Apple Logic Pro /LogicExpress Discussion

On Jan 24, 2005, at 8:26 PM, GAmoore@... wrote:

> But going back to my original idea which you didn't respond to - if 
> you are right that both professional and amateurs want simple and 
> stable - why does Apple/Emagic keep ignoring 100% of their users and 
> make a prodcut that one is happy with?

It occurs to me that this is perfect opportunity to give a specific 
example of how the pursuit of the computer for the masses concept can 
get all screwed up and end up being what nobody wants.

This is long and complex so if you don't like that don't read it.

It is also information I have gathered through direct correspondences 
with the developers so if you do like that do read it.

Ever since OSX we using native systems have been given the alternative 
of using core audio, core audio, or core audio.

The more limited third party flaky ASIO has been eliminated from OSX 
and apple have decided in the interests of "the people" they would 
establish one generic universal format i.e. core audio.

Core audio is robust and feature laden, it does all kinds of really 
fantastic things, it is multi client, it even allows SR conversion on 
the fly of un-synced digital inputs so that an application running at 
44.1k can for instance accept an input from a device running at 48k.  
With core audio playing quicktime movie audio out the same interface as 
you are playing your DAW as easy as scratching your balls while 
brushing your teeth.

Thus Joe Average is pretty hard pressed to throw something at core 
audio it can't handle.

ASIO can't do anything like that stuff and it is third party and kind 
of flaky to boot, so Core Audio is obviously a leap ahead for "the 
people" right?

Well.....

Something about the architecture of core audio means the IO device 
drivers that the IO manufacturers have to write for it have to have 
something called a safety offset buffer.  This safety offset buffer may 
be anywhere from 12 samples (the lowest I know of) up to 64 samples 
(very worst case).

A 64 sample offset buffer means that 64 samples of latency get added 
each direction in and out so it ends up meaning 128 samples of latency 
get added to the signal path if you are software monitoring say an 
instance of guitar amp in garage band.  Add to that the fact that 
instead of one input and one output buffer like Mac "Classic" and 
Windows versions of Logic had the new improved logic and Garage band 
have an input an output, and a third (mystery) buffer in their data 
path.

So a buffer setting of 64 samples in Logic or Garage band actually 
means that Logic is adding a total of 192 (64X3) samples to the 
monitoring path on top of which a Core Audio hardware driver may 
potentially be adding as much as 128 samples of latency, and those are 
in addition to the 40 or so samples of latency the ADDA will add each 
way for a total of up to 80 samples of latency caused by the ADDA.

So that is

128 safety offset
192 logic IO and "mystery" buffer and
80 samples of ADDA latency for a grand total of:

About 400 samples latency minimum for a guy trying to monitor his 
guitar through Guitar Amp in garage band from his iBook (but in reality 
probably he is not going to get away with setting his buffer that low 
on an iBook and definitely won't if he is using a pluggo or any other 
intersting 3rd party plug).

The same 64 sample setting through the same IO in "Classic"  or LAW 
would give the same guy

128 Logic IO buffer samples and
80 ADDA samples latency for a total of

about 208 samples latency.

What is more the guy will be able to up the buffer on the windows 
machine to 128 and still end up about 334 samples or in other words he 
will still have 46 less samples of monitoring latency than the guy 
running at 64 sample IO buffer setting in OSX.  And since his sequencer 
will not be struggling to work at the low setting he will have much 
more power available for his plugins.

So how does this sucking situation come about.

It is because OSX is considered by the people who made it and are 
continuing to develop it to be a "general purpose" OS.  The reason for 
this is that buy enforcing the principle a general purpose OS they can 
create an OS that is useful to the broadest possible range of customers 
and by selling to a broader market can keep the cost per customer down. 
  They do not want the system to in any way be a special use or 
dedicated purpose OS.  Therefore OSX developers will tell you that 
while they could allow third parties to write a direct access driver 
along the lines of ASIO that would mean relinquishing control over some 
design of some low level code to some outside parties which could 
potentially lead to destabilizing the OS for the rest of the General 
use customers.

So the bottom line is the populist desire to provide the best OS 
possible to the broadest range of customers turns out in this instance 
to be *the exact* reason why OSX is not performing as good in the 
latency dept. for Joe Average DAW user as OS9 did or as windows did.

And monitoring latency is if anything more important to low end DAW 
users than it is to high end users since high end users are more likely 
to have dedicated mixing boards and outboard effects and are less 
likely to be depending on software effects during the tracking stage 
whereas low end guys are more likely to be running mixerless since that 
is the cheaper way to go..