Combined request-go line idea

[m-ou-se]

Just some idea: With a few resistors, transistors, and a diode, it would be possible to use a single GPIO pin on the arbitrator to get a transfer request from a device, and accept that request using the same pin, while also controlling the buffer enable line and the chip select of the arbitrator:

[kalkyl]

An alternative idea is to invert the REQ logic to idle low, which would simplify the hardware and maybe reduce the mental overhead a bit...?

Truth table:

DEV ARB         DEV DEV
1:REQ   1:REQ/nGO   ->  1:nGO   GATES
LOW In: HiZ     ->  HI  OFF Idle/Waiting for request
HI  In: HiZ/Out: HI ->  HI  OFF Request to send/Not accepted
HI  Out: LOW    ->  LOW ON      Request accepted

Notes: U1: XNOR gate (i.e. SN74HC266 or CD4077) U2: Tristate buffer (i.e. SN74LVC125) C1: sized to avoid false nGO triggers, depending on capacitive load The nCS, COPI and SCK tristate buffer gates were omitted in the drawing for brevity (read "lazyness").

EDIT: Updated drawing with correct logic gate, added C1 for de-glitching, added R2 for defined behaviour when the device is not plugged in. nCS driven from controller, through tristate gating buffer.

[m-ou-se]

That wouldn't work: With this AND port, the nGO pin would be low whenever the REQ is low. Meaning the communication is enabled while the device isn't even asking to communicate.

In my schematic, I tried to minimize both the number of pins that have to go the device (the lines that cross the leftmost blue line: just one go/req line next to the three spi lines), and the amount of extra components needed on the device (just one resistor).

[kalkyl]

No 1:REQ/nGO is in floating state, so both ports will be low, so the AND XNOR output / nGO input is HI...

[kalkyl]

Only when the arbitrator's 1:REQ/nGO is pulled low and the device's 1:REQ is hi nGO will go low

[m-ou-se]

in floating state

How could that work? How would the arbitrator tell if there's an incoming request if the pin is floating? It'll just randomly measure high or low if that pin is floating. Same for the nCS pin.

[kalkyl]

In idle state it will be pulled low by 1:REQ, but yeah true that nCS will be a problem as is...

[m-ou-se]

But if REQ is low in idle, the output of the AND port will be low, enabling the buffer and enabling the GO signal.

[kalkyl]

Oh sorry, i meant it should be an XNOR gate