ccww

These windows were available as an option in the US market and I ordered them from my local MB dealer. The two that are coming from Germany just weren't in stock in the US.

MBNUT1

Not sure how 9 dB increase translates into 8 times louder. In terms of noise I think that it translates to less than twice as loud.

Mathnerd

8dB is not 8x. With amplifier power, the dB scale is logarithmic, every 3 dB change results is a factor of 2 change, doubling or halving of the original value depending on if it is an increase or a decrease. So if you increase your radio amps power output from 50 to 100 watts you have made a 3dB change.
However with sound it takes about 10 dB change for a perceived doubling (or halving) of the perceived loudness.

S-Prihadi

Our human ears are not linear in its frequency response towards "extra loudness" within our hearing limit 20hz to 20khz. We are more sensitive between 2 - 5 hkz. It is called Equal Loudness Contour.
Now I am 53 years old and I can't hear pass 15Khz, this one is age related.

https://en.wikipedia.org/wiki/Equal-...d%20of%20pain.

However, I am very sensitive to frequency under 500hz frequency and sometime I can hear ( feel is a better word ) that much better than younger people.

So for us to feel/perceived certain noise increase as so very loud of an increase..... that make sense if the frequency of that noise is at where we hear best.

ccww

I should have been more precise. As Mathnerd points out, each 3dB increase is a doubling in acoustic power. So the 9dB change in noise that I observed is 2^3 = 8 times increase in acoustical power. While humans don't perceive loudness linearly in conjunction with acoustic power, I think the observation is notable in that simply changing from one road surface to another at 65mph can create an 8x difference in the acoustic power of the noise inside the car.

And a very interesting point by @S-Prihadi . Humans are better at perceiving certain frequency ranges than others. In particular we are very sensitive to frequencies around those of the human voice, or the midrange of our hearing. This has obvious evolutionary benefits and is rooted in the physics and physiology of the inner ear. The brain effectively normalizes these differences in sensitivity with respect to our perception of different frequencies. However, it is much easier for very low and high frequencies to fall below our perceptive thresholds. It is easy to experience this. Listening to music at a low level we can probably understand the lyrics just fine as we perceive midrange with great sensitivity, but the bass sounds weak and thin as it is too quiet for our perceptive threshold. Turn up the volume and suddenly the bass is rich and impressive. The proportion of bass to midrange has not changed, but our perception of the proportion has changed very significantly due to the bass now being above the perceptive threshold.

Indeed, I think that this effect may exacerbate the perception of differences in road noise on different pavement surfaces. Perhaps on a smooth road many of the lower frequency components fall below the perceptive threshold. Drive onto a cement or highly aggregated freeway and not only do the low frequency noises increase but they also start to exceed the perceptive threshold.

I've thought carefully about how I want to attempt to quantify the results of my experiments, both with the laminated window retrofit and the additional sound deadening, blocking, and absorption starting in the rear area. I think that overall dB measurements are far too crude. Perhaps the laminated windows will be great at blocking wind noise, but have no impact on road noise. That wouldn't be well reflected in an overall dB measurement and might even be lost in the margins of error. But if we can look only at the high frequencies associated with wind noise, then a difference due to the laminated windows could be well quantitated in a way that is consistent with the experience change.

There is a software program for iOS called AudioTools which has an optional Decible-meter octave band logging function. You can take a measurement over a given period of time and then see both the overall average dB with any weighting and you can see the average dB of any octave or 1/3 octave frequency band. You can change the weighting after the measurement and also review any point in time in the measurement. This will allow a very holistic and specific analysis of any differences in road noise. If the laminated windows have a significant impact in high frequencies for example, one could perhaps see a 6dB different at 2000-4000Hz but no difference in most of the bands. I plan to use this software to determine my baseline measurements and then measure again with the new windows and then with my other acoustical treatments. I will select several different stretches of road such as smooth road, concrete, highly aggregated etc and take several measurements over each stretch. This will result in greater confidence and also allow me to estimate the precision of the analysis method by comparing the different runs over the same stretch. Notably when I did this experiment with three sets of three runs and just my iPhone held in my hand and a basic dB-meter app, I got a precision of about ±0.2 dB!

I have not yet decided whether to purchase a calibrated measurement microphone for this project. They are available for iOS connectivity for about $230. A calibrated microphone would make a very large difference in accuracy, or if the dB measurements are close to their real values. However, even the microphone in an iPhone has remarkably good precision, or being able to repeat the same measurement and get the same result. Precision is really all that is needed here as I'm only trying to compare differences between measurements with the same setup, not register numbers valid for external comparison. However, I think that a calibrated microphone's more accurate octave-band charts would be interesting, if not necessarily much more useful.

S-Prihadi

Hi CCWW,
I think a stand alone long wired mic will do much good. MB WIS has shown equiptment similar to chassis microphone ( chassis ear ) when troubleshooting suspension noise.
I believe, if you can aim the mic to where you think the noise source is, it will speed up the research.

On the subject of : Calibrated microphone.
This is technically a calibrated microphone. https://www.minidsp.com/products/aco...urement/umik-1
You get the correction file for this mic.
However this is a USB mic, not balanced XLR.

ccww

Very interesting that MB has a chassis ear for dealer use. I would suspect that it is highly directional to find sources of noise. The microphones I am looking into are omnidirectional, although now you make me think that a directional microphone could lead to identifying acoustic hotspots. Of course auto NVH engineers use acoustic cameras to find problem areas, sadly far beyond my budget.

Thank you for sending that microphone option. I will have to research if it is compatible with the AudioTools app. The $230 option is recommended by the app developers and is auto-calibrating.

happyeds

wow?

Tump43

I must say I am impressed with the analysis shown in this thread. As a former AMG owner who swapped out the OE exhaust for a much louder/freer exhaust on my C55 in the quest for speed, I experienced very good results reducing interior cabin noise by 'dynamatting' (almost any relatively thick self sticking sound deadening product will do) the entire inside of my trunk -- especially the particularly thin plastic tub that holds the spare tire. I carried this solution over to my present, near stock E350 and have a very quiet car. The roads aren't great here in Westchester NY and the car is not quiet over every surface, but it is much quieter and more sedate. Only about ! to 2 hundred dollars in material and an hour and a half to install.

ccww

The final two acoustic windows have arrived at the dealership from Germany. I will pick them up tomorrow and take them for tinting which hopefully won't take too long.

In looking through EPC I have discovered something very interesting. Option code 839 ("Laminated Safety Glass with Acoustic Film") did not only include different windows. Rather there are 21 additional pieces of insulation and damping that correspond to the 839 code beyond what is in normal W212s.

The bass driver in each door receives a foam ring that surrounds the driver and presses against the grille in the panel to damp vibrations and perhaps prevent noise from the outside from passing through the opening in the speaker grille.

There are also various insulating pieces under the rear floor, along the transmission tunnel, and associated with the rear wheel arches.

Very interesting as MB now offers the Acoustic Comfort Package on various models including the W213 E which is advertised as including laminated glass plus additional insulation in the aforementioned areas. 839 on the W212 seems very similar, but never advertised as such and much cheaper. Also worth noting how many forum members have identified the rear of the car as an acoustic hotspot and considering the location of these extra pieces MB engineers seem to agree.

Anyway, I have ordered all of the original 839 insulation and damping pieces from Germany except the transmission tunnel insulators, which are no longer available. I do have some ideas for how to replace them, however.

I'm also waiting for my calibrated measurement microphone to arrive.

The current plan for data collection is as follows:

Use calibrated microphone and AudioTools to generate 1/3 octave band (or maybe octave band if 1/3 has poor repeatability) average dB graphs over several different stretches of road at defined speeds of 65-75mph. Will try to include very smooth asphalt, rough highly aggregated asphalt, and concrete at a minimum. I will take measurements for comparison in the four following conditions:

1) Car completely stock, no 839 components
2) 839 Acoustic Laminated windows and speaker surround insulating pieces in doors installed
3) Remaining 839 insulating and damping pieces installed in rear floor and rear tirewells. Will substitute or make my own substitute for the unavailable transmission tunnel part
4) With all 839 components and my previously planned installation of aftermarket MLV, CCF, CLD, and 3M Auto Thinsulate under trunk floor, behind trunk lining, and behind rear seat.

S-Prihadi

I am drooling.... CCWW

MBNUT1

Test plan approved!

ccww

Small update:

I picked up the rest of the windows today. Dropped them off for 40% 3M Crystalline tinting to match the rear and quarter windows. That will be done next Tuesday and I am hoping to install them on Wednesday. Hoping that it will be a one-day project and also that all the door speaker absorbers will have arrived as well by then.

Microphone will hopefully be here within a few days and I can start developing my testing procedure and post my baseline readings.

If you have any suggestions about the testing protocol or tests that you wish to see please let me know!

arcnspark

Regarding tire type, call Tire Rack and ask them the quietest tires for your car. I did it for my Porsche and they will be the next set to go on.

LILBENZ230

Out of curiosity, what tire did they say was the quietest? And what Porsche?

arcnspark

Continental Pro Contact
Car is 2000 996, stock tire sizes.
Factory x71(?) spore suspension

ccww

I've received the calibrated measurement microphone and spent some time experimenting with it over various test runs.

Immediately I realized that the microphone has a far greater low frequency (LF) extension than the microphone built into an iPhone. There is also a tremendous amount of very low frequency noise present in the samples. The very low frequency noise is unlikely to be strongly perceived nor can it realistically be addressed with any kind of acoustical treatments. These factors, as well as the previously discussed sensitivity of human hearing in the midrange, suggest that A-weighting which emphasizes the midrange is appropriate for overall dB figures moving forward. For example on one 30 second, 65mph run on average pavement the LZeq (Z-weighted or flat average level) was 95.5 dB whereas the LAeq (A-weighted average level) was 66.4 dB.

I plan on establishing A-weighted overall and 1/3 octave band readings for several different test stretches, or defined sections of road that I can repeatedly visit for measurements at a given speed. I want to test smooth and rough asphalt, concrete, and heavily aggregated asphalt. The test runs will be 65 and 75 mph depending on the type of road. However, before taking these measurements to establish my baseline figures for the unmodified car I feel that it is very important to understand how reliable the measurements are. This can help inform how many averaged runs are needed for each stretch and what kind of difference in measured levels can be considered statistically significant.

Today I conducted two sets of ten 30-second, 1 datapoint per second, 1/3 octave band average runs on defined stretches of a local highway at 65mph. Climate control was off and I used Distronic Plus to regulate my speed. If another car caused me to slow the run was discarded. The road surface was average, wind approx 6mph, outdoor temp approx 95deg F, tire pressures about 46 rear and 44 front. The onmidirectional microphone was held vertically oriented and located on the centerline of the car at the approximately at the same height and lengthwise position as my ear.

Decibels are a logarithmic way of expressing sound power, normally measured in watts. Naturally therefore conducting statistical operations related to measurement error on decibels and on sound power measured in watts may yield different results. This paper is informative on the topic. The paper presents two ways of calculating standard error (which is the standard deviation of the measurement set divided by the square root of the number of measurements) based on sound power instead of on decibels and suggests cases for their use.

I imported the runs' data into Excel and calculated the decibel mean and power mean (convert dB into W, average the W, convert back to dB) of the weighted and unweighted averages and of the 1/3 octave band averages. I also calculated the standard deviations and standard uncertainties of the dB readings. I then used both of the sound power-based methods suggested in the paper to calculate the standard errors.

Below the 2 kHz 1/3 octave band the difference between the dB mean and power mean were negligible. The standard deviation and standard errors also increased substantially above the 2 kHz 1/3 octave band, unsurprising considering the nature of auto noise and the lower amplitudes at those frequencies. The differences between the three methods of calculating standard error were negligible at all frequencies. They increased as expected with the increases deviations at higher frequencies, but the differences between the methods remained much smaller than the standard errors themselves, indicating that the choice of method is not particularly important with this data.

The short version of the three previous paragraphs is that for this data simply analyzing the mean, standard deviation, and standard error of the dB readings is satisfactory.

The good news is that the standard deviation of the overall A-weighted averages and 1/3 octave band averages at and below 2 kHz are low. This indicates that it will be possible to resolve relatively small meaningful differences caused by modifications to the car and that a large number of laborious data collections runs shouldn't be needed. I've attached a bar graph showing the standard deviation by 1/3 octave band for the 10 Northbound runs. The standard deviation for the A-weighted average for this dataset was 0.30 dB.


Most octave bands at or below 2 kHz have a standard deviation of about equal to or less than 0.5 dB. As a 95% confidence interval (CI) is more than sufficient for this fun project and a 95% CI can be found as ± 1.96*(standard error), theoretically achieving a the ability to quantify audible differences of 1dB should typically require only a few runs. For example, the overall A-weighted standard deviation is 0.30 dB. If four runs are used the standard error is 0.30 dB / √4 = 0.15 dB. This gives a CI of ± 0.29 dB, substantially below the commonly quoted 1dB threshold for audible change. I happen to be skeptical of the latter, but that is another topic of discussion entirely. Even with only one run, in which case standard deviation = standard error, the CI is ± 0.59 dB, an acceptable figure.

I've attached an example graph showing the average of the 1/3 octave band dB averages of the 10 Northbound runs. This data not frequency weighted. The distribution of the energy may be of interest to some. Please note that here I did use the power mean. Even though it is not needed for the statistical analysis and frequencies of interest in this project, it is the best way to sum average sound levels and here produces a more accurate figure.

Note that the A-weighted power mean was 67.6 dB and the Z-weighted (unweighted) power mean was 95.7 dB.



Any comments or suggestions on the analysis are greatly appreciated. I'm hoping to start collecting baseline data tomorrow for the unmodified car and complete that collection within a few days so that I can start the modifications.

ccww

I finally got a chance to spend some decent time taking baseline measurements. The 65 mph measurements are pretty easy to get since I can get them on 2 and 4 lane highways around where I live. Pick a stretch with good turnaround spots, measure from a defined starting point for 30 seconds, turn around, and repeat.

I knew that the 75 mph measurements on the freeway would be harder due to traffic and turning around. I didn't anticipate quite how hard it would be though! I drove over 200 miles yesterday working on measurements. I wanted to get 4 different datasets, two 65 mph and two 75 mph, but the 75 mph is just too time consuming to do repeatedly to track progress through different stages of the project.

So going forward I will have two 65 mph sections of road (smooth and moderate asphalt) and one 75 mph section of noisy concrete freeway. I realized the freeway section is simply too time consuming to do with many repeated runs over the exact same single stretch. The noisy concrete stretch that I wanted to use is about 10 miles between exits, making many passes unbearably tedious. Instead yesterday I did several runs each with multiple 30 second segments. From some analysis that I won't bore everybody with this time, I think that averaging the segments together for 1-2 runs is sufficient for the purposes of this experiment.

Perhaps the baseline dB results will be of interest to some here. The car is a 2015 E250 BT RWD with the 321 Sport Package, 17" non-staggered wheels with Michelin Pilot Sport A/S 3+ tires, and without any acoustical modifications or components from the 839 acoustics package as of yet.

65 mph - Very smooth asphalt - A-Weighted 62.5 dB ± 0.1 dB.
65 mph - Typical Oregon asphalt - A-Weighted 67.6 dB ± 0.2 dB
75 mph - Noisy Concrete - A-Weighted 72.4 dB ± 0.4 dB







It's rather difficult looking at the three graphs to see much of a meaningful difference. However, it is interesting to compare the two 65 mph datasets. Since they are the same speed and both on level ground, we can reasonably assume that the engine and wind noise aspects will be similar. That means that any difference between the two sound characteristics is due to road noise. Looking at a graph of the difference between them should show a spectrum representative of road noise from driving on aggregated asphalt. Remember that the uncertainty becomes much higher above 2 kHz.



The greatest contribution to the road noise in terms of frequency is broadly centered around a maximum difference at 400 Hz. Frequencies at and above 2 kHz don't seem to be associated with road noise. The negative association is likely spurious due to the high error above 2 kHz. The road noise significantly increases energy broadly at low frequencies, although it is important to remember that the very low frequencies are much less perceptible.

Another interesting comparison is between asphalt at 75 mph (from an dataset without a graph shown above, average A-weighted dB 71.8) and the concrete dataset at 75 mph. Note that since the concrete spectrum is more energetic, the graph shows the concrete spectrum minus the asphalt spectrum. This means that the bars with positive values show the additional frequency specific energy from driving on concrete as opposed to asphalt.



Here we see that concrete as opposed to asphalt primarily results in increased low frequency energy, particularly between 80-125 Hz and below 32 Hz. The very low frequency energy here is unsurprising as this stretch of road mercilessly finds every tiny rattle in the car. In the rest of the spectrum the differences are small and unevenly distributed.

Unfortunately looking at my schedule I am unlikely to be able to install the windows until the middle of next week at the earliest. Then more measurements and onwards to the 21 other pieces of insulation in the 839 package and after that my own additional acoustic materials. I am glad however that I took the time to get good baseline measurements, understand how precise they are, and use them to help understand the nature of the noise that I am battling. Hopefully the last two charts are of interest to you, @S-Prihadi , as you'd previously discussed identifying the frequencies involved.

If anybody is interested in a specific analysis or has a question that might be answered by the data I'm happy to do my best to attempt an answer.

S-Prihadi

Nice data, thanks C. I really like the last 2 graph.

ccww

I have discovered two more parts associated with the 839 (Laminated Acoustic Glass) options code. The rear fender liners have different part numbers for 839. The left side seems to be universal, A 212 690 63 30. However, the right side is dependent on engine so inquiring with your parts department is advisable.

Considering that road noise is the major concern in my project and that these address it closest to the source, I am optimistic that they will make a significant difference if their construction is notably different from the standard liners. I'm ordering them with my local dealer and one has to come from Germany. I'll be adding them to the experiment. They might make a relatively cheap and easy improvement possible for the others bothered by road noise without the need to remove the cabin carpeting, trunk lining, or door panels to install the other 839-associated components.

Hoping to install the rear door laminated windows tomorrow! Fingers crossed it goes smoothly!

Tom in Austin

Interesting, what is the "rear fender liner" exactly?

ccww

It is the large, black carpet-like lining that you can see in the wheelhouse from the exterior of the vehicle. Part 250 in the diagram.

S-Prihadi

Wow, different engine different rear wheel liner, interesting CCWW. Rear is hairy/felt like material.
I was thingking about this sound thingy when I removed front wheel-well/house hard plastic liner.
Why would MB waste money making vent like shape, but there are actually no holes in this vent like shape ??
Front wheelhouse liner has 2 portions. Front facing and rear facing. Front facing one has the vent like moulding.

Why no air holes ???

I suspect this is about reflecting noise....somehow.

It cost a lot of money for an injection moulding for plastic, to have these vent like shape but not serving air flow. It has to have some STRONG technical reason.
In a car industry savings money per unit of a car no matter how small is a big MUST, why waste $$ here ?

ccww

Very interesting, @S-Prihadi. I did not realize that those weren’t vent holes. I don’t know the purpose of them either. They could be acoustic, although their dimensions are relatively small compared to the wavelengths predominantly in question so I’m not sure what effects the shape could have. I’ll definitely be thinking about this one.

I should note that I found no 839-associated parts for the front wheel wells, or indeed the frontal area of the car at all. It does seem that most of the W212’s noise problem comes from the rear. Perhaps the firewall and engine noise capsule ameliorate the sound from the front axel.

S-Prihadi

Front wheelhouse well can not be hairy-felt/carpet like material, at least for front portion, I am sure because it goes to the engine bay as well and plastic is harder than hairy felt.
That plastic protect the lights and some other electronics between bumper and wheelhouse/well.

Like below :



However, the rear portion, I think may benefit some adding of acoustic materail because it is not firewall but after-the-door-cavity



I actually wanted to share with you and the gang about these front wheelhouse/well zones when plastic liners removed. For acoustic purpose. I forgot, my apology.
My alignment thingy have burden my mind