JimMacArthur

I built several 48-ley MIDI English concertinas a few years back (pictured) for quiet practice, and to save wear and tear on my old Stagi. I went with the bellows-and-air-pressure-sensor route. I custom-designed the keys to duplicate the force profile of the acoustic ones, because I wanted my practice tina to be as close to my acoustic as possible. In addition to MIDI, the concertinas have a built-in wav synthesizer which plays back samples of my acoustic, amplitude-modulated by the pressure sensor, which is a pretty coarse approximation of what the bellows pressure really does to the sound, but as I said, this was just for practice. I did take the time to make the instrument feel solid and look beautiful, because I feel it's important. I never intended to sell these, so I could go with very labor-intensive construction techniques. Then, two things happened. My kids grew up and there was much less need for quiet practice. But also, I found myself favoring my acoustic more and more, to the point where the electronic ones just took up shelf space. I must say that it really surprised me -- I'm hardly an acoustic snob -- but, dammit, the electronic simply doesn't have the soul of the acoustic. So I'm just posting by way of a warning -- I designed something with the same tactile response as my acoustic, but I suppose I fell down in the sound generation part. A wav player is a decent starting point, but I suspect that nothing short of full-on physical modeling will create an instrument whose responsiveness approaches mechanical reeds.

Hi Reed Rat, Just a quick note to answer one of the questions you asked Conzertino, about the coil cord. I used that approach for my scratch-built midi concertina, which used an industrial bellows, not a traditional one, but the geometry was similar. I was fully expecting the cord to get jammed in the bellows folds, but it didn't happen. I think the fact that typical telephone coil cords are a bit slippery helps. For my setup, I measured the cord so that it would loop once around the inside of the fully compressed bellows. That was enough to give an extension of well over a foot, longer than the physical limit of the bellows. FWIW, I used a 6-wire coil cord, running two power wires, two ground wires, and CAN + and - on the remaining two wires. I'm not a huge fan of I2C, and my microprocessors supported native CAN, so that was a no-think.

Thanks for your replies; it helps a great deal with the design. Don, I agree with the class-compliant MIDI USB requirement. I'm using a microprocessor with built-in USB, and I'm told there's support for MIDI, but I guess I'll find out. Re the audio amp, I'm pretty happy with my compromise of a headphone amp and a line out to drive powered speakers. It's all built into this cute little codec from TI: http://www.ti.com/product/TLV320AIC23B?keyMatch=TLV320AIC23BPWR&tisearch=Search-EN-Everything This version even has a biased microphone in, for the kids to play with audio F/X processing. I figured that there are thousands of powered speakers out there, and if I build in an amp, it will have too much or too little power for someone's needs. Chris, your suggestion about the Beagle Bone goes straight to my heart. It's a great little board, with easily enough firepower, and I'm starting to integrate it into my scientific instruments, but it's a pedagogical disaster. The prof argues that for most of these classes, the kids have only an hour or two to get acquainted with their microcontroller, so even an Arduino is a stretch, and Linux boxes will be out of the question until I can build a nice Arduino-like IDE that hides all of the sausage-making. So as clunky as the two-board solution is, a stock Arduino controlling a synth that magically makes sounds in response to SPI or I2C commands is going to be what they want. The more adventurous can get inside the synth, but there will be only one or two of those per class. (Kids these days... ) Then maybe next year they'll be ready for RaspPi or Beagle Bone. I might toss the air pressure sensor, unless the prof wants it for breath control. I included it because it's really wonderful -- sensitive, expressive, and not crazy expensive. Don, I can't remember where we left off, but I went with the MPXV7007DP, which works great at 3.3V. But it does take up precious space, and "go with simple" is the right mantra, so I might have to murder this particular darling. On my practice concertina, I had originally planned to have the button actuators open holes to relieve the bellows pressure, but that design leaked too much. My kluge is a single bleeder hole that can be adjusted manually with a thumb screw -- not perfect, but the best I could do with the design constraints. Thanks again, and keep the suggestions coming as they occur to you. As always, any resulting designs will be free. Poorly documented, but free.

Don, Thanks for reminding me to support I2C. It's not one of my favorite interfaces, but it certainly solves many expansion problems, what with chips like the MPR121 out there. And thanks also for the Open Pipe link. I've got a student who wants to make a recorder/chanter controller the hard way, basically creating a time-domain reflectometer with an ultrasonic transducer at one end of an actual chanter, and a boatload of post-processing. Open Pipe might convince her that the low-tech capacitive sensor solution is good enough. -Jim

A while back, as some of you might recall, I built an electronic concertina for quiet practice. Recently, I was asked by a professor to turn the electronics into a flexible synthesizer module for use by her music class. That assignment is mostly straightforward, because I'm too busy to make major changes. The existing design is basically a low-power 4-channel polyphonic ROMpler with 100 MB of built-in NOR-flash for samples, as well as: micro-SD card for loading new samples MIDI IN and OUT Line and headphone outs SPI port so that board can slave to an Arduino or RaspPi. Built-in air pressure sensor and 3-axis accel/tilt sensor general-purpose analog and digital I/O. Built-in Lithium-ion battery with charger circuit. The existing design doesn't support USB, but that seems like such an oversight that I might have to remedy it. The remaining questions are about the general-purpose analog and digital I/O. The board has a fast enough ADC that I can add multiplexers to allow up to 32 analog inputs. And/or I could do something similar with digital inputs. And/or I could support a switch matrix with fewer I/O lines. The problem is, the more stuff I support, the clunkier the design gets, and the prof has already voiced her opinion that smaller is better. So I figured I'd run it by the folks on this forum to get some opinions. Given the above-mentioned beastie got dumped into your lap, how would you hook up to it? A bunch of individual mechanical switches? Hall-effect sensors? Something else? As always, thanks for listening.

That's very generous, Michael, but we'll need to make sure we protect your IP. In my box especially, those samples would be hanging out in soundfont files on a microSD card. Come to think of it, though, we could probably think of ways to encrypt the files well enough to discourage casual theft.

Yup. Several sources give the BLE midi spec at 6 msec. Bluetooth audio latency varies with the driver, but is usually over 100msec. I've got a headphone amp in my concertina, so I'm all self-contained, and if I want a wireless link, it will be transmitting MIDI data, not audio.

Jody, yes, yes, yes, I couldn't agree more. For me, the holy grail of this work is a better sounding, more responsive instrument -- one whose virtual reeds start to sound --reliably -- at the slightest pressure. I suspect that the long path will use a technique called modeling synthesis, but I'm starting to see some success with simple sample players. The trick is to record lots of samples -- at least one for every note -- with various amounts of bellows loading. If you only record one sample per note, it should should be with the bellows moderately loaded before the attack. Then the synthesis algorithm is straightforward: the synthesizer plays the sample when the button is pressed, and modulates it with (suitably remapped and filtered) pressure information. If the bellows has pressure when the button is pressed, the result is a loaded attack. If not, the result is a crescendo into the sustain part of the sample. It's simple, and it won't fool a discerning listener, but you can use the technique to make an instrument with far more dynamic range than an acoustic concertina. And you're right -- there's nothing in the MIDI protocol that restricts what we want -- the onus is on the synthesizer designers. Robert, I happen to work in one of the confluences of human knowledge (Harvard University), and have access to several experts on elastomeric foam, some of whom owe me favors. One of them will look for polyurethane foams which most closely approximate springs, although he expressed some doubt. Another wants to go another route -- he's looking into foams which change their electrical properties (conductivity, permittivity, permeability) with pressure. These already exist, but they're far too dense to have the right feel. If he finds something, we could combine the spring and the position sensing into one simple, low-cost design. Re circuit boards costs, you will find that board costs have dropped dramatically over the past few years, with the emergence of "pooling" services that combine several designs into one run. I regularly get a few pieces of a simple design for less than $20 a piece. In fact, the prices are so low that I now find it cheaper to have a board made than to blow an hour drilling holes into panels. That's why the end plates of my concertinas start out as $15 circuit boards before I dress them up. It might well be cheaper to swap our designs by email and have them made locally than to deal with customs.

Robert, I think that your approach is closer to what Bruce is looking for. My approach will never result in a low-cost instrument. Your use of foam, both for buttons and bellows replacement, is particularly interesting. I never would have though that foam would have a fast enough action. Then again, there are some new foams arriving on the market that might do the job even better. I will investigate. Re analog Hall effect sensors, I totally agree that it would solve many problems with key travel sensing. Analog Hall sensor prices have plummeted in recent years. Re SMDs, they're not so bad, once you get used to them. I would be happy to design a circuit board to your specs (that goes for all readers) but much less happy about shipping things to Germany. I have blown entire days filling out customs forms.

Thanks for the BLE info, Robert. That really helps me prune the decision tree. My status report: Hardware is finished. Software barely started. At the moment, it plays, but with simple wavetable synthesis. I'm hoping that I left enough firepower to do a decent synthesis job. Robert, I also went with a telephone coil cord to attach the two halves of the concertina, but I went with CAN instead of I2C, mostly because I didn't think to use I2C. The keys use inductive sensors -- those coils on the circuit board in the foreground sense metal washers at the ends of the plastic keys. This approach has a slightly lower material cost than magnets and Hall switches, but at the expense of assembly time and processor overhead. The white rectangular block in the upper left concertina is a 1.75AH LiPo battery. It supplies >>2 hours of play time, which I think is a good compromise for me, considering that I've never played for two solid hours in my life. And if the battery runs low, you can always plug in. But mostly I wanted to show you the OLED in the upper right. These displays are crisp, eye-poppingly bright, and not all that expensive. They do consume current, but considering how small batteries are getting, it seems a reasonable compromise.

My summer circuit design just got tricky: students asking for an integrated RF link. I hate my students. The obvious choice would be to use the Bluetooth Low Energy MIDI spec: https://developer.apple.com/bluetooth/Apple-Bluetooth-Low-Energy-MIDI-Specification.pdf although I share Roy's worry about latency of a BLE link. I guess the class will find out. Has anyone here played with BLE-MIDI yet, and if so, can they suggest a short path? Roy's going a custom route, but how about you, Conzertino? Did you use BLE or Classic Bluetooth? Anyone else? -Jim

For me, it's all about quiet practice. I live in a small flat with my two sons, and the only time I can practice is when they're around. They usually prefer not to hear my endless scales, and I can't blame them. For some others, it's about expanding the sonic palette beyond free reeds. I worry a bit about the musicality of that, but then again, the concertina has always been about experimentation. Charles Wheatstone, the inventor of the English Concertina, was one of the greatest engineers of his time, and I think that he would approve of our attempts to transmogrify it. One of the O.P.'s interests is in making an inexpensive, but faithful, concertina, perhaps by replacing the reeds with electronics. I and others have been exploring that space, but, IMO, we haven't found something that is both inexpensive and a decent replica of a concertina-playing experience.

Yes, it's a DSP running custom code. What I've got at the moment is a simple 4-voice "ROMpler", playing looped .wav samples from a microSD card, without pitch shifting or special effects. I don't even have the bellows pressure tied in to the amplitude yet. I also need to package the .wav files into soundfont format. But since this is now work that I have to do as part of my job, there's a fair chance that it will get done. Re DSP vs RaspPi, it was a tough choice, and I have to admit that I'm not sure I made the right one. The DSP is physically smaller, less expensive, consumes far less power, but requires specialized coding skills. The RaspPi can outperform it, and you can code it in C, but that C needs to be carefully optimized, and the caching and Linux can cause determinism and latency issues. I may yet decide that my course should use an SBC like the RaspPI (actually, I'd probably go with the BeagleBone Green -- better I/O and real-time control), but if I do, the offer still stands: if you guys can decide what you want, I'll lay out a PCB for you.

Bruce, We might want to differentiate between concertinas with MIDI interfaces, and those with synthesizers ("ROMplers", really), batteries, and headphone amps, which can be used stand-alone, for "quiet practice." I built a few synthesizing English concertinas from scratch, but the design is very expensive -- the raw material cost runs to hundreds of dollars, and the labor time -- well, I deliberately didn't keep track. A large chunk of the expense comes from my slavish devotion to getting the button feel just right, so that I could move back and forth from my Stagi to my synthetic without having to retrain my fingers. Over the summer, I plan to gut the electrical design and make a small board containing the core of my synthesizer, with inputs for keys, pressure sensors, etc., to be used to teach a class on electronic music in the fall. That board might be something that you could use as the basis for a from-scratch concertina design. Also, I would be happy to adjust the design to meet the needs of the the concertina community, but you would have to pay for the materials and the run of circuit boards. A typical run costs $1000 for a few dozen boards (blank boards without electronics installed), which would be doable if several concertinists agreed on what they wanted. In my opinion, the sticking point in any concertina design is likely going to involve the buttons. It took me years of playing around to find a design that I liked, and it's likely too expensive for what you envision. So I suggest that if you want to pursue this, you spend some time on the button mechanism. Once you have something you like, then maybe I can design a synthesizer for it.

FWIW, Rutland Plywood Corp suffered a catastrophic fire last year. No human casualties, but the factory is down for the count. They may have licensed the brand, but I haven't found an alternative source yet. I'm definitely interested in resin-impregnated wood products, so please post if you find something good.