Disable Open-Drain on SW I2C using PIC18F46K22

Pets · Joined: 06 May 2019 Posts: 6

I am implementing a software I2C bus on a PIC18F46K22.
- Single Master Network
- SCL is on PIN_A2
- SDA is on PIN_D3
The problem is that the I2C clock and data pins are open-drain which causes an issue with EMI filters.
I would like to force these pins to push-pull.
The PIC18F46K22 datasheet does not show a register (ODCONx) for push-pull / open-drain configuration.
How can a software implementation of a I2C network force a GPIO pin to use open-drain if there is no configuration register for open-drain?

Ttelmah · Joined: 11 Mar 2010 Posts: 19245

I2C is not push-pull. It _relies_ on the slave device being able to send
on the same data wire as the master, and the slave being able to suspend
operation for a moment while preparing it's reply, by holding the clock
low. Also controlling the clock is how ACK/NACk is signalled.

<https://www.silabs.com/community/mcu/8-bit/knowledge-base.entry.html/2015/10/19/why_do_i2c_pins_have-w3JL>

You cannot use push-pull driving with an I2C device. It won't work.

The PIC achieves open collector driving, by having the pin set 'low',
and then setting the pin to drive, or float as an input. Result drive low
only.

Describe your 'EMI filters.

I2C is only a 'local' bus. Designed for a few inches of transmission on
a board, or between close boards. If you need longer distances, then
you need to look at using a signalling standard designed for this, not
I2C.

asmallri · Joined: 12 Aug 2004 Posts: 1630 Location: Perth, Australia

asmallri · Joined: 12 Aug 2004 Posts: 1630 Location: Perth, Australia

Ttelmah · Joined: 11 Mar 2010 Posts: 19245

Agreed. However using this requires buffers at both ends of the bus,
not 'standard' I2C, with it's normal 3mA drive capability.
I have done I2C at 12v to a sensor head using P82B96 transceivers.
Big advantage also is that this type of configuration also supports
higher bus capacitances, which I suspect will be the problem with his
'EMI filters'.
However it is also worth realising that if the slave device supports higher
drive currents than 'standard', he might well be able to solve the
problem by simply using lower pull up resistor values. The standard
drive merrily supports 1500R at 5v, and if he is only using resistors
like 10K, this may be the whole problem...

Pets · Joined: 06 May 2019 Posts: 6

Thanks All...

Ttelmah - I agree with a different type bus... problem is this is not my hardware. I am in the process of requesting the MFG to change it to an RS485 bus.

As a temp fix I used the NOFLOAT_HIGH in the #use I2C statement

Ttelmah · Joined: 11 Mar 2010 Posts: 19245

This only affects the transition from low to high when the master is
driving the line. When the slave device is driving the bus still floats.

Try reducing the pull up resistors. What value has been used?.

temtronic · Posted: Fri May 31, 2019 2:46 pm

I'd like to know more about these 'EMI filters' ! MFR, make, model, etc....
If there is an EMI problem then EVERY pin of the PIC might be affected as well as ALL peripheral devices.
My energy control system can run on 15 miles (yes MILES) of copper wire and even near lightning strikes don't affect communications.

Jay

Pets · Joined: 06 May 2019 Posts: 6

Thanks All

Problem is this is not my hardware. I am in the process of requesting the MFG to change it to an RS485 bus.

temtronic · Posted: Sat Jun 01, 2019 5:01 am

I'm curious, what is the distance between PIC and the I2C peripheral ?
Jay

Ttelmah · Joined: 11 Mar 2010 Posts: 19245

and what is the value of the pullups.