Current sources vs sinks

[Analog Kid]

We're taught, from an early age, that this is the proper configuration for a current source:



and that this is the proper setup for a current sink:



But what if instead we use this as a sink?



and this as a source?



Question 1: Will these alternate configurations work?
Question 2: Will they work, but with some deficiency?

I'm really not trying to stir up a hornet's nest here: I'm just trying to wrap my head around this basic electronic concept.
It seems to me that really, either way should work, because if we think of the transistor as a device that passes current through from collector to emitter (conventional current flow direction, not electron flow), similar to a valve with some fluid flowing through it, why should it matter whether we tap the current at the top or at the bottom?

Now there is the matter of base current, which flows out of the emitter of the NPN and therefore adds to that current.
However, I_B is so small compared to I_E that it's basically "swamped" and can therefore be ignored, right?

This is basically the same as the high-side switch vs. low-side switch thing.
Just to be clear, I am excluding applications like precision current sources here; I'm proposing a typical current source that's just accurate enough.

[TimFox]

The distinction between source and sink:
A source supplies current to an external load from its internal power source.
A sink is a load for an external source.

Example: the traditional 7400 TTL output uses an NPN collector as a sink for current sourced by the next gate’s input emitter (from its power supply) to pull the output voltage down sufficiently for logic-low output.  For logic high, the totem-pole pull-up is a source to supply enough current (less than the low-state current) to obtain a voltage sufficiently positive to turn off the next gate’s emitter.

The four circuits you show are all sources, of different current direction, from the indicated power source “V+”.

[Analog Kid]

The four circuits you show are all sources, of different current direction, from the indicated power source “V+”.

OK, but that didn't really answer my question, unless the answer was implicit (in which case I missed it):
Will those all work?
Will they all work equally well?

And if they're all sources, what would a current sink look like?

[pqass]

The four circuits you show are all sources, of different current direction, from the indicated power source “V+”.

OK, but that didn't really answer my question, unless the answer was implicit (in which case I missed it):
Will those all work?
Will they all work equally well?

And if they're all sources, what would a current sink look like?

Yes the alternates will work (as a switch) BUT a base current must flow to turn on the transistor (WRT its emitter) and keep it on.
The only way to do that with a PNP collector on GND is to produce a voltage below ground out of the base resistor.
And with an NPN collector on V+ is to produce a voltage above V+ into the base resistor.

[PGPG]

Have you heard about emitter follower?

[TimFox]

Another distinction, specifically for current:  a good current sink exhibits a low impedance to the external current source, while a good current source has a high impedance to the external sink or load.
If you use a collector (PNP or NPN) to drive the load, that is a reasonable current source.
However, an emitter (PNP or NPN) drives the load from a relatively low impedance and is therefore better treated as a voltage source;  it also works as a current sink from an external source.
This assumes that the transistor is in its active region (not saturated).
A saturated transistor (switched hard ON) presents a low impedance at the collector.
Otherwise, calling the circuit a source or sink depends on the configuration and location of power sources, as I discussed above.

[tggzzz]

The four circuits you show are all sources, of different current direction, from the indicated power source “V+”.

OK, but that didn't really answer my question, unless the answer was implicit (in which case I missed it):
Will those all work?
Will they all work equally well?

And if they're all sources, what would a current sink look like?

That is reasonably well described in a book you have said you possess: TAoE, any edition will do.

An answer much deeper than Tim Fox has already given you will probably require considering semiconductor physics. Sorry about that.

Why don't you look at your copy of TAoE; you will then be in a position to ask a a more pointed question.

[Benta]

The distinction between source and sink:
A source supplies current to an external load from its internal power source.
A sink is a load for an external source.

Good definition.
Unfortunately, a lot of people define a "source" as current flow from V+ and "sink" as to V-. Let's call it "polarity thinking".

[Analog Kid]

I think the answer that @pqass gave is the one that's needed here.
I completely disregarded the biasing requirements for a source/sink circuit.
I believe this pretty much trumps any other discussions here, with maybe the exception of source/sink impedance:
The "alternate" circuits I posted simply won't work without outlandish biasing schemes (V below ground for PNP, V above V+ for NPN).

[TimFox]

The distinction between source and sink:
A source supplies current to an external load from its internal power source.
A sink is a load for an external source.

Good definition.
Unfortunately, a lot of people define a "source" as current flow from V+ and "sink" as to V-. Let's call it "polarity thinking".

That usage is similar to "source" and "sink" in generalized field theory, for example showing the velocity field for weather or the magnetic field from a bar magnet.
This was covered fully in the classic P M Morse and H Feshbach Methods of Theoretical Physics, McGraw-Hill 1953 (2 vols.)
I kept this book on my shelf at work:  when someone asked me a complicated question, I referred them to that book, and they didn't bother me any more.

https://www.physicsforums.com/threads/has-anyone-ever-finished-reading-morse-feshbach-and-courant-hilbert.1054844/

[Analog Kid]

Why don't you look at your copy of TAoE; you will then be in a position to ask a a more pointed question.

Again with the totally unhelpful "I'm not going to answer your question: read the book".

OK, wise guy: I have read what The Art of Electronics by Horowitz & Hill has to say on the subject.
Specifically from pp. 59-62 in my edition.
Yes, they have many good things to say about current sources.
However, they have really nothing to say about the difference betwixt current sources and sinks, which of course is the topic of this thread. All they write is this:

The other examples (using npn transistors) should properly be called current sinks, but the usual practice is to call all of them current sources.

Which of course doesn't really address my questions.

And as I wrote above, the answer given by pqass is a show-stopper that would stop my alternative examples from working without extraordinary base biasing schemes. Which I'm assuming is the real reason that the canonical forms of PNP vs NPN configurations are used.

[tggzzz]

Why don't you look at your copy of TAoE; you will then be in a position to ask a a more pointed question.

Again with the totally unhelpful "I'm not going to answer your question: read the book".

OK, wise guy: I have read what The Art of Electronics by Horowitz & Hill has to say on the subject.
Specifically from pp. 59-62 in my edition.
Yes, they have many good things to say about current sources.
However, they have really nothing to say about the difference betwixt current sources and sinks, which of course is the topic of this thread. All they write is this:

The other examples (using npn transistors) should properly be called current sinks, but the usual practice is to call all of them current sources.

Which of course doesn't really address my questions.

And as I wrote above, the answer given by pqass is a show-stopper that would stop my alternative examples from working without extraordinary base biasing schemes. Which I'm assuming is the real reason that the canonical forms of PNP vs NPN configurations are used.

You didn't bother to mention that you had read (some of) the relevant bits in TAoE.

Biassing transistors is always required, so I'm not sure why you regard that as "extraordinary".

TAoE uses the phrase "current sink" at least 70 times; I lost count. Most of those indicate/discuss biassing.

IMHO the best answer to your original question was given by Tim Fox. You were dismissive of his answers. Your subsequent responses indicate that you had asked the wrong question to get the answers you seem to like.

It is, of course, generally true that answering a question right (i.e correctly) is easier than asking the right question.

[Analog Kid]

Biassing transistors is always required, so I'm not sure why you regard that as "extraordinary".

I would regard requiring a PNP base bias voltage of below ground, or an NPN bias voltage of above V+ as "extraordinary" because those wouldn't be possible with a single supply voltage to the circuit.

[TimFox]

Single-supply circuits are now fashionable, but many circuits use split or dual supplies.
There are other ways to bias devices, even with a single supply.

[Analog Kid]

Single-supply circuits are now fashionable, but many circuits use split or dual supplies.
There are other ways to bias devices, even with a single supply.

Well, sure: I suppose one could use a voltage divider, f'rinstance. Or even a charge pump.

[pqass]

NPNs/N-FETs are just overall better than their P versions (due to physics). 
They're faster, beefier, have lower on resistance, more common, and therefore cheaper.
So it may be worthwhile to produce a bias voltage above V+ using a charge pump or DC-DC converter.
Some gate driver ICs provide a bias voltage so the same FETs/IGBTs can be used on both high and low side. 

[tggzzz]

Biassing transistors is always required, so I'm not sure why you regard that as "extraordinary".

I would regard requiring a PNP base bias voltage of below ground, or an NPN bias voltage of above V+ as "extraordinary" because those wouldn't be possible with a single supply voltage to the circuit.

You attach too much importance to a node being marked as "0V" or "ground".

For example, I have a piece of test equipment where, IIRC, some TTL outputs are high/1 if >-18V, and low/0 if <-19.6V. Others in the same equipment are >+2V and <+0.4V respectively. Need to be very careful when placing a scope probe shield on an IC pin

[Analog Kid]

Biassing transistors is always required, so I'm not sure why you regard that as "extraordinary".

I would regard requiring a PNP base bias voltage of below ground, or an NPN bias voltage of above V+ as "extraordinary" because those wouldn't be possible with a single supply voltage to the circuit.

You attach too much importance to a node being marked as "0V" or "ground".

Well, I'll admit that that's a shortcoming in my understanding of electronics.
I'm learning here ...

[TimFox]

Single-supply circuits are now fashionable, but many circuits use split or dual supplies.
There are other ways to bias devices, even with a single supply.

Well, sure: I suppose one could use a voltage divider, f'rinstance. Or even a charge pump.

A resistor can source or sink current,  depending on what you connect it to.

[Analog Kid]

OK, so would you accept this as an accurate illustration of that distinction?



It depends on whether the resistor is above or below the load, correct?

[Benta]

OK, so would you accept this as an accurate illustration of that distinction?

(Attachment Link)

It depends on whether the resistor is above or below the load, correct?

No. Your circuit is still supplying current to a load. Whether the current is positive or negative is irrelevant. It's a "source".

[Analog Kid]

OK, so would you accept this as an accurate illustration of that distinction?

(Attachment Link)

It depends on whether the resistor is above or below the load, correct?

No. Your circuit is still supplying current to a load. Whether the current is positive or negative is irrelevant. It's a "source".

OK, so can someone please show me what a sink would look like? Preferably in schematic form? (Crayon on cocktail napkin sketch perfectly acceptable.)

[SteveThackery]

No. Your circuit is still supplying current to a load. Whether the current is positive or negative is irrelevant. It's a "source".

Hey! This is new to me. So you're saying that the load has to be external?  In your post (#2) you give examples of a current source and a current sink, and you say that in both cases the load is external (ie in the next chip along).

Well, this requirement is news to me. But more importantly, it must mean that your assertion that all four of Analog Kid's examples are current sources is wrong, because none of the loads he draws are external.

So if the load is not "external" (whatever that now means), what are the circuits Analog Kid has drawn called? Obviously not current sources or current sinks.

[Analog Kid]

What they said.
What exactly does "external" mean?
I think it would be very helpful if someone could post examples showing a current source vs. a sink, and showing what an external load looks like.

[PGPG]

No. Your circuit is still supplying current to a load. Whether the current is positive or negative is irrelevant. It's a "source".

I'm not very good in English - it is my second language.
The terms sink and source I met first time in digital ICs datasheets.
How you explain that these outputs are characterized with their source and sink capability even between connecting to output a LED+R to GND or to VCC has absolutely no difference regarding the sense of supplying the current to load - both connections supply the current to LED and in both cases a power source is on the same PCB as IC and load (LED) is also there.

[tggzzz]

I suggest people re-read and think about Tim Fox's reply #5 (and hence his reply #1).

[Doctorandus_P]

A few years ago I built a self powered current source / sink.

It has:

• 1n4148 as a 600mV voltage reference.
• LM358 that buffers / amlifies / stabilizes the reference.
• Potentiometer on the voltage reference  ( Final current approximately between 10mA and 1.5A)
• Shunt resistor to GND
• LM358 (other half) to compare the reference with the shunt resistor voltage.
• Sziklay pair as current amplification driver
• Heatsink
• Some *&^%$#@! to get the thing stable. (That took me a few hours)

I know a silicon diode is not very stable / accurate, but the goal here was to make a self powered current source / sink. Therefore it was important to use low voltage parts. The thing from a voltage as low as 2.8V and up to the maximum voltage of the LM358 (30+V).

Because it's self powered, it just has two connections, and you can use it either as a current source (with an external power supply) or as a current sink, depending on where you put it in the circuit.

[TimFox]

OK, so would you accept this as an accurate illustration of that distinction?

(Attachment Link)

It depends on whether the resistor is above or below the load, correct?

No. Your circuit is still supplying current to a load. Whether the current is positive or negative is irrelevant. It's a "source".

OK, so can someone please show me what a sink would look like? Preferably in schematic form? (Crayon on cocktail napkin sketch perfectly acceptable.)

A quick sketch, cut-and-pasted from a description of the 7404 TTL hex inverter.  Note that this original form of TTL logic is an example of "current-sinking" logic, along with DTL.  Later "TTL" types, including 74LS, are usually DTL.

Ignore the redundant labels from the original figure, but concentrate on the horizontal connection from the output of the left circuit to the input of the right circuit.  This circuit normally operates in one of two states:  "V_hi" and "V_lo" for the voltage on that wire, with corresponding currents along the wire in opposite directions.  The two states are labeled with brackets for [hi] and {lo}.  I added designators for Q1, Q2, and Q3.  The two inverters could be on the same substrate or in two different packages.
I apologize for the bad image quality, but my crayon was dull.

Now, in the {lo} state (logic 0 at that node), Q2 is off and transistor Q1's collector is sinking current from transistor Q3's emitter.  Q3 is powered from the Vcc node on the right.  Since Q1 is saturated, it presents a low impedance to that current flow I_lo sourced by Q3, and the voltage V_lo is quite low.

However, in the [hi] state (logic 1 at that node), Q1 is off, and Q2 is a source pulling up the node to a higher voltage V_hi.  The current I_hi for this IC series is much lower than I_lo.  Q2 is probably better considered a voltage source, rather than a current source, driving a relatively high impedance.  Q2 is powered from the Vcc node on the left.

On the data sheet, the manufacturer guarantees certain voltage and current levels for the gate in isolation.
The suffixes below were changed to correspond to data sheet specs.
For the first gate's output supplying the second gate's input:
    V_oh > +2.4 V for I_oh < 0.4 mA
    V_ol < +0.4 V for I_ol < 16 mA
The requirements at the second gate's input for definite logic levels are
    V_ih > +2.0 V at 0.04 mA  and   V_il < +0.8 V at 1.6 mA
so there is some noise immunity at the high and low logic states.
From the low-state sink current of the first gate, and the low-state source current of the second gate, we quickly calculate that one output can certainly drive 10 inputs (10:1 fanout).

Incidentally, I have never seen the term "voltage sink", although "voltage source" is quite common.  If one needed that term, it might be a relatively high-impedance load fed from the source.
A CMOS gate output is usually a switched voltage source driving a capacitive load at the driven inputs.

[Analog Kid]

1. Let's at least be able to view that circuit of yours properly (I just needed to adjust the contrast & brightness of your image):



2. Too complicated, with too many extraneous factors; I still can't really wrap my head around what's a current source and what's a sink.
Can someone post a simpler explanation (w/circuit), please?
Pretend you're explaining this to a total n00b.

[Benta]

I'm with you, horrible example.
But what's so difficult about a current source (I) vs. a voltage source (U)?
U supplies voltage (and current depending on load)
I supplies current (and voltage depending on load)

The sink/source thing in digital ICs is a misnomer for output current capability, but unfortunately commonly used.

[TimFox]

An ideal current source delivers a constant current regardless of the voltage across the load.  (Output shunt resistance is infinite.)
An ideal voltage source delivers a constant voltage regardless of the current through the load.  (Output series resistance is zero.)
Real current and voltage sources do not achieve this ideal condition, but may be closer to one or the other ideal case.
Thévenin's Theorem for voltage sources and Norton's Theorem for current sources discuss the relationship between these two concepts.

[Analog Kid]

OK, let me try something else:

Say I give you the following kit of symbols:



Can you put these together to show us 1) a current source and 2) a current sink?
Feel free to add whatever other symbols are necessary (within reason!) for your examples.

And additionally, explain just exactly what an "external load" is?

TIA

[Andy Chee]

OK, let me try something else:

Say I give you the following kit of symbols:

(Attachment Link)

Can you put these together to show us 1) a current source and 2) a current sink?
Feel free to add whatever other symbols are necessary (within reason!) for your examples.

And additionally, explain just exactly what an "external load" is?

TIA

The load resistor symbol should probably be a variable resistor, rather than a fixed resistor.

As for "external load", think of it as the opposite of internal impedance 

[PGPG]

The sink/source thing in digital ICs is a misnomer for output current capability, but unfortunately commonly used.

In my opinion simply the same words happen to mean different things depending on context.

Save in context that the device will not kill a person means completely different thing than save in a context of cryptology algorithm.

(Ideal) Current source in circuit simulation is different thing than digital output working as current source (or sink). Saying that such output sources current doesn't mean it is current source from circuit analysis point of view.

I think I feel perfectly what 'source' means but only partially what 'sink' means but in digital output context I think of them as follows:
If you connect capacitor to the output (second pin grounded) than:
- when output sends a charge into it than this output sources charge (works as current source).
- when output takes charge from capacitor than this output sinks charge from capacitor (works as current sink).
I don't know if in English saying that something sinks something form something is acceptable use of word sink.

In my opinion using source/sink how they are used related to digital output is correct use. In this context these words simply means something different than when you use them in context of energy source and sink and in a similar sense.

[golden_labels]

Say I give you the following kit of symbols: (…) Can you put these together to show us 1) a current source and 2) a current sink?

Here:

And additionally, explain just exactly what an "external load" is?

Load that lies outside of the current source or sink. Which, for the topic discussed here, is usually⁽¹⁾ any load.

⁽¹⁾Not strictly true, because practical source implementations may need an internal load. But this is unnecessary detail here.

[Analog Kid]

First of all, thanks for playing along here. It's appreciated.

So I get the top example; pretty straightforward.
But the second one: what's going on there? The "sink" is, what? just a wire inside that box?
What exactly is the sink here?
What exactly does it mean being either inside that box or outside of it? What does the box itself represent?

[golden_labels]

First of all, thanks for playing along here. It's appreciated.

Throwing away reference points is intentional. Focus on the loops and where the energy comes from, not on the polarities or to which rail a component is connected.

But the second one: what's going on there? The "sink" is, what? just a wire inside that box?

Yes. An ideal sink is a short. See TimFox’s post.

What exactly is the sink here?

The entire blue box with word “Sink” in the upper part. As seen from the perspective of the circuit on the right.

What exactly does it mean being either inside that box or outside of it?

Inside the box is the source or the sink. Outside is the circuit, that considers this box to be its source or sink.

What does the box itself represent?

Either a source or sink, depending on what is written on it.

---edit
Perhaps I see, where the misunderstanding lies. Don’t look at what is inside a source or a sink. This is not what distinguishes them. The distinction comes from how the external circuit treats them.

The hydraulic analogy holds here well. Extremely well in fact, because both words (“source” and “sink”) come from the water meaning. Source is where the energy comes from. Sink is where waste energy is dumped to.

Observe: this is from the perspective of whoever/whatever uses the source or sink. Not from what is inside of that source or sink.

[TimFox]

In short, "source" and "sink" are distinguished by their function in the overall circuit.
The same subcircuit could be connected to function as either, depending on the requirements.

[David Hess]

Both configurations work, but not equally well.  Howland current pumps for instance usually use the emitter follower for a current output without problems.  Regulated power supplies with emitter followers have no trouble implementing current limiting.

The common emitter configuration has the advantage of inherently presenting a high impedance output, unlike the emitter follower which is the lowest output impedance configuration.  In circuit design, sometimes a common base stage is used in series with the output to further increase the output impedance.

[bson]

The first two aren't proper current sources or sinks, they are voltage followers and will only have a constant current with a resistive load.  If you try to control a current load like an LED or BJT transistor they won't work at all.  The current isn't constant, it's the voltage that's constant.  The latter two won't work at all as there is no voltage feedback from the load, on the emitter.  They can only work with a negative supply, but then you're back to the top two voltage followers.

[IanB]

We're taught, from an early age, that this is the proper configuration for a current source: ...

Surely, this opening statement is setting up a straw man argument?

Are we taught this from an early age? (I wasn't.)

Is the assertion even accurate or correct?

How did this topic get to two pages, without anyone questioning or debating the opening assertion?

I don't even understand the question. Does the terminology of "current source" or "current sink" even have sufficient context to mean anything without further qualification? For instance, a current source could be a battery, and a current sink could be a switch. Whereas a constant current source could be a combination of circuit elements that delivers a constant current to its output with varying loads. A constant current sink could be a source measure unit acting as a load (for example when discharging a battery at constant current and measuring its capacity).

When you have an undefined question, how can you answer it?

[TimFox]

One way to answer is to re-define or re-phrase the question into an answerable question.

[Analog Kid]

We're taught, from an early age, that this is the proper configuration for a current source: ...

Surely, this opening statement is setting up a straw man argument?

Are we taught this from an early age? (I wasn't.)

It was a, it was a, a joke, son.

How did this topic get to two pages, without anyone questioning or debating the opening assertion?

Because apparently I managed to unearth some unclear concepts and ambiguities, without even intending to.

[IanB]

Because apparently I managed to unearth some unclear concepts and ambiguities, without even intending to.

Perhaps I was being a little hasty, and I didn't quite understand what you were driving at. I apologize. I think your question is according to the picture below?

When transistor A is turned on (base low) it will charge up the capacitor (source current), and when transistor B is turned on (base high) it will sink current (drain the capacitor).

This configuration is different from the commonly documented "push-pull" output stage, the reasoning for which is beyond my knowledge.

It does appear to me that if transistor A is a PNP as shown and it is turned on hard, then it will charge the capacitor almost all the way to V+. And if transistor B is turned on hard, it will drain the capacitor almost to zero volts. I am not sure if A and B were swapped, it would lead to the same outcome, since the capacitor would block the base current in each case with A as NPN and B as PNP.

[Analog Kid]

That actually helps.

[watchmaker]

How did this topic get to two pages, without anyone questioning or debating the opening assertion?

Because apparently I managed to unearth some unclear concepts and ambiguities, without even intending to.

I am not sure you did.  It appears to me you conflated current source (a la nodal analysis) with constant current source/constant current sink.  In nodal analysis, all the power dissipative elements are the sinks, although no one calls them that.

Constant current sinks/sources are active circuits used in power management.

I think Ian is right, a lot of time was spent by others trying to put together the words you used into an answerable question.

[Analog Kid]

How did this topic get to two pages, without anyone questioning or debating the opening assertion?

Because apparently I managed to unearth some unclear concepts and ambiguities, without even intending to.

I am not sure you did.  It appears to me you conflated current source (a la nodal analysis) with constant current source/constant current sink.  In nodal analysis, all the power dissipative elements are the sinks, although no one calls them that.

Constant current sinks/sources are active circuits used in power management.

I think Ian is right, a lot of time was spent by others trying to put together the words you used into an answerable question.

Wellll, I didn't intend constant current sources: I believe the symbol I used in my later posts,



is the symbol for a generic current source, yes?

And you say

In nodal analysis, all the power dissipative elements are the sinks, although no one calls them that.

So they are sinks, even though nobody calls them that? See, that's just part of the confusion here, at least the confusion due to terminology if not electronic theory.

[golden_labels]

It would be a lot easier, if multiple meanings of “source” were not freely switched on each other post.

The symbol is an abstract, ideal element of a mathematical equation. Indicating that current through that symbol is fixed. Fixed not in the sense of being DC, but in the sense of independency from what happens outside of this element. “Perfectly controlled” if that sounds clearer.

It has little to do with sources as discussed when you put word “source” beside (or opposite) word “sink.” In that meaning they are abstract concepts and describe roles of elements in a system, from the perspective of this system. Note I used word “system”: these concepts both predate electricity and are widely used in other branches.⁽¹⁾

There is more than a single abstraction to describe things. Don’t confuse terms coming from different models. Even if they are used in the same context. If you were asked to fetch a glass from the fridge, you’d rather grab a cup-like container with juice instead of pulling out the entire glass pane shelf, right? Both are “glass,” but it’s obvious to you they’re two different categorizations and which one is relevant.

⁽¹⁾ Examples: graph theory (in particular transport networks or Petri nets), thermodynamics (heat sources and sinks), software engineering (event sources and sinks), or particular features of vector fields.