by Roger Nichols 4-1-92
The grass is always greener on the other side of the fence. The girl is
always cuter on the other side of the room. The snare drum on the other
guy's record always sounds better than the one you just recorded. The never
ending quest for that extra ten percent.
Drums and other percussion sounds add to the basic feel of a song. The fundamental
rhythm on which all of the other instruments rely. The three important factors
here are the pattern, placement (in the pattern) and sound of the percussion
instrument. All three factors are inter-related to the point that any change
in one greatly effects the others. A perfect scenario for endlessly chasing
your tail.
In the MIDI studio we are blessed (or cursed) with an unlimited number of
percussion sounds available to us. If we don't have just the right one,
we can record it ourselves and add it to the collection. I have 400 megabytes
of edited drum samples and at least four hours of DAT tapes that haven't
been edited yet. It now takes longer to audition all of the drum permutations
than it does to do everything else combined.
Usually when you start to write a song in your MIDI studio, you use a drum
set that already exists or create one that has only the bare essentials
to get you started. The initial groove seems fine. As you progress, you
keep telling yourself "This is OK for now, but I'm gonna fix that snare
and kick as soon as I get a chance." Well, you've put it off as long
as possible. It's time to face the music (pun intended).
If you are still working in the sequencer domain, you have an advantage;
all you have to do is substitute different sounds while the sequence is
playing, almost. Since we know that the sound has a lot to do with the feel
of the pattern, we have to choose a drum sound that appears to come in the
same place rhythmically as the drum we are replacing. You can put two sounds
side by side on an editor screen and they look like the peaks are in the
same place, but when you listen to them it is a whole other story.
One of the snare drums could sound early because of it's length. Where the
sound ends makes you think that it starts in a different place. If the sound
is too short it tends to sound early. If the snares ring longer, it makes
it sound as though the snare is more laid back. This effect is more pronounced
with sounds that are processed. Processed snares and other drum sounds use
the decaying part of the sound to accentuate the effect. This makes the
timing of the sound's decay as important as the sound's attack.
Placement of the new sound is critical. If you want to make the sound happen
earlier, you can make the sequencer shift it ahead a slight amount. This
amount usually has the resolution of the smallest increment available to
that particular sequencer. If the resolution of the sequencer is 480 ticks
per quarter note, then moving the snare one tick earlier at a tempo of 120
beats per minute will advance the snare by 1.04 milliseconds. This is of
course the best case scenario. Most sequencing software won't have this
high resolution and a lot of dedicated hardware sequencers don't allow adjustments
of one tick. Moving the snare one increment earlier on an Akai MPC-60 will
make the snare about 13 milliseconds ahead of where it was.
The way Donald Fagen works is to run the audio from each drum and sequenced
instrument through individual digital delays. He can then move any instrument
by any amount. Most of the delays he uses are made by Roland and allow increments
of .1 milliseconds up to 10 milliseconds where the resolution drops to one
millisecond increments. If he wants to have available .1 millisecond increments
above 10 milliseconds then he puts two delays in series. There are delays
available that allow increments of around 20 microseconds (which is 1/50th
of a millisecond). Donald has a stack of these as well.
I am not saying that these are necessary devices for everyone. If you put
together a sequence and everything sounds right, then that is just fine.
If you bust your hump trying to get a machine groove to feel good and you
aren't having much success, then try moving a few things around until it
locks into the groove you're looking for. Put every instrument through it's
own delay and crank them all to 15 milliseconds. Now you can make some sounds
earlier and others later.
The biggest problem with using all of these delays is that they do degrade
the quality of the sound passing through them. They also add little whirs
and wheezes of their own to the sound. If the sound the delays add bothers
you then there is a solution.
When you are ready to print the sequence on multi-track tape, stripe SMPTE
on the tape and then use the SMPTE to drive the sequencer. Make sure you
set the start offset so that the sequence starts where you want. Never start
the sequence at 0:00:00:00. Always let at least 10 or 15 seconds of SMPTE
go by before the sequence starts.
Now, let's pretend for this example that we have a 5.7 millisecond delay
on the snare drum and a 2.1 millisecond delay on the kick drum. Take the
snare drum audio out of the delay and run it as you normally would to the
tape machine. Insert the delay on the SMPTE signal between the tape machine
and the SMPTE reading device that is driving the sequencer. Make sure that
the SMPTE is not clipping the delay. About half way up the meter should
be fine. Make sure that there is no regeneration or any other effects active
on the delay, we want straight, no frills delay. In the case of our snare
drum, make sure that the delay is set at 5.7 milliseconds. Start the tape,
punch up record on the snare track and let the sequence print the snare
drum. If you have other instruments that were set for the same amount of
delay, you can print them too. After the snare is printed, do the same thing
with the kick drum, but change the SMPTE delay to, in our example, 2.1 milliseconds.
Continue with this method until all of the instruments have been printed
to tape.
See how easy this is? Maybe some day we will have Super MIDI and sequencers
with high enough resolution so that we won't have to do it this way, but
let me tell you, we've come a long way since 1978. There were no sequencers
to speak of in 1978, especially universal ones that you could use with any
synthesizer. Everything we sequenced back then was based on the timing of
the 1/8th note click on tape. I guess the resolution was two ticks per 1/4
note. We had to build our own delays to get small increments. It worked,
sort of.
Replacing Snare Drums For Fun And Profit.
OK, sports fans, what if you already put everything on tape and your spouse
got your sequencer in the palimony settlement and you need to replace that
ugly snare drum with one that doesn't rip the top of your head off every
time you play it back? Well, that's just what were going to cover next.
There are basically two ways to replace or add to an existing snare drum:
1. Build a tempo map and play it back to trigger the new snare.
2. Trigger the new snare in real time from the existing snare.
Right here, let me say that if the drums on tape were played by a real live
human person type drummer, I would prefer to add the sound of the new snare
to the sound of the existing snare rather than replace it entirely. This
greatly improves the chances of the finished track sounding like a person
played it.
Tempo Maps.
If the snare drum you are replacing plays a regular pattern, such as two
and four, all the way through the song, then you can use a wide variety
of devices to create your tempo map. These include, but are not necessarily
limited to, the Roland SBX-80, the Roland SBX-1000, the Aphex Studio Clock,
Syncman Pro and Opcode's Studio 5. Each unit operates a little differently,
but basically they read SMPTE to figure out where they are and you feed
them audio from the snare drum so that they can determine where the beat
fell. Some units have the ability to mask out unwanted sounds that occur
between the snare beats. If the unit that you are using does not have this
capability, then you may have to gate the snare drum or otherwise process
it to mask out the junk.
Next, write a simple pattern into your sequencer that will play the new
snare drum on every quarter note. This is because the tempo mapper thought
it was hearing quarter notes when the snare was fed to it and that the first
snare that it heard was the downbeat of the song. Make sure that it is quantized
to play exactly on the beats. Hook up the device used to make the tempo
map so that it reads SMPTE from the tape machine and sends MIDI song pointer
to your sequencer. Set the sequencer to "External MIDI" and play
the tape.
In reality what is happening is that the sequence is being played back at
a variable tempo. Let's say that the tempo of the original song was 120
beats per minute. The tempo displayed by the sequencer when playing back
should read 60 beats per minute (because of the half note snare part). When
we get to a bar where the snare drum was rushed, the tempo will jump up
to 60.13 bpm or 61 bpm. If the next snare beat was right on the money, the
tempo will drop down to 59 bpm so that the average comes out to be 60 bpm
for the duration of the song.
Now that we have the new snare playing along with the old snare, we have
to determine if the new snare is in the right place. If you listen to both
snares at once, you like where the new one lands and don't hear any flams,
then cool. Print it. If you are not exactly sure whether the new one is
right, then you can try a few different things to see how they line up.
One method is to make the new and old snare drum the same apparent level.
It doesn't matter what the meters say, just make them sound like they are
both at the same volume. Now pan them to opposite speakers. If the snare
appears to come from the right speaker, then the snare panned to the right
is earlier than the one panned to the left. If the image seems to come from
between the speakers, then they are probably happening at the same time.
I have seen a few people use the Russian-Dragon to determine which snare
is early. The only problem is that it is not calibrated so that you can
tell how early or late one input is to the other. The Bee Gees used to record
the new snare on the multi-track tape and then by scrubbing the tape across
the heads (like finding an edit point) mark both snare drums and measure
the difference with a ruler. At 30 inches per second, 1/4 inch would be
about seven milliseconds. I use a dual trace oscilloscope to see the difference
between the events.
If the new snare is late, because of MIDI slop or the difference in sound
quality or whatever, the same methods used earlier to shift sounds around
can be employed. Shift it earlier either by the sequencer or the offset
in the tempo mapping device and then run it through a delay to line it up.
Then move the delay over to the SMPTE line and print it to tape.
If the snare part is not as straight forward as our first example, we can
resort to another method of mapping where the snares lie (where the snare
lies? If only chickens lay, should we call it lieback instead of layback?
Never mind.)
There are a few devices around that will produce a MIDI event when given
an audio trigger. The ones that I have used over the years are the Simmons
MTM (MIDI-Trigger-MIDI), the Akai ME-35T and most recently the Alesis D4
drum module. Some of them have internal processing to mask out the junk,
while others make you do it yourself. Whenever you get the masking straightened
out so that as few stray triggers as possible exist, play the tape, record
the MIDI events into your sequencer and delete any of the bogus events from
the sequence. An added advantage to using this method of mapping the snares
is that you can record the dynamics of the live snare. The MIDI devices
I just mentioned follow dynamics rather well.
All audio trigger to MIDI devices have delay associated with this process.
The MTM has about a five millisecond delay between the time the audio event
happens and the MIDI event is transmitted (all three bytes, channel, note
and velocity). The Akai ME-35T is pretty quick with only two milliseconds
delay. The Alesis D4 delay depends on which trigger type is selected. If
you use one of the low trigger types, intended for triggering from pads,
the delay is between three and four milliseconds. If you use type 21 thru
25, the ones that Alesis recommends for triggering from analog tape, the
delay is between six and seven milliseconds.
You have to remember, though, that the delay amount doesn't really matter,
as long as you figure out what the delay is for your unit. You are going
to compensate for it when you print the new snare anyway. Once you have
your map cleaned up, shift it earlier by whatever the delay is. If your
sequencer shifts the sequence earlier by 13 milliseconds and the delay in
the D4 is six milliseconds, then crank seven milliseconds into the delay
fed by the new snare. Everything should line right up. Do a little delay
shuffling to compensate for the new sound, move the delay over to the SMPTE
line and print it.
Real Time Triggering.
This means that you don't want to build a tempo map into a sequencer, you
just want to play the tape, feed the snare drum into something, and come
out with the perfect snare sound. OK, if you figure it out, I want to be
the first to know how you did it.
All trigger units have delay, some just have more than others. If you are
going to use the original snare in addition to the new snare, then some
delay in the triggering may not matter. If the initial attack is used from
the original snare and the added snare is used for the meat of the sound
and doesn't really have a sharp attack of it's own, then a delay of a few
milliseconds will slip right by. I have used the Alesis D4 (with it's 500
built-in drum sounds) in real time to add fatness to a snare that was already
on tape and the six millisecond delay didn't show at all. The resultant
snare composite sounded great.
The Akai ME-35T has no sounds of it's own, so must trigger a sound module
via MIDI. The Simmons MTM must either send out MIDI to an external sound
module or send trigger pulses designed for Simmons Drums. With these devices
you have to add the MIDI delay of the sound module to the trigger-to-MIDI
delays. In the case of an Alesis D4 triggering a sound in an Akai S-1000,
this could add up to ten to 14 milliseconds. Now anybody can hear that,
even with one ear tied behind his (or her) back.
If you are recording on analog tape, you can compensate for the delay by
turning the tape over and delaying the trigger source over to another track.
When the tape is turned back over to travel in the right direction, the
trigger source is early. You can then run the trigger source through a delay
on it's way to the drum trigger. Now you can slip that new snare right to
where it is supposed to be.
If you are recording digitally, you can't turn the tape over 'cause it won't
play. You can only lock up a second machine with a copy of the trigger and
offset it earlier. I have used the Fostex D-20 SMPTE DAT machine for this
purpose, but you could use a two track analog machine or just about anything
that allows SMPTE synchronization. The other choice would be to use a device
that triggers with no delay.
The Wendeljr drum replacement module has the fastest real time trigger,
measured at three microseconds. Yup that's a 3/1000th of a millisecond.
The Wendeljr, however does not follow dynamics, so you have to either automate
the dynamics during mixing or use multiple Wendeljrs when replacing. There
was a prototype Wendel dynamics unit that would trigger in 60 microseconds
but it was never produced. Wendeljr could fake dynamics by switching between
two different samples during fills, eliminating the machine gun effect common
to drum machines. Wendeljr is not made any longer. Fostex was distributing
the remains but I think they are all gone. Some rental places have them,
but they always seem to be rented out.
Bob Clearmountain sells a little box that alternates between samples during
fills by sending out separate MIDI events. I haven't used one yet, so I
don't know the delay situation. It contains no sounds of it's own, but the
idea is to trigger two separate sounds in a MIDI sound module.
The Forat F-16 (drum sampler) is pretty fast when dynamics are turned off
and can be used in real time. It slows down considerably when following
dynamics. I have seen a few drummers with them in their racks.
I know that there are other samplers that trigger from audio, other tempo
mapping devices and other audio to MIDI converters, but the principles remain
the same. Just substitute the machines that you want to use and go to work.
Well there you have it, everything that you ever wanted to know about replacing
drums, and some things that you probably didn't want to know. This should
give you enough background to tackle just about any problem you come up
against. Good luck!