What is your favorite most versatile op-amp?

[maxwell3e10]

I've been looking for a versatile op-amp that can work close to ideal in a wide range of circuits.
Here is my list of requirements:
Voltage supply +/- 15V
Input voltage offset < 1mV
Input bias current <1 nA
Input voltage noise < 10nV/sqrt(Hz)
GBW >10 MHz but <100 MHz, not prone to oscillations
Slew rate >20 V/usec
Unity gain stable
Output current >50 mA

Ideally available in PDIP-8 package and as a dual op-amp. So far the best I found is OPA2132. What are other good candidates?

[OM222O]

can I ask why you care about the output current? if you need high current, just use a fet or a bjt as a buffer 
I know this doesn't meet your requirements but 99% of my projects are low voltage and single supply (0 to 5v), so I always choose rail to rail op amps.
my go to is the MCP600x (x can be 1, 2 or 4) which seems to be extremely stable, it has never had stability issues in my experience where other amps have (I remember a particular application where the TLV9002, OPA2333 and a few more had issues with oscillations but the MCP6002 worked like a charm). I know you should design your products to not have stability problems and what I described earlier is "bad practice" but I mean, if it works it's good enough for me 
For higher precision applications I usually go with the MAX4238 which dave used in his uCurrent. I bought a few to play around with and test them just for fun, but they performed great and are my other go to for precision applications 

[maxwell3e10]

For single-supply applications the requirements are different, rail-to-rail is more important and easier to use a single transistor buffer. I tend to work on bipolar circuits, so rail-to-rail has not been an issue.

[Conrad Hoffman]

You didn't mention cost. I think that's expensive. I also like the AD845, but it's even more. For many things the audio opamps like the LME series work well, say LME49720, but I don't know if you can get the LME series anymore (Mouser seems to have 'em). They tend to have good output drive. Don't forget the Japanese parts like NJM4556 and similar. It just depends on what kinds of circuits you build; that's why there are so many choices. I once thought the good old LF411 could do nearly anything, but it's not that fast. Needless to say, I keep more than one part type around!

[exmadscientist]

OPA192/2192/4192 will do for just about any "jellybean" op-amp application. It meets all of your desired specifications and then some, except for not coming in DIP-8. (Something I, personally, could not care less about.) Its downsides are that it's on the pricey side, it's got a bit more input capacitance than you sometimes want (I think due to the special sauce in the offset trim? Might be worth reading the relevant patents here...), and it has the usual RRI offset voltage wobble located at input CM voltages around a couple volts below the positive supply (though it's a small wobble as far as these things go).

The cost is no problem for prototype and lab work, but means that the OPA192 is not quite a "production jellybean". However, TI does offer a pretty wide set of cut-down versions that trade off one or more characteristics like offset or speed in exchange for cost, so you can compromise where it doesn't matter.

[maxwell3e10]

OPAx192 looks good indeed. Its $3-4 on Arrow, that's not bad. Its inevitable that one has to stock-up on DIP adapter boards these days, most top-performance chips don't come in DIP packages.

[T3sl4co1l]

Since you did not specify proximity to rails or cost, I would recommend looking for something LT (now ADI) or OPA (Burr-Brown, now TI), those specs are easily met by parametric search.

As for "favorite most versatile amp", that seems like a very misleading subject.  It also has the problem of the double superlative -- how could I even have a "favorite" and "most versatile" that's also an amp?  Amps don't work that way!  What combination of attributes could such a beast even have?  Very high GBW at very low supply consumption, say?  But even then, high GBW is a liability, so there must be some choice there.  And the laws of physics constrain the value of GBW * supply consumption, for a given design, so you can't expect anything special there, either.

Nah.  None of that.  You simply use the part that suits the application!

Tim

[maxwell3e10]

If one drops the PDIP requirement, the other two chips that look interesting and  even better than OPA2192:
OPA2197- cheaper at the expense of higher voltage offset
OPA2189 - zero drift amplifier with good dynamic response and also cheap
OPA2156 - high slew rate and low input noise.

As for "favorite most versatile amp", that seems like a very misleading subject.  It also has the problem of the double superlative -- how could I even have a "favorite" and "most versatile" that's also an amp?  Amps don't work that way!  What combination of attributes could such a beast even have?  Very high GBW at very low supply consumption, say?  But even then, high GBW is a liability, so there must be some choice there.  And the laws of physics constrain the value of GBW * supply consumption, for a given design, so you can't expect anything special there, either.

Nah.  None of that.  You simply use the part that suits the application!

I like as much as the next guy to scour the parametric search sites for just the right chip in a particular design. But sometimes its nice to pop in a chip without thinking about it and know that it will be more than sufficient .

[David Hess]

Parts like the Burr-Brown OPA627 and the Analog Devices equivalent meet all of your specifications except output current.  There are some other JFET input parts with similar performance although keep in mind JFET input bias current doubles every 10C so the input bias current requirement will be a problem at high temperatures.

That high output current requirement conflicts with the precision requirements because of the temperature gradients high power operation creates.  The only way to meet all of the specifications is with a composite design with a power booster separated from the input stage.

[maxwell3e10]

These are not the requirements for one specific design, more like an amalgam of reasonable requirements one can expect to get across the board- a versatile chip one can pop in many circuits.

[David Hess]

These are not the requirements for one specific design, more like an amalgam of reasonable requirements one can expect to get across the board- a versatile chip one can pop in many circuits.

The simultaneous precision and output current requirements conflict to make one or the other unreasonable. 

The supply voltage and output current requirements also conflict with the DIP-8 packaging requirement; parts with that kind of output current are limited to lower supply voltage, typically about +/- 8 volts.  (1) The datasheet for the LT1010 power buffer has a good discussion about the power limitations of the DIP-8 package.  Another place to find discussion of this issue is in the datasheets for +/-15 volt current feedback amplifiers.

The LT1010 works well as the output stage of a composite amplifier but even it would not meet your output power requirements without help unless used in the TO-220 package and perhaps not even then.  There are some current feedback amplifiers with 50 milliamp output current but they have the same problem.

(1) This comes up when driving a double terminated 50 ohm transmission line which presents a 100 ohm load.  +/-5 volts into 100 ohms is +/-50 milliamps.  Adding another 3 volts for headroom results in the +/-8 volt supply requirement.

[OM222O]

David Hess is right, that was also the first thing that I pointed out  just using a buffer for the power stage seems like the logical option as OP AMPs aren't really meant to drive any loads.

[maxwell3e10]

And yet there are old popular chips, like LM6171 in a PDIP package, and new designs like OPA2156, which have 100mA+  current capability.

Of course there will be power dissipation limitations, but often one doesn't need to drive high current continuously. And maybe literally the same chip can't be used to drive high current and have low voltage offset, but its nice to have a circuit that uses the same type of chip both for the input and output stage.

The basic question is to find an op-amp that can cover as much parameter space volume as possible without sacrificing too much in any other characteristics. 

[David Hess]

And yet there are old popular chips, like LM6171 in a PDIP package, and new designs like OPA2156, which have 100mA+  current capability.

The LM6171 is a current feedback amplifier in disguise with a buffered inverting input.  This is a great topology for high speed but not a precision one.

The OPA2156 is typical of misleading TI datasheets.  The precision specifications only apply with a high impedance load and low power dissipation.  No settling time specifications is given with a high output current because it is 1000s of times worse than the datasheet specification as it would be with any integrated operational amplifier.  So you can use it for high output current or high precision but not both at the same time.  With that high noise, it is not suitable for low frequency precision applications anyway but that is typical of CMOS processes.

The above may seem like nit-picking but consider the lessor known advantage of super-beta input parts like the old 308 and "modern" LT1008/LT1012/LT1097.  Their very low input bias current means that a high impedance feedback network can be used minimizing power dissipation in the output stage yielding better precision.  Lower noise parts like the OP07/LT1001 or OP27/LT1027 require much lower impedance feedback networks to avoid compromising their lower noise but this requires higher power dissipation in their output stage compromising precision.  In this respect, the "universal" part might be the LT6010/LT6011/LT6012 which is a rail-to-rail output version of the LT1012.  Note that no such single supply or rail-to-rail part like this can exist because single supply operation conflicts with input bias current cancellation.  Of course FET inputs do not require input bias current cancellation but they have their own problems.

Also notice that the OPA2156 specifications exclude input capacitance versus common mode voltage and do not mention source impedance in non-inverting applications.  That is deliberate.  In any real circuit, the non-zero source impedance interacts with the variable input capacitance to produce much higher distortion than they show.  There are ways to fix this but I am not sure that they apply to a rail-to-rail input part or standard CMOS process.

The OPA2156 is a great part for general purpose use but I would still prefer it to be available in a single version to take advantage of circuits which require access to the power pins.  That is a requirement I would add for any part to be most versatile.

[maxwell3e10]

Many of these precision parts, like LT1012 and LT6012 have terrible slew rates, so while they maybe good for precision DC applications, they are not good for any dynamic application. A good general chip will not be ideal for any application (not lowest noise or highest slew rate or lowest distortion), but will be pretty good for all of them.

What do you think about OPA2189? It seems to have better specs than any other chopper op-amp I have seen?

[David Hess]

Many of these precision parts, like LT1012 and LT6012 have terrible slew rates, so while they maybe good for precision DC applications, they are not good for any dynamic application. A good general chip will not be ideal for any application (not lowest noise or highest slew rate or lowest distortion), but will be pretty good for all of them.

That is a common theme which applies across all parts.  The emitter or source degeneration of the input differential stage has to be minimized to provide high precision.  But this results in high transconductance which ultimately limits slew rate.  JFET and CMOS transistors inherently have lower transconductance so result in higher slew rates but lower precision.

This really shows up with "video" operational amplifiers which add resistive degeneration raising the slew rate but also lowering precision.  An early example of this is the 318 which has a 15 MHz gain-bandwidth product and 70 volt/microsecond slew rate.  These same parts have higher current output stages to drive 100 or 150 ohm loads but with their lower precision, this is acceptable.

The LM6171 uses a different topology which does not suffer from this limitation and is much faster but it requires a complementary process and matching between complementary devices compromises precision.  I have designed audio amplifiers with a similar topology.

What do you think about OPA2189? It seems to have better specs than any other chopper op-amp I have seen?

Its specifications are better than any I have used.

Chopper stabilized operational amplifiers can be tricky to apply however.  Watch out for its input current noise which looks to be 75 times greater than an old "noisy" LTC1151.  That is consistent with larger area input transistors for lower voltage noise.  Chopper stabilized amplifiers should be used with low AC impedances limiting their versatility; they are not general purpose parts.

Pretty much if you do not have a big feedback capacitor and a big capacitor from the non-inverting input to ground, then you are using the chopper stabilized amplifier wrong.

[maxwell3e10]

I tested OPA2189 dynamic characteristics at high frequency. It works pretty well, with a slew rate of 20V/usec and clean response. See below a comparison of 3 op-amps, each driven by a 16Vpp square wave. The frequency of the wave is 500 kHz (330 kHz for OPA2132).

I was looking for some weird effects associated with the chopper modulation. What I found is that for a non-inverting gain of 100 if the frequency of the sine is exactly 67.5 kHz, the gain has a periodic modulation. This only happens in about 1 kHz range, must be some interference effect with the chopper frequency.

Also OPA2189 is sensitive to high-impedance path on the input due to current injection, as others have also pointed out. This can lead to rather large offsets that depend on the capacitance, so one has to be careful to realize uV stability. When that is not crucial, then OPA2197 looked pretty nice and also inexpensive.

[Gyro]

This is the second of these 'best op-amp for everything' threads in the past couple of weeks (https://www.eevblog.com/forum/chat/favorite-op-amp/msg2342253/#msg2342253). I don't get it - surely the op-amp you want is the one that's best suited to the specific application. If you try a 'one size fits all' approach then there are going to be massive compromises in all but the most mundane application.

You need to look closely enough at the datasheets to spot hidden 'undesirable features' is tricky enough as it is. Maybe a what's your favourite high impedance, DC precision, fast slew rate, high output drive? might have some value, but even then, there are all sorts of requirement combinations that would lead to a non-optimal match.

A parametric search is the only way to do it properly. That's what I would advise the OP to do for each of his applications.

Trying to tie down to one 'favourite' is just silly.

.... It would ultimately save the manufacturers the effort of producing many different parts though.

[MarkF]

Maybe the title should read "Op-amp Short List"

It would be nice to have a list of a dozen op-amps to pick from for general use.
There are soo many available that it's overwhelming.

[Gyro]

Agreed, that would be of more value, particularly if done by category.

[exe]

A parametric search is the only way to do it properly.

I'm yet to see a good parametric search, not to mention across multiple vendors (I don't qualify digikey as good search, but it's a good starting point). Also, many things are not mentioned in datasheets, that's why I'm always interested in more data, measurements and experience about opamps.

[H713]

Clearly anything less than an original API 2520 will not do.

All kidding aside, from an audio standpoint, the good old 5532 is difficult to beat without paying 5-6 times as much, and it is very low noise.

[Kleinstein]

A parametric search is the only way to do it properly.

I'm yet to see a good parametric search, not to mention across multiple vendors (I don't qualify digikey as good search, but it's a good starting point). Also, many things are not mentioned in datasheets, that's why I'm always interested in more data, measurements and experience about opamps.

The parametric search at Ti and AD is not that bad, once one gets used to it. For high end OPs there are not that many more sources to search for.

Sometimes it is also the bad sides of some OPs, that a worth to remember. So something like well known cross over distortion of the LM358/LM324,  phase reversal when leaving the common mode range or the TLC272, that does not like the input near ground and the output near the negative supply at the same time.  I just found out that with the TLE2071 the supply current goes up quite a bit if the output is at the negative saturation  .

[Zero999]

Clearly anything less than an original API 2520 will not do.

All kidding aside, from an audio standpoint, the good old 5532 is difficult to beat without paying 5-6 times as much, and it is very low noise.

Some of the parts from New Japan Radio look good, at least on paper, as I haven't tried them.  NJM2068, a general purpose but better replacement for the NE5532 and the NJM2122 which is lower noise, but lower voltage.

[magic]

NJM2068 beats the 5332 on noise, price and power consumption but it's said to be less linear at high frequencies and high gains or with less than a few kΩ load impedance. I think the general consensus is that nothing beats 5534/5532 in every respect at a comparable price. Uncompensated 5534 also has 3x more gain at audio frequencies, 5532 was some 90% of what NJM2068 spec claims. I could swear I've measured NJM2068 too but I can't find the results now.
But I have seen those NJM2068 used in some semi-pro soundcards and also stuck them into a Xonar STX in place of the original NJM2114 (JRC's own "improved" 5532) and got barely any difference in RMAA results. I guess that means good enough for undemanding applications.