02:24 <wpwrak> roh: ah yes, you're right. it was that test/reset input, not the "power" button

02:24 <wpwrak> anyway, some effect. this was one of the great engineering triumphs at openmoko :)

02:25 <wpwrak> others:

02:26 <wpwrak> - the LED controlled by a BJT (transistor) without base resistor, thus loading the GPIO several times more than just driving the LED directly would have

02:27 <roh> heh.. why did we not use a transistor with internal base resistor again?

02:27 <wpwrak> - of course, the dash crystal traces, my favourite (i spotted that one :)

02:28 <wpwrak> roh: i think that was the solution in the end :) of course, it all created the minor annoyance that we have some devices with the LED active-high and others with the LED active-low

02:31 <wpwrak> - the famous EEPROM loss of the debug board, caused by massive caps (well beyond the limits of what USB allows as well), that made the board crawl out of reset so slowly that the USB chip was trying to access the EEPROM before it was operational

02:33 <wpwrak> - the "wired-or" diode on nRESET (GTA02, i think), with a reverse current in the order of 1 mA (and a forward current suitable to start a car), causing the "receiving" net to contaminate the "sending" net. the proper solution was to do away with that diode circuit entirely (it had more, smaller issues) and to just use logic gates

02:35 <wpwrak> these, plus that unintended power-on after power-down are my five favourites. there were more things, of course

02:36 <wpwrak> the Glamo graphics decelerator might deserve mention, as well

05:22 <wpwrak> i must say that the idea of, when idle, disabling the boost converter and just feeding off the silo caps is eerily tempting. that would make battery life nearly infinite :)

05:22 <wpwrak> alas, would introduce that kind of bad MCU-must-not-twitch dependency

05:22 <wpwrak> (infinite) that is, when idle. of course, if busy radioing around, or with the display on, that would change