When I started designing things, I used a 20MHz scope to debug a 2MHz processor. Sometimes it was a cheap 5MHz Heathkit scope on a now ancient LM741 op Amp. Digikey still sells the same National Semiconductor part in DIP-8 and TO-5 metal can packages. Hard to believe.
Anyone? Anyone? Anyone?
Speed means something a little different now. I recently ran across an interesting article on the Analog Devices web site; "A Practical Guide to High-Speed Printed-Circuit-Board Layout." It has a lot of good layout hints (duh) but what caught my eye was the discussion on paralleling capacitors. Back in the day, for the most part, you'd just put a .01uf cap on each logic chip and be done with it. Even with power supplies, it was pretty much just "put as much as you have space and budget for. If you've got ripple, put in more". OpAmps required a little more work to determine capacitor values but in a lot of areas you could get by without much engineering when selecting capacitors.
Recent forays into motor control have opened my eyes to how different things are now. It's pretty much the same with switching power supplies - that's really what a PWM motor driver is anyway.
We used to parallel up capacitors just to increase the value. Now, however, there are other reasons. Most PWM circuits really need low ESR (effective series resistance) capacitors. With too much ESR, you'll lose some volts in the caps, they heat up and maybe even explode. Other bad things can happen too, but that's enough of a representative sampling. Putting caps in parallel does the same thing with ESR that putting resistors in parallel does with ordinary R. That's the main reason switching power supplies have six or eight or more electrolytics instead of just one big one.
The other reason to put caps in parallel has to do with differing frequency responses. Again, something I didn't worry about too much. This one is important for high speed analog circuits and PWM circuits such as motor control and power as well. You can combine two or more different value capacitors, including a mix of electrolytic and ceramic to cover a range of frequency responses. Put the smallest (both in physical size and value) closest to the chip and with the shortest path to ground. The article goes into a lot more detail, but that's the gist of it.
Yes, but can you parallel park?