Quick Opamp slection guide

[anishkgt]

Hi All,

I was reading something about choosing Op-amps and this link (http://www.nutsvolts.com/magazine/article/how_to_choose_an_operational_amplifier) has some good info about it for beginners.

There are limits to the input voltage you can apply to the amplifier and expect it to work properly (which is different from the Fry Points). This is called the common mode voltage range (CMVR). Many (probably most) new amplifiers have “rail-to-rail” inputs which permit you to use any voltage up to and including the V+ and V- voltages. Conversely, many amplifiers (especially the older ones) limit the input voltage to a fixed voltage less than the V+ and V- supplies (typically about 1.5 volts).

a quick question about it, is it referring to the inverting and non inverting inputs ? I am using the LT1078 for my project and was trying follow it with this guide but could not find the CMVR anywhere in it.

Could somebody please advise me the other name for it in the datasheet (http://cds.linear.com/docs/en/datasheet/10789fe.pdf) ? secondly is there a limit to how much volt can be applied at the inverting and non-inverting inputs ?

[Kleinstein]

When in normal operation mode, the voltage on both inputs is essentially the same. So there is only one common mode voltage range for both inputs. In the LT1078 DS the parameter is called input voltage range.

There is also a similar parameter in the absolute maximum ratings.

[Brutte]

The input that reaches both rails would have to be made with two separate stages (higher and lower) that interleave each other somewhere in between rails.
Rather complicated stuff, but doable.

Conversely, many amplifiers (especially the older ones) limit the input voltage to a fixed voltage less than the V+ and V- supplies (typically about 1.5 volts).

I really doubt it. AFAIK it is either 1.5V below V+ or 1.5V above V-, depending on the construction, but never limited at both sides.

[David Hess]

The input that reaches both rails would have to be made with two separate stages (higher and lower) that interleave each other somewhere in between rails.

There are at least two other ways.

Some CMOS operational amplifiers have an integrated charge pump to provide a bias supply for the input differential pair allowing a rail-to-rail input range;  Maxim makes at least one part like this.

National (bought by TI) has a series of CMOS operational amplifiers which use back gate biasing to shift the threshold voltage of the input differential pair over the common mode input range allowing a rail-to-rail input range.  These have terrible common mode rejection but are still useful in low precision circuits.

As far as the original question, the common mode input voltage range of the LT1078/LT1079 is specified as Input Voltage Range in the electrical characteristics and as Vcm under Common Mode Rejection Ratio.  It can also be found in the Common Mode Range versus Temperature graph.

Update: The Maxim operational amplifier with the charge pump that I was thinking of is the MAX4162.  I remember it being a pretty big deal in the trade journals when it came out.

[Brutte]

integrated charge pump

Ok, so input stage has separate power supplies (charge pumps).

An internal charge pump provides two internal supplies typically 2V beyond each rail.

OTOH input stages use ~uA so there is not a big deal to integrate small charge pumps inside. Interesting and complicated design.

Input Common-Mode Voltage Range Extends 250mV Beyond Either Supply Rail

As for back gate biasing - I have never heard about such technique before - thanks for info.

As far as the original question

CMVR spans from GND-0.5V up to Vcc-1.5V. Just like with typical bipolar PNP input type.

[David Hess]

As for back gate biasing - I have never heard about such technique before - thanks for info.

National did this with the LMC6462/LMC6464 (micropower) and LMC6482/LMC6484 which are still in production.  I used a lot of the LMC6482/LMC6484 parts.

Like the Maxim part with its internal charge pump, they made a splash in the electronics trade magazines when they were introduced which included technical articles.

CMVR spans from GND-0.5V up to Vcc-1.5V. Just like with typical bipolar PNP input type.

Sure, the LT1178/LT1179 is an even lower power implementation of the LT1077/LT1078/LT1079 which is a micropower implementation of the LT1006/LT1013/1014 which is a precision implementation of the LM358/LM324.  Linear Technology has newer versions of these micropower parts now in the form of the LT2178/LT2179 and LT2078/LT2079 although I am fuzzy about what was changed.

[paulca]

Found this thread after the LM741 ruined my evening again.  Was trying to create a voltage ramp circuit to feed into diodes and transistors etc. to probe their behaviour across a linear voltage sweep.  I found a circuit that used an opamp to amplify the first 2V of an RC ramp to give a more-or-less linear ramp.

Spend ages wondering why it wasn't working, long story short, I remembered that the LM741 only gets to about GND+1.3V.  I tried dumb things like dropping the voltage on the feedback with diodes, but in the end gave up.

Obviously for this to work I need an opamp that will go less than a transistor/diode activation voltage above it's lower supply rail.

The guide above is great but it would be nice to have a basic list of "beginner op amps" for basic purposes.

So if you had to pick a basic (close to) rail to rail opamp for general tinkering what would it be?

EDIT:  So I looked up the op amp used originally in the circuit I was working from, it was this:
http://www.ti.com/lit/ds/symlink/tl081.pdf
From the circuit:
https://www.circuitlab.com/circuit/xw65e2/voltage-ramp/

First he is running it at 0-23V which is beyond the amps specs of -18 +18V (or does it fall back to the differential voltage for + only)
Second I can see the common mode input voltage range is -12 to 15V but as that isn't relative to Vcc/-Vcc how do you know how low/high it's output will go?
The Peak output voltage is: ±13.5 but again, at what Vcc is that for?  ±18V?
The other thing is that Circuit Lab did not simulate the 1.3V minimum voltage of the LM741, the circuit simulated perfectly when I exchanged that part.

Baffled.

[David Hess]

My recommendation is to invest in a floating dual polarity tracking supply so you can use whatever operational amplifier you want.  Single supply operational amplifiers have various performance compromises and despite what they say, their outputs do not go rail-to-rail although their inputs may.

[Brutte]

voltage ramp circuit to feed into diodes and transistors(..)Spend ages wondering why it wasn't working, long story short, I remembered that the LM741 only gets to about GND+1.3V (.. )Obviously for this to work I need an opamp that will go less than a transistor/diode activation voltage above it's lower supply rail.

Ok, but mind this is output-related issue while the rest of the topic is input-related.

I am looking at LM741 datasheet and schematics and IMHO it can source anywhere in the range of <Vss up to Vdd-1V. It can easily drive anode of 0.7V diode that is tied with cathode to Vss. 0A up to about 25mA where the current limiter triggers.

The "output voltage swing" parameter (Vss+1V : Vdd-1V) only applies to the sink&source region where the test parameters apply. But the diode tied to GND would never be sourcing anything so eventually an op-amp's output does not have to sink anything with only a diode tied this way. Effectively the lower part of the output driver would be idle in such arrangement. Lack of sinking near Vss can generate some problems with discharging diode's internal capacitance at tens of MHz band but that frequency seems out of reach for this opamp. Not tested or simulated - just speculation.

[paulca]

voltage ramp circuit to feed into diodes and transistors(..)Spend ages wondering why it wasn't working, long story short, I remembered that the LM741 only gets to about GND+1.3V (.. )Obviously for this to work I need an opamp that will go less than a transistor/diode activation voltage above it's lower supply rail.

Ok, but mind this is output-related issue while the rest of the topic is input-related.

I am looking at LM741 datasheet and schematics and IMHO it can source anywhere in the range of <Vss up to Vdd-1V. It can easily drive anode of 0.7V diode that is tied with cathode to Vss. 0A up to about 25mA where the current limiter triggers.

The "output voltage swing" parameter (Vss+1V : Vdd-1V) only applies to the sink&source region where the test parameters apply. But the diode tied to GND would never be sourcing anything so eventually an op-amp's output does not have to sink anything with only a diode tied this way. Effectively the lower part of the output driver would be idle in such arrangement. Lack of sinking near Vss can generate some problems with discharging diode's internal capacitance at tens of MHz band but that frequency seems out of reach for this opamp. Not tested or simulated - just speculation.

My bad, it is that it won't go down to the negative rail, which in my case is GND.  The best I could get out of it was +1.3V and if I ramped up the rail voltage that rose to 1.45V with a 12V rail IIRC.  There was nothing connected to it.  I tried it with nothing, a current driver NPN a 100Ohm resistor and I could not get it to go below 1.3V.  If I had a negative supply I could give it then it would be fine.  I considered offsetting the input but was so annoyed that I couldn't figure out how to remove the offset at the other end.

I needed a 0V to 9V ramp. 

I've done some digging and googling to find some candidate rail-to-rail single supply op amps.  Ironically the one TI recommend in their single rail guide is one of the hardest to find and quite expensive the TLV247x.  Farnell, nor CPC stock them.  Their partner stocks them but only in 25+ quantities.

Alternatively I could run the op amp off two P3 batteries wired back to front, but that sounds like a hack.

What is the actual datasheet parameter that tell me how close to it's rails it will go?

[Vtile]

Use another opamp to split the rail?

[paulca]

Use another opamp to split the rail?

What confused me about this was the coupling cap required.  I was thinking my DC ramp was just a DC signal and so I didn't know how to add and subtract the offset to get the mid rail VGND.

I did a few calculations (an online one) and a 22uF (or higher) cap will allow frequencies well below 0.5 Hz, so it should pass a 1 second 0-9V ramp easily enough.

Thanks for the reminder.  I'm learning all the time.

[Brutte]

The best I could get out of it was +1.3V and if I ramped up the rail voltage that rose to 1.45V with a 12V rail IIRC. There was nothing connected to it.  I tried it with (..) a 100Ohm resistor and I could not get it to go below 1.3V.
What is the actual datasheet parameter that tell me how close to it's rails it will go?

The "output voltage swing" is the name of the parameter where it can source and sink. But you can go beyond that if you only want to pull or push. AFAIK that full range is not specified and opamp is not qualified there.

Your observation and 1.3V is really intriguing.
I do not own LM741 but I do have OP07 which has somewhat similar output driver (but much more complicated input b.t.w.). I am powering that with 12V and the:
-unloaded output swings [Vss+1.4V : Vdd-60mV] (datasheets says Vss+2V:Vdd-2V).
-with a 100R resistor in between output and Vss the output swings [Vss+1V : Vss+2.8V].
-with a 100R resistor in between output and Vdd the output swings [Vdd-2.8V : Vdd-10uV].
(28mA current limiter is cutting in).

Seems that I have to rethink the subject. It looks like depending on the output stage construction some op-amps outputs won't go close to some rail even when sinked or sourced externally.. 

[paulca]

See my last post in "DC Load" thread.  I'm about to find out how one copes in the microwave.

[paulca]

Now I've chilled out a bit, let me confess I have a really hard time reading some data sheets, op amps in particular.  I'm not familiar with the acronyms.  Looking the voltage swing for the LM741 am I reading it right that:

For a Vs (Supply voltage) of +-15V it's minimum output is -12V and if the load is < 10K > 2K it's -10V.  A whole 5V off the lower rail.  If the load is < 2K it just doesn't tell you.  That explains a lot.  If that were proportional on a 0-9V(ish) supply it could bottom out at 2V.  Which is about what I was getting in voltage follower mode tonight.

[Vtile]

uA741 is as far as I know the oldest opamp (1970?) still in production and it were the device that made the use of these a breeze. It really is meant to be driven from double supply and not single rail. That is the reason the boundaries are so big in this device, but hey would have you learned those with rail-to-rail device.  

[paulca]

but hey would have you learned those with rail-to-rail device.

True.  I was looking with a better insight into rail to rail devices.  It looks like they exist in both rail to rail input and outputs, but I'm still not convinced I am reading the limitations correctly.

How does rail splitting work when you are just using it as a DC amplifier?

Say to take a DC voltage of 0.0V to +0.1V and amplify it by a gain of 40 to get DC +4V, can that be done with coupling caps and a virtual ground or do I need a genuine rail to rail amp?

[Vtile]

I'm really not an electronics expert, but as far as I have understood ie. virtual ground opa circuit should work in limits of the opa used to create the VGND, what I have understood it also should work better in low frequencies than high. I haven't had need to experiment with one so for that solution you need to get someone that can answer off-hand and in detail.

Here is one opa rail-splitter circuit I have copied to my notes from some handbook. The supply needs to be a floating (ie. battery) to allow more freedom to ground the input signal in point you need in your amplifier circuit.

[paulca]

Here is one opa rail-splitter circuit I have copied to my notes from some handbook. The supply needs to be a floating (ie. battery) to allow more freedom to ground the input signal in point you need in your amplifier circuit.

Thanks.  I have that circuit already, it gives me an impedence separated / buffered fixed voltage at 1/2 Vcc to use as a virtual ground.

I have also successfully used it to rail split the LM741 to amplify an audio signal. 

The thing that confuses me is that audio etc. is an AC signal, so a coupling cap can be used to separate the virtual ground and real ground by blocking the DC and allowing the AC to pass.

The bit I'm not sure about is what happens if I put a fixed voltage DC signal as an input to such a circuit.  My instinct (which is more often wrong in electronics) tells me the coupling cap will just charge up to the new voltage over a period of time and the amp will just slam onto the rails.

I'm gathering for DC amplifiers I really do need a rail-to-rail op amp or a dual supply to feed the Vcc- of the op amp with a voltage well below my lowest input.

[David Hess]

uA741 is as far as I know the oldest opamp (1970?) still in production and it were the device that made the use of these a breeze.

The 741 was introduced in 1968 and designed in response to the LM301 (1967) which is still produced in the form of the LM301A (late 1968) which was improved to compensate input bias current over temperature.

Having used both, I prefer the LM301A because its input common mode range includes the positive supply and its COMP pin may be used as an output clamp.

The TLV247x series only runs on up to 5 volts.

The CMOS TLV2371/2/4 is about the least expensive TI option up to 15 volts but it is not a very good operational amplifier which is a common problem with rail-to-rail input designs.  I have used a lot of LMC6482s and LMC6484s which are also CMOS and limited to 15 volts but also not very good.  The good ones are much more expensive.

Circuit design can make up for lack of a rail-to-rail output.  Usually input operation is only needed to one rail and there are parts good to the negative or positive supply.

[paulca]

Circuit design can make up for lack of a rail-to-rail output.  Usually input operation is only needed to one rail and there are parts good to the negative or positive supply.

These videos look interesting.  https://www.youtube.com/playlist?list=PL9B4edd-p2ahtUMzAMv_cp_sWJt6RNk-2

The one I watched explained the rail-to-rail input stage problems.

I think I have come to the conclusions that for DC amplification for current sensing etc I need to use a rail to rail and accept it has limitations or create myself a dual supply somehow.  For AC stuff I can split the rails and AC couple both ends.

Learnt a lot about opamps these last few weeks.  Mostly through trial, error and frustration.  I expect I have a lot to learn still and more frustrations to come.  That's half the fun though.

[Vtile]

https://www.maximintegrated.com/en/app-notes/index.mvp/id/1199
... There is ICs for creating real negative rail so the gnd should stay intact. ie. MAX829.

https://para.maximintegrated.com/en/results.mvp?fam=chargepumps&828=Inverter

[David Hess]

I think I have come to the conclusions that for DC amplification for current sensing etc I need to use a rail to rail and accept it has limitations or create myself a dual supply somehow.  For AC stuff I can split the rails and AC couple both ends.

This is one of the places where the application matters a lot.  Current sensing in the supply lines almost always requires an input common mode range which includes one or the other supply voltage but not both so a rail-to-rail input operational amplifier is not really needed.

"Single supply" operational amplifiers with an input common mode range to the negative supply are very common so that is not a problem.  However there are also operational amplifiers which have an input common mode range which includes their positive supply voltage like:

LM301A - Effectively a 741
TL031/TL051/TL061/TL071/TL081 - JFET Input
LF351/LF355/LF356/LF357 - Also JFET Input

Of the above, the LM301A and TL031/TL051 (cheap precision low power and fast JFET input) are the most useful in my opinion.  Few of them and none of the ones above have an output which goes to the positive rail however adding a couple diodes, a zener diode, or an LED in series with the output and a pull-up resistor or current source can solve that.