Design Article
Balancing gain-bandwidth product vs quiescent current for dissipation optimization
Namrata Pandya, Product Marketing Engineer, Microchip Technology Inc.
8/25/2011 8:52 PM EDT
Though conceptually simple, operational amplifiers implement parametrically complex circuits that can pose numerous challenges to your IC-selection process. Many familiar mains-powered system designs include power budgets dominated by digital circuits, with analog subsystems representing only a small fraction of the total dissipation.
In these cases, your top priority for operational amplifier selection is likely one or more signal-chain performance parameters. These might include AC performance terms, such as distortion and broadband noise performance, or DC terms, such as input offset and offset drift.
But, in energy-constrained applications for which you must wring out every last bit of unnecessary power dissipation, the temptation is to start by looking for the operational amplifier with the lowest quiescent current. Unfortunately, this intuitively reasonable approach all too often identifies numerous candidates that meet the application’s power requirement but not necessarily its GBWP (gain-bandwidth product) needs.
For a given circuit topology, GBWP and Iq (quiescent current) go hand in hand—essentially in direct proportion. The reasons for this behavior are several and tied to the detailed topology of specific amplifiers.
At the top level, however, consider that the operational amplifier you choose must charge and discharge internal capacitances at signal speed. The resulting displacement currents flow from the internal bias current of the amplifier, which determines the net Iq. Therefore, for a given topology, as bandwidth increases so must the amplifier’s Iq.
A helpful figure of merit
The challenge for low-power design, then, is not simply to find low-power operational amplifiers, but to find operational amplifiers that most efficiently provide bandwidth. A simple figure of merit to assess operational-amplifier bandwidth efficiency is the ratio of GBWP to Iq.
In an example in the article, a performance comparison and figure-of-merit calculation for four devices of similar architecture—in this case, the Microchip MCP644X, MCP640X, MCP628X and MCP629X—shows a figure of merit that varies barely more than an octave, while the GBWP and Iq vary by a little more than three orders of magnitude. Indeed, at the lower bandwidths, the figure of merit for these devices is nearly constant.
In practice, your selection process will focus, of course, on competing devices of similar GBWP, not a group covering several orders of magnitude. First, though, you must determine the answer to one key question: "GBWP: How much is enough?"
This article explores the tradeoff and balance of gain-bandwidth versus quiescent current and in detail, with analysis, equations, and examples. To read it as a pdf document, click here; to read it as a Microsoft Word document, click here.
About the author
Namrata Pandya is a Product Marketing Engineer for the Analog and Interface Products Division of Microchip Technology Inc. in Chandler, Ariz. She is responsible for the strategic marketing of operational amplifiers, as well as tactical marketing support for Microchip’s Analog and Interface products in the South Pacific and ASEAN countries. Prior to joining Microchip in 2007, Namrata spent two years with Cypress Semiconductor in San Jose, CA in Product Marketing. She earned a BSEE degree from Mumbai University in August 2001 and a Masters of Electrical Engineering degree from San Jose State University in December 2006.



kendallcp
8/31/2011 3:52 PM EDT
There are some statements here that I'd take issue with if you're looking for good signal handling performance such as AC linearity (for instance on how close you should be working to the unity-gain frequency of the device).
One suggestion though is a very useful one, and I've used it quite a few times. Instead of getting all your gain out of one opamp, use two amplifiers, which can have much lower GBW and Iq. The example shown splits the gain equally between the two amplifiers, but it's generally best to put somewhat more in the first stage if you have any leftover GBW margin. That's because the signal level there is lower, so the linearity is better, so the loop gain doesn't have to work so hard. More feedback is useful in the second stage.
Always look out for decompensated amplifier options too. If you don't need unity-gain stability, there's no need to accept the loss in GBW suffered in achieving it. Decompensated low-power amplifiers are often just as good as compensated higher-power amplifiers when gains are high.
Sign in to Reply
KedarGodbole
9/3/2011 2:07 AM EDT
Kendall,
Though the article goes into some analytical gymnastics, it is a severe oversimplification of the matter.
The most important factor to consider when thinking about the amplifier current consumption is not gain bandwidth.
It is slew rate, isn't it.
For most class A output stages the output current and slew rate have the direct relationship.
Unless the signal is very small, a designer can only ignore slew rate at their peril.
For a class A output stage slew rate SR = Ib/CL, Ib is the output stage bias current, and CL is load current where that simple.
Also as you already know, with most ADCs you would want the signal to span a significant fraction of the input range. That means your signal especially at the input of a converter is going to be large.
I wish the article went into more details about the modern output stages, whether they are class A or class AB.
I will ask a few folks sure to know in a few weeks.
Cheers, Kedar
Sign in to Reply
ultimateanalog
9/4/2011 2:53 PM EDT
To be honest, this artical is just too superficial and even made several mistakes.
Almost everyone knows the Iq is proportional to GBWP, but the real question is how to determine the enough bandwidth for your applications.
For example, in the section of "GBWP: how much is enough?", when your amplifier in your application is followed by ADC, the bandwidth is actually determined by FPBW(full-power bandwidth, FPBW = SR/(2*pi*Vout-pp)), not by GBWP/Gn where Gn is noise gain because ADC is usually set to the full input scale range. In another word,for large signals, you need to consider SR to determine bandwidth in stead of the GBWP and noise gain. BW = GBWP/Gn is only applied to small signals.
This article should not be called as design article. I am looking forward to the technical paper from applications engineer or design engineer. Marketer is just not the one.
Sorry to say this, buy I have to point it out.
Sign in to Reply
ultimateanalog
9/4/2011 3:19 PM EDT
In my reply above, the FPBW should be equal to SR/(2*Pi*Vout) = SR/(Pi*Vout-pp), moreover, usually op amp data sheet only specifies the typical SR value, not the minimum one. You need to pay attention to this, for the worst case analysis, the SR could change by +- 30%, even more.
The real op amp expert would tell us more.
Sign in to Reply
kendallcp
9/5/2011 6:21 AM EDT
Well, I agree that the article has some deficiencies, but didn't feel it was appropriate or fair to pick it apart line by line. To be honest I don't have great expectations of articles like this. Some of the comments above deserve a reply, though.
The relationship between GBW and slew rate is topology-dependent. For traditional op-amps it's dominated by the input stage; rail-to-rail output stages can add some complication.
Really, you don't want to get anywhere near the slew limit of an op-amp in operation. By the time you hit the SR limit, the amplifier has /stopped working/ as an amplifier. Check out my piece at http://www.eetimes.com/design/analog-design/4019542/Use-it-or-slew-it
Sign in to Reply