News

After spending (part of) my holidays analyzing the vibrations on a snowboard, I’ve decided to take out my e-bike for a ride.

E-bike sounds

E-bikes become more and more popular these days, and a lot of different brands exist. An important criterium is their operating range. But also the e-bike noise is an important aspect for both manufacturers and end users. 

When talking about “noise”, I do not only mean the decibel (dB) value of the overall noise level. This dB value is something you typically measure with a sound level meter or via an ISO sound power test. I wanted to get a better understanding of the acoustic quality comfort that e-bike riders experience.  

So I took my backpack and the Simcenter SCADAS XS with its binaural headset to record some sounds. As the measurements took place outside, I added the windcaps to protect the built-in microphones from disturbing wind turbulent noise.

Binaural headset with windcaps

Since the SCADAS XS has an additional 12 channels available, I decided to take it a step further and to instrument the e-bike with more sensors. After all, why not? 

Instrumenting the e-bike 

Next to the binaural headset, i added the following sensors: 

• 7 accelerometers at different locations on the e-bike: handle bars, suspension, saddle and frame
• An extra microphone with windscreen close to the electric motor 
• GPS signal for speed, position and altitude
• GoPro camera to document where i was riding

Instrumented e-bike

All of this sensor information allows me to not only look at the sound (quality) aspect of e-bike, but also to analyze its vibration behavior. How will the suspension perform on different road surfaces? Will the bike frame deform a lot? and how about the ride comfort at both the handle bars and the saddle. 

Sensors for e-bike ride comfort

Sensors at the e-bike suspension and electric motor

Once all the cables were safely taped to the bicycle frame, it was a matter of seconds to plug-in the different sensors to the SCADAS XS, push the “Read TEDS” button to get all the sensor settings and to put the pocket-sized XS in my backpack and pedal away.

The Simcenter SCADAS XS in action

The e-bike test ride

Rather than planning very strict measurement scenarios, I decided to take advantage of the great weather and simply drive around. As this is the “real-life” usage of the e-bike, it covers a lot of different user scenarios. It includes cycling on different road surfaces: Belgian blocks (cobblestones), asphalt road, gravel, etc. to focus on the vibrations, deformation and ride comfort. Going uphill and downhill reveals the effect of different loads on the electric motor.  

Just before starting each measurement, I quickly checked all sensors and signals via the wirelessly connected tablet.  

                                                                                                     Checking the signals using a wirelessly connected tablet

Once that was OK, a single push of the button starts the measurement. All data from all sensors is securely saved on the internal memory card inside the SCADAS XS. The internal battery of the SCADAS XS allows me to take measurements for a full day in full confidence. The only thing I needed to do is push the pedals (with a little help from the electric motor). 

What can you do with all this data? 

As I measured a mix of different sensors, a lot of data is available that gives detailed insight into the dynamic behavior of the e-bike in both the sound and vibration domain. 

As a first step, simply visualizing and replaying the time data in the Testlab Neo Desktop shows the correlation between certain events and signals. Selecting the sensors in the pivot table automatically adds the corresponding signals in the appropriate displays in a nice full-screen layout. How intuitive can it be!

 Dashboard view using Simcenter Testlab Neo Desktop

Notice that you cannot only select time signals from the individual sensors, but also combine it with the video data from the GoPro or GPS data that is automatically visualized in Google Maps. You can zoom in to specific segments, look at the frequency content and filter away certain annoying sounds during replay.

Vibration comfort is a big deal 

The measured vibrations can be used to visualize frame and suspension deformations but also to quantify the ride comfort using ISO human body vibration standards. Some examples are shown below:

• Using the measured vibrations on the bicycle frame
  • time domain animation
  • operational modal analysis
    
    

 Animation of the e-bike frame and suspension vibrations

• Using the measured vibrations on the front suspension

• • time domain animation 
  • operational deflection shapes  
  • characterizing the damping efficiency of the suspension as a function of frequency 

Front suspension characterization

• Evaluation of the ride comfort (based on ISO2631 human body vibration filters)
  • handlebar vibrations for hand-arm vibrational comfort 
  • saddle vibration for whole-body vibrational comfort 

                                                                                                                    Vibration Dose Values (VDV) according to ISO2631

It is clear that riding on Belgian Blocks is way less comfortable than riding on a smooth road. Both the animation as well as the ISO standardized VDV values further quantify and confirm this.

But does the e-bike sound right?

We’ve covered the vibration aspect of the ride comfort, so let’s now have a look at the e-bike noise (dis)comfort.

I compared the acoustic signals measured at the driver ear position between driving on a smooth road and on Belgian blocks. Living in Belgium, i can assure you that these cobblestone paths are quite popular around here. 

Comparing the sound quality of the e-bike on different road surfaces

Next to replaying the sounds and filtering away annoying noises during replay, further processing can be done on the sounds. I used some popular sound quality metrics such as Articulation Index, Prominence Ratio and Time Varying Loudness. For completeness, i also added traditional SPL and Spectrum analysis. 

As expected, the noise caused by the (very) uneven road surface is masking the electric motor whine. While the prominence ratio for the smooth road measurement is 2 dB higher, all the other metrics report that the cobblestone road has a very negative impact on my acoustic comfort. So, while the motor whine is still present, it’s not strong enough to be perceived.

I will let you in on a secret though. Exterior acoustic measurements without a dedicated, and preferably secluded track can be quite challenging. That’s especially important when dealing with low noise sources, such as electric motors. I’ve mentioned at the beginning that this was a nice spring day, and unfortunately, I was not the only one enjoying the good weather. 

Lovely birds disturbing our measurements

That is why, for a detailed sound quality evaluation, I’ve decided to conduct additional measurements inside. 

Which e-bike sounds best? 

I asked some of my colleagues to bring their e-bikes to the office, so we could benchmark different models. In total 4 e-bikes were tested, 3 of which accidentally had the same brand of an electric motor, but differently integrated. 

                                                                                                    4 different e-bikes on the test

Let’s first compare the sounds from the 4 e-bikes:

We see (and hear) some clear differences in the measured sound pressure levels, but Bike D has some unusually strong high-frequency content. Let’s have a closer look. 

                                                                       Interactive audio filtering

Using the interactive audio filtering quickly reveals the issue. It seems that the owner of Bike D is suffering from a very prominent break squeal noise, with peak amplitudes around 12.500 Hz.  Can’t hear it? It might be linked to your age, high frequency hearing loss is unfortunately very popular once you hit 50. In order to make a fairer comparison, let’s filter this frequency out and listen to the effect on the sound quality of Bike D: 

 Filtered sound for e-bike D

Removing the break squeal noise helped to obtain a more representative sound sample. While level metrics such as loudness, articulation index or sharpness were not strongly affected, we do see the effect as a 15 (!) dB drop in the prominence ratio. That’s quite a fix.

Below are the filtered e-bike sounds:

 Filtered e-bike sounds

Bike B and C seem to have very similar sound quality performance. With high articulation index values, prominence ratio at around 7.5 dB and 9 sones of loudness, both e-bikes can be considered very silent and pleasant sounding. Bikes A and D, on the other hand, affect speech intelligibility and result in higher loudness and tonality levels. Bike A especially might be considered as most unpleasant. 

Do real cyclists agree with those metrics?

The last step of a sound quality analysis process is to conduct a subjective listening test. Jury Testing collects subjective feedback from a group of listeners in order to understand their preference towards one sound or another. This is extremely helpful when benchmarking competitive products, designing new sound quality targets (e.g. how much loudness is too much?) or trying to create combined sound quality metrics that represent a certain product feature (pleasantness, robustness, sportiveness, etc.). 

Here’s an example of a typical listening test on our e-bike sound samples: 

Each time, a randomly created pair of sounds is played to the juror. The scale of 5 buttons allows to select the preferred e-bike sound when asking “Which of these e-bikes would you take on a long ride?”. This type of test (called the Sheffe’s method), gives more freedom when answering, since the scale has 5 options: 

• Definitely A 
• Rather A than B 
• Both would be OK 
• Rather B than A 
• Definitely B 

The same question is asked for the next pairs of e-bike sounds. At the end of the listening test, a preference list shows the ranking of the different e-bikes from most to least preferred. Other methods available in jury testing are A-B comparison, semantic differential, ranking or category judgement.

The final step is to verify that the (objective) sound quality metrics (or their combinations) correlate with the (subjective) preference as given by the jurors.   

Great things come in small packages: the SCADAS XS 

Isn’t it amazing how many different types of analysis you can do, simply using the measurements we did using the hand-held Simcenter SCADAS XS? That’s what we call measurement power: maximal testing flexibility using our pocket-size sound and vibration analyzer.

Source: Simcenter Blog, article written by FrankD and Aga Oltarzewska

To learn more about Simcenter SCADAS and to keep up with the latest from Dipolo, subscribe to our newsletter.