I have some long term projects that will require high speed ball bearing, related to jet engines. This is very slow because I lack a lot of tools. I recently acquired a Proxxon MF70 mill, but I still lack a lathe, though I will get progress in this project too in the next months.

Last week while discussing on twitter with @nixie_guy, we had a look at Youtube videos made by Igor Negoda (Игорь Негода, youtube, instagram), where he is doing exactly what I will need to do (relevant video at the end of the article, to avoid spoilers :p). This guy is awesome, he built a fully working miniature jet engine from bare metal stock (beware, many jet projects on youtube are crappy toys, they only spit flames when fed with compressed air but will not sustain combustion/rotation, which is a joke), and flew it in a model plane, that he also made from scratch.

What is a ball bearing

Do your google homework !

A bearing is a device that can support a rotating shaft, allowing it to spin without (or with less) friction. There are many sorts or bearings, for a jet engine we are interested in deep groove ball bearings, that will support high radial loads (forces directed perpendicular to the rotating axis), and low/moderate axial loads (forces directed along the rotating axis).

A deep groove ball bearing has two cylindrical races with a groove, and a set of balls that run in these grooves. The balls are constantly kept separated by a cage made of steel, brass, or polymer. Without the cage, the bearing would fall apart because there are not enough balls to fill the space (the goal is to reduce friction).

A 608 bearing (very usual, used in all roller blades and scooters and skateboards) has an outer diameter of 22mm, an inner diameter of 8mm, a thickness of 7mm and 7 balls measuring 5/32th of an inch (or 3.96875 mm, yes, it seems that even metric bearings have kept imperial ball sizes).

A smaller usual bearing is the 623, OD 19mm, ID 6mm, thickness 6mm, 7 balls measuring 1/8″ (3.175 mm).

Note: it seems that finding the size and count of balls used in ball bearings is not so easy, here is a good doc (National Precision Bearings) that lists the number and size of balls for a great number of usual ball bearings (and a local mirror)

Ball bearings “became popular” during the recent “fidget spinner” frenzy, but despite this I could not find a lot of interesting information on high performance bearings.

What is a full complement ball bearing?

A full complement bearing is a bearing with no ball cage, and as a consequence has more balls. In fact, the race is full of balls. This has multiple consequences:

How to build one

Tools: small flat screwdriver, various hand tools, some large flat container to make sure you dont loose small balls

The first step is to acquire two similar normal ball bearings (here, 623) and disassemble them:

Remove ball cages

You will notice that the balls in a normal bearing use around half of the race space. This is normal and required for easy manufacturing.

Clean the grease from the races and the balls with some absorbing paper.

Now we will modify the race to allow insertion of more balls.

The trick is to grind a canal on the side of the races to allow more balls to be pushed there, while avoiding too much perturbations to the races.

Measure your balls, then find a small cylindrical dremel corumdum or silicon carbide grinder of a similar or just bigger size. For a 608 bearing, balls are 5/32 of an inch, or 3.96mm. Choose a 4mm grinding tool. Smaller bearings will have balls 1/8 of an inch, or 3.17mm. Choose a 3.5mm tool.

Note: You can change the size of a bigger grinding tool with a diamond tool used to dress grinding wheels. That’s cheap on ebay. Here is a 5mm ball tool reduced to a 4mm rounded cylinder:

Now, we have to grind a cylindrical slot in the outer race, anywhere. I have used my Proxxon mill, but a handheld Dremel tool will work, too. On this image we can see my 3.5mm silicon carbide tool: Don’t grind too much, or the race will be damaged and the balls will not slide perfectly in the groove. This is absolutely critical! I made this mistake and had to discard the deep milled outer race. Also, after grinding, make sure to remove every bit of metallic dust from the races and… from everywhere!

On this image, the front ring is OK, but the back one was too heavily ground, and prevented balls from rolling at high speed. The difference is very subtle.

To help you vizualise, here is a CAD image of an OKAY cut: Balls can still slide in the races.

But this cut is too deep, it reaches the center of the race and will interfere with ball movements. Each time a ball reaches this point, it will not be guided anymore in the main race, and will hit the ground slot.

Here is a bad cut in real life:

Next, we have to mill a similar slot in the inner race (the hub). You can clamp the hub in a screw for easy handling. Again, don’t grind too much or the balls will hit the slot and won’t roll properly. The one seen below is rather deep, still okay.

Now we have to reassemble the bearing. It’s easier said than done, and you will let the balls fall a few times before finding a working method! Holding the two races concentric in your fingers, you have to fit 6 balls, one at a time, preventing the others from falling. Magnets can help, but I did not do it to avoid skicky balls. It can be done with a bit of patience. Try to keep the two ground slots aligned and in the top part of the assembly. Using a small screwdriver, push all 6 balls together opposed to the slots. Now it should be close to impossible to fit more balls, except one in the space formed by the two slots. Put a seventh ball here. The assembly should not fall apart and look like this:

This ball now has to go in the races. But it will not do that easily: if it does, that means you drilled the slots too deep, and the bearing will not run smooth. You can use a light hammer and a cylindrical tool with a small recess to push the ball in its race.

Here is a video showing an insertion: This one is WAY too easy, I was doing this on the race visible above, which has a too deep slot!

Fortunately, this deep slot also allowed me to remove all the balls from the modified bearing, which should not be possible if the assembly is well made.

When I redid this on a ring with a shallower slot, I had to use a real hammer instead of the handle of a screwdriver.

Use the same process to fit as many balls as possible, this means 11 balls in a 608 or 623 bearing. At each new step, make sure the bearing rolls correctly, align the slots, insert a ball, push it in the races with the hammer and tool. Here is an intermediate step:

And voila! A full complement bearings with 11 balls! If you worked well, you can now test the bearing by sliding it on a finger, you will notice that rotation lasts much longer than with a normal bearing. Nice fidget spinner!

If the reassembled bearing does not run smooth, there are several possible causes:

If you plan to use this bearing in a real device, it has to be lubricated. High load prefers grease, high speed prefers oil. The balls will stick and the bearing will not run for a long time when launched with a finger.

Now there are some more details to understand!

High speed rotation tweaks

If you intend to use this bearing for a high speed application, it will not be possible. The milled slot in the rotating race will create an imbalance, that will be source of vibrations. At 200000rpm, this is not acceptable and may destroy the bearing and the device using them.

But there is an easy solution: Mill TWO slots, diagonally opposed. Do that as precisely as possible. I can do it with my proxxon mill and a rotary table. The goal is to remove the same amount of metal on both sides so the center of inertia is exactly on the rotating axis and no vibration can happen.

This is why you can see two slots, not one, on the bearing modified in the previous section.

Igor Negoda milled an opposite slot with a dremel hand tool and found this solution not quite accurate, so he used another solution, both more complex and more radical. He mounted the inner race ring on a mandrel in his lathe, and ground one full side of the race enough to get ALL balls in the race at once, but not enough so they will fall apart, Then, he installs the balls in the outer race in a holder, and presses the inner race with his mill.

Now let’s have a look at the master in his workshop

I did not invent a lot here, I merely interpreted his actions and tried to replicate them with my own means. So here is Igor’s last video (in russian with crappy subtitles) where he presents his own technique: dual dremel slots, and, finally, the lathe method!


Have fun, and post videos and pictures, in comments and in twitter, if you attempt such a build!

Future plans are coming, I have SiN balls in the mail!