TCA4311A: Schematic Review

Hey Sam,

Are SMBCLK_1 and SMBDAT_1 going to be connected together to SMBDAT_2 and SMBCLK_2 through the TCA4311A on PCB2? Or are they two separate buses?

Typically the hotswap buffer goes on the "card" holding the slave and not on the backplane (I assume PCB2 is the actual backplane here) so the TCA4311As are not exactly performing the hotswap feature on PCB2. It may be better if we remove those TCA4311As unless you are using them to buffer the backplane's capacitance.

The 100k pull up and the .1uF cap on the enable seems quite weak, we may want to get rid of the cap on the enable pin and switch to 10k pull up so the device is enabled quicker. The device itself will connect the bus (card to backplane) after it sees a stop condition or the backplane idle for 100us.

It looks like you have the device flipped backwards on PCB1, the hotswap feature only works on the 'in' side of the device so SDAin and SCLin should be facing the backplane (PCB2).

Lastly could you send me an email at duynguyen@ti.com? I would like to see if your design-in period is far enough out that you could use a different device (one which has yet to be released) as I am not allowed to provide information on this device on a public forum right now.

Thanks,

-Bobby

Hello Bobby,

Thank you for your reply.

SMBCLK_1/SMBDAT_1 and SMBCLK2/SMBDAT_2 are connected to an SMBus MUX, e.g. PCA9545A, so they are two separate buses. The MUX is controlled by the master SMBus device on PCB2 and it is used because the smart batteries share the same address.

Yes, PCB2 is technically the backplane as this is where the Vcc is sourced and the master SMBus resides.

In this design: if two batteries are powering the system, and then one battery is removed, all of the TCA4311As are still powered and connected to the SMBus. Should the TCA4311A on PCB1 be disabled when its battery is removed?

I referenced Figure 14 in the TCA4311A datasheet, so the two TCA4311As are connected back-to-back using their "out" sides across the length of the cable. The battery can be hot-swapped at the "in" side of the TCA4311A, however, the cable can also be disconnected at the "out" side of the TCA4311A.

I think my confusion comes from the fact that the slave is the battery, which can be hot-swapped from PCB1, and PCB1 can also be hot-swapped from PCB2. Is there a better way to approach this?

Thank you,

Sam

Hey Sam,

"In this design: if two batteries are powering the system, and then one battery is removed, all of the TCA4311As are still powered and connected to the SMBus. Should the TCA4311A on PCB1 be disabled when its battery is removed?"

If the devices are still powered and you insert the battery, the hotswap function will likely not be used (exception should be if the device is disabled then enabled). The hotswap function works when the device is initially powered on OR you disable then enable the device. That being said, the typical use case would be you inserting the daughter card into the backplane. If you design the daughter card correctly, you should have 4 lines (GND/Vcc/SDA/SCL) such that GND connects first, then Vcc, then finally SDA/SCL (Note you need 4 lines but 3 staggered). Once this sequence occurs, the device should have had enough time to get a stable GND connection then turn on and get its precharge voltage ready before daughter card's SDA/SCL connects to back plane's SDA/SCL which would help to prevent glitches from occurring on the main bus.

Now with what I presented above, because you are using an I2C switch (PCA9545A), I do not actually expect that you would need this hotswap feature because the switch itself can provide hotswap support if you use it correctly. By this I mean, you have the channel disabled, insert the card, then enable the channel to communicate. Because the channel is disabled during insertion, the main I2C bus will not glitch and you can support a live insertion. If you can provide this kind of procedure you probably do not need the hot swap function of the TCA4311A.

Figure 14 actually is not referencing a hotswap application but instead, suggesting a long distance cable communication (note the long distance bus arrow in between the two devices in that figure). I personally do not like this device for that kind of application I suggest customers look at alternative devices if they need cable buffering because the Rise Time Accelerators (RTAs) in the device cause a larger di/dt resulting in inductive kick back events or harsh ringing to occur.

"I think my confusion comes from the fact that the slave is the battery, which can be hot-swapped from PCB1, and PCB1 can also be hot-swapped from PCB2. Is there a better way to approach this?"

Is the master only sitting on PCB 2 or are there two+ masters?

My current thinking is you may be able to support the live insertion with the PCA9545A alone if you can guarantee only one channel is enabled and have a way to inform the master when a slave has been connected to PCB2.

-Bobby

Hello Bobby,

Thank you for your reply and apologies for the delay.

The battery connector is designed so that it is possible to connect ground last, not ideal, but I cannot use another connector. There were instances of the PCA9545A being damaged and the TVS diodes on the SDA / SCL lines shorting. I believe that this is a result of the battery power return current travelling through the protection diodes of the PCA9545A and the TVS diodes, before the battery ground connection was made. I have added low-capacitance Schottky diodes on the SDA / SCL lines to protect the TVS diodes and PCA9545A, which seems to be working.

There is only one SMBus master, and it resides on PCB2 and it is connected to the PCA9545A.

1. That is interesting. So, if the PCA9545A channel is only enabled when the battery is detected and it is disabled when the battery is disconnected, then that should suffice for the hot-swapping?
2. What happens when a channel is not disabled when the battery is hot-swapped?
3. Would it make more sense to use a buffer on either side of the cable (PCB1 and PCB2), such as the PCA9517-Q1, instead of the TCA4311A? Do you have any more information as to why rise-time accelerators should be avoided? Is it potential EMI problems?
4. Do you recommend enabling the buffers only when the battery is connected to PCB1 AND PCB2 is connected to PCB1? 

Kind Regards,

Sam

Hello Bobby,

Thank you for answering my questions. I have got the information that I need.

Kind Regards,

Sam