LTC4316IMS I2C Address translator

Hi Ole,

   Looking at the low level voltages in your waveform, it appears that when the PSU is added, it is introducing a significant pull-up current and that the other I2C devices are struggling to pull the bus down sufficiently.

   I assume your waveforms are measured at the output (downstream) side the LTC4316.  There are 2 different low voltages in your waveforms.  The three lower levels in the middle are from the LTC4316 pulling the bus down with the new address.  The higher low voltage levels are the start, write bit and stop and are due to the Host driving the bus.  These are likely larger than the LTC4316's low level due to a combination of the series resistance added by the LTC4316's pass transistor, the bus resistance and the host's drain-source resistance of its output driver.

   When the PSU is added, the increased pull-up current results in the higher VOL levels.  The LTC4316's output driver is rated to have a worst case VOL(max) of 0.4V when the pull up current is 4mA and so the total effective pull-up current with the PSU is likely much greater.

   Does the system work and the waveforms just look "ugly"?  if so, perhaps nothing needs to be done but this probably isn't the situation since you've reached out here.

  I'm assuming that changing the PSU or motherboard for different ones is not possible.  I'm also assuming that reworking the I2C circuits in either is also not possible.  Thus we need to find an external solution...

  Right now, I'm thinking that adding a device which will partition the bus and can handle the larger pull-up current of the PSU might be a solution.  The ADuM1252 might be such a part. While its an I2C isolator, which you may not need for your application, it does have a very strong (5 & 50mA) current sink capability which may be enough to be compatible with the PSU.  There are other options, but they don't have as strong a sink capability and not knowing more about your system, I'm just suggesting the strongest part.

  Assuming your I2C bus topology is something like:

Host on motherboard <-> LTC4316 <-> PSU, I'd put the ADuM1252 as: Host <-> LTC4316 <-> side 1 ADuM1252 side 2 <-> PSU

If you wish to retain the isolation feature of the ADuM1252, then each side needs to be powered separately and referenced to the applicable sides' bus.  If you are just using this part as a buffer, then it is OK to connect the side 1/2 supplies together and the side 1/2 grounds together.

Eric

Hi Ole,

   Looking at the low level voltages in your waveform, it appears that when the PSU is added, it is introducing a significant pull-up current and that the other I2C devices are struggling to pull the bus down sufficiently.

   I assume your waveforms are measured at the output (downstream) side the LTC4316.  There are 2 different low voltages in your waveforms.  The three lower levels in the middle are from the LTC4316 pulling the bus down with the new address.  The higher low voltage levels are the start, write bit and stop and are due to the Host driving the bus.  These are likely larger than the LTC4316's low level due to a combination of the series resistance added by the LTC4316's pass transistor, the bus resistance and the host's drain-source resistance of its output driver.

   When the PSU is added, the increased pull-up current results in the higher VOL levels.  The LTC4316's output driver is rated to have a worst case VOL(max) of 0.4V when the pull up current is 4mA and so the total effective pull-up current with the PSU is likely much greater.

   Does the system work and the waveforms just look "ugly"?  if so, perhaps nothing needs to be done but this probably isn't the situation since you've reached out here.

  I'm assuming that changing the PSU or motherboard for different ones is not possible.  I'm also assuming that reworking the I2C circuits in either is also not possible.  Thus we need to find an external solution...

  Right now, I'm thinking that adding a device which will partition the bus and can handle the larger pull-up current of the PSU might be a solution.  The ADuM1252 might be such a part. While its an I2C isolator, which you may not need for your application, it does have a very strong (5 & 50mA) current sink capability which may be enough to be compatible with the PSU.  There are other options, but they don't have as strong a sink capability and not knowing more about your system, I'm just suggesting the strongest part.

  Assuming your I2C bus topology is something like:

Host on motherboard <-> LTC4316 <-> PSU, I'd put the ADuM1252 as: Host <-> LTC4316 <-> side 1 ADuM1252 side 2 <-> PSU

If you wish to retain the isolation feature of the ADuM1252, then each side needs to be powered separately and referenced to the applicable sides' bus.  If you are just using this part as a buffer, then it is OK to connect the side 1/2 supplies together and the side 1/2 grounds together.

Eric

Hi Eric

Sorry for late reply.

Your comments started some thoughts and I found out, that the PCB in ERROR was not built to latest BOM.

In fact, the PMBUS_IN_OUT looked like this:

Where it should have been:

The I2C_Translator should have looked like this:

But it looked like this:

Fixing the resistors to latest BOM made everything work:

Thank you.

Glad to hear you've got things working!