisecond.

It's good that you used a Mac, because the resolution of the readily 
available timers in Windows suck. Something like 10ms resolution has 
been my experience. I don't remember if MIDI-OX or the like improve on 
that by using a timer that's deeper in the matrix; too lazy to check.

> 3. go and pour myself a beer.  samuel adams' imperial
> pilsner is surprisingly great, if you like really hoppy
> beers.

I'm gonna have to try that one. ChrisM, a member here on the list, 
turned me on to India pale ales and I especially like Pyramid's IPA when 
I'm in the mood for 'top of the hops'. Don't let him convince you to try 
"Golden Monkey", it will be your downfall.

> 797 ticks with 21ms between them
> 191 ticks with 22ms between them
> 13 ticks with 20 ms between them

Imho this is "good enough timing" and certainly better than alot of 
things I've seen. If you wanted to get really serious about seeing what 
the timing skew is, you could wire up a dedicated microprocessor with a 
UART module in it and log the exact times between those 0xF8 ticks 
(there are six of those per 16th-note) at a resolution much, much 
smaller than 1ms. Might be a righteous hack on a rainy day.

I think for the best timer, a manufacturer would want the MIDI ticks to 
be run by an on-chip timer that (highest-priority) interrupts everything 
in the instrument's CPU in order to get that 0xF8 sent off down the 
wire, and use a really tight crystal to drive the CPU's internal clock. 
The MD is really complex, so who knows how they are issuing clock 
messages, but I think your results show that it is performing at a 
professionally acceptable level.

In the words of Daniel Hansson, Elektron, "Dump on!"

,
Mark

o.s. upgrade 

uw 1.3b

mono 1.06b


psi

<->

--- niall munnelly <aleph@...> schrieb:


---------------------------------
i thought i'd put claims about the stability of the
machinedrum's clock stability to the test.  jump down
to the
"findings" section if the details bore you.

environment:

1. MD connected directly to a motu micro lite USB,
with no
other devices attached.
.....


This test shows nothing... You cant measure with
something that is worse jitterwise than the objekt you
want to measure.


	

	
		
___________________________________________________________ 
Telefonate ohne weitere Kosten vom PC zum PC: http://messenger.yahoo.de

On Sun, Feb 05, 2006 at 01:08:25AM +0100, Silvia Tubig wrote:
>
> 
> This test shows nothing... You cant measure with
> something that is worse jitterwise than the objekt you
> want to measure.

assuming that the jitter of the USB device is worse is
question-begging, unless you're already privvy to the spec
of the MD and the interface.  if that's the case, please
share.

given the uniformity of results recorded with and without
sequences playing, it shows that the clock is "good", but, because of
the limitations inherent to the test, can never definitively describe how good it is.  i think most people get that.

the rest has to be done with the ears.  if nothing else, it
lends some credence to the precision of the MD's clock,
something echoed as dogma on this list with very little
quantification {precise or not}.

-- 
yours,
niall.
.. .  .   .    .     .       .           .             .                 .
aleph null.                             a simple insinuation around silence.
http://syncretism.net
.. .. gpg public key - http://www.aleph-null.net/niall.gpg .. ..

Ist die machinedrum langweilig oder nicht? :-)


	

	
		
___________________________________________________________ 
Telefonate ohne weitere Kosten vom PC zum PC: http://messenger.yahoo.de

langweilig? haha. nein, nicht langweilig, es macht spass!