ADI ADA4620-x joins the party of OPAx140 for precision JFET Op-Amp

[laichh]

Finally, ADI joins the party of OPAx140 for precision JFET Op-Amp with its' new ADA4620-x.

https://www.analog.com/en/products/ada4620-1.html

Overall, the ADA4620-x is quite competitive to OPAx140 except the common-mode input voltage range of Vcm < V+ - 4.4 V!

[iMo]

There is a wiring in "Figure 123. Squeezing a High Voltage High Impedance Signal into a 5 V Precision Differential Input ADC" people often discuss here..

[David Hess]

Overall, the ADA4620-x is quite competitive to OPAx140 except the common-mode input voltage range of Vcm < V+ - 4.4 V!

The OPA140 is Vcm < V+ - 3.5 volts, so only 0.9 volts worse.  Analog Devices gives a separate set of worse specifications for Vcm > V+ - 4.4 volts.

[macaba]

This is a decent 2nd source for this class of opamp, and kudos for having defined behaviour in that (V+)-4.4V region, as OPA140 can do strange things in the (V+)-3.5V region. Adding this one to my "to do" list.

[Kleinstein]

The upper part is quite a bit worse and there can also be quite some step in the transition between the normal and high CM part. I would consider the high voltage part more like limited use to avoid a possible phase reversal. Nomal operation should avoid that range. The same is true for most other RR amplifiers (except maybe some AZ types).

ADI already had a good low noise precision JFET amplifier for quite some time, the ADA4625. A downside is the rather high supply current.
Here the new type seems to be really good.

[laichh]

The upper part is quite a bit worse and there can also be quite some step in the transition between the normal and high CM part. I would consider the high voltage part more like limited use to avoid a possible phase reversal. Nomal operation should avoid that range. The same is true for most other RR amplifiers (except maybe some AZ types).

ADI already had a good low noise precision JFET amplifier for quite some time, the ADA4625. A downside is the rather high supply current.
Here the new type seems to be really good.

The ADA4625-1 has an input bias current of 15 pA (typ) & 75 pA (max), so not really a precision Op-Amp to me. A good metrology grade DMM like 3458A has a 20 pA of maximum leakage current at 25 °C.

[Simmed]

the offset deviating from zero
how bad is this ?
x140 doesnt exactly have a comparison chart

[Kleinstein]

The offset vs output current is likely an effect of heating from the power loss in the OP-amp. For this reason precision amplifier should not drive much current to keep the power consumption constant. It is not only the mean temperature, but already temperature gradients on the chip that can effect of offset.

For the comparison there are different properties that may be relevant for an OP-amp. Sometimes a pA of input current makes a difference and often not. There can be other factors too, like the CM range, linearity (with a low power amplifier the output cross over can be an issue), EMI sensitity, input capacitance, voltage gain (here the ADA4620 looks good with light load, but not so much with more load), drift over time and temperature,....
Some of the points may not even be in the datasheet.
So there is rarely a true 1:1 replacement. What amplifier fits depends on the circuit.

[David Hess]

ADI already had a good low noise precision JFET amplifier for quite some time, the ADA4625. A downside is the rather high supply current.

ADA4625 is not in the same class as the OPA140 or ADA4620 for input bias current and common mode rejection, which are useful for precision non-inverting buffers.

If lower current is important, and as Kleinstein points out above it can matter for precision loss from self heating, then the OPA145 is a low current (0.45mA 330nVpp 5.5MHz) version of the OPA140 (1.8mA 230nVpp 11MHz).

[Alex Nikitin]

The offset vs output current is likely an effect of heating from the power loss in the OP-amp.

No, I think it is a simple result of a limited gain, the graph is cheating a bit, showing the same polarity change for both polarities of the output current, the effect is almost linear showing the output resistance of a few mOhm order, if you look at the open loop gain graph it it shows ~120dB into 1kOhm, which for a follower configuration would give ~1mOhm output resistance. As the output voltage is fixed at 0V, the input (=offset) voltage must change... .

Cheers

Alex

[magic]

Unity gain closed loop output impedance is supposed to be 0.65mΩ, which would correspond to a linear 26μV offset voltage change over 40mA change in output current.

So this readily explains the difference between the left and the right end of the plot, but the symmetric rise away from zero may be thermal.

[Echo88]

At a quick glance it looks quite good with some improved specs compared to the already nice OPA140.
Lets see how much theyll charge for it.
Cant make sense of the fact that the dual version apparently has a better CMRR-spec than the single version, perhaps thermal symmetry influence?

[Kleinstein]

The typical data a based on measured samples. The layout for the single and dual version can be slightly different and this fine details can make a difference at the details. This can be thermal and other parasitics.  I have seen better data for the dual version also with TL032/TL031.

[David Hess]

The offset vs output current is likely an effect of heating from the power loss in the OP-amp.

No, I think it is a simple result of a limited gain, the graph is cheating a bit, showing the same polarity change for both polarities of the output current, the effect is almost linear showing the output resistance of a few mOhm order, if you look at the open loop gain graph it it shows ~120dB into 1kOhm, which for a follower configuration would give ~1mOhm output resistance. As the output voltage is fixed at 0V, the input (=offset) voltage must change... .

The open loop gain calculated from the transconductance and load resistance of each stage, and the output resistance and load resistance, comes up short because ultimately gain is limited by thermal feedback from the output transistors to the input transistors.

This is why supporting high output current is not important in a precision operational amplifier, as it cannot be taken advantage of anyway without sacrificing precision, and the highest precision parts are at the lower end of the power spectrum.  Often a lower power part delivers better performance despite a small comprise in written specifications.

The situation is considerably improved by adding an external buffer to unload the output of the precision operational amplifier.  Past designs sometimes used a discrete matched input pair for the same reason.  For a long time hybrid operational amplifiers had an edge in precision over monolithic parts.

[magic]

This looks like a gradient issue. It's too large for open loop gain transconductance alone, but open loop transconductance limitation just happens to explain the asymmetric part of it. It's also too large for simple heating: 18V·20mA = 0.36W, 0.36W·115°C/W = 41°C, so this unit would slightly exceed the worst case limit of 1μV/°C.

So now idea how TI parts might compare without testing.

[iMo]

At least the model is there.. The AFE from the DS - it works somehow in the -10..10V input range, but the -12..12V inp range with the expected issue beyond +10.5V inp, and a pretty large currents off the opamps, though..
Frankly, it would require a better schematics - even for the 20bit ADC as depicted in the DS..

[laichh]

At least the model is there.. The AFE from the DS - it works somehow in the -10..10V input range, but the -12..12V inp range with the expected issue beyond +10.5V inp, and a pretty large currents off the opamps, though..
Frankly, it would require a better schematics - even for the 20bit ADC as depicted in the DS..

The proposed SE-to-Diff circuit in Figure 123 has too much resistive Johnson noise, maybe 20-bit is the max practical resolution, or slightly short of 6½ digits. A THS2630 + RES11A20 (2 kΩ : 1 kΩ, 2x in parallel to half the resistance) could be better.

[iMo]

This is a simplest idea, imho, which may work with a 20bit ADC, best when the first ADA4620 will be bootstrapped (because the CM).
The dividers have to be of a larger resistance - not to load the 4620 too much..
A couple of kOhm at their outputs would heat up the opamps significantly (their temperature will vary based on their output voltage/current).

[macaba]

This is a simplest idea

What is the AC CMRR like?

[Kleinstein]

For the invering signal path one would not really need a separate inverter, divider and buffer. One could direcectly habe the inverter with attenuation.  This would mean 1 amplifier less in the path.

With point of the good precision of the ADA4620 is that it has good CMRR and thus not so much need for a bootstrapped supply. If a bootstrapped supply is used, one could as well consider an amplifier that is made for low voltage like 5 V. With the rather large required headroom (e.g. 4.4 V) to the positive supply the ADA4620 is not that suitable for 5 V supply.
The DS has a curve for measured CMRR as a function for freqency. It looks rather good, but still not so clear how good other units are.