PCA9517: A side is not low enough while B side is low level

Mike Hsu

Intellectual 300 points

Part Number: PCA9517

According to the datasheet, B side has offset of VOL about 0.51V and A side VOL is about 0.1V.

As you see waveforms below, B side SDA keeps low (0.51V) but A side did not drop to VOL occasionally. Why?

CH1 A side SDA

CH2 B side SDA

CH3 A side SCK

Schematic

over 2 years ago

0 Clemens Ladisch over 2 years ago

Guru 316860 points

In that situation, the A side is driven only by the I²C device connected to PRI_SMB_xxx. Apparently, it has weak output drivers. Try changing R28/R29 to 4.7 kΩ.

0 Tyler Townsend over 2 years ago

TI__Mastermind 47668 points

Hi Mike,

Please follow Clemens' suggestion. Changing pullup resistor from 2kohm to 4.7kohm will allow for a weaker pull-up to VCC, helping an output driver on the I2C drive LOW to GND. Please let us know if changing the pull-up fixes the issue.

Regards,

Tyler

0 Mike Hsu over 2 years ago in reply to Tyler Townsend

Intellectual 300 points

Hi Tylor

1. R28/29 changed to 10K or 4.7K ohm, I2C still drive not low enough like previous waveform, any suggestion?

2. R28/29 changed to 10K ohm. Is it ok for bit 9 not drive low below 0.4V according to I2C specificatoin ?

0 Mike Hsu over 2 years ago in reply to Tyler Townsend

Intellectual 300 points

Hi Tylor

I have found another issue (see waveform below). Why is it happening?

Is this the load susceptibiliy errata metioned in the datasheet?

CH1 A side SDA

CH2 B side SDA

0 Clemens Ladisch over 2 years ago in reply to Mike Hsu

Guru 316860 points

As I said above, this happens when the A side is driven only by the I²C device connected to PRI_SMB_xxx. The problem is with that device, not with the PCA9517. Please tell us which device it is.

A short spike on SDA is no problem; its level is sampled only when SCL goes high.

0 Mike Hsu over 2 years ago in reply to Clemens Ladisch

Intellectual 300 points

A side is connected to another SoC

B side is also connected a server host

0 Clemens Ladisch over 2 years ago in reply to Mike Hsu

Guru 316860 points

Can that "another SoC" be configured for more drive strength?

0 Mike Hsu over 2 years ago in reply to Clemens Ladisch

Intellectual 300 points

Do you mean like increase pull up resistors for R308 and R257 of U145?

We are using two PCA9517

0 Clemens Ladisch over 2 years ago in reply to Mike Hsu

Guru 316860 points

No; I mean to change the configuration of the two I²C pins of the SoC_I2C. (Assuming that that device uses GPIO pins for I²C, and that those can be configured.)

In any case, that voltage is not a problem.

0 Mike Hsu over 2 years ago in reply to Clemens Ladisch

Intellectual 300 points

The bit 9 is ACK sent from the SoC_I2C to the CEM host.

I do not think I can change the drive strength of SoC_I2C.

How do I get A side data voltage level lower ?

0 Clemens Ladisch over 2 years ago in reply to Mike Hsu

Guru 316860 points

The current flows from P1V8, through R308/R257, and into the I²C pins of the SoC_I2C. THe PCA9517A input is high impedance and cannot affect this; your only choices are to change the resistor or the SoC.

0 Tyler Townsend over 2 years ago in reply to Mike Hsu

TI__Mastermind 47668 points

Hi Mike,

Have you had a chance to implement Clemens' suggestions about resistors R308/R257?

I would also like to mention some thoughts on U24, the I2C mux. I am assuming that U24 is not a true buffer, therefore the degree of separation between the B-side of U145 and the A-side of U12 is likely some small series resistance when U24 is passing I2C signals. If this is the case, please note that resistors R28 + R597 and R29 + R598 will actually appear in parallel pulling up to P3V3.

In the case that you are using 2kohm pull-up resistors on both sides of U24, this comes out to an equivalent pull-up resistance of 1kohm. Even if you replace one of the pull-up resistors with a 10kohm resistor, 10kohm || 2kohm = 1.667kohm which is a fairly strong pull-up resistance to P3V3.

My suggestion is that you try replacing resistors on both side of U24 with 10kohm pull-ups. 10k || 10k = 5k, which will bring the pull-up strength down allowing for the I2C drivers on the bus to be able to hopefully pull the bus to a LOW state.

Regards,

Tyler

Interface

Interface forum

PCA9517: A side is not low enough while B side is low level