a couple of highside current sensors

We call that one swell foop.

Yeah, well that's a little TOO elaborate for me; the use-another-power-supply route seems less of a problem when some of the solutions are seen.

Jim Thompson's two-transistor solution (with input through the low-Z emitters instead of bases) has some interesting features, too

It's another variant of handstand...

What I'd really like to see, is a logarithm-of-I scale, so I could make a lab supply that had good-ish resolution but didn't need range switching. My first way to do that, though, was to compare analog current monitor voltage against a triangle wave, and sample/hold the exponential decay of the asociated square wave into an RC circuit, to present to the meter. Or, to thermostat-control a well-characterized base-emitter voltage against the scaled current into the colledtor, and display BE voltage.

Both are slightly too elaborate to get past my laziness, so I just continue to range-switch the silly (only ONE meter, but two voltage and two current ranges) supply on my bench It gets pegged, a lot, because the white paint stripe on the knob is kinda... dirty.

LT1001AM maybe, guaranteed < 15uV offset, < 0.6uV / degree C drift, < 2 nA bias current

OP07A is a cheaper alternative

Needs to be RRIO and fairly low power. I like OPA197 lately, but it's fairly expensive.

It has to handle the bus voltage too, but that could be fudged.

it's not bias current, but Icc that has to be low and controlled.

Stick an LT1001A in the sim to see the issue.

RL

The models do a lousy job in simulation. Somebody forgot to cap Icc.

RL

Ah yea I see that.

With that in mind the most appropriate device in Analog Device's stable looks to be the LT1637, like a lot of things AD they slam you for small quantity but it's about $1.80 in 100s.

The inputs are over-the-top up to 44 volts so could run it off a lower supply voltage and cut the bias current down even further. works good in the sim

TI does some ADC equipped volatge and current sensors INA3221 is a triple

Oddly if you use the LTC1637 "over-the-top" amp and configure its supply rail in the same circuit like this:

the sim claims the amp draws no Icc thru the positive supply. At all.

The current sense feature still works fine at least down until the current thru the 100 ohm resistor becomes on the order of the normal supply current, then it craps out.

That's weird...

Definitely something to do with the model, the voltages on the transistor Q2 don't even make sense.

A very nice idea

But, isn't difficult to find a high voltage positive rail opamp that has low operation current?

Cheers

Klaus

=0

Maybe trying to use this circuit that has limitations? The voltage across R1 at 1 amp is 0.1 volts, it's a 0.1 ohm resistor with nearly no current in the amp input, how could it be anything else?

The rest of the circuit is the typical balanced voltages on different value resistors, R1 and R3, resulting in currents in the ratio of the resistors. This is normally done with a FET which means the same current in R3 match es the current in R2 resulting in another gain step. So the voltage at the output should be I*R1*R2/R3. However, in order to save one transistor Lar kin uses the negative rail connection of the op amp for the current sense w hich pollutes the signal with the op amp operating current.

In the circuit with the transistors it fails to work further by introducing unequal offsets in the BE junctions due to unequal currents. For the curr ents in the two BE junctions to be equal, the resistor currents and voltage s should be equal. You can only make the claim that the two resistor volta ges are equal if you can show the BE junction voltages to be equal, i.e. ci rcular reasoning. That's only a reasonable approximation if the voltages o n the resistors are the lion's share of the total voltage drop meaning sign ificantly more than 0.7 volts. This can also be approximated if the beta o f the two transistors are equal, but that requires a matched pair. Is that really any better than just using an op amp and a FET in the standard circ uit that doesn't have any of these issues?

Here's one with the FET included, SOT23, under a buck, just add two resisto rs!

Got to be problems with the model. LT1001A draws the typical 1mA, not enough to get the abysmal results of the sim.

Perhaps it's input offset voltages reacting to the low source impedance. Input pin currents are off the wall.

RL

Opamp max supply voltages keep creeping down. That's annoying. A lot of our stuff runs off a +24 supply, and some recent boxes use a 48 volt wart. High voltage opamps are rare.

Of course highside current sensor chips are common, but I'm a circuit designer. And I don't like to add new parts to stock (parts that will eventually go EOL) when I can make something from what we have already.

My current gumdrop is OPA197, but its supply current is about 1 mA. So one might add a transistor. At that point, there are better circuits.

This adds the transistor and deletes the opamp!

A problem with the LT1001 etc. is it's not rail-to-rail IO; the inputs have to be at least 1 volt below the V+ AFAICT

Frustratingly AD makes a lot of high-voltage precision op amps with rail-to-rail outputs but not that many with rail-to-rail inputs.

On a sunny day (Fri, 21 Aug 2020 07:44:29 -0700) it happened snipped-for-privacy@highlandsniptechnology.com wrote in :

Long time ago I bought some Hall sensor modules to measure high current (well what is high?) DC:

For AC I like current transformers like this:

There are surface-mount Hall current sensor chips. One of them just sits above a big PCB trace that carries the current.

Instead of a current transformer, you can use a surface-mount shunt resistor and a signal transformer, standard small cheap parts. The dual-winding inductors are good. An advanced trick is to add a resistor in series with the primary and drive the secondary into a summing point, essentially run the transformer zero-flux.

I can confirm that the model for LT1637 does work in the sim.

RL