Figure 3: The final physical chip as seen in a microscope.
We suspected that this second problem was cause by bad MP3 files, or rather that the MP3 decoder was quite sensitive in regards to what it would accept. It it was to read a number of bytes and then “crash”, this phenomenon would appear: the circuit would lock up and put the processor on hold so that it seemed like the song would run on forever as it was waiting for the decoder to request new bytes for all eternity.
By replacing our own MP3 files with the test files distributed along with the project, ev- erything suddenly started working. We could not determine for sure what properties of the MP3 files that crashed the decoder, but we had indications that parameters like sample rate, bitrate and number of channels (mono prefered) were sensitive things.
A remaining error was that the last letter in the LCD was not erased, so after switching from “pause” to “stop” or “play”, the display would display “stope” or “playe”. This could easily be fixed by adding a space to the strings in the firmware.
We recieved five circuits manufactured with serial number 67000 01/B (there were two de- signs in each circuit, /A and /B, but in each circuit only one of them was bonded). The Electroscience customer number A37370 was also written on the capsules. Of these five, four proved to work and one was broken. On this circuit we removed the lid and studied the circuit in a microscope, see figure 3.
9.5 Current dissipation and critical voltage
During normal operations at 3V DC the circuit would consume some 1.5 mA current. When we reduced the voltage the current dissipation decreased until we reached a consumption of 0.5 mA at 1.76V. Here the circuit stopped working entirely. It is thus possible to lower the current dissipation somewhat by lowering the voltage.
However it came to our knowledge that other circuits manufactured would only dissi- pate around 0.5 mA, so what we won in flexibility and configurability, we loose in power dissipation.