Analog Angle Blog

Good, better, and best: What’s your performance metric?

I was doing some research recently into the history of telephony and came across a technical term I had not seen in a very long time: erlangs. Back in the day when landline-based telephony ruled the world with its dedicated, circuit-switched loops between a local central office and the user’s phone, or between central offices via “trunk” lines, erlangs were a big thing.

So, what is erlang? It’s a unit of telecommunications traffic measurement representing the continuous use of one voice path. In practice, it’s used to describe the total traffic volume of one hour. For example, if a group of users made 30 calls in one hour and each call had an average call duration of 5 minutes, then the traffic figure is worked out as shown in the Table below.

The importance of the erlang as a vital metric in its heyday cannot be overstated. There were countless related articles, tutorials, analyses, and textbooks produced along with reams of field-research data looking at erlangs and phone traffic from various perspectives. The legendary Bell Labs devoted a significant amount of its resources to everything and anything erlang related. Now, of course, erlangs are rarely cited—it’s almost as if they are candidates for yet-to-be-built Museum of Obsolete Engineering Metrics.

But as erlangs have receded, we have many older ones we still rely on; even the humble op amp has a long list of standard parameters which a vendor is expected to call out on the datasheet (see reference below). Of course, there are also new metrics and figures of merit (FoMs). This makes sense as applications, technologies, and designer and end-user priorities evolve and advance in expected and unexpected new ways.

So, what’s the difference between a metric and a FoM? It’s a somewhat fuzzy topic with overlap. I like to think of a metric as a basic specification or performance parameter such as maximum operating voltage or quiescent current. In some cases, more or less is always better; in other cases, it’s not really an issue. For instance, do you need a higher maximum operating voltage once the rating exceeds the system voltage? In contrast, a FoM is often somewhat more complicated, as it’s often a ratio showing the relative balance or tradeoff between two specifications such as watts/cm3.

There are many commonly used FOMs, such as:

  • MLOPS/mW: It looks at the floating-point computational performance versus power dissipation of CPUs, especially DSPs.
  • Miles/gallon (or km/liter): It speaks to the basics of fuel and operating efficiency of an internal combustion engine.
  • Bits/sec per Hz: It measures the throughput efficiency of a channel, usually with respect to Shannon’s sampling theorem.
  • Watts delivered per cubic inch or centimeter for power modules and supplies.
  • Breakdown voltage per linear distance (V/in or V/cm).
  • Energy density by weight (joules/kg or watt-hours/kg) and volume (joules/m3 or watt-hours/m3) and power density by weight and by volume (just substitute joules or watt-hours in the numerator with watts in those ratios).

One metric getting increased attention relates to electric vehicles (EVs) and their recharge time. While absolute range is obviously an issue, many users are also concerned about this equivalent to gasoline fill-up time. It’s a function of vehicle and the charger arrangement more than the batteries themselves and is commonly expressed as miles/minute (or km/min).

As technologies and priorities evolve, figures of merit also change. What metric or figure of merit do you cite most often? Are there some older ones which you feel are no longer very useful or indicative? Are there newer ones which you feel are more relevant or which you have devised?

Related Content

1 comment on “Good, better, and best: What’s your performance metric?

  1. Seblegrand
    August 20, 2021

    thanks for this interesting article

Leave a Reply

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