Would it be possible to place two LTC4310 devices on same bus, such that MCU connect to both isolators and access remote sensors on both ends?
Will appreciate your comments,
The LTC4310 work as a pair. There are two, and only two LTC4310 devices on a transformer. The link between LTC4310 devices cannot be shared by additional devices.
There can be two(or more) parallel sets, with the local LTC4310 devices talking to the same master.
There is also an LTC4331. The LTC4331 may be interesting for this application depending on how remote these sensors are. The LTC4331 pair can be implemented with isolation, and the local LTC4331 can share the master side bus with other local side LTC4331 devices and/or other I2C slaves.
Thank you for your comments,
Actually I just found some devices more suitable for the applications since LTC4310 and LTC4331 seems to be designed for long wires extensions.
I was thinking to use one LTC4312 on the master side and for slave board use LTC4307 similar to Fig.5 of LTC4312.
Would you know by chance if this design might work for just a few meters of cable? I haven't found any specs on this regard, also would you think this way might have some issues to pass EMI related certifications.
There are certainly emissions advantages sending a differential signal over cables rather than a single-ended signal. The link between the LTC4331 pair is differential and there is the SLO function available to reduce the slew rate of the transmitted signal. The LTC4331 would be well suited for an application connected by few meters of cable.
The LTC4307 or LTC4312 aren't galvanic isolators. There also isn't a simple way to drop in an external transformer or external caps like the LTC4310 or LTC4331. Is galvanic isolation a requirement?
I understand the advantages you mention regarding emissions and galvanic isolation.
The wiring is inside the same equipment so there is no need to have galvanic isolation. I am only concern about getting EMI certification passed using the single ended solution.
My only problem with LTC4310 or LTC4331 solution is that we have two slaves that are separated a couple or maybe three meters of wire and LTC4312 +LTC4307 seems less complicated and less bulky solution, specially for the slave board.
I think we might get the bus running at 100KHz to reduce emissions.
I think two pairs of LTC4331 would be well suited for this application. The transformer/caps for galvanic isolation is optional for the link. It's a 4x5mm package and needs only the external termination resistors. Sure, on the slave side its a bit bigger than the LTC4307 in 3x3 and the pull-ups, but its going to be easy to get the application going with the '4331. I highly recommend giving the '4331 a try.
The I2C signal isn't meant to be sent off board on wires. But there a lots of designs that really stretch what I2C can do. The bus buffers are going to help by segmenting the capacitance on the master side and the slave side. Slowing the data rate and the slew rate over the wires to the slaves are going to help. You can disable the LTC4312 buses when you are communicating with local slaves to keep the data rates faster.
There are some details on cascading bus buffers in the LTC4312 datasheet.