Which small ceramic capacitors have the worst microphonics?

[Joerg]

Does anyone know which tiny MLCC have the worst microphonics? Brand, series?

I want to do some tests to see how well these things can sense low pitch
sound, mainly via changes in capacitance but also piezo effects. Main
reason is that I need a fairly low impedance, high capacitance and a
very small size. Ideally less than 0.010" width and height but for right
now I can test with larger ones. 0.050" or so would be fine for testing
the concept, maybe even a little larger. A capacitance higher than
0.01uF would be nice. The frequency response doesn't have to go beyond
100Hz.

--
Regards, Joerg

http://www.analogconsultants.com/

[Jim Thompson]

On Sun, 14 Apr 2013 09:55:19 -0700, Joerg <inv...@invalid.invalid>
wrote:

Wouldn't that most likely be X7R?

...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |

I love to cook with wine. Sometimes I even put it in the food.

[John S]

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

[John S]

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.

Cheers

[whit3rd]

On Sunday, April 14, 2013 9:55:19 AM UTC-7, Joerg wrote:
> Does anyone know which tiny MLCC have the worst microphonics? Brand, series?
>

Well, the worst microphonics will be for quartz crystals and ceramic resonator
equivalents (because they are poled). Microphonic response at frequency F
depends on the crystal being asymmetric (so that it knows up from down).
A non-polarized MLCC would see sound input at F and give feeble electrical
output at 2F. Raise the bias voltage on the MLCC , maybe beyond the recommended
applied voltage, to see electrical output at F.

Quartz crystals and ceramic resonators are usually packaged so you cannot
stress them with external soundwaves, but the internal elements are sensitive.
Ceramic resonators are permanently 'poled' with no need of a bias voltage source,
and alpha-quartz has asymmetric crystal structure which amorphous polycrystalline
quartz lacks. Sadly, you can't make a halogen lamp envelope into a good
microphone.

[Joerg]

Yes, that's what I probably do. But it is often surprising, one brand
has 2x the microphonics and all else is pretty much the same. So I was
wondering if people knew something, such as "brand XYZ always has tons
of microphonics" or something like that.

[Joerg]

I can't have any bias voltage in this case. But to my surprise I just
read a report where a C0G cap is as "bad" in microphonics as the
ferroelectric X7R sibling. This turf seems to be full of surprises.

[Joerg]

Jim Thompson wrote:
> On Sun, 14 Apr 2013 09:55:19 -0700, Joerg <inv...@invalid.invalid>
> wrote:
>
>> Does anyone know which tiny MLCC have the worst microphonics? Brand, series?
>>
>> I want to do some tests to see how well these things can sense low pitch
>> sound, mainly via changes in capacitance but also piezo effects. Main
>> reason is that I need a fairly low impedance, high capacitance and a
>> very small size. Ideally less than 0.010" width and height but for right
>> now I can test with larger ones. 0.050" or so would be fine for testing
>> the concept, maybe even a little larger. A capacitance higher than
>> 0.01uF would be nice. The frequency response doesn't have to go beyond
>> 100Hz.
>
> Wouldn't that most likely be X7R?
>

One would think so because it's ferroelectric but I found a Kemet study
showing it may not be so in all cases.

[Klaus Kragelund]

Perhaps the brands from the unknown (read Chinese) manufactors may be worse since they might not have inhouse control to minimize that parameter.

Is this for a one-off or a production item. I wouldn't sleep nights if it were for a production item.

Regards

Klaus

[Joerg]

Long term it is for production but right now only a feasibility study.
If it works we can then determine a pathway to obtain caps of consistent
behavior. If necessary a custom run. Some deviation is fine because
every unit will be calibrated.

I always sleep well :-) ... Drives others crazy, I can say good night,
turn around, and 15 seconds later ... zzzzzzz

[klaus.k...@gmail.com]

I'm the same way, even during periods of high stress on the job.

But, I have been told that could be a signal that one has sleep apnea, and it could be dangerous.

Some specialist say that if you can fall asleep fast, that is because you get to little sleep and you should get more sleep. I get along fine with 6 hours per night, but I may just be kidding myself....

Regards

Klaus

[Joerg]

I thought all Vikings could do that :-)

> But, I have been told that could be a signal that one has sleep
> apnea, and it could be dangerous.
>

Yep, especially if you have cases in the immediate family. The only way
to find out would probably be in one of those sleep diagnosis centers
that some hospitals run. Essentially you have to stay overnight.

> Some specialist say that if you can fall asleep fast, that is because
> you get to little sleep and you should get more sleep. I get along
> fine with 6 hours per night, but I may just be kidding myself....
>

6h is quite low. I try to get in 7h which isn't so easy anymore. Since
California had a business exodus most of my clients are now 2-3
timezones ahead. This means I now have to get up latest by 6:30am. And
then the dogs want to be walked as well.

[mook johnson]

hasn't been my expereince. I once has some large like .5X x 1" stacked
X7Rs from AVX in the lab as motor bus capacitors. When I ran the motor
I could hear the RPM of the motor from the cap back across the lab. :)

Replaced them with Kemet, Novacap and another custom cap supplier, and
they were very similar.

Can you gain some sensitivity by tuning the mounting structure (pcb I
assume) to have a high mechanical Q at the frequency if interest?

[Joerg]

I've had differences and those most likely had to do with the number of
layers and things like that. Some caps were taller than others.

> Can you gain some sensitivity by tuning the mounting structure (pcb I
> assume) to have a high mechanical Q at the frequency if interest?

Unfortunately not in this case because I have to measure from DC for
30-40Hz.

[miso]

Much like chip design, you will probably find different parts of the PCB
have more flex, which in turn effect the cap. Usually the corners are
considered to be the quietest.

If you can't put the critical cap in a corner and if the PCB is
rectangular. there is probably more flex in the long direction of the
board, so place the caps perpendicular to the long edge.

There may be a way to parallel caps so the microphonics cancel.
Obviously that would take more caps of a smaller value with different
orientations.

If you dig into chip design books and papers, you run into this kind of
black magic.

Perhaps flexible (elastomer) mounting of the PCB?

[Joerg]

I actually want microphonics in this case and the cap won't be on a PCB.
Some tests yesterday with what I have here showed that while slight
pushing onto the surface of the cap does generate a detectable signal I
could hardly measure any capacitance change. Capacitance change would be
the only way to get DC out of it.

[Klaus Bahner]

On 14-04-2013 21:54, Joerg wrote:
[snip]

>>
>
> I can't have any bias voltage in this case. But to my surprise I just
> read a report where a C0G cap is as "bad" in microphonics as the
> ferroelectric X7R sibling. This turf seems to be full of surprises.
>

All of these ceramic materials *are* ferroelectric, otherwise you
wouldn't get the high epsilon_r. Slight changes in the composition of
the compound affects temp coefficients and other properties, hence the
classification into X7R, C0G etc., but they are all ferroelectrics.

I think it is impossible to deduce microphonics from these general
material properties at all. For example I would assume that manufactures
use tricks like varying the orientation of the layers in a multilayer
cap to minimize the overall microphonics effect.

So most likely you have to contact the manufacturers and try to get into
contact with one of the component design engineers knowing the gory
details - which most likely is easier said than done :-(

Another loose ideas could be the IEEE Trans. Ultrasonics, Ferroelectrics
and Frequency Control. I seem to remember an article on capacitor
microphonics there, but a quick search didn't find anything. Most likely
my memory doesn't serve me well here.

Or look for PVDF (Polyvinylidene flueride) sensors, usually used as
piezo, but the should make a good microphone, too. Perhaps there is
something on the market which can be mounted on a PCB. The price tag
might be much higher than a capacitor, though.

Regards,
Klaus

[Joerg]

Klaus Bahner wrote:
> On 14-04-2013 21:54, Joerg wrote:
> [snip]
>>>
>>
>> I can't have any bias voltage in this case. But to my surprise I just
>> read a report where a C0G cap is as "bad" in microphonics as the
>> ferroelectric X7R sibling. This turf seems to be full of surprises.
>>
> All of these ceramic materials *are* ferroelectric, otherwise you
> wouldn't get the high epsilon_r. Slight changes in the composition of
> the compound affects temp coefficients and other properties, hence the
> classification into X7R, C0G etc., but they are all ferroelectrics.
>
> I think it is impossible to deduce microphonics from these general
> material properties at all. For example I would assume that manufactures
> use tricks like varying the orientation of the layers in a multilayer
> cap to minimize the overall microphonics effect.
>
> So most likely you have to contact the manufacturers and try to get into
> contact with one of the component design engineers knowing the gory
> details - which most likely is easier said than done :-(
>

My experience with that is that they won't release much, this is usually
closely guarded internal reseach.

> Another loose ideas could be the IEEE Trans. Ultrasonics, Ferroelectrics
> and Frequency Control. I seem to remember an article on capacitor
> microphonics there, but a quick search didn't find anything. Most likely
> my memory doesn't serve me well here.
>

Good idea. I am a member of that society but since the Transactions no
longer come on paper it's easier to miss stuff when you don't make
regular checking a habit. A paper kind of pushes itself under the nose
for the evenings but the computers are off in our house after 6:30pm or so.

> Or look for PVDF (Polyvinylidene flueride) sensors, usually used as
> piezo, but the should make a good microphone, too. Perhaps there is
> something on the market which can be mounted on a PCB. The price tag
> might be much higher than a capacitor, though.
>

In this case it can be several Dollars per capacitor. But PVDF would not
offer very large capacitances when small (a few hundred um by 1mm or so
in real estate). I remember some PVDF experiments we did in medical
ultrasound. We never warmed up to that material. It had great bandwidth
which is important in that market but shows quite poor acoustic
efficiency. So we kept PZT.

[Ian Field]

"Joerg" <inv...@invalid.invalid> wrote in message
news:at05c4...@mid.individual.net...

You can easily test samples yourself.

A couple of years ago I submitted a TL431 based electret mic booster to
Elektor magazine, which was published.

Among my experiments; I stripped a spare electret mic to get at the special
JFET inside, I examined the microphonic properties of various components
including bog standard capacitors of various types.

In case you have difficulty locating a copy of the article:

Give the TL431 a Vcc of 9 - 12V and a cathode load of at least 200 Ohms,
connect the electret from cont ip to GND and set the K to 1/2 Vcc with a 47k
from K to cont ip.

Measure the resistance - the final resistor gives large nfb, both AC & DC,
you need to shunt the AC nfb to get very high gain with stable DC set point.
connect a 100uF cap to GND and a 1k2 from K to the + of the cap, subtract
that 1k2 from the measured resistance and put a fixed resistor equal to the
remainder from cap + to cont ip.

If yoy break open the electret mic capsule, you can remove the electret
dielectric and attach wires to connect your capacitor.

Pretty much all types of non-electrolytic capacitor are microphonic when
tapped with a screwdriver - you will soon see, some more than others.

If the finished circuit picks up radio stations, your electrolytic has low
ESR - put a 100 Ohm resistor in series with it.

[mike]

On 4/14/2013 9:55 AM, Joerg wrote:

You don't disclose the application, and the devil is always in the details.

I have a couple of concerns.

A device that small and stiff is gonna have about zero sensitivity
below 100Hz. To make it work, you'll have to "bend it" by attaching
it to something bigger and flexible. Operational bending is bad enough.
Shock/vibration/resonance is gonna be a reliability nightmare.

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.

[Joerg]

mike wrote:
> On 4/14/2013 9:55 AM, Joerg wrote:
>> Does anyone know which tiny MLCC have the worst microphonics? Brand,
>> series?
>>
>> I want to do some tests to see how well these things can sense low pitch
>> sound, mainly via changes in capacitance but also piezo effects. Main
>> reason is that I need a fairly low impedance, high capacitance and a
>> very small size. Ideally less than 0.010" width and height but for right
>> now I can test with larger ones. 0.050" or so would be fine for testing
>> the concept, maybe even a little larger. A capacitance higher than
>> 0.01uF would be nice. The frequency response doesn't have to go beyond
>> 100Hz.
>>
>
> You don't disclose the application, and the devil is always in the details.
>

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.

> I have a couple of concerns.
>
> A device that small and stiff is gonna have about zero sensitivity
> below 100Hz. To make it work, you'll have to "bend it" by attaching
> it to something bigger and flexible. Operational bending is bad enough.
> Shock/vibration/resonance is gonna be a reliability nightmare.
>

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.

> 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.
>

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".

[Joerg]

I don't know what "cont ip" is. But I have a working measurement setup
here, that's not the problem.

> If yoy break open the electret mic capsule, you can remove the electret
> dielectric and attach wires to connect your capacitor.
>
> Pretty much all types of non-electrolytic capacitor are microphonic when
> tapped with a screwdriver - you will soon see, some more than others.
>

I know, that's the reason for my posting. Which ones are the worst?
Because worst is best in my case.

> If the finished circuit picks up radio stations, your electrolytic has
> low ESR - put a 100 Ohm resistor in series with it.

EMI is not a problem, we'll run this in a modulated fashion.

[mike]

On 4/16/2013 7:25 AM, Joerg wrote:
> mike wrote:
>> On 4/14/2013 9:55 AM, Joerg wrote:
>>> Does anyone know which tiny MLCC have the worst microphonics? Brand,
>>> series?
>>>
>>> I want to do some tests to see how well these things can sense low pitch
>>> sound, mainly via changes in capacitance but also piezo effects. Main
>>> reason is that I need a fairly low impedance, high capacitance and a
>>> very small size. Ideally less than 0.010" width and height but for right
>>> now I can test with larger ones. 0.050" or so would be fine for testing
>>> the concept, maybe even a little larger. A capacitance higher than
>>> 0.01uF would be nice. The frequency response doesn't have to go beyond
>>> 100Hz.
>>>
>>
>> You don't disclose the application, and the devil is always in the details.
>>
>
> 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 have a couple of concerns.
>>
>> A device that small and stiff is gonna have about zero sensitivity
>> below 100Hz. To make it work, you'll have to "bend it" by attaching
>> it to something bigger and flexible. Operational bending is bad enough.
>> Shock/vibration/resonance is gonna be a reliability nightmare.
>>
>
> 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.
>
>
>> 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.
>>
>
> 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.

[Joerg]

mike wrote:
> On 4/16/2013 7:25 AM, Joerg wrote:
>> mike wrote:
>>> On 4/14/2013 9:55 AM, Joerg wrote:
>>>> Does anyone know which tiny MLCC have the worst microphonics? Brand,
>>>> series?
>>>>
>>>> I want to do some tests to see how well these things can sense low
>>>> pitch
>>>> sound, mainly via changes in capacitance but also piezo effects. Main
>>>> reason is that I need a fairly low impedance, high capacitance and a
>>>> very small size. Ideally less than 0.010" width and height but for
>>>> right
>>>> now I can test with larger ones. 0.050" or so would be fine for testing
>>>> the concept, maybe even a little larger. A capacitance higher than
>>>> 0.01uF would be nice. The frequency response doesn't have to go beyond
>>>> 100Hz.
>>>>
>>>
>>> You don't disclose the application, and the devil is always in the
>>> details.
>>>
>>
>> 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.

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.

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.

[miso]

> 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.

[mike]

On 4/16/2013 6:10 PM, Joerg wrote:

>> 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.

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

[Joerg]

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.

[Joerg]

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.

[mike]

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...

[Joerg]

mike wrote:
> On 4/17/2013 9:22 AM, Joerg wrote:
>> mike wrote:
>>> On 4/16/2013 6:10 PM, Joerg wrote:
>>>
>>>>> 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.
>>>
>>> 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.
>>
> 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.

[mrob...@att.net]

Joerg <inv...@invalid.invalid> wrote:
> 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

[Spehro Pefhany]

On Wed, 17 Apr 2013 15:40:06 -0700, the renowned Joerg
<inv...@invalid.invalid> wrote:

>
>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
--
"it's the network..." "The Journey is the reward"
sp...@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com

[Joerg]

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.

[Joerg]

Spehro Pefhany wrote:
> On Wed, 17 Apr 2013 15:40:06 -0700, the renowned Joerg
> <inv...@invalid.invalid> wrote:
>
>> 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.
>

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.

[rickman]

On 4/14/2013 1:17 PM, John S wrote:
> On 4/14/2013 12:12 PM, John S wrote:

>> On 4/14/2013 11:55 AM, Joerg wrote:
>>> Does anyone know which tiny MLCC have the worst microphonics? Brand,
>>> series?
>>>
>>> I want to do some tests to see how well these things can sense low pitch
>>> sound, mainly via changes in capacitance but also piezo effects. Main
>>> reason is that I need a fairly low impedance, high capacitance and a
>>> very small size. Ideally less than 0.010" width and height but for right
>>> now I can test with larger ones. 0.050" or so would be fine for testing
>>> the concept, maybe even a little larger. A capacitance higher than
>>> 0.01uF would be nice. The frequency response doesn't have to go beyond
>>> 100Hz.
>>>
>>

>> My suggestion, Joerg, is to get the greatest capacitance in the smallest
>> size to start with.
>
> 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?

--

Rick

[Robert Baer]

Yeah; off-datasheet uses are a PITA, depending what attribute is
needed, a very minor fab change can kill your product.
Been there.

[Joerg]

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.

[Frank Miles]

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.

[Joerg]

> 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.

[josephkk]

On Sun, 14 Apr 2013 12:12:46 -0500, John S <Sop...@invalid.org> wrote:

>On 4/14/2013 11:55 AM, Joerg wrote:
>> Does anyone know which tiny MLCC have the worst microphonics? Brand, series?
>>
>> I want to do some tests to see how well these things can sense low pitch
>> sound, mainly via changes in capacitance but also piezo effects. Main
>> reason is that I need a fairly low impedance, high capacitance and a
>> very small size. Ideally less than 0.010" width and height but for right
>> now I can test with larger ones. 0.050" or so would be fine for testing
>> the concept, maybe even a little larger. A capacitance higher than
>> 0.01uF would be nice. The frequency response doesn't have to go beyond
>> 100Hz.
>>
>
>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.

?-)