high-side current sensor

INA280 looks pretty good.

I was going to suggest the INA139 (which I've used) but it only goes to 40v...

INA169 maybe?

The LT6106 is an easy to use device, but again, 36 volts only. It only requires a shunt resistor, a series resistor and a final gain setting resistor. You could design the same circuit yourself as the chip is really just an op amp and a transistor (PNP or PFET). Know any good opamps that work up to 48 volts?

On a sunny day (Mon, 10 Jan 2022 19:26:09 -0800) it happened snipped-for-privacy@highlandsniptechnology.com wrote in snipped-for-privacy@4ax.com:

I have some of those Hall modules somewhere HX03-P 3-50A and HX10-P Manufacturer LEM:

all I can say

So, is this regulated voltages, nine of 'em, and you want the output currents from nine low-V regulators, or is it input to the regulators, and is that input somewhere in the vicinity of 48VDC? Are these nine currents going to include initial charging of filter capacitors? How big are those capacitors? Are some of the regulators switchers?

As for 'intelligent cutoffs', do you want to do that at the high-side? A secondary low-current power supply grounded at the high rail might be economic, if you want to have nine sensors and nine cutoffs all near the +48V source.

You don't need a high voltage opamp - just hang a cheap 5V riro opamp from the +48V rail with a 5V zener and resistor, the pnp c-b junction does all the standoff.

60V rated Zetex/Diodes ZXCT1084 is another option.

piglet

MAX4080 ---> 76V

I realy like it....

Olaf

Does it work OK? Looks nice, but it violates my Never Buy Maxim rule. Mouser has stock.

I could maybe do it with 5 or 6 discretes for about 10 cents.

I've used loads of INA169, they work fine. And they work at temperatures well past their specified maximum which may be an indication of their robustness.

48V is within range of a bunch of those cheap high monitors.

eg. ZXCT1082/83/84/85/86/87 Diodes Inc. (nee Zetex) but you need a shunt resistor, of course. AD has some in theory but availability?

There are those Hall sensors, TI, (Allegro?) etc. Availability?

Last time I did this for a flight instrumentation package I used a RRIO op-amps rated for the supply voltage (OPAxxx) but I don't think there's anything easily available safe for 48V direct supply and if there was it wouldn't be cheap x 8.

Maybe use a 5V DC-DC to generate a rail for op-amps @48-5V, use your most common inexpensive low-Vos RRIO 5V dual or quad op-amps and PNP transistors or p-channel MOSFETs to transmit the current to near GND. Should be cheap enough and relatively low parts count.

Current mirror using pairs of high voltage PNP transistors?

This ain't bad:

With 50 or better yet 100 mV across R1, it should be fine. It's sure cheap. R4 performs four distinctly useful functions.

Official current monitor chips seem hard to get lately, and are priced several times what seems normal.

I dont use it in high production, but in many small circuit to measure current for prototyping. Work good!

Olaf

I just posted something like that. It should be good enough. If the voltage drop across Q2 is pitched a little high, so there is a little positive measurement offset, we can math that out.

There will be eight user-programmable power supplies, each running off

+48, all off a common +48 bus fed from a kilowatt bulk power supply. Users can potentially install modules and program and load the supplies such as to cave in the main source, which would be really ugly. That's against the rules in the manual, but we need to protect things if they do it.

So I want to know all 9 currents, all at 48 volts. Given that info, we can decide that the main supply is overloaded and shed as many load boards as needed to keep the main supply happy. And later recover. All in a sensible documented way. We are debating algorithms for that.

We will have a Zynq fpga/ARM chip, with an integral ADC, so we can program the FPGA to mux and digitize the currents furiously and leave the results in 9 registers. Then some FPGA or c-code algorithm can do the shutdown algorithm.

I'd expect that we could do it in c code, say 1000 times a second. My programmers tend to push back against doing fast realtime stuff, but they will have two 600 MHz ARM cores for Pete's sake.

Nice! Certainly works, and is very simple and cheap. A bit temperature sensitive (even with matched transistor temperatures). (~2:1 from 0 to 80°C)? Maybe that's good enough for a circuit breaker type function.

You could use 1/4 of a 10 cent LM324 per channel but it would complicate the power supply.

Yes, some are stupidly expensive. And if the customer has designed them in, they'll probably be forced to pay whatever is asked.

Tempcos should be OK. One of the functions of R4 is to reduce self-heating in Q1.

There are several basically identical 5-pin current sense chips... all with different pinouts!

There seem to be lots of >50 volt PNPs for around 2 cents each. We have stock here on one that cost us - gasp - 8 cents.

I've use the above current sense form (more or less) from time to time. My use is typically active bias of class A RF/MMIC amps.

1st time I used it was for biasing a 4W GaAsFET from Celeritek. In fact, Celeritek suggested the basic form. Some polish had to be added. It does temp drift a bit, as already mentioned by Spehro.