Transistor coming off the rails

Bench Press

With the help of the borrowed bench power-supply I manged to narrow down the -26V rail issue to board G3. Without that board the total current was less than 200mA, but with it the power-supply would switch the voltage down instantly. The current was probably lower than normal because the clock signals generated on G3 were off , which in turn meant that the other boards were not active.

Bench Power-Supply On Loan

The next step was to isolate which part of G3 was causing the problems. I started by lifting resistors to the -15V and -8.2V regulators and Master Clock Oscillator but the problem remained. Then I realised that the transistor on the -15V regulator was also connected to the -26V and I would need to remove that too.

Once that was out the problem stopped immediately. So now I had a prime suspect.

Checking the resistance between collector-emitter showed it was blown. Practically short circuit between the two pins. As I don't have a replacement to hand the next step was to connect the other side of the bench supply to the -15V rail and verify that there were no more problems there.


On doing this I was relieved to see that the unobtainium parts - the YM25400 octave dividers - were still running. However, I'm not out of the woods yet. The Master Clock Oscillator on G3 not looking good, to put it mildly.

The effect of this on the dividers that it feeds varies but it's possible that the bench power-supply is the problem here. I will reserve further judgement and investigation till I have the -15V supply back in place.
Perhaps I'll just piggy-back off the G4 -15V supply and see what that does. My HW engineer colleague agrees and points out that the bench power-supply is switched-mode and this could be an issue.

Transistor Sourcing

http://www.radiomuseum.org/tubes/tube_2sa509.html
 
The 2SA509 "Silicon PNP Epitaxial Transistor for audio power (1W) amplifier applications".  As this is a regulator though it's not used in that way on the SS30.
A voltage divider is used in place of the more usual zener diode to set the voltage at the base. The divider gives -9.5V. which means a difference of 16.5V base to collector. A voltage of -26V here would switch the transistor fully on, it being most negative with respect to the emitter.  Whilst if it was gnd/0V the base and emitter would be at the same potential no current would flow. Therefore the voltage  of -9.5V at the base results in approx -15V at the emitter.


It's possible to get the original parts - which are Toshiba - but these are pricey. There are equivalents which are cheaper but these all seem to be in the US and cost a ridiculous amount in P&P. I have found an eBay seller with the right part at a reasonable price, but not much information. I've contacted them to check that the pin-out is the same.  

The answer is blowing in the fuse.

A very frustrating and worrying night last night.

I re-stacked all the G boards and fixed the broken -26V wire. But as soon as I powered up the fuse blew again.

So, I started removing wires and measuring impedance and trying things but after another couple of fuses had gone I gave up for the night, dejected.

It seemed at one point that I had identified which rail on which G board was the problem. But, when I left it disconnected and reconnected some of the other boards I was back to square one with another blown fuse.

From that it appears that no single board is the problem alone. When a few are connected simultaneously the overall impedance drops and the current rises, killing the fuse.

The voltage is lower than -26 but it's always been like that. Back in this post I measured -23.6V

And if the impedance is above around 2K ohm all is well. When I keep adding rails though, it drops to around 1K ohm and the fuse is blown.

Luckily I work in an electronics company and after a quick chat with some hardware engineers (I'm software) a bench power supply was tossed my way and I have another way to investigate.

With a bench supply I can just use this to power the boards investigate and save blowing more fuses. I can set the maximum current to 500mA, or less, and then add rails till it clicks off. I can also measure the current being used for each board and try and work out which is taking the most to help trace the fault. My plan is to test each board in isolation first, checking the current as I go. I should find one that is taking more current than the others - although they will not all be equal, so I have to use some judgement. I was also advised to check for heat. Anything getting warmer than expected is probably the source of the trouble. I can also test all the ICs on each board and hope and pray that none are blown [shudder].

This is all very annoying. I seem to have precipitated this fault when fiddling around with the noise problem which was a silly mistake and should have been diagnosed without all the messing about I did. It's also worrisome because if I've blown one of the custom Yamaha chips the replacement might be costly and difficult, or even impossible.

Don't cut the black wire!

Sometimes the obvious answer is staring you in the face.

I found that I had cut the ground wire to the G boards at some point. Why? I forget. The perils of stretching this project over so long is things like this slip your mind.
Anyway, I found the two ends tied together.

Knot what I was hoping to find

As for why I didn't check that the earth wire was connected to ground before, I can only say I need to think a bit harder!

I reconnected the earth to the G boards and the problem was solved. Just the faint hiss of old electronics was left.

G Board output connectors cleaned and ready to reconnect.

I set about reconnecting the outputs and all was well until I realised that I couldn't get any tones to come out anymore. Nothing from the clock dividers, nothing from the tone generators at all. Then I checked the power supply and the -26V rail had gone. Just that rail though. After checking the PSU (during which I managed to spike the ground and trip my whole house's power, even with an RCD connected!) I could see that the fuse for that line was blown. I replaced it but it just blew again. I disconnected the PSU from the boards and it then held up okay. I couldn't find any obvious shorts to ground - maybe there was something touching somewhere - but I'd run out of 500mA fuses, so I finished reconnecting the outputs and started stacking the G boards back together. During that I found a loose -26V wire too. Hmm!
I've ordered some more fuses so next I will reconnect that wire, test the connections for -26V, check for shorts again and then see if the issue has gone.

If that works then everything should be okay, finally. A fully working SS30 in a rack case without MIDI! And that means I can, at long last, start on the MIDI interface.

In the background I have been rethinking the front panel again. More, much more, on that to follow.

It's oh so... noisy

Annual Apology

13 years it's been since I started this blog. Many things have changed since then but my desire to have a Yamaha SS30 in a rack mount with MIDI is undimmed and just as subject to my whims.

Anyway, 9 months since the last post and 11 months since the last update on the rewiring; where was I?

Bring the noise (down)

Last night I worked on locating the source of noise when no notes are being played. To begin with I confirmed that there is still an unacceptable level of noise but gave up trying to locate it in one place. My scope might show a specific signal but my ears discerned general bleed-through from many tones. Maybe all of them!

After a bit of messing about and reacquainting myself with the boards (it has been a while) I decided to be bold and disconnect the outputs from the G boards until I found the source if the problem. Eventually I concluded that not only are there tones bleeding through every output on every G board - to a greater or lesser extent - but the noise is also present on the earth for each output too!

Once I had disconnected all the outputs and all the corresponding earths the output noise-floor dropped to a more acceptable level. I would like to able to put a figure on the noise before or after but I a) don't need to when I can hear it without any trouble and b) can't be bothered with setting up the extra cabling etc and trying to find an app to do it. As I'm in the garage, as far from my studio as possible on my property, it would have to be a phone or iPad app - but I digress. With the SS30 volume control at 50% and the mixer gain at 0 and the fader set to 0dB I can hear noise. Once the G boards were all disconnected it was not silent, but the kind of quiet I would expect given the SS30 is old and the mixer is no spring chicken either. 

Reconnecting the earths for each part of each G board in turn I could hear the various components of the noise return. Even in the same board the sound of each part varied. This means that as the noise is getting on to the ground track it doesn't come through as strongly at the far end as it does the nearest node.
 
All of which is to say I have a serious problem. But what is the source? What is the root cause? 

The blame game

My initial, gloomy diagnosis is that the vast number of grounding capacitors (four per tone) are all to some extent leaky, and all that is adding up and causing the noise. 
This is an irrational fear though. Why should they be leaky and why should that mean the ground is noisy? There is a bit of a myth in the synth community about re-capping. Sure, tanatlum caps in the power supply should be replaced in case they short. And electrolytics can dry out or leak and I have had to replace one on my Roland MKS-50. However these caps are low voltage and connected to ground. If they shorted out the worst that should happen is that the audio would go down the plug-hole run to ground or be attentuated. So, I should probably not assume the worst.

Therefore, my second thought was that blaming the caps might not quite work as an explanation because it's not a likely explanation. But, there is a more fundamental reason why blaming these caps makes no sense. To understand why,  you need to understand the whole method of generating keyed, string-like tones from square-waves. 

 The tones start as square-waves from clock dividers. That signal is then put through a series capacitor. This cap achieves two things. Firstly the signal is DC biased by whatever voltage is on the other side of it and secondly the wave is shaped into charging curves as the capacitance charges and discharges. This how the square-waves start to become the 'spikey' wave-forms needed for string sounds. 

Offset VCA

The DC voltage on the other side of the capacitor is -7V when the key is up. When the key is down it is 0V.  That means that when the key is up the signal is all offset to less than 0V and the current is all flowing - more or less, as it oscillates -  in the negative direction. When the key is up the signal is centered on 0V so the signal flows half in the positive direction and half in the negative.

 The next device in the circuit is a forward biased n-p diode. When the key is up nothing should get through that diode. With -7V offset on the signal, nothing can get through. Bit at 0V the positive half passes through the diode. In fact the amount can be between -7V and 0V as the attack and sustain controls set the raye of change. This means that the amount of signal gettig through the diode is controlled by the keying circuit effectively making it a VCA and explains why the SS30 is partialy polyphonic. Pretty neat huh?


This is just the first of two diodes in series, with a grounding capacitor after each. Even if the capacitors had become short circuit nothing should be getting through the first diode until a key is pressed and the DC bias voltage rises to 0V.

When I probed the output side of a the first diode on a circuit on G4 I saw plenty of noise, but it seemed to be unrelated to the signal on the input. If anything it seemed to be getting in through the capacitors or bakc through the diode. From where though is unclear. On the diode I probed, the signal on the input side seemed to be safely under 0V, but maybe others are not so well behaved. 
If we assume that somesignal is getting through the diode it's either becaise the diode is not functioning properly or the -7V is not low enopugh and some positiev current is getting through. On the other hand, if the diode is really blocking the noise it must be coming though from the capcitors via ground. But how does it get there? On the input side of the diode is the keyting cictcuit. The offset signal is heading straihgt back to the K boards. Is it entering ground there? The K boards down't even have a 0V connection until the key is pressed down and connected to earth. The attack and sustain controls do feed in some voltage but what that is I'm not sure. They could easily be disconnected to check.

Capacitors are supposed to pass some signals to earth but the issue seems to be that whereas the earth is supposed to absorb this unwanted signal, it is instead passing it on. Maybe the earth being used is not grounded properly.

Next time I hope to have some answers. 

Dave Formula's Yamaha SS-30 - Sold on ebay

This was a month ago, but I just spotted that Dave Formula sold his SS-30. He seems to have been selling off all his old keyboards (and a motorbike) over the past few years.



This the eBay link (for as long as it lasts)  http://www.ebay.co.uk/itm/-/291879123496?roken=cUgayN

Auction blurb

And this is what he has to say in the auction.

Yamaha SS30 String Synthesiser - with a unique & collectable provenance

The SS30 was renowned as one of the most sophisticated of the 1970's string machines and the selling price at the time reflected this. It wasn't produced in large numbers and therefore has something of a cult following among collectors and connoisseurs of this period.

It is a remarkably versatile string machine, as the recorded examples below will demonstrate, with the ability to transform a track, when used as a full section, or as individual instruments: the cellos are particularly effective, solo, when used in staccato fashion. All in all, it records beautifully!

I've had this keyboard since late 1978, and as the photos show, it's in very good shape. I've just got it back from a thorough inspection and service and it's working as well as the day I bought it.

Although I toured extensively, the keyboard was expertly cared for by my keyboard tech throughout that period.

This would be perfect for both the player who wants to utilise the sounds of this unique instrument and/or a collector who 

would appreciate the increasing value of this vintage, investment classic.

I used the SS30 on a range of hit albums and singles by Visage and Magazine, including;

Visage

Visage         

https://www.youtube.com/watch?v=NKeGSxUrKPk - SS30 used throughout

The Anvil   

https://www.youtube.com/watch?v=x0YbPX2BWgg  -  SS30 used throughout

Magazine 

Secondhand Daylight  

https://www.youtube.com/watch?v=4KHEZirNWYQ - listen to the SS30 cello patch on

The Thin Air(around 23mins) as well as the rest of the string sounds on the other tracks.

https://www.youtube.com/watch?v=4KHEZirNWYQ

The Correct Use Of Soap

https://www.youtube.com/watch?v=pzkZy7J9-mo     - SS30 used throughout

Magic Murder and The Weather

https://www.youtube.com/watch?v=6AdYroYSW6k - SS30 used throughout

I also used the this Yamaha SS30 on Magazine's reunion tour in 2009, both on the live dates as well as TV shows, Later with Jools Holland and the BBC Electric Proms.

It sold for a handsome £934, which seems fair considering it's "unique & collectable provenance" and great condition. 

Blurb thoughts

Let's look at some of the sales pitch.

"renowned as one of the most sophisticated of the 1970's string machines"

  Well, of course. It's clear from my researches that, along with the Solina, the SS-30 was the professionals choice of stringer.

"It wasn't produced in large numbers and therefore has something of a cult following among collectors and connoisseurs of this period."

It doesn't come up very often, that's true. And there are no doubt Ultravox, Visage and Magazine fans who would love to their hands on one. But the serial number 2137 has been for sale for quite some time on eBay. There's a general point here about the claimed rarity of items that's not worth labouring.

"It is a remarkably versatile string machine".."with the ability to transform a track, when used as a full section, or as individual instruments: the cellos are particularly effective, solo, when used in staccato fashion."

It's very interesting to see that comment about the staccato action. Normally stringers are considered more for their pads. With a MIDI interface the possibilities for staccato playing or sequencing would be more obvious still.
 He's not wrong about the transformative power of the sound either...

 "I've had this keyboard since late 1978"

 Late 1978 and serial number 3020. Therefore by late 1978 Yamaha had made over 3000 SS30s. I've been keeping track of the SS30s that come up for auction and this is the highest serial number so far. Alas my SS-30's serial number was missing in action when I got it. Hmm!

"I've just got it back from a thorough inspection and service and it's working as well as the day I bought it."

Good to hear! Mine has been through a lot in it's nearly 40 years, not least living in an out-house for many years. The fact that it works (almost) perfectly despite being nothing but a pile of boards and wires is a testament to it's quality.

"I also used the this Yamaha SS30 on Magazine's reunion tour in 2009, both on the live dates as well as TV shows, Later with Jools Holland and the BBC Electric Proms."

This is a minor point but I did wonder about this. In this post from the Ultravox forum in 2009 Dave's Keyboard tech is looking for an SS-30 to take on tour. I guess they wanted to have a spare. 

Altogether now: Chorus! - A stereo chorus?

Josh, call the Bucket Brigade

I was re-reading the superlative Synth Secrets articles by Gordon Reid again to see what he had to say about Bucket Brigade Delay Devices. I was interested because I have a Roland System 500  (lucky me) and it's 572 module is based on the System 100m's 172 module, which has a delay section. 

"the audio delay has independent knob control of delay time and resonance (or feedback) for short chorus-like modulation delays." 

Josh from Malekko Heavy Industries has posted on the Muff Wiggler forum that:

"The original 172 used an MN3004 and really was a chorus. For the 572 i used a (much more available yet still 15v) MN3007 and overclocked the crap out of it. From 0 to around 5 will be that chorus detune range but after that it goes into "delay" territory (yes i know choruses are delays too...you know what i mean)". 

I do know what he means. That statement "The original 172 used an MN3004 and really was a chorus." caught my eye because the SS30 uses three MN3004 BBDs for it's chorus. Although it's called Orchestra in the SS30.  I wanted to look at how a single BBD becomes a chorus and how I might use the 572 in that way. I then started thinking about the SS30 chorus and that's where this post is going to end up. But first, the single delay 572...

What's your secret Gordon? 

The synth secrets part 62 covers choruses and explains how a single delay line can be used to create a basic chorus effect.

"We can patch Figure 3 very easily using just four modules from a modular synth: a multiple to split the incoming signal into two paths; an LFO to modulate the rate of the Echo unit (which combines the delay line and clock generator in a single module); and a Mixer to recombine the two audio signals."

Figure 3: Adding the unaffected signal to the pitch-modulated signal.

The important point is that the delay time has to be in a the relatively short "10ms to 50ms" region and be modulated by the LFO by only tiny amount and at a slow rate - from around a busy, vibrato style 600ms to a much slower, imperceptibly shifting 7s.

Gordon goes on to say that with this arrangement:

"you'll obtain a serviceable chorus, reminiscent of the cheapest '70s string synths."

So, not great! He then explains that whilst your modular synth modules might not be cheap - and the Roland 572 is not - the effect of a single delay is not "deep" or "lush" because our ears can easily perceive the effect of just two frequencies together.  To get a more convincing chorus you need multiple delays just as a true chorus consists of multiple singers or instruments.

 You've got to have a system, Roland

The Roland 172 module uses a single MN3004, as you can see in this schematic. The misleading thing, perhaps, about this module is that whilst it's called an audio delay the range it delays by is quite limited, 0 to 7ms. This is not what you normally expect from an audio delay if you think in terms of obtaining an echo effect. Remember what Josh said though: "The original 172 used an MN3004 and really was a chorus." A short delay which you modulate slowly and mix with the original signal is a basic chorus; this is what you get with the 172 and is made even more explicit in the design of the 572.
The 172 LFO has a range of 0.4Hz to 10Hz and it's even hard wired to the delay time modulation. As long as there's nothing plugged into the external CV input that is.

In the 572 Josh has made the, so called, chorus design of the delay module more explicit still and added control for the mix of dry and wet signals.

Let's get some feedback on that.

Of course, the short delay available in BBDs can be used for flanging or just, err, short delays, but with just one delay the chorus is limited, right? Yes, but don't forget that resonance control! This control adds "resonance (or feedback)". The signal out of the delay is fed back into the delay so you can overlay the signal again and again. To what effect though? Are we obtaining a more chorus-like effect by doing that? Yes and no. The modulation is slow, so whilst the signal being fed back is from a earlier part of the LFO's cycle it's not that much earlier. It's more like a phase shift of the original signal. Unless the feedback was also delayed and that delay modulated it's not adding true complexity. Our hearing can discern such simple relationships and whilst it's an interesting effect it's not adding the right kind of depth.

Because you're Gordon, I mean gorgeous.

In the largely monophonic world of modular synths the aims are different to those of a polyphonic instrument where chorus is a vital ingredient. The economics of making each note sound as rich as possible is a challenge in a polysynth. Complexity that sounds convincingly natural and pleasing means more cost, so polysynths and string ensembles need some help. And if you're using the simple divide-down architecture of most string synths that help is even more vital.

As Gordon goes on to explain, more delays are only part of the solution. Having more LFO's and some phase shifting of the LFO waves are how the more advanced designs implemented the lush chorus of 70's legend. Naturally, this adds cost again so to get a reasonable compromise phase shifters and mixer circuits are employed to create the illusion of more LFOs. As Gordon explains: 

 "This involves the use of just two LFOs (which cuts costs) and four phase-shifters (which are cheap), and generates three instances of a single complex delay modulation. As before, these are out of phase with one another, typically by 120 degrees, and the result, while not quite as lush as you can obtain using six independent LFOs, is nonetheless gorgeous."

 

Figure 13: The classic three-phase chorus unit.

Chorus or Orchestra?

We now come to the SS30 which is almost exactly as described above. The only difference is in the relationship between the LFO and phase shifters and how they are mixed.

Below is a diagram of the SS30 Orchestra section.

SS30 Orchestra section

The LFO signal to the first delay line (A) is a simple mix of the two, phase shifted LFOs.
The LFO signal to the next delay line (B) is also as simple mix of the two LFOs but with a different phase relationship.
The LFO signal to the last remaining delay line (C) is a mix of all the phase shifted LFO signals.

It's not obvious from the schematic what phase each LFO is shifted by (an exercise for the reader perhaps) but they are all different.

One of the LFOs is fixed at 0.6Hz and the other is variable from 0.6Hz to 6.2 Hz. The SS30 control simplifies this and provides a slow and fast option.

And there you have it. The SS30 Orchestra chorus is a tweaked version of a classic design. Presumably this particular variant was arrived at after trying a few different combinations.Yamaha doubtless had similar designs in their analogue electronic organs of the same era.

LFO customisation

At this rate...

The switched LFO speed provides the first opportunity to customise the SS30 Orchestra section and add extra control in the SS30-M. The variable LFO is based around the Yamaha custom iG00150 VCO chip.


As you can see from this design from the Yamaha CS-15, reproduced in the Yamaha IC Guide Book, the VCO is capable of a range of frequencies with sine and sawtooth outputs.  The CS15's LFO has been designed to provide 0.2Hz to 100Hz. 

 In the SS30 the circuit is similar but with switched control of rate instead of the variable control and a lower and narrower frequency range. 

I'll have to do the maths, but basically putting a linear pot in place of the the speed switch (SW1) should make the range variable. For fun, more extreme effects it might be worth having VR3 as a switchable control too, to get more coarse control of the VCO. It's not clear to me yet what the full range of that VCO actually is. I need to play around with those trimmers and see what I can do with it.

Wave hello (to saw-tooth?)

With a bit of hacking it will also be possible to make the waveform switchable from sine to saw output. I'm not sure how useful that would be though. The Moog ClusterFlux chorus has such a wave option and there's a modification for the CH-1 Boss Super-chorus pedal that goes from the normal triangle to saw-tooth at the higher rates. No idea what it sounds like though!

Modulating the modulator via modular modulations?

Another possibility is a CV input to modulate the LFO externally or straight LFO inputs to replace the built in ones. If I was removing the chorus section altogether to just use that it might make sense but things can get out of hand and the front panel will start to grow and grow.

Stereo - Typical chorus

The section of part 62 of Synth Secrets  - More Creative Synthesis with Delays - that got me thinking most was that on stereo chorus. 
It's quite simple to take a 3-phase mono chorus and create stereo chorus and you can see how below.

Figure 14: The output stage of a triple-path stereo chorus unit.

As Gordon says:

"In this scheme, you can leave out the 'straight-through' signal, because the dual inputs to each mixer will be chorusing differently, and — far from contributing to the result — the original might actually damage the impression of width and depth."

I think this will thought this would be easy to do on the SS30. The 'mixer' in this case is simply a 12K resistor on the output of each delay and a 22K on the straight-through signal. The mixed signal then goes through the Depth control. If I tap-off the delay outputs and mix them with similarly rated resistors I can create a stereo chorus output. In practical terms it might be best to remove the 12K resisters, take wires out to a new board and have a switch circuit to handle a stereo or mono signal and control the mix of signals via the Depth control. To keep things simpler the current mono output would stay the same and a new stereo output would be added. This would need a new pair of output amplifiers though.

BBD devices 

TC350 - 185 stage - This ITT device was used in the Solinas, Crumars, Elkas, Logans, Vermonas, Hohners and the modern recreation Elkorus

MN3009 - 256-stage - low noise
MN3002 - 512-stage - standard type 
MN3004 - 512-stage - low noise

MN3005 - 4096-stage - low noise
MN3007 - 1024-stage - low noise
MN3008 - 2048-stage - low noise

Mono Choruses

SS30 - 2 x LFO - 4x phase shifters - 3 mixers - 3x clock generators (50KHz) - 3x delays (3x MN3004)

Boss CE-2 - 1x LFO - 2x phase shift (1xMN3101) - 1 x delay (1xMN3007)
http://www.electrosmash.com/boss-ce-2-analysis 

CS70M - (3x MN3009)

Solina - 2 x LFO - 2x phase shifter + 1x inverter (0 degrees, 120 degrees and 240 degrees.) - 3 x clock generators - 3x delays (3xTCA350)

Korg Polysix - 3x LFOs - 3x delays (3x MN3004 with loop-back)
http://www.synthfind.com/korg/korg-polysix-klm-368-chorus-ensemble-board/

 Stereo Choruses

Boss CE1 - 1x LFO - 2x clock generators - 1x phase shifter  - 2x delay (1xMN3002)

Roland VP-330 -  2xLFO with 2x mixers 4x delay (4xSAD512D)
"the Roland VP330 (1978) has a thinner string ensemble sound generated by just two delay lines with dual LFOs." Synth Secrets 62

Elkorus - 2xLFO - 3xPhase shifters - 3x delays (TCA350)

Engraving objects - Front-panel Designer, Inkscape and HPGL files

An apology

This post was drafted some months ago. Since it was all too difficult and now that Schaeffer are offering printing to their front-panels now I didn't really come to any conclusion. Engraving tools, HPGL files and using Inkscape is a frustrating process and I just gave up. I'm posting this anyway, in case someone is looking for advice on using Inkscape with HPGL files and FPD, and hoping it might be of some use.

Oh no! Logo no-go?

I was thinking about having something more interesting than just the labels on the front panel. But what? How about a logo? If I use Schaeffer panels in theory anything can be engraved and in-filled with colour.

I found someone had created a Yamaha logo for CNC

They posted it as a dxf file so using Uniconvertor I made an HPGL file.

Front-Panel Designer can import HPGL engraving objects, so that works easily.

The cost is

HPGL engraving	54.44	62.98	Yamaha 2 HPLG.plt	1.83 €
Other	-	-	Engraving infill	5.69 €
Other	-	-	1 tool changes	1.03 €

Engaging in engraving

But this kind of engraving is easy because it's a single-line drawing. What if I wanted something more elaborate?

This is the Yamaha logo and logotype

If I convert this file to just show the outline though..

In this file the outline has not been cleanly closed off and it only makes sense when you view the 'fill' without the 'stroke'. To engrave this design you need to start with a clean outline and then have lines inside which are cut out.

You can do this in Inkscape but FPD recommends using Corel Draw as it has better support. But Inksape is free so... Let's try with Inkscape.

Inkscape

First there are some top tips in this document MANUAL FOR HPGL FILES WITH COREL DRAW
There is an explanation on engraving and cutting tools.


Something I missed earlier when doing this is that you need to set the stroke style's 'Join' to rounded. Doing this gives you a realistic picture of what the engraving will look like.

Also there is a fundamental problem with cutting a line in-side another line. Although the outside of an angle will have a rounded edge the inside will be straight and sharp. When you draw a line inside the corner that comes to meet that inside join it will be rounded and you get a gap. There's no way to prevent the gap so you have to fill in the triangle with another but of engraving or cutting to remove it. Alternatively go for designs with no acute angles inside.  

The S shape used below is a simply hard to do and there are much easier designs with fewer problems.


Starting with a single character 'S' in the Faktor font favoured in some of the Yamaha advertising of the same period as the SS30 let's try and engrave and cut a custom font.
This is how the text appears in Inkscape to start with.

Then I set the fill to nothing, the stroke colour to black and the stroke width to 0.2mm (which is the smallest engraving size on FPD).

That's not quite correct though. The thickness of the stroke works from the middle outwards.

A thicker stroke would start to unnoticeably distort the shape. This is at 3mm.

To be accurate the stroke needs to be centred 0.1mm inwards. Here's a close up showing the outline in black (at 0.05 px) and the proposed engraving in read (at 0.2 mm).

You can see clearly that the engraving is going outside the bounds. So let's fix that.
The Path->Inset command moves the selected path inwards by a set amount. You have to set the amount in the Preferences (Edit->Preferences->Behaviour->Steps->'Inset/Outset by'. For 0.2mm I've set it to 0.01.

Now the engraving line fit's inside the outline but it's lost the sharp edge. If we engrave this the tool will follow that curve instead of sharply turning the corner.

If you try and do this inset another way you get a different problem though. In blue I created another outline in blue, set it to 0.2 mm wide and then reduced the size of the outline by 0.1mm.

This still have sharp edges but as you can see it's not the same as doing an inset and it's not inside the lines. Reducing the size is nothing like doing an inset command. 
To get exactly the right size it is better to start with a shape which is 0.1 mm too small and then the engraving will keep it's sharp edges. But there are other problems with the inset function in Inkscape as we will see later. 

What happens when we actually engrave an inset outline? I took a copy of the drawing (deleting the blue engraving first) and offset the guide outline to the right from the engraving. Then I saved it as HPGL format.
This is what it looks like in Inkscape. The inset outline on the left the original shape on the right.

And then imported to FPD

You can see the problem. The line created with inset has lost it's smooth shape.

Now we know that the outline is better of being scaled correctly with allowance for the engraving tool width and not using inset we can turn our attention to the in-fill.

The cutter tool options are 0.6, 0.8, 1.0 , 1.5, 2.0, 2.4 and 3.0 mm.  Bigger tool means less cutting which is cheaper, but the bigger the tool the bigger the contour radii.

Here's is a close-up look at the outline cut with the 0.6mm tool.

The shape isn't too bad but the outer edges are rounded off. How will that fit inside the outline cut by the engraver?

Firstly let's create an inset path which is the first cut inside the engraving. The cut will be 0.6 mm and therefore 0.4 mm inside the outline (0.1 mm + 0.3 mm).

 

Not too bad, but you can see a gap. When this is cut and engraved that gap will be left over metal. This is because the inset has miscalculated and not followed the outline properly.

Let's convert to HPGL and see what happens.

I'm HPGL

We now hit problem with Inkscape and HPGL files. Each line in the HPGL file can be assigned a pen (from the idea that this was a file format for plotters). In FPD each so-called pen is assigned a tool (engraver or cutter). Unfortunately when you save the file in Inkscape you can only choose one pen and not multiple pens with a different one for each line.
Instead to create the engrave and the cut line you have to create two files. One file is just the engrave line and one is just the cut line. When you save the file you assign each one a different pen so that when you import then to FPD you can get two files loaded and placed at the same time with different pens. Actually not at the same time exactly. You have to place each one in the same position individually, which may be an issue if the reference points don't match up.  However, you can manually combine the two files. Open each file in a text editor, like Notepad++, and then append one after the other in a new file. Then you can open both as one file and you will have two pens, one for each. The pens are identified as SPx where x is the number of the pen, so SP1, SP2, etc.

In any case, bringing the file into FPD the result is poor.

I have done more experiments where the shape is much larger and the result is a bit better. But that is no use. You cannot reduce the size of the outline and it's still not perfect.

To be clear using the automatic inset function is not possible. But there is another way.

There is a path function called Linked Offset. Using this function you can manually drag the shape an inset. This has a much neater result and fewer errors. The only thing is that you have cannot measure the inset with 100% accuracy. You can move the amount of offset but also the position from which the offset is being dragged. This give a fair amount of variation and can create problems but is still better than the automatic option. Selecting this tool also creates a copy of the selected path so it's quite quick to use. There are some strange effects though. There seems to be a bug which causes all kind of artefacts to appear and make a mess of things. They can't be removed and appear scattered around the original shape for no apparent reason.


  And here I ran of interest in doing this....

Yamaha's Quality Design and Manufacture

I was browsing around the web looking at Yamaha organs (uh-oh!) today and came across this image which tickled me.

It's from an fact-sheet/flyer of their new organ around 1978, so contemporaneous with the SS-30.

Design

Manufacture

A switch to variable controls?

A bit on the slide?

As I've been too busy to get into the garage again these past few days (and will be until next week) I was instead meditating on the idea I had during the front-panel rewiring exercise.  The idea was that for some of the switches I could instead have a variable control. I liked the idea of using sliders for some of these as you can more easily switch them on/off.
Let's see how that might work.

Orchestral Manoeuvrers 

Here is the Orchestra section.

Orchestra Section - Controls

 The Depth control set's the mix of signal that's going to the output unaltered (dry) and the amount that is going through the of the effect (wet).  The only difference between the minimum setting and having the instruments bypassed is that when with the minimum setting a small anout of signal will still get through from the Orchestra block.

The other controls are all switches, but what would happen if they were variable controls too?

Speed Control

SPEED is a switch which alters the voltage to a Yamaha IG00150 LFO chip. It's speeds '1' and '2' could be any value between those two voltages but with the switch they are fixed. It could even be a voltage beyond those limits.  Replacing the switch with a suitable potentiometer would be a useful addition.

Switch controls

The instrument switches are more problematic. Each tablet switch is dual-ganged and has two 2-way switches.
For each instrument (CELLO or VIOLIN) when the switch is 'off' the Orchestra effect is bypassed. In bypass mode the signal goes through T1 switch, through a resistor and then on to the T2 switch and onwards to the output.
When the switch is 'on' the signal goes from T1 switch to the Mixing Amp, then it's split to go to both the Orchestra section and the depth control. The depth then balances between the original dry signal and the wet. Then it's back to T2 and onwards to the output.
In other words, when it's bypassed the Orchestra inputs and output are disconnected. Otherwise you would get both Cello and Violin input when you only select one and the noise from the choruses would leak through even if nothing was being input.
Therefore the level of each instrument into the orchestra could be controlled but there will need to be some care taken to ensure that 'off' still means off. The easiest way would be to retain a switch and add a variable control. A rotary pot with a switch would allow both to co-exist but I want a slider then a button would be needed. 

In the mix

Here is the mixing section:

The best form of attack is a switch?

The ATTACK switches set the period of the 'attack' phase of the envelopes - i.e. how long it takes each key to rise up to full volume. When not engaged the AT signal to the key switch circuits is switched to ground. In the keying circuit that means that when the keys switch to ground the transistor controlling the signal to the VCA switches immediately to ground too and the attack is period is fast. However when the switch is engaged and switched to SLOW the AT signal is disconnected from anything. Now a capacitor in the key switch circuit comes into play. This capacitor sits between -7V and the emitter of the transistor.  The ground from the key switch will now cause the emitter to rise to 0V but because the collector is no longer at 0V there is no current through the transistor. In order to bring the signal to the VCA up to 0V now the capacitor has to be charged up which slows the increase in voltage from -7 to ground. Phew!
In short the switch governs if the charging is significant to the switching, but the time period of the charging is set by the RC constant of the capacitor. As the AT line is not connected to anything in the SLOW mode there's no point trying to replace the switch with a potentiometer.
I'm not sure my circuit analysis is totally on the money there but it might work with suitably large pot' if the so some experiments are in order. I might have to see if there's a non-trivial but still feasible way to do this with some sort of circuit to control the current. Maybe... 

Mix selector

The selector switches turn the signals from the 'Mixing Amp & Filter' circuits on or off. 'Off' means the signal is switched to ground. 'On' means it is routed on to the next part of the chain.
Therefore it's a simple matter to replace each switch with a suitably rated potentiometer which would allow each instrument to be balanced rather than just switched.    

Summary

Most, if not all, of the switches could be replaced by controls, but should I? Putting aside the fact that I've swung back to not reusing the original controls again, are these useful changes? I think changing the Orchestra speed is useful. No respectable chorus effect would really want to limit the user to two speeds so this is definite improvement. Similarly all effects allow you balance how much signal is fed in so the orchestra switches would be better off as variable controls (or both). For the attack control is would be nicer to be able to control the time. again no self respecting synth has only a switch and the sustain time is variable after all, so that is preferable too. The mix of instruments is easy to do but is it worthwhile? I do think it is. It seems that no stringers offer this feature though. They were seen as organs and followed the switch per instrument approach (where they offered anything at all). But why not? Synths almost always allow you to mix the tones, even when they come from the same oscillator. In summary then, I should seriously consider this option.

Rewiring update 2

Cello, again!

Back one day later with more updates!

I fixed the issue with the Cello. Well, almost. The issue with broken wire to the volume control was easy to fix. However the signal had another break further down the chain. At the Orchestra section the signal is either passed through the chorus it bypasses the effect. In other words the it's a 2-way switch*. The problem was that the open backed switch for the Cello has had it's contact bent at some point and it was only operating as a one-way switch. The Cello Orchestra (wet) worked but not Cello without Orchestra (dry).  I bent the contact back and it's work okay now. No great, but okay.

Vibrato delay fix

Next, I wasn't sure that the Vibrato was working right so I did some measurements around the LF board and found a couple of broken wires. An earth wire and a control line for the Vibrato Delay. After that fix the Vibrato seems okay too.

The 100Hz Mystery

Yesterday I said it sounded like there was a noise floor of all the voices bleeding though. Last night after fixing the Cello and the Vibrato wires I again turned up the volume to listen to that noise. I can't be sure, but I think it has changed. Now I hear a single tone coming through. Like the stuck note I have on A# on K4 board, but this time much quieter. To be clear maybe my ears were deceiving me previously but now, for sure there is something there at one tone.

100Hz noise

There could be a bleed of everything else under that, but as you can see there's something quite loud. It's around 25mV at 100Hz and crucially it varies frequency with the pitch and detune controls. That means that it must be coming from the voices and not the power supply or some other circuit. I traced it back and it is present all the way back to G1 board.  I started there as it seemed likely that this would be coming from a whatever note is around 100Hz, and that is the lowest frequency board. However, I could also see it on the G4 board outputs. This isn't surprising actually. The outputs are mixed together (crudely without buffers), but I would have expected to see something stronger at the source, probably on G1. I didn't, so I will need to go deeper.

Make like a banana...

Something else I noted is that as I switch the split-gate upwards it lessens. So at split 0 it's loudest and at split 3 it's quietest. It's also present from all sources - Cellos and Violins.The split gate is quite fiddly to describe without a few pictures and tables but some voices are always open to be switched. The lower part the G1 board 16U and 16L are always present. In split 0 G1 16U/L are present as part of the Viola, mixed with the G1 8U/L. At split 1 (and above) G1 16U/L is present (alone) as part of the Cello sound. 

On the down low

If we assume some note in the lower part of the G1 board is the source of the noise then it should go away if only the Violins are switched in and the split is 0. This sort of matches the effect I got when playing around. It didn't go away completely though. I could get a trace of it in G4 so that does make sense. It could still be bleeding back even though the Viola and Violins are mixed separately.

How low can you go? 

Therefore the theory that it's coming from G1 makes sense, except for one thing. If I played through the notes on G1 I should have found the same note somewhere on there. I think I did, actually, but it wasn't quite right. I played note G and got something much louder but sounding the same. But I couldn't get a trace that matched. I think that the lowest G is G3 not G2 and my ears tell me it's more like 200Hz than 100Hz. I must redo this but I think it might just be because that G is a fundamental of 100Hz I'm hearing. G2 is 98Hz and G3 is 196 Hz. I think I'm hearing the first fundamental at ~200Hz because of the natural high-pass filtering in my listening environment from strip lights etc. (and yes, I tried turning them off to make sure it wasn't EMC). What I measure is 100Hz, but what I hear is a mix of harmonics from that pulse trace - I think! 


All very interesting (in a way) but if the fundamental is 100Hz then where is it coming from? The lowest note is C1 at 130.81. If it was the low G and I was wrong about the lowest note frequency it would probably be because the key circuit was broken. But I also shorted that key to -7 and it made no difference.

Pulse, a ting?

Finally the shape of this tone is not 'right'. Firstly it's a pulse, not a square -which is the source of the tones - or the more saw-tooth like shape - from the wave shaping (the subject of another post). If some wave shaper on the G board is faulty and bleeding through the raw oscillator then maybe a pulse is the result. The most obvious fault would be a shorting capacitor, but how that would get to a pulse I don't know.

Enough. Let's see what I find, next time...



* Idea - Replace the switches with a mix control? In fact all the switches could be replaced with such controls. Then instead of switching things in or out they could be balanced instead. Sliders would be faster for a quick switch in/out.

Front-panel rewiring update

Progress 

Temporary front-panel rewiring

Progress has many enemies: perfection; comfort; greed; fear; indecision; indifference and (it says here) free-time(?). 

I can't point to any of those and blame them for the lack of progress on the SS-30M. Free-time is the thing I lack the most.  Okay, maybe comfort. I don't like working in my garage when it's below 10 degrees. No, the real enemy of progress has been Eurorack.

Eurorack

 I made a rack case

Eurorack case sanding

Eurorack case rear

Eurorack case ready for loading

And I made some DIY modules

DIY Mutable Instruments Braids

And I got a Roland System 1M and then a System 500.

My Eurorack set-up taking shape

I even started another blog about the System 500 - http://system500.blogspot.co.uk/

Now all that is (mostly) done (for now) I am keen to get the SS-30M show back on the road!

Front-panel rewiring

You probably won't remember that last year I was wiring up the temporary front-panel when I stopped. Most of the controls had been removed some time back and the challenge was to reattach them all. In the bigger picture this would allow me to test the SS-30 and see if there were any major problems before starting on the MIDI interface.

Rewiring was quite slow because although each wire is colour-coded I still wanted to check that it came from where I was expecting. With the black ground wires this was quite hard as most are still bunched together and it's hard to tell where they go to. Note that there isn't just one ground for all the controls. They are all grounded to different points on the F, LF and OR boards so there are several black wires. I think I got them all in the right place in the end.

After a couple of evenings work everything seemed to be back in place and I powered up to see what was happening.

A sharp pain

The first thing wrong was that there was a stuck note. A# on the top K4/G4 boards was on and staying on. I traced it back and couldn't see any wiring issues with it. I fixed another couple of broken connections on K4 and then disconnected A# from the G board. The output should have been -7V but it was 0V. I poked around on the circuit for a while and after a bit it sprang into life. Hmm! Well, it is a very dirty board. K4 sits at the top of my stack of K boards and has collected a lot of dust and fluff over the years. I will just have to wait and see if it fails again, but for now all is well.

Yo! Yo! where's my Cello, Ma? 

The next obvious problem was that although I could switch in the Violin section (Viola, Violin1 and Violin 2 were present and seemingly correct) the Cello switches weren't making any difference to the sound. I switched through the KYB.SPLIT options to no avail and it was clear that the Cello audio was missing in action.
I traced it back to the Cello Brilliance control. At that point the Cello signal comes from the Cello Amp circuit and splits. The signal goes into the Brilliance control as part of the low-pass filter circuit. The control sets the resistance to the signal going in to the capacitor to ground. If there's nothing going in, no filtering - brilliant! Another wire goes direct to the Cello volume control and from there onwards to the Orchestra section. When I looked, the two wires had split apart and only one was connected. The signal in to the volume control was waving in the breeze.
It was too late to fix that so it's next on my to-do list.

Bleeding hell?

All the Violin section was working though and I checked the Pitch, Vibrato, Brilliance and Orchestra which all seemed well. I was particular happy to hear the familiar woozy chorus taking over as I increased the Orchestra depth. I was tempted to immediately try and patch in another synth and have a play with it as a stand-alone effect. I resisted though and pressed on with my checks.

The only other problem I found was that the noise floor was seemingly quite high. I say noise but in fact it's the sound of all 49 keys bleeding through at once. The only thoughts I've had so far are that either this is normal or that there's something wrong with the -7V or 0V around the key switches. If -7 isn't low enough you will get a bleed through. I need to measure and think about that. Maybe I have 49 capacitors that need replacing :-(