solved: Rail to rail *input* opamp for Vdd=36V?

[Zero999]

Hello incf,

It might be that using a vanilla opamp and some current mirror voodoo will do the trick. But of course the path is a bit steep and final specs could be disappointing.

If you need mutiple source for a dedicated part (in this case, the current sense amplifier), you may want to select 2 or 3 parts from different manufacturers and put multiple footprints on the PCB. You only populate one of those, depending on parts procurement.

Here's the list of Current Sense Amplifier from LT/AD (uni and bidi versions, up to 105 V, zero drift, 'Over-The-Top' feature, which is even better than 'rail to rail').

I have to admit that I'm not sure offhand how the opamp's CMRR spec would affect it's ability to serve as unity buffer/current mirror

If I understand the formulas right, a CMRR of 80dB on a unity gain buffer would give a offset voltage of 100 microvolts per volt - I think?

So from 10V to 30V a common mode offset voltage of 2mV might develop (equivalent to an error of +/-0.2 amps across a 10 milliohm shunt)

It's the input offset voltage which is the main problem for the current mirror circuit. The CMRR isn't a big deal because both inputs stay around the same voltage. The supply rejection ratio is more of an issue, if the input voltage doesn't stay constant, but that can be improved with the zener diode.

While I like the concept of the LT6015, it doesn't fit the bill.

I'm limited to the cheap commodity type parts with multiple suppliers.

There are plenty of high voltage JFET input parts which have an input common mode range that includes the positive supply.  The venerable LM301A bipolar part also works to the positive supply.

Which ones? so far I've got

• TL071/072/074
• TL081/082/084
• LM301A

edit: gee... the '301 really is a lousy amplifier. The large signal gain is only 15k not the usual >1M - I feel like having so little gain would introduce offset error (2mV @ 30V I think)

The stuff about clamping is nice to know, but irrelevant to this application.

The low gain shouldn't be a problem with the current mirror circuit, because the gain of the circuit is close to unity. The output is connected to a source follower, which means the voltage on M1's source is the same as the gate, plus a fixed threshold voltage. Changing the input voltage by 100mV, causes the op-amp's output to change by roughly the same voltage.

I was going to suggest the  ZXTC1009, because it's very easy to use, but it's not a generic IC and is only rated to 20V, so will need some tricks with a cascode to get it to work at higher voltages.
https://www.mouser.com/datasheet/2/115/ZXCT1009-93594.pdf

Schematic for educational purposes only. M1 has a potential divider which delivers half the supply voltage to its gate. The source voltage follows the gate voltage and is always a couple of volts above the gate. The voltage across the ZXTC1009 is therefore half the supply voltage minus a couple of volts. No current flows through the MOSFET's gate, so the drain current is the same as the source. The ZXTC1009 changes its current draw dependant on the the supply voltage and is sensed by R4. It works on the same principle as the current mirror circuit in my previous post.


A cascode circuit similar to the above can be used with the current mirror circuit I posted, to improve the supply rejection ratio. Engineering is all a compromise. If you want better accuracy, lower offsets, higher supply rejection etc. then it means more costly ICs and/or more components.

[incf]

I've committed myself to using something like the TL081 in a current mirror.

It's performance in the simulator is "adequate". I only wish those ancient amplifiers were widely available (ie. multiple suppliers) in SOT-23-5 packages.

Also sort of wish that low offset/chopper stabilized versions of those ancient amplifiers were ubiquitous enough to meet the criteria for use.

[Kleinstein]

There are plenty of chopper stabilized OP-amps and many of them have rail to rail inputs. Only the choices with more than 30 V supply range is a bit limited.

The types are usually also manufacturer specific, though rather similar types can often be found from different sources.
It is no longer common to have 2nd sources, but there can still be similar parts with a different name. Even with the same number, but a different manufacturer there can still be differences. E.g. 74LV4054 from Ti and NXP have quite different on resistance.

The simulations usually don't inlcude the offset and offset drift part.

The old types may be too large to fit in a SOT23. So one may have to look at modern parts.
The TL081H is available in sot23, but it is a quite different part.

[Zero999]

I've committed myself to using something like the TL081 in a current mirror.

It's performance in the simulator is "adequate". I only wish those ancient amplifiers were widely available (ie. multiple suppliers) in SOT-23-5 packages.

Also sort of wish that low offset/chopper stabilized versions of those ancient amplifiers were ubiquitous enough to meet the criteria for use.

The TL081H is available in SOT23-5, but you don't get the offset trim, which is presumably why you're using it rather than the TL082.
https://www.ti.com/lit/ds/symlink/tl081a.pdf

And, please do breadboard it. Don't rely on SPICE models which are imperfect.

[David Hess]

If you use the circuit posted by Zero999, then the operational amplifier does not need to be high voltage:

https://www.eevblog.com/forum/beginners/rail-to-rail-*input*-opamp-for-vdd36v/msg5861273/#msg5861273

The transistor sees the full voltage, but the zener diode provides a regulated low voltage negative supply to the operational amplifier.  Now the only requirement is an input voltage range which includes the positive supply.  The OPA186 is such a chopper stabilized part which is inexpensive and available in the SOT-23 package, but there are several options.

[incf]

If you use the circuit posted by Zero999, then the operational amplifier does not need to be high voltage:

https://www.eevblog.com/forum/beginners/rail-to-rail-*input*-opamp-for-vdd36v/msg5861273/#msg5861273

The transistor sees the full voltage, but the zener diode provides a regulated low voltage negative supply to the operational amplifier.  Now the only requirement is an input voltage range which includes the positive supply.  The OPA186 is such a chopper stabilized part which is inexpensive and available in the SOT-23 package, but there are several options.

You (and Zero999, et. al) are right, I suppose a single zener + power resistor is probably worth the flexibility and increased accuracy.

[Alex Nikitin]

With this kind of accuracy requirements I would go for a Hall-effect based sensor.

Cheers

Alex

[incf]

With this kind of accuracy requirements I would go for a Hall-effect based sensor.

Cheers

Alex

I did not know that those modules existed (and are not as expensive as I would have thought). While I can't use them, I will remember them in the future.

[KE5FX]

You (and Zero999, et. al) are right, I suppose a single zener + power resistor is probably worth the flexibility and increased accuracy.

Debatable whether you even need the zener.  The opamp's PSRR shouldn't be a limiting factor, so it doesn't need a regulated or stable voltage.  It's not driving any current, so it shouldn't need much supply current.  With a so-called 'micropower' opamp, you should be able to use a 10:1 resistive divider or something similar.

[incf]

Debatable whether you even need the zener.  The opamp's PSRR shouldn't be a limiting factor, so it doesn't need a regulated or stable voltage.  It's not driving any current, so it shouldn't need much supply current.  With a so-called 'micropower' opamp, you should be able to use a 10:1 resistive divider or something similar.

I have to admit that the thought of supplying a >1k impedance supply makes me hesitate a bit. (eg. 10k and 1k across 32V for power dissipation [real values would be selected for voltage in addition to power dissipation])

I don't regularly do PSRR calculations

Supply voltage is 12V to 32V
The circuit I had planned would produce 0 to 3V across 100k on the low side. (30uA)
A chopper amp like the 5.5V rated TSZ121 consumes 40uA "max" (30uA typ)
I suppose that gives 30 to 70mV of drop which "seems like" nothing given the 80dB PSRR.

With such a wide operating voltage range, I probably would need a zener. I think the output voltage needs to get above the threshold voltage of the current mirror FET (maybe a BJT based current mirror circuit might tolerate low voltage operation better)

[KE5FX]

Yeah, a supply voltage range near 3:1 does make the zener more attractive.  I missed that part earlier.  Could likely still be done (some opamps are very stingy with their power consumption) but the upside may not be there.

[incf]

Suppose the supply of an op amp goes from 5.5V down to 2V (ie. 32V down to 12V through a divider) at DC, or maybe over the course of 10 minutes.

At DC can things like PSRR or CMRR increase the offset voltage beyond the stated max?


Do PSRR specs apply to those sorts of DC changes? Or is it a strictly AC specification?
Same question for CMRR?
(Supply voltage rejection ratio vs PSRR?)

I "feel like" PSRR is more of an "AC thing" (aka. >1Hz - but maybe I've been spending too much time reading datasheets for chopper op amps)

It probably depends on the type of op-amp (ie. chopper amplifiers might do PSRR specs differently than general purpose amps)

edit: I need to reread TI's "Op-Amps for Everyone"

[Zero999]

Suppose the supply of an op amp goes from 5.5V down to 2V (ie. 32V down to 12V through a divider) at DC, or maybe over the course of 10 minutes.

At DC can things like PSRR or CMRR increase the offset voltage beyond the stated max?

The power supply rejection ratio is 90dB at 10Hz, which means any change in power supply voltage is divided by 10^90/20 = 32k.

I repeat, you really need to test this in real life.

[David Hess]

Suppose the supply of an op amp goes from 5.5V down to 2V (ie. 32V down to 12V through a divider) at DC, or maybe over the course of 10 minutes.

At DC can things like PSRR or CMRR increase the offset voltage beyond the stated max?

Yes, because the input offset voltage specification applies under specific conditions, including power.  If you rely on the +/-15 volt specification but operate in a single supply application with a common mode voltage close to one rail, then the CMRR adds to the input offset voltage.

Do PSRR specs apply to those sorts of DC changes? Or is it a strictly AC specification?
Same question for CMRR?
(Supply voltage rejection ratio vs PSRR?)

CMRR and PSRR apply at DC and AC, and decrease with increasing frequency.  However if the graph shows the same curve versus frequency for CMRR, PSRR, or Aol (open loop gain), then they were measured incorrectly.  A lot of old datasheets have this problem.  For instance the CMRR curve should be much better than the Aol curve.