I am using the well-known I²C level shifter from that appnote from Philips.


The 3V side is supplied by an LP2950-3.0 regulator and is loaded only moderate (a few quite lazy 74's).

The problem I'm obserivng is that the 3V side is being pulled to 5V.

I loaded the 3V side with a resistor and calculated that the level shifter presents a resistance of only 2.7 kOhms. I consider this rather low and I wonder:

  1. Is that normal behaviour or did I make some mistake?

  2. Isn't this harmful for the devices on the 3V side? There are several ICs that according to the datasheet are explicitly not 5V tolerant, not to forget the regulator.


I've just checked quickly, but your diagram looks good to me. I suspect it might be some other problem, maybe assembly of the parts or something along that line. Could you please check the following for both the SCL and the SDA line, using an ordinary Multimeter:

  • Are you able to measure one diode drop (approx. 0.5-0.75V) across the parasitic diode in the MOSFETs from the 3V side to the 5V side?

  • Do you get a reading of nearly infinite resistance (or at least some kOhms) when connecting the Multimeter's test leads the other way round, i.e. from SCL's or SDA's 5V end to their 3V ends?

Please note that these single MOSFET transistors are very (!) sensitive to ESD discharge, so even if you assembled the right parts the right way, they may have some hidden type of damage.

  • \$\begingroup\$ I am measuring a voltage drop of about 0.4V between the two sides. The resistance from 5V to 3V is about 2.7 kOhms. I know that these tiny things are really sensitive and break often (which brings up the question of an alternative), but I have three of those level shifters on a manufactured (i.e. not self-etched) board and already switched the MOSFETs with new ones and they behave all the same. \$\endgroup\$ – AndreKR Dec 7 '10 at 23:07

OMG, epic fail. I just found out that I grabbed from the wrong box and placed all BC857 where there should be BSN20.

However, it could be that I was right (and the BSN20 wrong) in the first place and made the mistake not before I replaced them. I don't know, because I am out of BSN20 and put in BSH108 (which have better specs anyway). Now it works fine.

  • \$\begingroup\$ Feel free to tick your own answer as the accepted one; you don't get points for it but it's probably the most helpful; remind everyone to make sure you are using the right parts :P \$\endgroup\$ – Nick T Jan 25 '11 at 2:48

The gate threshold of these parts can be as low as 0.4V at 25C according to the NXP datasheet.

I would guess that loading on the +3V side of the MOSFETs is creating enough Vgs to turn the FETs on in the linear region. What is the measured gate-source voltage? If it's too high you'll have to adjust your pull-ups or choose a MOSFET with a higher Vgs threshold.

  • \$\begingroup\$ It's 30-40 mV for the level shifter in the above schematics. I have another one on the board missing the 3V pullups (they are in an external module not yet plugged in), there the Vgs is about 600 mV. \$\endgroup\$ – AndreKR Dec 7 '10 at 23:22
  • 1
    \$\begingroup\$ What happens if you remove the two MOSFETs altogether? It sounds stupid, yes, but it's a sure-fire way to prove that the 'leak' is through the MOSFETs somehow. \$\endgroup\$ – Adam Lawrence Dec 8 '10 at 23:29
  • \$\begingroup\$ When I remove all the MOSFETs and the regulator the two sides are complitely isolated (applying voltage on either of them has no effect on the other side). \$\endgroup\$ – AndreKR Dec 11 '10 at 20:44
  • \$\begingroup\$ Now, you could put in diodes (*4148, for example), in the direction of the MOSFETs' inverse diodes (i.e Anode=Source, Cathode=Drain) and see what happens. \$\endgroup\$ – zebonaut Dec 14 '10 at 7:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.