Front-End Voltage Limiter for Instrumentation Amplifier

[foetusmachine]

I've been racking my brain trying to limit the input voltage to an instrumentation amplifier as part of a millivolt meter project.

An "over-voltage" occurs when I don't have a DUT in place and the kelvin clamps are sitting doing nothing. Essentially the 3.3v, 100mA current source I have connects directly to the InAmp inputs, which I would like to avoid, given I have a x10 gain set. To be clear, i'm dealing with a DC input here only.

I have tried all matter of diode clamping configurations, but given the low voltages I'm dealing with, I struggle to get anything to work with a degree of accuracy and a definitive cutoff.

Finally, I resolved a solution that utilizes a single, quad op-amp that can be operated off a single rail. Thanks to a couple of online resources - Analog Devices note AN-581 and https://ultimateelectronicsbook.com/op-amp-voltage-reference/



I still need to integrate into my broader circuit, but I'd love to get some feedback on the circuit.

You might not be able to see this in the ouput plot, but there is a ~3.12mV offset between the input and output at the very bottom end, which diminishes virtually nothing at 250mV. Any thoughts on how I might reduce this offset?

[Kleinstein]

Most amplifiers are OK with a slightly higher voltage. So chances are that just 2 anti-parallel diodes (e.g. 1N4001) to ground could be good enough.

The circuit shown is an active clamp for one direction. It has the problem that most OP-amps can only deliver some 20 mA - so it could not stop the 100 mA current source. The filtering / buffers for the trigger point voltage are not really needed. Just a simple divider will do.

[foetusmachine]

Thank you @Kleinstein.

It's been remarkably difficult to get a definitive (and clear) answer on the input/output limitations of the LT1789-10 InAmp I'm proposing to use. It's a single sided supply rail and x10 gain, so any input over ~370mV concerned me as the output was always simulated as being clipped at ~3.7v. I never knew if this would damage or lock up the inamp. I'd be interested in your experience here.

Thank you for your suggestions. I simulated the 1N4001 antiparallel diodes on the input and it limits the voltages to 800mV.

In regard to the circuit I posted, I see what you mean about the buffering and how that can be removed and therefore a dual OP2177 could be used. But also you're right about the current. I simulated my current source and the simulation freaks-out - I think the current sinking ability of the OPx177 is being overloaded.

[Kleinstein]

The data sheet for the  LT1789-10 gives a current limit of +-20 mA for the inputs and the circuit shows clamping diodes to the supplies. The amplifier is perfectly fine with the output driven to saturation. So this is absolutely not a requirement.
The internal clamping diodes may have a lower forward votage than the external diodes. So one may need external clamping diode and than an extra series resistor (e.g. 1 K range) before the amplifier. So worst case that resistor sees some 0.5 V and thus 0.5 mA of current to the amplifier inputs. This well withing the +-20 mA limit.

With only a low supply I am not so sure the relatively expensive  LT1789-10 is the best choice. It depends on the application. I would have a look at the INA317 as a possible alternative.

[Terry Bites]

I'm wondering why the inputs need precise clamping?
If its to protect against open circuit causing the INA drifting to rail, an ideal diode is overkill.

Most, if not all instrumentation amps have internal clamps that limit the excursions to a 0.3V beyond the supply rails.
As Kleinstein points out, a pair of back-to-back didoes will limit the differential input voltage. You can use jfets as very low leakage diodes here.
Diode connecte bipolar transistors can provide low leakage too.
You can generally get away with a series resistance at the inputs to limit the clamp current. Check datasheets for max input current and design conservatively. 10mA is typical.
Sometimes the effect of the series resistor and Ib adversely affect the input offset and CMRR. The solution is to add external low leakage clamp diodes (or jfets, eg J309) and use a smaller series resistance. There is also the posibilty of checking the INA output with a comparator an implementing input over range prorection with analog switches.

Have a look at the precsion clamp circuit here: www.analog.com/en/technical-articles/op-amp-precision-positive-negative-clipper-using-lt6015-lt6016-lt6017.html You can set a differential clamp range down into the mV.

[magic]

It's most likely unnecessary, but the OP doesn't believe so because Linear didn't give a clear and unambiguous spec on permitted input voltage.

More details in OP's earlier thread about the same circuit...

[foetusmachine]

Another concern I have in mind is downstream of the InAmp.

It's feeding into a 0-200mV EDIT:0-2v panel meter. I have a couple of these on order for testing. I'm not sure of the protection these devices will have. One supplier suggested input should be limited to 10% over (i.e. 220mV EDIT: 2.2v).

With the antiparallel diodes limiting InAmp input voltage to 800mV, this is still 8v being clipped by the rails to ~3.7v.

Maybe it's fundamentalist or just over the top, but I just felt like if I was able to keep each part of the circuit in-check, it would limit potential issues.

[foetusmachine]

Another issue...

While the 1N4001 antiparallel diodes work great to get at clamping the input voltage to 800mV, they also ensure significant current is being drawn even when the DUT is not in place. Not great for a battery powered project.

Providing resistors to limit the current works, but it also means the voltage does not drop as much:

Eg:

1N4001 + 1k resistor = 2v input limit + 1.5mA lost.
1N4001 + 10k resistor = 2.3v input limit + 316nA lost.

[Kleinstein]

Most panel metes are OK with significant (like 5x or even 10x) the nominal current - however it depends on the exact meter.

For the current consumption one would have to turn of the drive side to really reduce the current consuption. This could be from the clamp part or the drive voltage.

[LM21]

Why do you short U1 output to ground? Or how are your grounds connected?

[foetusmachine]

Which circuit are you referring to for U1?

[Kleinstein]

The point is probably the very first ciruit. Here U1 creates a virtual ground, that can be confusing. The amplifier is not actually shorted out it is just the ground symbol used for the virtual ground.

[foetusmachine]

Sorry, I see which U1 you mean. U1 creates a virtual ground to allow the ther OpAmp to swing lower. Not required If I had a split rail power supply.

[foetusmachine]

Just suggesting I plough-ahead with an active clamping system, i've taken on some of the suggestions above and modified the circuit, now showing the current source:



Now clamping to the 240mV with a slight over-shoot (290mV worst case) at the front end:



1.7mA being sunk by the op amp when the DUT is open circuit.

[foetusmachine]

While this is working, I think I'll start a new thread to discuss the power supply arrangement.

I'm okay with the two batteries, but I'll need to ideally have both batteries supplying the entire circuit. Otherwise, the current source will deplete the V3 battery much faster than the V1 battery. Ignore the fact that I show 5v for now.

This then creates complexity around the grounding references across the circuit.

[ArdWar]

Considering that you're basically measuring a low impedance source. You may not even need an In Amp here and probably can work fine with a Diff Amp instead, and they almost always can take quite large voltage between their inputs.

You can then protect/limit the amp output instead with the usual resistor and a zener. Or maybe a TL431, it's almost perfect textbook application circuit for that.

[foetusmachine]

Considering that you're basically measuring a low impedance source. You may not even need an In Amp here and probably can work fine with a Diff Amp instead, and they almost always can take quite large voltage between their inputs.

You can then protect/limit the amp output instead with the usual resistor and a zener. Or maybe a TL431, it's almost perfect textbook application circuit for that.

Awesome suggestion. Thank you. I swapped out the InAmp and plugged in a diff amp. I haven't researched details much yet, but i just plugged in an LT6376, which seemed to be the cheapest in the list of Analog Devices that seem to fit the bill and it works great so far with much less additional circuitry.

I'll need to do more reading and trawl through the datasheets, but i'm going to give this a go.

[Terry Bites]

All you really need to stop an INA drifting off with no input is to just add input resistors to ground or a CMV or bias points of choice.
I assume this all accdemic now and active clamping is not a need but an interest?