Analogue-sequencer

Hi Colin,

Just a query, seeing that the P3' resolution is 48ppqn,
is a higher resolution possible ?

Best,
A1k

> >Just a query, seeing that the P3' resolution is 48ppqn,
> >is a higher resolution possible ?

> A higher resolution would only have any meaning for a MIDI 
> recorder such 
> as Cubase -  but not a step sequencer which, by its very nature, is 
> quantised.

A much higher resolution with MIDI doesn't really make sense.
A MIDI note message takes roughly 1ms to send.
At 120 bpm, 48 ppqn clock ticks are roughly 10ms apart, so if you send 6
note messages on one clock tick, all on different MIDI channels so each
message takes 1ms, the last of those 6 note messages is going to reach its
destination closer in time to the next tick than the one it supposedly
occured on.
In practice this isn't as bad as it sounds, because humans can't resolve
timing variations much less than 5 ms anyway.
But it means that increasing your timing resolution over 48ppqn is just
going to mean your note messages are more ticks late than they would have
been if you'd had fewer ticks. ;-)

Best regards,
Colin Fraser
Sequentix Music Systems Ltd
http://www.sequentix.com

analog1k wrote:

>Hi Colin,
>
>Just a query, seeing that the P3' resolution is 48ppqn,
>is a higher resolution possible ?
>
>  
>
A higher resolution would only have any meaning for a MIDI recorder such 
as Cubase -  but not a step sequencer which, by its very nature, is 
quantised.

Paul

-- 
---
Paul Nagle / Soft Room Music / Bogus Focus Records / Binar / Headshock / The Joint Intelligence Committee
        www.softroom.co.uk / www.BogusFocus.com / www.JointIntelligenceCommittee.com

>>In practice this isn't as bad as it sounds, because humans can't 
resolve timing variations much less than 5 ms anyway.<<

[flame-helmet  on]

I find myself slightly astonished by this factoid; 

first because the one thing everyone complains about when using midi 
is it's timing; whether or not there is any scientific basis for it, 
midi always feels like it's just about keeping up with the music you 
are trying to make.

secondly, because this is one of those trite aphorisms trotted out by 
the scientific end of the music technology community whenever midi is 
criticised; the truth is that the core midi technology was created to 
a price, & using techniques that were pre-existing. we all know that 
if dave smith & the rest had their time over, they'd have used a much 
higher baud rate & probably cheapernet instead of german hifi plugs.

one hears the same arguments about bit depth & sample rates, when 
plainly the truth about 16/44.1 is that it is the best that could be 
achieved in time for philips'/sony's market deadlines. philips 
originally presented akio morita with a smaller disc containing less 
than an hour of music at 14 bit...
everyone knows vinyl is better. :-)

thirdly, 5mS is an age- easily enough that if you delay the left 
channel by 5mS relative to the right, it will affect the perception 
of direction. it corresponds to a sound being roughly eight feet 
further away from your head. 
this can be a big issue in the context of midi programming, not so 
much in the sense of directional hints but in the whole feel of a 
piece's timing.

[flame helmet off again]


duncan.

Thanks for info...

I always thought 96ppqn was a much tighter timing, it
seems things get sloppy [correct me if I'm wrong] at
higher resolutions?  Besides I never work above 90bpm,
more like between 50 and 80bpm [depending on what you're
actually using the P3 for].

Thanks guys for the input.  ;)

Cheers,
A1k


--- In analogue-sequencer@yahoogroups.com, "Colin f" <colin@c...> 
wrote:
>
>  
> > >Just a query, seeing that the P3' resolution is 48ppqn,
> > >is a higher resolution possible ?
> 
> > A higher resolution would only have any meaning for a MIDI 
> > recorder such 
> > as Cubase -  but not a step sequencer which, by its very nature, 
is 
> > quantised.
> 
> A much higher resolution with MIDI doesn't really make sense.
> A MIDI note message takes roughly 1ms to send.
> At 120 bpm, 48 ppqn clock ticks are roughly 10ms apart, so if you 
send 6
> note messages on one clock tick, all on different MIDI channels so 
each
> message takes 1ms, the last of those 6 note messages is going to 
reach its
> destination closer in time to the next tick than the one it 
supposedly
> occured on.
> In practice this isn't as bad as it sounds, because humans can't 
resolve
> timing variations much less than 5 ms anyway.
> But it means that increasing your timing resolution over 48ppqn is 
just
> going to mean your note messages are more ticks late than they 
would have

> been if you'd had fewer ticks. ;-)
> 
> Best regards,
> Colin Fraser
> Sequentix Music Systems Ltd
> http://www.sequentix.com
>

> first because the one thing everyone complains about when using midi
> is it's timing; whether or not there is any scientific basis for it,
> midi always feels like it's just about keeping up with the music you
> are trying to make.

In my experience, a lot of this comes down to manufacturers' use of  
slow (hence, cheap) microprocessors in their instruments. It used to  
be fashionable to talk about "MIDI latency", claiming that it could  
be several milliseconds, which clearly doesn't make sense since  
there's nowhere for the MIDI bytes to sit for that long. But it's  
entirely possible for processors to introduce delays in dealing with  
MIDI messages.

(Sometimes with good reason: large chords will be staggered since the  
best you can achieve is 2msec per note per cable, which is clearly a  
design drawback in MIDI itself, and there's nothing that the  
originating device can apart from marshall the bytes out one at a time.)

> thirdly, 5mS is an age- easily enough that if you delay the left
> channel by 5mS relative to the right, it will affect the perception
> of direction.

Agreed. Once you get to the 7-8msec range you can "feel" a delay  
between hitting a key on a keyboard and hearing the attack of a note.




   nick rothwell -- composition, systems, performance -- http:// 
www.cassiel.com

> > thirdly, 5mS is an age- easily enough that if you delay the left
> > channel by 5mS relative to the right, it will affect the perception
> > of direction.
> 
> Agreed. Once you get to the 7-8msec range you can "feel" a delay  
> between hitting a key on a keyboard and hearing the attack of a note.

As with many functions of the human brain, the perception of timing of
auditory events is not a simple process.
As Duncan mentioned, the brain uses time difference between left and right,
as well as the filtering effects of head shape, to extract directional
information from the same sound arriving at both ears - but the brain must
still perceive these events as the same sound occuring at one moment.
Ear to ear time delays are tiny - no more than half a millisecond or so, but
the brain also has to deal with multiple-path arrivals of the same sound,
which can be spread over a longer time up to 10 or 20ms.
There are theshold levels around which our perception of what we're hearing
will change, and these are different for different people.
According to papers I've read, and tests I've done myself, time delays
between distinct acoustic events only start to become perceptable around the
5ms mark (if you ignore effects such as comb-filtering).

If you're careful not to load up MIDI with too many simultaneous events, and
prioritise percussive sounds, you can avoid causing much damage to feel.
Personally, I recommend the use of devices optimised for transmission of
monophonic note data from a small number of tracks, say 8 or so...

To come back to the point about timing resolution, if your timing resolution
is already finer than the brain's resolution of the relative timing of
sounds, what would be the advantage in increasing the resolution further ?

Best regards,
Colin Fraser
Sequentix Music Systems Ltd
http://www.sequentix.com

> I always thought 96ppqn was a much tighter timing, it
> seems things get sloppy [correct me if I'm wrong] at
> higher resolutions?  Besides I never work above 90bpm,
> more like between 50 and 80bpm [depending on what you're
> actually using the P3 for].

Accuracy of timing is a separate issue from the resolution.
You can set up a timer to trigger your sequencer code 96 times ppqn, but if
the OS isn't too concerned about giving your code top priority, there could
be quite a degree of variation in when it actually runs.

Best regards,
Colin Fraser
Sequentix Music Systems Ltd
http://www.sequentix.com

>>As with many functions of the human brain, the perception of timing
of auditory events is not a simple process.
As Duncan mentioned, the brain uses time difference between left and
right, as well as the filtering effects of head shape, to extract
directional information from the same sound arriving at both ears -
but the brain must still perceive these events as the same sound
occuring at one moment.....[snip]
According to papers I've read, and tests I've done myself, time delays
between distinct acoustic events only start to become perceptable
around the 5ms mark (if you ignore effects such as comb-filtering).
 
If you're careful not to load up MIDI with too many simultaneous
events, and prioritise percussive sounds, you can avoid causing much
damage to feel.....[snip]
To come back to the point about timing resolution, if your timing
resolution is already finer than the brain's resolution of the
relative timing of sounds, what would be the advantage in increasing
the resolution further ?<<

don't those last two paragraphs contradict each other somewhat? :-)

midi isn't, so far as I am aware, especially optimized so that
percussion parts on channel 10 enjoy priority in the queue of events
to be processed. it should be, but midi just isn't that sophisticated. 

in fact, it hasn't even kept pace with the devices that use it.
(dave smith's own SCI samplers had a midi x2/x3 option so that a pair
of them could exchange samples at higher-than-standard rates using
sample dump protocol. this didn't spread, sadly). 
samplers & synth modules have much faster processors in them now. in
fact, the p3 has also got a faster chip than when I bought my first
one. but we're still stuck with 31.25 kbaud for all 16 channels, & as
much data as you can generate by way of controllers. let's leave sysex
operations out of this for now.
it just isn't fast enough to deliver nuance reliably. it just isn't.
& 1/4" tape at 15ips still out-runs cd or dat.

d.

> don't those last two paragraphs contradict each other somewhat? :-)
 
> midi isn't, so far as I am aware, especially optimized so that
> percussion parts on channel 10 enjoy priority in the queue of events
> to be processed. it should be, but midi just isn't that 
> sophisticated. 

Prioritising percussive sounds involves putting them on lower track numbers.
MIDI channel number doesn't matter.
 
> but we're still stuck with 31.25 kbaud for all 16 channels, & as
> much data as you can generate by way of controllers. let's leave sysex
> operations out of this for now.
> it just isn't fast enough to deliver nuance reliably. it just isn't.

The usual answer to the limited bandwidth is to throw more 'bands' (i.e.
MIDI busses) at it.
But MIDI was pushed into doing a lot more than they planned at first -
sample dumps over a serial line just seem mad to me.
For a handful of notes and controllers, it still does OK, and what's the
alternative ?
Maybe it's a victim of its own success...

> & 1/4" tape at 15ips still out-runs cd or dat.

I think I prefer the ease of handling of CD or DAT to analogue reel to reel.
But you can't beat the smell of decks and vinyl ;-)

Best regards,
Colin Fraser
Sequentix Music Systems Ltd
http://www.sequentix.com