21:18 <electronic_eel> whitequark: I was just browsing through the selftest code to get an idea how to fix/improve it, but then I stumbled upon something else

21:18 <electronic_eel> the OE signal goes from the fx2 to the ice40. do we really need it that way?

21:19 <electronic_eel> couldn't we just tap the enva and envb signals on the ice40 instead? I don't really see an application where you would enable the ice40 outputs but not any of the ports

21:20 <electronic_eel> that would save us one additional pin on the fx2. we were struggling to free some for revD, so this move might give us one extra

21:21 <electronic_eel> or did I misunderstand the use/importance of the OE signal?

23:08 <whitequark> electronic_eel: er

23:08 <whitequark> that OE signal is for the FX2 bus

23:29 <electronic_eel> ok, so I probably misunderstood that. So it is not the one switched by REQ_IOBUF_ENABLE?

23:31 <whitequark> that request only changes ENA/ENB pin state

23:32 <electronic_eel> hmm, I'm confused by the code in dac_ldo, iobuf_enable() there switches PIND_OEQ_N

23:37 <whitequark> let me check things again

23:38 <whitequark> ok so, first, take a look at GlasgowPlatformRevC1

23:38 <whitequark> there is no OE input other than on the FX2 bus, right?

23:39 <whitequark> oh

23:39 <whitequark> sorry, I see what you mean now

23:41 <whitequark> electronic_eel: you are 100% right to be confused

23:41 <whitequark> because that part *is* confusing :)

23:42 <whitequark> PIND_OEQ_N does not exist on revC at all

23:42 <whitequark> it was driving the nOE input on FXMAs on revAB

23:42 <electronic_eel> ah, ok, that explains it

23:43 <electronic_eel> thanks

23:43 <electronic_eel> so no new pin to spare :(

23:43 <whitequark> yeah

23:44 <whitequark> IIRC we did settle on some solution though?

23:44 <electronic_eel> yes, we did find a solution IIRC

23:44 <electronic_eel> but would be easier if we had one more

23:49 <electronic_eel> do you know offhand why the selftest enables nOE after the tests in https://github.com/GlasgowEmbedded/glasgow/blob/master/software/glasgow/applet/internal/selftest/__init__.py#L175

23:49 <electronic_eel> while setting the io-voltage to 0

23:50 <whitequark> yes

23:50 <whitequark> the reason it manipulates _iobuf_enable at all is revAB

23:50 <electronic_eel> yeah, that part is clear, nOE is only on A and B

23:51 <electronic_eel> but why not disable it at the end of the int/ext/pull tests?

23:52 <whitequark> iirc nOE is normally always enabled or something

23:52 <whitequark> yes, it is enabled in iobuf_init_dac_ldo()

23:53 <whitequark> in fact it is only exposed for selftest

23:54 <electronic_eel> ah, ok, so you just set it to disabled in selftest to get the pins-int test not to interfere with the outside

23:54 <whitequark> yes

23:55 <electronic_eel> ok, thanks for explaining. understanding the details of selftest is key to fixing it

23:59 <whitequark> I think the main thing we need to decide is whether to keep revAB support in selftest

23:59 <whitequark> elsewhere it doesn't hurt, in selftest it's troublesome