2000-12-01 by Daniel G.
> Why do you want it in one unit? Won't two do? I think you can go
Well, cost is one factor. HD recorders no doubt have some very complex
software running them already -- it would be cheaper to add a midi out
port and arrangement functions to an existing OS, than producing two
separate products, I would imagine. That is, assuming the physical
user interface was there already (arrow keys and buttons for
maneuvering through menus), the marginal increase in cost for adding a
sequencer (in software, plus the serial chips for midi out ports)
would be less than the cost of an entirely new group of engineers to
design a separate product, and this I would hope could be reflected in
the cost. But as of yet I don't think any company has found that it
would be a seller. If someone is serious and can afford an HD
recorder, it seems a little silly that they wouldn't be using a
computer already, but that's precisely what I'd like to avoid.
Timing is another factor. I am probably just being anal here, but I
would very much like to have a system where the midi and audio are
synchronized on a hardware level. As far as I know this is not done
with any system currently around. While midi ports are usually on the
same card as the AD/DA, they have no internal timing and rely on the
operating system (win9x or macos etc) for the timer tics. But this
isn't good because while audio in and out can usually be
synchronized (the OS tells the card both "play now" and "record now"
within microseconds, and only has to swap out the full buffers every X
seconds [where X = buffersize/(sample_rate * sample_size * channels)],
midi is a more troublesome issue because system timers usually have
poor resolution and, in non-realtime operating systems, aren't
guaranteed to be on time. (with most *nix systems, the timer
resolution available to apps is 10ms, and I believe it's the same in
win95 and 98, although I've been vaguely told that 16-bit apps can get
better timing performance since they don't have to share things in the
same kind of way. I'm not sure about macOS performance) So a little
box dedicated to recording and sequencing would, I hope, perform much
better in these areas.
I'm in win98 now, and if I create 4 beats of straight 32nd notes in
<sequencer name omitted because it's a big name, but crap, and I've
no intention of paying for it> at 120bpm, I can hear the timing
irregularities easily. This is pretty much what it sounds like when
the OS has an accuracy of about 10ms. I know the difference because I
wrote my own sequencer for linux sometime back, and I finished up a
song and lo and behold, my 32nd hihat roll sounded like crapola. I
wrote up a quick midi-only driver in RTLinux (realtime linux -- an
extention to linux with precise timing), and the same section sounded
fine. A lot of people blame midi for these sorts of things -- the
artist 'BT' is one of them -- but they don't understand that it's the
underlying software that's the problem, not the protocol itself. Midi
is @#%!@$ wonderful, in my opinion.
So that's why I want it all in one machine (but not a computer)
although technically a two-unit setup might not be too bad -- if only
too expensive.
> Hmm, that could be harder than you think. It's possible, sure, but
it's also
> time-consuming and it might get into the way of creativity. And I
thought
> you wanted a >8 track recorder?
Well, the idea is to sketch out a song with a full 16 midi channels (I
really love the XG variations for this purpose), and then, once the
rhythms and things sound nice, go back and record each track with a
nice synth. I want to basically compose a whole song with some crap XG
sounds, and then do the final rendering with the plg150an for every
instrument. Kind of a poor man's analog symphony :) So I want 8 audio
tracks, yeah, but just for playback. The actual recording only needs
to be stereo.
> By the way, you can always buy ProTools. Or perhaps you can't, like
Heh.. Well if the "professional" software available for the mortals'
market is any indication of what to expect, I'd probably not like
protools very much. I find the user interfaces for cubase/cake/logic
et al to be pretty much revolting.
Anyhow, the solution I've come up with is, well, "DIY". I've found the
hardware-sync redemption in the form of the Turtle Beach Tahiti sound
card, which has an on-board motorola dsp56001 chip which can be
programmed to one's liking. Currently I am going through the 700(!)
page manual for this beast so as to learn how to program the ports to
which the AD/DA and midi are attached. I'm going to simplify the
existing firmware provided by TB in that I only need 44100/16
sampling, and that I'm going to make sure it's all synchronized
on-board (so as to avoid sending separate play and record and
midi-start requests to the card) Midi timing will also be handled on
the card -- the app will just send delta-prefixed midi events, and the
card will wait for the delta to expire before sending the event out
the port. That way I can use the card on a non-realtime system and
still have accurate timing. Why this approach has never been tried
before is beyond me (there will be a slight pause before playback
begins, during which the sequencer will have to preload say, 0.5 to
1.0 seconds of audio and midi, thus ensuring that it is always a
little bit ahead, but it's certainly not unbearable.) I believe Emagic
now sells some external boxes which perform this function -- hardware
midi timing -- they call it 'AMT' or somesuch.
As for the sequencer, I'm going to have to write that too, since my
last one.. well, sucks. But if anyone can spare me from this evil
effort and let me know of any hardware seq+hdrecs, please help!
Programming is neat and all, but I'd like to get to making music
already.
-dan
