What common ways are there to get a precise current?

[ELS122]

What common things have very precise current ratings? for making a "precise" current source.

I was thinking maybe leakage currents in semiconductors, if you get a fairly constant voltage with something like a 431, maybe you can get very precise leakage current trough a BJT or a diode or whatever?

Maybe quiescent currents in op amps? how stable is that?
You could make the temperature more or less constant by using ice water that would be around 0 degrees C.

Maybe having a straight piece of wire perpendicular to earth's magnetic pole and measuring the degree deflection on a compass and tuning the current until you get x degrees deflection, and if all the dimensions are precise, so is the current flowing trough it?

[TimFox]

The usual method to generate a precise DC current is to pass the current through a very high quality resistor to get 100 mV or more DC voltage, and then use normal voltage measurement techniques across that resistance ("shunt") to regulate the voltage that drives the load to stabilize the current.
My go-to device for DC current is a Keithley 225.
Relying on leakage current through a semiconductor device is problematic, since leakage currents increase exponentially with temperature.
Wires and compasses:  Hans Christian Ørsted did this in 1820, to demonstrate the magnetic field induced by the current, and the well-known d'Arsonval moving-coil and pointer meters exploit the effect to measure current (but not directly to high accuracy or precision).
There are "current diodes" that are really JFETs with zero gate-source voltage (short circuit), and I have used more complex devices from IXYS  https://ixapps.ixys.com/Datasheet/DS98703A(IXC-SERIES).pdf  and  https://www.ixys.com/Documents/AppNotes/IXAN0053.pdf  (the latter can use an external resistor to set the current). 
Specifically, I have used the latter IXYS part for a constant-current plate load on a vacuum triode to enhance PSRR, although they do make higher noise than a simple resistor for the same current.

[ahbushnell]

Check this out.

https://www.allaboutcircuits.com/technical-articles/how-to-design-a-precision-current-pump-with-op-amps/

[TimFox]

Notice that in each of the circuits in that article, there exists a sense resistor to generate a voltage to be compared with a reference voltage or input voltage.

[mikerj]

What common things have very precise current ratings? for making a "precise" current source.

I was thinking maybe leakage currents in semiconductors, if you get a fairly constant voltage with something like a 431, maybe you can get very precise leakage current trough a BJT or a diode or whatever?

Leakage currents tend to be very unstable and change with applied voltage, time, temperature etc.

Maybe quiescent currents in op amps? how stable is that?
You could make the temperature more or less constant by using ice water that would be around 0 degrees C.

It's not a well defined parameter (other than min/max) so no guarantees it will be stable.

Maybe having a straight piece of wire perpendicular to earth's magnetic pole and measuring the degree deflection on a compass and tuning the current until you get x degrees deflection, and if all the dimensions are precise, so is the current flowing trough it?

You just re-invented the moving iron current meter (galvanometer), which can measure current but not produce stable current itself.  A quality moving coil meter would be more sensitive and repeatable, and modern digital meters can be even more sensitive.

As Tim says, the usual way is to use a stable resistor to sense the current (convert to voltage) and them compare it to a voltage reference with some kind of differential amplifier (e.g. op-amp) which then modifies drive current until equilibrium is reached.  Such a circuit can be made very easily with a TL431 voltage reference driving a suitable transistor, check out this app note: https://www.ti.com/lit/an/snoaa46/snoaa46.pdf

[ejeffrey]

Generally, precise voltage + precise resistor -> precise current is the answer at least for general lab use.  The details depend a lot on whether you want 1 amp or 1 microamp, what compliance and bandwidth you need, and what (if any) power budget you need to stay within.

[ELS122]

What common things have very precise current ratings? for making a "precise" current source.

I was thinking maybe leakage currents in semiconductors, if you get a fairly constant voltage with something like a 431, maybe you can get very precise leakage current trough a BJT or a diode or whatever?

Leakage currents tend to be very unstable and change with applied voltage, time, temperature etc.

Maybe quiescent currents in op amps? how stable is that?
You could make the temperature more or less constant by using ice water that would be around 0 degrees C.

It's not a well defined parameter (other than min/max) so no guarantees it will be stable.

Maybe having a straight piece of wire perpendicular to earth's magnetic pole and measuring the degree deflection on a compass and tuning the current until you get x degrees deflection, and if all the dimensions are precise, so is the current flowing trough it?

You just re-invented the moving iron current meter (galvanometer), which can measure current but not produce stable current itself.  A quality moving coil meter would be more sensitive and repeatable, and modern digital meters can be even more sensitive.

As Tim says, the usual way is to use a stable resistor to sense the current (convert to voltage) and them compare it to a voltage reference with some kind of differential amplifier (e.g. op-amp) which then modifies drive current until equilibrium is reached.  Such a circuit can be made very easily with a TL431 voltage reference driving a suitable transistor, check out this app note: https://www.ti.com/lit/an/snoaa46/snoaa46.pdf

Galvenometers rely on a magnet which could be weak, strong.
And springs which could be loose, or tight.

Meanwhile a compass would rely on the magnetic strength of the earth's magnetic pole, which I'd imagine is pretty constant.
Then just calculate how strong of a magnetic field would be needed to deflect the needle off the pole by x degrees...

[ejeffrey]

Earth's magnetic field isn't stable either and varies with location and time, especially location.

You can get precise measurement of currents using two coils and measuring the force between them, that is going to be much more accurate than the earth's magnetic field and doesn't depend on the strength of a permanent magnet.  This is the operating principle of the Ampere balance, for instance, but it's not really practical for most applications.

[TimFox]

"DCCT" direct-current transformers (flux gates) are useful to measure the ratio between two currents.
One form of current balance passes the test current through two windings, and balances the force between them in a gravitational balance against calibrated weights.
The old-school -hp- 428A current meter uses a flux gate around an unbroken wire, balancing the flux from the external current with a servoed current to give a null signal, which is the error signal for the servo.
The balancing current is measured in the normal way with a shunt resistor.

My education used traditional instruments to teach us what was really happening (as opposed to merely looking at digits):  galvanometers were useful as null indicators when balancing a bridge, but we didn't rely on the nominal calibration factor of the galvanometer itself.
Voltage measurements were referenced through appropriate bridges to "standard cells".

Not only does the Earth's magnetic field vary "all over the map" (literally, look up "declination"), but the strength of the magnetic dipole in the compass needle is not very stable, since proper use of a compass for navigation does not depend on its strength.
Even so, the compass needle, mounted on a very low friction bearing, does not respond to the strength of the local magnetic field, but aligns itself in the direction of the horizontal component of the local field, independent of its magnitude.

Back in the day (late 1890s) precise measurements of current used an electrolytic cell, measuring the mass of silver deposited from a silver nitrate solution and Avogadro's number to measure the charge (integral of current) in Faradays, which are defined as the charge of Avogadro's number of electron charges (1 Faraday is roughly 96,500 Coulomb).
https://www.nist.gov/si-redefinition/ampere-history

[ELS122]

"DCCT" direct-current transformers (flux gates) are useful to measure the ratio between two currents.
One form of current balance passes the test current through two windings, and balances the force between them in a gravitational balance against calibrated weights.
The old-school -hp- 428A current meter uses a flux gate around an unbroken wire, balancing the flux from the external current with a servoed current to give a null signal, which is the error signal for the servo.
The balancing current is measured in the normal way with a shunt resistor.

My education used traditional instruments to teach us what was really happening (as opposed to merely looking at digits):  galvanometers were useful as null indicators when balancing a bridge, but we didn't rely on the nominal calibration factor of the galvanometer itself.
Voltage measurements were referenced through appropriate bridges to "standard cells".

Not only does the Earth's magnetic field vary "all over the map" (literally, look up "declination"), but the strength of the magnetic dipole in the compass needle is not very stable, since proper use of a compass for navigation does not depend on its strength.
Even so, the compass needle, mounted on a very low friction bearing, does not respond to the strength of the local magnetic field, but aligns itself in the direction of the horizontal component of the local field, independent of its magnitude.

Back in the day (late 1890s) precise measurements of current used an electrolytic cell, measuring the mass of silver deposited from a silver nitrate solution and Avogadro's number to measure the charge (integral of current) in Faradays, which are defined as the charge of Avogadro's number of electron charges (1 Faraday is roughly 96,500 Coulomb).
https://www.nist.gov/si-redefinition/ampere-history

Well if location is known, isn't there a map of measured magnetic pole strengths?
The compass needle magnetic strength doesnt matter since the reading will be relative.
Adding a magnetic field 90 degrees off from the magnetic pole will move the compass needle more towards the coil magnetic field, the stronger the coil field the closer the needle will align itself to that field rather than the earth's field.
But I guess using weights would be more precise if indeed the earth's magnetic pole strength changes over time of day.

[TimFox]

How old is your map?  The magnetic poles migrate greatly over historic time, changing the field strength at your measurement point over just a few years.
Besides, the actual field strength is only roughly 1 gauss, not very useful to measure against the field from a wire.
A flux gate is a far better measurement of total field from two wires, or the current in one winding required to balance the external magnetic field.
The actual maps of the Earth's field do not worry about the magnitude, only the navigationally-important difference between the local (magnetic north) direction and true (geographic) north ("declination").

In one of the silly TV "documentaries" about the Bermuda Triangle, it was pointed out that the "line of zero declination" passes over Lake Michigan and runs to the Triangle.
You can calculate the declination (as a function of location and historical time)  at this website  https://www.ngdc.noaa.gov/geomag/calculators/magcalc.shtml  (free registration required).
To prove that linkage, it was noted than once a single-engine plane flew east out of Chicago over the Lake, and was never seen again.
Can't argue with that--it's too stupid.

[ejeffrey]

Well if location is known, isn't there a map of measured magnetic pole strengths?

Yes, the earth's magnetic field changes on a daily basis as well as with location.  Whats your real goal here?  If it's to produce or measure an accurate current, just use a voltage reference + resistor.

[tggzzz]

Well if location is known, isn't there a map of measured magnetic pole strengths?

Any map will be out of date. The magnetic north pole is currently moving at about 35/miles per year.

[ELS122]

Well if location is known, isn't there a map of measured magnetic pole strengths?

Yes, the earth's magnetic field changes on a daily basis as well as with location.  Whats your real goal here?  If it's to produce or measure an accurate current, just use a voltage reference + resistor.

To get a precise current.

The position of earth's magnetic pole doesn't matter either, like I said already, you position the wire/coil relative to magnetic north, 90 degrees off from it.
a flux gate or whatever will need a reference, but with a compass the reference is earth's magnetic pole.
Maybe there's some other more precise magnetic reference that could be used instead, to get better precision.

Obviously you could use a precision voltage reference, a precision resistor. But I'm talking about simple ways to get a precise current.

[tggzzz]

My education used traditional instruments to teach us what was really happening (as opposed to merely looking at digits):  galvanometers were useful as null indicators when balancing a bridge, but we didn't rely on the nominal calibration factor of the galvanometer itself.
Voltage measurements were referenced through appropriate bridges to "standard cells".

Yes indeed.

Youngsters can't conceive that, with the addition of a metre rule and some resistance wire and a known voltage, we could use 2% voltmeters to measure voltages to 0.1%. (It was easier when I got to university, since they had more advanced and expensive equipment.)

For the next couple of weeks I have a digital voltmeter where the voltage reference is the breakdown voltage of a neon bulb. That's surprisingly stable, when compared with the internal portable Weston standard cell - which is still working well!

And that raises a more important issue. When someone mentions "very precise" without giving limits or use cases, orange flags start waving vigorously.

Let's use lovely numbers, not loathsome untrustworthy adjectives

[tggzzz]

Whats your real goal here?  If it's to produce or measure an accurate current, just use a voltage reference + resistor.

To get a precise current.

Numbers not adjectives, please.

Are you sure you only need the current to be "precise"

[TimFox]

How much precision would you expect for a previously-determined magnetic field strength (on the order of 1 gauss, not very big)?
Even the direction is iffy:  where I grew up in northern Minnesota, the maps said "declination uncertain" due to the iron ore deposits.
The rotational period of the Earth is reasonably well determined and constant, but the generation of its magnetic field is a complicated process, with lots of junk in between the iron core and the surface.
Fluctuations in the magnetosphere (external to the Earth, at 10 to 100 times the Earth's radius) due to Solar effects can cause short-term fluctuations in that region up to 10%, so there will be fluctuations even at the surface.
NOAA states: "Earth's magnetic field intensity is roughly between 25,000 - 65,000 nT (0.25 - 0.65 gauss)".  https://www.ncei.noaa.gov/products/geomagnetism-frequently-asked-questions

You would do better to use a AA Duracell as a voltage reference and a cheap carbon resistor to make your current.

Back to basics:  physics defines current in terms of the force per meter between two infinitely-long wires carrying the same current.
In engineering, it is important to have repeatable standards that are very close to the physical definition, agreed to international bodies: 
At present, the Volt is defined by a "cryogenic Josephson Junction", related to a frequency defined in terms of the second.
(Microwave frequencies can be measured to incredible accuracy against the cesium-beam standard, roughly good to 10^-10.)
Also, the Ohm (ratio of volt/ampere) is now (since 1990) defined by the "quantum Hall effect", using an internationally agreed value for the "Klaus von Klitzing constant"  R_(K-90) = 25812.807 Ω.
(I leave the meaning of that to you youngsters.  When I was a lad, the ohm was defined in terms of several resistors kept at the NBS before it became NIST.)

If you want something more portable than these lab systems, try using any of the good voltage reference ICs and a good precision wirewound or metal foil resistor, calibrated at the factory.
In the back room at Fluke, etc., they have these expensive lab systems for primary calibration.

edit:  emphasis added

[PCB.Wiz]

Obviously you could use a precision voltage reference, a precision resistor. But I'm talking about simple ways to get a precise current.

Surely, the obvious way is also the simple way ?
Voltage references are available at low prices, because they are made in very large volumes.
Resistors have a precision/price curve, so you can simply choose how much precision you want to pay for, in your very simple circuit, that is also simple to calibrate and check.

[bdunham7]

Obviously you could use a precision voltage reference, a precision resistor. But I'm talking about simple ways to get a precise current.

I think the things you are talking about are neither simple nor precise.  If you want a meaningful answer, are you simply looking for a single, precise current under one set condition or are you looking for a circuit that will provide that precise current over a range of loads?  Typically when you talk about a current source, you also specify a compliance voltage and the expectation is that the circuit will source or sink the specified current at any voltage within that limit.  Then, how exacting do you want this current source to be?  The simplest to construct circuit I know of is simply a linear voltage regulator with a resistor in series, along with some sort of unregulated power source like a battery or PSU.  Most precision current sources in practice are simply a variant of this circuit.

[EPAIII]

While this is a tempting circuit, it seems that it's accuracy depends not only on the precision of the resistor, but also on the exact operating point the LM317 wants between it's output and the adjust terminal (1.2 to 1.3 V according to the TI datasheet so +/- 8%). That 8% makes the advertised 0.7% temperature range of the LM317 look downright great.

So it could be somewhat stable (neglecting temperature effects) for an individual LM317, but it could be all over the map for even several from the same production batch. Of course, if you add a trimmer pot and do a calibration with an external meter things do get better but probably still around +/-1%.

I am sure this is why more complicated circuits are often used. I like that two OP amp circuit a lot better.

Obviously you could use a precision voltage reference, a precision resistor. But I'm talking about simple ways to get a precise current.

I think the things you are talking about are neither simple nor precise.  If you want a meaningful answer, are you simply looking for a single, precise current under one set condition or are you looking for a circuit that will provide that precise current over a range of loads?  Typically when you talk about a current source, you also specify a compliance voltage and the expectation is that the circuit will source or sink the specified current at any voltage within that limit.  Then, how exacting do you want this current source to be?  The simplest to construct circuit I know of is simply a linear voltage regulator with a resistor in series, along with some sort of unregulated power source like a battery or PSU.  Most precision current sources in practice are simply a variant of this circuit.