I2C Level Translator

The SCL and SDA lines of I2C are open drain / open collector so as long as your 5V part has TTL levels (0.8V and lower is a zero, 2V and higher is a one), all you need is two pull-ups to 2.5V. Make the pull-ups stiff, 1K - 2.2K.

Would work -- except that the processor is 5V CMOS levels, so it wouldn't see 2.5V as high.

Hence the need for level translators in the first place.

Den mandag den 9. maj 2016 kl. 20.35.24 UTC+2 skrev Tim Wescott:

we use a tca9509 from a 1.8V master to 3.3V slaves, though I'm not sure if it works in the opposite direction

I2C is open collector. You can connect a 2.5 volt bus to the 5 volt bus through a quickswitch without interfering with the logic in any way. You might need to use a diode in series with the power supply to the quickswitch part which already drops the 5 volt supply just enough to limit the low voltage side to 3.3 volts. One more diode drop gets you pretty close to 2.5 volts.

Hi,

Or see figure 45 of the I2C Bus Specification which shows how to connect devices with different voltage levels using MOSFETs.

Here is a link to the document:

Best regards, Martin

Yep, I've done it that way also. I used BSS123 FETs.

As Philips originated and standardized both, you'd be better off reading up on the difference and limitations built into the methods than blaming the parts. No profit in it.

Your scope may tell you more than bus monitoring tools can, if there are issues in a specific app, even if the solution ends up being in firmware. Timing and impedance levels, rather than bit banging.

RL

So, have you actually used it? Did it work?

I understand the principle involved -- I want to know if anyone's done this and been happy.

Yes sir, I have successfully used the GTL2010PW (same family, 14-bits). TI second-sources these parts as SN74GTL20XX.

Oops, should have typed GTL2010 as 10-bits.

That's a quickswitch except the quickswitch is powered from the 5 volt side and they lower the voltage to the gate to limit the voltage that can be driven through to the low voltage side.

I've used a lot of these in multiple designs. Never had a problem.

Regards, Allan

Here is the same part at NXP.

Regards, Allan

the

BTW, these are (as other posters have pointed out) just a pair of mosfets with some biasing circuitry to limit the voltages on one of the ports. (Hint: the PCA9306 "EN" pin will be regulated to the lower supply voltage

• Vgs(th).)

This is perfectly fine for low speed I2C with its open drain drivers and resistive pullups. Some of the faster I2C modes have active pullups and won't work at full speed through this class of voltage level translator.

Regards, Allan

I have also used the mosfet method with no problems at 30kHz.

John

Very much like the quickswitch devices. I'm pretty sure the quickswitch devices I've used have a diode drop from the higher of the references to drive the gates, so the low voltage is fixed. I wonder how they set the gate voltage in these devices since that seems to be controlled by the Vref on the low side. The gate voltage should be a bit higher than the Vref to allow the high driver to drive the low side to the full Vdd without driving it too high.

I guess they use a third mosfet, with G shorted to D, to measure the threshold voltage. The 220k resistor (to the higher rail) determines the drain current. The I2C pass fets have their gate volages set to the lower voltage rail + the threshold voltage.

Allan

Makes sense, thanks.