I2C (smbus) level transceiver problem

[Gyosun]

Hello,

I am using the AR8035 phy connected to the LS1046A custom board.

I am using PCA9517 for level conversion between cpu(1.8v) and phy(3.3v).

After applying the power and after a certain period of time, ifconfig fm1-mac3 down/up, there was a phenomenon that the up did not work, and when I checked it, I found that there is a problem with the level converter between the cpu and the phy.

There was no problem after testing after replacing PCA9617 as a pin compatible replacement.

When I take the waveform, there seems to be no difference as follows, but PCA9517 is fail and PCA9617 is pass, but I cannot find a clear cause, so I ask a question.

Can I use PCA9517, which was reviewed as a replacement from a long-term perspective,
If so, I would like to know the answer for what reason was solved.

Please understand that the translation is not smooth.

Thank you,

Best Regards,

Gyosun.

[ufedor]

Provided traces are not clear enough.

Please provide them in full resolution.

[Gyosun]

[ufedor]

PCA9617 has significantly higher operation frequency - refer to the presentation:

https://www.nxp.com/video/nxp-semiconductors-i2c-pca9517-to-pca9617:NXP-I2C-PCA9517-TO-PCA9617

It is required to lower MDC frequency to 400 kHz for PCA9517.

[Gyosun]

Hi ufedor,

Thank you for quick response.

When using 9517, I had symptoms even if I lowered the A/B side pull-up resistance to 1.5k and the MDC clock to 245khz.

Can't I use 9517 for smbus? Which part is the problem in terms of the waveform above?

The decision to replace is being delayed because the cause is unknown.

Any suggestion of a clear solution would be appreciated.

Thank you.

Best Regards,

Gyosun.

[ufedor]

Please point on provided traces which exactly problem happens with PCA9517.

[Gyosun]

Even if I lower the MDC to 245khz, problems arise.

AR8035 low input is 0.8v max and PCA9517 vol max is 0.6v.

It comes out up to 0.8v. Could the VOL level be an issue?

[ufedor]

Please provide the PCA9517 connection schematics.

[Gyosun]

I attach a schematic diagram.

[ufedor]

Use 3kOhm resistors for R169 and R170.

[Gyosun]

I changed it to 3kOhm, but nothing improved.

We will send you the measured waveform and test details.

Is vol a problem? Is there any way to lower it with a combination of resistance?

MDC is also set to 400khz or less so there should be no problems, but I don't know why?

[ufedor]

Please disconnect EMI1_MDIO_3V3 from AR8035 and capture a trace.

[Gyosun]

The MDIO was separated and captured from the AR8035.

[ufedor]

Where exactly on the schematics the signal was probed?

[Gyosun]

EMI1_MDIO_3V3 was separated from AR8035 and measured in 3kOhm pull-up state.

[ufedor]

1) In the PCA9517 datasheet pull-up resistors are 10kOhm.

2) Have you compared traces at the AR8035 for proper and failing cases?

[Gyosun]

1) In the PCA9517 datasheet pull-up resistors are 10kOhm.

I've tried it with 10kOhm, but it's the same symptom.

2) Have you compared traces at the AR8035 for proper and failing cases?

As a substitute, PCA9306 of other company was assembled and tested, and it worked normally.

However, since the board has already been built, we want the PCA9517 to work.

[ufedor]

Have you tried to replace the PCA9517 by another PCA9517?

[Gyosun]

I tried to replace it many times.

[ufedor]

From the provided traces it is not clear where the problem moment is - i.e. data is identical.

[Gyosun]

Found a suspected phenomenon.

If you look at the waveform of the MDC clock when it is abnormal as follows, it is not clean.

At this time, it seems to have an influence on the data received value by recognizing the clock.

If you have any opinions and countermeasures, please reply.

Thank you.

Best Regards,

Gyosun.