Bass Shaker/Tactile Transducer performance measurements

[reality_boy]

As part of my job I make audio effects that drive bass shakers and haptics for a video game. In order to do the best job I can, I wanted to develop a way to measure the performance of different transducers to both better understand there performance and to try and find an 'ideal' shaker to develop against. There is surprisingly little information out there on the performance of transducers or how to test them properly, so this has been a trial and error process for me. After a lot of work, I am starting to get reasonable data out of my tests. I'm hoping to share them here with the hope of coming up with better ideas on how to test shakers in the future.

For our needs were concentrating on frequencies below 200 Hz, the focus is on complimenting game audio with a tactile experience, so were trying to stay below the frequency where shakers start to produce more audible noise than discernable vibration. Bass shakers tend to have a very strong resonant frequency, so doing a single tone frequency sweep results in an unrealistic performance curve. That is to say if you feed a real life signal into the shaker, you will get far less power out of it then you would with a pure tone. This is easy to test out for yourself with a few tone generators. I experimented with driving the shakers with a white noise, pink noise, and even just white noise with a low pass filter at 200 Hz. All produced a fairly weak signal. The shakers are not efficient enough to replicate such broad spectrum signals. In the end I ended up playing 20 evenly spaced pure tones at 10 Hz intervals (10, 20, 30...200 Hz). This seems to generate a reasonable response, without saturating the shaker. I applied a random phase offset to the tones to try and break things up, so we don't end up generating a poor mans triangle wave.

For measurement of the vibration I'm using an M5StickC with custom firmware that reads the included accelerometer at 1 KHz, and runs a FFT on the output. This allows me to measure vibrations from 0-500 Hz fairly accurately. I then wrote custom software on the PC to both drive a signal through the sound card, and measure the resulting output. From there it was not hard to do an amplitude sweep and plot the resulting data out. I'm currently feeding all this through a ButtKicker BKA1000/4 amp. This amp is an older class A amp that is relatively linear up to around -20 dB input power. All my tests are designed to keep the signal below this point, so as to avoid any distortion from the amp or from saturation of the signal.

For the output data I'm currently logging the input signal gain in relative dB, and the output is measured in a dB scale that maps acceleration into something that approximates an audible dB scale. So 30 dB is approximately the threshold of feeling, and 100 dB is dangerous (will produce numbness with prolonged exposure). I'm not sure if it makes more sense to try and calculate amp voltage. I have a dummy load large enough to use to profile the amp, but I don't have a really good way of verifying the exact voltage that is being output with the given shaker. Basically I'm wide open to ideas here. Another issue is that I'm not capturing the full dynamic range of the shakers. The very tiny TT25-8 is outputting a signal close to the many times larger ButtKicker LFE, while in practice, properly mounted to a chair, the magnitude of the output does not compare. I'm not sure if the hanging rig just is way too light, or if there is something off with my procedure.

I have an older non technical paper on my setup, if you are curious. https://drive.google.com/file/d/1YaIsVrascj3dGshutN9zYxGpeE6TY5HS/view?usp=sharing

Here is the preliminary data.

 

[amirm]

Very interesting work. Thanks for posting it. I have bought some of these and hope to also quantify them as sources to generate vibration in audio gear.

 

[staticV3]

I don't have a really good way of verifying the exact voltage that is being output with the given shaker.

A basic TrueRMS multimeter should be plenty good enough to measure your Amp's output voltage with the bass shaker as load.

I use an Aneng AN870 to great effect. It has a flat frequency response from ≤8Hz to 2kHz:


Plus it has AC ranges at 0.02/0.2/2/20/200/2000V, so plenty of resolution no matter the amplitude.

 

[EERecordist]

If I were trying to solve this problem, I would look up perceptual quality metrics as a measure or to form measure.

Then I would seek out expertise from Dolby, THX, Disney, gaming developers large enough to have research - Microsoft would be one. Some may have developed tests before you. I would discuss with AES and SMPTE. Then I would talk to people at shaker and gaming chair companies like SubPac. They may have done the perceptual quality research. An additional clue is the outgoing QC for shaker manufacturers.

From that I would develop an industry standard and test. Standard coupling to the measurement device is non-trivial. Measured Pascal coupling to perception by modus may be different.

As I read your question, you are studying sub-cochlear stimulae. That is going to be processed in the brain from the peripheral nervous system and be aesthetically formed in the brain differently. But the peripheral nervous system stimulae may be associated with the cochlear response upstream from our complex trained audio cortex in the brain. I would speculate subsonic response in gaming or music it is highly conditioned to the individual.

 

[Timcognito]

How are they loaded? Are the moving anything like 100 lb weight?

 

[restorer-john]

How are they loaded? Are the moving anything like 100 lb weight?

A typical basement dwelling video gamer would be surely over 200lbs?

 

[Timcognito]

A typical basement dwelling video gamer would be surely over 200lbs?

Yes but the chair and his legs take some of that not sure of the distribution. I think that was some tung in cheek humor from the the bottom of the world about the bottom on the chair.

 

[reality_boy]

A basic TrueRMS multimeter should be plenty good enough to measure your Amp's output voltage with the bass shaker as load.

I use an Aneng AN870 to great effect. It has a flat frequency response from ≤8Hz to 2kHz:
View attachment 352118

Plus it has AC ranges at 0.02/0.2/2/20/200/2000V, so plenty of resolution no matter the amplitude.

I did this using a 200 watt 8 ohm dummy load (its a bit small for full power) and my cheap digital multimeter. I was able to come up with a formula that approximates the output voltage for a given input dB. However it falls apart at the ends (mostly the top end) and of course the signal I'm actually driving this with is not a pure sine wave, so the power is lower. Plus these range from 2-8 ohm impedance, and I'm not sure how I should be correcting the data for that. Should I be using watts? The only real reason to add proper units is so I can compare from amp to amp (or others can compare with there amp). That may not be as important as knowing the approximate frequency response, regardless of the input power.

Oh and I initially tried measuring with a 2, 4 and 8 ohm dummy load, and at various frequencies across the 20-200 Hz range I'm interested in, but the variance was not that great so I stuck with a 60 Hz test tone because I'm more confident my budget meter can accurately measure that.

Truthfully this amp is the wrong one to use. I have a smaller Emotive UPA-500 amp that I usually use for development work that is far more linear. But that can only get to 120 watts @ 4 ohm, while the BKA1000/4 gets to 1100 watts @4 ohms. The largest shakers need more than 120 watts to saturate them, and I need extra headroom so I'm not pushing the amp to 100% and adding in distortion. I need to break down and get a more modern amp that can output 1500 watts @ 4 ohms, so I can drive them all to saturation in a perfectly controlled environment.

 

[reality_boy]

How are they loaded? Are the moving anything like 100 lb weight?

These are unloaded. They are on a 3 point hanging rig that lets the shaker resonate freely without any impedance from the mount. It is similar to measuring a speaker outside of its cabinet in an anechoic chamber. I have setups to measure the vibration on my chair, and on a test platform with weights attached. However these are very inconsistent, in particular how I sit on my chair has a big impact on the frequency response.

This is an open problem, for sure how you mount the shakers makes a big difference, and of course trying to move a 2000 pound sim rig vs a 250 pound person in a chair makes a big difference as well. However measuring how my rig changes the response of a shaker is not as useful to people who don't have my chair (or weight distribution). That is part of the beauty of all this, I can experiment with different mountings and see how it effects things at the shaker and at the base of my chair. There is a lot more experimenting that could be done.

 

[reality_boy]

Very interesting work. Thanks for posting it. I have bought some of these and hope to also quantify them as sources to generate vibration in audio gear.

From my limited set of shakers I would use the ButtKicker Advance if your goal is low frequency vibration. The Dayton BST-300EX is a good close second if you are on a budget. The larger ButtKicker LFE is sluggish, in my experience, and has poor higher frequency response, although it can make your teeth rattle. And the budget Clarks that I have lack any sort of bottom end. They are probably fine for contact speakers but not really for bass shakers. Maybe the much more expensive platinum model has low frequency response.

I find they work very well to recreate real world vibrations (in our case vibrations from the engine and road), but they are hardly high fidelity. I hope that the market starts focusing on fidelity, and maybe in some small way these tests can help move the needle a bit.

 

[reality_boy]

Here is some sample raw data from my test, before processing. This is on the Clark TST239 transducer, with a fairly weak input signal. The graph is the FFT from my M5StickC transducer. I'm driving it with 20 pure tones from 10-200 Hz. You can see at this point that there is not a lot of noise showing up in the higher frequencies, but when I drive it harder the noise appears. One of the metrics I want to add is some sort of quantification of that noise. Oh, and 25 dB is the noise floor of my accelerometer.
--
Edit, fixed the broken image link

 

[reality_boy]

Oh, and I wrote all this for work, so I can't hand out the code, but I did post my FFT library for the M5StickC on hackaday a while back.

Good Vibrations | Details | Hackaday.io

<p>At work I needed to precisely measure the vibration frequencies of a device. I pulled out a <a href="https://shop.m5stack.com/products/stick-c?variant=17203451265114" target="_blank">M5StickC</a> that I had laying about and wrote some custom code for it. This is a small ESP32 based micro...

hackaday.io

It is possible you could use a contact microphone to quantify bass shaker performance, but I fear it would have poor low frequency response. I'm sure there are analog accelerometers out there that could be feed directly into an oscilloscope or recording software, but I'm no expert on that. The cheap mems accelerometers are not that hard to deal with, especially now that everything comes attached to a microcontroller these days.

 

[dennis h]

What EERecordist said...
I second looking into AES papers...
And thank you for starting this thread.