A discussion on Precision OpAmps

[Ole]

Hiho,
since I am planning a project that requires some pretty precise OpAmps, I have been on the lookout for them.
The OpAmp that I´m mostly going to use (as of now) is the LTC2057, as the xDevs Article on the KX-Board mentions them
and they do have decently low noise statistics.
The reason I am interested in the LTC2057 is it´s capability to utilize a +/-15V supply safely.

But my little project is not the only reason why I want to start this topic. It´s also because I could not find another similar thread.
So if anyone has a recomendation or has noticed that my choices are "suboptimal" please tell me.

Cheers
Ole

[Andreas]

Hello,

I also use often the LTC2057 but also the ADA4522-1.

Depends on what you are doing.

What I have observed with the LTC2057 is that the chopper frequency may have a interference with my ADC (LTC2400) + 2:1 Divider (LTC1043 based ~400Hz).
At certain temperatures I get excessive noise at the ADC output.

Ok with a LTC1050 it is even worse.

The ADA4522 I did not test up today. But since it has internally 2 different chopper frequencies it is likely that there is less interference with my ADCs.

with best regards

Andreas

[Kleinstein]

With AZ OPs one has to be aware of possible interference from spikes in the input current. This can effect ADCs or in some cases other OPs.
The supply should also have good decoupling and ideally filtering as they are possible EMI sources.

Also take care if the source is high impedance, as the AZ OPs can have quite some current noise and the data for the current noise are not all valid. The LTC2057 and ADA4522 data seem to be valid, but others like AD8628 and MCP6V51 are too optimistic.

With a +-15 V supply the power consumption and thus heating can become an issue if highest precision is needed.

[MiDi]

Overview of modern AZ op-amps for +-15V supply:
LTC2057 (1 Ch)
LTC2058 (2 Ch)
ADA4522 (1, 2, 4 Ch)
ADA4523 (1 Ch)
OPA189 (1, 2, 4 Ch, MUX-Friendly)
MCP6V51  (1 Ch)



Edit: Added MCP6V51 & corrected ADA4522 current noise

[Kleinstein]

A nice table.

If cost and availabiltiy is an issue, the MCP6V51 would be an option too: somewhat comparable voltage noise to the LTC2057, but lower power. The DS value for the current noise is wrong (more like the shot noise from the input current). My crude measurement of the noise current is at around 250 fA/sqrt(Hz) and this somewhat higher than for the LTC2057, but still OK.
The other problem with the Microchip data-sheets is the slightly odd definition of the bias current as the average including the sign. So the the comparable value would bias plus 1/2 the offset current. So expect more like 160 pA typ and 750 pA max input bias with the normal definition.

The ADA4522 current noise specs are 800 fA/sqrt(Hz) , not 80.

[KT88]

The precision that an opamp can provide depends a lot on the application and the purpose it serves in it:
Are you interested in time domain or frequency domain signals?
AC or DC?
Bandwidth?
The frequency range.
Source characteristincs - high or low impedance, noise...
Level of precision?
Load? ADC, cable...
Temperature range?
Power constraints?
Budget?
....

Cheers

Andreas

[Ole]

First of all, thanks for the input.

Second, KT88, I am more interested in the DC side or up to maybe 1kHz. The Load that the OP´s would need to drive are 5kOhm Resistors at worst and another OP at unity gain (Av=1) at best.
I intend to operate the entire instrument at an internal temperature of ~35-40°C. As for Power constraints, the instrument is not intended to run via battery power, so a very low power consumption is not essential.

Generally the lower the ouput voltage noise is the better for my usecase.

Cheers
Ole

[macaba]

I think we'd have to see the circuit to give the best recommendations. For example, there are some circuits where AZ isn't ideal and something like OPA140 is better.

[KT88]

@Ole:
Still the source impedance, input signal level --> gain, and accuracy (16bit?) are required...

Cheers
Andreas

[Ole]

source impedance, input signal level --> gain, and accuracy (16bit?)

Well as you may or may not have guessed I intent to build a multimeter,
so the source impedance can vary between near zero, if I was to measure a Power Supply
and near infinite, if I was to measure a specific resistor in a device.
But only the OP´s directly on the frontend are in charge of the very high impedance

The maximum input signal lever should be a 10V difference between two cooperating OP´s
As for the gain and accuracy, the gain will most likely not exceed 1000 and the accuracy should be as high as possible.

Also I´m reworking the circuit so that other people can understand it without headaches (hopefully)

Cheers
 Ole

[David Hess]

source impedance, input signal level --> gain, and accuracy (16bit?)

Well as you may or may not have guessed I intent to build a multimeter,
so the source impedance can vary between near zero, if I was to measure a Power Supply
and near infinite, if I was to measure a specific resistor in a device.
But only the OP´s directly on the frontend are in charge of the very high impedance

Multimeters and other high input impedance instruments do not use chopper stabilized or automatic zero operational amplifiers because of the high current noise; they are simply not suitable for high input impedance applications.

The usual solution is to use a voltage follower, and then *if necessary* correct the offset of the voltage follower as part of the integration cycle, so essentially make an automatic zero amplifier that operates at a low frequency synchronously with the conversion cycle.

Additionally, if the offset is corrected at the input voltage, then the common mode rejection is corrected as well.  This is why chopper stabilized and automatic zero operational amplifiers have high common mode rejection, and why CMOS voltage followers which have atrocious common mode rejection can be used with automatic zeroing.

I think we'd have to see the circuit to give the best recommendations. For example, there are some circuits where AZ isn't ideal and something like OPA140 is better.

The OPA140 would be my first choice.  It has the right combination of input bias current, offset drift, noise, and common mode rejection ratio.  In the past the LT1012 low input bias current bipolar part might have been used in the highest precision applications.

[macaba]

Agree with David - if talking about DMM ADC input buffer; OPA140 for up to 6.5 digits, bootstrapped (to follow input signal) OPA140 for 7.5 digit upwards. It is a wonderful device for lowering part count here.

[Kleinstein]

There are DMMs that use a zero drift ADC at the input (e.g. Keithley 2000, 2002, likely DMM6500, Sigilent SDM30x5, Prema5000), thought often no on its own. The input bias current is a problem and thus the choice of amplifier is somewhat tricky and may need selection. The maximum limits for the input bias are often not good enough, but typical or selected slightly better one can be OK.
Even though the input impedance of a DMM is high, they are not really made for high impedance sources. So the >10 Gohms input impedance is more like made to get less than 1 ppm error for a 10 Kohms source.

Using a JFET amplifier (like OPA140 as a modern chip) and seprate switching the input is an option, but also not that easy. There are similar or possibly even worse switching spikes from doing the choper type offset correction under software control.

The choice of amplifier also depends on the ADC used.

[SoundTech-LG]

David,

Would LT1008 also have been a good choice in the past, as the main difference (to LT1012) seems to be internal/external compensation?

Thanks!

Martin

[Kleinstein]

There were a few DMMs (e.g. Datron106x, Fluke 8505 and related) with a BJT based input amplifier, though more like discrete transistor pair.
The LT1008 / LT1012 would perform somewhat similar. For todays standards the bias current is a bit on the high side, but still not way off.

[Ole]

The choice of amplifier also depends on the ADC used

One of the two ADC´s is planned to be a AD7177, and the other being a ADS1235-Q1.
Yes, the multimeter has a role model or two.
I do intend to perform HI and LO sensing, so for each function I would need two* OP´s with a very low bias current if possible.
(The exeption is the low current ranges where I intend to use a inverting I/V converter)

The HPM7177 uses OPA189 for the input stage, most likely for the very low noise I/O,
so those might be worth remembering for the internal amplifiers.

[KT88]

except in low-cost applications it is never a good idea to use the input amplifier as the ADC driver. If one is already concerned about the current noise a zero-drift amplifier produces, the noise injected by the ADC is orders of manitude higher than that - considering the internal buffers are bypassed for accuracy reasons.
An opamp output appears as an inductor to a signal (or noise-) source attached to the output. That means the noise will be pushed into the signal source only slightly attanuated at best.
The only proper way to mitigate these issues is to have a dedicated ADC driver and appropriate filtering in between the amps.
This problem gets even worse if the input amplifier has a gain (much) greater than A=1.

[Kleinstein]

The SD ADC chips have a differential input and a range of some +-2.5 to +-5 V.  To use the full range and get best INL one would need to drive the input differentially, so more than just a simple amplifier. For a 10 V range one would also need some attenuation, not just gain

[RoGeorge]

MT-054: Precision Op Amps - Analog Devices
MT-055: Chopper Stabilized (Auto-Zero) Precision Op Amps

[Ole]

I´ve looked at the input amplifiers of a few other DMM´s, and found a few interesting OP´s that could be used:
AD795 (from the HM8012, 1-2pA Ib but -100µV Vos and 1-3µVpp Noise)
AD706 (from the 34401A, 50-200pA Ib and +30µV Vos and 0.5µVpp Noise)

An Idea that I had was to attach a Peltier Element to the Input Amps and alter their temperature to the point, that their offset voltage is nearly zero.
The primary problem with that, besides the thermal EMF caused by the leads, would be that the offset drift and the offset are rarely of the same polarity.

Edit:
Additionally I had a look at the LT6018, which is nearly pefect (0.03µVpp Noise and +-8µV Vos) exept for one thing: 40nA typ Ib.
I might Edit this post again if I find some other interesting infos.

[Kleinstein]

For an DMM input amplifier (unless using an old chip like ICL7106) one would not care about the absolute offset voltage, as this can be subtracted in software anyway. So critical parameters are more like the input bias current (usuallaly < 50 pA are aimed for, ideally < 10 pA) and the voltage dirft.

The HP34401 (old version) uses an AD706 in the input amplifier, but this is only for some current sources and the main amplifier is still old style with a dual JFET. The drift compensation is with switching at the input with most of it inside the Hybrid. Today one could consider an OPA140 or similar, though there the CMRR and linearity is limited.

The OPs like AD706 may need an extra adjustment for the bias current and the offset drift can still be borderline.

[KT88]

You will run into a trade-off at some point of time...
Bipolar amps offer a very low voltage noise but pretty high Ib - they are only viable for low impedance sources.
JFET (and MOS-FET-) amps offer excellent Ib and current noise.
AZ-amps - there are many different architectures - offer excellent voltage noise without 1/f noise and low voltage TC at the cost of moderate current noise.
A way to extend the usability you could roll your own AZ scheme - it would allow to sync the frequency with your ADC and also reduce the AZ frequency which would reduce the average charge injection into your source.
I would look into an AZ amp like the ADA4522 with some RC filtering to smooth the noise from the input as well as BW limiting the input signal. As AZ amps have some sampling happening, too high frequencies will cause some interference (Intermodulation distortion, IMD) https://www.analog.com/en/analog-dialogue/articles/demystifying-auto-zero-amplifiers-part-1.html
Bandwidth limiting the output is recommended as well.
Also: the longer you integrate the signal the better SNR gets (oversampling, process gain).

[RoGeorge]

...critical parameters are more like the input bias current (usuallaly < 50 pA are aimed for, ideally < 10 pA)

In terms of bias current, some are specified for the fA range. 
LMP7721 3-Femtoampere Input Bias Current Precision Amplifier

[Ole]

some are specified for the fA range

So similar to the ADA4530 and the LMC6044.

Well, I´ve looked into the datasheet of the LMP7721 and it might be the right canidate for the input,
though I do need to find a way to reduce the LF noise on the output, or increase the integration time or sample number.

Cheers
Ole

[KT88]

@Ole: What is the max. input impedance you expect and what voltage noise level do you need referred to the input?