filtering high-frequency noise from power bus

- J
- Jim Alexander
  
  Contact options for registered users
posted
17 years ago

Mon, Oct 16, 2006 5:45 PM

I have a device involving several 4538 CMOS monostable multivibrators that has been operating fine for 6 months or so, but these multivibrators have suddenly started triggering spontaneously. I hooked up a scope, and found nothing unusal on the leading-edge trigger line, the one I am using, but found that these spontaneous triggers are correlated with some high-frequency noise on the 5V VCC bus. The unused trailing-edge trigger input is tied to VCC, which is what the spec sheet calls for. The noise has a period of about 20ns (50MHz) and a typical duration of a few hundred nanoseconds, and the voltage ranges from 10 to 50V peak-to-peak. I think this noise is causing these unintentional triggers.

I confirmed that this noise isn't coming in via the regulated 5V power supply, so I'm guessing that it's induced via EMF from an unknown source. I'm looking for advice for getting rid of this noise (as is probably obvious, I'm inexperienced with handling RF interference). Is there some way I can damp out the noise on the whole power bus, or am I better off filtering at the affected inputs? Or is shielding a better approach?

--

________ Jim Alexander __________________ jalex@cis.upenn.edu ________________
I have yet to see a problem, however complicated, which, when you looked at it
in the right way, did not become still more complicated.      -- Poul Anderson

- P
- PeteS
  
  Contact options for registered users
Vote on answer
posted
17 years ago

Mon, Oct 16, 2006 5:55 PM

In general, and without knowing details, a 0.01 uF to 0.1uF cap for each two chips is a good idea. If you have extra induced noise, then one small cap per chip close to the power pin should help.

Cheers

PeteS

- C
- Chris
  
  Contact options for registered users
Vote on answer
posted
17 years ago

Mon, Oct 16, 2006 6:34 PM

Hi, Jim. If you got a 50V spike on the power supply of 4000-series CMOS, they'd be toast in a nanosecond. A voltage spike of lesser magnitude on the power bus could easily trigger the one-shots, though. Look at your measurements carefully, and make very sure about what you're actually seeing with your scope. Always believe what your instruments tell you until you've proved otherwise -- especially if it doesn't make sense.

Tracking down noise problems to make things work can be an excruciating experience, for sure. And getting the power supply right is definitely a first step.

I'm assuming you've got a power supply external to the board, at some distance away. Before anything else, hook your power supply up to another load, and see if it still misbehaves. What you're describing sounds a little bit like a glitchy switching power supply. It may need to be replaced.

Assuming the power supply is OK, put a good elecrolytic cap (100uF to

470uF, made for low loss at higher frequencies) at the point where the power supply connects to the board. Then you want to use a good 1uF tantalum for every row of several ICs, and an 0.01uF to 0.1uF ceramic cap across the power pins of each IC. 4538 monostable multivibrators are very susceptible to glitches in the power supply, so you might just want to protect every IC that way.

Once you've done that, you should use your scope (put it on 10X) to look at the power again, and see if it's better. If you've still got problems, try replacing the ICs.

By the way, make sure *all* unused inputs are tied down -- a floating reset input could do this, too.

Long term, though, I'd recommend you look at the inputs to the ICs -- they're not protected against ESD or inputs beyond the supply rails, and that may be messing things up. If you're using these inputs to recieve signals from off the board, this could be a big problem.

Please feel free to post again with more information.

Good luck Chris