PCA9615 SCL (single ended) failures

[rudyh]

I am using the PCA9615 part to extend an I2C bus. The PCA9615 converts I2C to a differential bus.

There is a control box and a remote slave box. There are existing buffers in each box. (Analog Devices ADuM1250  I2C Isolators) The slave box normally would be within arm's reach of the control box. A short cable joined the two. This worked well but now we need to have a solution for larger distances between the control and slave box. I used the NXP PCA9615 to extend the distance.

What I have is the following.

Control I2C <--> ADuM1250 <----> PCA9615 <=========> PCA9615 <----> ADuM1250 <---> Slave I2C

3.3 volt power _____|     |______________________________________________________ 5 volt power

The 3.3v side of the ADuM1250 is isolated from the right side. Everything to the right of that is powered from +5v from the slave box. I am using a cat 5 cable to join the two PCA9615. +5v and common are supplied through the four wires of the cat 5 cable. The other four are for the differential SDA and SCL lines.

The problem I have is that the SCL line of the left PCA9615 is failing. A permanent low. With an Ohm meter I get 340 Ohms to ground, with the chip out of the circuit and not powered. On the production test jig, the led is on solid and no communications happens. This has happened to two different modules at the beta site. We are trying to schedule a trip to the site. 

Has anyone been using this chip in a product? Any similar failures? I have included a schematic of the described circuit.

[rudyh]

Some further information. I said earlier that I measure 340 Ohms to ground on the damaged SCL pin. Then the circuit is powered the resistance will sink current from the pull up resistors. The voltage seen at the pin is 2 volts on one board, and 2.7 volts on the second board. Grounding that SCL line with pass the low through to the differential side. And if I generate a low on the differential side, that low gets passed through to the single ended SCL line and it sinks the current on that side. So the failure has the affect of putting a ~300 Ohm resistance to ground/common. Still a fault but the signals are being passed through.  

[guoweisun]

HI

To optimize your schematic:

1:Please change the 470ohm into 600ohm and change 147ohm into 100ohm.

2:Also increase the pull up resistor of SCL/SDA pca9615 A side into about 10K.

From your description the PCA9615 A side of SCL damaged, can you repeat your failure situation? That's considered as electrical damage? if your using condition doesn't match or above the rating limit value of SCL pin in  PCA9615,that also can damage it. If possible you can monitor the SCL signal and catch the waveform.

I don't understand why you use the level shifter IC, both of this IC sides you have 5V power input.

[rudyh]

1:Please change the 470ohm into 600ohm and change 147ohm into 100ohm.

There is little effective difference between 470 Ohms and 600 Ohms and it has no bearing on the problem at hand. 147 Ohms is only a little high. I might have 120 Ohms specified in the production builds, The values chosen were based on a number of factors. Noise immunity was decided to be a bigger factor over line matching. The distances are short and the clock frequency is only 100kHz. The minor mismatch in the line impedance is just that, minor. Using 100 Ohm resistance would not be optimum. Don't forget that the pull up and pull down resistances will also be resistance seen by the line. Using 100 Ohms would also result in a mismatched situation and will decrease the noise immunity. 

2:Also increase the pull up resistor of SCL/SDA pca9615 A side into about 10K.

Do you know that the SCL line of the PCA9615 is specified to 30mA? The PCA9615 is a Fast Mode Plus (FM+) rated device. The pull up resistance could be as low as 260 Ohms and still be withing specification for that device. Even if it were not a FM+ device, just a standard device that was rated at the minimum I2C standard of 3.0mA, the pull up resistance could be 1600 Ohms. Why do you think 10K would be of any use? That high a resistance would cause problems due to slow rise time, and also be more sensitive to noise pickup.

I don't understand why you use the level shifter IC, both of this IC sides you have 5V power input.

The level shifter is not it's only designed function. It also is capable to drive higher bus capacitance. I is designed to be a buffer as well. I had explained that the original design was for a master control box and an external slave that was connected with about two meters of cable. The ADuM1250 on the control side is to perform level shifting, provide isolation, and to provide higher drive to the external line. On the slave side the level shifting and the isolation is not a requirement, but the buffer capability is still needed. 

[guoweisun]

For the questions2:

100kHz frequency IIC doesn't fast plus mode in your system,are you sure you need it?

large sink current will also bring the higher offset of SCL.

Also bring the large over voltage/under voltage of SCL signal because the parasitic capacitive/ inductance in the PCB.

The negative voltage level on SCL also have damage risk!

The question is you can't repeat your failure condition,suggest to test the waveform of your system all questionable point to verify just mentioned last reply.

For the questions3:

I am not familiar with NLSX4373,after read datasheet I can't find any word marked as capacitive load isolation.

[guoweisun]

What's your issue state now?

Did your fix it? Do you need support next?

[rudyh]

We went to the beta site that had the SCL line damaged on two separate modules. I replaced the modules with ones with clamping diodes on the single ended SCL and SDA lines. There has not been any failure reported yet. I am hoping that solved that problem.

On the problem relating to the reset line question. I have not had time to see if adding a capacitor to ground prevented the fault condition. I did have our production department do the modification as I didn’t see any reason for it to cause a problem. I hope to do some testing at some point but I am too busy right now with a different product.

Rudy

[guoweisun]

OK,expect  for your update about this issue next.

If possible,you can apply do a failure analysis about damage ICs request from your buy channel.

[rudyh]

The pull up and pull down resistors on the differential side are 470 Ohms. The resistor between the differential pair is 147 Ohms. The cable length of the cat 5 differential cable is 25 feet / 7.5 m.  The cables between the PCA9615 are 2 meter long and are a flat cable. The transfer speed is 100 Khz.

The failures have only happened at the beta site. (we have scheduled a trip for Thursday) Our setup at work has had no issues over a month of operation. The beta site had worked properly for three weeks and then failed. A replacement module was sent out, installed, and worked for a number of days before it then failed.

Removal of the ADuM1250 is not practical at the beta site.  

[guoweisun]

Hi

For debug please remove ADuM1250 both sides of your system and connect Master IIC to PCA9615 directly.

Please also give us your detail value of resistors of your schematic.

Please let us know your transfer speed of IIC and the distance between two PCA9615.