Laser Displacement Sensor

Follow up:

Does anyone know of a setup with such a sensor (1D), that is mounted on a XY table (looking down on the product), to measure the resonance over the entire PCB

Off course, the sensor needs to be mechanically isolated from the structure, so this can be challenging

Cheers

Klaus

Hi Klaus, Are you trying to sense the x-y motion or motion in the z direction. If you could mount a little mirror on the pcb you could use a split photodiode to sense motion in the x-y plane.. hmm I guess z-axis too. I'm thinking of the mirror at 45 degrees so the laser makes a 90 degree turn.

Maybe there is a more clever method?

George H.

Hi Klaus,

Using a 640 x 480 pixel camera and a disc target, you can get better than 10 um resolution in x,y. The center of the target is measured by calculating the centroid of the target image. Seismic motion is in this range though, so you may need to isolate things if you need a clean signal. Even the concrete floor will move. Taking the measurement at night with no automobile traffic and no people walking in the building can help. You may still need something like pneumatic isolation for the sensor platform.

Here's a picture of such a measurement on a static target, which of course shows the noise floor of the system.

A link to some math on measuring the disc fiducial is here:

Alon Efrat and Craig Gotsman, (1994), ?Subpixel Image Registration Using Circular Fiducials?, International Journal of Computational Geometry and Applications, February.

Seems to me that it would be quicker and cheaper to just shake a prototype and see what parts fly off.

In lots of situations, you can see the parts blur before they fatigue and fly off.

Kodak, in Rochester, NY, had a shake-table big enough to handle car-sized equipment. When I was at GenRad we'd take our stuff there and view it with a strobe while shaking at different frequencies... scary to watch the neck of a CRT flexing >:-}

Kodak had an interesting program for that table... it simulated a truck driving from your dock through various road conditions, over rail-road tracks, etc. We learned to use pour-in-place foam packing. ...Jim Thompson

I have used a Keyence meter for measuring the motion of a piezoelectric stage. It worked very well, although the stage was only moving at 120Hz.

I wasn't the guy who bought the Keyence meter, but my impression was that Keyence was very forthcoming about what their product did and did not do, whether it was good or bad. If you talk to one of their sales engineers and he says it'll do something -- it'll do that thing.

onents flying all over the place during vibration tetsting

uess better than 10um resolution.

nting a MEMS sensor)

a XY table (looking down on the product), to measure the resonance over th e entire PCB

ture, so this can be challenging

diode to sense motion in the x-y plane.. hmm I guess z-axis too. I'm think ing of the mirror at 45 degrees so the laser makes a 90 degree turn.

Right now we are looking at the deflection of the PCB in the z-direction du e to a component that came loose during vibration. We are investigating dif ferent techniques to either shift the resonance or dampen the vibration, bu t to qualify the tests I would like to be able to measure the absolute move ment of the PCB, along with measurements of the table level.

Cheers

Klaus

It sounds very interesting, but we are looking for something that shows the results right away without too much data manipulation.

Cheers

Klaus

That's what we have done. Now we need to investigate to increase the g level when the first part flyes off

Cheers

Klaus

I work for a rather large company, and we are using that equipment. We can vibrate up to 50G and close to 1000kg

We just need some more high tech stuff to measure the vibration

Cheers

Klaus

You've probably already thought of all the following:

I assume you can't just get a nice small piezo accelerometer and glue it to the board?

It may be a good idea to do some resonance-frequency studies before you put the thing on the vibe table. Come to think of it, when you do vibe stuff you'll want it mounted the way it'll be mounted for real -- a board flopping around on a test fixture doesn't necessarily match that same board flopping around in your real system.

todiode to sense motion in the x-y plane.. hmm I guess z-axis too. I'm thi nking of the mirror at 45 degrees so the laser makes a 90 degree turn.

due to a component that came loose during vibration. We are investigating d ifferent techniques to either shift the resonance or dampen the vibration, but to qualify the tests I would like to be able to measure the absolute mo vement of the PCB, along with measurements of the table level.

OK rethinking, just mount the mirror flat and bounce the light off it, 45 d egrees in and 45 degree's out. As the mirror moves up the light beam moves up. If the laser beam is ~1mm then 10um is a 1% change. But you do a lot better than that because you can take the difference of the two signals fr om the split photodiode. And with a laser beam you can make the PD plenty fast.

George H.

Maybe a strain gauge glued to the board. (maybe we're not allow to stick something to the board?)

George H.

It is possible, but the weight is going to affect the result, I would like to make non-invasive measurement

Yes, the real PCB is mounted on standoffs, with no standoff located the position of the failed component

We want to measure both on the real life standoff, to measure the level and frequency response with random excitation, then probably draw back to a stiff fixture testing with both random and sine excitation

Cheers

Klaus

Well, the resolution can be had by using a contact microphone; if you just couple it to the board with a stick, that means you'll be wandering over the surface looking for a 'strong' indication. If the mass of the microphone is an issue, you could also put (grounded) conductive paint dots on the board, and monitor capacitance to a fixed electrode, by oscillating the capacitor in a tank circuit and looking at the FM component.

And, you can find the pattern of displacement by scattering sand onto the board, and looking for where the stuff sits versus where it dances... see "Chladni plate" for more info on that.

That sounds VERY complex. Whatever would you do with the data? What kind of fixture is stiff if you're talking about excitation that drives components off a board? And, you DO realize that solder, or any other metal, is elastic at small stress, but plastic (and subject to fatigue) at high stress: you are NOT going to have an amplitude-independent frequency response.

t couple it

e looking for

ould

itance

looking

board,

ni plate" for

and frequency response with random excitation, then probably draw back to a stiff fixture testing with both random and sine excitation

d of fixture

board? And,

s, but plastic

It is complex and normally we just increase the amplitude in discrete steps and record what falls off and the cause. This time I want to dig a little deeper, so that we have some more knowledge to use at the next prototype se ries and to pass information on to the mechanical engineers. (instead of ju st saying "it falls off, duh :-)")

Reproducing vibration tests is hard, but if we don't try to understand it d eeper, it is not easier the next time around

Cheers

Klaus

For starters a vibration test with the entire product, recording the level on the failed component, random vibration

Next, vibration test with a very stiff fixture. So we test only the component without the attenuation/gain of the standoffs and PCB. This test would if possible show the mechanical resonance of the component

Finally, try to reduce the gain of the product + PCB at the resonance frequency of the component.

Cheers

Klaus

On Thu, 9 Oct 2014 05:43:10 -0700 (PDT), Klaus Kragelund Gave us:

Yeah? What test frame rate? All 50kHz?

The resolution would require a laser, and an optical sensor and a very fast read/sample circuitry. Like an expensive, real time, fast LIDAR type thing.

Instead...

Just use a high speed camera, like they use at the railgun facility or the impactor test facility.

Then you can construct a selected frame video of the entire vibratory frequency sweep range, and watch at which frequencies damage "resonates". No need for micron resolution to quantify the movement. You'll see it, and know at which frequencies it occurred at.

The power required to vibrate an object with an amplitude s is

P = M omega**2 s.

Vibrating an object through 10 um at 50 kHz requires a power of

P = M * 4 E10 1e-5 = 400 kW/kg.

I suspect you need something closer to 10 nm sensitivity.

Cheers

Phil Hobbs