I recently upgraded my Pista for an MXL2. Looking at my data from Sebring I seem to have excessive noise on the accelerometers. Is this a mounting issue or do I need to add some smoothing?

I don't remember my Pista sensor showing this much "noise", and yes I know Sebring is bumpy. I'm seeing 1G high frequency spikes in the X & Y Accelerometers:




Any suggestions on cleaning this up?

Thanks,
Greg

 

Be willing to bet the sample rate is at the default 50 Hz and yes, smoothing is recommended in Analysis.

I reduce sampling to 20 Hz and jack smoothing up to 7-8-9 in my analysis.

YMMV

 

Thanks Peter,

Yep, sampling is set for 50Hz. When I set filtering in RS2 analysis to 10, the plots change only a little, about 50% less peak noise.

Besides reducing the amount of data collected at the slower sampling rate that would miss some of the peaks, wouldn't it still pick up the amplitude of some of the peaks, just at a lower frequency? Wouldn't I see the same amplitude spikes, or does changing the sampling rate also affect the internal filtering in the MXL2?

Do see this much noise in other cars unfiltered?

Thanks for your help!!

Regards,
Greg

 

The filter in RS2 is a moving average filter. The filter that was in the MXL2/MXS/MXG as a dash level filter when they were released was more like a Butterworth or Chebyshev filter. This is import to realize how the filter is working. If you set the average to 10 when you have 100 Hz sample rate, it is not as significant as if it was a 10 Hz rate.

The sensors in the AiM products are pretty sensitive and do not have the high level of device level filtering that some other systems have. It's a balance to select the right logging rate and smoothing to get the useful information but not obscure the details. I personally use a balance between the GPS measures and the actual accelerometer data. It depends on what you are looking for as to what is best to use.

How is your dash mounted? The mounting setup can create a lot of noise. Remember, most guidelines to mount an accelerometer involve hook and loop fastener (Velcro) to absorb some of the noise.

 

Matt's provided good information.

I neglected to consider the sensor filtering in the device. Please investigate testing several levels of sensor filtering as well, this may help the output from the unit and leverage the smoothing in analysis so the waveform is more usable.

For whatever reason, the new-series AiM dash accelerometer output seem to be noisier than any of the other systems I sell, including MoTeC and Bosch.

These dashes all come with rubber vibration damper mounts (4mm x .7mm thread) and AiM obviously wants users to implement these, though I have seen FEW installations in the field use them. How is yours mounted?

I recommend only 3M High Strength Dual-Lock fastening "velcro," but I've not seen that used for dashes (and wouldn't recommend that for something your pushing buttons on laterally!), only on standalone loggers, like the EVO4 and EDL in my car.

 

Thanks! That helps a lot. I suspected the amount filtering was tied to the sample rate from configuring my fuel level sensor. I'll have to play with it at my next event.

It is amazing the filtering that is now possible with DSPs/multicore CPUs vs. the old analog solutions.

It's mounted to the steering column with a bracket in front of the 944 dash. I don't think it's reinforced at the top so that might accentuate any vibrations.

The car is in Milwaukee and I live outside of Chicago, I'll have to wait for my next trip North and get a picture from the shop.

Thanks,
Greg

 

We are having a similar problem with our FSAE car and a MXL2. It's running a steel space frame chassis and a single cylinder KTM 450 engine so it's a really noisy environment and I am having trouble cleaning up the date.
You can see traces comparing the GPS accel and the dash accelerometers and the dash accelerometers are basically worthless right now.
The MXL2 is mounted to an aluminum plate with the included damper standoffs, and the plate is mounted to the front roll hoop with rubber insulated hose clamps.

There is well over 1 g of vibration noise and the highest filtering in the AIM software isn't helping.
There are methods with matlab to denoise the signal, but I don't really want to have to export and try to figure out how to re-import every time we run.

Has anybody developed a math channel to denoise signal data?



Lat Accel



Long Accel