switchable 12V pull-up resistor

[jwmelvin]

I'm using Teensy 3.1 in an automotive interface that includes a connection to the OBD2 k-line. That specification requires a terminating 510 Ohm pull-up resistor from the k-line to +12V. If there is an external device connected to the bus, it will have a terminating resistor; when there is no such device, I'd like the Teensy to be able to enable the pull-up.

From what I gather, this requires some fashion of high-side switching, since the k-line idles at +12V. It seems I could accomplish this with a NPN transistor feeding a PNP transistor, or through MOSFETs with the same arrangement. There are of course details that I'm trying to figure out. Do you have any advice about a good way to accomplish this? It seems like there are some high-side driver ICs that might do what I want, although the current isn't very high for just a pull-up, so they are overkill. I'd like to keep it compact as possible (surface mount is good) so it will fit on my board. (I could do it on a daughterboard, which may be the most cost-effective way now that my main board is already a few revisions in.)

Thanks,
Jason

[jwmelvin]

So maybe someone is willing to comment on this schematic? (Adapted from post 11 of this thread.)

It uses an NPN transistor to drive the P-channel MOSFET. Since the car operates at ~14.4V, the resistor dividers would mean the gate is about 7V when the transistor conducts, if the K line were at ground. But what about when K idles (nothing pulls it to ground)? In that case, the idea is that the pullup resistor (R1, the one I'm trying to selectively switch into the circuit) pulls the K line up. Won't that floating K line prevent the MOSFET from working correctly? I know I'm missing something obvious due to my ignorance. Does the MOSFET care only about the gate-source voltage, making the drain irrelevant?

[PaulStoffregen]

This should work fine, assuming you connect something that is a load (consumes power, never sources power) between K and GND.

The TP0610K is specified at 10 volts on its gate. You should probably reduce R3 to something well under 800 ohms, or just replace it with a wire.

[jwmelvin]

Thank you Paul. I really appreciate your help (and great products). I had thought that reducing R3 may be in order, just wanted to stay well clear of the 20V max voltage differential. I will go with your recommendation.

[PaulStoffregen]

Another way to do it with only bipolar transistors....

[jwmelvin]

Thanks. Do I not need a series resistor between the Teensy and the NPN?

[Jp3141]

No -- the 10k controls the current in the NPN quite well (but 10k isn't ideal...).

Basically, the Teensy pulls the NPN's base to 3.3 V. Therefore the emitter is at 3.3-0.7 = 2.6 V, and so 2.6/10k = 260 uA flows to GND. This also flows in the NPN's collector and to the PNP's base which amplifies it by the PNP's gain. Say this is 50 -- then the PNP will conduct 50*260u=13 mA.

However the 510 ohm needs 14.4/510 = 28 mA. This means that actually the PNP can't pull the 510 fully to supply.

In circuits like this it is good practice to consider 'worst case conditions' - the resistors have 5 % variation, the PNP beta may not be that good, the 14.4 V could rise to 16 (or higher). Working backwards, you need:

16/510*1.05 = 33 mA in the PNP collector
this means 33/50 = 0.7 mA in the base.
This means the NPN emitter R is (3.3-0.7)/0.7/1.05 = 3.5k
-- so choose a 3.3k resistor.

It might also be better to have a pull-up R on the PNP (between base and emitter) -- 10..100k would be OK there and not affect much else.

Also, the NPN dissipates nearly 500 mW - depending on your setup, a small amount of heatsinking might be useful

[jwmelvin]

Thank you that helps my understanding considerably. Is there any reason to use the PNP instead of a p-channel MOSFET?

[PaulStoffregen]

Oh, yeah, good point on the resistor value if the PNP gain is low.

The mosefet circuit should be fine, and uses less power.

[Jp3141]

A p-channel FET would need a R from its gate-source to turn it off when the NPN turns off (else the gate is 'floating' -- you can't say what votlage is on it).

You want to get good VGS on the device (say 5 V); on the other hand you don't want >> 10 VGS (many p-channel can take 20 V briefly though).

So -- going back to the original 10k; it has 2.7 V across it. If you use twice this (say 22k), then you'll get 2.7/10k*22k = 6.0 VGS on the FET which will be good. Note that at battery < ( 6.0 + 2.7), you won't get that much, but it'll still be ~OK.