What are some recommended IC opamps for DC voltage reference circuits?

[elecdonia]

I've been working with DC reference voltage circuits, mainly based on SZA263.  Soon I will be working with LM399 and 2DW23x devices too.

Now I want to get a supply of modern high-performance IC opamps to use in my reference circuits:

1)    What is a good choice when zero offset ( < 1 uV)  is required?  (My assumption would be chopper-stabilized)
2)    What is a good choice when near zero bias current is required for circuits that must have a very large input resistance?
3)    Is there a good solution that covers both 1 and 2 at the same time?

I'd also like to get recommendations for high-performance low TC trimmer potentiometers.

Thanks in advance!

[RandallMcRee]

It will be interesting to see what the more experienced folks have to say. I had a similar need and after looking at many opamps (just datasheets) ended up settling on the ADA4522.
http://www.analog.com/media/en/technical-documentation/data-sheets/ADA4522-1_4522-2_4522-4.pdf

An important criteria for me was offset voltage drift. Most applications can trim out or simply ignore offset. Also, I was using +-15v supplies, there are not many chopper amps other than ada4522 that have a wide supply range.

Anyway they seem quite well-behaved other than accidentally shorting a dual and having one of the channels die. Data sheet claimed it would be ok. My bad, not like I was actually relying on that feature, just carelessness.

[montemcguire]

I've also decided to use an ADA4522-2, in my case for a DC servo amplifier stage. It has a good combination of low offset voltage, low bias current, and low 1/f noise. To thoroughly optimize this, I made a spreadsheet of the various candidates, and then plugged in some of my circuit's component values to be able to sort on the chip's performance in my circuit. For example, knowing the circuit resistances allows me to combine the bias current and voltage offset specs to get a total offset value that can be more easily sorted. I then created a 'feature vector' scoring system to weigh the remaining features - specs like distortion, 1/f noise, cost, etc. against each other to determine a figure of merit that could also be used to sort the candidates.

Long and short, different circuits can have different requirements, so what's ideal for one use may not be ideal for another. So, while we both like the ADA4522-2, I think it's worth the effort to build a table of the basic specs and then incorporate your own figure of merit calculations to see which parts objectively perform best in your circuit.

[george.b]

I recently ordered some LTC2051s. Haven't done anything with them yet, as they have not yet arrived, but I'll use them in a milliohm meter. 0.5µV typical offset (3µV max), 10nV/K typical offset drift (30nV/K max), sounds pretty good to me.

Edit: link to the datasheet - http://cds.linear.com/docs/en/datasheet/20512fd.pdf

[Kleinstein]

Az OPs are an obvious choice if a very low offset and drift is needed. Also the input bias is really low - so they can be good choice for high impedance circuits too. One has a choice between very low noise (like the ADA4522 / LTC2057 / max4425x ) and low bias but more noise (e.g. ICL7650, LTC2050,MAX4238).

However with AZ OPs one has to be careful about the input current spikes. It depends on the input impedance how these current spikes behave - this more than just input bias and input current. So also capacitance at the input and supply decoupling can be important to even if one is interested in near DC performance only. With several AZ OP in one circuit, there is also a chance of interaction between them. Also current noise specs on AZ OPs are thus somewhat suspicious, as the bias is not just a continuous DC current, but more like a sequence of spikes.

The ADA4522 and a few others are special in that they include some RF filtering. This could reduce the influence of the current peak a little.

So if you don't need an AZ, it might be safer to use a good conventional OP, like OP277, OPA1177, LT1097, LT1012. Here it is compromise of voltage noise and bias / current noise.

[Conrad Hoffman]

Something to watch for is the output current capability of chopper amps. Some are surprisingly low, and won't drive low value feedback resistors, which you want for low noise.

[David Hess]

Even if you are using non-chopper precision amplifiers, it can be a good idea to unload the output with an emitter follower or buffer so that changes in the output stage power dissipation do not get back to the input transistors.

Besides offset and bias current, consider noise and especially 1/f noise where it is better to avoid adding it than to try and filter it.  Chopper amplifiers have the advantage of flat 1/f noise.

The OP-07 has become the jelly bean of precision operational amplifier simply because it is so cheap.

[elecdonia]

The OP-07 has become the jelly bean of precision operational amplifier simply because it is so cheap.

Yes I'm familiar with the OP-07 and have used them myself.   Although the OP-07 has been around for a long time it is still an excellent choice for for many applications.

When time permits I intend to create a spreadsheet of opamps that are useful for DC applications and will post it to this forum.  The OP-07 will definitely be on this list.  Low price, wide availability, and high performance are characteristics that will move a device up to the top of this list.

[David Hess]

The OP-07 has become the jelly bean of precision operational amplifier simply because it is so cheap.

Yes I'm familiar with the OP-07 and have used them myself.   Although the OP-07 has been around for a long time it is still an excellent choice for for many applications.

When time permits I intend to create a spreadsheet of opamps that are useful for DC applications and will post it to this forum.  The OP-07 will definitely be on this list.  Low price, wide availability, and high performance are characteristics that will move a device up to the top of this list.

There are so many improved OP-07 parts now that its only advantage is price but that is still an advantage.  I end up using it as a high precision replacement for the 741 and 308.

[Edwin G. Pettis]

For trimmers, for the absolute lowest TCR and best stability, the Bourns 3250W wire wound trimmer is the only choice, none of the other trimmer types come close to its performance.  Granted it is a bit expensive but shopping around they can be found for about $10-$12 each.  They are also the only military qualified trimmer on the market.  One important note, trimmers require a minimum amount of current flowing through the wiper, the 3250W has one of the lowest requirements for this current compared to other types.  The lack of current causes instability in the contact over time, particularly if they are adjusted infrequently.  There is also the 3290W trimmer as well, I'd do a search for them on the web as the mainstream distributors are rather high priced.

[MisterDiodes]

Use chopper-style AutoZero / Zero Drift amps very, very carefully - notice that the current noise spec is usually listed for low freq.  For instance on the '2057 the datasheet makes no mention of the current noise at chopper amp freq, which is quite  substantial.

Connect a semiconductor device directly to those AZ amp inputs (like a photodiode, Zener, LTZ1000, etc.) and now you've introduced a whole new set of problems with those AZ input spikes.  The datasheets will gloss over that part and hurry to tell you that the input spikes won't be much of a factor on the amp output.  But I'm usually worried about what's happening to the sensitive device getting exposed to chaotic current flows at the input side of the amp.  If you don't mitigate these effects, those spikes can result in long term stability / reliability problems you'd rather not have.  Not to mention all the noise you're injecting onto the power supply rails.

Also note that the LTSpice simulator will -not- show you the AZ amp chopper noise spikes- as it does for a lot of other switcher parts: LTSpice just takes a shortcut and assumes ideal amplifiers without calculating every switching event on the amp inputs.  This is one of the reasons it can complete a simulation reasonably quickly; everything is a trade-off in SPICE-style simulators.  This is a case where the circuit simulation doesn't really tell you what's happening in the real world.

You just have to spend some bench time and carefully investigate noise effects - a lot of times it easier to trim the offset of a bipolar amp with a pot and plan on doing a touch up trim now and then.  Unless there's a lot of vibration the pots work well and if done correctly you might only ever do one adjustment per your lifetime anyway.  Take datasheets specs specs with a small grain of salt: You will also quickly learn where the datasheet is overly optimistic or conservative.

Every application is different and sometimes AZ amps can help - but they also introduce a lot of problems that aren't revealed on the datasheets. DON'T take a critical datasheet value at face value without investigating effects in your circuit.  Sometimes a good bipolar amp and perhaps a good PWW trimmer pot (Bourns) works just fine, with less overall noise for your application and better overall reliability.

[mimmus78]

@MisterDiodes

If we use LT1006 and trim it Vos (Input Offset Voltage) can we expect less drift over time than what's specified (0.7 µV/Mo)?
In a LTZ1000 reference circuit, what's about using LTC2057 for the heater part and LT1006 for the zener current?

[Kleinstein]

Trimming the offset of an BJT based OP can slightly improve the TC. This way even the old 741 might reach the <1µV/K range. However external offset trimming usually is not improving aging drift, at least not much. It more like adds additional drift from the trimmer.

The OPA234 (more or less TI's equivalent to the LT1006) looks like it has slightly better specs on the aging drift. However there is always quite some guesswork and safety factor in this number. So not sure the part is really better.

With the LTZ1000 circuit the drift of the OPs is not that critical. The drift of the final output voltage is something like 1/100 to 1/200 of the OPs drift. So using an AZ OP is not needed but is only calling for trouble: the input is nonlinear and can rectify the current pulses. So it would need extra filter caps and thus a modified circuit. So even if only used for the temperature control, there is a chance for additional drift / sensitivity, though at least the output load should not have such an influence.

[MisterDiodes]

@MisterDiodes

If we use LT1006 and trim it Vos (Input Offset Voltage) can we expect less drift over time than what's specified (0.7 µV/Mo)?
In a LTZ1000 reference circuit, what's about using LTC2057 for the heater part and LT1006 for the zener current?

The '1013 is by far the best amp to use with LTZ, and LT (it's all Analog Device now too) apps engineers will confirm that - and they have a lot of amps to choose from.   The LTZ datasheet circuit is about the best there is, and that has been refined over years and years across many manufacturers and millions of device-hours .  You really don't want a chopper amp used around an LTZ (especially as current driver) when you're trying to keep noise out - even though it probably works short term it's not the very best solution for lowest noise and highest stability of the Zener substrate.  You want that current flow as quiet and steady as possible on the LTZ zener.  The 8-pin DIP package helps also if you have to build a board that is subject to mechanical stress (for instance motor vibration) or bigger temperature changes.  Not every LTZ lives in a cushy lab environment - some have a harder life. 

The Voffset of the '1013 has no or very little effect on the LTZ zener-side output voltage (it is a current driver) - we've never seen any need for any Vos trimmer on that amp.  The heater side is pretty forgiving anyway.  That's the beautiful thing about the LTZ1000 ref circuit - so much of its parts can drift quite a bit individually, but without a lot of effect on the output voltage.  Most of the output voltage drift will be traced right to the LTZ die itself, and that is far beyond your control. 

The only place we've ever used a trimmer is on the Vboost circuit that follows the LTZ circuit, maybe.  BUT not all applications need a "trimmed" reference voltage - a lot of times you just need a -steady- reference voltage, and usually very low noise is more important than ultra-low low ppm/year drift.  At least in real applications in our industry when you're running sensitive circuits downstream of the LTZ section. 

[Gyro]

Trimming the offset of an BJT based OP can slightly improve the TC. This way even the old 741 might reach the <1µV/K range. However external offset trimming usually is not improving aging drift, at least not much. It more like adds additional drift from the trimmer.

The OPA234 (more or less TI's equivalent to the LT1006) looks like it has slightly better specs on the aging drift. However there is always quite some guesswork and safety factor in this number. So not sure the part is really better.

With the LTZ1000 circuit the drift of the OPs is not that critical. The drift of the final output voltage is something like 1/100 to 1/200 of the OPs drift. So using an AZ OP is not needed but is only calling for trouble: the input is nonlinear and can rectify the current pulses. So it would need extra filter caps and thus a modified circuit. So even if only used for the temperature control, there is a chance for additional drift / sensitivity, though at least the output load should not have such an influence.

Just remembering from the old NS AN184 (Refs for A/Ds), The LM121 had a predictable TC depending on offset: 3.6uV/'C per mV of offset. It was used for fine tuning their 'ultimate' LM199 reference circuit.

[elecdonia]

 

The LTZ datasheet circuit is about the best there is, and that has been refined over years and years across many manufacturers and millions of device-hours .

Thank you MisterDiodes!

The next thing I will do is to read all the data sheets and app notes I can find for voltage reference devices, including the LTZ.  That will be a terrific source of data for recommendations about OP devices and other parts that are useful in voltage reference circuitry.

[Kleinstein]

...
Just remembering from the old NS AN184 (Refs for A/Ds), The LM121 had a predictable TC depending on offset: 3.6uV/'C per mV of offset. It was used for fine tuning their 'ultimate' LM199 reference circuit.

That 3.6uV/'C per mV of offset is about normal for BJT based OPs, not just the LM121. The offset usually behaves like a small fraction of a diode's drop.