Tuesday, January 02, 2018

Attack Formation

When I wrote this post about the concept art  I said "I've assumed I can replace all switches with potentiometers".  That assumption was wrong in the case of the Attack control though. In the last post I mentioned that "it's more of a re-design job than a modification".I shall now try to explain briefly, why that is.

The SS-30 Attack control is a switch and according to the user manual :

With this tablet switch on, the rise of the sounds of the Violin system will be softened.
 Softened is the adjective because although it is slower with the switch on the change is subtle. The normal attack is not particularly hard, but the Slow setting is only slightly longer. As we'll see below when I calculate the the rise time, it's not the typical attack range setting that you get on a synthesizer.

Another bow to the string?

Billy Currie put his finger on the reason for this choice in the interview he did with GForce Software. "You were able to create the feeling of a bow" he states. I had a bit of a look around at the synthesis of bowed strings but all I could really come up with was that you need a bit of attack to stop it sounding like an organ. Fair enough, but Curie's point is a bit more subtle than that. The Slow setting provides a softer attack option which his Elka Rhapsody 610 could not. In any case, an even slower attack is not typical of a bowed string. You cannot, for example, slowly increase the volume of each string. Therefore, an even slower attack wasn't needed, because that would stop it sounding like a bowed string (or bowed strings), and a shorter attack would have the same problem.

That's the design choice Yamaha made dealt with, but what about redesigning the attack circuit? Can it be done?

Not easily, no. I had a stab at analysing this circuit before, in the post A Switch To Variable Controls. I wasn't quite correct* in the description there, but I was right that no control could easily be added. Let's see why.


Charging into attack

Keying Drive circuit
 In the diagram above I've highlighted in red the part which controls the attack setting. Tr24 is a transistor which controls the current that charges the 6.8uF capacitor when the key is pressed. The same capacitor discharges through Tr23 when the key is released - this is the sustain control.

The key to the attack time is the charging time of the capacitor. Capacitor charging time is constant for a fixed voltage and is the product of resistance and capacitance. As the capacitance will not change the charging time is controlled by altering the resistance.

The collector of Tr24 is common across all the keys. This node is either open-circuit or grounded depending on the Attack switch.

Attack Switch - BL is the BLack ground wire, which is fed from VR3.
 The base of Tr24 is connected to the key switch. It's normally -7V (via the supply rail and the diode D12) and then switches to 0V when the key is pressed.

The emitter of Tr24 is normally at -7V (again via the supply rail and) so, as this matches the base voltage, normally there's no current flowing through the transistor. When the key is pressed and the base is set to 0V the base-emitter junction acts as a diode and current flows across it and thus the capacitor is charged . What about the collector though?

Let's start with the collector open-circuit. In this case we can treat the base-emitter as a diode and simplify the circuit somewhat. In this case the resistance is circuit with the capacitor is 12K Ohm and 2K2 Ohm,  in series, which give 14.K Ohm in total.

Now what happens if the collector is grounded? In this case we have provided a path from ground to the capacitor directly through the transistor Tr24. When the key is pressed the base switches to 0V as before and a small current is produced across the base-emitter, but now the transistor is switch 'on' and current runs through from the collector. In terms of the resistance to ground the 12K resistor has been bypassed and the total resistance is now just 2K2 Ohm. As the resistance is lower this will produce a shorter charging time, so the normal attack setting is grounded - as you can see in the diagram above.

We can now calculate the attack time for both the normal and Slow settings.

Normal  - R = 2K2 C = 6.8 uF RC = 1.496e-2 = 14.96ms
Slow - R = 14K4 C = 6.8 uF RC = 9.792e-2 = 97.92ms

Note that I haven't accounted for the Vce voltage drop of 0.2V. With that the normal attack time is slightly longer at around 15.3ms.That assumes that the supply rail is exactly -7V and to be more accurate will require some actual measurements. The important thing is that the attack time changes from approximately 15ms to roughly 100ms for the Slow setting.

The long and short of it

Now we know how it works, what about making the attack time more controllable? Can it be adjusted to be shorter or longer? The answer is that the circuit can be modified only to make it adjustable between 15 and 100ms. The RC constant of the circuit is limited by the 12K and 2K2 resistors and no matter what you do with the collector voltage those two impedances will still be there. The attack cannot cannot be shorter or longer than the setting already obtainable, without completely changing components on all the 49 switch circuits.

With 2K2 after the emitter the minimum attack time will always be 2200x6.8e-6=14.96ms. The resistance can't be set any lower so the CR time constant of the capacitor charging will never be lower than that  If I remove the 2K2 resistor from every key circuit, and replace it it with a short-circuit, I could replace the attack switch with a 2K2 variable resistor and vary the resistance from 0 to 2K2 Ohm, giving an attack time of 1-15ms - or thereabouts.
Similarly the 12K resistor into the base of Tr24 sets the maximum attack time. Increasing the resistance between the collector and ground will initially provide an increase in the time constant from15ms upwards, but the higher it gets, the more the total resistance is being determined by the two impedanaces in parallel. In practise the total resistance will never get back to the same value as when the collector is open-circuit and the maximum attack time will always be just a bit lower than the 100ms.
The only way to make the attack time fully controllable, via the collector would be to replace the 12K resistor with much higher value - like 1M - and replace the 2K2 resistor with 0 Ohm link. Then the collector resistance would be in complete control of the CR charging constant.

I don't think I'm going to do that. It would probably take several hours, and that could still be be worth my while, however the risk of damaging a PCB track when there's 4x49=196 joints to de-solder is, well, it's not good odds.

In conclusion, where there's a will there's a way but there's no will to do this one. The attack control will remain as a switch.

Organ Donor

Which brings me to a confession. I bought an entire organ to get hold of two switches.

I may have gone too far this time. It's quite something to discard the wooden case of the SS-30 only to later by a dual-manual organ because I need a couple of switches.

Not just any switches, of course...

These Yamaha switches are just the right sort  - being black and tablet, not too large and the right vintage - to go into my front panel. There will be more about this, and the organ, later.

Wednesday, December 20, 2017

Amazona.de Article

It's nice to see the hard work researching and documenting my journey into the SS-30 get some recognition...

Der Yamaha SS-30

Every now and then I decide to Google the SS-30 and see if anything new has come up. Usually, it's just an auction, but sometimes there is a new article mentioning the SS-30.

Yesterday this came up:

Here's the Googlish translation:

Amazona is a German website dedicated to synthesisers and this article is written by keyboardist Costello. I'm very pleased to see that he has quoted this blog and credited me throughout!

Costello's SS-30 in close-up

Found in translation

The Google translate does a decent job and it's quite readable. Even where it's a bit hard to parse it still makes sense.
"Those with the organ made rock, with the string part epic-spun "pothead music" - no play under 7 minutes."
My favourite quote though is:

"Steven Norgate dedicated his own website to the instrument, where he documents his self-imposed goal of the (very elaborate) MIDI-coding of the string synthesiser"
Don't forget about the crazy the rack mounting!

Costello talks through his early desire for a strings machine which is then rekindled by seeing Ultravox live in 1981. He is a fan.

Then he goes through the competitors and the market for stringers before getting down to a detailed description of the functions - laying out its strengths and weaknesses.

He notes that the violins are better suited to the Orchestra effect whilst the Cellos are surprisingly good without and work as a bass synth part. This is where the selection of which one to send through the effect is crucial.

He then reviews the use by Ultravox and Dave Formula and via him to the current market value and his own recent purchase. With a nod to the unobtanium chips the review is over and I have to say its the best yet - if only because I have covered most of the same ground already! 😉


Just a Limit

Let's take those limitations one at a time:

Shrill violin

I can't argue with that, but there is the Brilliance control and with the Orchestra and with a good measure of reverb it's basically fine. Costello's comparing with the Solina and other stringers - so it's all relative.

Only does strings

True. A few times I've thought about what it would take to try and tap-off a sqaure-wave voice from the dividers. Apart from the ridiculous amount of wiring involved it just isn't worth the effort though. There would be no enevelope either.

Only fast/slow control for the attack

 True, and I've looked very carefullly about how to modify that, but it's more of a re-design job than a modification. I have a whole post about that limitation in the works.

Noisy Orchestra effect

Not a problem unique to the SS-30 though. Gating and eq'ing can help. It might be possible to clean it up with some newer components, but I haven't ever thought it was bad enough to warrant that kind of intervention.

Mono only output

This is something I have thought about addressing already, in this post. It's quite possible, but will need a bit of extra output buffer amplification.

No separate output for violins and cellos.

I've only considered this briefly in the past, but couldn't see much to gain from it. Costello is clearly interested in putting different effects on each voice though, so should I reconsider it? I might. It depends on the levels, but there is an obvious place to feed the voices from, just before the Orchestra sedtion. And I'm going to be rewiring there anyway, so it makes sense...

Orchestra depth

Also, in the comments section, another person previously quoted in the article, 'iggy_pop' notes that "the depth of the chorus modulation (never enough)". As I noted in the Orchestration post the so-called depth setting on the SS-30 is mix depth, not the modulation depth. However, I can add a control for this LFO/modulation depth, but it would not increase the amount of modulation. Do I need to see how it might be boosted as well as cut?  I will take a look at the voltage out of the Phase Mixing section to see how much is being swung on those outputs.
If he was actually talking about the LFO rate, then I can probably do something there as well. The rate is limited compared to the Yamaha CS-15, which uses the same oscillator chip.

It's A Demo

But wait, there's one more treat in store! Costello has created fifteen tracks using the SS-30 to demonstrate the versatility of the old beast. Picking up styles from the likes of Pink Floyd, Genesis, Ultravox, John Carpenter, and other late-seventies influences, he uses external effects from the same era to show what the SS-30 sounds like in a variety of settings.
There is a mixture of full tracks and parts showing off certain features, such as the detune. The use of a good reverb is well noted and is something I have been considering for the long-term.


Costello's SS-30 with a bunch of vintage and vintage style FX pedals

Wednesday, October 18, 2017


Let's look at the Orchestra. Firstly I was interested in a new DIY product from Oakley Sounds which leads me on a diversion around the Roland Juno chorus. Then I talk about the controls I will need for the SS-30M.

Oakley SRE330 - Enhanced Stereo Ensemble and Chorus Module

I just saw this new DIY chorus project available from Oakley Sound and it got me thinking about the Orchestra (AKA Ensemble or Chorus) part of the SS-30 again.
They also used an alpha Juno for the demos, which is also kind of interesting as I have an MKS-50.

The Oakley design has 4 delays which can be stacked or and split for stereo in various combinations, It has two basic settings.
"The three phase setting replicates the action of classic string machines originally from Europe such as the Solina and Logan String Melody."
"The Multimode setting is designed to replicate the actions of various classic synth chorus units and string machines from Japan."
A-ha! Like the SS-30? Not quite, no.

The multimode has three settings
  • Quad Ensemble
  • Dual Ensemble
  • Stereo Chorus

The SS-30 is mono and the quad ensemble uses all four delays, whereas the SS-30 has just 3. In fact the SS-30 is closer to the three-phase mode, albeit with fewer phases.

You may recall from this post that the SS-30 has two LFOs with two phase shifters
So, that's two LFOs, although one is variable and only two phase shifts but the mixing is complex. It's not even clear from the Oakley description how they mix the phases in 3 phase mode, so I can't actually tell how they compare.

Overall it looks like a very nice project and one that I would be prepared to consider doing, but probably after the SS-30M is done.

Juno what chorus the Alpha has?

As mentioned above I have an Roland MKS-50, which is a rack-mount Alpha Juno. The chorus on the Alpha Junos/MKS-50 is apparently a different beast to that on the older Junos. Gordon Reid summed it up as "more flexible but less rich". And rich is the desirable thing, really. You can change the rate of the LFO (or LFOs?) but is that better than being able to switch between two LFOs? I would say the chorus on the Alpha is in keeping with the more restrained and cleaner sounds which were in vogue at the time. But, why is it like that?

The Alpha's chorus is a stereo-pair of single BBD delay lines. The MN3101 is the clock and the BBD is the 256-stage MN3009. And the Juno-6? The same! They use the same exact chips. There are two identical circuits, so you get stereo outputs. The JX-10/MKS70 had more-or-less the same design as Alphas too. So what was different?
The Juno 6 service manual helpfully makes it clear that there are three triangle-wave LFOs. Two at chorus speed of 0.4 and 0.6 Hz and a vibrato which is added at a lower level and runs at 8Hz. The schematic is a bit loopy by I think the same modulation LFO mix is is fed to both channels but out of phase.
On the MKS50 things are not as clear. There seems to be a single signal split into two out-of-phase signals driving the clock. The signal's origins are more obscure though. It appears to be digitally sourced from the CPU via a gate array and a multiplexer.
The gate array has a D-to-A converter built in and its here that the data from the CPU is converted into the LFO signal. On the block diagram, the multiplex is labelled as a sample and hold (S/H) and there is some kind of active filter after that - which makes sense, as there is no continuous CV to make up the LFO. It is smoothed out by the filter, I guess. The notes refer to it as chorus LFO and Chorus Rate CV.
In notes there is also basic diagram of what to expect.

Those arrows seem to indicate some sort of pulse-width is adjusted but I guess this is actually the frequency/wavelength.

It's possible to add a chorus input to the MKS-50 but there's a little debate about the wisdom of doing it and I haven't seen a really popular modification for it. Apparently there was a professional modification around, so enough people must have thought it was worth the effort of tapping into it to make that a thing. There's no info on what they did though.

Conducting The Chorus

In my mock-up, concept art panel design I had replaced the Orchestra section's Violin and Cello switches with faders. Looking back I'm not sure I knew what that was actually meant to do. As there is also a Depth control what was I thinking? Would the depth control also be required? What would the effect of the control be exactly and wouldn't those faders be pointless if the depth was set to zero? I honestly wasn't sure. And then the switches started to misbehave...

I don't  think I had ever looked at the schematic for the orchestra controls in any great detail and when the Violin switch started to behave strangely I realised it was doing more than I had imagined.

Mix Depth

To start with, in the Off position the switches bypass the Orchestra completely. In this mode the depth setting has no effect on the sound at all.

This might seem pointless at first sight. It's generally less confusing if all controls work at all times and there's no context in which they aren't having some effect. In this case the context is whether the switches are on or off and the key here is that there are two switches and two sources that pass through the orchestra effect. There are, of course, four options here and only one leaves the Depth control redundant. Still, it does return us to the question of what my two faders were supposed to be for.

Next, if one switch is on and the other is off there's a way to have one voice dry and one wet. This is why there are two switches.

If either switch is in the On position the Depth control sets the mix of wet (effected) and dry (unaffected) signals for both Cello and Violin. This means that unless one is completely off, there is no way to set the mix independently. You cannot have one at 50% and one at 100% wet, for example. Now I think I can see the point in the faders. If they controlled the amount of mix for each voice then there would be more control. However the depth control would be redundant. And yet there is another option.

LFO Depth

Referring back to the Oakley Sounds SRE330 there is a depth control there too. However this Depth control is not the mix, it's the amount of LFO being applied to the BBD. So, could I rethink the depth control as LFO depth?

There are two LFOs. And then they are mixed together and the result is three signals which go from LF board to the OR board. I could easily intercept these wires and send them through a 3-gang potentiometer. Or I could have a depth control for each of the two LFOs - but then I would have to pull up some resistor and I;m not sure it's what I need anyway.
I reckon the LFO depth would be a good change. Along with the LFO rate settings I thought through in the previous post on the chorus I'm starting to get a better idea of a more fully featured Orchestra/Ensemble/Chorus effect. 

Switches to Faders?

If that's it for the mix depth control, there could be two fader controls to set the mix of wet/dry of each voice. Can I achieve this just by rewiring the existing connections though?

Unfortunately Yamaha took a design decision which makes that harder than you would think. The switches are actually double-pole (two switches which are actioned together at the same time). The first pole, on both switches, switches the incoming Violin and Cello signals to either a mixer which then passes them on to the Orchestra, or to another mixer which then passes them onto the other pole of switches. At this pole the mixed signal either goes directly to the output stage or to the Depth control.
To be clearer if only Violin goes to the Orchestra, Cello goes to the output stage directly and Violin goes through the Orchestra and to the depth control. The depth control is then mix of the dry Violin and wet Violin from the Orchestra.
I need to somehow have the completely dry, bypass signal at one end of a potentiometers and the Orchestra at the other. The incoming signals will then be sent to one or the other, or some mix of both. Then the output of the bypass would go direct to the output and the Orchestra would also go direct and and the switches and original depth control would be completely redundant.

Looking again at the diagram I simply removed the switches and Depth control then drew in what I wanted to do. 

After the Buffer section and before the (pre-Orchestra) Mixing Amp and Mixing sections I have replace the switches with two potentiometers. I will have to experiment with the values here.
After the Mixing section I have labelled the path to the next section as 'Bypass'.
The output from the Mixing amp is split two ways. One goes to the Orchestra board and the other to the mix Depth control. Now the one to Depth control line redundant. Instead the signal back from the Orchestra goes direct to the next section where it is mixed with the Bypass signal.

I will need try this, but it's mostly just disconnecting the switches and Depth controls then adding the sliders. It's not a big job. My main concerns are how well the inputs are balanced between bypass and whether having the orchestra always switched in makes and difference to noise levels.

External Input? 

One idea which has been in my mind for long time - and is again fired by the Oakley SRE330 - is to have an external input on for the Orchestra effect. I makes perfect sense and in terms of the value of the finished article will mean I have a top notch chorus effect available even when I'm not using the SS-30M strings. All it needs is an input jack and maybe a level control. It's something else to try, and whilst I have the wire's loose it make sense to at least try.

Summing Up

In summary:

  1. Need to test out the ideas in the previous post about adjusting the chorus rate and making it controllable - LFO Rate
  2. Insert a control on the amount of LFO reaching the BBDs - LFO rate
  3. Remove the switches and Depth control and replace with a mix-depth control for each voice. 
  4. Test an external input

Friday, October 13, 2017

Live Update

Real-Time Operating System

I have played live! For the first time (well, almost) I went out and played to a small group and it went fine. This was a totally new experience for me, playing not only live but with mostly my own compositions.

Shadow Factory is my studio project

Now it's done I'm considering my next move. I think I want to record some of the tracks I played properly and then I'll probably try and keep up a regular process of updating and writing new stuff for a potential new gig at some future time.

Here's what I took, and you can see the set-up below.

  • MAM - Freebass 383 (TB-303 voice clone)
  • Roland - System 1M
  • Roland - System 500
  • Mutable Instruments - Braids (home made)
  • Expert Sleepers -  Distings MKIII (unused in this set)
  • Make Noise - Function
  • Mutable Instruments - Kinks
  • Mutable Instuments - Ripples filter (home made)
  • Plankton Electronics - Jellyfish V2 Analogue Delay
  • Doepfer - a138m Matrix Mixer
  • Fostex - DE1 Dual MultiFX
  • Behringer - Pro Composer
  • Roland - A880 MIDI Patchbay
  • Arturia - Beatstep Pro
  • Arturia - Keystep
  • Korg - Volca FM
  • Korg - Volca Sample
  • Korg - Beats (modified snare and indivdual outs)
  • Behringer - Xenyx 1202 Mixer

Real-Time Operating System is my live project

Friday, September 15, 2017

2017 - Wrapping up for another year

Time's up! The summer here is over. Yep. Three months and we're done.

I could have done more, but the rest of life gets in the way - which is only right. This is a hobby project and one amongst many other interests. I've also had to carry out a few unforeseen repairs which slowed things down a bit.

There is a bit more warm weather left and I could get a heater for the garage but something else has come up which I need to focus my energy on. I'm breaking the habit of a life-time and playing music live. It's less than a month away now and I really have to get my act together. By the time that's done it will be mid-October and, well, we'll see.

What I have done this year
  • Fixed noise and other issues
  • Built interface and played the SS-30 via MIDI
  • Started a front-panel design.
What I also wanted to have done this year:
  • Sort out the power supply for the interface
  • Fix up a solution for the power supply, MIDI and audio connectors on the back panel.
  • Fit everything in the enclosure and fix temporary front-panel to it as well
If I'd managed that I could have taken the SS-30M inside for the winter and, y'know, actually played it. Well, some of the front-panel controls aren't working very well. The Orchestra section switches in particular are failing. This is my fault for not looking after them properly but it's another hurdle.

What I did that you can't see yet (aaaah!)
  • Bought components for the front-panel 
I'll save that for another few posts on the subject of knobs, sliders, switches and, err, cannibalising old Yamaha organs (ahem). Oh yes!


Let's build this thing!

First board with Lite-On opto-couplers placed ready for soldering.

The first problem I encountered was that my lovely connectors were too big for the wires on the K-boards. I had to go and order another batch of wires and then add new wires from the K boards to the connectors. This took a lot longer than I was hoping but I had no real choice.

Lotsa, lotsa wires.
K1 and K2 boards wired to connectors on the first interface board.

Much soldering and wiring later I had the first board (err) wired up and - somewhat to my surprise - all working perfectly.

The first interface board wired up to the MIDI decoder


I ploughed on to the second interface and suddenly I had a MIDI controlled SS-30!

MIDI Interface almost complete

 So, I'm almost there. I have to sort out the power to the decoder now though. So far it's been run of a bench PSU and I hope to get it running off the SS-30 supply. Thats' another post though.

I meant to get a recording of the whole thing being played, but it's not done yet so I'll end this post on that tantalising prospect.

'Opto-mum' Currents?

This post is a bit behind events now (more on that to come) but here it is anyway.

'Opto-mum' Currents?

Before I launch into building the coupling boards that will interface the MIDI decoder to the key switches I need to actually do a bit a electronics design with equations and everything! The first task is to ensure that opto-couplers have enough current on the input to make sure there is enough current on the output. The input is an LED, so if the current is too low the light emitted by the diode will not be sufficient to turn on the photo transistor fully. In that case the voltage drop across the phototransistor will be too low and the switch will only be partially on. It might be enough to hear something, but I don't want a quiet SS30.
The second task is to limit the current to each input so that the cumulative load of the switches is not drawing hundreds of milliamperes. With 49 keys the potential is there to simultaneously draw more current than the power-supply can handle and blow the fuse. The power rails on the SS30 are rated up to 500mA and if possible I want to use the +15V supply for the MTP8 and the couplers.
Therefore I need to find the optimum current for each optocoupler input - not too low and not too high.

MIDI Decoder 

The MTP8 draws around 15mA for it's own logic etc.
The maximum current per output in the MTP8 is 500mA. But, each group of 8 outputs is limited to 2A which translates to 250mA per output with a possible maximum for 49 keys of 12.25A(!). All that means is it can comformatbly handle large currents. But I want as low a current as possible.

K Boards

The K boards key-switch 'input' is held at -7V through a 3.3K resistor. That means we should getting around 2mA through each switch.

KS-M-C3 is what goes to the key to be switched to ground

The measured value I got from one I picked at random was over that so my calculation must be mistaken somewhere.

And, hey, look! That's my new multimeter.

2.37mA means that the voltage is actually higher. I usually measure around 7.7V which results in a current of 2.33mA, which is much closer. In any case I measured approximately 2.4mA so that is my target output current.

Opto-Coupler Characteristics

Transistor outout optocouplers have many charascteristics but the main one I'm interested in the CTR. The Current Transfer Ratio - expressed as a percentage - simply decribes the relationship between the input current to the LED and the output current. This is essentially the gain of the coupler. All I have to do is make sure that the gain is 100% and they will be no loss through the output, thus mimicing the elecro-mechanical switch I'm replacing. But, what if the gain is a lot greater than unity? Does that mean that somehow the current through the going to be higher? Err, no. The current cannot be any higher than the 2.4mA I measured, but you can have a CTR above 100% because the input current can be lower than the output. This is the kind of coupler I need because if I had to have 2.4mA available for every input I would need 118mA to cover the entire 49 keys. I want to target 50mA or 1mA per switch. 

When I set up the single octave test I had just grabbed some opto's from Maplins - Vishay IL74s - without looking too closely at this CTR characteristic.
The IL74 - data sheet here - has a quoted typical CTR of 35% for 16mA. But the datasheet also shows that it ranges above 100% when the current is greater than 20mA. To be honest I worked through the datasheet and it was a very long and boring process which required this guide on how the CTR graphs should be interpreted. Vishay take the long road and I'm not interersted in that here, so instead here's the shortcut.
I was using 1K current limiting resisistors with a 15Vsupply in that test. The input current (If) was then measured as 13.34mA (the voltage drop across the resistor being 13.34V). A lot higher than I was looking for but that comfortably switched the keys on. The effective CTR in that case was then (Ic/If) 2.34/13.34 x100 =  17.54. Experimenting with other resistors and supply voltages I managed to get good results with 9V through 1K2. This was 2.59mA If and 2.08mA Ic giving 80% CTR. However I couldn't do any better and I want to get the input current down to more like 1mA. This would require a CTR of  234%.

The LiteOn LTV-847 has minimum CTR from 50% (for 5V/5mA) but a maximum of 600%, which seems much better than the Vishay part, and they are quite cheap, so I ordered these.
Avoiding the datasheet again I worked through a series of resistsors from 1K to 10K and measured the input current and output voltage drop. I was looking for the point where the input current is low but not so low that the output current is throttled and the voltage drop across the output collector emitter is rising.

Since taking these measure I seem to have lost the info what the supply voltage was. Was I used 9V or 12V or what? Ah well. When I go the 6K8 resistors I decided they were too high and although the voltgae drop shoudl have been relatively little i t was much higher. May be I changed the input voltage? In any caseI decided to change to 5K6 and as I had enough 'in stock' I used those instead. It was time to start building!

Thursday, August 31, 2017

Billy Currie talks String Synths and Ultravox

Big, exciting updates on the SS-30M progress coming soon!

But in the meantime, here's a very interesting video from GForce Software. Billy Currie (of Ultravox fame) talks about stringers with reference to the Elka Rhapsody and SS-30 in particular. He describes how both were used on the smash hit Vienna for different parts, their relative strengths emulating real strings and the feel of each.The video is to promote the Virtual String Machine/Re-Strings plug-in emulation of strings machines,

In the video Billy is at the GForce studio and is playing the mother of all string synths - the Freeman String Symphoniser.I might have to write something about that one day (and correct a few of my mistakes).

I've written some commentary on what he's saying as it's just a general chat really and some of the points fly past pretty quickly. I then took the opportunity to write an analysis of how the SS-30 and Rhapsody were used on Vienna "of course you've got to talk about Vienna".

So Much For Solina

Ultravox were doing experimental things so they decided to get a strings machine. I'm not sure quite what that means, but I think this would be part of a general move towards electronics. It was the Solina they got first though. Billy doesn't like it because it was 'so weak'. He was not impressed! "Very middle of the road" and a "Mantovani" type of sound.

So then they got the Elka Rhapsody. The Elka had an "emotional" character. This matched his interest in German classical music gained at music college - Schoenberg, Bartok etc. And by extension then Kraftwerk and the "emotional and lost" feel. In Particular on The Ascent - from Rage In Eden

This string's something to me. 


For Vienna the SS-30 was used on the chorus because it was violinistic (that this a word, btw). What Billy wanted was the right envelope because as a violin player he wanted to play it like a violin.  It was obviously to Billy's liking when compared to other stringers for this reason. He sings and mimes the way a violinist would play to better describe the effect.
As he says: "You could do a very slow fade in" and "You were able to create the feeling of a bow". The SS-30 attack is actually quite short, even in the 'Slow' setting, but we're dealing with the articulation of a bowed string, and not a slow fade in.This was in contrast to sustaining the notes as suggested to him later by the guys from Mellotron (Streetly?).

I think it's quite hard to hear what he's talking about on the first two choruses though. He's specifically referring to the melody, which is picked out on piano. The strings are following the piano part but the piano dominates your attention. The strings are more washy. You need to pay a bit more attention to notice that they are following the piano. I think that's because they first sustain into chorus and then out to the "Oh, Vienna" line. You are somehow distracted from the way they articulate in the melodic phrase. The final chorus is more definitive though and this time you are left in no doubt.

In a way it's similar to how ABBA used the SS30 on Gimme Gimme. It's right there, in the hook, but as an undertone. 
In this video you can see Billy playing the part on the SS30, which gives you a better idea of what he was doing - the sound is quite poor on this though.

He doesn't really mention the bridge into the chorus, which is arguably a slower attack. This is more of a swell though and I wonder if the volume pedal was used here.

Middle String Section

 The middle section is preceded with an instrumental version of the verse. The piano takes the lead here with a single high pitched sting over the top. It's not clear which stringer but the higher parts tend to be Elka.
Then the middle section proper starts with a cello from the SS30 - "Great sound". This is the simple descending figure underpinning Billy's violin part. The attack is slow and it tucks in behind the beat to counter the synth bass line. 

I have to agree that it does sound very authentic. Very wooden.
 Then the orchestral part which comes in next is the Elka. I assume this was chosen here because of the feel. The sound is somehow 'shinier'. And then you can contrast it with the SS30 which comes marching in for the climactic final chorus. The SS30 sees the track out with one last sustained chord.

There you have it a pop masterpiece helped along by the SS-30.

Tuesday, August 01, 2017

Connecting you to the Switch Board


I have found the solution to my key switching PCB worries. After the quick lash-up I did on a breadboard I started to fret about what kind of permanent, solderable board to use. A standard Veroboard or matrix perfboard was the obvious answer, but only with a lot of soldering. I didn't want to have to cut dozens of wires and tracks and link everything together by hand, or rely on solder links. I also wasn't keen to design custom PCB. Apart from the costs the learning curve for creating a PCB is not one I wanted to commit myself too. If only there was a PCB just like a bread board, I thought. Well, apparently, there are! There's several on the market in fact, but I found one that matches my needs the closest.

This is the SparkFun Solder-able Breadboard - Large PRT-12699

"a real FR4 fibreglass board, with soldermask, plated through-holes, and the layout duplicates the connectivity of a solderless breadboard"

You can lay this board on top of a breadboard and insert the components through the PCB to test everything as normal. Then you can lift the PCB up with all the components already in the right place and you just have to solder them in.


  • 63 split rows
    • Like the breadboard there's "twin rows of 5 holes each, spaced 0.3" apart, to accommodate DIP ICs."
    • And there's an additional line in between the rows, but I might not need that.  
  • The holes are 0.1"/2.54 mm pitch (apart)
  • The hole size is 0.040"/1.016mm
  • The rails can be hooked up at either end 

Board Layout

I will need two of these large sized boards which will accommodate twenty-four and twenty-five key switches. With 16-pin, four-channel opto-isolators there will be three per octave (with one extra on K1 for the extra ). That's six 16-pin ICs per board needing (6 x 8) forty eight rows. K1 will need an additional single-channel 4-pin device. This will leave room to spare on both boards.
I might use some of this extra space for power rail circuit to the MTP8. A regulator is all that's needed if I can use one of the rails from the SS-30 supply. But that's another post...

The design then is simple. Every other pin on the opto-couplers will be an input or output wire and the alternates will be a resistor to one of the rails. What could be easier?

I have created a diagram of the layout in Excel  - here's the first board. Note that as this has 25 keys I'm going to reuse a spare connect on the connector and then have to cut a track (column h, row 9) on the board and link it to the correct one.

Getting A Good Connection

The only construction concern I have is how to get the wires to the board. The 8 strand wires from the K boards are around 0.7 mm with insulation but the conductors are approximately 0.5mm. That's probably 24 AWG or maybe less.

The options are:
  • Solder wires directly through the PCB holes 
  • Use some sort of pin or tag soldered to the board to solder them to 
  • Terminal blocks to hold the wires (screw or spring clips)
  • Use some sort of IDC connector
  • Solder the wires to another smaller board which is connected to the main board with some sort of  male to female headers.

Wire through the PCB 

The simplest and easiest option is to push the end of the wire through the hole on the PCB and solder like a component leg. The main drawbacks of this approach are that if I need to de-solder it there is more stress on the holes. Also the wires will probably be subject to stresses during construction and maintenance which could break off and needed re-soldering - as I have seen time and time again on the original boards. I could probably mitigate the stresses quite a lot by threading the wire through the holes. Including the rails there will be 5 available holes to go through and the stress on the join between wire and solder should be eliminated altogether.

In one's cups

On the K-boards and G boards the very fine wires are soldered to pins with cups (or 'solder wells') in them. The parts list is a bit vague here and I don't think there are even listed.

Here's a satisfyingly detailed piece of vintage instruction on  soldering such terminals.

Digikey (for example) list a part like this as PC PIN receptacles of type solder cup . The solder termination is like those you find inside connectors (DIN, Centronics, etc. see video above.

I can also use the threading technique to remove tension so there is double the advantage over simply soldering through the board.

One downside of this approach is cost. The ones at Digikey come out at over £40 before tax for the number I need. I can also see the potential for a lot of work there too. 

Terminal blocks

Avoiding soldering the wires altogether is another option. Screw and spring clamp terminal blocks could be fitted to the boards and are relatively cheap. Strips of 12 can be had for as little as £1.70 before tax.

This makes disconnection a very easy job compared to soldering and again there is strain relief possible.
 On the whole though these are not my favoured choice. The thin wires from the K boards may not hold in the clips very well and I'm not a fan of these sorts of terminals. That said, the MTP-8 uses these and I'm not currently planning to replace them.

IDC Connectors

IDC connectors are a possibility because you can get the wires into the connector without soldering or paying for fancy crimp tools*. For the kinds of wires in use this might be a good option. As the Molex notes, err, note:
"overcoat seven strand cable is the easiest and often performs as well as solid wire."
Now, because each connection is going direct to the optocoupler ICs's pins via the board, and I don't want to cut any tracks or add bunch of link wires, it will need to be double spaced (2 x 2.54mm) 5.08mm. Most IDC is single spaced at 2.54mm though. If you want 5.08mm spacing the pins are bigger because it's designed for higher current purposes. This is a drawback of using a prototype board. I'm looking for a connector that will match the board not designing the board around the ideal connector.

The photo above is 5.08 pitch but is for 18-22 AWG, so it's probably a bit to large for the wires.

 I can solve this by getting one that is double length and then using half the contacts though. That then creates an issue of sourcing the right part as the longer ones with more contacts have more limited options. For example you might find a 24-way single row IDC female header, but it's probably for the wrong wire size, or no-one stocks it in quantities under 100. Also some of the more fancy solutions will be less cost effective if they are 50% redundant. For example there are AVX parts in the 8284 series which are almost ideal but the total cost would be over £100!

* Why are crimp tools so eye-wateringly expensive anyway?

Other options

There are other options including replacing the existing wires with ribbon cable somehow so that it's a snap to plug then in and out again. I really don't think there's a better option than those above though. 


The MTP8 has terminal bloc connections so I don't necessarily need more at the other end for those wires. Also, I will use fatter, stronger wires for those connections so that will probably be fine to solder through the board holes directly.

The thinner wires from the K boards are the bigger concern.
They are already soldered at the K-board end, so I would prefer not to do it at the other too. Once they are soldered in there's no quick way to disconnect if needed.  However, because they are so small, even after tinning the ends with solder then may not hold in clamps very well. I would probably have to add a pin to the end of each which will be tedious to do. 
Therefore IDC is the best way to go (short of crimping) and I eventually decided to get these connectors - Autocom from Stelvio Kontek - available from RS Components

Autocom HE14 90° IDC Connectors

  • Single row IDC connectors with right angle exit for flat or single wire cables
  • No special preparation of cables required
  • Can be assembled by hand
  • Double blade contacts
  • Rated at 3A/1000V
  • Flammability rating UL94 V-0

12-way Autocom Connector

These come in the correct pitch of 2.54mm and in single rows. Individual wires can be inserted in each hole and then the two halves are simply squeezed together.

Any 2.54mm pitch header would probably work but they come with their own range of headers which key in to the connectors.

Concept Art

Here's a mock-up of what I want the panel to be like.

I've assumed I can replace all switches with potentiometers. I've opted for slide potentiometers instead of the tablet switches except for the Attack setting which I've made rotary in common with the Sustain.

I've grouped the Cello and Violin settings together, which differs from the original where the tuning and vibrato setting are followed by Sustain and then keyboard split and then the switches.

This is just a first pass at the design concept and there is a list of problems. The logo is probably too large, there is no power switch, power indicator, or MIDI activity LED. There may be other things I decide to add during development too. The whole layout might change again to get a better balance of elements, or some other whim of mine.

Nonetheless this is roughly what it will look like.