Noise on power supply net

Hi All,
Ok, more on the dreaded color reader problems... ;-)

Now that I have the scope, been troubleshooting this thing a bit. One
thing I am seeing is that there is a lot of noise on my power net.
While the chip's specs say it should have a ripple of around 40mVs, I
am also seeing a noise signal around every 100ns of 200mV p-p. I
haven't been able to figure out what is causing it. It is there even
when the mcu is in break or asleep, so it doesn't seem to be from
therre. The LED power supply is shut down, so it shouldn't be from
there. The only things power up are the opamps and the digital pot.

I have a 10uF next to the PSU chip, another 10uF on the same rail by
the other PSU chip. I have a full copper pour for both ground and
Vcc, and 1uF bypasses next to each chip (and two by the MCU.)

So, any ideas? That noise pulse is only around 100us wide, but fairly
regular. If it was the opamp oscilating, I would have expected more
of either a square or sine wave, instead of just regular pulses...

Thanks,
Charlie

 

Hi Charlie,

--------------------------------------------------^^^^^

Hmmm... units error? "100us" pulse occurring every "100ns"
Perhaps you meant 100us occurring every 100ms?

 

Actually, they are every .1ms, and only 100ns wide...

 

Since you had clarified that it's 100usec repetion rate and 100nsec
pulse width a suspicion arises: The charge pump in idle mode, and you've
got one in there that supplies the board (U7).

http://ww1.microchip.com/downloads/en/DeviceDoc/21752a.pdf

Those things are hysteretic and theoretically should have more than one
pulse but who knows. Oh, did I mention that I don't like charge pumps?
....

One trick to figure this out is to make two probes. One a coax with
about 1/4" center conductor exposed and BNC on the other side. Wrap some
tape around the tip so you can't touch and bzzzt anything. The other is
the same but 3-4 turns soldered to the tip, maybe 1/4" diameter. You can
take a 6ft-8ft BNC cable, cut it in the middle and make those two
probes. Now plug them into your scope and hunt. Once you found the
culprit keep those probes, Murphy says you'll need them again. At least
I do, so I bought some rugged commercial ones.

Sometimes the results are rather surprising. Once at a client I found
the source of similar stuff to be <gasp> a Tektronix TDS220 scope. In
another it was an unmarked huge building across the street, with armed
guards and lots of antennas on the roof.

 

Good ideas. My scope has these little springy thingies that go on the
end of the probe for that. Using those, the noise is down to only
100mA at all the chips, and I don't see the big, regular spikes
anymore...

Wonder if I was picking up my laptop sitting next to it...

Charlie

 

Jeorge,
What do you prefer to do a little boost like this? Do you just roll
your own?

Charlie

 

When it has to be very cheap and it's for a mass product I often roll my
own but that costs a lot of engineering time. Otherwise I try to find
something reasonable that can be simulated and that generally boils it
down to a chip from LTC. Not sure about your requirements and I haven't
looked in detail, maybe this one:

http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1003,C1042,C1031,C1060,P2130,D3957

Two things are important in a noise-sensitive application:

a. It should not have hysteretic control, causes too much ripple and
thus noise.

b. Should ideally not have burst mode. While that reduces idle power it
does increase noise, often big time.

So in the example you should prefer the B version because that does not
have burst mode. Beats me why they called it "B"

 

100mV? That would still be a lot. In your application a whole lot.

 

Most annoying was an Agilent spectrum analyzer of noble heritage and
corresponding price, for EMC analysis, that produced ... you guessed it
.... EMI. Lots of it, from its display. I had to ask a client engineer
where the next hardware store was. Bought some chimney spark screen and
draped it over the whole thing.

 

Just curious, did you get this noise issue fixed?

[...]

 

Hi Joerg,
Not yet. I at present have put a .1uF cap on the net, and just now
tried replacing one of the 1uF bypasses by the MCU with a 47uF tant.
It reduces the noise to what appears to be a 20-30KHz sawtooth about
60mVs high. I think I will probably redesign to a chip like the one
you suggested using an inductor for the next version.

My next effort is to try averaging a lot of samples (maybe 20-30) to
see if that stabilizes the measurerd value.

Charlie

 

Sounds like the MCU to me. Try adding one or two 1000 pF (yes picofarad)
caps just as close to the MCU power/ground pins as possible.