DIYMobileAudio.com Car Stereo Forum banner
Status
Not open for further replies.
141 - 160 of 1,726 Posts
Discussion starter · #141 ·
Ok, sorry to leave this for so long.

I need to clarify my statements a little bit. When I said I think that I think similar products are going to test similarly, I mean that products that have a 4mil thick aluminum layer, and 1.5-2mm thick butyl layer are going to test almost identical. I do however think that a product with a 8-10mil thick layer of aluminum will test differently than the products with 4mil thick aluminum. Then theres the asphalt products and the stuff from STP, which is very different than all of the other products. Of course, this is just my hunch, and could be proven wrong. But I wanted to put it out there.

The resonant frequency of the metal on the first test made it very difficult to measure at a wide range of frequencies. This is bare metal, no deadening. I was actually unable to get anything out of it except at 100hz. That tells me the Q of the resonant frequency is very high, something I will have to investigate more. After I added deadener, I was able to measure at both 80hz and 100hz, telling me that the Q flattened out a little, but it also showed a very direct decrease in vibration. I'll be able to quantify it later once I get the equation from the accelerometer manufacturer, but the peak to peak voltage almost halved from bare metal, to deadened.

Like rton20s said, its not going to be possible to compare these results to in vehicle results with an equation. That's why this is just the beginning. There are plans to go much much farther in depth than the simple testing done here between brands, including real world car tests. That's why its so important to me to be able to automate these tests as much as possible.


I will get some video up later today of everything running, I have to get some things done. But I will get it up tonight.
 
Ok, sorry to leave this for so long.

I need to clarify my statements a little bit. When I said I think that I think similar products are going to test similarly, I mean that products that have a 4mil thick aluminum layer, and 1.5-2mm thick butyl layer are going to test almost identical. I do however think that a product with a 8-10mil thick layer of aluminum will test differently than the products with 4mil thick aluminum. Then theres the asphalt products and the stuff from STP, which is very different than all of the other products. Of course, this is just my hunch, and could be proven wrong. But I wanted to put it out there.

The resonant frequency of the metal on the first test made it very difficult to measure at a wide range of frequencies. This is bare metal, no deadening. I was actually unable to get anything out of it except at 100hz. That tells me the Q of the resonant frequency is very high, something I will have to investigate more. After I added deadener, I was able to measure at both 80hz and 100hz, telling me that the Q flattened out a little, but it also showed a very direct decrease in vibration. I'll be able to quantify it later once I get the equation from the accelerometer manufacturer, but the peak to peak voltage almost halved from bare metal, to deadened.

Like rton20s said, its not going to be possible to compare these results to in vehicle results with an equation. That's why this is just the beginning. There are plans to go much much farther in depth than the simple testing done here between brands, including real world car tests. That's why its so important to me to be able to automate these tests as much as possible.


I will get some video up later today of everything running, I have to get some things done. But I will get it up tonight.
Hey Chris, thanks again for doing this! Once you have things going smoothly, I'd like to send you a 1 sqft cut of our next product. I'm interested in getting feedback on it. Can you PM me your address info once more?
 
Discussion starter · #143 ·
PM sent Tracie.

Ok, well I tried a few more things today. I burned a cd with test tones spaced at 5hz intervals from 20-120hz. Same results, it goes from not picking up anything on the scope, to clipping the output from the accelerometer, before falling off again. Same frequency for the peak, right around 100hz. I'm not sure how I'm going to deal with that just yet. I have some ideas. I know the current power supply for the accelerometer is pretty noisy, I'm going to try to Dell power supply 3.3V out since it seems much cleaner than the simple board I'm currently using for the accelerometer. I may all try running it off some batteries, if that's still too noisy. I also used the Omnimic today with the microphone about 5mm away from the metal panel. I hate to say it, but this may be the way it ends up going. If so, I need to build a chamber around the back side of the speakers to isolate the sound from them, so its only picking up noise from the panel.

These are my pics, not rton20s', from a canon cyber shot taken today.








Here is the averaged response from noise from the Omnimic. I need to redo it with less smoothing, but you can see the HUGE peak, 20db on the right side and 30db on the left. I should have tested this yesturday before and after I applied the "not to be named" deadener.



Ugh, video is still "processing". I'll get it up as soon as its done.
 
Discussion starter · #144 · (Edited)
Hmmm. Epic fail. Even after watching numerous how to's on embedding videos, I still cant get this to work.

Anyone want to hold my hand through this?


Link is here Test Run of Equipment - YouTube



Also, anyone know how to eliminate the stupid pops when burning test tones to CD from audacity?
 
Export the wav files to disk then burn them with a -1 sec delay between each track. Try CDburnerXP (free).

Tried decoupling the mic from the table? It might pick up vibrations transferred through the table...

Sent from my Samsung Galaxy 3 via Tapatalk.
 
Discussion starter · #149 ·
Hanatsu, thanks. I usually actually add the delay between tracks when using the home computer, but didn't have a program that could do it on the laptop.

I will try that decoupling the mic from the table. I wanted to try that, but all my materials are in storage. Maybe this weekend I can get out there and re-try it.


Dean, thanks for the PM, and for posting the video.

Correct me if I'm wrong, but the IASCA and Cleansweep disks are both 1/3rd octave right? I have the Focal tools disk, which has 1/3rd octave tones, and it wasn't enough resolution. Before adding a piece of deadening, the resonant frequency seemed to be somewhere between 100hz and 125hz.

My dad worked on the program all night and said that it will be able to generate the tones itself, so I'll deal with the disk I have for now.
 
It seems like using a sweep would be better than individual tones because time domain is important too. The maximum steady-state amplitude of a resonance is interesting, but without taking time into consideration, you cannot see how quickly the resonance builds or decays (how quickly the motion is transformed into heat, and thus how effective the damping is).

Room EQ Wizard for example generates its own sweep and will calculate just about anything you could ever want. Each sweep is quick and the program and makes comparison of multiple trial runs (or multiple test products) very easy. Among a ton of other stuff, the program will find the frequency, amplitude, and duration of all resonances for you. That's the interesting stuff, isn't it?

Also, your non-treated system having a single dominant resonance is not unexpected.

-J

(I love this project by the way!)
 
Discussion starter · #152 ·
It seems like using a sweep would be better than individual tones because time domain is important too. The maximum steady-state amplitude of a resonance is interesting, but without taking time into consideration, you cannot see how quickly the resonance builds or decays (how quickly the motion is transformed into heat, and thus how effective the damping is).

Room EQ Wizard for example generates its own sweep and will calculate just about anything you could ever want. Each sweep is quick and the program and makes comparison of multiple trial runs (or multiple test products) very easy. Among a ton of other stuff, the program will find the frequency, amplitude, and duration of all resonances for you. That's the interesting stuff, isn't it?

Also, your non-treated system having a single dominant resonance is not unexpected.

-J

(I love this project by the way!)
Hey, I'll pm you later today, I ran through a bunch of things today with omnimic, all with sweeps, but until I can satisfy a few problems I have with using the mic (mostly related to isolating it from the sound coming from the actual speakers) its just playing around. I'll post some snapshots of what I got though. The tones are ONLY for the accelerometer, although there will be some time domain stuff going on there too, with phase comparisons between the speakers, and the accelerometer.


I totally expected the single resonance, and for it to be big when un-treated. But, both the frequency response graph I posted and the video of response from the accelerometer on the scope was with with the metal treated. It has 30% coverage right now with a CLD sheet, in the center of the metal. The CLD currently on there has the second thickest constraining layer of all the samples I have.
 
I am not familiar with the Omnimic. What signal does it use? I see tone burst mentioned on one of the graphs. The scale of the frequency response makes it a little hard to see, but that resonant peak near 100hz is much more sharp than the result of any of my tests. My experiment had the sheet metal suspended on string though, not firmly clamped to an enclosure from all edges.

You are having problems isolating the microphone from the speakers? It looks like the rear half of the speakers are open to air and might contribute noise to your measurement. However it looks like your mic is vesy close to the sheet metal so the high signal to noise ratio should overwhelm any interference from the rear wave of the speaker since it is relatively very far away.

Maybe the signal from the speaker is being passed through the sheet metal and picked up by the microphone in addition to the re-radiation from the sheet metal. That could skew the results dramatically.

So, why not forgo the mic and use the accelerometer exclusively? If you can hook it up to the scope, it should have enough voltage to drive a preamp and be captured by your computer and analyzed by any software you choose. I assume this would be the most direct way to measure the vibrations of the sheet metal without any acoustic interference.
 
Discussion starter · #155 · (Edited)
For all measurements above, it uses sine sweeps, short sweeps for the freq response, impulse response, and waterfall, longer sweeps for the energy time curve and bass reverberation decay, and long bass sweeps for the bass decay.

The accelerometer will put out 3 volts when I move to a battery to lower the noise level, I hadn't yet looked into whether or not it will work with my sound card. The other probably seems to be sensitivity. In order for it to read anything but the peak, it needs to have some eq, to allow it to read the low intensity areas while preventing clipping at the resonance peak. With the program for the scope, that would be built in, but I'm not sure if that can be done with Omnimic or REW. Of course, if the resonant frequency changes dramatically enough, that built in EQ could also cause a problem with accurate measurements.


As for the speakers interfering with the mic, even with as close as the mic is, they still mess with the results. I held some pillows over the speakers today while running sweeps, and it definitely affected measurements, and I wasn't doing a very good job of muffling the speakers on the outside. I might have my wife help grab some shots tomorrow without and with pillows, since I need two sets of hands. I have an idea on how to isolate them better, but it requires more wood, which has to wait until I get paid next friday. Then I can see if the sound coming through the sheet metal is enough to skew the results.




Edit) Just talked to my dad, hes working on the program right now and will be testing certain parts of it tomorrow. He is basically working on a program that would run the scope, output the signal to measure, compare phase between the signal (at the speakers) and the accelerometer output, and the intensity between the two, all while logging everything into access. It would then be put into excel and graphed however I wanted. He encouraged trying to work on getting the mic to work better as well, just as a sort of checks and balances system, basically saying that if one varies wildly from the other, something is wrong with the test.
 
Hello, guys.

Looks like its going to be a very interesting test, but I would like to add my $0.02 to possibly make it even better.
First of all, it is good to see that at least one of the Russian suppliers has made it to US car audio market (STP that is).
In Russia, most of vibration damping material marketed for car audio take its origin from construction industry. And they share measurement techniques.

It is well known that damping properties are frequency ant temperature dependent. Less known that eficiency also depends on metal thickness and vibration mode shape.

TOOSTUBBORN2FAIL, your test setup looks overly complicated and prone to systematic errors:
1. steel panel has square shape. This shape is not the best if we want to measure damping in wide frequency range because vibration mode shapes wil be different and will effect results in almost unpredictable way.
2. Panel clamping has low reproducibility and possible high level of damping, which in turn will mask damping of CLD.
3. I don't see clear raw data processing strategy and all these transient measurements are completely uneccesary.

I would strongly recommend to reconfigure test according to Oberst beam method
1. You would need to cut you panel into stripes (length depends on the stripe thickness, but 1:100 ratio would be a good starting point). Stripe shape is good because it has primerely lateral mode shapes of different wave length. Mode shapes uniformity will help with results consistency in wide frequency range.
2. Clamp the stripe vertically by the top end in really rigid vice.
3. Put the accelerometer at the bottom end of the stripe
4. Excite stripe vibration is some way (sweep tone, white noise, even hit with the metal bar at different stripe points - it all works!)
5. Record spectrum response (holding max. achieved levels).
6. Display spectrum in linear scale for frequency
7. Identify resonances (you only can measure damping at resonances)
8. Identify resonance frequencies - Call it F0 (it is good if CLD has more added stifness rather than added mass, i.e. treated stripe frequency goes up)
9. Identify frequency range where response is -3 dB from max resonance response - call it delta_f
10. Calculate Q factor of the system.

Using this method you wil be able to measure damping in wide freq. range (which is really imposrtant for CLD effective envelope evaluation)

P.S. May be I'm too obviouse here with this explanation, but I really wish you all the success with this test and looking forward to see some reasonable (and usable) test results.

Regards,
Andrey.
 
Hello, guys.

Looks like its going to be a very interesting test, but I would like to add my $0.02 to possibly make it even better.
First of all, it is good to see that at least one of the Russian suppliers has made it to US car audio market (STP that is).
In Russia, most of vibration damping material marketed for car audio take its origin from construction industry. And they share measurement techniques.

It is well known that damping properties are frequency ant temperature dependent. Less known that eficiency also depends on metal thickness and vibration mode shape.

TOOSTUBBORN2FAIL, your test setup looks overly complicated and prone to systematic errors:
1. steel panel has square shape. This shape is not the best if we want to measure damping in wide frequency range because vibration mode shapes wil be different and will effect results in almost unpredictable way.
2. Panel clamping has low reproducibility and possible high level of damping, which in turn will mask damping of CLD.
3. I don't see clear raw data processing strategy and all these transient measurements are completely uneccesary.

I would strongly recommend to reconfigure test according to Oberst beam method
1. You would need to cut you panel into stripes (length depends on the stripe thickness, but 1:100 ratio would be a good starting point). Stripe shape is good because it has primerely lateral mode shapes of different wave length. Mode shapes uniformity will help with results consistency in wide frequency range.
2. Clamp the stripe vertically by the top end in really rigid vice.
3. Put the accelerometer at the bottom end of the stripe
4. Excite stripe vibration is some way (sweep tone, white noise, even hit with the metal bar at different stripe points - it all works!)
5. Record spectrum response (holding max. achieved levels).
6. Display spectrum in linear scale for frequency
7. Identify resonances (you only can measure damping at resonances)
8. Identify resonance frequencies - Call it F0 (it is good if CLD has more added stifness rather than added mass, i.e. treated stripe frequency goes up)
9. Identify frequency range where response is -3 dB from max resonance response - call it delta_f
10. Calculate Q factor of the system.

Using this method you wil be able to measure damping in wide freq. range (which is really imposrtant for CLD effective envelope evaluation)

P.S. May be I'm too obviouse here with this explanation, but I really wish you all the success with this test and looking forward to see some reasonable (and usable) test results.

Regards,
Andrey.
Where the hell were you six pages ago? :p


Bret
PPI-ART COLLECTOR
 
Discussion starter · #160 ·
Hello, guys.

Looks like its going to be a very interesting test, but I would like to add my $0.02 to possibly make it even better.
First of all, it is good to see that at least one of the Russian suppliers has made it to US car audio market (STP that is).
In Russia, most of vibration damping material marketed for car audio take its origin from construction industry. And they share measurement techniques.

It is well known that damping properties are frequency ant temperature dependent. Less known that eficiency also depends on metal thickness and vibration mode shape.

TOOSTUBBORN2FAIL, your test setup looks overly complicated and prone to systematic errors:
1. steel panel has square shape. This shape is not the best if we want to measure damping in wide frequency range because vibration mode shapes wil be different and will effect results in almost unpredictable way.
2. Panel clamping has low reproducibility and possible high level of damping, which in turn will mask damping of CLD.

I did notice this, and have been working on a work around for this but haven't been able to test it yet.

3. I don't see clear raw data processing strategy and all these transient measurements are completely uneccesary.

I would strongly recommend to reconfigure test according to Oberst beam method
1. You would need to cut you panel into stripes (length depends on the stripe thickness, but 1:100 ratio would be a good starting point). Stripe shape is good because it has primerely lateral mode shapes of different wave length. Mode shapes uniformity will help with results consistency in wide frequency range.
2. Clamp the stripe vertically by the top end in really rigid vice.
3. Put the accelerometer at the bottom end of the stripe
4. Excite stripe vibration is some way (sweep tone, white noise, even hit with the metal bar at different stripe points - it all works!)

The paper linked suggests using non contacting electromagnetic transducers for vibration generation, stating that it significantly reduces the accuracy of the results if the test sample is contacted by the method of vibration generation. It also suggests the same for the measurement, suggesting that an accelerometer can be used, but that it would degrade the results.

5. Record spectrum response (holding max. achieved levels).
6. Display spectrum in linear scale for frequency
7. Identify resonances (you only can measure damping at resonances)
8. Identify resonance frequencies - Call it F0 (it is good if CLD has more added stifness rather than added mass, i.e. treated stripe frequency goes up)
9. Identify frequency range where response is -3 dB from max resonance response - call it delta_f
10. Calculate Q factor of the system.

I've seen a lot of studies stating that the half power bandwidth method of calculating damping ratio can be significantly inaccurate, often significantly overstating the damping ratio. There is a correction for it, but I have not been able to find somewhere where I don't have to purchase it (and unfortunately, I don't have any additional budget right now).

Using this method you wil be able to measure damping in wide freq. range (which is really imposrtant for CLD effective envelope evaluation)

P.S. May be I'm too obviouse here with this explanation, but I really wish you all the success with this test and looking forward to see some reasonable (and usable) test results.

Regards,
Andrey.
In the beginning, I really wanted to pursue this method, but the more I looked into it, the more it became priced out of my range for reliable results. Granted, I admit, there are still some issues with my setup, but as a comparative comparison, I will make sure all possible fixes are made before final testing and results are obtained.

Damping ratio can also be calculated by comparing the phase shift between the exciter and the substrate, which is what I'm working on now.
 
141 - 160 of 1,726 Posts
Status
Not open for further replies.
You have insufficient privileges to reply here.
Top