| houkime | sorry, was a bit deep in both IPC and IRL absolutely crazy stuff. | 11:36 |
|---|---|---|
| houkime | Joerg-Neo900, ok, so, vapor-phase soldering. | 11:37 |
| Joerg-Neo900 | yeah | 11:37 |
| houkime | Joerg-Neo900, soldering is not the only purpose of the courtyard, it is also about positioning accuracy of PnP and ease of repair (where cool soldering methods are not always availible). | 11:39 |
| houkime | I made myself a quick memo from all IPC I've read | 11:40 |
| houkime | now to make missing courtyards. | 11:40 |
| houkime | wow, it seems like 2 LED controllers are actually one of the hardest components in the whole v2. | 19:26 |
| houkime | 0.5 mm pitch BGA, and 6x6 balls, meaning likely 2 layer breakout with very small vias | 19:27 |
| houkime | i actually consulted Xilinx design guide on BGA breakout, and for such a thing it calls for 3/3 traces and 0.1 via hole diameter. | 19:28 |
| houkime | *0.1mm | 19:28 |
| houkime | however i compared this to a CPU and it is also .5 pitch | 19:29 |
| houkime | so this kind of thing is needed no matter what. | 19:29 |
| houkime | the only problem is that not every ocb manufacturer is able to make 0.1mm holes | 19:30 |
| houkime | *pcb | 19:30 |
| houkime | ok, will just do what metacollin did on LOWER and place microvias and necked traces | 19:37 |
| houkime | however for CPU later this will imply stacked microvias. | 19:38 |
| houkime | bug in schematics - on keyboard scanner chip unconnected rows or columns should be pulled up via resistor to vcc. | 21:51 |
| houkime | for LED controller - there is actually an up-regulating internal LDO but it seems to be not used and pins for related components unconnected. Most probably intentional, made a note in sch. | 21:57 |
| DocScrutinizer05 | we operate LEDs from Vbatt, no? So no charge pump needed for the LEDs we got there | 22:04 |
| DocScrutinizer05 | there is a note in schematics | 22:04 |
| houkime | DocScrutinizer05, yeah, 3.5V. That's why i though it is intentional. | 22:05 |
| DocScrutinizer05 | about LED 7 8 9 iirc | 22:05 |
| houkime | in keyboard scanner there another bug. IRQ is active low but it is not pulled up | 22:05 |
| DocScrutinizer05 | check if the pullup is in CPU or chip | 22:06 |
| houkime | it might be pulled up on bb-xm but idk how to check that | 22:06 |
| houkime | probably in the bb-xm somewhere | 22:06 |
| houkime | will check the docks | 22:06 |
| houkime | *docs | 22:06 |
| DocScrutinizer05 | search for beagleboard-xM schematics | 22:06 |
| houkime | it seems like this IRQ is connected to pin AG4 in processor which is mmc data pin 2. Seems like a random choice | 22:21 |
| houkime | processor on bb-xm has only 3 or so actual interrupt pins | 22:21 |
| houkime | + GPIOs that can be used as interrupts | 22:22 |
| houkime | but if one uses GPIO as an interrupt one probably wants to provide a pullup | 22:22 |
| houkime | need to check how this is supposed to work. | 22:23 |
| Joerg-Neo900 | CPU has no internal pullups? | 22:25 |
| houkime | Joerg-Neo900, for real interrupts there probably are. For GPIOs made into interrupts - don't know yet. | 22:26 |
| Joerg-Neo900 | I'm pretty sure each HPIO has optional weak pullup, maybe even strong pullup and weaj and strong pulldown | 22:27 |
| Joerg-Neo900 | GPIO | 22:27 |
| Joerg-Neo900 | some pullup mandatory for "open collector" mode | 22:28 |
| Joerg-Neo900 | btw the "IRW" from LED controller is pretty much useless | 22:28 |
| Joerg-Neo900 | IRQ | 22:29 |
| Joerg-Neo900 | needed pnly foer *very* special gimmicks that never were used or done in N900 | 22:29 |
| Joerg-Neo900 | like syncing audio to LED patterns | 22:30 |
| Joerg-Neo900 | another "funny" detail: with N900 LP5523 "IRQ" you *may* fry the CPU | 22:31 |
| Joerg-Neo900 | when you program LP5523 to output VDD there which is VBATT | 22:32 |
| houkime | IRQ going to strange places problem is not about LED controller but about keyboard controller - TCA8418 | 22:44 |
| houkime | U1801 | 22:44 |
| houkime | It goes to "13A" on v2 board, which is pin 13 on the main expansion header | 22:46 |
| houkime | which goes to AG4 | 22:46 |
| Joerg-Neo900 | kbd controller is special, we planned to use the TPS65951 IIRC | 22:46 |
| houkime | *pin 13 on the 28-pin unit of main expansion header. | 22:47 |
| houkime | It may be that AG4 is in GPIO mode and MUX is set to 4 | 22:53 |
| houkime | docs say that MUX selection is independent for each pin so it is kinda cool | 22:54 |
| houkime | If so that will be GPIO_134 | 22:55 |
| houkime | which indeed can be used as interrupt provided that it has a pullup | 22:56 |
| houkime | which might be internal | 22:57 |
| houkime | checked DM3730 CPU docs. there are both pullups and pulldowns on GPIO | 23:02 |
| houkime | ok | 23:02 |
| houkime | however, unused rows on TCA8418 still need to be pulled up. | 23:03 |
| houkime | will add some. | 23:04 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!