A guy asked me how to optimize a bridge circuit. His design had eight op amps and over a dozen precision resistors. I showed him how to minimize the number of (expensive) resistors and precision op amps. I got it down from eight to four to two. Then I even figured how to get the number of op amps down to one.
But before we finished, I had to inquire how many bridge circuits he was planning to make. That's important. If you're only going to make five or 10 or 20 of them, you will probably want to just get the circuit built and running and evaluate the system you are going to run. The cost per unit is not important. The time you waste optimizing it could be very important.
If you're planning to make thousands of them, you'll want to put in some extra engineering effort to figure out how to optimize and minimize the cost of each component—which usually isn't important if you're going to build only a few. The cost of parts, the assembly labor, and the trim-and test work will be quite important for a large-volume design.
So when a guy asks for help, I have to explain this problem. Do engineers understand this? The good, experienced ones do. The new ones (the ones who have to ask a lot of questions) have to learn. Often, trying to be very frugal and using the cheapest parts is poor economy. Spending more for some precision parts may save you a lot of grief.
It's very much the same as a guy asking, "How can I get the best low-noise amplifier?" As I always respond, "What's the bandwidth, what's the source impedance, and what's the size of the smallest and largest signals?" Without that information, you can't optimize anything. So the planning is really important.
PLAY BY THE RULES—IF YOU KNOW THEM
When I was back at Philbrick, we had several plans for optimizing a potted module. One plan was to optimize parts cost plus 150% of assembly labor at 2 cents per second. Back in 1970. But one day, I was told my new circuit was badly designed because I was violating the new rules. "What new rules?" I asked.
They forgot to tell me that they had changed the rules. It was now 150% of parts plus 450% of labor and assembly costs— and they hadn't bothered to tell me the rules had changed. The word infuriated isn't strong enough to describe my mood. How the heck am I supposed to do my job with poor information?
Another time, they changed the "rules" so that the cost of a jumper went up (or down), while the cost of a double-sided pc board went down (or up). But they didn't tell the engineers. So I confronted our manager, Richard. "These new rules mean I and said, "No, that would be wrong. Let me get that fixed."
Well, in a couple of weeks, Richard was gone, and the question wasn't solved. And a month later, I was gone. I walked out on the last day of 1976. If you're hired to optimize new designs, but the rules keep changing—and they forget to tell you the rules—hey, you have to walk out.
PLAY BY THE RULES, PART 2
My friend Arnold had designed a very good high-voltage amplifier with ±100-V output swing, the Teledyne Philbrick 1022. But the guys in marketing decided they needed a low-price version to fill a high-volume need.
So Arnie took all the rules and figured out how to use a more spacious layout (so things weren't packed in so tight) with lower assembly costs and lower-cost parts to make the Model 1032, with definite cost improvements.
But after it was put into production, the Manufacturing Department decided to re-interpret the rules, and the Model 1032 was listed as more expensive than the 1022. Arnie couldn't win. He bailed out, too.
Check out Wideband Amplifiers by Peter Staric and Erik Margan. Available from Springer.com for about $159, this 612-page book covers all aspects of high-frequency amplifier design. It is now in its second edition after correcting a few typo errors. The first printing sold out. I like its attitude and its insights. It covers theoretical and practical applications, computer-aided design and filters, real semiconductors, and integrated circuits. It also covers bipolar transistors, FETs, fast op amps, and current-feedback amplifiers. I'll recommend it for anybody working in this field.