Scolded by a Child

This past week, the manager of the digital design group decided to give a good scolding to us analog guys. His frustration is due to the analog design changing on a regular basis, which sometimes require associated changes to the digital logic. He wonders why we analog guys can't do a bit of up-front feasibility study (let's say, 10% of our design time) and then finalize our designs' pin-outs, specs, and footprints up front. After all, the digital guys are able to do it. When Mr. Halal responded that analog is more complex, we got the "yeah, I knew you were going to say that" response. The manager may have known how we were going to respond, but he didn't seem convinced.

In the digital world, functionality is implemented in software-like code call RTL. Thus, when one finds a functional bug in a digital design, changes to a text file fixes the problem. Functionality aside, there are three other major specs that needs to be met -- power, area, and clock speed. Three variables. That's it. I'm not saying it's trivial work; it isn't. But let's go through an example of what an analog designer needs to worry about when designing an amplifier:

- frequency response
- gain
- phase margin
- gain margin
- biasing
- headroom
- DC offset
- slew rate
- power
- input impedance
- output impedance
- compensation capacitance
- area
- matching
- common mode
- common mode feedback stability
- common mode rejection
- power supply rejection
- gain compression
- intermodulation
- parasitic coupling
- ground loop
- package inductance
- electromigration
- over-voltage

I'm sure there are others I haven't thought of. To top it all off, changing one spec can ripple through and affect ten other specs. Sometimes, the inability to meet one spec means one must throw the design away and start over. And that's just an amplifier. There are many, many other analog circuits that go into a complete system. That's why analog design remains an art created by manual labour while digital circuits are synthesized, created by a software tool using RTL code as its input. By the time enough feasibility has been done on the analog design, the design is pretty much more than half-way finished.

I don't expect the digital manager to fully grasp the complexities of analog design. But at the very least, he can drop his condescension thinking that he knows it all. It's like being scolded by a child who's telling you that rocket science is easy because he just made a rocket yesterday out of paper-mâché.

I'm not big on EE class warfare. But in light of his most unwelcoming comments, perhaps I should just forward him this lovely poster (grabbed from The Amp Hour):