TCA9539: Anti-interference of the I2C

user4057332 said:

Part Number: TCA9539

Anti-interference of the I2C

The I2C  of TCA9539PWR has better anti-interference performance than PCA9539 when  there is Low-frequency（eg:50KHz） interference in the environment.

[Bobby] Can you provide how your test procedure is done? 50kHz interference does not provide much information

Problem:

1、Why is the  i2c of TCA9539PWR  more resistant to interference than PCA9539?

[Bobby] From a design perspective, they are both completely different devices due to a difference in process technology.

2、Difference between i2c input filter of TCA9539PWR and PCA9539.

[Bobby] The input deglitch filter should still be similar in a schematic sense. The level of doping on the silicon and likely the layout may have changed however.

Comments above in BOLD.

-Bobby

User environment:

Modulated low frequency signal (eg 50KHz) in RF signal

Noise : 50KHz , As shown below：

1、Why is so much difference between TCA9539 and PCA9539?

2、If PCA9539 is used, how does the peripheral circuit filter out noise?

3、When new design, how to choose the right model when working in a strong interference environment.

Hey User,

Your second scopeshot with TCA9539's SDA does not make sense.

If you are injecting a frequency onto the SDA line, you should either see:

1) the injected signal (if it is between Vcc and GND)

2) if you overvoltage (or generate very fast edge rates)  but don't necessarily kill the device immediately, the signal clamp high (engaing the ESD protection). I suspect this would kick in at around 7V.

3) confuse the state machine, latching the line low

4) undervoltage the device but don't necessarily kill the device immediately, the signal will clamp to about -0.7V

As you can see, none of the events here are occurring.

Your trigger in your o-scope is sitting ch2 and at negative 2.20V, This is outside of the actually the event boundaries so you may be missing what is actually happening. I would suggest redoing your test on the TCA9539 with the trigger set to be within the event and with the scope also set to normal mode. I suspect your second scopeshot may not be capturing what is actually happening.

"2、If PCA9539 is used, how does the peripheral circuit filter out noise?"

I2C standard only states slaves to have a 50ns deglitch filter built internally (the filter does not actually sit on the SDA/SCL pins but rather buffer the input into their buffer). It shouldn't influence/attenuate what is happening on the I2C pins.

"3、When new design, how to choose the right model when working in a strong interference environment."

I2C devices operate at low frequencies. You can't just filter the I2C lines without repercussions to the actual signal. I2C specs max cap load is 400pF on a line. This limits your cap. I2C is an open drain architecture meaning we can't drive highs fast. Series resistances on the line therefore cause huge VoLs inorder to get the desired filtering even if you were below 400pF of cap.

To put it simply, I2C is not a great protocol for applications where low frequency filtering is required. You may want to consider a unidirectional communication protocol which utilizes a push pull architecture and sits decades (single filters roll off at a rate of -20dB per decade) away from the fundamental frequency you are trying to filter out.

-Bobby