randy's Recent Posts
Thanks for the comments!
Despite its simplicity on the outside the Soundplane will not be at all cheap to make. But we are working hard to make it as affordable as possible.
At http://madronalabs.com/DIY I've put a demo video and a paper showing how it works. The technology is not like an Omnichord. But I hope we can count on Devo as a customer anyway. :-)
Hi there! We're still hard at work on the Soundplane A. We have been busily designing parts and putting them together, and it's starting to resemble finished hardware. Let's take a look...
I've recently taken delivery of the flex circuits that are the Soundplane's carrier antennas—basically its only moving parts. There will be four antennas on the Soundplane with 16 carriers each, for a total horizontal resolution of 64 taxels. Here's one of the antennas in closeup.
This was kind of an unusual design because the shapes of the antennas are so critical. Though everyone seems to use it for making PCBs, Eagle would not really, as far as I could tell, have been a good tool for drawing this kind of thing. Adobe Illustrator was, and I already had it. So I looked around for a way to convert Illustrator files to Gerber, which is the format the circuit board people use. I was in luck! Some guy in Seattle had already written a Perl script for this purpose called pdf2gerb. I modified it slightly to output rectangular line ends instead of circular ones, and made some Gerber RS-274X files, hooray. Another open-source tool, gerbv, lets me inspect the Gerber files before I send them off. I was able to get gerbv compiling pretty quickly using Fink on Mac OS X.
Typically, flex circuits are little guys, stuffed into your camera or cell phone. These antennas are considered really big pieces for the process and consequently a little expensive. But flex is the only way I found to reproduce them with the very close tolerances we need. Our prototypes have been made by IPC Calflex in California.
In this picture you can see all the carriers from the top. Instead of the veneer surface we are developing, I have covered the carriers with a matte finish plastic. Now that I see it, I like the look of the Soundplane this way, and I'm guessing that some people will want this as an option.
In the background you can see one of the ubiquitous blue tarps that we Cascadians use as shelter during the winter months.
Here's the flex from the side, bringing the carriers up from under the pickup board that amplifies the received signals and sends them to the DSP. It's a really compact design and there's no way we could have pulled it off without the FFC (flexible flat circuits) and ZIF connectors.
And meanwhile at Brian's, here's the first rev of the DSP board itself, hooked up to a logic analyzer for testing. Our board is based around a Texas Instruments CPU with a combination of high bandwidth and low power. Internal to the Soundplane we have 32 carrier signals and 16 pickups signals all running at around 60 kHz. That's around 50 Mb per second. And amazingly, we will be able to run off USB power if things go according to plan. We have a DC jack on the prototype, in case things don't.
For assembling these boards, we are working with Schippers and Crew here in Seattle. Friendly, fast turnaround, quality work—I've been really impressed. And, their shop is just a short bike ride from my house.
Brian has checked out all of the board's subsystems now, and is making the rev.2 design. When we get those boards back, we can start writing the firmware that will calculate the pressure grids and send them over USB.
The analyzer itself is a pretty cool product. Made by Saleae (which I have no idea how to pronounce), it looks very approachable, but is a powerful tool for the serious hardware hacker. Its software is Windows-only for now, but they claim Mac and Linux support are coming soon.
There are a ton of details, but they are all coming together. The case design is getting a minor revision, and the metal plate that reinforces the USB jack needs to be finalized after the DSP board is finalized. And I'm still working on getting the veneer surface just right, but that's probably another post in itself. Stay tuned for more, including more DIY info and previews of our first synth software, coming soon.
Hi A.C.--
Let me take a step back and explain the relationship between FFTs and audio channels.
There needs to be one audio channel---one sinewave carrier---per carrier antenna. These have to be on electrically separate surfaces. By placing a single pickup antenna near a combination of these carriers, the distance from the pickup to each carrier can be calculated from the combination of the carrier signals induced. The pickup would take a single audio input---it is this signal that is decoded by the FFT.
If you want to play around with this idea I would start by making a 1 by 2 setup, with two carriers and a pickup, and see what you can do. With two carriers as long triangles and one pickup over them, you would have a nice ribbon controller. I have been meaning to write this up as a DIY project for a while. I am still meaning to...
Hi,
You can set up the controller in any kind of m x n configuration where you have m rows and n columns. Less than 8 will work fine, but the square grid approach will stop working as well when the squares are big. If you have only four or fewer steps on a particular axis, it will probably make sense to reshape the electrodes to use a tapered pattern, so that the voltage varies smoothly across a larger area. I've been meaning to write up more instructions on this approach.
Imagine that you have a long thin rectangle, divided into two long skinny triangles by a diagonal through it. Now induce an AC voltage in this strip at some point along it. If your two triangles are a and b, the position can be calculated as v(a) / v(a) + v(b). More info and a drawing are in the Radio Drum section in my Masters thesis.
I have tried out the MOTU Ultralite Mk. 3 and the RME Fireface interfaces. Both had pretty much identical performance. The MOTU is a bit trickier to set up the gain structure of, but it's $500 as opposed to $1200 or something.
Well, it's not quite as simple as one sensor strip per touch. A given touch will spread both mechanically and electrically, and appear on more than two sensors. Telling two close touches apart while keeping the precision of each is a complex problem in both hardware and software design. So we have a requirement (piano-like resolution) but we don't know how far past it our available time and resources may take us.
Hi Smuff,
A jack for continuous control seems like a good addition. But this may be a user mod.
One of my goals for tracking objects is to support a grid size comparable to a piano keyboard.
Parts of the software are still up in the air, but you're taking about many of the same ideas we are. It seems important to make the fundamentals open-source. in my view this is the best way to ensure our customers that their hardware will be supported for a long time.
Thanks for the note! I fixed the link.
The copper strips are just connected to the positive sides of the audio inputs and outputs. Ground is connected to the shield on each cable to a pickup or carrier from the interface. Ideally the shields would end where the copper strips themselves start. Right now there are some lengths of unshielded cable inside the 8x8 prototype. It works anyway.
The schematic for the Radio Drum (Figure 2 in the paper) shows an op-amp circuit that will work for an active version of the device.
Hi Chuck, and thanks for writing.
Durability is definitely an important design goal for the Soundplane. I think that the ability to use a wide range of gestures from light ones to really heavy ones is important.
The Soundplane A is probably going to cost more than a monome. It has a lot of analog electronics and its own DSP inside. Recall that the prototype uses up an 8x8 audio interface---the resolution of the Soundplane A will be more like 64x8.
I am definitely hearing the need for lower cost controllers though. Keep in touch and keep checking out our DIY section...
Very cool work! Did you say you are using a dishcloth as your dielectric? OK then.
Sorry about the formatting thing. For now you can just use HTML like <p> and so forth to format. I know, maybe the site is a little more minimal than it needs to be right now :-)
Thanks very much for sharing---I hope to see more of you playing your instrument in the future.
Hi, thanks for the reply! Most of the feedback I've gotten so far has been in agreement. We hope to include some extra A/D channels so people can customize their own instruments, so that covers all the bases in a way.
Since making my "Multitouch Prototype 2", I have been waiting for a really good name to come along for these devices. Waiting, because, after naming a few musical projects etc. over the years, I've found that thinking really hard about a name doesn't do much good. What's needed is time...
...because you can't plan those moments when things crystallize and sometimes you just have to get out there and do seemingly unrelated things to have them.
At NIME this year, I had the pleasure of hearing a panel of electronic music luminaries agree that pressure-sensitive multi-touch surfaces are an idea whose time has come. So it's a good time to be coming out with our product. But it also means other people will be coming out with similar things.
I think that there's room for multiple products in this market, especially if they are differentiated by technology, price point and aesthetics. One condition that seems required to sustain all of our projects in the long run, though, is a growing pool of people writing music and musical applications for them. At NIME 2009, David Wessel pointed out the brutal truth that "there are a lot of controllers by the side of the road." Most new musical inventions get used mainly by their creators for a while, after which the creators move on to the next thing. This makes sense, because inventors like to invent. But in order for a new instrument to survive, it needs to get into the hands of a critical mass of players and composers. Composers need players. Players need music to play. Sound designers need a variety of rich mappings to start from. All of this activity is mutually sustaining and obviously, the more instruments out there, the better.
So let's recognize that at least a few of us will be making different kinds of the same thing. The only multi-touch, pressure sensitive controller currently available is Haken Audio's Continuum Fingerboard. Most compositions or sound-making tools written for Fingerboard could be played on our Soundplane A and vice versa. And coming soon will be more and probably cheaper controllers based on projects like the IMPAD designed at NYU. Assuming they're big enough, and have a high enough sampling rate, all these instruments can be used to play the same pieces of music.
At the time of the piano's rise in popularity in the late 1700's, a few manufacturers were competing to improve the action. The resulting improvements led to important pieces, particularly Mozart's Piano Concertos. And of course, these pieces led to more pianos being made.
Touch-sensitive surfaces may well become the piano for the 21st century. Because the computer can map flexibly between gesture and resulting sound, mappings can be designed that work for players at different skill levels from beginner to expert. A huge variety of applications are possible, from Guitar Hero-like games to instruments for the virtuoso to sound design tools for film. No one instrument maker is going to write all of these applications. So there's a potential opportunity here for developers, but to be rewarding, the instruments have to be out there.
So: "soundplane." Generic and descriptive. It's made for making sounds (otherwise we wouldn't need a kilohertz sample rate) and it's flat. I hope that more applications and compositions than we can provide are written for soundplanes, and yes, that other people make soundplanes. Really. Our first one, taking a cue from Steinway, is the Madrona Soundplane (model) A. Following our DIY instructions, you can make your own Soundplane 8x8. And so on.
Probably this modest proposal, my attempt to name this class of instruments, will fail. These things arise organically, and are settled on after a long time. Christofori wouldn't have predicted that his gravicembalo col piano e forte" would become simply the "piano." But I'll be open to changes, and actively supporting the community that will eventually lead to more instruments, more pieces, and more people playing music.
NIME 2009 on the CMU campus had a really high good-to-sucks ratio. Here's my really brief report on the highlights...
Unlike, say, NYU, the environment in Pittsburgh did not have a lot of culinary or cultural distractions nearby. This was both a bad thing and good. Bad, because after finishing making slides at 02:30, being waken up repeatedly by fire alarms at 04:30 and managing to get about thirty minutes more of decent sleep before getting up an hour before my presentation, I found out there was no source of coffee nearby and had to take a 20 minute speed walk to the local coffeeshop. Good, because every single presenter and performer who felt like tying one on after the last night of the conference ended up at the same joint, the fun and sinister Panther Hollow Inn, pictured above.
David Wessel's performance on his Slabs controllers was one of the highlights of the conference. The Slabs are very sensitive multitouch controlers made from multiple touchpads, custom sensing electronics, and a custom-designed FPGA-based brain that provides 96 streams of control data as audio signals. Read the paper. See the video.
Like the Soundplane, The Slabs are a refined multitouch surface—I much admire the obsessive pursuit of control intimacy evident in the work. Their construction as a multiplicity of discrete elements lends itself to really different performance mappings compared to a homogeneous surface. On the Slabs, you are always sure which element you are on, so they seem to lend themselves better to mapping a collection of different voices predictably. Only one finger at a time can be detected on an individual slab, though. So the Soundboard seems better for mapping distinct touches to voices of a synthesizer that can be moved freely across the surface.
On Thursday morning we were treated to a teleconference with three prominent personalities of electronic music: Roger Linn, John Chowning and Max Mathews. Max ate his breakfast wherever he was and interjected just a few words, but really good ones. Linn said that the most crucial area for expressive control lies "in between silence and whispers." This is welcome support for my opinion that signal-based control is going to be a theme of the next generation of interfaces. Everyone seemed to agree that we are seeing the beginning of a renaissance for electronic music, and furthermore that pressure-sensitive multitouch controllers will be important. This gives me comfort that if I'm wrong about this whole thing, at least I'm wrong in the company of some really smart people.
A lot of people came by my demo table and played with the Soundplane 8x8 DIY project. I saw consistent surprise at the sensitivity to touch of what is, electronically speaking, a pretty primitive device. I know, I was surprised too. At least three people seemed quite serious about making their own. I hope they do, and that I can be of help.
The sheets of reinforced, flexible veneer arrived the Monday before the conference, so I managed to get the look-and-feel prototype of the Soundboard together just in time. I'm glad I took the thing out for a show-and-tell at this stage. People really seemed to like the feel of the surface, and on the "what goes on the blank space" question, "nothing" outvoted the next most popular suggestion, "a knob," by 3 to 1.
I seem to be establishing a pattern of going to every other NIME, having hit 2005 and 2007 and 2009. So maybe I'll see you in 2011 in Oslo.