MOS current sensing with IR25750LTRPBF: alternative?

[DiTBho]

hi
I am trying to design a power-module with a P-channel MOSFET, but I need to measure the current flowing trough it.

I can use a big MOSFET with 35mOhm RDS_ON, and measure the current by measuring the Voltage between the drain and the source, such big MOSFETs "should" have a rather stable RDS_ON when the current is 2A of the 100A (2%) they are able to provide.

Anyway, it might be a wrong assumption, the RDS_ON might double with the temperature.

A couple of application notes suggest the right solution is something like the "IR25750LTRPBF" chip, which looks great and brilliant, and it's exactly what I need, unfortunately it's no more in production.

Any alternative? Otherwise I will try to reproduce its working scheme with discrete components 

[T3sl4co1l]

Right, to get the actual current you need a PTC to compare it to; or an NTC that's perfectly matched, but good luck with that.

Cute IC, easy to see why it's obsolete though.  Very simple function, probably something one or a couple customers needed a few million of, then designed it out and *pfft* there goes the demand.

There are a few MOSFETs with a sense pin, which I think is usually something like a small piece of the source terminal, so a known fraction of the load current flows through it (when its voltage is kept near the main source anyway) which can be sensed with any old shunt + current sense amp.  These might not be available in P-ch or the ratings you need... no idea.

N-ch is a good idea just for the bump in performance and cost savings, even if the circuit becomes wonkier (bootstrap drive? high CMR sense amp?).

Tim

[Pineapple Dan]

Have a look at Allegro ACS hall effect sensors. Didn't ever hear of the IR25750LTRPBF, pity it discontinued

[exe]

I heard hall sensors require calibration (sometimes done automatically), and they don't like magnetic fields. So, how much precision do you need?

[T3sl4co1l]

It's pretty uncommon (not at all impossible, of course) to have enough ambient magnetic fields to worry about, and Hall effect sensors are good to whatever it says on the tin -- typical 5% or so, with 1% or a bit better being available, and beyond that you really should be considering a different technology as the baseline noise is just not very helpful when you're dealing with that much sensitivity.  (That is to say, the dynamic range is about 40dB.)

For greater accuracy, VAC's line of saturating-ribbon servo sensors comes to mind?

Meanwhile, the bandwidth is not very good, if you need much more than 10s of kHz (Hall sensors go up to a few 100kHz, don't think I've seen more than that) you should really consider a shunt resistor, or current transformer if it's pulsating.

Note that, in a switching converter, while you can't put a CT on the inductor itself (in CCM), you can put a CT each on the two switching devices (diode/transistor), with a half-wave burden as well (diode into resistor).  With a shared burden resistor, the sensed current will be continuous, while the current through each transformer is discontinuous (at least up or down to a modest duty cycle -- it fails when there isn't enough time to reset between pulses, so duty should stay in the 10-90% range, say).

Also also, if this is slow switching and 100% duty, consider a gate drive with charge pump power, or even a photovoltaic driver.  Doesn't solve the current sense, but can handle one of two annoyances with high side N-ch application.

Tim

[thm_w]

Yeah quite a nice IC. A few are left in china, but not anywhere in NA that I see. So not something I would use.

What sort of budget, voltage, currents are we talking about?

There are many types of smart FETs as mentioned. Some like the VND with current output: http://www.st.com/st-web-ui/static/active/en/resource/technical/document/datasheet/CD00191501.pdf
Although its not very accurate, maybe better ones exist.

[langwadt]

ina169 or similar and a shunt?

[DiTBho]

What sort of budget, voltage, currents are we talking about?

2-3A, the budget is not problem.
But I need the load to be "high-side".

I don't like/want to use hall-effect sensors.

[DiTBho]

how much precision do you need?

+/-1mA over 2A of max load.

[DiTBho]

The current on the load changes with a frequency of max 100Hz. It can be considered enough "slow".
The 12bit ADC will sample it at 2Khz, with an iterative mean window of 16 samples.

So 16 samples -> mean -> measured value.

[exe]

There are devices like INA220 that do current and voltage monitoring, and have i2c/spi outputs. I don't remember which one I wanted to use, so can't suggest anything except that "consider searching 'power monitors' on https://www.ti.com/amplifier-circuit/current-sense/power-current-monitors/overview.html ".

Beware of the specs, last time I checked a power monitor ic, it was more like for 3.5 digit dmm. So, with single range absolute precision of 1mA for for 1A range is kinda on the border of what such solutions can provide.

PS Oh, I found it! Consider INA238/INA239. But, again, check specs

[Wolfram]

how much precision do you need?

+/-1mA over 2A of max load.

Precision or resolution? For +- 1 mA max error, you have an error budget of 0.05 %, or 500 ppm. For reference, the Rdson of a 100 V MOSFET changes by about 1 % per degree celcius, so you'd need to know the actual junction temperature to a precision of 0.05 degrees, caused by both changes in ambient temperature and significant self-heating due to the current. And that's after you calibrate out the 30 % uncertainty in Rdson from the factory, not including any aging effects.

You have a better chance with a dedicated shunt resistor, though with an error budget of 500 ppm is still going to be very tricky. The error budget will have to be spread across temperature coefficients of components (these can be compensated out with some effort, but only as well as you can accurately know the temperature, and temperature coefficients also have uncertainty) and offsets from any amplification. Component tolerances can easily be the biggest error, but these can be calibrated away.

Cheap current sense resistors are available with a Tc of as low as 20 ppm/C, so if you calibrate out any component tolerances and know the resistor temperature to within a few tens of degrees, it could be doable. Keep in mind that every other resistor in the design also has a temperature coefficient, so if the ambient temperature changes significantly, you also need to account for this, given that the thermal distribution can be different depending on if the heat comes from your shunt resistor or from the external environment.

[DiTBho]

Precision or resolution?

Resolution, I still don't know about the precision of this engine, but the customer wants +/- 1mA

[DiTBho]

I cannot find a P-mos with a sensing gate.

Do you think, guys, is it possible to design a current-mirror configuration where the big-P-channel mos feeds 2A to the load while a small mirror-p-mos feeds the shunt resistor with a fraction of that current?

Will it work? Impossible? or just bad idea?

[JPortici]

I think this is what happens with intelligent high side switches like the TPSxH series (i used TPS2H160 in a prototype for a simillar application and now that i have to make a production run every member of the family - 1,2,4 channels various rds on and current sense option - is out of stock)

[exe]

I cannot find a P-mos with a sensing gate.

Do you think, guys, is it possible to design a current-mirror configuration where the big-P-channel mos feeds 2A to the load while a small mirror-p-mos feeds the shunt resistor with a fraction of that current?

Will it work? Impossible? or just bad idea?

Afaik it has to be a monolithic device. Also, from what I've seen, precision and drift of such solution is quite poor. I think shunt is the best current sensing option in terms of precision, bandwidth and drift at least up to 5-10A.