TCA9617A: TCA9617A/B A-side pins (SCLA and SDAA) state

Hi Bobby,


A version may be low impedance if VccB is not powered.
Does it means TCA9617A, A version, may keep Low, even if external Pull-High?

Hi Bobby,

More questions as below,
1. If VCCB pin was not powered, what are the A-side pins (SCLA and SDAA) state? H-impendence in 9617_Channel A internal? Or else?
2. If the both of Vcc Pin power on sequence is always different, does the TCA9617 has to work normally after both of power rail is ready? Will possible to get any abnormal working?

Thank you

Annie,

"1. If VCCB pin was not powered, what are the A-side pins (SCLA and SDAA) state? H-impendence in 9617_Channel A internal? Or else?"
-The I2C pins on this device are: an internal buffer which is high Z or an output stage which is a pull down FET, if the FET is on then will pull to GND but if off it will be High Z.

If you have A powered but B unpowered the state will be unknown for the A version.

"2. If the both of Vcc Pin power on sequence is always different, does the TCA9617 has to work normally after both of power rail is ready? Will possible to get any abnormal working?"
-VccA should never be higher than VccB even during turn on.

Thanks,
-Bobby

Hi Bobby,

About your answer below, could you please help to provide a more clear answer on it?
What is unknown? Is it floating or low impedance at A side?

-The I2C pins on this device are: an internal buffer which is high Z or an output stage which is a pull down FET, if the FET is on then will pull to GND but if off it will be High Z.

HI Nguyen,

 

 Please help to check this further. Thanks

 

1.      If VCCB pin was not powered, what are the A-side pins (SCLA and SDAA) state? H-impendence in 9617_Channel A internal? Or else?"

-        The I2C pins on this device are: an internal buffer which is high Z or an output stage which is a pull down FET, if the FET is on then will pull to GND but if off it will be High Z.

If you have A powered but B unpowered the state will be unknown for the A version.

Questions: In same case, how about 9617B?

 

2.      If the both of Vcc Pin power on sequence is always different, does the TCA9617 has to work normally after both of power rail is ready? Will possible to get any abnormal working?

-        VccA should never be higher than VccB even during turn on.

Question: Does the 9617A has to recovery and work normally once VCCB powered? even VCCA power ramp-up is faster than VCCB.

 

 

Hey Helios,

"Questions: In same case, how about 9617B?"

B version should remain in high Z.

"Question: Does the 9617A has to recovery and work normally once VCCB powered?"

Yes, after VccB goes above POR the device is expected to work normally.

"even VCCA power ramp-up is faster than VCCB"

-The SDA/SCL pins may NOT be high impedance when this occurs, so this means the device could potentially pull the line low before VccB turns on and glitch the I2C lines.

-Bobby

Hi Bobby,

For TCA9617A, previously you mentioned VccA should never be higher than VccB even during turn on.  Considering the case that  VCCA is supplied before VCCB, will the chip be back to normal function after VCCB is supplied?

Thanks!

Antony

Hey Anthony,

Yes, VccB controls the POR circuitry therefore this means that the device is supposed to reset itself once VccB goes above the POR threshold.

-Bobby

Hi Bobby,

Thanks for your reply.  On customer board, they may have B side in Standby and A side in power on stage.  So, there could be long period of time that VCCA is supplied but VCCB is not.  For such scenario, will there be any leakage current or any risk such as IC damage?

Thanks!

Antony

HI Bobby,

In parallel, Quanta still requests more precise explanation about the state of SCLA and SDAA when VCCA is supplied but VCCB is not supplied. Previously you mentioned it's unknown state. Does it mean it's floating? Or we can only say it's unknown so there're many possibilities?

Thanks!

Antony

Hey Antony,

Please see below:

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

"Previously you mentioned it's unknown state. Does it mean it's floating? Or we can only say it's unknown so there're many possibilities?"

-This device was not intended for VccA to be larger than VccB. This is partially what I meant when I said the state is unknown: the POR did not initialize yet so it has not properly turned on, at the same time the A version of this device had issues with keeping the device high Z when turned off though I believe this was also associated with the logic on the enable pin (if properly turned on by powering B side then this would not be a problem).

Looking at this again, from the image above, the ESD cells turn on because VccA  > VccB meaning there is a possible path for current to travel through this device (something I had missed until you started asking about leakage current).  

Sorry for the confusion,

-Bobby

Hi Bobby,

Some more questions from customer boards here.

As shown below, they uses two TCA9617A with A-side signal connected together.  As previously mentioned, they would have a scenario that P3V3_Stbyis supplied, but P3V3 is still 0V, and we understand that the leakage from A-side to B-side @ U3C1 could be expected behavior.

Due to leakage mentioned above, the hi-level voltage of SDA and SCL of the top TCA9617A is impacted by the bottom TCA9617A.  Currently they see two different results regarding the I2C access through the top TCA9617A as below.

1. The I2C access is still workable, and the hi-level voltage of SDA and SCL of the top TCA9617A  is around 1.4V.
2. The I2C access is not workable, and the hi-level voltage of SDA and SCL of the top TCA9617A  is around 1.3xxV.

So, the questions from Quanta would be

1. We see different hi-level voltage at SDA-A and SCL-A when the leakage is observed when VCCB is not supplied.  Does it mean the leakage is getting more serious when the leakage period (VCCB is not supplied) is longer and in the end hi-level voltage at SDA-A and SCL-A  become lower?  Can we also say the impedance between A and B side are getting lower as well?
2. In TCA9617A, the VIH is 0.7xVcc.  Now Vcc is 3.3V with SDA_A/SCL-A level around 1.4V which is far below 0.7xVcc.  Why I2C access is still workable?

Thanks!

Antony

Hey Antony,

"Does it mean the leakage is getting more serious when the leakage period (VCCB is not supplied) is longer and in the end hi-level voltage at SDA-A and SCL-A  become lower?"

This sounds like the Vforward of the diode is changing slightly due to temperature because of the leakage current.

Diodes typically will have a lower forward conducting voltage as they heat up. Please refer to the graph below taken from https://na.industrial.panasonic.com/blog/photomos-input-circuits.

I suspect the leakage current will change slightly before the temperature stabilizes.

"Can we also say the impedance between A and B side are getting lower as well?"

(VccA-Vf)/Ileakage=Z

Vf seems to be dropping which seems to imply if the leakage current is constant then the impedance is actually getting larger.

"In TCA9617A, the VIH is 0.7xVcc.  Now Vcc is 3.3V with SDA_A/SCL-A level around 1.4V which is far below 0.7xVcc.  Why I2C access is still workable"

Am I correct in the assumption that the top TCA9617A is the one you are referring to here? Can you provide scope shots of both working and failing cases?

Thanks,

-Bobby

Hi Bobby,

Please see the waveform captured at A-side and B-side as below for the 1.4V case which they mentioned the I2C access is successful.

A-side

Bside-1

Bside-2:

Antony

Hi Bobby,

THere's something I need to correct as below about the waveforms provided yesterday.

1. Please check the bottom of this email for a more complete topology.
2. As previously explained, the siuation is "P3V3 not supplied" which causes a lekage happens at TCA9617A-2 A/left side, and impact the voltage lvel at TCA9617A-1 B/left side simultaneously.
3. So. among the 3 waveform provided in last post, the first (top) one is actually measured at TCA9617A-2 B/left side,  The 2nd and 3rd ones are measured at TCA9617A-1 A/right side.
4. In summary, at TCA9617A-1 B/left side, the votlage level is around 1.4V and the voltage level is around 3.3V at TCA9617A-1 A/right side for this case.

What customer observed in this case is,the controller can still recognize FRU and TMP75 correctly.  So, the questions includes,

1. Why TCA9617A-1 seems still work normally when the B-side signal is only around 1.4V?  (What is the tolerance of our VIH and VIL?)
2. Regarding TCA9617A-2, previously you mentioned that when the impedance is getting bigger, we see a lower voltage level at TCA9617A-2 A-side.  Do you think you can conclude such impedance between A and B side would get bigger during such leakage condition?

P.S.: For another case with TCA9617A-1 B/left side voltage level around 1.3V, the controller is not able to recognize FRU and TMP75 correctly.

Thanks!

Antony

Hey Antony,

Thanks for clarifying, signal path seems clearer to me now.

"Why TCA9617A-1 seems still work normally when the B-side signal is only around 1.4V? (What is the tolerance of our VIH and VIL?)"
-I don't have an immediate answer for you yet however this device is expecting the SDA/SCL signals to be above ViH and below ViL. The device was never meant to have it's SDA/SCL lines sit between ViH and ViL. If the signal sits between ViL and ViH then I do not know what the device will in interpret the signal as (is it how or low?) as this is a grey zone in the operation of the device.

I am currently trying to get access to the internals of the device to get a better understanding of what is going on here however I am running into some roadblock on my side. Even if I do get the access I need, it may be difficult for me to pin point what is going on at the low level of this device as you are pointing out that there are two results (device does work at 1.4V and does not work at 1.3V).

What I plan on checking is how the input is compared. (Schmidt trigger referenced to a voltage divider? or is there something else?) This may shed light on how ViH is determined.

"Regarding TCA9617A-2, previously you mentioned that when the impedance is getting bigger, we see a lower voltage level at TCA9617A-2 A-side. Do you think you can conclude such impedance between A and B side would get bigger during such leakage condition?"
-We need to verify if the current is remaining constant or is actually changing. If constant then we can say the impedance is getting larger. If the current gets smaller then we can also say the impedance is getting larger. The impedance would only be smaller if the current actually increases during this.

Is there any reason why the customer cannot swap out TCA9617A-2 with our B version of the device? This swap should fix the issues they are seeing currently.

-Bobby

Hi Bobby,

Customer did plan to replace TCA9617A by TCA9617B, but their end customer needs some explanation to what we see now.

Regarding the increasing or decreasing of leakage current that you mentioned, is there a way to know it should be the increasing case or decreasing case on customer board?

Please keep us posted regarding your investigation about VIH and VIL.

Thanks!

Antony

Hey Antony,

"Regarding the increasing or decreasing of leakage current that you mentioned, is there a way to know it should be the increasing case or decreasing case on customer board?"

the easiest way is to use a digital multimeter (ammeter) while the I2C lines are high and watch the current to see if it is getting larger or lower. You can do this on either SCL/SDA line.

"Please keep us posted regarding your investigation about VIH and VIL."

-Will do, I don't think the issue with my access to the UNIX servers will be resolved very quickly though. I have placed an order for the A version of this device to see if I can put it on a board and play with it in the lab. Hopefully this will give me an better Idea of what is going on during the "Leakage Current" event and ViL/ViH.

-Bobby

HI Bobby,

Regarding "increasing" or "decreasing" leakage current under such scenario, we're still not able to judge which case it shall be based on our chip design, right?  Customer needs to measure this on their board at this moment, correct?

Thanks!

Antony

Hey Antony,

Can you provide the pull up resistor values for chips #1 and #2?

"Customer needs to measure this on their board at this moment, correct?"
That would be the easiest way as they have the problem units. I ordered some parts and have an EVM to test with, I plan on seeing if I can replicate the issue and watch the leakage current.

Thanks,
-Bobby

Hey Anthony,

Has this issue been resolved or do you still need more information/support?

-Bobby