Using $20 Bluetooth stereo conduction speakers to emulate a Beech Band, Tuning Fork, Metronome Watch, or other similar vibrotactile device (Not to be used to emulate a chest mount Cue-1)

I. Introduction and Disclaimers

II. How It Works

III. The Set-Up

IV. Cautionary Safety Notes

V. DIY 3D-Printed Button Isolator

VI. Choosing Which Brand to Use

VII. Sample ChatGPT Prompts to Create WAV Files

VIII. Post Your Feedback Here

I. INTRODUCTION AND DISCLAIMERS

I came up with the idea of using a pair of stereo Bluetooth conduction speakers as synchronized vibrotactile devices as part of a project for a friend who was looking for a way to use electronics to continuously reproduce the therapeutic vibrations from a tuning fork. It then occurred to me that conduction speakers could be used to reproduce any type of buzz pattern. Using two synchronized stereo vibrators might provide additional benefits compared to single point vibrators like the Beech Band.

While I have tinkered a bit with conduction speakers for this project, I have not yet seriously experimented with various patterns to use as therapy for my own PD. I am very happy with the results from my 8-channel Blue Buzzah vibrotactile gloves (and Blue Buzzah vibrotactile shoes), and I have not been further motivated to develop this additional 2-channel buzzer idea. So, while I can confirm that the conduction speakers described below can actually reproduce synchronized stereo buzz patterns, I can’t personally say whether this type of single/double buzzer intervention is effective for PD.

Perhaps someone here might be interested in developing the idea further?

IMPORTANT: Please read the cautionary safety notes below in section IV before trying a conduction speaker as a vibrotactile device.

II. HOW IT WORKS

A typical Bluetooth music speaker is designed to move a speaker cone in air to produce sound waves as shown in the top right diagram. A Bluetooth conduction speaker is the same, except it is designed to move a button surface instead of a cone. The button must then be put in contact with a resonant external object (like a window) which can then move air in place of the cone. (See: https://www.youtube.com/shorts/4mK91RBIkRw )

Conduction speakers set at low volume for safety may be used in the same manner to transmit vibrations to thick fleshy areas of skin to emulate a Beech Band, metronome watch, or tuning fork.

III. THE SETUP

1. Use ChatGPT to create a WAV file of the vibration pattern you want to try out. Unfortunately, ChatGPT cannot make MP3’s which are more compact than WAV’s. See Section VII for example prompts to submit to ChatGPT to make your WAV files.

2. Use your phone and some app that allows you to import user wav files to loop-play your WAV file through your Bluetooth conduction speakers, just like you might play any song file. I use an app called Anytune because this app allows me to also adjust file playback speed and frequency independently while the file is playing, eliminating the need for having to make a new file on ChatGPT every time I want to audition different playback speeds and frequencies. I have the pro paid version of Anytune, but I think the free version does everything needed here.

3. Affix the speaker to the part of your body you want to vibrate. Avoid head, brain, eyes, ears, and heart/chest. Use low volume setting. I used either athletic tape or elastic ankle wrap to affix the speaker. They can also be held in place against an ankle by a sock.

IV. EXERCISE CAUTION

Note that I am not a doctor and only have a layman’s understanding of medical issues surrounding vibrotactile therapy. The notes below are my own common sense bits of advice for myself, and are not meant as safe informed advice for others.

USE AT OWN RISK.

CONSULT DOCTOR BEFORE USING CONDUCTION SPEAKERS FOR VIBRATION THERAPY.

AVOID HEAD, BRAIN, EYES, EARS, AND HEART/CHEST FOR OBVIOUS SAFETY REASONS

NEVER EVER USE WHILE SLEEPING.

You might see “conduction speakers” also referred to as “bone conduction speakers” and get the erroneous idea that they are designed to be put on your body. But this unfortunate term comes from “bone conduction headphones” which produce much much smaller vibrations that are designed to reach the inner ear through bone/body conduction. “Conduction speakers” set at high volume are capable of producing very strong vibrations that much more powerful than those from bone conduction headphones. Strong vibrations like these are made to be placed on inanimate objects that act as passive resonators. They are not designed to be used safely on the human body. Caution therefore must be exercised to make sure the speaker volume setting is always very low before applying to your body. That said, to put things in perspective, the vibration strength from the conduction speakers - even at maximum setting - are still much lower than a hand sander, vibration plate exercise machine, or jack hammer.

The Beech Band uses ERM vibrators that make X/Y-axis vibrations that shake the skin surface tangentially, instead of towards and away from underlying bones. I am pretty sure the Cue-1 also uses X/Y vibrations. In contrast, conduction speakers produce z-axis vibrations that move towards and away from the body - like tuning fork bottoms and the buzzers in the Stanford vibrotactile gloves. This z-axis motion is great for localizing very small vibration points on fleshy areas like the fingertips, but if placed near a bony area at high volume, you will get “bone conduction” throughout your skeleton. If placed over a body organ, for example the heart, z-axis vibrations are focused directly downward towards the organ. Even when placed on my ankle bone, if I turn up the volume too much, I can feel the vibrations getting into my chest. So, my preference is definitely to stay away from bones, and to only buzz with low amplitude around fatty or muscular areas in my arms and legs.

Tuning forks also z-axis-vibrate the skeleton slightly when pressed against bone as shown here: https://www.youtube.com/watch?v=ws9kYRD2FxY . However, the vibrations of a tuning fork are far lower amplitude than can be achieved by a conduction speaker with the volume up. Metronome watches can reach fairly high vibration intensities, but such watches are not capable of continuous vibration.

As a quick test, I compared the feeling of one of my small Blue Buzzah buzzers to a conduction speaker. Holding a small Blue Buzzah buzzer against my chest, I could feel the vibrations in my skin, but the sensation did not noticeably penetrate deeply to my heart. Testing the conduction speaker briefly against my chest at high volume, I immediately felt strong vibrations penetrating deep into my chest. The sensation was very uncomfortable. Have you have ever worked with a hand sander for too long and felt your hands go numb? That is what I imagine could happen to my heart if I held the conduction speaker on my chest for an extended period. One might simply resolve to keep the speaker volume way down, but how can a user reliably determine the safe level of vibration near vital organs? Plus, if a user gets a positive result at a low level, there is a natural temptation to turn the dial higher to see if an even better result might be achieved. So, as far as I am concerned, I am always going to stay away from any kind of buzzing near my heart. The same caution applies for me when it comes to buzzing near my brain, eyes, ears, and any of my vital organs.

In conclusion, while conduction speakers can reproduce an infinite number of patterns and waveforms, be aware that they are not designed and tested to be used safely anywhere on the human body. Using high volume settings can result in large concentrated vibrations that may cause medical problems (for example, inflammation), especially if you buzz near your brain, eyes, heart, or any other vital organ. Use at your own risk.

V. DIY 3D-PRINTED BUTTON ISOLATOR

The vibrating button on a PocketTunes conduction speaker is nearly 1 inch in diameter. (See bottom right diagram.) Vibrating buttons of this size create large diffuse areas of vibration sensation when placed on the wrist or finger. I made a cylindrical 3D-printed isolation shell that enabled me to use a 5.1 mm button with a PocketTunes speaker as shown in the diagram. The resulting vibrational sensation was identical to the small buzzers we use in our Blue Buzzah. The amplitude was much smaller, and more importantly, the vibration sensation was much more localized.

According to Dr. Tass, the localization of vibrational sensation is very important in vibration gloves. I have no idea whether the same is true with single buzzer devices like the Beech Band.

VI. WHICH BRAND IS BEST?

I experimented with five brands on Amazon and settled on PocketTunes as my best alternative.

Pros for PocketTunes:

1. At 35 grams per speaker, these were the lightest/smallest I could find. As a comparison, another brand I purchased (the Okjew) was much heavier at 75 g. My beefy Apple Watch Ultra II weighs 72 g with band.
2. A PocketTunes order comes with two speakers that can operate in stereo for under $20.
3. The button is the smallest of any model I found (buzzes are more localized).
4. There are more buzzing pattern possibilities with a stereo pair like this.
5. These speakers have the smallest power level of all the models I considered. So, they are less likely to provide excessive vibrations if a user accidentally turns phone volume up too high.
6. Using at low volume gives many hours of continuous use.

Cons for PocketTunes

1. Even though power is smallest available, it is still way more than needed. It’s OK as long as user keeps volume low, but when volume is high the vibration can travel throughout your body. May be dangerous if put on head or chest.
2. Even though the button is smallest available, it is still larger than I prefer for localization.
3. I wish they were even smaller like an actual Cue-1 or Beech Band
4. I don’t think they turn themselves off if you forget to turn them off after a session

Honorable mentions for two other brands:

1. The R1Y brand features an internal micro SD card slot that enables the speaker to play mp3 files without the need for a smartphone. Its big drawback is its whopping size at 134 g. Also, the R1Y cannot not link to another R1Y in stereo, precluding any type of 2-channel vibration pattern.
2. The Duramobi brand (weight 55 g) has a multiple speaker mode that allows up to 100 speakers to be paired in mono. It also includes a stereo pairing option, but additional speakers must be purchased separately. It was not clear whether the Duramobi allows multiple stereo pairings.

VII a. USING CHATGPT TO MAKE A TUNING FORK TYPE WAV FILE

Note that ChatGPT cannot make WAV files by itself. Instead, when you direct ChatGPT to create a WAV file, it writes a Python computer code which ultimately creates the WAV file when ChatGPT runs the code internally in its “sandbox”. To get ChatGPT to seamlessly write and run the Python code to make your wav file, you will need to have a free signed-in account. If you only have a free account, you will quickly exceed the free level of computation, and ChatGPT will only make around 2-3 WAV files for you per day. “Three WAV files”, sounds like a lot of files since you only need one to operate a speaker. But, if you start experimenting to see how various patterns feel, you will quickly use up your free daily computation allowance. That said, I get by somehow with a free account.

Note also that the length of the file you create will impact upon your daily allotted the computation load. Files without randomized elements (like the tuning fork emulation below) can be short, since the speaker is producing a monotonous repeating waveform. Long treatment sessions can be achieved by simply looping the playback on your phone playback app.

If you are using a vibration pattern with random elements as found the second example below, you will need longer files to avoid becoming habituated to a small number of the repeated randomized patterns. Unfortunately, ChatGPT has a file size limitation that limits WAV files to around 2 minutes. You can get around this limitation by piecing together several files in the Anytune app. Or, if you have a Python compiler, you can ask ChatGPT to just write the code that you can execute on your computer.

Here is a prompt for a simple 128 Hz tuning fork emulation:

“Hello ChatGPT! I want you to make a WAV file that has the following characteristics:
The name of the file should be 128Hz_Tuning_Fork. The length of file should equal 30 seconds.
The WAV file itself should consist of a sine wave of frequency 128 Hz.”

Here is a prompt for a tuning fork emulation that vibrates at new random frequencies that change every 5 seconds:

“Hello again ChatGPT! I want you to make a new WAV file that has the following characteristics:
The name of the file should be Random_5s_Tuning_Fork. The length of file should equal 2 minutes.
The WAV file itself should consist of a sine wave whose frequency is changed every 5 seconds. The successive frequencies should be chosen randomly between F_LOW = 100 Hz and F_HIGH = 140 Hz.”

VII b. USING ChatGPT TO MAKE A BEECH BAND TYPE WAV FILE

Below is an example prompt to get ChatGPT to make a wav file that emulates a Beech Band buzz pattern. The ON/OFF times and FREQUENCY value I include below were measured from a Beech Band audio recording I found on the web.

A nice thing about the conduction speaker process is that you are not stuck with using the parameters that are fixed in an actual Beech Band. You can tweak the FREQUENCY, ON TIME and OFF TIME to suit your own personal preference.

“Hello again ChatGPT! I want you to make a WAV file that has the following characteristics:
The name of the file should be Beech_Band_Period_1121_ms. The length of file should equal 2 minutes.
The WAV file itself should consist of a pulsed square wave of frequency 115 Hz. Each pulse should have an ON TIME equal to 543 ms followed by an OFF TIME of 578 ms.”

Your imagination is your only limit to the type of patterns you can use to experiment. Here is a prompt for a Beech Band Pattern with random frequency shifts. (Note: The reason you might want to have random frequencies is to avoid the problem of “habituation”, where your brain eventually becomes desensitized to a repeated stimulus):

“Hello again ChatGPT! I want you to make a new WAV file that has the following characteristics:
The name of the file should Beech_Band_rand_freq. The length of file should equal 2 minutes.
The WAV file itself should consist of a pulsed square wave. Each pulse should have an ON TIME equal to 543 ms followed by an OFF TIME of 578 ms.” Each pulse should have a randomly chosen square wave frequency between F LOW equal 170 Hz and F HIGH equal 250 Hz.”

For illustration purposes, here is the same file with the pulse length shortened while keeping the overall ON + OFF pulse period constant:

“Hello again ChatGPT! I want you to make a new WAV file that has the following characteristics:
The name of the file should Beech_Band_rand_freq_square. The length of file should equal 2 minutes.
The WAV file itself should consist of a pulsed square wave. Each pulse should have an ON TIME equal to 500 ms followed by an OFF TIME of 621 ms. Each pulse should have a randomly chosen square wave frequency between F LOW equal 100 Hz and F HIGH of 120 Hz.”

VII c. USING CHATGPT TO MAKE A METRONOME WATCH TYPE WAV FILE

“Hello again ChatGPT! I want you to make a new WAV file that has the following characteristics:
The name of the file should be metronome_pulses_1000ms. The length of file should equal 1 minute.
The WAV file itself should consist of a pulsed square wave of frequency 60 Hz. Each pulse should have an ON TIME equal to 50 ms followed by an OFF TIME of 950 ms.”

VII d. USING CHATGPT TO MAKE A CUE-1 TYPE WAV FILE

The Cue-1 is probably the most advanced and most researched single point vibration device. There is a lot of info on the web, for example see this NoSilverBullet video: https://www.youtube.com/watch?v=QIsYmQai_pY . And this free PubMed article: https://pubmed.ncbi.nlm.nih.gov/40809990/

According to the above article, the Cue-1 delivers a proprietary waveform that is delivered optionally in two modalities: “High-Frequency Vibration (80–250 Hz) and Low-Frequency Cueing (pulse and rest length between 100–2000 ms)”.

I would guess that the high frequency vibration mode is essentially like the continuously vibrating tuning fork discussed previously, except the waveform is not a simple sine wave as with a tuning fork. Presumably, the Cue-1 designers developed their own special waveform that is more effective for PD. Someone trying to emulate a Cue-1 in “high frequency” mode could try the previous tuning fork waveform with frequency between 80 Hz and 250 Hz. To easily experiment with the type of waveform, a DIYer could try square waves, triangular waves, or sawtooth waves. For example:

“Hello again ChatGPT! I want you to make a new WAV file that has the following characteristics:
The name of the file should be 200Hz_CUE1_HFmode. The length of file should equal 1 minute.
The WAV file itself should consist of a sawtooth wave of frequency 128 Hz.”

The subjects in the above study link used the low-frequency cueing mode with ON=800 ms and OFF=800 ms. This mode is like a Beech Band ON/OFF pulse except the actual waveform that is being pulsed is Cue-1’s special proprietary waveform whose frequency and shape are unknown to us. Someone trying to emulate a Cue-1 in “low frequency cueing” mode could try the previous Beech Band waveform with randomized frequency since we don’t know the actual Cue-1 pulse wave frequency. To easily experiment with the type of waveform, a DIYer could try square waves, triangular waves, or sawtooth waves. For example:

“Hello again ChatGPT! I want you to make a new WAV file that has the following characteristics:
The name of the file should CUE1_rand_freq_LFC_triangle. The length of file should equal 2 minutes.
The WAV file itself should consist of a pulsed triangular wave. Each pulse should have an ON TIME equal to 800 ms followed by an OFF TIME of 800 ms. Each pulse should have a randomly chosen triangular wave frequency between F LOW equal 100 Hz and F HIGH of 140 Hz.”

A final note about the Cue-1. This device was specifically designed to stick to a user’s sternum. The mass of the buzzer is only 17 grams. A PocketTunes speaker’s mass is over twice as heavy at 35 g. The PocketTunes is also much larger. But, more importantly, the Cue-1 was specifically designed and tested to only produce low intensity vibrations that won’t damage a user’s heart, even at the highest setting. This is not the case with DIY PocketTunes speakers that can produce much stronger vibrations if volume is turned up, which could be dangerous. If you want a vibration device for your chest, consider supporting the Cue-1 company and purchase one of their actual Cue-1 devices. My understanding is that they come with a no-risk return policy.

VII e. USING CHATGPT TO MAKE A PSEUDO RVS TYPE WAV FILE

The PocketTunes conduction speaker on Amazon comes with two buzzer units that can be played in stereo. While we need four buzzers to emulate a real Tass RVS-type pattern, we can make a two-buzzer stereo pattern that employs some of the same general ideas:

“Hello again ChatGPT! I want you to make a new stereo WAV file that has the following characteristics:
The name of the file should be PSEUDO_RVS. The length of file should equal 2 minutes.
The WAV file should consist of a series of unique pulse sequences.
Each sequence should be made up of 12 sine wave pulses followed by a rest period of length 1.336 seconds. Each sine wave pulse should have a frequency of 250 Hz. Each of the twelve pulses should have an ON TIME equal to 100 ms followed by an OFF TIME of 67 ms. You should use a random number generator to decide whether each sine wave pulse will be written to the left or right channel.
When you write a pulse to one channel, you should leave the alternate channel blank for the entire 167 ms pulse ON+OFF period.”

And here is that same prompt that employs randomized frequency:

“Hello again ChatGPT! I want you to make a new stereo MP3 file that has the following characteristics:
The name of the file should be PSEUDO_RVS_randF. The length of file should equal 2 minutes.
The MP3 file should consist of a series of unique pulse sequences.
Each sequence should be made up of 12 sine wave pulses followed by a rest period of length 1.336 seconds. Each sine wave pulse should have a frequency chosen using a random number generator between F_LOW = 180 Hz and F_HIGH = 280 Hz. Each of the twelve pulses should have an ON TIME equal to 100 ms followed by an OFF TIME of 67 ms. You should use a second random number generator to decide whether each sine wave pulse will be written to the left or right channel.
When you write a pulse to one channel, you should leave the alternate channel blank for the entire 167 ms pulse ON+OFF period.”

VII f. USE CHATGPT TO MAKE YOUR OWN TYPE OF WAV FILE

What ideas do you have for a buzz pattern you would like to try out? Some new pattern you found described in a PUBMED research paper on vibrotactile devices for PD? An electric tuning fork that randomly cycles through the 10 top new age healing frequencies you read about online? A metronome watch that clicks at randomized intervals? A Beech Band that uses randomized normalized gaussian noise pulses of randomized length? A Cue-1 buzzing with the musical 5-note sequence from the extraterrestrial Close Encounters movie? ChatGPT makes it easy to try out any vibration pattern that can be played by a stereo pair of speakers:

“Hello again ChatGPT! I want you to make a new WAV file that has the following characteristics:
The name of the file should be….
You fill in the rest!

VIII. SHARE YOUR FEEDBACK!!!

If anyone tries the idea of using conduction speakers as vibrotactile devices, please respond here and let us know what kind of buzz pattern you used and what kind of results you obtained!

Good Luck!

IX. AN EXCITING BODY BUZZER POSSIBILITY

At one point in my vibrating glove journey I wondered if I could enhance my Tass vCR glove therapy by making a Blue Buzzah system that buzzed widely disparate points on my body instead of fingertips. By buzzing left/right foot, hip, wrist, and jaw, I obtained an immersive 3D effect that felt like it took me to a level beyond Tass nominal vCR glove buzzing. There were two problems with this “Body Buzzah”. First, it took 20-30 minutes to put on the network of wires. And worse, the wires were impossibly confining. I tried to go on a hike and kept pulling at the wires, sometimes even pulling buzzers out of their tactors.