Decoupling
localroger
Posts: 3,451
in Propeller 2
I was in a conversation with a design engineer for a fairly large international company a few weeks ago, and our conversation drifted toward supply line problems and in particular capacitors.
And he went on a major rant about the amount of decoupling needed by modern CPU's, and the difficulty of sourcing proper capacitors. Tantalum is a strategic material and other materials are distinctly inferior. He said the processor they are using in their current product, a 160 MHz single core with a lot of ARM-like features, requires 10 to 20 decoupling caps. He said that people making cheap stuff like toys and games would make their prototype, then start pulling caps until it stopped working, then put a couple back in. Making industrial devices which have to perform to spec he couldn't get away with that.
I bring this up because it seems like P2 is designed from the outset not to need that much decoupling. And that could be an important advantage for some applications.
And he went on a major rant about the amount of decoupling needed by modern CPU's, and the difficulty of sourcing proper capacitors. Tantalum is a strategic material and other materials are distinctly inferior. He said the processor they are using in their current product, a 160 MHz single core with a lot of ARM-like features, requires 10 to 20 decoupling caps. He said that people making cheap stuff like toys and games would make their prototype, then start pulling caps until it stopped working, then put a couple back in. Making industrial devices which have to perform to spec he couldn't get away with that.
I bring this up because it seems like P2 is designed from the outset not to need that much decoupling. And that could be an important advantage for some applications.
Comments
The good news is many other passives can be eliminated, especially resistors, due to the dacs and what is now built into each smart pin
I wonder about the tantalum thing, though, certainly that was the case some years ago, but ceramics have come such a long way
P2 currently has Cpd ~ 1.6nF + 83pF/COG, so will have quite significant switching currents, and EMC should benefit from 4 layer board.
VIO caps will be determined by the pin activity/loads.
It may be that smaller caps are better for VDD decoupling ?
eg see https://www.eetimes.com/author.asp?section_id=36&doc_id=1320354 etc
It would be easy to replace a bank or two with 10nF (perhaps || 1uF) and see whether this affects ADC noise at all
Just looking at 22nF / 15nF / 10nF, I see price minimum is at 0402 or even 0201 packages, in 16V or 25V, and Samsung 15nF 0402 data shows a |Z| min at 80MHz & < 1 ohm 10~800MHz
lowest price in (>= 0402) is EMK105B7223KV-F Taiyo Yuden 0.022UF 16V X7R 0402 $0.0036/10k