Monday, 2019-06-17

*** tpb has joined #tomu00:00
*** AmosSam has left #tomu01:33
*** AmosSam has joined #tomu01:41
*** xkapastel has quit IRC04:10
*** rohitksingh_work has joined #tomu05:19
*** AmosSam has left #tomu06:51
*** AmosSam has joined #tomu06:51
MadHackerxobs: pcb rework? Had to change things much?07:41
*** AmosSam has left #tomu11:27
*** AmosSam has joined #tomu11:29
*** AmosSam has left #tomu12:06
xobsMadHacker: reworking EVT1 boards to make them usable. It requires four bodges: two 0-ohm resistors to short, a spi flash chip to program and install (into a footprint that's slightly too small), and a crystal to install (into a footprint that's upside down)12:06
*** AmosSam has joined #tomu12:07
*** rohitksingh_work has quit IRC13:14
MadHackerOuch on the upside-down crystal. I've done that a few times.13:31
MadHackerIf there's a smaller package available, it makes the wire routing for the bodge much easier.13:31
xobsYeah.  Turns out if you mount it on its side, you only need to run a power wire.13:31
MadHackerUsing only 3 of the 4 pins?13:31
*** awe00 has joined #tomu13:31
*** awe00 has quit IRC13:32
xobsYeah.  The 4th pin is usually either "disable" and/or "NC", and can be left floating.13:32
MadHackerAs you say. I've seen some peculiar variations, but the normal ones are fine left floating.13:33
MadHackerMy Openbench logic sniffer has the clock crystal swapped out by one with reversed pinout, but I used a smaller package and just superglued it upside-down in the middle of the pads and tacked little bare wires (looks like bond wires, almost) from there to the board.13:34
MadHackerIf you're gentle, you can probably roll over the SOIC pins for the flash into almost more like a J-leg pattern and have it fit the pad nicely.13:35
MadHacker <-- the peculiar variations. VCXO in a normal crystal package, 4th pin offsets the frequency.13:45
tpbTitle: GVXO-331L | 3.3V SM VCXO - Low Current Consumption| Golledge (at
xobsThe crystals I picked use the 4th pin as a "Standby" signal to stop the clock.13:45
MadHackerA lot of them just have it as a tristate output enable, rather than an actual standby.13:45
MadHackerThe power consumption of those little 4-pin modules can be surprisingly high.13:46
MadHackerI used one on a board and found out the hard way (from the flat battery) that it drew about 15mA no matter what.13:46
xobsOuch.  That's almost more than all of Fomu.13:46
xobsThough the little linear regulators aren't the most efficient things.13:47
MadHackerThe regs are really variable. I've had some that basically draw zero unless you load them, and some with 20-30mA drawn from the input no matter what. :/13:47
MadHackerUnfortunately with almost identical part numbers.13:47
MadHackerOne of those "read the damn datasheet, twice" lessons that was, again. Nearly buggered up a big event opening ceremony that way.13:48
xobsHardware is hard that way.13:49
tntxobs: which vreg do you use on fomu ? Some x2son ones or something smaller ?13:49
xobstnt: Yeah, though I found two different vendors with footprint-compatible alternatives.13:50
xobsI don't remember if they ended up putting or on13:51
tpbTitle: fomu-hardware/LDO-MIC550x-300mA-Single-Output-LDO-in-Small-Packages-DS20006006A.pdf at pvt · im-tomu/fomu-hardware · GitHub (at
tntAh yeah, that;s the footprint I was thinking of.  I just used it on a PCB of mine and only realized how really small it was when came time to handsolder it :p13:54
xobsI used the SOT23 package for the EVT board to make it easier.13:55
MadHackerQuite a big difference on the current limit trip points on those; the TI one can trip at 250mA but the Microchip one won't trip until at least 400mA. Dropout's a lot higher on the microchip one too, at least under load.13:55
tntice40 consumes basically nothing in anycase.13:56
MadHackertnt: The *core* consumes basically nothing. The I/O pins can eat plenty juice.13:56
xobsMadHacker: true, but the target current is about 20 mA.13:56
tntsure, but the fomu has capacitive touch pads and that's it.13:56
MadHackerSurely 3x that from the LEDs?13:57
xobsThe LEDs are current-limited to 4 mA each.13:57
MadHackerFair enough. :)13:57
tntAlso, aren't the leds connected to vbus ?13:57
MadHackertnt: You'd be over-voltage for the iCE40 I/O on the red if you did that, I think?13:57
xobsTechnically you can pulse them at up to 24 mA each for 10% duty cycle.  But I'm not doing that.13:58
tntMadHacker: I was wondering that at first, but I've actually checked that the ice40 doesn't have the diodes to Vio rail on those 3 special pins.13:58
xobsThe LEDs are connected to 3.3V.  Which is over-voltage on the red, but since it's current-limited it should be fine.13:59
tntxobs: that's the whole point of those IOs, they're current sinks, not normals IOs.13:59
tntconstant current sinks even.14:00
MadHackerThere's still generally a limit on the input voltage though, so when they're in their off state, the input voltage can rise to the supply voltage.14:00
MadHackerSo 3.3v is fine here, but 5v would be over the voltage limit on most iCE40 pins. Haven't checked the special rules for the LEDs.14:00
xobstnt: right. I worried about that for a bit.  But hooking up a scope and measuring it made me feel better.14:00
tntxobs: :)14:00
MadHackerChecked, max is still 3.6 for those pins. So if you have a higher-voltage LED supply you might do bad things.14:04
tntMadHacker: you still have the Vf of the led. The ice40 DS says that  "Vdd - Vf < 3.6V" ...14:04
MadHackerYep, although the Vf is the minimum leakage Vf, which is usually less than the "normal operating conditions" one.14:04
MadHacker(you can get a microamp through an LED at a lot less than the normal Vf; that won't blow things up but minus that protection diode you mentioned it'll cause electromigration and gate oxide stress and so on still)14:06
MadHackerRelatedly, some of the microchip PICs have internal shunt regulators, which let you run them off things like 48V DC supplies, just using a series resistor on the power input pin to bring the current into the shunt's control range.14:09
MadHackerWeird having a 5V *output* pin on a microcontroller, and doubly weird it being the same pin as the input pin.14:11
tntThe up5k eval board from lattice uses vdd=5v for the rgb led, but adds a shottky (Vf=0.3v @ 100mA so at low current, it would have virtually no drop either ...).14:11
MadHackerInteresting that they felt the shottky necessary. Wonder if the minimum Vf in the LED datasheet made them nervous.14:13
MadHackerOr if it's just reverse-polarity protection.14:13
*** rohitksingh has joined #tomu14:15
tntI just measured the Vf of an "Off" led connected to the rgb pins of the ice40 and it was 1.4v drop, I guess just from leakage current.14:20
tnt(I guess technically my meter also added a tiny bit of current with 10M input impedance)14:20
tnt(That's also using the same led as the fomu)14:21
MadHackerAye, so that's right at the absolute max spec on the part.14:26
MadHacker(if it were on 5v)14:26
MadHackerActually a bit above; Vbus is allowed to be up to 5.25.14:26
*** earthnative has quit IRC14:45
*** emeb has joined #tomu14:49
*** rohitksingh has quit IRC14:52
*** cdmatter has quit IRC15:11
*** AmosSam has left #tomu15:21
*** AmosSam has joined #tomu15:26
*** rohitksingh has joined #tomu15:26
xobsBy the way, I've added initial RGB Support to the experimental Fomu MicroPython port:
tpbTitle: micropython/ports/fomu at fomu · xobs/micropython · GitHub (at
*** rohitksingh has quit IRC15:42
*** AmosSam has left #tomu15:50
*** AmosSam has joined #tomu15:52
*** rohitksingh has joined #tomu15:53
*** rohitksingh has quit IRC16:22
*** xkapastel has joined #tomu16:59
*** alexhw has quit IRC17:57
*** esden has quit IRC18:32
*** esden has joined #tomu18:33
*** xkapastel has quit IRC19:58
futarisIRCcloudxobs: Thanks. I'll try and a test it later today.23:06
*** AmosSam has left #tomu23:38
*** AmosSam has joined #tomu23:38
xobsfutarisIRCcloud: I also got an updater written, discovered an issue with Foboot on certain devices, and added the ability to read the SPI ID.  (These were all part of a yak shaving exercise.)23:59

Generated by 2.13.1 by Marius Gedminas - find it at!