Current mirror question: hybrid mirror (PNP/NPN)?

[Analog Kid]

Probly over my head here, but here goes:

Let's say a guy would want a current mirror that would reflect one current used as a source onto another current used as a sink: would this work?



I ask this somewhat naively; my guess would be that the V_be levels might be off because of the reverse polarity of the Qs, but I really have no idea. Again, the idea is that the transistor on the left (PNP) would source current to something, while its partner on the right (NPN) would sink the identical current.

[moffy]

You have effectively two diodes in series with a power supply across them, not good.

[Analog Kid]

Yikes. OK, scratch that idea.
So how would a guy accomplish this?

[Someone]

So how would a guy accomplish this?

Put the current source and current sink in series.....

[Manul]

So how would a guy accomplish this?

Why not use two current mirrors, one NPN and one PNP? One of them can set the current for another one. So it will translate current sink into current source or vice versa.

[tggzzz]

Let's say a guy would want a current mirror that would reflect one current used as a source onto another current used as a sink: would this work?

This kind of question can be quickly answered by a Spice simulation. Draw your schematic, select components from library, select type of simulation, run simulation.

Key points about any simulations: results depend on quality of schematic and component models, single numerical answers rather than general analytical results.

There are many versions of Spice available, including LTSpice, MicroCap, QSpice, TINA.

[magic]

SPCIE will have no problem with this circuit...
The only question is whether it will be kA or MA flowing through those transistors. Any bets?

[TimFox]

When sketching a simple circuit idea such as this, start by adding the expected voltages at each node and currents through each branch, with simplified assumptions such as V_be = 0.7 V and beta = 100.
If it seems reasonable, then try Spice to get a better answer.

[TimFox]

So how would a guy accomplish this?

Why not use two current mirrors, one NPN and one PNP? One of them can set the current for another one. So it will translate current sink into current source or vice versa.

This is a practical approach:  note that the direction of current flow in a NPN/NPN or PNP/PNP current mirror.

[Analog Kid]

Turns out the problem I'm dealing with here is a topological one, meaning the problem of how to tie in a current mirror to a certain connection point in a circuit.

So: here are the two forms of the classic current mirror using 2 BJTs:



The first one, the PNP mirror, sources current to circuitry below the mirror; the other one (NPN) sinks current from above.

My problem: I need to source the I_set current to a component, but then have the I_mirror current sunk from another part of the circuit (hence my "hybrid" in the topic).

According to one suggestion here I could couple two current mirrors together, one as source, the other as sink. I'm thinking (but aren't sure) that this arrangement would do that:



(I'm assuming that the current of the 2nd mirror would be set by the current sourced by the first one, so no set resistor would be required.)

Now all these schemes so far are using the current mirror as a 3-terminal device, with either a common V+ (PNP/source) or ground (NPN/sink) terminal.



But what if we could construct a current mirror as a 4-terminal device, so that one side could be used as a source and the other as a sink (or vice versa):



Is this possible?
I don't really like the 2 cascaded mirror plan, as it would seem to introduce even more error into the equation. But if this is the only way to accomplish this, then I guess so be it.

One thing: discretes only. No ICs!

[tggzzz]

You need ICs, since you need closely matched devices.

You can also use MOSFETs, e.g. see the applications in https://aldinc.com/pdf/ALD1106.pdf

Advanced Linear Devices have some interesting devices http://www.aldinc.com/ald_mosfetarrays.php

You mention "I don't really like the 2 cascaded mirror plan, as it would seem to introduce even more error into the equation.", but don't quantify the allowable errors. It might be relevant for you to specify the currents, voltages, and required tolerance.

[Analog Kid]

First things first. I'm trying to get an answer to the topological question here, namely the "4-terminal" idea. If you have anything to say about that it would be appreciated.

Accuracy is a few items down the list here.

[tggzzz]

First things first. I'm trying to get an answer to the topological question here, namely the "4-terminal" idea. If you have anything to say about that it would be appreciated.

Accuracy is a few items down the list here.

You mentioned accuracy.

Try simulating your circuit with a Spice. You will quickly get some answers to your questions, and will learn a lot.

[Analog Kid]

[Try simulating your circuit with a Spice. You will quickly get some answers to your questions, and will learn a lot.

I'd try simulating it (with LTSpice, which I've used), but I can't simulate what I have no idea how to construct, hence my questions here, hoping those who know a lot more than I do might be helpful. I thought that was the idea of forums like this one. Maybe I'm wrong about that.

[tggzzz]

[Try simulating your circuit with a Spice. You will quickly get some answers to your questions, and will learn a lot.

I'd try simulating it (with LTSpice, which I've used), but I can't simulate what I have no idea how to construct, hence my questions here, hoping those who know a lot more than I do might be helpful. I thought that was the idea of forums like this one. Maybe I'm wrong about that.

You have schematics that can be entered into a Spice and simulated.

You can even use ideal components: Spice has voltage-controlled-current-sources and current-controlled-current-source components as used in some of the pictures you supply.

[Analog Kid]

You have schematics that can be entered into a Spice and simulated.

Not for my notional 4-terminal device I don't. Or haven't you been reading what I wrote?

Sheesh.

[tggzzz]

You have schematics that can be entered into a Spice and simulated.

Not for my notional 4-terminal device I don't. Or haven't you been reading what I wrote?

Sheesh.

Spice contains all the components for this picture of yours:

[magic]

Somebody already gave you the solution: connect the reference current sink in series between the load and ground.

And if you can't do that because the sum of voltage drops across the generator and the load exceeds your supply voltage, then how on earth could any floating transistor circuit work between them?

[Alex Nikitin]

Like that? Only as an illustration of a topology  . R2 can be replaced by an active device to control both currents at the same time.

Cheers

Alex

[Analog Kid]

You have schematics that can be entered into a Spice and simulated.

Not for my notional 4-terminal device I don't. Or haven't you been reading what I wrote?

Sheesh.

Spice contains all the components for this picture of yours:

No, it doesn't: what about the link between the two sides, whatever components are needed to make the right-side current track the left side?

Look: if you're really trying to help here, then why don't you just give me a circuit--you know, with components, a schematic, doesn't matter if it's drawn on a napkin--that will perform this function, as I'm sure you're capable of, OK? Otherwise, maybe just keep silent and let others help here.

[tggzzz]

You have schematics that can be entered into a Spice and simulated.

Not for my notional 4-terminal device I don't. Or haven't you been reading what I wrote?

Sheesh.

Spice contains all the components for this picture of yours:

No, it doesn't: what about the link between the two sides, whatever components are needed to make the right-side current track the left side?

It is your circuit; you tell us!

It sounds like you are thinking of "current controlled current sources", in which case see https://www.eecg.utoronto.ca/~kphang/teaching/spice/part2.html#6.8

[Analog Kid]

Yes, I think it's fair to say I'm looking for just that, a current-controlled current source.

Any chance I can get a schematic rather than a bunch of Spice parameters? That would be much more useful to me at this point.

[tggzzz]

Yes, I think it's fair to say I'm looking for just that, a current-controlled current source.

Any chance I can get a schematic rather than a bunch of Spice parameters? That would be much more useful to me at this point.

Spice source code is an ASCII netlist, originally on punched cards. Draw a Spice schematic. You will see that it is converted into an ASCII netlist; originally the conversion was done by hand. A schematic is merely syntactic sugar.

Try LTSpice, and all will become clear.

It isn't difficult, but it will be a valuable learning experience for you. Yes, that requires you to do some work, but you will gain from that.

[Analog Kid]

Somehow I knew you'd come back with that.
What's with your insistence that I use Spice to model this?
Like I said before: you're not my Electronics 203 professor.
Thanks but no thanks; I don't roll that way (entering code into Spice).

[tggzzz]

Somehow I knew you'd come back with that.
What's with your insistence that I use Spice to model this?
Like I said before: you're not my Electronics 203 professor.
Thanks but no thanks; I don't roll that way (entering code into Spice).

Why? "Give a man a fish and you have fed him for a day, but teach a man to fish and you have fed him for his lifetime" - old proverb.

"You can lead a horse to water, but you can't make it drink" - another old proverb.