adventures in DIY music

Saturday, 22 June 2019

Rear Panel: Boss RSD-10 digital sampler/delay

From left to right: standard Boss barrel connectors for 9 volts DC, negative tip, in and out; Effect Remote on/off footswitch, Pad control input, Trigger control input, Keyboard control input, Tape control input on RCA phono, Audio Out on phone and RCA phono, Level switch, Audio In on phone and RCA phono.

As in the RDD-20, the Effect Remote switch expects to see a press-to-break, normally closed switch, which grounds the contact when a plug is inserted. The front panel Effect switch must be on for this to work. Pressing the footswitch turns on the effect only while pressed. While this is very handy, for my purposes I wanted control over when the audio was reaching the circuits, so I converted this to an input mute switch, allowing me to drop audio in to the delay or sampler input with a tap of the foot, leaving hands free for time adjustments and keyboard triggering.

The Pad input detects a trigger to start the sample playback, as well as a volume level for that trigger by sending a CV to the compander circuit.

The Trigger input just detects a sample start trigger. Volume is fixed.

There are two pitch control inputs, Keyboard and Tape. Tape won't work without Keyboard also being plugged in - with nothing plugged into the keyboard jack, the pitch of the sample playback is determined by the front panel pot. So how does it detect pitch? There's no midi, and it doesn't use a pitch CV in the sense that a modular or analogue synth does. A chip known as a PLL (phase-locked loop) changes the frequency of its VCO when it detects the frequency of the Keyboard signal. To do this, it needs to see a fairly clean, pitch stable, monophonic waveform with minimal harmonics, preferably a sine wave. The Keyboard input also triggers the start of the sample, the volume, and uniquely, the playback time (gate).

Roland had two purposes in mind for the Tape input. The first is a rather complicated way of using a cassette as a backup device for the sampler and it's settings, by recording the initial key pitch, followed by the sample itself, onto different channels of a standard stereo cassette tape, while switching settings on the front of the machine*.  Secondly, under the heading "As a Sequencer",  there is this single obscure paragraph in the instruction manual:
"By recording the keyboard sound onto the cassette deck and feeding the playback sound to the RSD-10 through the Pitch Control Tape input jack, the sampled sound will be automatically played back".
I'm assuming that the "sequencer" refers to the sequence of pitches on a tape. Of course, any suitable audio source would work here.

This machine, like so many other Roland and Boss effects from this era, is based on the "long chip", the custom RDD63H101 CMOS gate array. But it is the external control circuitry of this machine that  enables the magic to happen. It is a fascinating little bit of eighties technology, another example of Roland ingenuity.

* I've never attempted this, however, there is of course a way of doing the same thing with a modern DAW that works quite reliably. You put the audio tone on one track, and the sample on another, and make sure your triggering audio tone starts bang on with the sample you are loading (this works in Mode B Manual Rec/Play).

Thursday, 20 June 2019

Rear Panel: Boss RDD-20 digital delay

From left to right: DC power input 9 volts standard Boss barrel connector negative tip, DC power output for daisychaining, Modulation Bus polarity switch and phone jack (input/output), Effect Remote on/off footswitch, Delay audio signal output on phone jack and RCA phono, Mixed audio signal ouptut on phone jack and RCA phono, Level (Unigain) switch, Input audio signal on phone jack and RCA phono.

The Boss Micro Studio system of half-rack sized modules from the mid-eighties used the same 9 volt, negative tip, barrel connector as the Boss effect pedals, and here they conveniently gave you a daisychain power output on each unit, with the current draw of that particular unit printed below, so a sufficient supply/current capability could be calculated. Maximum total draw is recommended not to exceed 200 mA.
The Mod Bus allowed a 0 - 5 volt CV to control the VCO that provided the master clock for the main controller chip (the custom Roland/Boss "long chip" used in so many of their effects around this time). Several other units in the series sported a Mod Bus connector - the idea was you could link two units (perhaps more?) and have one as the "master" (Mod depth turned up) and the other as the slave (Mod depth turned down), changing the polarity as necessary, for stereo effects.

The Effect Remote jack expects to see a typical Roland type footswitch, normally closed, press-to-break type switch. While this is handy, I wanted an easy way to implement a "dub delay" function, where the unit is used as a send for special effects, rather than say, corrective or fixed delay uses. The simple but surprisingly effective Tone control in the feedback circuit makes this unit great for this sort of thing. I disconnected this jack from it's stock purpose (by de-soldering R63), and ran a wire from the tip to the audio input. Thus, when a footswitch is plugged in here, the audio input is grounded. Pressing the switch breaks the contact allowing you to briefly drop in sounds at a touch, for those classic dub effects.

Tuesday, 18 June 2019

Rear Panel: Korg SDD-3000 digital delay

This is the original rack unit from 1982.
From left to right: ground terminal, ID plate, captive mains lead, Direct output jack, "+"Mix/Mono output, "-"Mix output, Delay Mod CV input 0 - 5 volts, Bypass footswitch, Hold footswitch, Prog Up and Prog Down footswitches, audio input jack.

The +Mix jack can output a mix of direct and delayed signal, the mix determined by the front panel Level Balance pot, and the phase of the delayed signal can be inverted to out-of-phase also by a front panel switch. The - Mix output can have a similar mix, however the phase of the delay is always inverted. This allows appropriate stereo set-ups for mono compatible recording, or live PA scenarios.

The Delay Mod input has a dramatic effect on the delay time. This can be scaled using the front panel Intensity pot. This gives scope for creative external control e.g. using your midi-CV converter to send quantized  5 volt changes can allow beat-matched pitch-shifting type effects on melodic material, or create interesting polyrhythms or fills from a drum input.

The Program shift Up and Down inputs also allow creative mangling e.g. synchronising different flange settings on different points of the rhythm.

The audio input jack has a convenient feature. It can stay patched in, but when a plug is inserted into the front panel input, this signal is switched out.

Saturday, 15 June 2019

Rear Panel: Roland SDE-1000 digital delay

From left to right: ground terminal, captive mains lead, ID plate, Delay time pot x1 to x1.5, Modulation CV input jack 0 - 5 volts, Preset shift control input jack, "Play Mate" control input jack, Hold on/off control input jack, Delay on/off control input jack, Delay audio output jack, Mixed audio output jack, "Unigain" switch for input and output levels, Audio input jack.

The Delay Time pot directly affects the VCO in charge of the master clock pulses to the main controller. For some reason, turning this up will increase the unit's tendency to go into nasty high frequency oscillation at high feedback levels. Also, the quality of the delay signal is noticeably better at lower settings. Moving this pot will affect the time of the 4 programs.

The Modulation Foot Control has 5 volts available to directly control delay time via a pot or a volume pedal, or you can send it 0 - 5 volts from another source. This can be scaled from the front panel Mod pot. Unfortunately the range of the external control is somewhat underwhelming.

The Preset Shift jack takes a momentary switch/footswitch to step through the four programmable presets. This would seem to give scope for some creative program changing. Say for example, to enhance a rhythm track, you could have different preset resonant flange effects synchronised with the beat. Unfortunately, Roland put a spanner in the works there, by getting the CPU to mute the delay output every time a preset is changed (this muting also occurs when the delay time is changed from the front panel, and is there to suppress glitches. The glitches, for me, are often where the happy accidents occur). Looking at the schematic, I identified the line from the CPU to the logic that controlled this muting. I disconnected the CPU pin, and tied the logic side to 5 volts, disabling this "feature" (without disabling the power-on mute etc).

The "Play Mate" footswitch function, on first tap, resets the display to all zeroes, then the following two taps allows the CPU to calculate the time between them and set that as the delay time, i.e. tap tempo.

The Hold jack takes a footswitch, and plays the contents of the memory over and over, uninfluenced by front panel control except, happily, the Timex2 switch.

The Delay on/off switch is very handy for external sequencer control, especially in high feedback scenarios - you can chop it rhythmically! Note there is no corresponding front panel control for this.

The Delay output jack obviously just has the delay signal on it. When a plug is inserted here, the delay signal is switched out from the Mixed output, leaving just the dry signal, for those live "stereo" configurations.

The audio input can take line or guitar level signals. The pre-amp is a push-pull emitter follower type with a gain of about x24.

Sunday, 19 May 2019

Farfisa "Compact"

The Farfisa "Compact" organ line was manufactured from 1964 to 1970, and there were many variations. Their unique sound stems from the way the transistor divider circuits act on the high frequency "square" wave oscillator, resulting in narrow, rounded, negative-going pulse waves.
The model I have here is probably from around 1966 (hard to say for sure as the serial numbers aren't consistent). Although they made a big deal about it's "electronic transistor"  circuits, it also contains two twin-triode valves. Of the four valve stages available, one is for the preamp, two for the reverb driver and one for the reverb recovery.
To service one of these beasts, the legs need to be in the open position, to allow access to the 5 wing nuts in the wooden base, so that the internal frame that holds the keys and the electronics can be lifted up out of the flightcase.
!!!Now's the time to re-check you unplugged it from the mains!!!
The trick here is to find the four screws holding down the PSU chassis, with it's gigantic transformer  and two valves sitting atop. Just loosen these enough to allow them to slide in their slots, moving the PSU towards the front of the machine so that the pots on the rear side clear the body cut-out. Then you can lift the internal frame up and rest it on the rear edge of the case, allowing access to the voice boards and (sorta) the key contacts. You can't really put this thing on your bench unless you do a much more comprehensive  dismantle, so you'll need plenty of room to get it as close as possible to an area you can work off.
Above: with the internal frame resting on the rear edge of the case, you can see the 3 groups of 4 voice boards ("master oscillators"), one for each of the twelve notes, and the filter boards to the right of these.

Above: the large transformer and the two ECC83 valves sitting atop the PSU chassis, on the left with the bass (pedal) trimpots board just in front of this.

This baby had been 'in the family' since the late seventies, and had given trouble-free service all that time. However now it wasn't making any sound except buzzing. Inspection immediately revealed a few electrolytic capacitors had given up the ghost, and lifting up the PSU revealed the charred remains of what I think was a spark suppression capacitor near the power switch.

Replacing these, and a few other electros in the PSU, and we had organ sounds again, but they weren't quite right...

When diagnosing faults with the sound of the organ voices, it helps to isolate the oscillator divisions. Start with all tabs off, except one of the 4' voices. Play all the notes chromatically.
There might be an obvious crackle suggesting the contacts need cleaning. Or, there might be a dead key altogether. Is it the contact, or the voice electronics? Switch the 4' tab off and the 8' tab on. Did the dead key stay dead, or did it just move up an octave? If it moved, it could be that specific divider output on the master oscillator boards for that note. I had a dead "A" that moved with the footage selection. This was traced to the 1uF electrolytic cap at the output of that division, on the master oscillator board for "A". Each division of each of the twelve notes has such a cap on the output. 
Shortly after fixing that, all my "D" notes went wonky, sounded horrible across all octaves. Getting the scope onto the boards I found the shape of ALL of the D divisions looked like this

Whereas the shape of the healthy notes looked like this

The top frequency of the D oscillator, the square wave from which all the lower notes are divided, looked OK, much like the other notes. So whatever was wrong, it had to affect ALL the divider circuits. I couldn't see, from looking at the schematic of the master oscillator board, what that one component could be. 

But inspecting the voice board itself revealed a cap that wasn't marked on the schematic, a 50uF 12 volt electrolytic that was bypassing the 8 volt rail (top line of the schematic above). This rail is a common connection to ALL the emitters of the divider transistors. Ah ha! Sure enough, replacing this fixed the sound, and the pretty pulse waves were back again.

Above: the "D" oscillator board. The blue electro was the unmarked cap.

Key contacts will likely need cleaning. Each key (except for the bass octave) has a contact for 4', 8', and 16' footages. Cleaning these contacts will test your patience, I assure you. Make sure you allow a LOT of time. Beforehand, you will need to fashion some kind of tool. Now, this could be something as simple as an old-school pipe-cleaner with a couple of judicious bends made in it. That didn't work for me, I needed something more rigid. But it needs to be very narrow, to get past the crowded linkages at the rear of the key assembly. Yes, 3 contacts per key. And you can't really see what you're doing. Be very careful, if you damage one of the delicate springs (that live in that tiny space where you can't see) that actually make the contact, it will be a major nightmare to replace!           
If anyone knows of a shortcut...?

Above: by carefully sliding out the PCBs on their harnesses, you can gain slightly easier access to the rear of the keys to clean the contacts... if you can make it past the white plastic linkages.

So having fixed the oscillators and cleaned the key contacts, I then realise the reverb has gone. The reverb in this machine is a source of great character. The springs are suspended inside a box that itself is suspended below the base of the flightcase, beneath the keys. Hitting the box produced the usual "reverb explosion", however no organ sound was making it to the springs. I deduced from this that something on the driver side must have failed, while the pick-up circuitry was still functional. Two halves of valve 2 are involved with the driver amplification, and half of valve 1 with the pickup recovery amp (see below). There was a seemingly very healthy 100 volt signal at the driver side output to the transducer. Still, I tried replacing the valves - no change. The driver and pickup transducers are actually ceramic gramophone cartridges. In the place of the stylus, is a wire that is soldered to the springs.

These sort of cartridges appear to have some sort of piezo material inside them, that decays over time. I couldn't find anything like them for sale anywhere.However, Jaycar sold "buzzers" that contained piezo elements, so I cracked one of these open and experimented with it, after bringing out some new connections from the PSU so that I could have the box on the bench. (About here is where you realise it's going to be much easier if you adapt the reverb wiring to accomodate some Molex connectors, rather than crouching under the machine with a hot soldering iron in one hand..) There were definitely some "sweet spots" for the transducer, and the wire needed to be tight against it. I tried it both in driver and pickup positions.

I settled on using the disc as the new driver. It sounds to me like there is some "plate" effect as well as a fair bit of distorted "direct" sound going on with this arrangement, but in any case, it sounds great, certainly just as characterful as the original!  In fact, I liked it so much I converted the output jack in the PSU chassis that had been dedicated to the bass or pedal octave, to an audio input jack for the pre-amp and reverb, summing it with the internal organ sounds, making that dirty, springy, valve-y goodness available to all!

I'm grateful for the all the assistance I received with these repairs, from the heavy hitters on the Sound On Sound DIY forum!

Friday, 17 May 2019

Silent TR-77

The TR-77 went silent. All, except for the metronome, which sounded normal. There is a single IC in this machine, a Motorola MFC6040 "electronic attenuator" (essentially a VCA) that is switched on or off by a flip flop on the control board, triggered by the touch plate. This action also simultaneously resets the tempo oscillator, so that at a touch, the machine should spring to life starting from the top of the pattern (essentially the pattern is always running, you don't hear it when the VCA is switched off). The metronome sound is added to the audio buss AFTER the VCA...

The CV in measured 1.4 volts despite triggering the flip flop - from my previous investigations I knew that wasn't right. Audio was present at pin 3, but barely audible at pin 5.
I managed to find a replacement part at great cost. Apparently these things were also used in certain pinball machines. Sure enough, replacing the IC fixed the problem. Now the voltages changed as they should, and the drums became audible again.
Pin out for the MFC6040: note the transverse ridge across the body from pin 1 to 3.

This system enables the celebrated "fade out" feature, where the CV is controlled by a capacitor discharge, rather than an abrupt switch, allowing the drums to fade out just like on a record! Priceless.

Wednesday, 1 May 2019

Roland SH-2000

This synth came out in 1974, Roland's third synthesiser, after the SH-1000 and the SH-3. An analogue monosynth with minimal real time programmability, it features 30 preset voices selectable via flip switches that run along the front panel below the keys. Selecting a voice will override any selection to the left of it. Each preset flip switch selects, via a complex matrix of diodes, an oscillator range (footage), a waveform, vibrato, filter modulation parameters, and an amplitude envelope for each voice. What brings these voices to life is the “Touch Effect” i.e. aftertouch, which can be switched to bring in all or any of the following: Volume, “Wow” (filter sweep), “Growl” (filter fast-LFO mod), Vibrato, and Pitch Bend up or down. There also sliders for filter cutoff (which will affect the preset voice) and resonance (which won’t, except for one setting), and a rate control for the vibrato LFO. Portamento can be switched to a “long” preset time or a programmable time via a slider. There is also a wonderfully wacky Random Pitch effect, where the pitch CV is detached from the keyboard, and determined randomly via the noise generator, at a rate determined by the vibrato LFO - Lost In Space-esque!

So on initial impression, this machine looks like a curio, a bit goofy and definitely dated. But the sound… mmm, the sound…

If you find one of these, the first thing to do is flip ALL the paddle switches up (off), and play the keys. There should be NO sound. If there is, there’s a fault somewhere. When I first got this one, that was the case, and I mention it because I recall not being that impressed with sound of this synth, and that turned out to be the reason why. If there is a sound with all paddles off, that sound will be there with any of the paddles on, and will “pollute” that preset. On YT you can find people demonstrating the predecessor, the SH-3, and on that synth you have manual control over the footages and waveforms via sliders. Here you get a great example of the principle of “less is more”:- notice on these videos how as more waveforms and footages are added, the sound becomes bland and home-organ-like. The presets in the SH-2000 for the most part sound fabulous, strong and musical, with excellently chosen parameters especially for the filter and envelopes. Adding these presets together more often than not detracts from the sound, rather than enhancing it.

So what makes this synth sound so special? The interwebs will tell you it is down to the Moog transistor ladder filter that Roland “borrowed”, but of course it ain’t that simple.

For a start, what kind of VCO is that?

I had a poke around with the scope to see what was happening.

Scope shots below: note upper/yellow trace is always VCO output at connector 1 on the VCO PCB.

Frequency about 1.5kHz, low F - Pin 3 of the 555

Frequency about 11kHz, high F - Pin 3 of the 555

Low F - pin 1 of IC 101

High F - pin 1 of IC 101

Low F - capacitor C102

High F - capacitor C 102

My first impression was that it was based around the 555 timer chip. These chips can make VCOs, but they’re not very good for musical purposes. Advice from the Synth-DIY mailing list suggested the VCO is actually based around the thyristor, the SCR (silicon controlled rectifier) in the top right corner, and that the 555 is there to help linearise or scale the frequency response. It appears that this sort of VCO, as opposed to the regular saw-core reset integrator type as used by Moog and other American companies, were used in a few of Japan’s very earliest synthesisers. Roland’s aforementioned models, along with Korg’s 700, 800 and MS-20. By about 1977, it seems Roland had dropped this idea and changed to using the more conventional type of VCO design, as evidenced by the SH-5 and SH-7. However it was to show up in one more product that I’m aware of - their most famous creation, the TB-303!

The criticisms you will read about these “linear” type of VCOs are usually that of 1) range and 2) waveform purity. If you’re designing a commercial synth that may be required to emulate keyboard instruments, or provide “solo lead” type sounds, these are probably important considerations. If however you are looking for interesting or unique timbres to go with a piece of music you’re working on, they’re much less so. Regarding “range”, often you find that a particular instrumental part really only works musically over about an octave or two anyway. This is why there are orchestras! ;-). Regarding waveforms, this machine surely confounds the premise that you need pure or carefully constructed waveforms to create great sounds. The square wave output from the VCO goes to multiple flip flop dividers (via a LM3216 chip), then the different divisions are simply summed via a matrix of diodes and resistors. This provides a selection of rather coarse stair-stepped sawtooth waves, and pulse waves (see the videos below). Presumably they are rich in harmonics, giving that transistor ladder something to get it’s teeth into.

The audio path is then completed, the signal from the filter, via some level-selecting transistors, going on to the VCA, that itself is controlled by the CV from the preset envelope board. There is no user control over the envelopes, but the aftertouch can be switched to send a voltage to control the VCA directly (this turns out to be of great use when modding for external control).

So selecting a voice preset brings into play various parameters such as footages, waveshape, vibrato, filter modulation and level, VCA envelope and level etc etc.. These parameters are selected by transistor switches, controlled from the matrix (think of the matrix as a hard-wired, read-only memory). About 30 of these switches are based on FETs, and the other 14 or so use BJTs. The FET switches need a gate voltage more negative than about -4 volts to stay fully off. Higher than this, and there is a variable amount of leakage possible. So a switch might fail due to the FET failing, but it could also appear to fail if the gate voltage wasn’t negative enough. If the switch that fails controls, say, the 16’ sawtooth wave, then that is what you will hear with all paddles in the off position. If it controls one of the three “transpose” positions, this footage will be added to the voice. If the FET switch controls an envelope parameter, then the voice might just sound a bit too muffled, or staccato… you get the idea.
FET switch detail

In the case of this particular machine, I was scratching my head at the voltage measurements I was getting at the gates of various FETS - of course, the common link turned out to be the power supply. Although, when I checked it initially as a matter of course, the electros looked fine and the loaded supply voltages seemed rock solid, when I finally replaced the electros and reflowed all the joints many of the faults just disappeared, and the gate voltages were much more consistent. Immediately the sound of the presets improved dramatically. There was still a tone present with all presets off, which I later tracked to a faulty FET switch governing the “Chorus” sound (which is added to the audio bus, along with the “Noise”output, AFTER the point where the wave/divider board output is added).

Having got the thing fully functional, and then being completely gobsmacked at how good it sounded, I then simply had to find a way to control it externally from my sequencers.
After breadboarding up a circuit based on 3/4s of a quad op-amp, and fiddling with some resistor values, I had a Hz/volt pitch CV and 10 volt gate control for notes, and an input tapping in to the Touch Effect, along with a separate control over voice volume.
The incoming Hz/volt CV had to be inverted and scaled. This was applied to pin 8 of the VCO board. A 10 volt gate applied to pin 9 seemed to be the easiest way to implement note on/off, given that it was important to not just trigger the envelopes, but also the envelope CV for the PWM ("Chorus"), see below. This little circuit gives the "Fuzz Guitar" presets their buzzy twang at the start of a note. (There is also a dedicated fixed LFO just below this, for the more conventional PWM effect as used in the Planet and Singing Voice presets. Check the scope video for a demonstration).

Accessing the voltage buss for the "Touch Effect" gives you control over voice volume, as well as the modulation parameters "Wow" (filter sweep), "Growl" (a fixed fast LFO mod of filter), Vibrato, and Pitch Bend up or down. I applied the CV to a wire attached to the wiper of the sensitivity pot, see below

If you do it that way, you can use the pot to drop some of the voltage to ground to help fine tune the effect but if the pot is turned all the way down, there goes your external voltage. The CV, Gate, and Touch controls are buffered by op-amps, which need to be switched out if you want to use the synth from it's own keys and controls. Roland have very conveniently provided a 3 pole, 3 position switch on the back panel (originally for output level adjustment) that can be used for this purpose.

Above: space enough on the back for a small proto board.

Above: re-purposing the back panel switch.

Since there is no dead simple way of getting user control of the envelopes, I also elected to add separate external control over the VCA, via the Volume Touch switch connection. I put this on a switched jack, so that with nothing plugged in, the aftertouch control of Volume behaves originally. Having the minimum voltage here go a little below ground (say about - 0.5 volts) means the voice cuts off to complete silence.
This four-way external control makes it a good fit for a Kenton Pro Solo (remember it needs Hz/volt scale), but it's a lot of fun with one of those Korg SQ-1 step sequencers as well.

Above: there's enough room on the back panel for the 4 mini-jacks.

Videos below: Top trace = audio out
Bottom trace = divider output

Presets: Planet, Singing Voice, Fuzz Guitar 2 and 1

Presets: Clarinet, Flute, Oboe and Bassoon