AirHarp Shield

[lyratron]

How to build an AirHarp - Part 1: Hardware
http://www.youtube.com/watch?v=DRsAbDTGRqc

How to build an AirHarp - Part 2: Firmware
http://www.youtube.com/watch?v=pHOSCIRoJG8

Also, http://www.lyratron.com now features a web store in which the AirHarp Shield 1.2 PCB can be purchased!! Thanks for motivating me to do this, Adafruit!!!

Yours in tinkering,
~ Peter

[lyratron]

Oh yeah - here's the "Instructable", which includes both vids:

http://www.instructables.com/id/How-to-build-an-AirHarp/

8D

[itcamefrommars]

I've mentioned a few times in other posts about using ebay for getting components cheap for first few batches of kits.
I'm gonna start a post and will edit with a link here!

Here's the post :http://forums.adafruit.com/viewtopic.php?f=26&t=22530

[adafruit]

sorry for the delay, we said we'd take a look at this ok so this is your current BOM

* 1 Arduino Uno ($30 retail)
* 1 Maxbotix LV-EZ1 ultrasonic rangefinder ($25 retail)
* 1 AirHarp Shield PCB (~$15 from BatchPCB)
* 5 Pushbuttons, 12mm ($.40 /ea. in quantity 100)
* 34 Break away header pins ($2.50 for a stick of 40)
* 1 LED, 50,000 millicandela blue + 220 ohm resistor (<$1)
==================================================
TOTAL COST: $75.50

but there's a lot of things we can cut the price down on. first up, instead of a full arduino, we suggest like others have, either an AVR with a 'bitbang MIDI' implementation (there may not be one out there) or an atmega32u4 such as the breakout we have. It has mounting holes and built in usb MIDI. lets say you decide to buy more than 10 from us at a time and you buy them during our 10% off coupons - so thats 20% off 20 - $16 each - you can port arduino code fairly easily to the 32u4 as paul has done the work of making a fork of the IDE.

Next, the maxbotix is a fine fine rangefinder BUT as you have mentioned you may be better off using a DIY version by making the 32u4 do the 'heavy lifting' of measuring the delay - its certainly fast enough. http://www.kerrywong.com/2011/01/22/a-sensitive-diy-ultrasonic-range-sensor/ is one example. we suggest keeping at it, to get your version working well using filtering techiques, as its a major component of the cost. Now your sensor is $5 not $25

PCB - $15 is really just the prototype cost, this will probably be $5 once you make 10-50 PCBs. see http://www.ladyada.net/library/pcb/costcalc.html for some better ideas on how much it should cost.

pushbuttons - 0.40 is about right if you're buying 100 buttons at a time

break-away header - http://www.jameco.com/webapp/wcs/stores/servlet/ProductDisplay?langId=-1&storeId=10001&catalogId=10001&pa=68339&productId=68339&cid=PDF jameco has them for $0.40 each

LED plus resistor - there's an LED on the 32u4 board but you can get blue LEDs for about $0.25 each - heck we have them for 35 cents each http://www.adafruit.com/products/301. resistors are a penny each, so ignore that number.

So the cost is $16 + $5 + $5 + $2 + $1 = $29 say $30 - a 60% cost reduction! Eventually you could reduce the price even more by getting more PCBs, integrating the 32u4 onto the PCB (and soldering it for customers) and buying in bulk. You could get it down to probably $20 or less in cost if you had to make 100 of them

[lyratron]

Dear Ada,

I'm greatly indebted to you for the time and expertise you've lent me! I'm sure you'll agree that time is the most valuable of all commodities. I deeply appreciate it and I hope to return the gesture some day.

I'm a little afraid of the 32u4, but you've convinced me to give it a shot. I'll be purchasing one of those from your shop before the day is done (along with a boatload of other stuffs!)

It looks like the ADC sampling rate is the same as the 328 (15ksps), so little chance of using DSP to circumvent the need for some dedicated analog signal conditioning hardware (that's what my ghetto ultrasound experiment was all about - hence the ghetto-ness). Needless to say, I'm a lot more comfortable with digital than with analog. In fact, analog is a dark and mysterious world I would just assume avoid!

Thanks for the tip on cheaper PCB fab houses! I'm setting up accounts with them as we speak.

The more I think about it, the more this idea of yours makes perfect sense. $30 is way more manageable than $75! Assuming I can achieve this, would you recommend offering them for $42 (30 * 1.4), or $58.80 (30 * 1.4 * 1.4)? I recall reading in one of the links you provided me that failing to account for a retail markup could curtail any possibility of retail distribution (since retailers would then be undercut by their source and unable to charge a fair margin). I'm pretty clueless here, 'cause A) there's no precedent for how much an "airharp" is worth, and B) I'm completely new to running any sort of business!

You guys are so awesome. For what it's worth, I intend to frequent your shop for every possible part from now on. I see now that you do a lot more than just "sell parts". Yours is a business model the world desperately needs!

Gratefully Yours,
~ Peter

[adafruit]

thats the spirit! we really do suggest trying out the teensy or 32u4, as its smaller and cheaper and the UNO/midi thing is a real pain. for pricing, we suggest two margins to start, so lets say you include a nice laser cut grippy and a miniB USB cable (check monoprice.com for low cost cables!) - $75-$80 for a completed thing is fair given that you'll quickly be able to lower YOUR cost by buying more. REMEMBER THAT MARGIN IS CALCULATED WEIRD! 40% margin means multiplying by 1.66! so $30 * 1.66 * 1.66 = $83!

Since this is a fairly simple project, you may even start by saying something like "first 25 orders are fully assembled and tested! future versions may be kits" - this will get you quick customer feedback without the issue of kit assembly errors. what its worth, thats really impossible to know. but personally, we think that somewhere around $75-$100 is worth it for a supported, good quality, (assembled?), working airharp. the neat thing is people will be able to plug it in, and modify the MIDI messages if they want to using the IDE. thats a bonus!

and remember, if you have extra margin, and the project doesnt work out for some reason (it happens) you can always say "Clearance sale! these are now half off!" and you can recoup your initial inventory - nobody can resist a sale

for sonars, we suggest looking here for ideas
http://www.futurlec.com/Ultrasonic_Sensors.shtml
they have very slow shipping and customer service but maybe thats OK for you to start the maxbotix doesnt use a dsp, it uses an even slower PIC from what we recall, but sonar projects are complicated. that said, if you can get it working then thats the star of the project because sonar is waaay better than IR or other light-based techinques

as you make these, you'll find ways to reduce cost but keep the quality high by adding yer ingenuity. if you open source your design, we can continue to help you find better choices for your components!

[lyratron]

Thanks so much! You're a godsend!! I... didn't even know that bit about the margins! Wow, I do have a lot to learn about business.

BTW, I'm working on a production version of the light harp as well:

http://www.lightharp.com

A less expressive instrument than the airharp to be sure, but way cheaper (no ultrasound) and arguably a bit more straightforward to play.

Do you know of any board houses that will do a PCB that's 2 feet in length? Otherwise I was thinking of designing tiny little 1" square boards that each mount a photoresistor, LED, resistor and maybe a JST plug, allowing anyone with an Arduino to build a light harp for about five bucks in about as many minutes. Oh yeah - the AirHarp firmware is also the light harp firmware! Just REM the line that says:

int DEVICE_PROFILE = COMBO_AIRHARP;

and un-REM:

//int DEVICE_PROFILE = STICKHARP_PRO;

I was too lazy to fork the code, so I just used a billion software switches instead.

[adafruit]

dont know about long PCBs - you'll have to contact them to see what the max size they can make! big PCBs can really add up tho.

[TheFallen]

If you were looking to DIY your own ultrasonic ranger have a look at the schematic for the SRF04 with more info here. It uses a RS232<->TTL chip to generate a high voltages for the output transducer and the negative voltage for the analog circuitry.

[adafruit_support_bill]

Do you know of any board houses that will do a PCB that's 2 feet in length?

I've got a 1" x 26" PCB here that was done by Candor in Canada: http://www.candorind.com/index2.asp

[lyratron]

I've got a 1" x 26" PCB here that was done by Candor in Canada: http://www.candorind.com/index2.asp

YES!!! That's exactly what I'm looking for! Thank you!!

[lyratron]

If you were looking to DIY your own ultrasonic ranger have a look at the schematic for the SRF04 with more info here. It uses a RS232<->TTL chip to generate a high voltages for the output transducer and the negative voltage for the analog circuitry.

Thanks for those links! That looks pretty advanced to me (about on par with the Maxbotix module itself). I was really hoping to cut some corners if possible (or come up with a novel, elegant way to do proximity detection that's optimized for the needs of this project). I know for instance that if I had a fast enough ADC, I could do a basic FFT on the sampled waveform and that should, in theory, allow for triggering accuracy approaching that of analog filtering without any additional hardware. I wish the Arduino's ADC's were just a little faster...

I'm not trying to put Maxbotix out of business here. They make an excellent product and they have every right to ask $25 for such a marvel of engineering. My system has to be fairly precise, but absolute accuracy is not a requirement, nor is a range capability beyond 2 or 3 feet! Due to the inverse square law, return bounce strength at this range should be about an order of magnitude greater than at 10 feet, so I'm not sure the signal even has to be amplified. In my experiments, I discovered that switching the Arduino's analog reference voltage to 0.1v (from the default 5v) allowed for sufficient amplitude resolution to detect return bounces, but the slow ADC made FFT analysis extremely unreliable. I'm gunning for a software solution because A) when the only tool you have is a hammer, everything looks like a nail, and B) it would be a 1-time expenditure of time, energy and cost, whereas a hardware solution would necessarily involve repeatedly purchasing bags of components and building hardware from those components - a cost I would have to pass on to the customer.

[TheFallen]

I'm just saying, hypothetically your ideal sensor would:
range would be 0 to 100cm but precise enough to get centimeter-ish accuracy?
readings every 50milliseconds
and cheap.

As for FFT, why do you need to do that? you'd what to do maybe 10 40kHz pulses and then wait for some sort of response which will between 0 to 3milliseconds after your first pulse. will be a little complex but it would be a lot cheaper.

[lyratron]

The FFT would be for purposes of rejecting false triggers from extraneous sound sources (anything from a passing truck to the "click" of one of the airharp's own buttons). This is usually done with an analog bandpass filter I believe, but it should be possible to do it in firmware given a sufficiently fast sampling rate. Between the cost of a single ADC and the cost/hassle of designing and repeatedly assembling complex analog circuitry, I'm heavily inclined toward the former.

I'll do it the analog way if I have to, but it may take a while since my experience with analog is very limited. Also, I may have invented a new type of optical prox sensor the other night (~possibly!~), so there's that. I'll have to run some experiments before making any bold announcements. If it does work, it will be outrageously cheap! ;?)

[adafruit]

you probably dont need an FFT, if you do some basic low pass filtering (ie averaging) and/or looking for outliers.