We have developed a PCB with I2c bus controlled by a T.I. Piccolo microcontroller. We have several I2C devices and have sometimes problems with a pressure sensor running at +5V . This is a known issue where sometimes the pressure sensor freezes the bus and keeps SDA low especially when we put noise on the pressure sensor wiring with sparks from a piezo lighter. Therefore we made a switch where we can switch off the supply from the sensor so it will run again when this happens. At the same time the PCA9507 is disabled by keeping the enable line low during 40msec and start the I2C communication again after this (and some delay time). However sometimes after disabling the PCA9507 and switching on the sensor supply again we see no pulses at the I2C bus SDA and SCK at the sensor side (= B side of PCA9507) while the pulses are present at the A side (microcontroller) while the buffer is enabled. At the +3V3 buffer we see pulses at both sides at the same time.
So it looks like the PCA9507 is put in a kind of 'lock state' after disabling because the SDA and SCL lines remain high on the B side while the A side is switching (the PCA9507 power is not switched off , only the power of the sensor). Even when we disconnect the pressure sensor so there is nothing left on the bus the SDA and SCL lines are still high. The only way to get the buffer out of the lock situation is to switch off the +5V (VDD_B) for a short moment so the buffer is working normal again. Do you have a suggestion while sometimes (by ESD discharges on the I2C lines) the buffer could remain in a lock situation? We would like to solve this in software because a hardware patch means changing 50 boards in a machine at the customer and this is a difficult job.
