Analog Design
 

Analog Design

We actually know where it all came from in the beginning. You have to understand where it came from, if you want to understand where it's going - that's old school.
"Analog is Everywhere" runs the advertisement of a well-known chip manufacturer that years ago employed several of our principals as Field Applications Engineers.

There's rarely such a thing as a purely digital circuit anymore, since the speeds of buses have increased to the point that everything is ultimately being examined for its risetime or eye pattern. And the world remains an analog beast: if you're going to interact with it, sooner or later you'll have to meet it on its own terms.

To help you in this, We will:
  • Tackle the most difficult of Laplace Transforms.  If it happens in the "Land of Oz" that is the left half of the s-plane, we can tell you what it'll mean back in Kansas.  And if it happens in the right half plane, we'll tell you before it blows up in your face.
  • Find suitable tradeoffs between stability, responsiveness, risetime, damping, bandwidth, etc.  More of any is not always better.  Better is better, and we don't lose track of what problem we're actually solving.
  • Operate in both continuous and discrete time domains.  Our understanding of signal processing starts with mathematics.  The fact that the number crunching may be digital is an implementation detail.  The fundamental underpinnings of it all are in the math, and this falls out of the analog world, and incidentally enhances our ability to do DSP design.
  • Recognize where information is lost when crossing from analog to digital domains.  Similarly we can spot where false information can be generated going from one domain to the other. Fourier Theory describes what's real and what's a false image that may show up where you least expect it.
  • Implement control systems (analog or digital) and do the continuous or discrete-time analyses required  in determining their responsiveness and stability using techniques such as Root Locus of Evans and Routh-Hurwitz Matrices.
  • Build SPICE models that make sense. Using a computer tool cannot substitute for a fundamental understanding of how circuits behave.
  • Have a hard look at the "device physics" of a transistorWe understand what's happening at the silicon level in p-n junctions and bipolar and FET transistor structures. When you need just the right pair of matched transistors for your Gilbert cell or the accuracy of your clamp circuit depends on using the B-E junction of an NPN transistor as a diode and tying off the collector, we'll know what to do.
At Focus Embedded, analog is nothing new. Our team members have been putting together analog circuits since the days when they accounted for the vast majority of new designs.

True, we do SPICE modeling. If you need statistical analyses of your circuit, complete with a fallout calculation for component variation, we can build the model that guarantees that the 100,000th circuit board runs just as well as the first. And if you have an existing board that's failing, we can dive into that analysis as well.

More importantly, though, we can find closed-form solutions to problems with nothing more than pencil, paper, and higher mathematics. Good design isn't about having the computer technology to generate incorrect answers more quickly than a human can spot them; it's about understanding what's going on at a fundamental level and building the correct model in the first place.

We've been doing analog design work long enough that we're aware which components are non-ideal and which of their parasitic effects can impact system performance. When it matters that an inductor is also partially a capacitor and partially a resistor, we'll be aware. When Johnson noise is enough to affect a measurement in a highly sensitive instrumentation amplifier, we'll find it and ferret it out – or describe it so accurately mathematically that software can spot it easily to remove it later.  And when the copper traces on a PCB become very much a part of the circuit, we'll know.

We've designed power management circuitry, so we know how to keep things green by matching complex impedances properly (and doing so dynamically if the source or load is a moving target).

Because we do circuit design, embedded software design, and programmable device design, we'll be able to tell you where you need one and where you can very likely do as well or better (possibly for much lower cost) with one of the others. Getting back and forth between "s" and "z" domains means doing the old-school math to which we're accustomed. Presuming one to be the preferred solution because it's all you know is a formula for missing better alternatives.

We have some rare talents at Focus Embedded, who can find an analog problem — and provide an elegantsolution — before you go to market.