search

tigard-tools / tigard

An FTDI FT2232H-based multi-protocol tool for hardware hacking
Other
560 stars 65 forks source link

revise mode selection switch #5

Closed securelyfitz closed 4 years ago

securelyfitz commented 4 years ago

Right now the mode selection switch is a DPDT but we're only using half of it: SPI/JTAG: DO and DI separated SWD/I2C: DO and DI connected with resistor

We have three features worth considering for the second half of the switch:

  1. Adding an I2C pullup on SDA Theoretically this is necessary for I2C to work, but due to FT2232H's poor i2c implementation it doesn't actually matter. There may be corner cases where it does, but for those cases, a user can put a resistor between pins 1 and 4 of the JTAG header to get this done.

  2. Disconnecting MISO from the SPI/I2C header in I2C mode Theoretically this causes issues for i2c chips that use pin2 as their address input. I haven't encountered issues, perhaps because most devices latch their address bits at powergood and they're ignored after. The fix to this is not connecting that pin to the target system, defeating the convenience of the I2C header.

  3. Swapping between TDO and TMS on pin 2 of the SWD header This allows using the swd connector as a 10 pin arm jtag connector as well. (edit: fixed wrong pin)

Options 1 and 2 are choosing to protect against a unknown risks in the least frequently used interface that's avoidable with inconvenience. Option 3 is adding support for a frequently used feature for a large audience.

I'm leaning toward option 3, but only if both @esden and @fharding1 feel that it's worth the risk that avoiding options 1 and 2 would mitigate.

securelyfitz commented 4 years ago

Wait, does this work using only one switch: SPI/JTAG: DO connected to MOSI SWD/I2C: DO connected to DI via swd resistor

Who knew switches were so hard. Maybe i'm just getting old.

securelyfitz commented 4 years ago

More details: MISO really does need to be disconnected, but luckily that can be done with only one switch like mentioned above:

SPI/JTAG: DO connected to MOSI SWD/I2C: DO connected to DI via swd resistor

My current thought is that option 2 is mandatory but doable on the first switch, and option 3 is the best choice for the second leg of the switch.

securelyfitz commented 4 years ago

updating - i fixed an error in the initial post. tms/swdio is pin 2 of the swd connector.

fharding1 commented 4 years ago

My current thought is that option 2 is mandatory but doable on the first switch, and option 3 is the best choice for the second leg of the switch.

I think I agree, just to clarify:

The first pole will connect DO to either MOSI or DI via the SWD resistor, and the second pole will connect pin 2 of the SWD header to either TDO or TMS.

securelyfitz commented 4 years ago

Just to remove ambiguity - When switched, both the SPI header and the JTAG header (and i guess swd too) should have the correct wiring. The JTAG header is primarily for jtag, but in general, the wiring harness will also be used when trying to flywire connect to spi tespoints - not the 8-pin header which will most likely be exclusively for clips and sockets.

So: BD2 enters the ftdi via the level shifter, via what is now DI TDO MISO SWDIO After change, it should be just DI

First Pole: When in JTAG+SPI mode, the switch shorts DI to TDO MISO When in SWD+I2C mode, the switch shorts DI to SWDIO, which connects to DO TDI MOSI via resistor, currently R16 this affects the SPI, JTAG, and SWD headers

Second Pole: When in JTAG+SPI mode, Pin2 of the swd header is shorted to TMS When in SWD+I2C mode, Pin2 of the swd header is shorted to SWDIO this only affects the SWD header. SPI and JTAG are unaffected

Dilemma: Ideally in I2C mode, there's no R16 resistor, but we need it for SWD. Since we need the resistor, ideally, DI TDO MISO SWDIO would be SDA. However, on UNIT and V0.0, DO TDI MOSI connects to spi header pin 8 and this works fine as SDA

I don't think this is an issue. I also don't think that truly fixing it has a solution that doesn't require more parts. Really, we could to a 4-way JTAG-SPI-I2C-SWD switch even though it be mostly redundant, with the advantage of reducing # of unique parts at the cost of space.

fharding1 commented 4 years ago

Dilemma: Ideally in I2C mode, there's no R16 resistor, but we need it for SWD. Since we need the resistor, ideally, DI TDO MISO SWDIO would be SDA. However, on UNIT and V0.0, DO TDI MOSI connects to spi header pin 8 and this works fine as SDA

I don't think this is an issue. I also don't think that truly fixing it has a solution that doesn't require more parts. Really, we could to a 4-way JTAG-SPI-I2C-SWD switch even though it be mostly redundant, with the advantage of reducing # of unique parts at the cost of space.

Could we add a SPDT solder jumper to the back of the board that normally routes DI through R16 but also allows you to cut the jumper, solder jump the other throw, and connect it directly?

image image

securelyfitz commented 4 years ago

i like this option. It does complicate the schematic - worth documenting in the schematic. But allows those who never use SWD and heavily use I2C to get the best I2C circuit.