If I hold a key down, it should report the initial keypress and then ignore because the same key is being detected. Sometimes, if I hold down the 4 key, for instance, it returns a 3 and then bounces back and forth between 3s and 4s. It happens with other keys, too and it's always key and key-1. What's up with that?

What's the center-to-center spacing on the buttons?

WARNING - there is no reverse voltage protection on this keypad, so be sure not to mix the supply pins. I, unfortunately, had to learn that the hard way...

It did not run correctly on any of my boards (Leonardo, Due). Thus, I dumped the library, determined the analogRead values for every key (which might be different per board), increase them a little bit (about 10 points) and use the following program adapted from DFrobot example code to use the five buttons on the Romeo, but added a threshold of 20, because the keyboard is noisy:

int key=-1;
int oldkey=-1;
int adc_key_val [12] = {45,85,125,170,215,260,300,345,380,425,470,510};
int adc_key_in;


void setup()
{
 Serial.begin(9600);  
}

void loop(){ 
  delay(50);   
  adc_key_in=analogRead(0);
  key=get_key(adc_key_in);
 if ((key != oldkey) && (key !=-1)){ Serial.println(key);}
 oldkey=key;
}


 int get_key(unsigned int input){
 int k;
for (k=0; k<12; k++){
  if (input<20){k=-1; return k;}
  if (input<adc_key_val[k]){k=12-k; return k;}
  if (k>=12){k=-1;return k;}
  }
}

Hooked this up and am having trouble with the default library settings. Basically the 1 key doesn't register a key press at all, 2 key reports as 1, 3 reports a 2, 4 reports 3, 5 reports 4. Then there is a gap (no 5 key), and the 6 key through 12 key work fine. I'm running this on a RedBoard with a measured Vcc of 4.987 volts. Looks like the VScaleKey table may need to be tweaked further for production variances. Here are the Vout readings for each key: 1 - 2.488v, 2 - 2.280v, 3 - 2.073v, 4 - 1.867v, 5 - 1.660v, 6 - 1.453v, 7 - 1.246v, 8 - 1.038v, 9 - 0.831v, 10 - 0.623v, 11 - 0.414v, 12 - 0.207v, Open - 0.061v. Is anyone else having similar problems with this board?

More than 20 years ago Microchip published an I/O app guide, a little 3x5in. sized printed pamphlet (remember those?) that was included with the development boards. Most of Microchip's business was low pin count parts back then, so I/O was at a premium. This guide showed how to get more bang from each I/O - using one I/O to drive 2 LEDs, PWM for DAC output, etc. This is very similar to one of those app note circuits except as I remember that one used a simple voltage divider rather than a current source. Funny how nothing ever really changes.

Just port the Arduino Library to NETMF (Netduino Board) with a KeyPress Event and synchronous ReadKey() method that will block until a key is pressed - I don't have the Keypad to test (but 90% sure it will work) so, am wondering if anybody have a Netduino and the Keypad to test the code for me before I upload it CodePlex

if you remove the opamp current source it would work across all voltage ranges (the buffer can take) ... your comparing it against the supply voltage anyway so it would save a ton of pain!

With 12 keys on analog voltage, I'm thinking: map the 12 tones of the musical scale for an analog keyboard/synth. There's a MIDI application in this. I want one.

Interesting concept! I wonder if a 'digital' version is planned?

An MCU (micro controller) in place of the op amp, reading switch closures, using key-scanning of a matrix, could then output serial data of key closures. This would require only two signal pins (plus supply and ground) for the popular I2C (or TWI) protocol.

Having an MCU allows programming of behavior, and resolution of the multiple button press issue. Or, to distinguish between a button "tap" (< a second) and a button "hold" (> a few seconds)

according to my calcs this design will only work when limited to +/- 10C of temperature change. due to the diode drop changing with temp. maybe add some temp comp, either in software or hardware. the old analog mono synths used a similar circuit and had temp comp to combat this issue.

if all you need is 1 key at a time it saves a lot of pins. wonder about temperature effects however.

Can it detect multiple button presses?