PCA9518: Conditions under which PCA9518 input/output latches

Hi Ryu,

This device has powered-off impedance only on the I2C pins, not the enable pins.

 I have not seen this issue before, but I also don't know why powering EN first should cause a 30 second latch since you would basically be powering the device through the ENx pin through the internal pull-up resistors. Usually latching does not occur for this long of a duration, and if it occurs for this long of duration there is usually some type of damage to the device. 

This is more likely a stuck-bus issue of some sort. Do you know the voltages that are present on the I2C bus when you apply EN signal first before power supply stabilizes?

Regards,

Tyler

Hi Tyler,

Thanks for the reply.
I had a 500mV input on the bus pin and I thought this was the cause.
However, when I latch up for 30 seconds it returns for some reason.
Is there some condition for latch up?
I would like to know if there is also a condition for it to return.

Sincerely.
Ryu

Hi Ryu,

Can you provide a scopeshot of what you're seeing?

The device itself has a static voltage offset of around 500mV which can look like a step on a scope sometimes. I'm wondering if you're mistaking the offset voltage for some kind of latch up condition. (See VOL on page 6 of the datasheet in section 6.4 of the electrical parameters, it shows with low or high current the VoL is around 0.52V)

-Bobby

Hi Bobby,

I think it is latched up because 0.5V is output even though a High signal is sent.
I would like to know the conditions under which it latches up.
Also, please tell me how to restore it.

Sincerely.
Ryu

Hi Ryu,

Are there other devices present on the I2C bus? Does this error continue to happen with the PCA device removed? 

Are you able to provide scope captures per Bobby's request?

Regards,

Tyler

Hi Tyler,

Nothing but PCA is connected to the I2C bus line.
I have also confirmed that High is output correctly when PCA is disconnected.
I am attaching a partial waveform.
All pins are pulled up, but only SCL3, SDA3, SCL4, and SDA4 are powered separately and have a floating voltage of about 0.6V
Please tell me about my question as well.
I would like to know the conditions under which it latches up.
Also, please tell me how to restore it.

Sincerely.
Ryu

Hi Ryu,

I understand that no other device is connected to the I2C bus. 

Ryu Yamashita said:

All pins are pulled up, but only SCL3, SDA3, SCL4, and SDA4 are powered separately and have a floating voltage of about 0.6V

What do you mean by this statement? Do you have a schematic to help explain what you are describing here? 

Also the voltage being seen in the scope capture looks like the VOL of the PCA device ~500mV. What is being driven on the I2C bus? 

Does this bus hang up eventually fix itself? 

Ryu Yamashita said:

I would like to know the conditions under which it latches up.

We do not have conditions in which we spec for "latch up." Also, latch up is a more severe electrical condition. Usually when latch up occurs, the device experiences permanent electrical damage due to the high current endured during a latch up. It looks like the bus is hanging indefinitely, not latching. 

Have you tried this same test with another PCA device? Can you conduct an A-B-A swap and see if the issue still exists?

Regards,

Tyler

Hi Tyler,

Is there a mention in the datasheet about a possible lockup, but the conditions are not known?
If you know the logic, please let me know.
Also, when I send a pulse during the lockup state, the lockup is released.
Is there any possible reason for this?

Sincerely.
Ryu

Hi Ryu,

This is the clock line getting stuck low? 

Do you know if there are any devices on the bus that can clock stretch?

I'm wondering if there are any devices powering up and getting a bad PoR that cause them to clock stretch upon power up. 

Does this go away if you remove the device and power up?

-Bobby

Hi Bobby,

Both clock and data lines are latched Low.
Symptom improves when device is removed.
Please provide an answer to the question.
Is there a mention in the datasheet about a possible lockup, but the conditions are not known?
If you know the logic, please let me know.
Also, when I send a pulse during the lockup state, the lockup is released.
Is there any possible reason for this?

Sincerely.
Ryu

Ryu Yamashita said:

Is there a mention in the datasheet about a possible lockup, but the conditions are not known?

We are not aware of any known conditions like this (We haven't seen a lockup on PCA9518 where all signals low up during power up). We have seen some PCA devices have issue if the device powers up from a non GND state. Your scopeshot looks like its starting from 500mV. Is it possible for you to try to repeat the issue by powering up from GND? Maybe you can tie a bleeder resistor to Vcc like 20k then power up.

Ryu Yamashita said:

Also, when I send a pulse during the lockup state, the lockup is released.

A pulse to where? Which pin?

Do you have a schematic we can review? 

-Bobby

Hi Bobby,

Pulses are sent to SCL and SDA of Bus0 that are latched up.
A schematic is attached.

Sincerely.
Ryu

Ryu,

Tyler and Bobby are both out of office at the moment. Please allow some time to respond, it may be until next week.

Regards,

Eric Hackett

Hi Eric,

Yes, I understand.
I am waiting to hear from them.

Sincerely.
Ryu

Hi Ryu,

Have you had a chance to try Bobby's suggestion on bringing VCC close to GND before conducting a POR? It seems that the VCC level rests at about 500mV before being powered up. Can we try powering from GND instead by implementing a bleeder resistor of about ~20kohm?

Schematic shows PCA device and corresponding pull-up resistors on both sides of the device. Are there any other I2C devices present on the I2C BUS0? 

Regards,

Tyler

Hi Tyler,

Thank you for your reply.
This symptom subsided when I suppressed the voltage float (common power supply for device power supply and BUS3,4).
Why does the latch-up occur when there is voltage floating?

Also, there are no other I2C devices connected to BUS0.

Sincerely.
Ryu