Softpot Ribbon Controller

[Scott Stites]

I'm pretty well finished with the core circuit. Here's a schematical block diagram of it - I haven't put in the values yet - I won't release that until I work out the range stuff, which for now is a simple buffered voltage off of the rail. It will ultimately be a regulated reference voltage.

The SoftPot itslef proved to be a challenge to fully understand and tame for this application. The generalities were easy enough, but the nitty-gritty stuff took a bit of thinking. It was easy enough to apply voltage to it and make it go up and down, it wasn't quite so easy (at first) to acquire the voltage, hold it, and produce gate/trigger signals. But the really challenging part was dealing with what the SoftPot does in that nether region between gently pressing and releasing. It wasn't thinking as linearly as I wanted it to. Pressing it, then releasing softly, then pressing again would cause the analog shift register to grab bits and pieces of the convoluted fluctuations the SoftPot went through. Think of a finger controlled S&H circuit processing a noise generator, and that's pretty much the effect.

After some thought, and no small amount of flat out swearing, the solution finally occurred to me - it was staring me right in the face. The analog shift register was designed to grab events that happened in the very near past and turn them into a present occurence. It was relying, in large part, on chance. So, why not focus it to determine if the Softpot was having a momentary hissy and just ignore that hissy? I did that by comparing the output of the first S&H and the second S&H and ensuring that the two levels were within a certain window of voltage over a certain amount of time. If the difference falls out of that window, S&H 3 ignores the results, and the first two S&H's try again. Only until they agree is the voltage considered a valid voltage, and S&H3 is allowed to sample it. The window itself must be wide enough to ensure a continuous slide over time (estimate how fast a person could slide anything across the ribbon, then use that as the criteria for the threshold). At first, I miscalculated how long the SoftPot might hang onto a certain erroneous level - I severely underestimated the time. I finally figured the right clock frequency, and the SoftPot is as solid as a rock and as smooth as a baby's behind. I've been using the circuit for two nights in a row, and I'm very, very pleased with the results.

This is the core of the Appendage - it takes a pressure event and translates it into three unique control voltages (continuous, intial value and differential) and produces the timing signal for gate and trigger. Everything past that is downhill. These signals can be manipulated to do an astounding number of things, simultaneously.

The way I see it, if you want a standard ribbon controller, go buy the Doepfer. From the videos I've seen, it's pretty cool. The Appendage, however, is going to be a different beast. I envision it as a cross between a ribbon controller and a Buchlian gestural controller.

Position, relative position, rate of movement, direction of movement (split even into different directions) will be able to control any parameter your synthesizer has, simultaneously. Mixing these control voltages is fascinating. For example, the amount of control mixing differential with initial provides is surprising. Mixing slide with initial creates a sort of hyper-range. Slide with slide bends the curve radically. I've even discovered a neat trill function (quite by accident).

I'm already thinking of the DAC circuit, which is down the road. It need only grab the S&H3 and S&H4 voltages for infinite hold (everything is really derived from those two voltages). I figure a half note step quantizer will fit in as well quite nicely.

Cheerios,
Scott

---

appendage_prelim_1.PNG
Description:	Appendage Core Circuit
Filesize:	55.15 KB
Viewed:	929 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

[Uncle Krunkus]

You're doing some great work here Scott!!

[Scott Stites]

It gets better, Uncle K.

Today I breadboarded a second circuit from my schematic of the first. I did change the positive logic ICs to M2L (CD40106) so there wouldn't be any underused ICs taking up space. Other than that, the circuit was identical, and, much to my relief, it worked right off the bat.

This gave me an "unfuzzy" breadboard to work with. When I put it together, I left some extra room around the analog shift register section. One of my goals was to improve the performance of S&H 3 and S&H 4 (the two S&Hs that provide the voltage outputs). I was getting an abysmal 90 mV/minute droop, which I couldn't do much about until I had some room to get in there and work. I put in a CD4047 and an extra LF398 on S&H3, calculated values for pulse width, routed the path accordingly, and turned on the juice. I then opened up the VCA and played a note, waiting to hear the droop. I waited. And I waited. And I waited some more. I began to doubt that the VCO was under control of the thing - I hit another note. Nope - it was controlling the VCO. Dumbfounded I hooked the DMM to the S&H3 output, hit a note and watched my watch and the DMM.

It was a full three minutes before the voltage dropped 1 mV.



Never in my wildest dreams did I think I would ever get a sample and hold to do this on the breadboard. This thing does not need a DAC, unless you're planning on holding a note or a control for an hour or so.

Needless to say, I didn't waste any time rigging up S&H 4 with the same circuitry. The Appendage has landed - it's now performing beyond all my expectations. Even though the SoftPot is still loose on the bench, it feels like a musical instrument. I can't remember the last time I was so elated over a circuit.

Here's a sample I recorded a few minutes ago. Again, the SoftPot isn't mounted to anything, so I can't really freely whack on it like I want to, but if you've heard the previous samples, you should get a feel now for the improvement in the thing.

It's CV2, which is the fixed voltage, controlling two VCOs. The trigger output alone is controlling an EG. The EG is long attack/release and is opening the VCA and modulating the filter. Both VCOs are going through my 2040 clone filter and through the VCA. I've got the differential output controlling the cutoff of the filter as well. When I slide up, the VCO pitch doesn't change, but the filter opens up. When I slide down, same thing - filter closes, VCOs don't change. Differential only changes from zero when there's a slide, or a hammer on while the pressure is still applied -it only changes relative to the point where the note was initially played, which means you have a considerable amount of control over the filter as you play.

CV2 is unique, because it is sampled only once per note - so it's a fair example of the new S&H, though a true sample demonstrating that would be a verrryyy long one-note sample.

This was all recorded in one take, through my Lexicon MX200 straight into the recorder.

Cheerios,
Scott

---

Appendage_Mark1.mp3
Description:	The Appendage has Landed	Download (listen)
Filename:	Appendage_Mark1.mp3
Filesize:	4.01 MB
Downloaded:	1375 Time(s)

---

[Uncle Krunkus]

That's one small wiggle of the finger for man,
One giant slide for mankind!!!

[Wilba]

Great work, Scott!

A while ago, I tried getting the Paia dual note ribbon controller circuit working, connected up to a MIDIbox so it would output MIDI note events and pitch wheel slides. (Yes, effectively a Doepfer R2M clone, but with two notes on one ribbon). While I got the circuit working, and the MIDIbox code done, I was plagued by these spikes on releasing my finger... and don't know enough analog electronics to go further with it.

I am really excited that someone else is doing stuff in this area, and I look forward with much interest to whatever circuits you share, as I'd love to get back into it... I can't wait to finally control my MIDIbox SID synth with 500mm ribbon controllers!

[State Machine]

Good audio example Scott. This really brings me back to the age of true musical experimentation. The droop rate you reported is mighty impressive. I am sure many will be interested in building something like this.

[ringer]

Hi Scott,

Nice sample, I am blown away. I have been playing with membrane surfaces for a while. One of my problems with them is if you position them vertically, is the offset voltage at the bottom. I like to control a VCA with them instead of an envelope. You put your appendage down at the bottom and raise it to the top, for say 10 volts, then you bring the appendage downwards to the bottom to get 0 Volts, there may be a small voltage leak which keeps the VCA turned on.

Have you tried controlling a VCA with your design? I would be interested in the results,many thanks.

Ringer

[Scott Stites]

Wilba - I must chime in with State Machine and say !

I'm pretty sure this would work for interfacing with the MIDI box. It's certainly possible to get a two-note response with the Softpot, but I didn't start off in that direction. I have thought about putting that sort of functionality in, bearing in mind using this glitch-gar method, it would require a bit more circuitry.

Uncle K: Pretty nice tag line!

Bill: I have to say, I would never have been able to get this far without that nice package you sent a couple of weeks ago. It would have been impossible without it.

Ringer: Yep - I have controlled a VCA with it (to control the level of an LFO modulating the VCO - a lovely patch with a ribbon controller). This design has positional "memory", so that that CV1 and CV2 outputs put out a voltage relative to where the pressure is applied, and when pressure is withdrawn, the voltage stays at that level. The "bottom" of either of these two CVs will always have some voltage. The differential output is what you would use to control a VCA. The differential output is relative to where you first apply pressure. Offset doesn't figure into it.

Differential starts at *zero* volts no matter where you touch the ribbon. If you slide up from that point, the voltage increases positive. As you slide down to the point you originally touched the ribbon, the voltage slides to zero again. If you slide down past the point where you touched it, it slides more and more negative. It does have memory, so, say you open the VCA, then remove the pressure, the VCA stays open. Then, the next time you press the ribbon (anywhere) the voltage falls back to zero and the process repeats. It is possible to switch it to a non-memory mode - IE, when you remove the appendage, it immediately falls to zero. If you lag the voltage, it would glide down perhaps a little more satisfactorily rather than just snapping off. Of course, with memory, you just glide your appendage down to close the VCA then remove the appendage.

Implementing the differential permutations I plan to incorporate, you could open the VCA by sliding up past intial contact, and/or sliding down from initial contact. You could even open one VCA by sliding up and a different one by sliding down. That's the power of DIY - you get to do exactly what you want to do.

Cheerios,
Scott

[loss1234]

wow

cant wait to build this when some schematics come out!!

i have been wanting a good ribbon controller forever!

thanks for all your great discoveries

[Wilba]

Thanks for the welcome!

[Scott Stites]

[Scott Stites]

[Scott Stites]

To extend the monologue...

Here's an idea for an Appendage controlled touch sequencer, which would be sort of like the Serge touch sequencer, but rather different as well.

First of all, it's based on an 8X2/16X1 sequencer I worked on shortly before I diverted into the Klee. You could use it as such. The addition of the Appendage would take it to another level.

Using window comparators, the Softpot is divided into eight touch zones, separated by "guard bands". Touching within each touch zone will activate the sequencer step so that either one or both of the pots would produce control voltages. Touching in the "dead zone" between the "touch zones" would not trigger the sequencer step. This ensures that you can accurately trigger whichever step you want to. Wiggling the finger on the touch zones will (if you want) wiggle the voltage for vibrato or filter or whatever. This would be somewhat like the Serge, which has dedicated plates for each stage.

The difference is you would have continuous slide for each "note" or "stage"; the "dead zones" would not be "dead" for slides - only for triggering the initial stage. That means you could do much more than wiggle your finger on a certain stage. You could hit stage 3, for example, and slide it up beyond the range of hearing if you cared to. Or down. Hitting the next stage would start the voltage at that stage's voltage and you could do the same. Obviously the more left you go, the less you can drop the voltage, and the more right you go, the less you can slide up.

It may be possible to get 16 across, but that might make it a bit trickier to set up (narrower touch bands).

Taking it a step further, you could have, say, four rows of pots. On the left or right side of the rows of pots you could have a smaller soft pot, mounted vertically. This soft pot could be used to select the active row (and slide up or down, etc). You could have an 8X4 matrix of 32 pots, each one individually accessible by manipulating the horizontal and vertical soft pots.

This ain't nothing compared to an idea that was PMed to me last night.

Cheerios,
Scott

Hey - guess I'll have to load the pic later. Can't upload now? I'm getting an error....

Edit: Ah, there we go.

---

appendage_touchseq.PNG
Description:	Appendage Touch Sequencer Application
Filesize:	12.28 KB
Viewed:	485 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

[Scott Stites]

Tonight I put in a small sub-circuit to linearize the response of the last couple of inches of the SoftPot. Without this addition, things are pretty well linear up until around the last two or so inches, and then things climb for the last 20% of the range. It appears the response is somewhat exponential. It was, anyway, not anymore!

That was the last link in this version. Now I'm going to try for reduction of parts though none of the blocks can be done without - last night I spent some time seeing if I'd made any changes that would render any of the circuit non-essential. I did this by by-passing the functions, but they all work together in lock-step to make appendage sliding a pleasure and not a glitch-fest.

I think I can reduce the parts count in the ASR portion by going to simpler S&H topology for S&H1 and S&H2 - just as long as they can carry a voltage accurately for a split second or so.

I'm thinking I'll leave the hooks in this board to accomodate the two-note variety via a second board, which would also hold the voltage permutations, retriggering, and slide velocity, etc. This circuit could be held on a single board, but the circuitry for what I have in mind for two-note operation would put it over the edge. This version makes a great ribbon controller, and also I think it's a nice interface for smaller control ribbons to mount on keyboards, etc. And it'll be required for the two note version - the Double Appendage? .

Anyway, one of my goals is to make it with parts that are easy to get - it contains the following ICs:

TL074
CD4013
CD4017
CD4093
CD40106 (74C14 works fine, too).
LF398
ICL7556 (low current version of the 556, which would also work, but I don't recommend it for anything. You'll need more caps if you want to use a 556).

It uses comparively few discrete parts (resistors, caps, diodes, transistors).

It will provide the following output voltages, max 10V each:

Raw Continuous Slide CV
Raw Fixed Note CV
Raw Differential CV (auto zeroing bend)
Ranged Continuous Slide CV
Ranged Continuous Fixed Note CV
Ranged Differential (auto zeroing bend)
Gate
Trigger

With the following controls for the ranged outputs:

Continuous Slide Range (potentiometer - not sure if there's enough room to put in a calibrated range rotary switch - mebbe)
Fixed Note Range (potentiometer - again might be switchable)
Differential Range (potentiometer, again with the switch note).
Fine Tune Pots for all of the above
Offset
Fine Offset
Bend level pot for ranged fixed output (basically mixes differential with fixed note) - an absolute must - much more controllable than the continuous slide in many cases, and it's cool, trust me).

The number of controls and outputs used are up to the builder. This could be a module or a full-blown rack (considering what's coming down the pike). Or it could be a self-contained unit. In the case of a module or a rack, I think a single stereo 1/4" TRS connector would work for interfacing the ribbon to the module/rack (like on a foot controller). Then you could strap it on, wank away, and steal the groupies from the guitarist.....

It's also the basis for a number of specialized controllers that I can't stop thinking about (like the touch sequencer, etc).....I need sleep......

I'm certain it'll work with either +/-12 or +/-15V, though +/-12 might require some different resistor values - remains to be seen.

Cheerios,
Scott

[Coriolis]

ohgodohgodohgod, keep it coming Scott!

You are singlehandedly realizing all of the control features I've ever wanted, but could not put together myself!

I can already see it before me: An 8-step touch-sequencer/programmer whatchamacallit with 4 rows of pots (Octappendage?), supplemented with the original appendage perhaps? Or do I only need the Octappendage?

Droool...


C

[Scott Stites]

One Appendage would do it for the touch controller idea - the sequencer portion would contain the circuitry to sense what the SoftPot was up to and act accordingly. You could do all kinds of groovy things - control the sequencer, jam stages in while it's running, control the clock, bend the sequencer around, or, of course, play the sequencer like a bendy little keyboard.

Each "key" on the strip, could of course put out it's own gate/trigger so you could tap out gnarly rhythms on your drum modules while the sequencer plays. If your drum has a CV input for shell pitch, you could do some pitch bendy accents. Schtuff like that.

The idea is that an appendage board will do enough that it can be applied in all sorts of control situations.

[Scott Stites]

I've re-worked the circuit so that it uses a different voltage acquisition scheme - the Analog Shift Register has been replaced by a single S&H that monitors the softpot voltage and its own output. It's looking for a signal that is stationary for a relative period of time (within a certain window of time and voltage). If it determines that a voltage falls within that window, it passes it on to the slide sample and hold and "hold" sample and hold. If not, it scans again. Everything is happening so fast, there is no apparent latency. It's very responsive and very clean.

I've changed the terms CV1, CV2 and Differenential to "Slide", "Initial" and "Bend".

Slide: You press the ribbon, which initiates a trigger and holds the gate open as long as pressure is applied. The CV level (amount of voltage) is proportional to where on the ribbon the pressure is made. While applying pressure, sliding to the right slides the voltage up and sliding to the left slides the voltage down. Releasing the pressure holds the voltage at the last value that was generated before release.

Initial: You press the ribbon, which initiates a trigger and holds the gate open as long as pressure is applied. The CV level (amount of voltage) is proportional to where on the ribbon the pressure is first made. While applying pressure, sliding to the right or to the left will not changed this voltage up or down. It remains fixed at this initial value when pressure is removed and only changes when a different point is pressed on the ribbon. This is analogous to the operation of a keyboard.

Bend: You press the ribbon, which initiates a trigger and holds the gate open as long as pressure is applied. The CV level (amount of voltage) is set to 0V, regardless of where on the ribbon the pressure is first made. While applying pressure, sliding to the right slides the voltage more positive and sliding to the left slides the voltage more negative. Sliding to the initial point of contact returns the voltage to zero. Releasing the pressure holds the voltage at the last value that was generated before release.

Obviously all of these voltages occur simultaneously. The ranged voltages are suited for controlling pitch (though they can be used for modulation purposes as well). The unranged voltages are suitable for modulating anything in the synth. The modulation inputs will modulate each of the ranged voltages, which is a powerful tool. Not shown are the controls that will allow interaction between the voltages. Mixing Bend with Initial allows a variable amount of bending to a fixed note. You can have just a small range for initial, and a wide range for bend, or vice-versa. A mid-range setting for initial and a limited range for bend modulation really produces a nice saxophone type of pitch bend, for example.

I've never had anything open my modular up like this thing. I was controlling a CGS Wave Multiplier with it last night - bend into the folds input and slide into the offset input, while initial+bend was controlling the pitch of the VCO passing through it. It was marvelous and very organic...

I'm hoping all of this will fit - I think it will - into a board about the size of the Klee Digital Board. I'd really like to see the voltage permutations included as well (which are not included in this diagram).

The next task will be a "duophonic" version.

Cheerios,
Scott

---

appendage_markII_block.PNG
Description:	Appendage Mark II Block Diagram
Filesize:	14.66 KB
Viewed:	620 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

[Scott Stites]

First of all the sample and hold - don't know what I was smoking that night, but it's actually a more realistic number: It's around 2.4 mV/min droop. This means it will hold a note for around 2 minutes before it hits the average 5 cent noticeablity point if controlling a V/Oct VCO. This isn't so bad compared to the around half a minute on the Moog Ribbon Controller,I suppose. In practice, I just can't hear a difference for much longer than that.

Got the ribbon plastered to a piece of aluminum. I've been poking at the circuit here and there, improving it. Screaming. Improving some more.

It's really rocking now - latency is around 2 mS as far as I can tell. That means if you can pound your fingers more than 500 times a second, it won't be able to keep up.

Here's a goofy/doofy sample I recorded to illustrate how fast and responsive it is. Actually, it's more of a sample of how frikkin' slow and uncoordinated my fingers are....

Anyway, it's using the slide CV, with bend added in, which creates a sort of "super slide" - you merely have to flick your finger one way or the other to really slide up or down. Sorry about the noise level - my recorder is too far away to see the bars, so I just kinda guess at the level....

I've started a web page to cover the Appendage, so I don't clog up em with a monologue.

Cheerios,
Scott

---

Appendage_Muddle.mp3
Description:	Mucking About with the improved Appendage	Download (listen)
Filename:	Appendage_Muddle.mp3
Filesize:	3.15 MB
Downloaded:	1296 Time(s)

---

[SineSurfer]

I'm feeling good vibrations heehee, nice work Scott

[bugfight]

woohoo!
awaiting the details impatiently...

[Coriolis]

[Wilba]

Is this the emoticon for waiting impatiently?

Karma is catching up with me, after making people wait impatiently on the MIDIbox forum for my PCBs, now I'm the one being teased and tortured.

Great stuff tho, but you already knew that.

[Scott Stites]

Careful what you ask for.

Here's my working schematic - it's a mess. All the connections are correct, but there are a few missing component values. No big secret, I just haven't recorded them on the schematic (I'll do that for a check-rebuild). What reference designators are there are meaningless.

Missing is the scaling and linearization, which I only have recorded in fragments. This portion acquires the sample and hold information and produces the gate and trigger signals (which are still at CMOS level in this portion - they'll need scaled down).

The acquisition range is from around 2V to around 10V (an 8V span). I could make it higher, but I'm designing this in mind for 12 or 15V operation. Putting the initial range at 2 to 12V would prevent 12V operation. Going from zero to 10 doesn't work out - it makes discerning a decent note-on event more difficult, but, more importantly, the two S&Hs begin to disagree below perhaps a volt and a half, which introduces offset error in the differential amp for bend (the "diff" CV).

So, between the buffered tap of the Softpot and the S&H's, I actually have the linearization, which introduces a calibrated breakpoint in the top end of the Softpot - this allows the portion that rises up sharply to be pulled back into a more even ramp. The output section (not shown) takes the voltage output and scales it for 0 to 10V. It's up in the air to put the bend differential there or leave it as it is. If it's here, it will need scaled on the output (not by much, right now bending from center raises 4V to right and lowers 4V to left). If it's made a differential between the scaled bend and slide CVs, then any error in the scaling will show up as an offset on the output of the bend. I'll probably leave it here and scale it separately - it's frikkin perfect feeding directly off the S&Hs.

Basically, this circuit uses a S&H as the judge and jury for the incoming signal from the softpot. Once the slightly delayed copy of the comparator signal goes high, it grabs the first voltage it sees when it's told to do so. It then gives that voltage 1.8 mS to stay level. If it stays level, the slide S&H and bend S&H simultaneously take a sample. The bend S&H is satisfied with that - it's the only sample it'll take during the event. The slide S&H takes a sample every two mS, unless the watchdog S&H detects a difference in voltage on its input and output. If it does, it tells the slide S&H to ignore things until the voltage is stable. As mentioned, the comparator signal is delayed slightly when it goes high - that discourages the CV S&Hs from taking a sample before a difference between the input and output of the first S&H is apparent. Hmmm...actually the value on that delay is 6n8 (I had it at 4n7 before). Mental note...

Anyway, this all happens very quickly. As an example, if the finger presses the SoftPot for a hundreth of a second, that press is still scanned five times.

The "long swig" S&Hs should remain LF398s. The shorter S&Hs may suffice with only a FET switch and an op amp buffer. I haven't tried that yet - I'm just happy the thing will produce gates, triggers, and CVs without a glitch in sight. It was NOT an easy process to get to this point. At least for me. The hard part was accounting for that region of pressing and releasing the SoftPot ever so gently. This circuit makes it a smooth, glitch-free process. Trust me, that SoftPot is doing a dance of madness in that region.

Anyway, the two CV S&Hs are two stage deals. The first LF398 of each S&H takes a brief, uS range gulp. The second S&H in each section savors the output of the first S&H for better than 12 mS (courtesy of the 7556). In the case of the inital note CV, that one gulp is enough. In the case of the slide S&H, it keeps gulping. The output of the second S&H of that stage changes dynamically, even though its sample time is much longer than two scans of the whole process!

Now, last night I did notice that actually the second stages are gulping promptly after the first stage has taken the sample (on the falling edge of the first stage sample pulse). I experimented with the slide S&H by inverting the "savor" pulse to the second stage so it was exactly simultaneous. There was no difference in performance, but I have a couple of extra inverters, and using them would allow me to get rid of two resistors and two capacitors.

I made the most beautiful, ethereal music of my life with it yesterday, at times using the Klee as the modulation input to the Appendage and the Appendage feeding the range of the Klee. Through a long, digital delay, of course. What do I record? The slide bass player on acid. This is typical of not only me, but I think probably anybody here, no?

Cheerios,
Scott

---

appendage_rev023.PNG
Description:	Working Copy of Appendage Schematic
Filesize:	194.89 KB
Viewed:	637 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

[Scott Stites]

Replacing a CD40106 and re-arranging with a 4081 would reduce the parts by 6 diodes, 5 resistors and 2 caps. Replacing the three "short term" LF398s with op amp buffers and FETs would subtract three 8 pin ICs and add a 14 pin IC, plus FETs and related resistors. That doesn't strike me as all that worth it in itself - the space saved would be equal to little more than an 8 pin IC, which would probably get eaten up by adding the FETs. That's even if it worked as well (which it probably would).

Anyway, keeping the LF398s puts the count of this section at 14 ICs - seven 8 pin ICs, six 14 pin ICs, and one 16 pin IC. Pin for pin, this is less than the Klee Digital Board. However, it's going to take some op ampage to do the scaling and linearization. Even at, say, five additional quads, it would be close to Klee Digital Board scale, I'd say.

[Scott Stites]

If I'm capable of pulling a M2L stunt with the CD4081 and get rid of another redundant CD40106 stage, that might leave enough 40106 stages to be able to re-work the clock and trigger into a configuration that allows the CD4093 to be eliminated. This would reduce the count by one 14 pin IC.