Hi. I need to select the highest of two reference voltages. The current needed is very low, less than 20 μA, but the voltage needs to be as precise as possible, and the references are very near the positive supply, which is 35v. There's also a negative supply there of the same value that can be used, as well as +6v and -6v.
I had thinked about a comparator driving a p-channel to connect one channel, and the other connected with just a 1k resistor, but it has it's own share of problems, like that annoying built-in mosfet diode (is there any mosfet without that?), and the max possible Vgs. Something like this (the schematic is incomplete, U1 is only there to show that the current needed is very low, and i didn't shown the comparator that drives the mosfet):

Maybe someone already did something like this? I accept suggestions, doesn't need to follow the schematic up there at all.
Thanks.

 

MOSFETs without body diodes do exist. You generally only get them in power devices.

How about using two comparators? One senses V1 > V2 and the other V2 > V1. These control a switch of some kind.

 

Here is a circuit that I think will work.

The comparator (op amp) is rail-to-rail with 44V max supply voltage. So it should work well. It also has very low offset voltage for good accuracy of selecting references.

I put a pair of P channel MOSFETs for each reference input to switch it. Then I drove the gates of those using a voltage divider, to protect against too high Vgs. You must choose the MOSFETs to have at least 20V Vgs but preferable to have higher like 40V. You do not need good Rds(on) here, even 10 or 100 Ohms is alright. The off leakage must be low, however.

So, the output of the comparator drives one of the switches. I also made an inverter to drive the other gate. This way, it selects only the higher of the two inputs.

Then finally, I put a small low-pass filter for the output.

For the simuation, I simply made the two references switch back and forth for which one is higher, and you can see the output is following the higher of the two inputs.

I put this circuit in public domain.

 

Thanks for the help.

I just remembered the "ideal diode" circuit, so i thinked about putting two of those with their outputs connected together, and simulations look good! (i attached it).

I think that with a dual opamp that can go to the positive supply it would be quite painless to implement. The diode will keep it from reaching exactly Vcc, but i think i have that small headroom if i use schottky diodes.

 

Wow, it is a good thought! It is certainly much simpler than mine, as long as you have that headroom. I think a Schottky will only set you down 0.2V or so. Good one!

 

Thanks.
About the MOSFETs, i didn't really used any yet, but i readed something about them. I have a few questions: Are there any known low-current P channel MOSFETs without that pesky diode? Can D and S be swapped, or does D always need to be negative vs S in a P-channel MOSFET?

 

Thanks.
About the MOSFETs, i didn't really used any yet, but i readed something about them. I have a few questions: Are there any known low-current P channel MOSFETs without that pesky diode? Can D and S be swapped, or does D always need to be negative vs S in a P-channel MOSFET?

This question comes up a lot. There are some MOSFETs with a separate lead for the body, which allows you to tie the body to a low (or high for a P-channel) voltage in your circuit. They are very rare, and they usually are for very small signal amplification. You will also see some JFETs where it is really symmetrical, the data sheet will even say "drain and source are identical and can be interchanged".

But for some reason, the normal construction of any MOSFET that can handle any reasonable amount of power always has that body diode, and body is always tied to source. I am always interested to learn more about why this is, too.

 

D and S can be swapped in a lateral MOSFET, which doesn't have a body diode.

The reason for the body diode is that power MOSFETs are vertical. Their structure is really a large number of MOSFETs in parallel. That type of construction demands (?) a vertical structure rather a lateral one, which forces you to have a PN junction between source and drain.

http://ecee.colorado.edu/~bart/book/book/chapter7/ch7_8.htm

That lateral power version exists too... it would be harder to parallel though like a HEXFET, since both source and drain are on the same surface.

 

Excellent info, thanks.
I thinked that only power mosfets had that diode, but i searched some p-channel mosfets in nxp.com site, and even low current mosfets seem to have it. For example, this mosfet can handle 1/4w only, and still is a vertical type, with the body diode: http://www.nxp.com/documents/data_sheet/BSS84.pdf

Anyways, i'll not be using them this time, but always was curious about them.

 

Well, it is a tiny package. The thermal resistance is huge, 500 K/W. A TO-220 is almost 10 times better, which directly influences how much power you can dissipate and current you can handle.

You also get a larger breakdown voltage for the given characteristics. It just makes sense to use vertical structures for anything related to power I guess. The body diode isn't a huge problem most of the time.