adventures in DIY music

Tuesday, 1 August 2017

Swing Conversion

(See table at the end of this post)
"Everybody play nice together."

Talking about rhythmic swing with musicians is just asking for trouble. Have a poke around the interwebs to see what I mean - lots of strongly-held views, with poorly-defined terminology.

But when it comes to rhythmic swing as a feature of rhythm machines, we know where that all started. It was introduced by Roger Linn back in 1979. So let's hear what he says about it, when asked what makes his drum machines so highly regarded:

“Swing – applied to quantized 16th-note beats – is a big part of it. My implementation of swing has always been very simple: I merely delay the second 16th note within each 8th note. In other words, I delay all the even-numbered 16th notes within the beat (2, 4, 6, 8, etc.) In my products I describe the swing amount in terms of the ratio of time duration between the first and second 16th notes within each 8th note. For example, 50% is no swing, meaning that both 16th notes within each 8th note are given equal timing. And 66% means perfect triplet swing, meaning that the first 16th note of each pair gets 2/3 of the time, and the second 16th note gets 1/3, so the second 16th note falls on a perfect 8th note triplet. The fun comes in the in-between settings. For example, a 90 BPM swing groove will feel looser at 62% than at a perfect swing setting of 66%. And for straight 16th-note beats (no swing), a swing setting of 54% will loosen up the feel without it sounding like swing. Between 50% and around 70% are lots of wonderful little settings that, for a particular beat and tempo, can change a rigid beat into something that makes people move."

(from an interview with Attack Magazine, July 2013)

Pretty straightforward if your drum machine has that function, but what if you are running more than one machine in sync, and you want them to swing together? Some old drum machines have very limited swing options, and some such as the TR-808 and TR-606 don't have any at all. Yes, they have triplet rhythms, but you can't swing the sixteenths for those in-between grooves that Roger mentions above. Well one way to do it would be to simply send them all the same sync signal, and embed the swing timing into the sync signal itself - voila! everyone swings together. This is something that devices such as those by Innerclock Systems allow you to do.

Above: the Innerclock SyncGen plugin can generate 8th or 16th swing timing to the hardware.

In my studio I'm currently running three separate audio sync tracks from my DAW out to my hardware. There are two Innerclock SyncGen tracks (one for an original and one for a newer model SyncLock) and one track of synthesized FSK to slave a Roland MC-50 midi sequencer.
 (Details about my synthesizing FSK for sync purposes can be found here)

Unfortunately, Innerclock, the DAW (Cubase) and Roland don't share the same terminology when it comes to applying swing.

Innerclock use the classic Linn terminology, from 50-75%, where 66% is a perfect triplet.

Cubase uses a swing percentage that goes from 0-100%, where 100% is a perfect triplet.

The Roland MC-50 doesn't use the term at all, and amazingly doesn't have a swing function per se. But you can achieve it using a quantization trick.

So the first problem I had was converting the Linn swing percent to the Cubase swing percent. That turns out to be pretty easy: subtract 50 and then multiply by 6 (reference). I checked it, it works in practice. OK, so how to get the MC-50 hardware to swing with the rest of the gear? One option is to make the FSK sync signal itself have swing, much like the Innerclock sync signal. How much? Maybe I could apply Cubase's swing function to the midi files driving the FSK synthesizer? Yes! By selecting the pulse that occurs on each even-numbered sixteenth beat division and dialling in 16th note swing percentage, that pulse ends up where it needs to be. However, you still then have to manually edit the pulses around it (each pulse takes up a 128th triplet division) to accomodate this, and to make sure that the second set of six pulses take up a shorter duration so that the first pulse of the following set starts bang on the regular beat division, shown below.

Above: Straight FSK pulses as midi notes (128th triplet duration)

Above: the first pulse on the second sixteenth has been swung by Cubase to 60%, and the pulses around it have been stretched or compressed as required.

So that's a bit of a pain, but at least once you've done it for the commonly used settings, you just save it as a template. I made up bars of FSK files for 54% 56% 58% 60% 62% and 64%, and they worked perfectly. 52% is too subtle for my ears, and 66% is simply triplets.
But.. of course, when the old hardware is used independently of a computer (which is one of the joys of this system!), it reverts to straight time again. Doh! 
So I needed to work out how much to delay these beats on the MC-50 for each of the settings of my other machines. That way it would swing just the same as when it ran as master, or slaved to straight sync code.
The Roland MC-50 midi composer has a resolution of 96 "clocks" or "ticks" or pulses per quarter note (ppqn). That's 48 per eighth note, and 24 per sixteenth note. To get the number of clocks needed to delay alternating sixteenth notes to achieve a certain swing percentage, multiply 48 by the percent swing (in Linn terminology) desired, divide by a hundred, then subtract 24.

e.g. for 54% , multiply 54 by 48, to give 2592, divide by a hundred = 25.9. Let’s round it to 26, then subtract 24 which leaves 2. So delay the second sixteenth note by 2 clocks from 24 to 26, and the fourth sixteenth note from 72 to 74.

e.g. “58%” = 27.8, rounded to 28, subtract 24 leaves 4. Therefore the clock positions of the “swung” notes will be at 28 and 76 clocks in the quarter note as seen in Microscope mode.

Of course, it would be exceedingly tedious to have to manually move notes in Microscope edit to achieve this. Luckily there is a fast way, called "iterative" quantization. You dial in a quantize resolution that will pull the notes in which you're interested in the right direction, but with control over just how much shift occurs.
On the MC-50 you hit EDIT, select  9. After quantising to rigid straight sixteenth time, choose sixteenth triplet resolution*, and adjust the “rate” of quantization (i.e. how strong the pull): a rate of 0.2 or 0.3 will give 54% swing, a rate of 0.4 gives 56%, a rate of 0.5 gives 58%, 0.6 gives 60%, 0.7 or 0.8 gives 62%, 0.9 will give 64%, and of course a rate of 1.0 gives perfect triplets. Note that if you haven’t quantized to straight time first, the outcomes of the different rates will yield less predictable results. For reference I've put all these numbers into a table for comparison, below.

*Why not eighth note triplet resolution? Because in the sixteenth triplet resolution, the “middle” sixteenth note in the quarter (i.e. the third) is preserved in it’s midway position between the “swung” notes.

So, who cares? Why do you need to obsess over these silly numbers? Because, as Roger says above, for certain rhythms, at certain tempos, you’ll find that just a certain amount of swing makes the thing “groove”. And if you can get all the "players" (humans and machines) in the room locked to that groove, that's a good place to start to make some great music.

Linn swing
from 50
Cubase swing
from zero
Roland SuperMRC
at 96 ppqn
iterative Q
"Rate" *
54% 24% delay by 2 clocks 0.2-0.3
56% 36% delay by 3 clocks 0.4
58% 48% delay by 4 clocks 0.5
60% 60% delay by 5 clocks 0.6
62% 72% delay by 6 clocks 0.7-0.8
64% 84% delay by 7 clocks 0.9
66% 100% delay by 8 clocks 1.0

*at sixteenth triplet resolution after hard quantizing to straight sixteenths.

Tuesday, 6 June 2017

Roland CR-8000 CompuRhythm

This drum machine arrived in 1981, the same year as the TR-606, and a year after the TR-808. They all share similar analogue voice circuits. The CompuRhythm's programmability is less intuitive than the other two, and the mainstay of it's rhythms are rock, latin, and cabaret dance presets. What's different about the CR is how these presets can be varied instantly via a couple of unique features. One is the "Arranger" buttons along the top row, that select a preset motif of a single instrument that can be superimposed on the pattern currently playing. For example, a cymbal on every 8th note, a handclap on the 2 and 4. More than one can be selected at once. The motifs for the hi hats are interactive - start the 16th note hi hat pattern then add the open hats on the off-beat eights for instant disco! The other feature is the "Register" button. This flips between two states: an "A" and a "B" pattern of your choosing, along with whatever "Arranger" options are selected. This button can be "played" in real time, or triggered via a jack input, allowing you to "cut" a mix of the two play states. Since the machine also sports a trigger output you can, by plugging a cable from the trigger out to the Register input jack, make the CR-8000 actually auto-arrange a new composite rhythm. These deceptively simple tricks, along with the fill-in implementation, make the CompuRhythm a lot of fun, and embody the "semi-automatic" ethos.

The CompuRhythm 8000 is superior to the 5000 in that:- it possesses the handclap sound (the most complex in terms of circuitry), it is externally sync'able via Din Sync with tempo shown by a segmental LED display, and, it has eight user-programmable 2-bar patterns and four programmable fill-ins. The 5000 is a preset only machine without external clocking options (although you can use the "Restart" input jack creatively to get around this limitation). The 8000's programmability is unfortunately not extended to three of it's voices - the Rimshot, the Clave, and the High Conga. The only time you can hear these is as part of a preset, which is one reason why I have wanted to look at external voice trigger options.


There are 14 voices: Bass Drum, Snare, High Tom, Low Tom, Cowbell, Rimshot, Clave, Hi, Mid and Low Congas, Open and Closed Hi Hats, Cymbal, and Hand Clap. Examination of the service manuals for the CR and its TR cousins from the same era, reveals they all share similar circuitry to create these voices. In particular, the cymbal and hi hat sounds are based on a filtered mix of square wave oscillators, giving a nice metallic sheen, as opposed to Roland's rhythm machines from the previous decade such as the TR-77 which used simple white noise filtering for these sounds. The membrane drum sounds (BD, SD, toms, congas) use one or two "bridged T-network" damped oscillators as shown in the TR-808 service manual Figure 11. Additionally, the tom and snare oscillators output have pink and white noise, respectively, mixed in.

Above image copyright Roland Corp
Figure 12 shows the single-transistor VCA that is used to process that mix of six square wave oscillators for the metal instruments.
The rimshot voice in the CR-8000 is outstanding - a wonderfully woody, cutting sound. Looking at it closer, one finds it is created differently here than in other drum machines, where it usually a T-network voice. Here it takes the output of the lowest two of those six oscillators (I measured these around 574 and 378 Hz) through a "swing type" VCA, as above. What does this do? An engineer on the synth-DIY list explained it to me (thanks RB!). Basically it is a very cheap and dirty ring modulator, that creates a lot of complex non-harmonic frequencies from just a couple of oscillators.
Check out the circuit for the rimshot voice below. The VCA in question is Q36. To the right of this is a resonant high pass filter that cuts all the lows out and emphasises around 1kHz.

Above image copyright Roland Corp
You can see the "lower" two square wave oscillators on the upper left, the sum of their outputs entering the base of Q36. Note the rimshot trigger has an input to these via Q37. This is to "reset" the oscillators at the start of the impulse, so that there will be a consistent attack to the sound. You can see on this scope shot, the upper trace is the square wave mix measured before C149, the lower is the rimshot audio at the combined audio output:

Again, compared to the TR-77 and it's ilk, these circuits are more sophisticated, and "realistic". The eighties instruments feature nice subtleties in the bass and tom circuits whereby resonant frequency changes with amplitude, in imitation of physical vibrating membranes (part of the reason a 909 kick drum can seem so powerful).

Voice triggering

"Accent" is Roland's term for emphasising the volume of certain steps in a rhythm pattern. This emphasis was needed because the early machines (TR-77, 33, 66 etc) had no "velocity" or volume variation in a voice, so their patterns were criticised for being mechanical and lifeless. The addition of just a single level of volume change between steps within a pattern makes a huge improvement in this regard.
There are different ways of implementing "Accent". In the TR-808, each individual voice's trigger pulse is increased on the accented steps. This results in not just volume increase, but also some tonal change in some voices, especially noticeable in the snare and toms. Accent in the CR-8000 is achieved further down the audio chain by using a trigger pulse identical to those used for the voices to cause a VCA (Roland BA662 OTA) to bump up the volume of the voices mix buss, the amount being controlled by a front-panel potentiometer. This VCA affects the mix of bass drum, snare, toms, rimshot, hi hats, cymbal and congas, but is placed before the addition of the handclap, claves, and cowbell to the master mix, and so these later voices are always left un-accented.  Implemented this way, the CR-8000 accent should not introduce any tonal change to the voices. It's not until you get external control of the voice triggers that you discover some of these voices really do respond in tone and volume to different pulse amplitudes.

Above image copyright Roland Corp

The voices and Accent are triggered by outputs of the CPU. These triggers are a negative-going pulse from a baseline of 5 volts, to zero volts. How long the pulse lasts depends on the tempo set, the variation is from about 4 milliseconds up to 70 millisecond -  pretty wide. This pulse duration seems to have no effect on the volume or tone of the voices, with one exception that I'll demonstrate.

To gain external control I disconnected the internal trigger buss and connected an Elby midiSDS-16 (midi-to-trigger unit), after configuring it to send 5 volt negative pulses of around 4- 5 ms duration, with midi velocity determining pulse amplitude.

Turning the Accent pot to maximum gives the widest amplitude response to midi velocity. In the following audio demonstration you can hear voices that are all triggered with midi velocity of 100. In order, you hear the snare, then kick, rimshot, hi conga, low tom, cymbal then closed hi hat. For each, the first eight hits are un-accented, then an accent pulse going from maximum to minimum is triggered simultaneously with the subsequent 16 hits. What you hear is mostly volume change, very little tonal change. The little there is might be some distortion at the total mix amp circuit on the loud hits.

The cymbal and the open hi hat are the two voices that have a little "tail" to them, and you can hear the accent pulse interact with this, in possibly useful ways.
In this demo, hear the initial cymbal hit, followed by accent pulses punching into the decaying tail. It plays twice. Following this, hear how you can get a kind of "pedal" hat sound, by putting the accent pulse just at the tail of the open hat. These applications suggest that the accent pulse is the one trigger point where it might be useful to have control over the duration of the pulse, not just the amplitude.

So much for the effect of Accent. What is the effect of changing the individual voices pulse amplitude? Using a midi file such as that shown below,

the following demo has each voice triggered with a decreasing velocity, but note: NO accents are triggered. In order, you hear hi conga, mid conga, low conga, cowbell, cymbal, mid tom, open hat, low tom, closed hat, snare, bass drum, clap, clave, and finally rimshot. At low velocities some instruments cut out early and won't play all the notes. Most have some volume change with velocity, and some such as the toms and snare have interesting tonal changes, such as the noise components dropping out quickly at lower levels. Tonal changes in the rimshot and congas are much more subtle, and are pretty much absent in the cowbell and claps.

And finally, here is a mess of drum hits to show the dynamic range the combination of accent and velocity can get you with these voices under external control.

dyno demo wav

Text, images and audio copyright © Adam Inglis 2017 except where indicated. 

Wednesday, 31 May 2017

Roland R8 into Frostwave Sonic Alienator

In seeking percussion sounds for a new Funboys track, we were really taken with the energy of this combo. The SA has a filter which allows you to find the sweet spot for your sound, which you then can obliterate, mangle, or just rough up a little, using the various modes and the sample rate reduction knob.

Tuesday, 30 May 2017

Roland Rhythm TR-77 - how it works.

(from March 2010)

The "Roland Rhythm" TR-77 was Roland's very first product, from 1972. It makes its distinctive drum voices from what are called "tuned resonant" circuits, which consist of a capacitor, an inductor and a resistor - very basic!. The patterns are preset rock, jazz and latin styles of 1 or 2 bars. There are no user-programmable patterns, because there is no "memory" - the patterns are hard-wired. That is, they are literally made from a matrix of wires between the transistor flip-flop counters, connected by diodes, to the inputs of the voice circuits, depending on which buttons on the front panel are pushed in (Pushing one in causes the others to disengage, but if you're careful you can manage to get more than one to stay in, giving you a mix of patterns.)
Here's a demo video: first I start the metronome, then touch the start plate to kick the rhythm off, then hit the fade button for auto fade out... back to the metronome and through a few combination of rhythms. Notice it takes a certain pressure to keep more than one button latched down!

My main motivation for investigating the gubbins of this old beast was to see if there was a way to synchronize it to a source of external clock. I also wanted to know if there were any good voice mods possible. But another reason was that I wanted to understand how this basic type of pattern sequencing worked, because my amateur electronic explorations hadn't really covered this so far.
Above: Photo from the rear with the heavy wooden top/end cheek/music stand assembly flipped up to the front. At the bottom right corner I've taped a cardboard box lid over the transformer and mains terminals (WARNING: LETHAL VOLTAGES HERE) for safety while working on the machine.
Drum Voices

There are thirteen voices: Bass drum, low conga, low bongo, high bongo, rim shot, cowbell, claves, tamborine, cymbal, high-hat, maracas, snare drum and guiro. The first seven are based on tuned LC circuits or "tanks", that "ring" when struck with a brief negative pulse of about 5 to 12 volts. Of these, the cowbell uses two detuned tanks. The tamborine, cymbal, high-hat and maracas are triggered pulses of filtered white noise (supplied by a common transistor), with different envelope times and filters. The snare is a mixture of both a noise filter and an LC tank.
Above: video demonstrating the voices (except guiro), triggered from the points along the front edge of the voice board. The other end of the wire in my hand is clipped to matrix pattern line no. 21 on the logic board. At 00:52 I start a latin pattern, and "solo" on some of the voices with the wire!
Above: Schematic of the voice circuits (copyright Roland Corp) from the service manual. The left column shows the LC tanks, the middle the triggered noise circuits with the guiro at the bottom, the right shows the noise source, pre-amp and the single VCA chip.

The guiro is completely different. It is itself a two-transistor multivibrator, i.e. an oscillator, who's output is summed with some white noise and fed to a bandpass filter. It has two control inputs that, unlike all the others, require a positive voltage of about 5 volts. One input triggers the sound itself, the other raises it's pitch periodically in the pattern cycle. These positive CVs are provided by pattern numbers 6 and 12 respectively on the logic board.

Above: guiro oscillator being triggered by positive CVs. First by a wire connected to matrix line 6 on the logic board. Then by a square wave from a signal generator at 1 hz, then 10 hz. You can hear slight pitch changes throughout. This is because the pitch CV input is hardwired to matrix line 12 (red wire joining the board near the pale blue cap) causing a permanent cyclical kick upwards in pitch (see chart below).

Most of the trim pots on the voice board simply change the output volume of the voices, which is important because there are only four separate voice volume sliders on the front panel (- bass, snare, guiro and a combined maracas-high hat-cymbal volume -) but not very exciting from a sound modification point of view. The snare pot increases the mix of tuned sound to white noise, the rim pot increases the aggressiveness of the sound and also seems to add a slight echo. The guiro pitch trim doesn't do much. Much more dramatic effects can be had by bridging the caps in some areas. The guiro can become quite gnarly by increasing the resonance of the BP filter (cap C371). I found by bridging the cap across the combined output of the "noise" instruments (cap C356, to 470pf I added a 220nf) there was a great improvement in the quality and clarity of their top end, so I kept this in.
Also, some sounds have a slight "accent" response to different amplitudes of trigger voltage, namely the bass, snare and rim shot.
Something to keep in mind with any machine like this that uses lots of inductors in it's sound generating circuits - it's a fantastic amplifier of HUM! Make sure you keep other units or wall-warts well away, especially from the rear right corner where the voice board is.

One last audio feature worth mentioning is the single IC to be found here. Because the TR-77 is ALWAYS running when the power is on, Roland used a VCA chip to gate the audio output on and off. A transistor switch triggered from the start bar causes a change from the OFF voltage (about +7.5 v) to the ON voltage (about +3.0v). Even better, by slowly changing this control voltage by charging a capacitor, it allowed the wonderfully wacky "fade-out" feature! Those seventies cocktail bars would've been gobsmacked - "it sounds just like a record!"

Above: the VCA chip
Above: cap 356 bridged. To the right is the snare trim pot. The black disc is one of the many inductors on the voice board that make up the tuned "tanks".
Sync and Timing

The timing information in the TR77 originates with the master oscillator, a two-transistor multivibrator, which generates a square wave (about 5 volts or more) who's frequency is controlled by the tempo slider on the front. This frequency is four pulses per quarter note (i.e. sixteenth note division), and is fed to the first of five flip-flop stages (also made of two transistors each), which each act as a divide-by-two counter. So the first counter, output Y, counts two sixteenths and so completes a cycle in an eight note time division. This is fed to the next counter, output X, which counts two eighth notes and completes a cycle in a quarter note. This chaining continues so that output W cycles per half note, output V cycles per whole note (1 bar) cycle, and finally output U cycles every 2 bars. These are all square waves of about 10 volts p-p. Touching the start bar or the restart footswitch causes the master oscillator and all the flip-flop's voltages to be slammed briefly to ground, causing the count to start over (see the video below). One other thing to note - each counter's square wave output has it's inverse waveform also output (e.g. for Y, there is Y').
Above: short video of the scope showing the effect of the reset circuit on the master oscillator (top trace) and the "V" counter (bottom trace - cycles one square wave per bar or 16 sixteenths). You can hear me hit the touch bar quickly starting and stopping the pattern a few times before letting it run.

Now here's the clever bit: it's by the ingenious combination of two or more of these various counter outputs that the pulse patterns are derived to trigger the voices. This is the "matrix"!
(It's shown on page 10 of the service manual). If you're tired of sloppy timing in your drum machine patterns, then maybe you need to start hard-wiring them like this!! The connections are via capacitors, diodes and resistors - very simple (see the edge-detector circuit drawn below).These patterns are presented at the numbered terminals along the front edge of the logic board, demonstrated in the video below..
Above: Some of the matrix patterns available on the logic board. The other end of the wire I'm holding is clipped to the cowbell trigger point on the voice board. No other instruments except a four/four bass drum pattern are sounding.

Page 11 of the service manual (below) shows the graphical representation of the pulses present at each matrix line over a two bar period (32 sixteenth notes).
Note line number 6 and line number 12 are specifically for the guiro as mentioned above. They provide positive 5 volt gates, not negative pulses.

Above: TR-77 logic timing chart (copyright Roland Corp)

External Sync Mod

Initially I figured I'd need a divide-by-six circuit to get the 24 ppqn of standard (Roland) din-sync down to the required 4 ppqn that the TR77 expects to see. I breadboarded up a circuit using CMOS chips, a 4027 and a 4013, but I didn't have a great deal of luck - sometimes it worked, sometimes it didn't. I suspect the chips didn't like the quality of the power I was feeding them from the TR77. There are NO voltage regulators in there! Just some basic filtering after the transformer. Also, there was the problem of how to set the start voltage accurately each time with the first clock pulse.

Above: Resistor R4 (clipped off the board) where I accessed the master clock input.
I converted one of the switched mono phone jacks at the rear of the unit into a clock input, and the restart jack as a run/stop input. The clock wire was taken to the logic board at resistor R4 (the other side of which is connected to the base of Q1 of the master oscillator). The run/stop was connected as shown below in (a)...

At (b) is a negative-edge detector circuit, found repeatedly throughout the matrix to provide the negative pulses for the voice circuits.

Above: where I put that transistor for the Run/Stop input.

External clocking needs a source of accurate square pulses in 16ths that will start bang on with the start of the rhythm. I use the arp clock output from a Kenton Midi/CV converter, or the pulse out from an Innerclock Sync Lock.

Possible Creative Mods

At least one thing is clear from having a close look at these old machines - even if you don't have access to one, it shouldn't be too hard to make the circuits from scratch yourself!
One popular idea has been to bring out the trigger inputs to the individual voices for external access. However, that means you need a source of multiple negative pulse triggers controllable from your sequencer. Which leads to the next idea, of also bringing out the matrix lines as well, giving you cross-patchable patterns for each voice. I'm sure you could even just twist a couple of wires together to combine the rhythm lines for any voice, as they all are diode protected.
Serious sound mods may require switched capacitors and/or inductors, not simply pots. Inductors like these could be hard to find - although I've heard they're still obtainable from suppliers of parts for old Thomas organs and the like. Mods like these could start to take up a bit of space. But by removing the wood cover and replacing it with perspex or sheet metal, one then gains a huge area of real estate for pots, switches, and even a patchbay or pin matrix, with room for sound-mangling circuits on the underside!

All non-Roland copyrights are copyright © 2010

Wednesday, 24 May 2017

Roland SH-09 slide pot rehab

When my usual method of cotton-bud-with-De-Oxit didn't fix these crackly slide pots, it was time to open them up and take a closer look.

Note the Resonance pot is inverted with respect to all the others.

Unbending the 6 lugs.

The wiper tines are extremely delicate. I chose to very gently splay them a little so they would be in contact with a fresh area of track, being careful not to deform their angle from the carriage. The contact surfaces were also very gently sanded.

A gentle rub over the conductive track with some 600 gauge superfine sandpaper has taken off the darker layer of oxide. Don't sand the resistive track! The oblique lines across the resistive track are normal, presumably a result of the manufacturing process.

In my case, although the resistive tracks had been cleaned quite well with the De-Oxit/cotton bud, the oxide on the wiper and the conductive track meant the crackles persisted. They are now working like new.... but for how long?

Saturday, 13 May 2017

Akai AX-60/S612 synth/sampler combo

I wasn't initially very excited when an Akai S612 sampler was bundled in with a second-hand deal I was doing. A whole one second of sampling time at 32kHz - woo hoo!! And then I read in the manual about the multi-voice output and it's use with the Akai synths of the day, and got curious.
When I finally found an AX-60 (they were rare even back then in the early 1990's) to partner up with it, I discovered the way these two units work together over midi and the proprietary voice interface was quirky and unique.

As a midi synth of it’s time (1986) the AX-60 had a pretty limited spec - no sys-ex, no velocity or aftertouch, response to CC#s 1 , 7 , and 64 only, and pitch bend. Yes, it had an arpeggiator, but this didn’t clock to midi, it required an old-fashioned voltage trigger for external sync. However, there was a big plus. It was bi-timbral, with a programmable split point, and separate (adjacent) midi channels for each timbre. You could allocate voices either side of the split in several defined ways. You could save the whole set-up - split/voice allocation/patch numbers/midi channels/unison and chorus settings - to one of eight “SPLIT PRESETS” for instant recall.

But it was when you hooked up midi, and the 13 pin DIN cable, to the Akai sampler that things got really interesting.
The multi-way connector allows the 6 voices of the sampler to enter the 6 voice channels of the synth individually. Hitting the “SAMPLER” button on the synth then does two things: it sends a midi mode change message CC# 126 to the sampler, putting it in midi mono mode, and it likewise puts the synth into (a kind of) mono mode. Now, note values generated by the synth CPU will be allocated to one of the six voice channels, and that note will be transmitted on a midi channel 1 to 6, that corresponds to the voice number. Thus, when a synth note is generated on midi ch. 1, Voice 1 on the sampler is triggered and sent to the synth’s voice channel 1 to be affected by the VCF, VCA and chorus circuits. This works for the arpeggiator as well - in fact, this is the only way to get the arpeggiator to transmit its notes from the midi out. (That last fact in itself gives you some interesting options when interfacing other synths with the AX.)

So what happens to this combo when you apply some of the other features?

In Unison mode, it works as expected, pressing a key generates 6 notes on 6 midi channels to the sampler to be processed by the synth. With chorus on, this can be quite a thick sound, but perhaps you wouldn't call it lush.
In Unison + Arpeggio mode, again as you would expect, each arpeggio note is transmitted in 6 note unison.
In Split mode (so says the manual), you can't use the sampler interface. Er... except you can! Although the S612 is not bi-timbral like the AX (it can only hold one sample in memory), the sampler voices each side of the split are treated according to the separate patch settings for the filter, EGs and chorus, so it's kinda bi-timbral in this scenario. The special split voice allocation settings where there are 6 voices one side and zero voices on the other (so the "zero" side of the split can control an external unit) will not work here, because bizarrely there is no midi transmission from that half of the keyboard until you switch the Sampler button off.
In Split+Unison+Arp modes combined, things can get a little hard to keep track of. In the 2/4  voice allocations, you can have Unison on one or both sides, and Arpeggio on just one, and the samples will follow along.

All in all, a powerful and fun combination. So, of course, you would like to control this duo and their interactions from your sequencer, wouldn't you? Well, there is another quirk of the AX-60 that will mess this up if you're not aware. When the Sampler mode is engaged, the AX transmits the mode change to mono mode, but in itself, it only becomes a transmitter in mono mode, not a receiver! It will receive external midi commands, but only on the channels designated while it was in Poly mode. Even more bizarrely, it transmits only notes in mono mode, not pitch bend or mod wheel. Those CCs get transmitted on the Poly channel (or both channels if in Split mode). Phew! 
But hang on, the S612 responds to both pitch bend and mod wheel in mono mode, doesn't it?... in fact, it seems to respond to those CCs on any channel between 1 and 6. 
So, reliable sequencer control of the combo is going to require some thought. I'll grapple with this topic in a future post.

Thursday, 11 May 2017

Funboys "In The Boot" EP

We serious gents, the professor, the doctor and the chocolatier, have our new EP (such a quaint initialism from the vinyl days) out on Club Sweat. And we didn't have to pay these people to say nice things about it here !!