Hi!

I´ve tried to analyze the plot attached for a while now, but it does´nt make any sense to me. I´ve never encountered a PSD plot before and I need a frequency (mean) to continue my calculation on vibrations, but the PSD just makes me confused. Could anyone extract some information from it to me?

Acceleration data: http://files.engineering.com/getfile.aspx?folder=7...
PSD for the acceleration data: http://files.engineering.com/getfile.aspx?folder=b...

Kindly //

I concur, that you cannot usefully derive a frequency(mean) from that plot.

I suggest you go back to whoeever asked you for a better definition. It is not an expression i have come across in 30 years of vibration analysis.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376: Eng-Tips.com Forum Policies http://eng-tips.com/market.cfm?

Thanks for your reply Greg. You just saved me a whole bunch of time :)

Are you asking for the resonate frequency? I've never heard of mean (average?) frequency.

Tobalcane
"If you avoid failure, you also avoid success."
“Luck is where preparation meets opportunity”

A resonate frequency would be perfect, if the psd provides the information needed :)

If your input PSD is as flat as NAVMAT-P-9492 with a broader band (50Hz to 1000Hz at 5dB down), the Fn would be the first highest peak from left to right. If the profile looks like a Harley Davidson engine, it is very difficult to determine the Fn. However, a good tester would do a .25g sin sweep first to tease out the Fn so that when you do the random vibe, you can compare the energy imparted into the part due to the random vibe at that frequency. If anything, ask if they can do a sin sweep.

Tobalcane
"If you avoid failure, you also avoid success."
“Luck is where preparation meets opportunity”

From the units used in the PSD it looks to me like a PSD was taken of the (time history - avg. accel of the time history). Your time history does seem to have an offset. Maybe someone thought it was best to get rid of the offset by subtacting the average acceleration from the time history before taking the PSD.

To carry out a PSD, you have to suppose that your signal is stationary and ergodic (it is the definition of a random signal)

Your time-history accel. is not enough flat ( many peaks).
In my opinion, your signal is not enough stationary to perform a PSD...

Is there more to the PSD plot? It appears to be falling off past 100Hz. That might indicate a corner frequency.

What exactly are you trying to calculate with your "mean" frequency? The only thing I can imagine using a single frequency is to determine whether a particular mechanical frequency overlays on top of an isolator frequency, which is undesirable.

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

Thank you for your help, and sorry for the late answer - everyone needs vacation :)

The plot is from a test vehicle, and I need the frequency to complete some calculations within piezoelectrics.

Again, what is the "mean frequency" supposed to represent?

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

IRstuff, Im a bit lost in my calculations of piezoelectricity. But hopefully Im a bit more educated within the area now :)
What Im looking for in terms of frequency is the bandwidth where the most of the frequency is. Since I need to order piezoelements that work in that specific frequency range.

Is that easier to outline from the PSD? Thanks in advance (again!)

I still don't understand. The standard bandwidth is essentially defined by the intersection of the DC response level vs. the sensor rolloff, which is a straight line with a 10dB/decade slope for power.

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

Ok, now Im completely lost :/

The only data I have is the acceleration graph and the PSD for it. They have been measured by someone else, and was given to me since Im doing studies that require an operating frequency. I need to pinpoint a frequency from which I could harvest the most energy from (piezoelectrics), and I have no idea how to read the PSD and what the strange unit (m/s2-rms^2)/Hz tells me.

Sorry, but my knowledge within electronics and vibrations are highly limited..

Perhaps, rather than getting mired in a subject you appear to know nothing about, you should simply describe what end result you are trying to achieve.

It seems to me, reading between the lines, you are actually needing to use the amplitude for a particular frequency to determine the amount of power your "piezoelectrics" produce? You then integrate with respect to frequency, and divide by the average bandwidth.

You do have a problem in that your PSDs are not rolling off, so there's no information about where the input energy actually starts to fall off.

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

A convential PSD does tell you where most of your power or energy is. The units on yours are a bit wierd. I think the PSD was taken from a time history for which the RMS of the time history was subracted (my guess based on the units). The DC value and the low frequencies (< 5 hertz) in your data are the most dominant but they may not be real. Your PSD is fairly flat with no real dominant frequencies. There's a couple of peaks around 16 hertz and 85 hertz. I'd think you'd want a piezo device with a flat response through 100 hertz minimum. If you want to go down to DC that will determine the low end of the response.

IRstuff: I do want to determine the power I could achieve from a PZT, bi- or unimorph plates mounted within a vechicle. With the hopes of harvesting the energy from ambient vibrations. Actually the amplitude would probably have been a issue after I managed to get my head around the frequency mystery. Average bandwidth might be the key-word I´ve been looking for? :)
I do actually have one more PSD from another measurement. I´ll post it here as soon as I get into the office tomorrow.


BrianE22: I tried to get hold of the guy who passed along the PSDs for me on order to see if they have other PSDs to get my hands on. I´ve checked other information graphs and I think the frequencies should be around 20Hz. Could the two peaks be about resonance frequencies?

Flat response through 100Hz minimum? And the circuitry I´ve planned on following uses a simple rectifying bridge to get DC, how does this determine the low end of the response?

ps. I really appreciate your time & thanks for the responses even though my replies tend to be a bit slow. Different time zones makes it hard for me to answer right away.. I´ll post the other PSD tomorrow

Well, you are missing a huge part of the problem, from what I can tell. The PSD is only marginally useful for what you're trying to do, since you do not have anything resembling the transfer function between the vibration source and the piezoelectric device.

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

Do you mean a physical transfer function, or an equation? I have a few articles available with equations for calculating the piezoelectric power with dependence of the vibrations from the source.
That´s the reason Im doing this, I can´t order any piezoelectric devices if I don´t have anything pointing out that I actually can get somewhere close the power output I need (with current being the limiting one).


Here is another graph with acceleration data and the PSD for it:

• http://files.engineering.com/getfile.aspx?folder=792425af-1d5d-4ce3-885d-83

Things like this are highly application, design, and location dependent. I don't see how it would be possible to predict anything without a solid, physics-based analytical model. Piezoelectricity is dependent on the applied stresses across the device. If is completely rigid body motion, then there would be no output.

Conversely, your "current" requirement may simply exclude this approach completely. Piezoelectric devices are not high-current sources.

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

Yes, and my task is to evaluate the implementation possibilities within an application with a small rechargeable battery, with the task to power some LEDs from time to time. Models for power output has been developed and there are a bunch to try out, which is my plan before I´ll give myself into putting together a prototype.

The piezoelectronic will be mounted as a cantilever, which bends back and forth due to the vibrations. This is where my need for average frequency bandwidth comes in :)'

ps. Im not trying to achieve high current, I just mentioned current as the limiting one which I´ll have to investigate closer than the other factors.

OK, then it seems to me that this "average" frequency approach is not correct. Assuming that there is a frequency response curve for current output for the cantilever/piezo structure, the response curve should be multiplied by the input spectrum, and the resultant should be integrated over frequency to get the total current.

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

The problem is that most piezoelements only harvests energy within a specified frequency, and the power output decreases heavily outside that spectrum. So I think that integrating the curve would give me false hopes of power output..

I´m currently halfway through the book "Energy Harvesting Technologies" by Priya & Inman, and I´ve read countless of articles. And the most of the equations rely on the electronics within the circuit and the surrounding kinetic energy (amplitude and frequency). Unfortunately the piezoelements are a bit too expensive for try-and-error solving, thats why Im bound to the calculations.

Btw, atm Im looking into this harvesting kit. And I´ve managed to sqeeze in two of them within my device (with CAD). Unfortunately it also has a higher frequency demand than I think I´ll have if the PSD would give me some actual frequency data.. So currently Im looking for a solution that offers a broad frequency band for energy harvesting.
I think that the average frequency bandwidth would lie around 20Hz, or am I totally lost now? :P
(My assumption is based on other data on the net about normal car driving accelerations & frequencies)

ps. It´s really hard to find anyone to rely on with discussions within piezoelectronics, so you are actually the one I´ve had the longest discussion with.

"piezoelements only harvests energy within a specified frequency, and the power output decreases heavily outside that spectrum."

That's precisely what I'm saying when I talk about the response curve. You simply multiply the response curve against the input PSD to get the weighted output PSD. The integral of that is the average power expected from the transducer.

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

If PSD is not appropriated for analysing a random vibration, then you can also use another indicator: the Extreme Response Spectrum. It is the same as Shock Response Spectrum. The only difference is that ERS is used for random vibration.
If your signal is supposed to be random but not stationnary, then ERS is a good indicator for analysing random vibation.

You are confused, by the way. The device's datasheet simply states that 52Hz is its "max rated frequency," although it's probably its most efficient at that frequency. Your PSD clearly shows response at, and around, the device's operating frequency.

Of course, your device only puts out 7mW at rated frequency and deflection, which is pretty miniscule, when you get right down to it.

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

I´d like to try this approach, but Im a bit confused with the details. The response curve you are talking about, is it available to get from some piezo institutes or do I need to order a piezoelectric kit and measure the curve myself?

For the PSD: do I get the equation of integration interest from doing a "reversed bode plot" on the PSD? (it´s many years ago I had a course with bode plots, but its the only thing I remember where I´ve constructed equations from a curve)?

"Of course, your device only puts out 7mW at rated frequency and deflection, which is pretty miniscule, when you get right down to it."
Is it possible to achieve these two requirements with three of the devices coupled together:
(1) overcome the required power to start charging a small battery (around 120mAh)?
(2) to charge it in a reasonable time (Reasonable time = matter of hours, not days)?
Or is it completely out of question?

Depends on what the consumption is and what type of battery it is. Some batteries allow for high current (rapid) charging, and others do not. Are you attempting to run the electronics at the same time? What is its consumption?

Even in an ideal case, you would need something like 10 of these piezoelectric devices in parallel to get enough current to charge a fully drained battery within 2 hours, with no additional load.

Seems to me that you've not presented a sufficiently enough detailed operational scenario and requirements to even begin to figure out what you need to do.

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

Seems to me that you've not presented a sufficiently enough detailed operational scenario and requirements to even begin to figure out what you need to do.

At my first try Im going with a Ni-Mh battery since it doesnt need extra circuitry to allow charging. The electronics are not going to be on at the same time as the battery is being charged (with and exception with the standby power of the circuitry ~1mA). The consumption of the electronics will hopefully not be over 50mA, but that is a bit uncertain for now with a choice of LEDs (the number of these has to be around 10 totally).

The 2 hours is a dream scenario, hopefully the idea is at least feasible and shows some charging activity which would allow more development and optimization.

The requirement of the lights today is that they can be activated 20times/day (8seconds/time) + one long term activation (60seconds), those are the main requirements for the energy source to handle.

Btw, I managed to get hold of the guy who produced the PSD today. He said that the PSD was constructed for one purpose only, and he has no idea why the data was sent to me since it cant be translated to something useful by anyone else. He is going to send me some PSDs and frequency data plots in the coming days that I actually can work with. But probably Im gonna need some help with those also :)

20*8+60 = 220 s
220s*50mA = 3 mAh plus 1mA standby

So there you have it. Your device puts out 1.4 mA at rated performance, and since you cannot have 100% charging efficiency, you'd need more than 8 hrs to recharge battery.

TTFN
FAQ731-376: Eng-Tips.com Forum Policies

I cant find it myself so I´ll have to ask, where did you get the 1.4mA from?