How to handle voltage mismatch on I2C/TWI bus as battery voltage drops?

[Stefan]

I need to connect an I2C/TWI device to the nRF51822. The I2C device is running at a fixed 3.3V (from a boost DC/DC), but the nRF51822 is powered directly from a CR2032 battery (starting at around 3.0V).

As the battery voltage drops to 2V, now the I2C bus is running at 3.3V but VDD for the nRF51822 is 2V. I know technically the input should never be more than VDD, but does the open collector of the I2C bus give a bit of leeway here? This is my first battery-powered design and it seems like a pretty typical situation - how are these I2C bus voltage mismatches typically handled?

I know I could run both devices off the same VDD but I really don't want to power the nRF51822 using the 3.3V DC/DC because I only need to turn it on occasionally.

[John Kelley]

If it's I2C, why don't you pull the lines high on the nRF51822 side instead of the I2C side? You can use the internal pull-ups of the TWI block for this (enable them only when needing to talk I2C) and 1.8v is still above logic-high for a 3v3 I2C part.

[Ole Morten]

This seems to be a good suggestion, and could very likely be worth trying. The same would be achieved by having external resistors on the nRF51822 side instead of on the peer device's side.