Tuesday, November 11, 2014

I Dream Of Wires

A great film and well worth watching even if you're determined not to get sucked into that whole different mess of wires.


Also a Garry Numan song and a made up (by the NME) band.

Anyway, as I may have pointed out once or twice there are a lot wires in the SS30. A lot.

I've had a couple of sessions since the G boards were brought back to life just rewiring the K boards.

First off, two of the G boards where disconnected from their corresponding K boards.

Lots of old, seventies solder. 

Finding a good method for stripping wires as thin as these can be a case of trial and error. Luckily I quickly hit upon using these tweezers to just yank the plastic sheath off.

Strip and twist, strip and twist.
Note the wires are tied with fine string dressing.

Ready to solder

And on the other side, another 49 wires from the K boards to the key switches.

The muck was there when we bought it. Also you can see more loose wires to be reconnected. That yellow one should be attached to AT1.

I had cut all the wires from the key switch PCBs on the assumption I wouldn't need them again. It was also to make handling the whole tangle of boards and wires less difficult. Now I decided to reattach them temporarily to make it easier to test things out.

I removed the switches as they were in a bad way.

And the finished job. A keyboard (if your fingers are earthed).

Monday, November 03, 2014

Why negative voltages and how are the tones created?

This post is just about why the switches use a negative voltage. It's not really important to the whole project but I wanted to understand the circuit and satisfy myself as to why it works that particular way. As I said in an earlier post I was looking for a better understanding of the circuits to help with the key switching problem.

Why are we forced to think about switching a negative voltages anyway?

The way the square-waves are transformed and the switching is woven into that circuit leads to the reason for all this below 0 volts brow-furrowing.

Where is all this negativity coming from?


Your actual schematic of the keying drive

Firstly, what is required of the key switch for each note is a method of not just switching the tones on and off but controlling their amplitude. Each note has it's own envelope (or volume contour if you prefer) going through an attack phase, where it increases in amplitude, then a sustain and release. So the key is a switch to start and end the envelope. That envelope is what drives the amplitude of the tones that you hear.

Trace of envelope from simulated keying drive circuit

Let's be positive!

If we had a positive voltage square-wave output from the divider chips and an envelope that goes from ground to positive - which would be more normal - how would we put them together? I would probably reach for an op-amp, but this machine was being designed in the mid-seventies. Maybe op-amps were expensive then.. Then I'd go for a transistor. I'd use the envelope as the control voltage in a simple VCA (voltage controlled amplifier). But what if my boss is already worried about the costs of this thing and transistors tend to demand biasing and all kinds of ancillary components to get a good result. Can I come up with something better cheaper?

Negativity wins the day

There are lots of solutions to the problem  which don't use op-amps or transistors but the actual one used is pretty neat - albeit one that introduces those fiddly negative voltages.

The solution lies in AC coupling and half-wave rectification.

Conscious coupling

Any AC signal, no matter what the DC offset, fed through a capacitor will create a signal on the other side balanced to whatever reference you choose. If it's not tied (i.e. connected through a resistor) to ground, or anything else, the signal will swing (equally) both positively and negatively. In fact, it will do that if it's tied to ground too. It's voltage swing will centre on what ever voltage it's tied realtive to. So, if you tie the other side of the coupling capacitor to a negative voltage, like -7 V,  you will get a signal which is centred on that voltage.
This is exactly what happens in the tone switching on the SS30. The square-wave output from the divider chips is passed through a coupling capacitor. This output is then tied to the envelope generator output which is -7 V in the off state, rising to 0V in the on state. When the key is up, the envelope is off, the output is tied to -7V so the signal centres around the -7 V level. In the trace below you can see the effect of this.

AC coupling - ties to ground

The lower trace is the source square-wave from the divider chip. It is 9Vpp (peak-to-peak) and -4.5 V offset from ground.  It is clearly all below the ground rail. The upper trace is the output on the other side of the capacitor, which is tied to ground. As you can see it's centred around the ground level.

You will also have noticed that the wave has lost it's square shape. I won't go into that in detail here but it has been high-pass filtered when passing through the coupling capacitor. More intuitively this capacitor has the effect of sloping off the neat, flat edges to rising and falling slopes as it charges and discharges.

The next trace shows what you get if the output is tied to -7.5V

AC coupling tied to -7.5V

In this trace the signal is all below the ground rail - like the original signal

The effect of the envelope on the signal is to shift it either completely below or partially above the ground level. That's all very nice, but the signal is never 'off', so what's the purpose of that shift?

Give us a (half) wave!

If you send a DC-balanced square wave through a diode you still get a square wave on the other side. The difference is that you only get the positive half. This is called a half-wave rectifier, which you may already be familiar with. Only the positive side of the signal makes it through the diode and the negative is blocked.
The traces below show this.

Square wave coupled and rectified

The lower traces are as shown before and the top trace shows the half-wave rectified signal.

By now you may have worked out that when the switch is off, and the envelope signal is -7.5V, tying the AC coupled signal well below 0V, the tone will not get through the diode. And furthermore, when the switch is on and the envelope rises to 0V a half-wave signal will make it through the diode.

So, we have a kind of switch controlled by the envelope. It doesn't matter that the signal is chopped and reshaped in this process. In fact we want to reshape it anyway. As noted in the magnificent Sound On Sound Synth Secrets series by Gordon Reid :

The unmodified waveform produced by a real string is similar to a sawtooth wave, so it's no surprise to find that the waveform selected by the designers of electronic ensembles was always a sawtooth or a similarly spiky waveform.


This is a simplified and cut-down diagram of one of the two circuits which are employed for each tone that goes into to make the note. The two tones share a common switch envelope, each have one of these circuits and are mixed at the end. There's more to follow that diode but this is the essence of the shift and rectifier which makes the notes go on and off.

Simplified tone switch example
Vss is -7.5 volts and the square wave is as described above.

This is the actual circuit for one of the tones from the G1 board.

G1 board tone switch and mixing example

There are two tones being switched and then mixed at the end , on the right of the diagram. The two tones come from the dividers at the top left. C1 is the coupling capacitor. The switching voltage comes in at the bottom left. Note that C1 and C2 values are specified here. These values vary as the tone frequency increases across the four G boards. The capacitors are affecting the shape of the rectified signals, so they are tuned to get the desired wave shape, and therefore sound, of each tone.

Just one last thing on switching negative...

That's nearly it, but there's is one more, unexplained, thing. If the output of the divider chips is passed through the decoupling capacitor then why are they negative going? As I noted above it doesn't matter what DC offset you start with; the coupling capacitor balances the signal to whatever reference you choose on the other side. So, why not start with a positive going square-wave?

Perhaps it's because you need fewer power rails. The amount of wiring in the SS30 is already extreme so limiting the power cables around the place is probably a good idea. The G boards use a single -26V supply with on board regulators to step it down to

Monday, October 20, 2014

More power?

The MIDI interface needs a power supply between 8 and 35 volts. The supply current needs to be 15 mA for the interface itself plus whatever is drawn by the output loads.

The loads are connected to a common positive supply switched to ground by the darlington pair drivers.

If the driver voltage is 15V I'd need the 24V rated Coto reed relay which has a coil resistance of 2KOhm. That would be a current draw of (15 / 2000) 7.5 mA each. Assuming worst case of all 49 relays being on at once that would be approx 368 mA.

That's worst case but the question is this: Can the existing SS30 PSU be used to power the interface?

The SS30 PSU has a +15V rail (this is all carefully worked out, you understand) which is rated up to 500 mA. 

What's not clear, is how much of that 500mA is used by the SS30. I know that the +15V rail is not used for the K boards (-7V) or the G boards (all voltages are derived form the -26V rail) but it is used on the LF, OR and F boards. If the interface took the best part of 400mA (absolute maximum) is it likely the SS30 is using under 100mA?

I'm going to have to get the whole thing working and then measure the current on that rail before I can decide whether I need another power supply.

Friday, October 17, 2014

G Board Restoration

Here are the SS30's main PCBs.

It's about 50% wires!

G boards on the left, K, boards at the back and the rest in the middle. During the hiatus I'd cut out a piece of MDF to fit everything to.

I'd checked the power was okay so I decided to start with G boards and see if the oscillator and tone generator ICs were all working. It's a good place to start as all the sound originates there so anything else would be hard to test until they were verified. Also a faulty tone generator chip might be difficult to replace. Yamaha said they had spares when I got the service manual from them way back when, but would they still have these chips? I;d need to find out early on.

I took everything out of the case to gain access.
Don't worry, I've got a wiring guide.
I prodded the scope probes around the G2 board, which is the top most of the stack, and straight-away found that there was a problem. Only one of the tone generator chips was outputting anything. The other IC had no input clock. I wasn't surprised, a quick look around the G boards and I could see several wires that had worked loose at some point.

-26V and -9V supply rails broken on G4
 These wires were all easily fixed, although I did have to check a few more than once on the schematic and wiring diagrams. One group of wires in particular seemed to be linked in a circular fashion. I'm not even sure what they do - something to do with suppressing noise I guessed - so I decided to move on and come back to that mystery later.

I checked the traces again and was disappointed to find that there was still no output on the tone generator ICs. Now I was worried. Damaged ICs would slow things down and a potentially be expensive. However, I've spent enough time fixing electronics to know that you need to be careful not to just swap out an expensive IC because it's appears faulty.

By the way, the SH-101 I'm also fixing does seem to have a broken oscillator chip, but I double checked everything around before condemning it. When I get the replacement I'll still be nervous until it's in and working though.

I went all round the circuit and traced the problem back to the oscillator on the G3 board. There are two identical oscillators built from discrete components, one on G3 and one on G4. This is great as I could A/B compare them to see where the problem started. The first problem I had was using my scope! I was comparing two wave forms and getting one at the right the frequency (around 500KHz) and the other was... well it seemed to be beyond the range of the scope! It was already late, so I went to bed.

Spot the mistake? (the chocolate wrapper is not a clue)
The next evening I started again and got the same problem. This time I went through and systematically worked out that the scope was not showing me the right trace if both signals were connected. I'd got the trigger set-up incorrectly. In the picture above you can see that only channel A is being used as a trigger source. The oscillators are not synced so triggering one from the other gives the problem you can see in the photo.
Once I triggered each input off it's own signal I got two comparable traces. The night before I'd quickly ascertained that the signal from the G3 oscillator was DC shifted negatively by several volts. Having excluded any other issues I was sure this was the problem. It's supposed to be between 0 and -15V, so what was causing this anomaly?

G3 Oscillator being negative.

I eventually disconnected a cable - which I'd just reconnected - that feeds the clock from G3 to G4. Now the signal was biased correctly. So, it was something on the G4 board.

There's something
 When I was fixing that cable I'd noticed this capacitor on the underside of the G4 PCB. It was obviously a modification and didn't appear on the schematic. It also wasn't present in the same position on the G3 board (where the G4 clock enters that board). So, I took it off. Now everything worked perfectly. This cap was clearly an afterthought. It was pulling the signal too low and served no apparent purpose. I suspect that once everything is running I may find out what it was for, but that bridge can be crossed if I get to it.

Test Points 1 and 2 (A and B triggering!)
So, I could now see clean, strong signals from both of the first two test points on the schematic, indicating all was well with the oscillators.

The next job would be to reconnect the wires from K1 and K2 to G1 and G2.

Thursday, October 16, 2014


It's been an exciting couple of evenings (yes, that's how I roll) as I've had the actual SS30 hardware out and powered up for the first time in years.

I had done some work a few years back that didn't make it on to the blog. I had put the power supply into a temporary enclosure. It's never going to fit in the rack case so one day it'll get it's own, proper enclosure.

The original back-panel (upside down), power switch and lamp

 Which is a re-purposed carry box

Inside you can see everything looking good

Note the Japanese characters on the PCB.

 I checked all the rails and was relieved to see them all looking healthy

But what about the synth itself?

Tuesday, October 07, 2014

Switching negative voltages - Concluded?

Keep it simple

 Back in this post http://ss30m.blogspot.co.uk/2009/04/how-are-you-to-switch-negative-voltages.html I was talking about the issues with switching negative voltages with the simplest circuitry possible.

The guy at J-Omega recommended opto-isolators and I like the sound of that but in the interests of trying to limit the number of pins and foot-print (can't we just use three pins and keep things small?) and the price (transistors are cheap!) what, if any, are the alternatives? Also, I don't know about you, but despite studying these blighters at university I can never quite remember the exact rules governing their use. In particular there are precious few examples with negative voltages and certain factors, that don't come up when everything is above the ground rail, were obscure to me.

Transistor biasing basics.

It seems almost counter-intuitive but you can't switch a negative-ground current with a positive-ground current. The problem is that to turn a transistor switch 'off' the gate (or base) voltage must be lower than the lower end of what you are switching. This is as well as having the 'on' voltage higher.
Normally switching a positive current to a ground is a doddle because your 'off' voltage is ground. No current flows (in enhancement type FETs, anyway) because there is in-effect no bias. When you apply a positive voltage the transistor is biased and the switch is 'on'.
With a negative current to switch the 'off' voltage has to be lower that the lowest voltage. In this case it will have to be lower than -7V (or thereabouts). Because the MIDI interface drivers use Darlington pairs which themselves are driven on from TTL logic levels there is no way to generate an off voltage that is less than 0V.

Previously I mentioned analogue switches which can (in some examples) do this kind of magic. Normally analogue switches are subject to the same limitations as laid out above and the logic switching voltage must be between Vss and Vdd. I referred to the Modern CMOS Circuits Manual book which mentions that some devices - the 4051 and 4053 - have a logic level converter. In effect a 0-Vdd input logic level can be translated into a Vss-Vdd level internally. Those devices are multiplexers though and take a binary 3-line logic input to decode which of the 8 switches to enable. This is no good for this design as we have a line per switch and moreover need to be able to switch all the keys on or off independently.

In summary neither analogue switches or simple transistor circuitry can be used to switch a negative current. I'm glad I checked though as it would have bothered me. I actually spoke with a few colleagues who are full-time electronic engineers and they all initially thought it must be possible, only to conclude that it wouldn't be. It's such an unusual situation that it's easy to get caught out.

Reed relays?

Metal Vs Silicon

 So, opto-couplers it is! Or maybe not. One suggestion a  colleague gave me was reed relays. The main advantage of a reed relay would be that in electrical terms the mechanical switch contact in the keyboard would be replaced by another metal contact switch and not silicon. Introducing silicon might affect the circuit in some way. Perhaps. Or perhaps not much, but the simplicity of this solution is attractive and it might remove some possible risks.

Slim Jims

 One other nice thing about relays is that some come in narrow, SIP (single, in-line pin) packages. So you can see that they could be lined up neatly and compactly. This Coto part http://www.cotorelay.com/datasheets/Coto%20Technology%209007%20Spartan%20SIP%20Reed%20Relay.pdf is 5mm wide*.With a maximum of 13 switches per line (from the four K' boards) and assuming that I would stack the boards vertically, that would be a minimum* of 6.5cm per board, which is around the length I would want it to be.

* Actually they are 5.08mm wide which is 2 x 2.54mm, the standard pitch for through-hole PCB components and boards. At a minimum you could sit them side-by-side, on adjacent lines on a strip-board with no gaps.

Tracking out

 The only down-side of this package is that that the coil pins are on the inside (pins 2 and 3) so you can't as easily track to the outside edge of the board.

To be clear, I'm working on the assumption that the board will have wires from the MIDI interface going in one side and wires to the K boards on the other. Because both of the coil pins are on the inside of the device if you line then all side-by-side there is no easy way to get one of those pins out to the edge.

It would have been great to just use veroboard for this and avoided any kind of tracking out. It's frustrating but they all seem to be designed like this so it looks like I will still need a custom PCB.

Quick enough

 There is one potential downside - timing. This Hamlin part http://www.hamlin.com/specsheets/he3600%20revised.pdf , which is currently the cheapest at Farnell, has a maximum turn on of 1ms. The next is half that at 0.5ms. To be frank I don't think the SS30-M is going to be played at high tempos and the in-built attack and release are already quite slow. A millisecond is not going to be noticeable.


So, reed-relays it is.Is it? Am I decided?
I think so but I need to work out the PCB and think about the options a bit more before deciding once and for all.

Wednesday, September 24, 2014

Tone Generation

Whilst looking for some clues on the key switching problem I came across this.



The Internet Archive has a mission to provide "universal access to all knowledge". Amen to that!

The document seems to have come from here : http://www.loscha.com/scans
which has a number of other interesting synth related documents, including another version of the IC guide book.

The manual contains details of the two chips used on each tone generator board. Ten years ago (blimey) I wrote about the polyphony. I mentioned in passing that the SS30 has two oscillators per note. This post expands on that and, although it still isn't the whole story of how the sounds are generated in the SS30, it is where it all starts.

The SS30 tones start with a pair of oscillator circuits built from discrete parts. These are free-running VCOs with control for the pitch and detune. They share a common master tune and vibrato input and one also has the detune control input . The output frequency of these VCOs is much higher than what you will end up hearing though.They both produce 500 KHz signals.

These two signals are then feed to a pair of YM25400 Digital Tone Generator chips as the master clock input.

Each cascades of these Digital Tone Generators output a divided down clock which is passed on to next generator board for further division. 500KHz - 250KHz - 125KHz - 62.5KHz

The YM25400 derives 13 tones (an octave + 1, C0 - C1) from the master input clock. Each YM25400 then feeds a further pair of LM3211 frequency divider chips.

In summary: the G (tone generator) boards convert a pair of master clock inputs into to a pair identical octaves from two YM25400s. These octaves are then divided down further by a two pairs of LM3211s. This creates another identical pair of octaves one octave lower than the first pair. You end up with two groups of 26 semi-tones (2 x 13).Only G1 uses the extra semi-tone to give octave  + 1, the other boards discard the extra C tone and just output an octave.

The two octaves are named using the eight-foot-pitch naming convention. The initial octave direct from the YM25400's at the 8' and the one below is the 16'.

By now you may be thinking we're going to end up with not with 49 tones for the 49 keys but more like two hundred tones! And we do, but not all at once. Firstly the outputs of the YM25400s and LM3211s are mixed in pairs. Each tone that come out of the G boards is a combination of two square waves - and quite a strange combination too, which I will cover in a later post. If I can actually figure out what is going on!

So, there are actually half that number, but still double the number of keys. The reason for that is the G boards generate Violin, Viola and Cello tones. The various outputs of the G boards are split, switched in and out and combined in various ways to provide the various options selectable from the front panel. At it's simplest you can play the bottom octave, the G1 board, as violin/viola or Cello. When playing as Violin/Viola you only use one half of the G1 boards output. And when you change the split and play Cello you use the other half of the G1 board. It becomes more complicated on the other boards. G2 is split twice, so you get Cello half way through an octave at F as well as at C, and G3 and G4 don't out put any Cello. It gets very confusing in the schematic but is quite simple in the end.