Wednesday, 16 June 2010

Yamaha CS01

A note to document the modifications made to my Yamaha CS01 synthesizer. I may be selling this synthesizer in the next few weeks, if you are interested in buying it please send me an email.

The standard mod for this machine is to disconnect the resonance switch and connect resonance to one of the breath control knobs at the bottom left of the machine, thereby providing variable control over the resonance amount.

When I first looked at the CS01 circuit diagram, my instinct was to use the breath control VCF and VCA knobs for LFO modulation, in the absence of a breath controller being connected, and use a 10k slider from somewhere else for resonance. The VCA EG depth slider looked like a suitable candidate. Since the CS01 is a single envelope synthesizer, and the envelope generator is always hardwired to the VCA (as well as the VCF), the VCA EG depth slider merely acts as a second volume control and can be reused for something else. Then I thought maybe the old resonance switch could be reused as an EG / Gate switch for the VCA, to allow for filter envelope sweeps without affecting the volume level.

I started modifying the CS01 by connecting the output of the LFO to the breath control modulation knobs. The point marked TP6 on the circuit diagram is an AC coupled version of the LFO output, which ordinarily is connected to the modulation wheel pot. A 1k resistor connects the breath control modulation bus to ground when no breath controller is plugged in. By removing this resistor, the AC coupled LFO output can be connected here as well.

This works and provides LFO modulation of VCF and VCA; however there are problems because the modulation depth is rather low. Observing the modulation wheel circuit for the VCF, I noticed the output of the mod wheel pot is connected to the VCF cutoff frequency input via a 10uF capacitor and a 8k2 resistor, whereas the breath control VCF pot is connected via a 39k resistor. I increased the modulation depth by replacing the 39k resistor with an 8k2 resistor (so that the breath control VCF knob is capable of the same modulation depth as the mod wheel) and then noticed that the range on the cutoff frequency slider had been shifted by the extra load pulling the VCF cutoff frequency input towards ground. So I decided to AC couple the breath control VCF pot output and added a 10uF capacitor in series with the 8k2 resistor.

As an aside, it is interesting to note that the CS01 has no summing amplifiers for either the VCF cutoff frequency or the VCA volume level. All voltage sources are simply connected together with resistors and fed to the appropriate input. What results is a weighted mean for the VCF (modulation wheel, cutoff frequency slider, breath control knob, EG depth slider and internal fc ADJ preset) and VCA (breath control knob and EG depth slider). Therefore any change to the breath control modulation depth will affect the range of the other controls. Going from 39K to 8K2 does not seem to affect the other VCF controls noticeably once the decoupling capacitor is added; however with the VCA it is not so straightforward.

Observing that the VCF EG depth slider is connected to the cutoff frequency input using a 33K resistor, but the VCA EG depth slider is connected to the VCA input using a 39K resistor, I decided to replace the 39K resistor on the output of the breath control VCA knob with a 10K resistor and AC couple the output by adding a 10uF capacitor in series. The resulting tremolo is certainly louder than with the original 39K resistor (which was basically not audible at all) but turns out to be volume level dependent. With the VCA EG depth slider all the way up, the tremolo can barely be heard, and with the slider nearer the bottom the tremolo is much more noticeable.

The cause of the problem lies in the fact that the VCF cutoff frequency input has an exponential response, whereas the VCA volume input has a linear response. Wherever the other controls on the VCF are set, the VCF LFO modulation has an appreciable effect, whereas the effect of the VCA LFO modulation depends on what volume level the LFO is varying around.

However this is inherent in the design of the CS01 and the problem would be noticeable with the breath controller in the original design as well, so I decided to leave it at that. I had shown that the breath control VCF and VCA knobs can be used as LFO modulation depth knobs, rather than reusing one of them as a resonance control, so now all I needed to do was add a 10k pot for the resonance, rewire the old resonance switch as an EG / Gate switch for the VCA and I would have a very capable CS01.

Then I hit on the idea of using modulation sources other than the LFO for the mod wheel and breath control knobs, and my CS01 became rather more interesting. Sitting in a noodle bar late one night (a shout out to Dojos noodle bar in Cambridge at this point), I realised there were at least three modulation sources it would be nice to apply to the VCO, VCF and VCA. In addition to the LFO, the VCO itself and the noise generator would make good mod sources. So I sketched out a system of two three position switches in which the modulation source for the mod wheel (VCO or VCF switchable) and the breath control knobs (VCF and VCA) could be independently selected. Rather than drive the breath control modulation bus with the LFO directly, I would select from LFO, VCO and noise generator with a switch, and the mod wheel pot would also no longer be driven only by the LFO, but select from LFO, VCO and noise generator with a second switch.

This was implemented easily, the VCO output being available from pin 15 of the tone generator chip and the noise generator output being available in decoupled form at the WN (white noise) pole of the waveform switch. I only needed to add a 10uF capacitor to decouple the VCO output and all three sources would be in a suitable form to drive the mod wheel bus or breath control bus.

Now the CS01 is capable of filter FM sounds and noise modulation sounds, and even of slightly bending its square and pulse waveforms away from their original pulse widths (by modulating the VCO with its own output), it is a much more capable machine indeed. With the mod wheel destination set to VCF, it is possible to apply two different modulation sources to the VCF at the same time, by selecting one on the mod wheel and the other on the breath control VCF knob, and get some incredibly complex timbres. Examples include LFO modulated filter FM and noise modulated filter FM, as well as doubling the modulation depth by setting the two modulations to the same source.

Since the VCA EG depth slider is useful in the modified CS01 for increasing the apparent VCA modulation amount on the breath control VCA knob (by turning the EG depth down!) I decided to keep the VCA EG depth slider as is and added a separate 10k pot for the resonance control. Removing the original 10k fixed resistor and associated wiring for the resonance switch, I looked for a suitable source of a Gate signal to provide an EG / Gate switch for the VCA instead.

The TR# output (trigger) of the tone generator chip, pin 9, is at -9V when idle and at ground potential when a key is pressed (note the positive earth, PMOS technology of the CS01). This is the opposite of what I want; the envelope generator varies from 0V at idle to about -2.9V when fully open. The collector of Tr 8 on the other hand, pulled to -9V by a 47K resistor when a key is pressed but shorted to nearly 0V by Tr 8 when idle, is a good starting point.

Clearly this point is to weak too drive a 10k VCA EG depth slider by itself, so I added a PNP transistor (happened to have a few ZTX500 lying around), connecting its base to the collector of Tr 8, its own collector to -9V and its emitter via a 15k resistor to the new EG / Gate switch. The value of the resistor was chosen to make the volume of the synthesizer in the new VCA Gate mode close to the volume in the EG mode with the sustain level on the highest setting.

The final modifications I made were to incorporate the additional waveforms demonstrated in the following YouTube video.

These modifications work by selecting more than one waveform at once on the tone generator chip. Also known on the MOS 6581 SID chip from the Commodore 64 computer, here the combined waveforms actually work as expected. The waveforms are generated by digital counters and/or adders and output to a digital to analogue convertor. When you select more than one of them, the result is a logical AND of the waveforms selected. The most interesting new waveform available by this modification is the result of selecting triangle and sawtooth together. A very complicated waveform with lots of high frequency harmonics present (see the YouTube video for an oscilloscope trace), it can sound quite like a combo organ with a little VCO vibrato applied.

Since the result of combining the square and pulse waveforms is simply the same pulse waveform, I decided it was not necessary to be able to select both square and pulse at the same time. Triangle and sawtooth, on the other hand, should be independently selectable and combinable with each other and with one of square and pulse. This can be done by adding an SPST switch and a diode between N4# and F# for the triangle waveform, and another SPST switch and diode between N5# and F# for the sawtooth waveform. Selection of square or pulse waveforms can be done with the original front panel slide switch. Note that PWM cannot be combined with any of the other waveforms, since the PWM position works by having no waveform selected.

Finally, I made some minor adjustments to the presets onboard the CS01. Having a pitch bend wheel which only bends upwards was a source of some annoyance, so I disconnected the return spring and adjusted the f0 ADJ preset for the lowest frequency range. Now the pitch bend wheel sits in the central position and I can bend down a minor third or up a minor seventh. For some reason Yamaha used a log pot for the pitch bend wheel; anyway that is the range you get with the pitch bend wheel centred! The filter cutoff fc ADJ preset can be adjusted for the desired range on the cutoff slider; with the new breath control modulation resistor it may be necessary to adjust this preset to get the range you want. Lastly, the resonance Q ADJ preset can be adjusted for as much screaming resonance as you want at the maximum end of the new resonance pot. Unlike the earlier Yamaha CS filters, this iG02610 VCF can be pushed well into self oscillation in its standard configuration.

The CS01 in its current form has two three position rotary switches, two slide switches and one 10k pot dangling on flying leads emerging from the back of the case. In fact it has two extra pots as well, because the LFO rate slider and VCF EG depth slider weren't working on my CS01 when I bought it, and as yet I have been unable to get them working again. It would be a good idea to mount these in some sort of project box, once I have found one of the right size.

Further ideas I have had for modifications include replacing the 22k minimum rate resistors for the glissando, LFO and PWM rate sliders with something smaller, to increase the range available on these controls. It would also be good to use the envelope generator as a modulation source for a few more destinations. Connecting the EG to modulate the master clock oscillator for the tone generator chip should be easy enough, and it may be possible to modulate the pulse width in the PWM waveform instead of the usual PWM LFO, although I am still researching how the PWM LFO works and this may or may not in fact be possible from the tone generator chip's external connections.

Monday, 12 April 2010

INET conference 2 and Korg X-911

Well it's not every day you get to meet George Soros.

I have Mr Soros and the Institute of New Economic Thinking to thank for two magnificent conference dinners and 56 presentations by economists and other thinkers of the highest calibre.

During some of the less interesting moments on the first day I sketched out some mods for the Yamaha CS01, which is a synthesizer I have still yet to obtain so I will try them out when I have one. During idle moments on the second day I sketched out some mods for the Korg X-911 and tried them out the next day. Unfortunately they didn't work.

I'm not entirely sure why they didn't work, but it seems to have something to do with the X-911's original purpose as a guitar synthesizer rather than a keyboard based synthesizer. None of the presets have release settings above zero, which is what I was trying to change. I'm going to try a similar mod on the Korg Sigma along with increasing the range of the Synthe VCO detune knob, and hopefully confirm that the basic idea does work, before returning to study the X-911 in some depth.

Friday, 9 April 2010

INET conference & latest synthesizer mod plans

First of all, this is meant to be a technical blog and therefore in general I will spare you the tedious details of my drab, wretched life*.

However, I've just had an extremely good evening at the Judge Business School and King's College, Cambridge and one aspect of it was the solidification of some ideas I've had for modification of a well-known digital/analogue hybrid synthesizer. I just need to complete my own buy side transactions related to this product before going public with my ideas.

What I will say is that there are quite a few 'consensus' views on how to mod this synthesizer out there in cyberspace and I hope to trounce them with my own, as always. Watch this space.

* Wry acknowledgement due to Tom Lehrer. There are other places in cyberspace where you can read of my (mis)adventures if you really want to.

Monday, 5 April 2010

Korg Sigma

A note to document the modifications made to my Korg Sigma. I will be making a YouTube video about my Sigma shortly. It is quite a famous Sigma already; it is the very same Korg Sigma that features in the following YouTube video:

Originally I bought it cheap because nine of the eleven Instrument tabs weren't working properly. I traced this fault to a broken 4013 dual D-type flip flop and replaced the chip.

Sigma is an extremely interesting synthesizer. Essentially a preset monophonic synth, it offers many of the same features as other preset monosynths - a 3 octave keyboard with aftertouch sensitivity, a range of preset sounds (19 in Sigma's case) and various performance controls.

Unlike other preset monosynths, however, it does not feature a single VCO-VCF-VCA signal path with a diode matrix or diode and resistor matrix to select the preset sounds.

Rather, Sigma contains a separate signal path for each of the 19 sounds, and these may be enabled simultaneously with large rocker switches. The sounds are divided into two sections, Instrument and Synthe, corresponding to the type of sounds intended to be available in each section. Imitations of acoustic instruments on the Instrument side and synthesizer sounds on the Synthe side. Be in no doubt, though, this is a 1979 analogue synthesizer. Both sections make sounds that befit such a machine.

Korg provided two separate VCOs, one for the Instrument and one for the Synthe sections, so that two oscillator sounds may be made with Sigma. Korg's VCOs generate four sawtooth footages simultaneously (32', 16', 8' and 4') and different footages proceed to the preset voice circuits for waveshaping, filtering and envelope generation specific to each preset sound.

Page 5 of the Korg Sigma service manual shows the circuits for the Instrument presets, and page 6 the circuits for the Synthe presets. The two VCOs are on page 4 along with various performance features such as the pitch joystick, aftertouch sensor, transpose rocker switches and the unusual Quarter Tone mode.

Each preset sound is equipped with a single potentiometer control for adjusting some aspect of that sound. On the Synthe side they are mostly envelope controls and on the Instrument side a mixture of tone (filtering) controls, envelope controls and one pulse width control. Every preset sound has its own single transistor VCA, an envelope generator and some form of waveshaping and filtering. Here is where Sigma gets very interesting.

The Instrument section has four preset sounds, Electric Bass, Tuba, Horn and Trumpet, which have their own KORG35 voltage controlled filter circuits for shaping the sounds. Electric Bass uses a pulse waveform with a fairly narrow pulse width and Tuba, Horn and Trumpet all use sawtooth waveforms. All four presets have simple envelope generators with very short attack times and long decay times, these are used to control each preset's VCA (a single NPN transistor) and to provide an amount of VCF cutoff frequency envelope modulation. The Electric Bass envelope generator decays to zero so this preset may be considered a percussive sound, whereas Tuba, Horn and Trumpet all feature envelope generators which decay to non-zero sustain levels, and so may be considered sustained sounds.

The other seven preset sounds in the Instrument section have organ-style fixed filters for shaping the sound, and simple envelope generators controlling their single transistor VCAs. Thus there is no dynamic timbre variation resulting from filter cutoff frequency changes in these presets. That's not to say they aren't interesting sounds, however. Fuzz Guitar uses a pulse waveform where the envelope generator modulates the pulse width before proceeding to fixed filtering and the single transistor VCA. Clavi uses a pulse waveform where the potentiometer control adjusts the pulse width.

Clarinet is a square wave with a tone control for crossfading between the outputs of one fixed filter and another fixed filter. Double Reed is a pulse wave with a similar tone control, String is a sawtooth wave with a control for the envelope generator's attack, Flute is a sawtooth wave with a tone control and Hammered Percussion is a sawtooth wave with a control for the envelope generator's decay.

On the Synthe side, each preset sound is provided with a KORG35 VCF, an envelope generator, a single transistor VCA and some waveshaping. A joystick control adjusts an offset voltage applied to all eight VCFs, and some of the presets feature envelope modulation of the VCF cutoff frequency as well. In fact only the 16' PWM and 4' sawtooth presets have no VCF modulation at all; the others either have envelope modulation of their VCF or in the case of the 16' sawtooth S/H preset, the VCF cutoff frequency is modulated by a sample and hold circuit which gives the preset its name.

Therefore, the entire synthesizer has twelve KORG35 voltage controlled low pass filter circuits. This is a lot more than a Korg MS-10 (one) or a Korg MS-20 (one low pass and one high pass), and yet those synthesizers fetch a lot more on the open market. No doubt that is because those machines offer a lot more apparent control over the sound with their variable controls and patch panels. Yet Sigma features 31 potentiometers and two X-Y joysticks over its entire control surface - it is not a synthesizer without control possibilities! Furthermore, Sigma can generate sounds that feature 32', 16', 8' and 4' footages simultaneously. No MS series instrument can do that.

A few things frustrated me about Sigma from the outset. One, the relatively simple VCF envelope modulation possible with A/R or D envelope generators. Two, the lack of VCF aftertouch modulation.

Further frustrations were the lack of VCF LFO modulation and the rather curious Delay Vibrato rocker switch which provides delayed vibrato for the Instrument VCO, and has three associated controls for the purpose (depth, speed and delay), but with no equivalent for the Synthe VCO. Finally, the reduced range of the pitch bend joystick on the Instrument VCO relative to the Synthe VCO.

So I decided, having already 'modified' my Korg Sigma by replacing its original broken 4013 chip with a new one in an IC socket, to go further and do something about the above limitations.

The first modifications I did were to increase the range of the pitch bend joystick on the Instrument VCO, and make the Delay Vibrato rocker switch apply to the Synthe VCO as well. For some reason Korg decided to make the range of the Instrument VCO pitch bend approximately half that of the Synthe VCO. This can be changed by replacing R197 (470K) on page 4 of the Sigma service manual with a 240K resistor, to equal the value of R163. Delay Vibrato can be applied to the Synthe VCO as well as the Instrument VCO by adding a 470K resistor between the output of the delay vibrato pass transistor Q13 and the inverting input of the Synthe VCO pitch modulation amplifier IC24 (pin 6).

To add VCF LFO modulation to the presets that feature KORG35 VCFs, I observed that the Instrument and Synthe circuit boards both have a VCF offset adjustment, which are provided as trimmers on the back surface of the machine. The outputs of these are buffered by inverting amplifiers and applied to pin 2 of the KORG35 VCFs.

The above picture shows the VCF offset amplifier for the four KORG35 VCFs in the Instrument section. By adding in a signal from the delay vibrato LFO here, one can add VCF LFO modulation. I did this by attaching a 10K pot to the output of the LFO (pin 7 of IC10 on page 4) and connecting a 100K resistor between the output of this pot and the inverting input of the VCF offset amplifier above (pin 2 of IC10 above).

The next picture shows the Synthe VCF joystick amplifier and VCF offset amplifier for the eight VCFs in the Synthe section. VCF LFO modulation can be added to the Synthe section in the same way as for the Instrument section, by connecting a 100K resistor between the output of the VCF LFO modulation pot and the inverting input of the Synthe VCF offset amplifier (pin 6 of IC1 above).

VCF aftertouch modulation can be added to the Synthe section VCFs by connecting a 10K pot to the output of the keyboard sensor amplifier (pin 7 of IC12 on page 4 - or after diode D26) and a 100K resistor between the output of the pot and the inverting input of the Synthe VCF FcM IN amplifier (pin 2 of IC1 above). The reason for adding it in here is that the FcM IN amplifier provides an extra stage of inversion. If it were added to the offset amplifiers already discussed, the result would be a decrease in VCF cutoff frequency when the keys are pressed harder. More useful musically speaking is an increase in filter cutoff frequency, hence it is added in here. Applying aftertouch modulation to the Instrument VCFs in a similar manner will require an extra inverting amplifier.

The above modifications have been implemented on my Korg Sigma and I am very happy with the result. Gordon Reid, author of many Sound on Sound reviews of analogue instruments, has seen the mods and given them the thumbs up. There are two remaining improvements I would like to perform on my Sigma.

First, the VCF LFO modulation pot provides more movement in the sustain phase of a note, but it cannot completely hide the relatively simple A/R and D envelopes applied to the VCFs. What is needed is a full ADSR envelope generator, and this can be applied to the Synthe VCF FcM IN input without any further internal modifications of the Sigma.

Second, and in common with other early Korg instruments including the 700, 700S, 800DV, 770, MS-10 and MS-20, Sigma has no VCF keyboard tracking. This is partly because Sigma (and Korg's other early instruments) features linear oscillator control voltages and exponential filter control voltages. To provide a VCF keyboard tracking control, a logarithmic amplifier is needed, such as the one in the Korg MS-02 interface. This interface converts linear (Hz/V) control voltages to exponential (V/Oct) control voltages. The linear keyboard CV output can be converted to exponential using such a log amp and applied to the Synthe VCF FcM IN input as well.

A word about resonance - one comment in the original Sound on Sound retroview of Sigma expressed the author's frustrations at the lack of a resonance control. In fact, each of the twelve KORG35 VCF circuits has a preset amount of resonance designed into the Sallen & Key filter circuit by the component choices made. To provide a resonance control for each of the VCFs would require twelve resonance controls, and a much more complicated VCF circuit for each. Sigma manages to pack in twelve VCFs into its highly unusual synthesizer architecture by keeping each circuit as simple as possible, and therefore I can forgive Korg for the lack of resonance controls; indeed I'm very glad they designed the synthesizer this way because it made Sigma more commercially viable in the first place. Already Sigma cost almost twice as much as the MS-20 when first released (185,000 yen versus 98,000 yen). This may explain its relative scarcity on the modern day second hand market.

To compensate for the lack of resonance controls, one can become familiar with the preset resonance amounts applied to each of the VCF-filtered preset sounds. This can be done either by listening to the sounds or by studying the component choices (capacitor values) for each of the KORG35 filter circuits. The presets having the highest resonance are the 8' sawtooth, 8' square, 16' S/H and Noise waveforms on the Synthe side. The presets having the lowest resonance are the Tuba and Horn sounds on the Instrument side.