TCA9803: Connection with hub and an external device through cable

Hey Elder Costa,

"1- Is the connection between TCA9803 and TCA9548 OK?"

-It A side of TCA9803 to the TCA9548 look okay

"2- Is it OK to connect the port B the way I did? As there is a cable involved, I am worried with reflections or ringing. Is the limited slew-rate of port B enough to avoid these problems? If not, is there a way to add AC terminators that would not break port B operation and still solve the reflection problem?"

How long is the cable? What kind of cable are you using (cat5?)? What is your operating frequency?

I2C is fairly robust in the sense highs are 70% and higher while lows are 30% or lower so minor ringing can be acceptable. (note some devices are not voltage level triggered but edge rate triggered so it depends on transceivers and the inbetween)

"3- Are the ferrites beads OK in this application?"

I've seen an app note which states ferrite beads can help with glitch filtering. My main concern is how much inductance the ferrite beads introduce as inductance  could actually cause ringing due to inductive kickback events. I don't think it would hurt to have and then just short it if you don't need it.

"4- Maybe off-topic but, are the protection diodes suitable for this application?"

I believe this is a good question to ask, particularly because it looks like you have cable communication and this device is unique.

From what I've seen, people usually don't communicate across cables with I2C but there are some instances. In my experience I have not seen problems with this in short distance cables..... A potential problem I see would be parasitic inductance in long cables causing ringing in the circuit from a high to low transition. This can cause signal to temporarily dip below the absolute maximum negative range of -0.5V (damaging our device). In this case, choosing an appropriate schottky diode would be suggested. P82B96 (a device made for long distance I2C) suggests using BAT54A to clamp the negative swinging and helps also reduce oscillation during low transitions. Selection of these diodes will depend on the expected peak currents and the ambient temperature. I have not personally seen reflections in I2C over cables before but that may be because people usually don't do this in the first place.

You will need to be careful with putting diodes onto this device (B side in particular) because this device is unique in the way it determines which direction a low is from. Buffers typically utilize a static voltage offset to help the device determine the low direction and prevent latching but this device instead utilizes current detection. Particularly if the current is flowing out of B side. If you place diodes on B side, you need to ensure the leakage current does not exceed the maximum Iext-out parameter of the part. If it does, then the device will source it's current into the diodes and see that the current is leaving the device (this occurs when someone pulls low) but because this is continuous what will happen is the bus will latch low because the current being sunk will not stop.

Looking at the diodes you've chosen, the leakage current is in the pA range while Iext-out max allowable is 200uA. So this should be okay.

If I am reading the datasheet correctly, a negative voltage of -1V will cause the diode to conduct current and may try to push current into our device which would try to push current back into our device signalling a high.... As long as negative ringing exceeding -1V does not occur then we may be okay.

May I ask why you are adding ESD diodes? Our device does have internal ESD diodes. The rating can be seen below:

Lastly, make sure you do not add pull up resistors on B side of our device on the remote board (or another buffer with static offset connected to ours).

Thanks,

-Bobby

Hey Elder,

2.3m isn't too long so I don't immediately see a need to worry about the signal integrity. I would advise you place "DO NOT POPULATE" resistors on B side of the TCA9803 just incase you run into issues (I don't think you will) and may want to use one of our other pin to pin replacements.

Lastly, If you believe the cable's parasitic inductance could be a concern I would include an schottky diode to be placed on your board (you may not need it but it helps to have).

Thanks,
-Bobby

Hey Elder,

"Re your comment "and may want to use one of our other pin to pin replacements.", what parts do you have in mind?"

-TCA9517 (non A version is fine) it's a standard operating buffer which uses a voltage offset of 0.5V on B side.

and TCA9509, this one is similar to the TCA9803 however it's "unique offset" is on A side instead of B side (like the TCA980x). I believe TCA9803 is still better though.

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

I was speaking with one of the more experienced members of our team yesterday about your post (and the TCA9803) and he told me when the device pulls down on B side, it has controlled slewing from high to low (I originally thought it was only from low to high). This means the device is much less prone to the inductive kick back events (this is great) because its it's di/dt is controlled (linear) versus a normal pull down (a capacitor discharging and exponential decay). Hes done some competitive analysis on this device in the lab, the device performs better (much less ringing and faster rise and falls) than competitors.There's mention of this in section 2 of this app note below:

From this, I would say TCA9803 is a good pick for short distance over cable communication.

Note: You may still see ringing (inductive kick back) on the remote board when the slave devices pull low (placing shottky diodes on remote board will help). From TCA9803 A side to B side, there should be little to none comparatively.

Thanks,

-Bobby