TCA9803: B-side Pull-up resister

Hey Ishiwata,

"Also, I would like to attach a pull-up resistor to the B side of TCA9803. This is customer matter.
The board has already been completed and pull-up resistors have to be connected.
Is a pull-up resistor connectable to the A side?"

Pull up resistors on B side of this device will not work well. The 1.2k pull up resistors in general are a bit strong so I don't think this would work.

"Is it possible to multi-drop connect B side of TCA9803?"

In your picture, this is okay. I must point out that the first picture has A side on the master but the new picture is B side on the master. Does this mean this is a different I2C bus or did you draw something wrong?

"Or is it possible to multi-drop connect the TL9600 T / R side?"

Yes, T/R side is okay but S side is NOT okay.

Thanks,

-Bobby

Hey Ishiwata,

Sorry for the late reply. I was out of office when you sent the earlier reply and just got back in today.

The TCA9803 does not have a normal/conventional static voltage offset. It uses a current source like you know. In a sense, it does have a static voltage offset because the current source is constant and the Rdson of the device is constant in a steady state sense. So the 3.3mA current multiplied by the Rdson of B side NFET of TCA9803 will generate a static voltage offset but this offset is much lower than most static voltage offset buffers because it does not use a diode.

Now, what needs to be understood is how the device detects which side is pulling low (this is necessary for the device to insure it does not latch itself low). The device can tell if B side is driving low because it can see if the internal current source is being sunk externally. If the B side is sourcing current then it means B side master/slave is driving low. When the current source is being sunk internally, it means A side is driving low. In a contention state where A side is driving low and B side master/slave wants to also drive low (example would be an ACK hand off) the master/slave needs to be able to pull enough current away from the TCA9803's NFET to recognize that B side is driving low. This means the master/slave needs to have enough drive strength to accomplish this (the parameter I am referring to is IiLc).

"The drive capacity of the master is 3 mA. In this case, is transmission possible with a 1.2kohm pull-up?"
1.2k is alot of current. You are adding an additional 2.75mA of current into the master/slave side plus the TCA9803 current source. This effectively means the master/slave now needs to drive 6.05mA of current instead of just the 3.3mA from the TCA9803 during a low condition. TCA9803 requires atleast 1mA of the current source to be pulled in order for it to recognize that the B side is driving low during contention. TCA9803 specs that the master/slave must provide a drive strength of 150 ohm or less to ensure lows can be passed during contention. A standard I2C slave/master will provide 0.4V at 3mA which is 133 ohm. By adding the 1.2k external resistor you may increase the Ron of the master/slave such that the drive strength is no longer less than 150 ohms. Because you spec'd the 'drive capacity of the master' at 3mA I assume that the drive strength is only standard and not especially strong like some FPGAs. This leads me to think you will violate the RiLc requirement.

"There is another matter, can I input only SDA to TCA9803?
The customer says that it wants to transmit SCL by another path.
If possible, are the unused pins (SCLA, SCLB) connected to VCC? Is it GND connection?"
This still seems odd to me but yes, you can make SCL high (externall pull ups on A side and leave B side floating)

Thanks,
-Bobby

Hey Ishiwata-san,

"Low can not be driven correctly if sufficient current is not supplied to the 9803 NFET."
Should read: Low can not be driven correctly if sufficient current is not sourced from the 9803 current source during contention.

"Therefore, it does not operate correctly if there is a current of 3 mA in the B-side counter device."
Should read: Therefore, it MAY not operate correctly if there is an EXTERNAL current source (like 1.2k pull up) on the B-side of 9803 device.

"Therefore, transmission with 1.2kΩ pull-up will not be possible."
A side 1.2k pull up should be okay, B side with 1.2k pull up may not be. It is dependent on Master/slave drive strength.

" And you are judging."
Correct, I heard when this device was designed that they did do testing to see if B side could handle external pull up resistors and for weak pull ups, the device worked. [I did not find data for this testing] For strong pull ups (like 1.2k) they started to see problems. I personally have not done testing but my thoughts are the issues are due to the drive strength of the master/slave not being able to sufficiently draw the IiLc from 980x.

Thanks,
-Bobby