Analog Angle Blog

Can We Have Too Many Op-Amp Performance Curves?

I was looking closely at the data sheet of a new op-amp recently (don't ask why). Like most good op-amp data sheets, it had lots of performance curves and charts. The vendor name and model number are not relevant here. The entire data sheet for this functionally simple and ubiquitous analog building block ran about 16 pages, and pages 4-11 had between eight and 12 such figures each, so we're talking about 80 figures here. (The other pages had some up-front basics, pin-out information, absolute maximum numbers, and package/ordering information.)

The curves in the figures showed combinations of typical, minimum, and maximum performance for many static and dynamic parameters versus all sorts of operating conditions, depending on the parameter of interest. For some, min/max is actually hard to measure, so typical performance is all the vendor can show. These parameters included temperature extremes, voltage rails, signal levels, bias current, offset voltage, common-mode performance, drive levels, output loading, transient response… the list goes on. You get the picture.

To the non-analog enthusiast, this may seem like too much of a good thing, given the huge number of op-amps already available and the never-ending flow of new ones. (See: Do We Really Need Yet Another Op-Amp?.) Considering so many op-amps, each with so many curves, can give you a selection headache. Soon you'll feel like the fellow in this mid-20th-century painting by Dadaist Max Ernst: Young Man Intrigued by the Flight of a Non-Euclidean Fly.

It's easy to think vendors are overwhelming the users with needless data, simply because it's possible to generate so many a -versus-b graphs with automated test equipment and the software to drive it. But in my experience, that isn't the case at all. Most engineers looking at the universe of op-amps (or other analog building blocks) have a set of first-tier performance issues to consider as they try to drill down and select a few devices for extra consideration.

Then comes the real challenge of knowing and checking the second- and third-tier factors that are critical to your specific application. Many of those characteristic curves are not relevant to your situation, but some certainly will be. Knowing which ones you need to study — that's one of the marks of a seasoned analog engineer.

Have you ever been overwhelmed by the mass of performance specs and curves on the data sheet of a basic part such as an op-amp? Conversely, is there a spec rarely called out in detail that you have to ask about — knowing that a verbal answer from the vendor does not provide the same assurance as seeing it in black and white on the formal data sheet?

Related posts:

7 comments on “Can We Have Too Many Op-Amp Performance Curves?

  1. samicksha
    November 6, 2013

    I guess feedback factor of an op-amp application defines the circuit performance in best way, it sets the gain received by the input-referred errors of the amplifier, often we see open-loop errors include offset voltage, noise and the error signals. In addition, a circuit's feedback factor determines bandwidth and frequency stability.

  2. RedDerek
    November 6, 2013

    It is interesting to compare the datasheets of an old opamp such as the LM741 to what you are describing – a 4 page data sheet versus 20+. I have learned to start looking at the curves to see what interesting things the new part could do in my circuit. The key is just to understand really what is going on to generate the curve.

  3. Alan S
    November 13, 2013

    Never too much data – almost.  Short of a book, as long as you can find what you want, just ignore the rest.  Maybe I don't care about distribution of Vos at -40° or BW and phase at Acl =Aol/2, but maybe I do.  Sometimes one has to know this obscure info so that when asked, I can say what the probability of centreline performance will be, rather than typical with hard limits.  Bring on the data!

  4. Winfield Hill
    November 13, 2013

    Never too much data indeed, that's the reality when you discover that the one thing you need to know is missing.  What's the input bias current and voltage offset as the common-mode voltage approaches the positive rail?  Missing?  Oops!  What's the output slew rate as a function of input error?  Missing?  Too bad.

    One issue not mentioned in the article is the idea of comparison curves, that is, plotting the same parameter for dozens of op-amps on the same graph so we can easily see how they compare.  A painful time-consuming taks to be sure, but potentially very useful.  That's what we've done in the new 3rd edition of The Art of Electronics, due out soon in 2014.  And even if your favorite op-amp, or the one you have under consideration isn't on our graph, you can pencil-plot it there yourself and quickly see how it stacks up against our selected exemplar parts.

    I'd love to see a manufactuer's data sheet with a few comparison-part plots on it, wouldn't you?

     — Win

  5. bcarso
    November 14, 2013

    Yes, the details thus are crucial.  I'm happy to see the pages of curves, although they could be worked into a more comprehensible set in many cases. And as Winfield points out, some really important things are missing most of the time.

    Remember the days when discrete component manufacturers like National Semi  routinely supplied characteristic curves?  There are still a few who do, but I gather that too many engineers were taking the typical data and supposing that it guaranteed that performance.  Plus, it was a lot of work to plot all that stuff, and with processes usually tending to improve, it became obsolete.

    When I designed the electronics of a small portable audio product for a client, I was given a difficult combination of very low cost, low power consumption, and adequate signal-to-noise and distortion.  The low-power op amps of that day were too expensive.  The complex equalization specified by the loudspeaker guy was a natural for DSP, but particularly at that time the conversion electronics and DSPs available would have badly blown the power budget.  So I used carefully-tailored gain cells made from lots of SOT-23 parts, and specified a reliable vendor with whose parts I'd gotten lots of experience.  The design was vetted by a friend, with about the only change being in the auto-on/off timer, where I agreed to live with the uncertainties of capacitor leakage rather than betting on the low leakage current of a transistor.  The product shipped in lage quantities with a very low return rate, despite the rather high parts count.

    The vendor of the filterless class-D amplifiers said a customer had tried to reverse-engineer the design and had given up.  That was satisfying to hear.


    Brad Wood

  6. Navelpluis
    November 15, 2013

    Count me in for as much specs as possible. New parts arrive with very poor specifications. For sensor design -as an example- having all specs is a *must*. My experience is that with each design I fight through my database of opamps. Very seldom I use something new, mostly 'the old beasts' can do the job. It is a combination of 3 dominant variables. I call this the 'triangle of misery' where we -as engineers- have to make a choice for best of all options. So definately a big *yes* for as many specs as possible.

  7. SunitaT
    November 30, 2013

    In the OP amps,  the most frequency reply of an operational amplifier is limited to Gain/Bandwidth produce or open loop current gain ( A V ) of the op amp being used providing this a bandwidth restraint, where the fastened loop reply of the operational amplifier crosses open loop reaction.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.