EEVblog 1443 - They Don't Teach This in School! (Coherence)

A follow-up video to the Electrodynamic Shaker, showing you the critical importance of Coherence measurement. Something you'll only get taught in the school of hard knocks!
Using the Ling Electrodynamic shaker, a measurement accelerometer, and a Dynamic Signal Analyser to set up a vibration test system.

00:00 - Electrodynamic Shaker
01:20 - The most inportant thing in vibration measurement
03:56 - What is a Dymanic Signal Analyser?
06:28 - It's all about the system response
08:12 - Your expensive calibrated setup is worthless!
08:44 - What is Coherence?
12:17 - If you've got a really crappy shaker
13:10 - Let's set up a vibration test jig
13:54 - How to power and accelerometer
16:44 - What happens if you leave it flapping around in the breeze?
20:14 - Traps for young players
22:39 - Let's sweep this sucker
26:50 - Tighten your nuts!
28:22 - Show me your coherence plot or GTFO!

Buy anything through that link and Dave gets a commission at no cost to you.

Donate With Bitcoin & Other Crypto Currencies!

#ElectronicsCreators #Vibration #Measurement

[flymypg]

We did this primarily to ensure accuracy and repeatability as part of our DoE (Design of Experiments) process. The worst part overall was to do reference characterizations on the many, many shake jigs we used. Each had a reference fiducial for attaching the accelerometer. We'd always validate the signal source, the driver amp, the shaker axes, the accelerometers, the signal processing system, and, of course, the shake jigs. Only after that can you mount the DUT (Device Under Test). And then, for reference runs, we'd have to repeat the entire process in reverse as we took down the test setup.

[motion-scope]

Great to see structural dynamics in electrical engineering!
Accelerometers can also be attached using beeswax, as it is easy to detach compared to glue.
The BIG problem with accelerometers (along with poor low frequency responses and triboelectrics in the cables) is that they influence the dynamics of the PCB with their mass, so you never measure true PCB responses, even though you get good coherence! Mass changes the mode shapes of the PCB and this changes everything.
This is why we use Vision based Modal Analysis, where cameras measure submicron vibrations without adding mass to structures. Also you get to actually see what is happening because you measure a spatial response as well (not just a single point), so you see individual components moving and you can pinpoint problems and even shift mode shapes.

[jacobfaseler]

Working as a defense contractor, this is really insightful! You always see the vibrational energy spectra at the bottom of source control documents, and the mech-e’s audibly dread vibe-qualification - but to actually see how it’s done brings a lot of light to both the how and why.

[TheStuffMade]

I used to do electronics and electrical designs to control hydraulics using by anything from the US coast guard to the Swedish Navy. While we had all kinds of certifications like Bureau Veritas etc. the various nations military always wanted to do their own testing including vibration testing and honestly I was surprised to see our electronics and electrical cabinets passing those tests, they were very violent. The most interesting issue we had was when we did an installation on some new mine sweepers with hulls made of fiber glass to prevent attracting and setting off mines and all the hydraulic gear we delivered was custom made from stainless steel to be non magnetic. However being navy ships they still had very powerful radio equipment on board, so having a fiberglass hull and powerful radio, it would make some of our equipment freak out whenever they used the radio. In the end I had to fly there to solve the issues by making sure all cables were shielded and terminated properly plus adding a handful of nano farad sized capacitors at strategic locations. I used a roll of aluminum foil to narrow down the most sensitive locations. Overall a fun experience and they signed off on the installation after a few days. This is of course part of a much longer story including bar fights, a black eye and being denied access to a foreign naval base.
Cheers,
Jake

[teawizardry]

We actually DID go over this in school! In an elective barely anyone took.... I almost forgot about this, and definitely didn't know about this application.

[marco]

This is the best video in a long time! Great job Dave and thank you for all the effort that you put into this channel during the years. I started watching
these videos as a teenager, now I am a Physicist.
I don't think I would be where I am now without your work. Keep it up!

[mcconkeyb]

Yup, when I was doing vibration testing we had 1 of these HP devices, but we never used it, as it was too simple. We used either a sine or a random closed loop control system. 2 very big boxes of electronics that were driven by a PDP-11 computer (yes this was back in the 1980's and the PDP-11 was a powerful computer back then). We could control the vibration at a selected point on the shaker or fixture. We could also set up multiple points to control the vibration, by 1 of several fancy mathematical functions (usually just an average). We normally just controlled the input to the fixture that was designed to hold the product/device under test (DUT). We would then characterize the fixture to determine if it was suitable for the testing that was planned. As long as the fixture didn't have any resonances at any of the test frequencies, then you can control the vibration input from anywhere on the shaker or fixture and the control system would ensure that the coherence is always 1 across the frequency range of the test.
I also did research into vibration testing. The 2 most interesting things were: a) Impulse driven device characterization, b) mechanical heterodyning of structures. Unfortunately I left to pursue other electronics design goals before really being able to write papers on either of these fascinating topics.

[peterldelong]

Very cool. I repaired those very large shakers from MB Dynamics and Ling back in the 80’s working for a certain US defense contractor. The MB Dynamics C210 used these pieces of spring steel in a ring around the armature to suspend it and they would crack and break which in turn messed up the coherence response. It took a couple days of tear down and re-assembly to replace those stupid spring steel parts, what a PITA. Thanks for the cool video, it’s something most probably never come across as EE’s.

[theagnihotris]

We did do this and I still use it on a daily basis. I work in vibrations/acoustics and our school had amazing curriculum. Now all modern softwares automatically calculate coherence and save it with your measurement data. Amazing stuff thanks for sharing.

[neil]

Dare I congratulate you on a very coherent explanation?

[milantrcka]

Big shakers typically use a feedback accelerometer to linearize the driver itself and to control the acceleration in a closed loop servo. Some 75 -80 kg device was designed to take 30 g crash safety. During a test the feedback accelerometer fell off. This resulted in a 100 g (not grams, earth gravity acceleration). A massive motor tore off, main chassis bent, a pitiful sight. We used to stand in a respectful distance during tests.

[billharris]

Thanks Dave for the fairly detailed overview on vibration table setup. You are right, being taught this type of information in school would be extremely rare.

Being from the US, I enjoyed your Aussie Techno-jargon.

Having retired from engineering in 2016, I must say, the job of vibe testing setup has gotten easier over the years, as well as the number of pieces of equipment used, and size of the equipment. Probably no one is using a water cooled valve shaker power amp these days.

That little shaker you have is really cute, I have never seen one nearly that small. My warped mind has already come up with an alternate use, such as (as a joke), covertly attach it to the wall of a non-technical department head, and randomly play assorted barnyard noises throughout the day, just barely loud enough to be heard.

[nameredacted]

From best to worst, methods of attaching accelerometer:
Bolted
Epoxy / Cement
CA adhesive
Magnet
Double-sided tape
Wax
Touch

Cost to calibrate accelerometer seems a bit much... Perhaps nobody wants to deal with a hobbyist? My accelerometers cost $500 (single-axis)-$2000 (triaxial/fancy), and calibration costs us $200 I believe. We calibrate yearly. But we have a calibration lab right in the building (yet they are a different company due to conflict of interest concerns).

[johnyoungquist]

In the early days before I could find a Ling actuator I made one from a large speaker magnet. I put a linear bearing down the center of the magnet to support the test table. I later used a voice coil driver from a large disk drive. It was about 6 inches in diameter. I would use it horizontally rather than vertically, Over the years I found a few actual Ling drivers. We use one here every day to test each every vibe sensor that we make. The vibe sensor uses a 3 axis accelerometer. We mount the device at 45 degrees to each axis so a single setup excites all channels with similar magnitude. So we check the gain of all channels with one setup too. the engineering phase of the design few if any of these measurement details matter much. Initially the goal is to excite the assembly over the specified range or larger to find anything that resonates and moves. It could be a tall leaded component or the entire PCB itself. A component could flex enough at resonance to fatigue and break off. It the entire pcb flexes traces and solder joints are at risk of failure. If you find any suspicious resonances you might just excite the system at that point for a long time to see if you can induce a failure. If the product is exposed to vibration at some characteristic frequency you might drive it there for a long time too. Modern SMDs are so small and stiff that their resonant frequencies are too high to matter but they suffer when the pcb flexes. So board stiffeners and rigid mounts are important. Once all this real work is done you can perform the test in the spec even though by now it means nothing. We often far beyond spec limits to see how much margin we have.

[InssiAjaton]

I found this quite interesting in several aspects. The first is that I once participated in a 3-1/2 day customer training occasion offered by Bruel & Kjaer. One of the things they demonstrated was that you had to lock the accelerometer cable to the same plate where the accelerometer was attached. Probably using bees' wax on both the accelerometer and the cable. Another thing was to use the smallest available accelerometer available. The same rule then was also applicable to microphone selection on sound level meter, especially if you wanted it to work beyond some 20 kHz. The other thing is the coherence measuring instrument that I had never seen in use. I do see plenty of similarities to my HP 3563A Control Systems Analyzer, though. It also has plenty of math built in. The most impressive feature it has is a synthesizing of the the PID circuit out of the Bode (& Nyquist / Nichols) results. I once did that and was duly astonished. Otherwise, I mostly have used just the Bode diagram for my control systems tuning.

[jessicav]

When a grid's mis-aligned with ano-ther behind, that's a

[tolikechicken]

I had five (four of them core) classes in my ME program where coherence played a major role and was heavily emphasized in our dynamics testing. But to be fair, it was THE go to place for structural dynamics and vibrations research and the faculty throughout the colleges history pretty much wrote the book on how to do this stuff so I don't think it's an average case!

[ArtyomGalstyan]

Wow this is useful. I am currently going through the DO-160 training course, where one of the tests is for vibration. This video really helps to understand a lot of things!

[SeanBZA]

You can do the same with regular speakers, which will show that different manufacturers of speakers do actually make them sound different, because they are all more or less non linear over the response. They might show a frequency response curve in a standardised fixture, but pretty much you can be sure that that is only approximate, and your unit will be different.

The more expensive ones come with either compensation in the amplifier, generated from a test speaker, that had it's response and coherence measured, or actually have positional feedback, so that at least the measuring point is going to be correct. That is why cones need to be stiff, and why there are titanium cones, and other exotic materials. But most use paper, because it is easy to form, and very cheap.

But to use this your source must be good, and pointless if the recording you play has been mastered in a studio with less accurate speakers, as then you only can get the response the studio had. Even more so if the sound has been compressed with lossy compression, the compression has to be nearly equal to the uncompressed to get a non audible difference. For me recording that turned out to be 320k 48kHz, as the recording did not sound different to the original.

[sonicgr]

I love old school HP instruments. I still have a photocopy of a manual for an HP digitizing oscilloscope I was using in my student years...