Troubleshooting 'singing' opamp!

[Charlie E.]

Hi All!
Ok, at my wits end, because I am afraid I just wasted $250 on a pcb
that isn't going to work. However, I am hoping one of ya'll can give
me some insight!

I am building a small board that has speech output. It goes through
two stages of an MCP6024 for a four stage audio filter, and then to a
power amp.

My hand build proto worked fine, as does the original circuit I
'borrowed' the circuit from. But, when I power this guy up, he sings
like a canary. The new circuit is all surface mount, SOIC and 0805
parts. All the passives are mounted with in an inch of the opamp, and
I have a full pour of ground on the bottom and VCC on the top.

I found one component error (when ordering a 4.7nF cap, I actually
ordered a 47nF cap...) but I had a TH part of the same value I have
tacked in its place. This didn't effect the problem, though.

So, any ideas on trouble shooting this? Before the boss shoots me????

Thanks,
Charlie

 

[Spehro Pefhany]

Hi All!
Ok, at my wits end, because I am afraid I just wasted $250 on a pcb
that isn't going to work. However, I am hoping one of ya'll can give
me some insight!

I am building a small board that has speech output. It goes through
two stages of an MCP6024 for a four stage audio filter, and then to a
power amp.

My hand build proto worked fine, as does the original circuit I
'borrowed' the circuit from. But, when I power this guy up, he sings
like a canary. The new circuit is all surface mount, SOIC and 0805
parts. All the passives are mounted with in an inch of the opamp, and
I have a full pour of ground on the bottom and VCC on the top.

I found one component error (when ordering a 4.7nF cap, I actually
ordered a 47nF cap...) but I had a TH part of the same value I have
tacked in its place. This didn't effect the problem, though.

So, any ideas on trouble shooting this? Before the boss shoots me????

Thanks,
Charlie

I note that they consider 60pF to be a large capacitive load...

Is there another op-amp you can use?

 

[Charlie E.]

Aha. This is a known problem. Don't use the quad opamp as the four stage
filter, especially if it is the shaky R-R one. Those things are
extremely sensitive to capacitive loads.


Sings in the audio band or somewhere in ~MHz area?


Don't know what kind of performance are you looking for (do you really
need MCP6024 for speech? May be LM324 would do), however you can
probably try OPA4350 as the direct replacement.



Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

Hi All,
Since I can't do the ascii art thing, I will have to see about posting
to my website...

Ok, easier said than done...

While I work on that, the singing is in the audio band. The output
load is a series .47uF cap. My present investigations are two caps
from their respective IN+ pins to ground. They ended up about 1" and
1.5" from the chip...

Thanks for the help!

Charlie

 

[Charlie E.]

Aha. This is a known problem. Don't use the quad opamp as the four stage
filter, especially if it is the shaky R-R one. Those things are
extremely sensitive to capacitive loads.


Sings in the audio band or somewhere in ~MHz area?


Don't know what kind of performance are you looking for (do you really
need MCP6024 for speech? May be LM324 would do), however you can
probably try OPA4350 as the direct replacement.



Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

Ok, now have a hidden downloads page on my website with a couple of
PDFs.

http://edmondsonengineering.com/downloads.aspx

Thanks for any insight you can give! The filter section is in the
middle of the bottom of the board, to the right of the SOIC14 opamp.

Charlie

 

[Charlie E.]

Icky! Sallen-key :-(

...Jim Thompson

Yeah, but it worked!

Actually, seem to have fixed it. Moved one of those caps (the
farthest one) and tacked it right on to the pin. It is a temporary
fix, but these first 10 boards are just for final software and
marketing demos. Will have to fix this on the production versions!


Thanks all for your help!

Charlie

 

[Charlie E.]

Your schematic doesn't indicate any power supply bypass or decoupling --
did you forget the most important caps of all? _Always_ bypass your
power supplies -- digital circuits will act weird without them, and
analog circuits will oscillate. You should have 10 - 100nF caps as close
to the op-amp power supply pins as you can get them, and if the power to
the amp isn't clean you should seriously consider decoupling the +V line
with a 100-or-so-ohm resistor (after verifying voltage drop, and the knee
of the RC network you're making).

The trace lengths that you mention elsewhere on C5 and C10 shouldn't make
that big of a difference.

Hi Tim,
Yes, there is a bypass of 1.0uF just above the chip. I think it is on
the schematic by the A section of the quad opamp. Do you think 1.0uF
is too much? I also have ground pour on the bottom and VCC pour on
the top...

Charlie

 

[Charlie E.]

1) Teach it Aida.
2) Cut some traces in one board to isolate the various stages.
Haywire (tack) input and out terminators to each stage; at minimum
one meg each place (leaded parts become rather useful here).
Test each stage separately.
Start at output stage and connect previous stage and re-test; stop
sequential re-connections like this near middle.
Then start at input, connecting second stage and re-test; then
connect next stage / retest etc and stop in middle.
At this point, if your luck is same as mine, all will be perfect
UNTIL you re-connect that magic middle I/O....
3) Teach it the Ring..

just an update on this MPP (Major Posterior Pain...)

I found that part of my problem was that, for some unknowable reason,
when these were being prototyped, the clock was being set incorrectly.
the PLL was being turned off, giving me a clock divided by 4. this
put the PWM frequency right into my passband, and may have been a part
of the problem. While trying to troubleshoot this, I started getting
debugging and programming problems i.e. cannot validate the chip. I
am now just about ready to drop the PIC and start over with a PSOC...

Charlie