Socketed semiconductors
Here's a long story to illustrate a point about sockets:
When Tektronix began to produce solid state oscilloscopes, they socketed ALL of their transistors. When ICs were developed and used in the scopes, they socketed ALL of the ICs and transistors. In fact, in the "new" 400-series portables (e.g., 465, 475, 485), they even socketed the four tunnel diodes in the trigger circuits. The number of sockets exploded with the advent of logic analysis. All of the logic analyzers with their S-TTL and ECL logic were fully socketed. Needless to say, service was easy on all of these instruments. All we techs had to do was pluck out a transistor and throw it on the curve tracer to check it and put it back if it was good.
But sometime you'd get a problem where you'd pluck out a transistor, it checked good, you'd put it back in the instrument and the instrument would start working correctly. Intermittent transistor? Intermittent socket? Cold solder joint? So, you'd replace transistor AND socket to make sure. At Tektronix, we strove to fix the problem right the first time so that there were no callbacks and good customer satisfaction.
Enter the Japanese. They were threatening the low-end scope market. The Japanese had a marvelous track record. They would introduce products on the low end of the price range and capture that market and then begin working their way up the product line until they had cornered the marked share on it all, whether the cheap stuff or the expensive stuff. Don't believe that? Just check out what happened to the U.S. market for cameras, stereos, TVs, calculators, radios, tape recorders, video recorders, automobiles, etc.
Several manufacturers (Tek, Hewlett-Packard, et. al.) got together to decide if they could counter this oscilloscope threat. With the exception of Tektronix, none of them had the desire to commit corporate resources to a fight that they could very possibly loose. Alone, Tek went ahead with their project to counter the Asian invasion. Their goal was to produce an oscilloscope that was inexpensive to manufacture, yet was still high-performance and reliable. They had earlier replaced the Telequipment (a wholly-owned U.K. subsidiary of Tek) line with the notorious T900 series that resembled Hoover vacuum cleaners. This new thrust was in a different direction and the 2200-series of oscillosopes was the result.
To save money, the entire scope was made on one large circuit board. This board was then scored and snapped to form two other boards. One corner was folded up at a 90° angle to become the front panel board and another corner was removed to become the attenuator/preamplifier board. Knobs were reworked to be one-piece knobs with no setscrews or brass inserts. All sockets were eliminated as were nearly all connectors. A new high-efficiency power supply was used. The result was a very light-weight, low-cost, high-performance, reliable scope.
The first thing the Japanese had to do was lower the cost on all their scopes to below their manufacturing cost in order to sell them under the Tek price. Then they had to do a major redesign. The result? Tek succeeded in countering the threat. You don't see Asian scopes that can compete with the Tektronix product line. There are a lot of Chinese and Korean scopes out there, but in comparison, they're junk.
Now to the moral of the story. Tek, through all of its years of research, discovered that fully 80% of the problems they were having with their products was due to intermittent or open sockets and connectors. The high-end 7000-series and 400-series portables were socketed and were plagued with intermittents. The lower-end 5000-series and TM500-series instruments had few, if any sockets and had few problems of that type. By eliminating sockets and connectors, the entire Tek product line improved tremendously in reliability and the company saved a lot of money by not using sockets, saving another manfacturing step and eliminating a lot of parts.
The only parts that Tek ended up socketing were ROMs and EPROMs that might need periodic updates and expensive (at the time) microprocessors.
Do I socket ICs? When making PCBs, absolutely not. "B" series CMOS are much more forgiving that the older "A" series, so don't need sockets. Now, I DO use sockets if I build a circuit on perf board with point-to-point wiring, simply because replacing a defective IC (they are sometimes bad right out of the sleeve) would other wise be a very destructive process with that method of prototyping. And if you're using wire-wrap, you don't have any choice but to use sockets.
Dean