Problems with green TWI (I2C) port on Arduino Braccio Shield

Question

Summary: The green TWI/I2C port on my Braccio shield I2C port hasn't worked for me so far. Looking for advice or information please.

Detail: I'm using the Arduino (Tinkerkit) Braccio robot arm for self education. It comes with a Uno based Braccio Shield. Tinkerkit no longer seems to be in business and documentation is a bit scarce. The shield uses KF2510 style connectors for servos (orange), inputs (white), serial (yellow) and importantly for my question a TWI/I2C port (green).

The green four pin TWI/I2C connector provides GND, +5V, SDA (A4), SCL (A5). However the SDA and SCL connections are reported to not be direct, but that "a very tiny chip (located next to the I4 connector) buffers between the green connector and the Arduino pins SDA and SCL" . A photo of the chip described is below.

Contrary to the reports of a "buffer" studying the schematic shows a level shifter PCA9306TI and two sets of 10K pullup resistors (R7,R8 and R11,R12) on the bus on both sides of the chip. VREF2 on the outboard side of the chip has a R9 220K resistor to +5V.

The other side of my I2C connection was an Arduino Pro-mini 5V. I was never able to get it working using the green port described, but it works fine directly hooked into SCL and SDA bypassing the level shifter.

My questions (sorry for the long lead up) are:

1. Has the green TWI port been level shifted to expect ~3V? (Only really came to this conclusion writing this all up clearly).
2. If I want to change the green TWI port to 5V I2C can I replace R9 220K with a zero ohm jumper?
3. The 10K pullups (R7,R8 and R11,R12) seem larger than anticipated (I had anticipated about 4.7K for 5v or about 2.4 for 3V)? Can they be removed and left open if I have a single set of pullups elsewhere on the bus? Or better suggestions?

Thanks!

Update 2021/01/30

Well my I2C debugging skills are improving researching this! As a test bed I set up a clean working I2C bus with two MCU's running the Arduino Example Master/Writer Slave/Receiver sketches by Nicholas Zambetti. I tested the I2C signal on the oscilloscope to make sure it was clean. All good. To that test bed I connected the Braccio shield (5V, GND, SCL, SDA). Connecting the Braccio shield changes the I2C signal in interesting ways, with partial strength (runt) pulses, and round shoulder slow rises. Connecting via the green TWI port seems to accentuate the runt pulse aspect. Also interesting the oscillscope was still able to decode the signal. Still working on figuring this out.

Pic 1. First test - SCA and SCL connected via pin headers in top left of shield picture above.
Pic 2. Second test - SCA and SCL connected via green TWI port Pic 3. Clean Signal (no Braccio Shield connection)

Answer 1

I fear we have a colloring issue. Your green is my blue ;-).

I have looked into the data sheet of the level shifter IC and TI only describes scenarios with a higher voltage on the right side and a lower voltage on the left side. Also the allowed voltage translations do not contain 5V to 5V. So I would say in equal voltage scenario the Level shifter may not work.

Answers

• Q1: The TWI port is always used with 5V (not ~3V). The Arduino that is supposed to use the shield might be a 3.6V version (like a DUE or M0). Perhaps the TWI port is only intended for this use case: A low voltage arduino with a 5V device. I admit: This should be noted in the documentation. So perhaps Braccio shields work outside the specification and may works under normal circumstances. But meanwhile I begin to understand why they are out of busines ;-).

• Supplement: the SCL and SCL signals are routed to two pairs of header pins as they are connected to different pins on different Arduinos. On a UNO you can not use A4 and A5 for analog as they are used for I2C for example. Do you use A4 A5 in any way? Or did you connect the second I2C pin pair in the power header when you had used the TWI connector? (rhetoric question ;-) of cause I will change the answer if you give me more information ;-) )

• Q2: The 220k resistor is necessary as it limits the current through the chip if VRef1 is lower than VRef2. Removing it with 0 Ohms would be a bad idea. As long as 5V is used on both sides, nothing would happen. But if Vref1 is 3.6Volts, you would waste energy and perhaps the chip would take part at a grill fest.

• Q3: Most I2C devices breakout boards already use their own pullup resistors. So lowering the pullups is not the best idea. -> too high current with LOW signals as the devices pull ups and the braccio pull ups are set in parallel. -- There is a chance, that the resistors are too large for devices that don't use their own pull ups, but at a 100kHz bus speed I don't think, this could cause an issue.

BTW: You do not need to exchange the pull ups, if you insist to try out different values. Just use parallel resistors to lower the resistance (10K || 10K -> 5K) [|| means connected in parallel] (R || Rp -> (R*Rp) / (R+Rp))