Week 9: Input devices

Now, it is time to program a microcontroller to measure something (Week 5). For this week, I decided to make a switch, which would turn on and off by clapping. The idea is to use this device later for my final project and enable any connected device turn on and off remotely.

Step 1: Detect the sound

First step is to built a sound detector. For that, I ihnerited the schematics from the Neil's sound input device. I took an empty PB3 pin as an output pin, which would indicate whether the clapping event has occured.

Modifying the board design to account for the 2x2 connector (connected to PB3 and ground), I strived to have no vias, and winded up with the following layout in Eagle:

Circuit design

PCB design

As one can see, the gain pin on the micrphone is not connected to any resistor, which according to the datasheet of SPU0414HR5H-SB microphone, leads to the maximum gain of 20dB, which is exactly what I need. The 10uF capacitor, coupled with 2.44kOhm microphone impedance, leads to the high-pass filter with the cut-off frequency of 6.52 Hz, which seems sufficient for our audio clap detection purposes.

After the arrangement of components is done, I proceeded with milling and stuffing the board.

Sound detection is working

After a few struggles with starting up the board (see the issues tracker below), I have finally gotten it working, and uploaded the Neil's code and the board passed the sound detection test. Now, it is time to learn how to detect a sound with the LED.

Step 2: Detect a sound

Now, to detect a sound, I took the contents of higher bits, (corresponding to 8 bits precision) in the Neil's code and compared them with the experimentally found threshold of "0x8D". I ignored the contents of the lower bit register (corresponding to the extra 2 bits of precision)

Modified Neil's code for the sound detection

The resulting sound detection performed on the connected to my 2x2 pins looks as follows

Sound detection

Step 3: Distinguish a clap

I would like to intitialize the board by performing two quick claps, which, sounds, according to the Audacity software, looks as follows:

Set of claps visualized with Audacity

We can see that my clap sequence consists of two peaks separated by ~200 ms with ~50 ms decay time. After studying the datasheet for Attiny45, I wrote a set of if statements with a timer, which check all the requirements for two claps followed by ~800ms of silence.

Clapper main code

Finally, the sound detection looks as follows

Selective clap detection

Issues that I faced

• Milling of the traces was not clean. On one side of the board, the gaps between the traces still contained some amount of copper. I just cleaned them with a small chisel.
• Milling of the outline did not work as expected. The toolpath for the outline for whatever reason crossed paths with the actual traces. I just made the outline frame larger, which resulted in a good second time milling.
• The board did not work as expected after a first run. Though I was able to program the Attiny45, the signal from a microphone was at the DC level. After checking all the connections, I have decided to resolder the microphone (with a solder paste and a heat gun), which fixed the problem. The lesson: if you use a heat gun, heat up your component sufficiently, such that all the solder paste is melted. (but don't overheat it, as it might eventually kill your component!)
• Once I started to slow down the clock speed, I have faced "rc=-1" problem. For quite a while, I thought that the problem was with the ATtiny45 becoming dead, and after resoldering a few chips, I have realized that the problem is in the relative clock speed of the avr programmer and the board. A great related discussion can be found here . In short, the board should always be at least 4 times faster than the programmer. If you have to slow down the board's clock, then you should either put ~3s delay before you declare the clock speed in your code (like I did), or slow down the programmer itself.
• More than 2 claps is quite challenging to make work properly. I have struggled with trying to come up with the appropriate timings for three claps (slow 700ms delay, followed by ~150ms delay), but eventually gave up, since the system often ignored my attempts to clap in accordance. It would be nice to instead, code the FFT transform and extract two periods, but I had no time for that (and I would probably need other microcontroller, with more memory)
• _delay_us() and delay_ms() commands for some reason underestimated time by precisely a factor of 10 the length of the real delay. This was true even for very short delays and I couldn't find the cause for this issue.

Files

• Eagle PCB design
• Eagle circuit design
• Traces
• Outline
• Clapper code
• Clapper make module