Because of the Thanksgiving holiday in the U.S., TI E2E design support forum responses may be delayed the week of Nov. 21. Thank you for your patience.

Designers sometimes find op amp data sheet specifications perplexing because not all performance characteristics have minimum or maximum specifications. You must occasionally rely on “typical” values in the specification table or typical performance graphs. But what does typical mean? How much can it vary?

There are no easy answers and it depends on the specification. Here are some guidelines on three characteristics which seem to get frequent questions:

Bandwidth—The gain-bandwidth product (GBW) of an op amp is primarily controlled by the current in the input stage and the on-chip capacitance values. The variance of these two variables could mean that GBW could vary by ±20% or so. While this might sound like a wide range, it’s generally pretty easy to design with ample GBW by selecting an op amp with sufficient margin. The closed-loop bandwidth of your application can be controlled with feedback components, if necessary. Notice that on an open-loop gain/phase graph (figure1) this variation appears remarkably small.

Slew Rate is affected by the same variables as bandwidth, internal currents and capacitance and the same issues apply. Op amp selection of 20% faster than your minimum needs will usually suffice. Perhaps you might want a bit more margin in critical applications. Most applications do not push amplifiers near their slew limit so this is not often an issue.

Voltage Noise—Broadband or flatband voltage noise of amplifiers depends primarily on the current in the input stage transistor(s).  Higher current reduces the noise in a square-root relationship. So a 20% variation in current makes approximately a 10% change in flatband noise density (figure2).

Low frequency 1/f noise (also called flicker noise) is a different matter and can vary much more. The noise amplitude in the 1/f region can vary over approximately a 3:1 range. A somewhat larger variation might be expected in JFET and CMOS processes. This noise region determines the peak-to-peak noise in low frequency band, often specified from 0.1 to 10Hz. So this noise spec can also vary over a similar range.

These are reasonable guidelines but there is no way to account for the wide range of amplifier designs and IC processes used. But some information is better than none and most designs can comfortably tolerate these estimated variations.

The margins appropriate for your application may vary according to the type of equipment you are designing, and, perhaps, the end-product testing you perform. The consequences of a marginal miss on a spec can affect the margins you design to. This engineering judgment is an important factor in good analog design. If you need guidance on a particular situation, you are welcome to ask for more information. Our E2E forums are the place to do it. We may be able to provide some additional information or estimates.

Are there other typical specs give you grief? If so, I’d like to hear about it. Your feedback might actually influence how we specify our next IC.

Thanks for reading and your comments are welcome.

Bruce

Anonymous
  • In all my years of reading op-amp datasheets I found that many, if not all, dynamic specifications (noise, GBWP, Gain/Phase margin, SR, THD+N anything vs. Frequency, etc.) are presented as typical values. On top of a typical value, it is usually only stated to be at room temperature - so what happens at cold and hot temperatures? Since the tail current drives most of these I presume a good rule of thumb would be /2 and x2 at cold and hot respectively. When you design ULTRA LOOOOWWW POOOWEEEERRR!!!! circuits, I tell ya, the customer wants every nA saved. Yikes!

    This is a good write up since these gems lie buried in the minds of designers.

    -Ken

  • John--  I'm a long distance literally and figuratively from the MSP430 product line so I can't comment with authority. The practice of offering various versions of a device with different tested limits and price points is very common in the IC business. Testing and assuring performance limits has real costs and we often provide more than one version to meet differing needs.

    Thanks for your example and comment.  Regards, Bruce.

  • It's a quite different system, but the difference between the F2xx and G2xx series of MSP430 devices is an interesting example of typical values compared with limits. Some G2xx devices appear to offer identical features to F2xx devices but the price of the corresponidng G2xx is lower. Many of the parameters are given only as 'typical' values for the G2xx but have limits for F2xx. I assume that this lowers the testing costs for the G2xx because the limits do not need to be checked.