Maker Pro
Maker Pro

Which small ceramic capacitors have the worst microphonics?

M

mike

Jan 1, 1970
0
Sensing a local pressure in liquid. It has a DC value and then I'll have
to measure up to 20-30Hz worth of changes. Very cramped space, hence the
0.010" width.

Would be interesting to see the mechanical configuration that gives you
a differential compression force on such a small area with large
wavelengths at 20 Hz. without bending.
I can't work on bending (which I am aware is the normal modus operandi
of a "singing cap"). I can only work with thickness changes which will
eb quite miniscule. Ideally with capacitance change because the signal
coming out of it via piezo generation is very tiny, at least here on the
bench.



Oh, I've done that a lot :)

One could buy half a truckload of reels and store them in a gigantic
nitrogen cabinet. That would last nearly forever. But the better method
is to first establish principle of operation. Then it's time to sit down
with a manufacturer. It would not be the first time where the response
is "You want to do WHAT?!". After they heard the Dollar numbers involved
that usually changed to "Oh, wow, let's see how we can do this".
Can't argue with that. Money talks!
Problem is that it only talks to the original people making the agreement.
When a cap vendor is making a million caps a minute, your puny order
for 10-million caps isn't worth the stamp it costs to reject your request.
And even if you get agreement, the guy's successor will back out
if it helps his bottom line.

I've had situations where the new guy in my purchasing augmented his bonus
by scrapping my "lifetime buy" of specially selected parts. Even if he
wanted to ask engineering, there's nobody left who knows why. I learn
about it after the production line shuts down and nobody wants to source
the "special" part. THAT 2N3904 is exactly the same as THIS 2N3904...
NO, IT ISN'T.

I can tell horror stories until you get bored...which is probably about now.
 
J

Joerg

Jan 1, 1970
0
mike said:
Would be interesting to see the mechanical configuration that gives you
a differential compression force on such a small area with large
wavelengths at 20 Hz. without bending.


It's easy. Hook up a cap to a low noise connection into a sensitive
meter. I used a Fluke 8845A. Which I bought because John Larkin said so :)

Anyhow, then clap your hands and you'll easily get a bump in the mV
range. But capacitive change is a bit more difficult although I have a
sensitive setup to measure that. Meantime I have a whole compressor and
chamber setup here in the lab.
Can't argue with that. Money talks!
Problem is that it only talks to the original people making the agreement.
When a cap vendor is making a million caps a minute, your puny order
for 10-million caps isn't worth the stamp it costs to reject your request.
And even if you get agreement, the guy's successor will back out
if it helps his bottom line.

We have a guy on board who is really, really good with contracts. He
could make sure this doesn't happen.

I've had situations where the new guy in my purchasing augmented his bonus
by scrapping my "lifetime buy" of specially selected parts. Even if he
wanted to ask engineering, there's nobody left who knows why. I learn
about it after the production line shuts down and nobody wants to source
the "special" part. THAT 2N3904 is exactly the same as THIS 2N3904...
NO, IT ISN'T.

I can tell horror stories until you get bored...which is probably about
now.


Not at all. I've live through those kind of goof-ups as well. Full line
stop, people hollering, crisis meetings, the works.
 
M

miso

Jan 1, 1970
0
I'm skeptical about component usage far outside the design parameters.
I've had enough production problems when the vendor actually improved
a published spec.
Exploiting an unspecified characteristic that's undesirable for
the typical application is asking for trouble.

Not to mention this is the kind of parameter a manufacturer will try to
improve, i.e. they would like to remove microphonics.

As I have said a hundred times, if it is not tested or GBD, you can't
count on the parameter to be stable. That is Component Manufacturing 101.
 
M

mike

Jan 1, 1970
0
It's easy. Hook up a cap to a low noise connection into a sensitive
meter. I used a Fluke 8845A. Which I bought because John Larkin said so :)

Anyhow, then clap your hands and you'll easily get a bump in the mV
range. But capacitive change is a bit more difficult although I have a
sensitive setup to measure that. Meantime I have a whole compressor and
chamber setup here in the lab.

Put a 30 Hz. low pass filter on that signal and look again.
 
J

Joerg

Jan 1, 1970
0
miso said:
Not to mention this is the kind of parameter a manufacturer will try to
improve, i.e. they would like to remove microphonics.

As I have said a hundred times, if it is not tested or GBD, you can't
count on the parameter to be stable. That is Component Manufacturing 101.

As I've said, it's about feasibility. Nothing more. If it turns out to
work, you'd be amazed what manufacturers will do if your company puts a
commensurate pile of money on the table. It's the American way :)

This is by far not the first time I do this kind of stuff. Sometimes the
unorthodox solution is what puts you in front as a company.
 
J

Joerg

Jan 1, 1970
0
mike said:
Put a 30 Hz. low pass filter on that signal and look again.


I've also had it on a waterfall spectrum display. The bulk of the
generated electrical energy was below 100Hz. But that is generation
which is not what I am ideally after. I am after capacitance changes.
 
M

mike

Jan 1, 1970
0
I've also had it on a waterfall spectrum display. The bulk of the
generated electrical energy was below 100Hz. But that is generation
which is not what I am ideally after. I am after capacitance changes.
OK, I'm surprised that a "clap" has much energy at all at 100Hz.
Learn something new every day.

I poked a random ceramic disc cap into the scope.
It's a Tek 7A22, so I've got sensitivity and filtering
out the ying-yang.
At the sensitivities required, I couldn't see any "clap" at all.
But I could see the effect of my hand waving several feet away.
With a random cap and insufficient shielding, my conclusion
is...well...inclusive.

Wonder if you could stick the cap in an arbor press and squeeze
it until you get the capacitance change you need. That might give
you some numbers on what you're up against.

<back from more experiments>

I'm fascinated by this.
I took the output of a sound level meter and plugged it into a scope.
Tek TDS540 this time.
A hand clap is all over in a couple of milliseconds...if you factor out the
room echos.
The FFT display showed a lot of low frequency energy, but was more
affected by the
windowing function selected than by the actual clap.
I've never determined anything useful at very low frequencies
from a scope FFT.

So, I plugged the sound level meter into the PC and CoolEdit.
Got pretty much the same result.
In the waterfall display, there's a lot of energy (color) showing at low
frequencies,
but it's unaffected by the clap. The clap shows up above 1kHz.
You can't sense something that isn't there.

All I know at this point is that your results differ from mine.
And mine line up with my intuition...yeah, yeah, wouldn't be the first
time my intuition led me to crash and burn. ;-)

Fascinating...I'm more interested in the pressure measurement than
whether caps are microphonic.
I'd like to measure differential air pressure with fractional Pascal
resolution near zero differential (CHEAPLY). Different animal, but might
learn something from what you're doing.

Keep us posted...
 
J

Joerg

Jan 1, 1970
0
mike said:
OK, I'm surprised that a "clap" has much energy at all at 100Hz.
Learn something new every day.

I poked a random ceramic disc cap into the scope.
It's a Tek 7A22, so I've got sensitivity and filtering
out the ying-yang.


I don't know the 7A22, got the 7A26 here. But you need to be able to
measure sub-mV.

At the sensitivities required, I couldn't see any "clap" at all.
But I could see the effect of my hand waving several feet away.
With a random cap and insufficient shielding, my conclusion
is...well...inclusive.

You have to use an SMT cap, not something potted in a disk-type
structure. IOW the layer stack needs to be directly exposed. I found
that leaded parts have very little microphonics.

Wonder if you could stick the cap in an arbor press and squeeze
it until you get the capacitance change you need. That might give
you some numbers on what you're up against.

Yeah, I'll probably do some more experiments but only after I buy some
modern very high density caps. Because I need a change in the thousands
of pF over a couple of psi.

<back from more experiments>

I'm fascinated by this.
I took the output of a sound level meter and plugged it into a scope.
Tek TDS540 this time.
A hand clap is all over in a couple of milliseconds...if you factor out the
room echos.
The FFT display showed a lot of low frequency energy, but was more
affected by the
windowing function selected than by the actual clap.
I've never determined anything useful at very low frequencies
from a scope FFT.

Oh, I did. The worst was microphonics. Client had tried on their own and
even a fancy expensive Stanford Research Analyzer saw ... nothing. So
here I was with a laptop and an FFT program. 7-8 Hertz showed up, not
too stable. Hmmm, what the heck could that be when every clock in the
system is crystal controlled? Told the client it must be the wind pulses
from the fan blades causing microphonics in the caps. "Now that's
voodoo, you are kidding, right?". So I gently leaned the palm of my hand
onto the fan blade center, slowed it down, and sure enough the noise
spectrum moved lower in frequency. When I stalled the fan it was gone.
Some jaws dropped.

So, I plugged the sound level meter into the PC and CoolEdit.
Got pretty much the same result.
In the waterfall display, there's a lot of energy (color) showing at low
frequencies,
but it's unaffected by the clap. The clap shows up above 1kHz.
You can't sense something that isn't there.

All I know at this point is that your results differ from mine.
And mine line up with my intuition...yeah, yeah, wouldn't be the first
time my intuition led me to crash and burn. ;-)

:)

Try to cup your hands so you get more energy at the low end instead of
the high pitched applause type of clap. Or pop an inflated paper bag
instead.

Fascinating...I'm more interested in the pressure measurement than
whether caps are microphonic.


That makes two of us. However, as a generator caps can never be very
good at low frequencies and there is no DC. Only capacitance change
could do that.

I'd like to measure differential air pressure with fractional Pascal
resolution near zero differential (CHEAPLY). Different animal, but might
learn something from what you're doing.

Keep us posted...


Sure. But I can only reveal what's covered in our patent applications at
this point. I've actually got a compressor, pressure chamber, pulsating
ports, and so on in the lab right now. I wish those compressors wern't
so freaking loud.
 
Joerg said:
I've actually got a compressor, pressure chamber, pulsating ports, and
so on in the lab right now. I wish those compressors wern't so
freaking loud.

One of the local used office furniture places has several old printer
enclosures in stock for a couple of dozen dollars... a steel box with
foam on the inside, for quieting old impact printers. Most of them even
come with a free power strip. :) If the compressor is on the order of
"blow up beach balls quickly and tires slowly", it would fit in one of
these. If it's more like "run 5 half-inch impact wrenches at once",
then the usual solution is to banish it to the great outdoors.

Matt Roberds
 
S

Spehro Pefhany

Jan 1, 1970
0
Sure. But I can only reveal what's covered in our patent applications at
this point. I've actually got a compressor, pressure chamber, pulsating
ports, and so on in the lab right now. I wish those compressors wern't
so freaking loud.

Unfortunately, compressors tend to be either very, very expensive,
3-phase and quiet (eg. a Quincy or scroll compressor) or don't give
much air (like a Paasche airbrush compressor). I think dentists have
some quiet ones too, probably not cheap.

The absolute worst are the oil-less cheap rattle and roar hobby grade
compressors.. very antisocial they are.


Best regards,
Spehro Pefhany
 
J

Joerg

Jan 1, 1970
0
One of the local used office furniture places has several old printer
enclosures in stock for a couple of dozen dollars... a steel box with
foam on the inside, for quieting old impact printers. Most of them even
come with a free power strip. :) If the compressor is on the order of
"blow up beach balls quickly and tires slowly", it would fit in one of
these. If it's more like "run 5 half-inch impact wrenches at once",
then the usual solution is to banish it to the great outdoors.

Matt Roberds

That's what we used (and sometimes built) for our old telex machines in
the ham radio days. But I just don't have the sapce for it here. So I
charge the compressor in the garage, disconnect, and haul it in. Gives
me some exercise as well.
 
J

Joerg

Jan 1, 1970
0
Spehro said:
Unfortunately, compressors tend to be either very, very expensive,
3-phase and quiet (eg. a Quincy or scroll compressor) or don't give
much air (like a Paasche airbrush compressor). I think dentists have
some quiet ones too, probably not cheap.

The absolute worst are the oil-less cheap rattle and roar hobby grade
compressors.. very antisocial they are.

And that's exactly the kind I have :)

But since I am a very social guy I charge it in the garage with the
doors down and then carry it inside. The setup is a bit leaky but it
holds for about an hour's worth of experimenting.
 
R

rickman

Jan 1, 1970
0
I should have given some reasoning behind this. Sorry.

The high capacitance in small size can only be achieved, AFIK, with high
dielectric constants. That means a lot of variation with applied voltage
and with mechanical variation as well. They are transducers that were
not designed as such.

I know I'm coming late to the party, but let me ask a question. Over
some years, the density of ceramic capacitors gets higher and higher.
Just a couple of years ago I found a part mis-spec'd on one of my boards
that I had to up the voltage without changing the PCB. So we stuck an
0805 on an 0603 footprint. Just recently I found I can now get the
proper voltage in an 0603 footprint and get back on the straight and
narrow with my assembly house.

So what are they changing that allows this? My understanding is they
are making the layers thinner which ups the capacitance in a given
volume, not changes to the formulation of the ceramic. But what allows
thinner layers? Is it small tweaks to the formulation?

I ask this because if you want to utilize a parameter that is not spec'd
by the manufacturer, you need to believe this won't change without
notice. If they tweak their formulation at a later date, this may mess
up a design depending on the non-spec'd and non-controlled parameter.
Or is this a one-off?
 
R

Robert Baer

Jan 1, 1970
0
rickman said:
I know I'm coming late to the party, but let me ask a question. Over
some years, the density of ceramic capacitors gets higher and higher.
Just a couple of years ago I found a part mis-spec'd on one of my boards
that I had to up the voltage without changing the PCB. So we stuck an
0805 on an 0603 footprint. Just recently I found I can now get the
proper voltage in an 0603 footprint and get back on the straight and
narrow with my assembly house.

So what are they changing that allows this? My understanding is they are
making the layers thinner which ups the capacitance in a given volume,
not changes to the formulation of the ceramic. But what allows thinner
layers? Is it small tweaks to the formulation?

I ask this because if you want to utilize a parameter that is not spec'd
by the manufacturer, you need to believe this won't change without
notice. If they tweak their formulation at a later date, this may mess
up a design depending on the non-spec'd and non-controlled parameter. Or
is this a one-off?
Yeah; off-datasheet uses are a PITA, depending what attribute is
needed, a very minor fab change can kill your product.
Been there.
 
J

Joerg

Jan 1, 1970
0
rickman said:
I know I'm coming late to the party, but let me ask a question. Over
some years, the density of ceramic capacitors gets higher and higher.
Just a couple of years ago I found a part mis-spec'd on one of my boards
that I had to up the voltage without changing the PCB. So we stuck an
0805 on an 0603 footprint. Just recently I found I can now get the
proper voltage in an 0603 footprint and get back on the straight and
narrow with my assembly house.

So what are they changing that allows this? My understanding is they
are making the layers thinner which ups the capacitance in a given
volume, not changes to the formulation of the ceramic. But what allows
thinner layers? Is it small tweaks to the formulation?

I believe it's both the ceramic formulation and also layer thickness, a
more controlled production process.

I ask this because if you want to utilize a parameter that is not spec'd
by the manufacturer, you need to believe this won't change without
notice. If they tweak their formulation at a later date, this may mess
up a design depending on the non-spec'd and non-controlled parameter. Or
is this a one-off?

In my case it's feasibility first. Then comes the business part of it.
Two avenues for that:

a. (preferred) Line up a deal with a manufacturer where they guarantee a
legacy process to be available for x or maybe xx years. Negotiate a high
enough per-piece price that makes it worth their while. Not a problem in
this case, we'd be prepared to pay several Dollars per capacitor if it
performs. Versus the usual $0.005.

b. Buy an obscene amount of caps and store them under nitrogen. CFOs
frown upon such ideas but in this case the cost would be relatively
miniscule.
 
F

Frank Miles

Jan 1, 1970
0
I know I'm coming late to the party, but let me ask a question. Over
some years, the density of ceramic capacitors gets higher and higher.
Just a couple of years ago I found a part mis-spec'd on one of my boards
that I had to up the voltage without changing the PCB. So we stuck an
0805 on an 0603 footprint. Just recently I found I can now get the
proper voltage in an 0603 footprint and get back on the straight and
narrow with my assembly house.

So what are they changing that allows this? My understanding is they
are making the layers thinner which ups the capacitance in a given
volume, not changes to the formulation of the ceramic. But what allows
thinner layers? Is it small tweaks to the formulation?

I ask this because if you want to utilize a parameter that is not spec'd
by the manufacturer, you need to believe this won't change without
notice. If they tweak their formulation at a later date, this may mess
up a design depending on the non-spec'd and non-controlled parameter. Or
is this a one-off?

One change over the years is a loosening of the voltage coefficient. I
dug up some information on some typical X7R parts from about 30 years ago
- the capacitance reductions were much less than what is common with
today's small SMT parts. It seems that voltage ratings for these low-
voltage parts has never been about "breakdown" - which has typically been
MUCH higher than the spec - it's been value tolerance. You can get
around this problem by specifying a part with a high-enough voltage, but
you may need a physically larger part. There probably have been changes
to the ceramic formulation, but they've not been solely responsible for
the size reductions.

And as others have said, depending on unspecified characteristics is
risky. The importance of risk avoidance is dependent on issues of
product volume and production longevity as well as design tractability.
 
J

Joerg

Jan 1, 1970
0
Frank said:
One change over the years is a loosening of the voltage coefficient. I
dug up some information on some typical X7R parts from about 30 years ago
- the capacitance reductions were much less than what is common with
today's small SMT parts. It seems that voltage ratings for these low-
voltage parts has never been about "breakdown" - which has typically been
MUCH higher than the spec - it's been value tolerance. ...


Yep. Many of today's high density X7R caps are almost as bad as Y5V when
it comes to loss of capacitances at full rated voltage. Easily 70%,
something that used to be unheard of with X7R.

... You can get
around this problem by specifying a part with a high-enough voltage, but
you may need a physically larger part. There probably have been changes
to the ceramic formulation, but they've not been solely responsible for
the size reductions.

I think there have also been improvements in production precision that
allows the manufacturers to specify a higher voltage for a given layer
thickness. Maybe more uniform, lower tolerances.

And as others have said, depending on unspecified characteristics is
risky. The importance of risk avoidance is dependent on issues of
product volume and production longevity as well as design tractability.

This is one of those projects where risk avoidance would get us nowhere.
 
J

josephkk

Jan 1, 1970
0
My suggestion, Joerg, is to get the greatest capacitance in the smallest
size to start with.

Smallness doesn't necessarily help. Physically larger parts thus longer
reaches are more likely to experience the stresses to produce strong
piezoelectric outputs. Actually i would recommend a modified
piezoelectric sounder as the base of the device. Perhaps unpackage some
small ones (under 1/2 inch dia case) and see what they can do.

?-)
 
Top