SN74CBTLV3125: hotswap application

Hi Parker,

Thanks for your prompt reply!

I think I understood the hotswap scheme, please correct if I am wrong.

1. For SN74CBTLV3125, when SN74CBTLV3125 powered off (VDD = 0), its internal switch's enable pin will enter Hi-Z mode. So data transmission through the internal switch may still continue as OE voltage may be still low in Hi-Z mode. Is it correct? But I don't understand why it's good for hotswap application, could you kindly explain?

2. For TCA9511A,  when add-in card with TCA9511A is plugged into backplane, the I2C bus in backplane may be occupied by other devices. In that moment, TCA9511A will wait, until the current I2C communication stops, then TCA9511A will transmit I2C data from add-in card to backplane. Also TCA9511A will precharge the SCL & SDA lane to avoid large inrush current during hot-plug. Is it correct?

Hi Parker,

For your last paragraph, I guess you are suggesting TCA9545A in this application.

In fact, my customers used TCA9545A before. But they found a issue of this device, if one I2C (TCA9545A output) had problems like I2C bus abnormally pulled low, the other three I2C outputs will also not work. That's why they use SN74CBTLV3125 here.

Have you heard this issue before?

HI Miranda,

The TCA9545A is an I2C controlled switch and also support's I2C signal transfer, meaning you have to first do a I2C write to choose the required channel and them communicate the signal through the channel. If the I2C line is corrupted, it will not allow switch selection in the beginning. I have notified the I2C expert team to comment on the TCA9545A capability here.

The SN74CBTLV3125 is a GPIO controlled switch, meaning you will need external control signals to control the switch and the switch by itself only allows signal transfer through it.

Thank you,

Regards,

Sandesh

Hi Sandesh,

Thanks for your explanation!

About your sentence 'If the I2C line is corrupted, it will not allow switch selection in the beginning', it well explained the TCA9545A problem that my customers met before. So TCA9545A may not be suitable in my customers' application.

How about SN74CB3Q3253? Customers are also estimating SN74CB3Q3253, but it doesn't support hotswap in datasheet. Will SN74CB3Q3253 has same problem as TCA9545A (if one I2C fails, the other three I2C will also fail)?

Hi Miranda,

"In fact, my customers used TCA9545A before. But they found a issue of this device, if one I2C (TCA9545A output) had problems like I2C bus abnormally pulled low, the other three I2C outputs will also not work. That's why they use SN74CBTLV3125 here."

The SN74CB3Q3253 is a passive device which either turns on or off depending on the OE state. If there is a problem downstream from the SN74CB3Q3253 then it would experience the same issue as the TCA9548A since neither device has a state machine to detect stuck buses and disconnect the downstream I2C chain from the backplane. If you want a device that can determine if there is a stuck bus that disconnects the downstream I2C chain from the backplane until the problem is resolved, then you would need to use the TCA4307. 

I also see that the line card diagram you drew actually disables some of the PCA9511A's hotswap features since you block the device from direct connection to the backplane from the card. The app note that Parker referenced :

Parker Dodson said:

However this device may not be the best for your needs - please refer to this application guide on your specific application from TI. It should clarify the issues that you are trying to prevent and will give more guidance on this type of application.

 scpa058.pdf

"The first design set up that needs to be done is the ‘IN-side’ of the TCA9511A needs to be designed to make connection with the backplane from the external card. This is because the ‘IN-side’ of the device does its bus idle/stop condition detection only on the ‘IN-side’ of the device while the ‘OUT-side’ only looks to see if the voltage is above VIH (Voltage input High). For this reason, pull up resistors should not be populated on the ‘IN-side’ of the external card’s I 2C bus as it would generate a false idle condition to the TCA9511A and also turn off the 1-V pre-charge circuit. Figure 2 points out which side is the ‘IN-side’ of the TCA9511A and non-populated pullup resistors on the ‘IN-side’."<- section 4

The IN-side of the 9511A uses the pre charge to offset some of the capacitive loading in rush current when the card is plugged in, if you have the SN74 infront of it, then you will lose the pre charge feature. The In-side also requires no pull ups on it because it uses the pull ups from the back plane to keep the 1V pre charge circuit engaged until the connection is made, if you put the SN74 in front, you then are required to populate a pull up resistor and this would also force the 9511A device downstream to skip its hot insertion detection feature since a pull up would engage its start up checks and connect the out side to the inside, most likely pre maturely. 

-Bobby

Hi Bobby,

Many thanks for your detailed explanation!

1. I understood the first part --  SN74CB3Q3253 has same problem as TCA9548A (one I2C fails, the other three I2C will also fail) . So SN74CB3Q3253 is not suitable in my customers' application. Besides, TCA9548A is a good part with Stuck Bus Recovery feature, but it's not a I2C expander that customers need. So I think SN74CBTLV3125 is still the best choice here.

2. Sorry I didn't catch up the second part. Do you mean you suggest to put SN74CB3Q3253 to backplane instead of line card? Customers know that they didn't use hotswap function of PCA9511A, here PCA9511A is just a buffer. Are you suggesting to change this structure?

3. My customers are using NXP PCA9511A, not TI TCA9511A, I am trying to persuade them to change to TCA9511A.