1. Home
  2. Electronics Design
  3. Practical current mirror

Practical current mirror

Gents,

I'd like to measure voltage of a low impedance source (10mOhm), which is in range of +/-200mV. The ADC cannot handle negative voltages and I can't do the obvious trick with connecting one of the input terminals to, say, VDD/2 and perform differential measurement because the shunt's terminal must be connected to GND. I don't want to go dual supply either, as it introduces more problem than it solves. So the basic idea is to implement a level shifter based on a PNP/PMOS current mirror.

Curiously, even at the Spice level the PNP variant using BC857B simulates poorly, the current waveforms are in sync, but shifted by a detectable offset (the collector current is in the 1mA ballpark). The PMOS variant on BSS84 simulates extremely well. Suspiciously well.

But that's a simulation. What should I expect in practice and how to do it properly? Are the real dual MOSFETs in SOT343/SOT23-6 matched well enough?

Best regards, Piotr

Reply to
Piotr Wyderski
Loading thread data ...

Why not use a resistive divider to raise it up above ground?

Cheers

Phil Hobbs

Reply to
pcdhobbs

Post the sim?

How about two rrio opamps, one inverting and one non-inverting?

Or a diffamp with around-zero common-mode compliance, and its output reference at Vcc/2? There are parts like that, or you can make it.

A bandgap could do a simple level shift, up into the ADC's range. Auto-zero it when the shunt current is known to be zero.

Or just shift the shunt voltage up with a couple of resistors.

Lots of modern delta-sigma ADCs would accept your differential shunt signal directly.

Flying capacitors are always fun. A flying capacitor voltage multiplier even more fun.

Ooh, how about chopping it, AC-coupled shifting up, and doing a software lock-in?

Most dual transistors these days are just two chips in the same package. Matching is mediocre, especially fets.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics
Reply to
John Larkin

Indeed, thanks to the low source impedance injecting some current does not distort the measured value too much and such a solution is perfectly acceptable for my purposes. Will use an SMD resistor ladder due to the tempco reasons. Thanks, Phil!

So the main issue has been solved, but how about the more general problem uncovered by the simulation? Is BCV62 or compatible worth using, if even the perfectly matched transistors Spice has plenty of produce rather disappointing results? Perfectly matched MOSFETs exhibit superior performance, but I don't believe the real ones are even close.

Best regards, Piotr

Reply to
Piotr Wyderski

BCV61 and 62 are both EOL, iirc.

The main issue is the poor thermal coupling between the two separate devices, but beta errors are also important.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com
Reply to
Phil Hobbs

Post the sim?

How about two rrio opamps, one inverting and one non-inverting?

Or a diffamp with around-zero common-mode compliance, and its output reference at Vcc/2? There are parts like that, or you can make it.

A bandgap could do a simple level shift, up into the ADC's range. Auto-zero it when the shunt current is known to be zero.

Or just shift the shunt voltage up with a couple of resistors.

Lots of modern delta-sigma ADCs would accept your differential shunt signal directly.

Flying capacitors are always fun. A flying capacitor voltage multiplier even more fun.

Ooh, how about chopping it, AC-coupled shifting up, and doing a software lock-in?

Most dual transistors these days are just two chips in the same package. Matching is mediocre, especially fets.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics
Reply to
John Larkin

I don't understand the problem. Why mess with discretes?

A simple op amp inverter will convert a negative signal below ground to a positive above ground. You can use a LM358 which has a common-mode range that includes ground.

Reply to
Steve Wilson

Sure, below. Connect the one you wish and compare the resistor currents.

Can they go slightly below GND?

Yes, that feature is definitively nice. I think it will also work with the SAR ones if the differential input stage is designed properly. The problem is that I want to detect anomalies in about a microsecond, which makes the ADC data stream a kind of a torrent. And there will be many of them. I could handle this tsunami, but I don't want to solve the problems which aren't there. A hybrid detector composed of a slow ADC for the mains quality inspection and a purely analog bag of window comparators becomes increasingly appealing.

BW must be around 1MHz, so chopping would be a bit complex, sufficiently faithful reconstruction even more.

What's their structure-to-structure withstand voltage? Can't find it easily.

Best regards, Piotr

----------------8

Reply to
Piotr Wyderski

But still easily available. Will buy a hundred of each just in case.

Best regards, Piotr

Reply to
Piotr Wyderski

Because they are much less picky on the power supply issues. A transistor doesn't have insanely low VDD_MAX, just a bundle of breakdown voltages and there is a huge collection of them.

It is simple if you can afford a dual isolated supply, otherwise things are getting interesting.

Best regards, Piotr

Reply to
Piotr Wyderski

Most have p-fet or PNP front-ends and some can can go a few hundred mV below ground. Chopamps usually go below ground too.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics
Reply to
John Larkin

LM13700 (transconductance amplifier) is a fairly good dual current mirror, and you can choose source or sink polarity by a pullup resistor on one of the differential input pins (grounding the other, and the 'linearizing' diodes).

MOS mirrors have a noise advantage, but lower transconductance than bipolar. If a MOS mirror is to be made 'ideal', it must be cascode-output style.

Reply to
whit3rd

You don't need a dual supply with the LM358. It accepts a wide range of supply voltages and the common mode range includes ground.

Ground the positive input terminal. Connect a 1k from the negative input to your -200mv source. Connect another 1k from the negative input to the op amp output.

Voila. You have an inverter with a gain of one.

Any other op amp with a common mode range that includes ground will work.

The LM358 has a gain/bandwidth product of areound 500KHz for moderate supply voltages. If you need more bandwidth, choose another op amp. There are many, but the price goes up with the bandwidth.

Reply to
Steve Wilson

Someone makes a chip that does exactly what you describe; it's a level-shifter for high-side current sensing that internally is a made up of a BJT current mirror and sources. I'm blanking on the manufacturer, I want to say TI but not sure. Having trouble finding it on Mouser after

5-10 min of looking but I know it's here somewhere...
Reply to
bitrex

Ah, it's Diodes Inc. Discontinued at Digi-Key but still have some available, doesn't seem to be discontinued by the manufacturer tho:

Reply to
bitrex

p
I

I was going to ask, why are you wasting time searching on Mouser? :^)

George H.

Reply to
George Herold

One reason is that Mouser's search engine is great and Digi-Key's is a hot mess

Reply to
bitrex

There are many far worse search engines. Some actually don't even know that 500mA is less than 1A, because 500 is obviously greater than 1, so the search results are often amazing.

Best regards, Piotr

Reply to
Piotr Wyderski

My application is actually low side alternating current sensing, because the MOS drivers require positive supply voltage and the shunt is between the sources (as the sim shows). But this chip can be used in a different part of the same device, so thanks anyway! They provide a schematic and it is indeed very close to my models.

Best regards, Piotr

Reply to
Piotr Wyderski

Duh?

I've never seen a category where Mouser's was more accurate and organized.

There are plenty of categories where both fall flat. Connectors, mechanical anything, you name it. But among the usual sorts of things like semiconductors?

On a related subject, can we all get some hate for Allied's system? Yeah. I haven't bought a damn thing from them in years...

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: https://www.seventransistorlabs.com/
Reply to
Tim Williams

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.