DIY Bone Conduction Headphones

EDIT 07/18/17: This post has been getting a lot of hits lately. Since I don’t have comments enabled for my site, if you have questions or would like to comment on anything, please feel free to email me directly. My email can be found on my CV (I won’t post it here so the spam bots don’t eat me!)

I discovered something fascinating while browsing around the other day: headphones that transmit sound directly to your skull.  This method of sound transfer has been dubbed bone conduction.  All you do is press the little transducers up to your temple, jaw, or skull, and the vibrations in the little electrical device transfer to the waves through the solid bone medium to your inner ear.  This way you can listen to things without blocking your ears with big cans or buds.  Rather than go out and purchase one of the little premade units, I decided to make my own DIY bone conduction headphones.

I have my own issues with headphones.  I am the type who strongly prefers earbuds.  Sure, you can get better sound out of headphones, but I find that the large strap and bulky padding smashes my ears and glasses together in an uncomfortable way.  Furthermore, the shape of my skull with causes the applied pressure of the headphones to pull my glasses out of alignment.  That puts pressure on one side of my nose or the other, and it makes my vision all out of alignment.  To make things just that much worse, my very fine hair is easily molded, a quality that makes it easy to get ready in the morning but causes instant hat hair.  Headphone bands give me this weird wave in my perfect, voluminous follicle coif.  If you look at the bone conductivity headphones available on the market, they all use a strap to keep them in place like a normal pair of headphones.  With the extra pressure required to push the transducers up to my jaw, I can only imagine that they would have even more issues.

So I thought to myself: how can I make a set of DIY bone conduction headphones with properties closer to earbuds?

Answer: use the straps that you wear every day, your glasses.

The Build

For this build, I purchased equipment from Adafruit which was very similar to the stuff recommended in the Ruiz Brothers build.  I used

I followed the build similar to the Ruiz Brothers instructions linked above with substitutions made for the parts I sourced elsewhere, such as the battery pack.

I glued the completed breadboard to the battery pack and called it a success.  My first tests (using the transducers on my apartment’s door to introduce my partying neighbors to Norwegian black metal) proved the build to be a success.

For future modifications to this, I plan on soldering a jumper on the gain pins instead of using the removable jumper.  I also would wire the headphones directly to the board rather than using the European headers included with the breakout board.

The amp circuitry and transducers for the DIY bone conduction headphones.

The Buds

As you can see in the picture of the build, there are these strange brown boxes wired to the device.  These are the containers for the bone conduction transducers.  These were designed in Fusion360 to fit the transducers quite snugly.  The Adafruit page offers a technical drawing of the transducer, but I found that my units were slightly off of this specification.  I wanted the transducers to fit in the housings without much room to rattle and lose volume and quality, so I went with my measurements.

These are the housings for the bone conductor transducers.

The hooks on the end of each housing are designed to fit onto the earpieces of my glasses.  The plastic glasses I wear have a slightly tapered profile.  This means with a little push of the parts onto the earpieces, they stay in place without sliding off.  The inside of each of the hooks on the housing are slanted, e.g. one end is more open than the other.  This means that the housings are not interchangeable, and fit on either the left or the right side of the glasses.  The cover of the housing fits into the remaining room around the transducer and these can be glued into place.  The slit at the bottom provides enough room for my cheap wires to go through or a bit bigger gauge if you are concerned about sound quality.

The parts fit onto the earpieces of my glasses.

The housings were printed on a Makerbot 5th Gen in dark brown PLA filament.  The print was scaled at 102.4% from the actual design size.  This scaling was based on a previous calibration curve of the printer PLA shrinkage and print accuracy.  Like I said, I wanted the tightest tolerance the little 3D printed parts could manage.

The little widget below is clickable, and it will let you move the 3D object around to see for yourself how it is designed.

This is the “left” container.

This is the cover which fits over both the “left” and “right” versions of the above model.

Just want to try the parts out for yourself?  The three .stl files you will want to download are below.






The Fit

With my little DIY bone conduction headphones complete, it is time to try them out.



A handsome man with his DIY bone conduction headphones

Look at that handsome fellow in the picture.  The transducer housings fit as expected on my glasses.  The sound transferred well to my skull.  If you have ever used bone conducting transducers before, think of it like a medium press sound transfer rather than a super hard press.  The brown color is still stands out a bit from the color of my glasses and my hair.  Finishing paint would be necessary to make it look nicer.

The sound quality was pretty good as far as these little transducers can go.  They won’t be replacing my in-ear monitors any time soon, but they are sufficient to listen to spoken word audio.

Bonus trial: my mother is partially deaf due to Meniere’s disease.  She tried my DIY bone conduction headphones and was able to hear things on better on her deaf side.  Further development of this technology may prove to be useful for sufferers of mild hearing loss who do not desire the full hearing aid.

Blue Canola Oil or Dye Another Day

Did you know that if things go just wrong enough, you can create blue canola oil?  I didn’t either until the other day when I decided to make some fries to go with some burgers I made.

Hungry Yet?

It was going to be a delicious meal.  Some big Idaho potatoes were into thick fries (the best kind when you make them at home).  I was making sliders with ground beef and just a little bit of sausage.  I had fresh buns, salad greens, Vidalia onions, and spicy mustard all ready to go.  The fries were timed just right with the patties so that everything would be ready to go at the exact same time.  It was going to be perfect.  I grabbed a metal colander to scoop out the fries (I couldn’t find a slotted spatula), dipped it in to grab my deep fried potato prize, but suddenly my cooking oil turned a bright turquoise.

I scooped out the fries and laid them to cool on a paper plate.  The color change had gotten to the fries.  They were contaminated.  I wasn’t sure what caused this horrific turn of events, but something told me that these were inedible.

Blue French Fries

Unless it comes with the suffix -berry, I’m hesitant to trust blue food.

I was stymied.  What could create blue canola oil?  I knew it had to be linked to the metal strainer, but how?  I always thought the strainer was steel, had it magically turned into copper?  Even if it did have copper in it, how did that suddenly dissolve into canola oil?

And then the solution hit me:


Remember a few posts ago when I was going over my Model M mods?  You know, this one.  Remember how I dyed my keycaps blue?  Yeah…. about that…

It seems that I somehow mixed in my crafting colander with my food-safe cookware.  The iDye Poly I used for that project was formulated to dye polymers.  Most polymers a craft oriented person would want to color are hydrophobic in nature.  It follows that a very hydrophobic liquid (canola oil) would dissolve the dye quite readily.

Since this little event, I have separated my metal strainers in a more obvious manner.  Accidentally using lab ware for human consumption is no joke.  Now the real question: what do I do with all this blue canola oil?


Blue Canola Oil

Fountain Pen Ink pH – A Comparison of Five Commercial Inks

A while back, there was some discussion about fountain pen ink pH.  During that discussion, I offered to test inks and my friend from the Geekhack forum, CPTBadass, offered to donate some samples for my tests.  I received samples from my good CPT in the mail, and I went about running tests.  This post is the outcome of that work.

The Samples

Four samples were procured from CPTBadass and one was supplied by the author.

Noodler’s Baystate Blue – a vibrant royal blue ink
Rohrer & Klingner Scabiosa – an iron gall ink
Noodler’s North African Violet – an intense violet ink
Diamine Soft Mint – an ink that appeared turquoise or aqua in the bulk solution
Noodler’s Polar Brown – a somewhat rust tinged brown ink

Upon receipt, the samples were transferred to 20mL plastic scintillation vials.  The original containers’ opening was too small to accomodate the probe used in the pH experiments.

1. Shake Tests


Samples in scintillation vials were given a quick shake and observed for fluid and surface characteristics.


Ink Observations
Noodler’s Baystate Blue Ink left a thin layer coating the wall that persisted for a long time.  This thin layer did not show any detail of the plastic surface.  It produced bubbles,
Rohrer & Klingner Scabiosa Ink left a very thin layer coating the wall that persisted for some time.  The layer highlighted plastic imperfections of the bottle by contrast.  It produced bubbles.
Noodler’s North African Violet Ink left a very thin layer coating the wall that persisted for some time.  The layer highlighted some plastic imperfections of the bottle by contrast.  It produced bubbles.
Diamine Soft Mint Ink did not coat walls.  Preferred cohesion and formed small islands of liquid on the plastic surface.  It produced NO bubbles.
Noodler’s Polar Brown Ink left a very thin layer coating the wall that persisted for a long time.  The layer highlighted plastic imperfections of the bottle by contrast.  It produced many/large bubbles.  Bubbles appeared to grow over time.


The surface coating indicates that there was interaction with the walls of the container.  Since the plastic was either polyethylene or polypropylene, we expect the walls to be hydrophobic.  The inks which wetted the surface well interacted with the plastic surface.  Since bubbles were formed, we can assume that the interaction with the wall was due to surfactant in the ink reducing the surface energy allowing this prolonged contact.  The lack of coating in the Diamine sample suggests that it is highly hydrophilic.  The lack of bubbles suggests that no surfactant was added to the ink during manufacturing.

2. pH Test


Samples were tested for hydrogen ion concentration using a Vernier pH probe connected to a computer running the logging software.  The probe was calibrated immediately before the experiment with a pH 4.00 buffer standard (Fisher – methyl alcohol, formaldehyde, and potassium hydrogen phthalate) and a pH 10.00 buffer standard (Fisher – disodium EDTA dihydrate, potassium carbonate, potassium borate, potassium hydroxide).  The lab temperature was 25C, and the standard pH did not need adjusted from the label value based on this temperature.


Ink pH
Noodler’s Baystate Blue 3.97
Rohrer & Klingner Scabiosa 2.33*
Noodler’s North African Violet 4.60
Diamine Soft Mint 4.00
Noodler’s Polar Brown 8.78


The majority of the inks were slightly acidic in nature, and the Polar Brown ink was slightly basic.  The iron gall based ink was quite acidic.  This ink pH is marked with an asterisk(*) since the pH was not stable.  The value continually drifted downward, trending toward acidity.  The value used here was the observed pH value after 15 minutes of attempted stabilization.  After the experiments were complete, the pH probe was observed to be slightly discolored.  To determine how this effected the results, the standards were measured with the discolored probe.  The 4.00 buffer measured 3.74, and the 10.00 buffer measured 9.88.  This indicates that the results are probably skewed downward by ~0.25 pH units.  The results were not corrected for this error, since it would be difficult to quantify the error without much experimentation.  The probe was cleaned by soaking in 2% nitric acid for 30 minutes.  After this time, the probe was deemed ‘clean’ and returned to storage.

The results of this experiment should clear up a few misconceptions.  Notably, none of the inks in this less than comprehensive test were pH neutral.  Some sources discussed previously make claims based on the acid or base nature of inks based on brand.  Clearly, this is not a wise assessment.  The brand with three different inks in the test (Noodler’s) showed a wide range of pH values.  Some outside sources suggest that acidic inks may damage pens.  Most of the inks tested were acidic.  No metal based tests were performed in this experiment, but this author believes that the correlation between metal pitting and acid content of an ink may be spurious.

3. Solvent Tests


In this test, a small quantity of ink was deposited onto a strip of printer paper (Staples Brand) and partially submerged in solvent.  The ink was applied to the paper by smearing one or two drops from a disposable plastic pipette on the paper.  All labels on the samples were written in pencil.  The ink was allowed to dry then tested.  Most strips were folded so they would stand and set in a petri dish with the testing solution for five minutes. These tests were broken up into aqueous, organic, and chlorinated solvent groups to prevent unwanted reactions.

The wet papers were then taped to a rail in the hood to dry.

One strip was tested under a 18.4 W long wave UV lamp for one hour.  One strip was left as a control.



The numbers represent the inks Noodler’s Baystate Blue, Rohrer & Klingner Scabiosa, Noodler’s North African Violet, Diamine Soft Mint, and Noodler’s Polar Brown respectively.  The inks were not applied evenly, and some spreading, especially in samples 3-5 was due to large drops.  The Polar Brown ink naturally feathers a great deal on this paper, and the Scabiosa and North African Violet inks showed some feathering.  The Soft Mint ink naturally spread, but did not show feathering.


1 2 3 4 5
minimal feathering feathering feathering spread and loss No Change (NC)


1 2 3 4 5
upward drift some upward spread upward drift upward drift edge reddening


1 2 3 4 5
some spread NC feathering upward drift NC


1 2 3 4 5
upward drift some feathering some feathering/drift some drift NC


1 2 3 4 5
upward drift NC minor drift NC edge reddening

Nitric Acid (~4%)

1 2 3 4 5
gone, yellow edge almost gone, blue edge almost gone, green edge gone NC

Hydrochloric Acid (~4%)

1 2 3 4 5
gone, yellow edge almost gone, blue edge almost gone, green edge gone/upward drift edge reddening


1 2 3 4 5
upward drift feathering upward drift/feathering minor upward drift yellow color separation

UV Exposure

1 2 3 4 5


There is a great deal of information to be learned from these tests.  What we are looking at is actually a very rudimentary form of thin layer chromotography.  The paper acts as a stationary phase and the solvent as a mobile phase.  The mobile phase moves our eluent along the capillaries within the paper fiber matrix.  If we let this go for a long time, we would see the individual components separate from each other.  There are better methods to perform that experiment though, and it was not in the scope of this one.  The goal of the test was to determine the stability of the ink when in contact with the solvents.  In doing this, we can learn something about the character of the ink.

Different solvents have different polarities.  The polarity of a molecule being the concentration of electrons on one side of an atom and exposed protons on the other side.  We can relate these by a polarity index.  Since “like dissolves like” if something is moved or dissolved by more polar solvents, then the material must be polar.  So we can see that the Diamine ink is more polar than the quite nonpolar Baystate Blue based on what different solvents do to it.

The Polar Brown ink interestingly slightly drifts in both polar and nonpolar solvents.  If you look closely (probably not visible in pictures), you can see a yellow component separate out in nonpolar solvents and a red component separate out in polar solvents.  These colors moving out of the bulk drop only slightly affect a color change in the brown.

I chose the two acids to test here on purpose.  We would expect the HCl to react mostly as an acid, whereas the HNO3 would react as both an acid and a strong oxidizer.  Since both of the acids had similar effects on the inks, we can see that none of the inks oxidize easily – which speaks well for the permanence of these inks.

Future tests would include solvents with polarity index of 0.

The lack of change due to UV light is good, though this author would like to point out that this was not a very strong source.  Future tests should attempt to use more powerful sources such as a laser.

A. Errata

All solvents were obtained from Pharmaco Aaper with purity >99.5% except the nitric acid, which was obtained from Fisher.

All water used in the experiments was reagent grade, purified by reverse osmosis to resistance of >18Mohms.

I would like to thank CPTBadass again for his donations and my advisor for being out of town this week so I could do silly experiments like this without getting yelled at.

Thank you for reading, I would be happy to address any questions or comments you may have about this work.

A 3D Modeled Keyboard Cap

After I got hooked by the allure of 3D CAD, I decided to practice my skills by making a 3D modeled keyboard cap.  I started working on something using OpenSCAD, but it proved difficult to get the sort of organic curves I wanted.  I wanted to get some experience using a GUI based-program, rather than just scripting.  Instead, I decided to try a program called Fusion360.  This program is a free variant of Autodesk for personal and low-impact use.  It sadly does not have a Linux build, but I have had luck with it on my Windows box (it has the more powerful gaming GPU’s on it anyway).

The Design

I based my design on the classic caps you would get on retro units, like the Commodore.  A modern variant is available from Signature Plastics (PMK) known as the SA profile.  At the time I originally made this model (April 2015), there weren’t any of these models freely available.  This should be similar to the SA keycap dimensions but not quite the same.

The 3D modeled keyboard cap from the side.

The 3D modeled keyboard cap from the bottom.

The Real World Versions of my 3D Modeled Keyboard Cap

I decided to have these 3D printed from a commercial vendor, rather than trying to print it on a home printer.  I wanted to get some better detail.

The first print.

The height is somewhere between Signature Plastic’s DSA and SA row 3 profiles.  You can compare the shape and size in the picture below of my 3D modeled keyboard cap  to a DSA (left, in black) and an SA (right, red) profile cap:

A comparison of the print to commercial versions.

I’m glad I did this test run.  The stem shrank in a way I didn’t expect, and the cruciform is a bit too big.  It slips on and off the stems too easily.  This is an easily fixable problem.

The bottom side.


I made a new version with a tighter cruciform.  I added some text on this one.  Nothing like a little American Psycho quote to brighten up a keyboard.  I ended up giving this to a colleague as a gift.

Feed me a stray cat.


If you would like to use this file, feel free to do so.  Check it out:

This is a 3D render, play around with it!

And here is a download link.


IBM Model M Mod

I decided that I should do my own IBM Model M mod.  In early March of 2015, I bought a Model M from a Geekhack user on the cheap.

The original IBM Model M I purchased with the cap covers off.

The keyboard was bolt modded, had a wonky membrane, and no cable.  I did a few mods to it.

The Caps

I dyed the caps aqua colored utilizing Jacquard iDye.  I boiled water and used an old metal colander to submerge the caps.  I oriented the caps in the same direction in the bath, and that gave them a nice gradient.

The keyboard with the case removed.

The electronics

Since few new computers come with AT-DIN connectors any more, for this IBM Model M Mod, I wanted to add a USB.  The original connector of this board was a hard-wired DIN connection.  I removed the old connector and added a Teensy for USB capabilities.  Desoldering the wires was easy enough, but I actually went through a few Teensys in the process.  I had a Teensy 2.0, but while soldering it I damaged one of the resistors/capacitors and the socket.  I did some work on it with the Teensy 2.0+, but I am going to use that for another project.  I bought a Teensy LC, but the ARM processor was giving me problems.  So I finally went with another 2.0.

I’m pretty proud of the usb connector.  I took the old DIN socket and ripped it all to hell with my Dremel.  I made little slots so the teensy board would slide in and super glued it to that spot.  I filled in the open area with epoxy clay so it looks fairly clean.

The clean surface of the USB connector.

I also desoldered all of the old ugly green LEDS and resistors on the breakout board.  I put in their place some bright blue LEDs and appropriate resistors for them.  Now everything looks bright and color coordinated.

I'm glad that I went with the bright LEDs. If you aren't blinded by your keyboard numlock, you aren't doing it right.

The Artisan Keys

I have a BS compatible Brobot from [Ctrl]Alt  as the escape key.  This picture is nice because it really shows off the color gradient on the other caps.

Purple Bro

I also have a pair of glorious multi-shot Krap bonus caps on the pause and scroll lock buttons.

One-off Krap caps - art if you like it.

Now I have a very attractive IBM Model M Mod.

The Finished IBM Model M Mod

SEM of Uranium Adsorption Sites

Before my dissertation project got underway, I was working on a project describing the kinetics of adsorption of uranyl ions onto a metal oxide thin film.  The thin film was impregnated onto pellets of silica and alumina, then I would submerge the pellets in a solution of uranium nitrate in water.  After being submerged for a while, the white pellets would come out with intense yellow-green coloration, as seen in the picture below:


pellets, before and after
Before and after uranium adsorption.

There is a lot of interesting stuff happening with the kinetics there as you adjust things like ion concentration, pH, mixing speed, etc.  We think that there is an ion exchange happening at the surface sites, but the kinetics get complicated.  There are clearly some Langmuir adsorption effects going on. But that isn’t what I want to show you today.  During my time doing my PhD, I was fortunate enough to get into a course on microscopy which was offered by the veterinary biology school.  When we weren’t busy looking at mouse blood and sickly tree leaves (those pics will come later), we had an opportunity to bring in our own samples to explore.  I decided I wanted to take a look at one of these pellets.  We mounted a sample onto a platen, coated it with some carbon, and took a gander at it through the scanning electron microscope.


One of these SEM images, it’s mesoporo-tastic!

Now here is the fun part.  This wasn’t your grandma’s SEM.  This one was equipped with an optional x-ray spectrometer for elemental analysis.  To generate an SEM image, we fire electrons at a surface and analyze the backscattered electrons and secondary electrons.  Not all of the electrons that hit the target generate the backscattering we need to ‘see’ a sample. Some of them wiggle and release the energy from the absorbed electron as a photon.  This means that while we are blasting the surface in the picture above we can analyze the photons that come off.  X-ray photons from atoms have particular binding energies which allow us to identify what is present in the sample.  In the sample I’m looking at, I can run a check and make a pretty graph of what is detected:

Elemental Analysis of the Thin Film under the SEM

redacted elements
Even from this, we can tell that the sample has already adsorbed uranium, and the thin film was mounted on alumina instead of silica.  I have redacted one of the atoms, since this work isn’t ready for real publication yet.

The instrument we use to check the elements in the sample come up on the plot as a color.  Since the instrument also saves each pass, I can look at them as overlays on the original SEM image.  I can see what goes where:


This handy-dandy gif shows each layer overlayed on the SEM image. The colors and order matches the peaks on the plot above.

That’s neat and all, but so what?  Well that is best part.  Did you notice how some of the colors in the animation above sort of glomp together?  I did. There is some significance to that grouping.  If we combine the overlays for the calcium and [metal X] onto one image, we see that these pair together nicely. This is because of the thin film we use: CaXOy.  The calcium starts out bonded to the XOy complex.  In the ion exchange mechanism, which is proposed as the primary mechanism for action in this system, the uranium/uranyl comes along and knocks off the calcium and takes over that spot.  When we see [metal X] and calcium together, we know that these are unreacted sites.  If we combine the overlays for both uranium peaks and the oxygen peak, these materials match up.  This suggests something important:

Uranium is adsorbing as the uranyl (UO_4) species, not the bare atom.

Surface Features Influence Adsorption

Now if we compare these two images side by side, we see that there are areas which are distinctly different.

calcium+[metalX] on the left, uranium+oxygen on the right The bottom image is the base without overlays to show features.

In the bright blue spots in the left image, there is little to no oxygen showing up on the right image.  This means that all of the oxygen in our metal oxide thin film as well as the oxygen in the structure of the alumina is sufficiently buried so that we can’t see it on the surface.  This also means that there is something about those locations that makes the uranyl not ‘want’ to bind there.  What is it about those sites?  I’ve included the original image here too so you can compare.  You can actually see that the blank areas have distinct features, even if it isn’t high enough resolution to make out the exact shape of those features.   There is clearly a preference for binding sites in this material.  This has ramifications for the kinetics experiments I was running a on the material.  If we classify sites that uranyl ‘wants’ to bind to as SITE1 and sites that uranyl doesn’t ‘want’ to bind to as SITE2, we can think about it like this:  The kinetics of adsorption would have a single mathematical model to describe how fast it adsorbs at SITE1, and it would likely have a distinct mathematical model to describe SITE2 locations.  But when all of the SITE1 locations run out, and the remaining uranyl is forced to bind at SITE2 locations, is there a smooth transition?  Is there some energy barrier separating these that we would see as a hiccup on the uptake curve?  There are some anomalies on the curves that still need explained.  Could this be the answer?

Unanswered Questions

Unfortunately, I didn’t have the time or funding to answer those questions when I got these data.  I presented my findings on the uranium uptake at two ACS meetings already(cited below), and I needed to move on to bigger projects.  So all I can do for now is to communicate to you, humble reader, how interesting this finding is.  These data may be published in the future when we finalize the uranium adsorption project into a paper, but until then I will leave this up for you to enjoy.


  1. Honeycutt, W. T., Hamby, H., Apblett, A. & Materer, N. F. Uptake kinetics of heavy metals from water using a high surface area supported inorganic metal oxide. in Abstracts of Papers, 247th ACS National Meeting & Exposition, Dallas, TX, United States, March 16-20, 2014 ENVR-272 (American Chemical Society, 2014).
  2. Honeycutt, W. T., Kadossov, E. B., Apblett, A. W. & Materer, N. F. Selectivity and kinetic behavior of heavy metal and radionuclides on supported ion-exchange adsorbant. in Abstracts of Papers, 249th ACS National Meeting & Exposition, Denver, CO, United States, March 22-26, 2015 I+EC-44 (American Chemical Society, 2015).

UHV Electron Beam Fluorescence of UOx

Let’s get this thing started. This is a new blog which will be used to act as a frontpage for my projects. Since I have a few years of interesting stuff worth posting, I’m going to post old things every week. Each of these posts will be categorized as “backlog” and will be dated if possible.

I will start this with a gif I found from one of my old experiments.

This video was taken using my cellphone of a sample inside our UHV-XPS.  What you are looking for is in the center of the image.  There is a dark square which has some glowing spots.  The dark square is uranium oxide powder which is held to the sample holder by an advanced fixative (double-sided scotch tape).  In this image, we are not running the x-ray source.  Instead, we are blasting the surface with electrons to clean it off for the analysis.  In this process, a beam of electrons are shot at the sample, and the energy imparted will remove any adsorbed contaminants one layer at a time.  When we shoot these electrons at uranium oxide, there is a luminescent phenomenon.  The incident electron knocks around the electrons of the molecule and  the molecule wiggles away the energy by ejecting a photon.  This one happens to have a pretty green glow, which unfortunately doesn’t come across in the gif.

Another interesting part of this is the way the glowing portion seems to go from one blob, to two blobs, then back again to one blob, and so on.  This appearance is due to the shutter speed of the camera.  The electron gun is programmed to raster the beam across an area.  Imagine rastering an area similar to mowing a lawn.  You push the mower back and forth in lines to cover the area of the lawn.  Our electron beam is sweeping across the uranium oxide like a high voltage power-washer.  It just so happens that the zig-zag of this sweeping pattern syncs up with the shutter speed of the camera in such a way as to make it appear like there are little dancing blobs.  Pretty neat!