The wafer in the PMMA mask.

So, while waiting for the turbopump to arrive, my mind was left meandering around, and thinking about the PMMA masking, and the difficulty to obtain very low molecular weight masking liquid…

And I thought…If Silicon is sort of transparent to 10600nm laser light…but PMMA is not…can’t I just use laser etching to patern low resolution features on my test wafers?

With that thought, I prepared a test vector file with lines separated 100µm – 200µm – 300µm – etc… for the laser cutter:

Then proceeded to engrave some acrylic I had around, and McGyvered the shitty microscope we had at home:

That is nice!

That looks very consistent to me! Also, linewidth seems pretty constant (no micro power fluctuations in the laser supply)

For scale, a 110µm copper wire was placed in one trench:

The process idea is as follows:

First etch the PMMA with the laser, but not trying to go all the way, just to the top of the silicon wafer. Then, using PMMA solvant, you eat away some thickness from the leftover PMMA, revealing the silicon on the bottom of the trenches.
After that, an anneal step and you can etch the wafer. This should enable 100µm features, with 100µm spacing.


You can’t cheat physics…

…and neither can I.

So, where have I been? Up to no good, apparently.

Up until now, I have skipped altogether the way I was going to pattern the chips. The thing is, you can either have not super expensive paterning tools and expensive resists, or expensive tools and cheap resists.
Obtaining the expensive resists is, well…expensive, plus they are very especialized and sensitive. Also, the optical setup needed is not an easy task either. While looking into it, I found out that acrylic (PMMA) can be used as resist, however, it requires e-beam paterning, wich in turn means you need an electron microscope, wich is not cheap.

Some countries have kind of a market for secondhand electron microscopes, however mine doesn’t. The thing is, if you don’t actually need to scan the thing, but just to shoot electrons at it, could you use any other electron gun?

In theory, you could. It is not a question of voltage (this paper talks about very low voltage PMMA paterning) but to get the assembly to the adequate vacuum conditions. It is not enough to have a good mechanical vacuum pump.

In case you don’t believe me, here’s what a CRT tube does in 40 microns of Argon vacuum:

Here’s the setup:

Having proven that I needed a High vacuum pump, I started working on an oil diffusion pump.

However, the pieces I practiced on where 0,5mm stainless, wich proven trycky, but doable. The final pieces where only 0,25mm thick, wich, unfortunately, altough possible in short runs, it did pierce the sheet on start, wich left the pieces unusable. Shame, they where looking gorgeous:

Now that I had to start again and spend more money on it, a consensus was reached between me and an expert (to be mentioned when, if he, permits it) to try to go for a turbomolecular pump, wich are quicker to reach and stop vacuum, clean in their operation and they look like miniature turbines, wich, to be frank, turns me on. XD


This Boring Company II


Finally got to install the temperature probes. I began by drilling/reaming a vertical hole from the heat face, so even if it was crooked, the start point was correct.

After that, a bigger hole was drilled on the other side to accept a ceramic separator that came with the probe:

To make space for the probe dome and cables, a sharp scrapper was used:

And it looks like this:

The ceramic retainer is attached to the metal plate with screws, with their head on the inside, so they can’t work loose. Also, dual nuts where used:

Now looks like a partial eclipse:

In the meantime, the plasma cutter arrived, and made my life much easier by allowing me to cut simple shapes with zero effort. So, a 3D model of what I wanted was made:

That was transferred to CAD, and then lasercut some MDF stencils to gide the plasma torch:

Other project stencils, but you get the point.

And panels where made.
Plasma cutting by hand (with stencils) is tricky for internal shapes, and thus, all holes had to be tweaked with a file for the modules to fit.

I am slowly getting better at welding (and thick sheet metal helps):

After power sanding:

First test fit alongside the oven:

The control box is upside down. xD

Nuts where welded to hold the cover:

And tabs where added to hold the box away enough from the oven side:

And then the integration began:

Power cabling:

A quick note on fiberglass covered power cable. You MUST crimp the ends before cutting the fiberglass, or it will unravel and create isolation issues. I did mine with some brass tube and parallel pliers:

Since I was not in the mood to make more holes in the box, I opted for the easy cable routing:

And no, the heatsinks can’t get hot enough to damage the fiberglass cable or the silicone cable, altough the second one might get sleeved up to the probes at some point. Let’s say it is workshop friendly as long as there are no kids around. XD

When you want a compact unit, things like this happen:

Testing the power section to ensure everything is ok:

I left it cooling overnight and then installed the timer. It is NOT connected to the power section, it’s just a convenient place to have a programable clock/alarm.

The controller also has a double circuit switch to deactivate the SSR’s, so the oven power is OFF but I can still monitor the temperature.
The oven now just requires some external quartz tube supports, but it can be used as is for now.

See ya!

The hunt for Pump October.

If not for the help of Niklas Fauth the race for the turbopump wouldn’t have happened. (all turbo maintenance and repair performed by him too)

As the building of the diffusion pump came to a screetching halt because my TIG welder can’t weld 0,25mm stainless steel, a brainstorming session was held, to decide if further continue with the diffusion pump, or employ the money and time to try to obtain a turbopump.

This model was found on ebay, bid on it and obtained at an amazing price. The ad mentioned it was removed working, and had a return guarantee of 30 days, so we had nothing to loose.

Upon arrival tough, The serial comunications controller was fried, so the driver had to be hacked, altough all monitoring of the pump was lost.

Furthermore, it did not reach full speed, so a dissasemble was in order:

This pump had obviously had a crash. The blades where dented and:

That amount of gunk should not be in a turbo…

After cleaning tough, the pump worked as well as a dented turbo can do (wich should suffice for my needs).

And this, kids, is how  got a turbopump.

To pick or not to pick, that is the tweezer.

How do you handle wafer pieces to submerge them in HF acid and other cleaning baths?

You use tweezers, but HF will eat away any metal ones, and I would not trust much else other than HDPE plastic that can withstand it. However, I did not find accesible (cheap) ones, so I decided to make them, it’s funnier and cheaper. Too thick stock coupled to a short endmill marred the edges a bit, but it proves the point, once I get thinner HDPE sheet, I’ll make some more.

It picks the wafer square by the edges, allowing it to rest in the ledge, making for a very secure hold, no need for double sided tape or other feeble methods.

See ya!


This Boring Company I

Finally! The oven build begins.

All materials have been gathered:

  • 6x Alumina bricks. (4 needed, 2 spares/testing)
  • 2x SSR 25A+ 2x ssr radiator.
  • 2x High temperature K probes.
  • 2x Temperature controllers.
  • 2x Porcelain connectors.
  • 2x Kanthal heating filament 2000W (60cm compressed)
  • Quartz tube (45mm OD x 450mm)
  • Diamond hole saws (6.5 / 48 / 61mm)
  • 3D printed drill guides.
  • High temp electric cable.
  • Lengths of 15mm square stainless steel tube, screw rod and some washers.

First, a holey test:


If one works slowly, the cores remain intact and you can remove them without having to remove the drill from it’s mount:

Next: Big drills.

A 48mm drill digs out half of the Quartz tube channel, and a 61mm one cuts the channel to allow the heater to serpentine.

I’ll have to figure a way to align the block so the cuts are straight before actually building the thing, tough. XD

Go, go, go!

First, find a way to hold the bricks in place. Justsome press fit wood blocks will do.

This is probably NOT the worst tool in the world. Positioning is within ±0.2mm and angular alignment is good enough. Since this is a single use only (so to speak) tool, I wasn’t going to bother doing the mounting perfectly:


In retrospective, I would make the pieces in such a way that they would fully reach to the base (and also have align markings) instead of this complex centering/mounting technique.

The complete tool (with short test brick):

Testing the drill-through capability before commiting with the good bricks. Everything looks nice.

The bricks are longer than the reach of the filament saw, so they have to be turned around and drilled from both sides. How good is alignment?

Good enough, I’d say:

Video of the process (not full length drilling because of reasons):

Once that is done, vacuum it good and insert the filaments (pre stretched so there are no shortcircuit between loops)

This looks so nice. ^^

Now. How should this be put together?


I really, REALLY, need a welding table. I have zero holding devices, and the aluminium plate I use, warps with heat.

I made some “trays” out of stainless, allowing for a small gap between them and the bricks, so I could use a thin ceramic matt (4mm, compressible)

This way, I cushion everything, reducing the likelyhood of breaks.

Some length of stainless screw get soldered to the top and bottom:

And some washers hold everything together:

Looking good:

Next, heater connection. Seats are cut into the body.


And into the back plate:

This will make a secure connection AND hide everything.

You don’t want to clamp the canthal wire directly as a single filament. Instead, you have to bend it a few times to decrease resistance in the clamping point and allow a better connection over time.

It is also time to test the quartz glass tube fitting:


One more beauty shot:

Now, add some legs and cables:

I’m using duall controllers, one for the top filament and one for the bottom filament, because I’m a control freak, mostly.

I could not resist to test it before finishing the electronics body. First, one hour @300ºC

Then 20 minutes at 600ºC

The filament glows nice at this point, and you can very well see the heat pulses:

Let it cool down, and then, ramp it up to 1000ºC

Closeup of the inferno:

Make sure you don’t have the probes swapped, it can lead to weird temperature readings/setups (one coil heating the other to keep temperature), don’t ask me how I know.

And that’s it!

All the important bits and pieces have been done, encasing the electronics is trivial enough not to deserve a continuation of this post.

Part II


Have nothing better to do? Watch a sputtering rig from start to finish:

It has time annotations so you can skip the uneventful bits.

Plasma starting is quirky:

  • With too low pressure, plasma won’t start.
  • With too high pressure, arcing will happen.
  • With too low voltage (below 300V), plasma won’t hold in the trap as pressure drops.
  • With too high voltage, but not low enough pressure yet, plasma will arc, then fade and won’t restart.

A current meter must be added to complete the graph, and see what is this thing doing. Next week, probably.

For now I ca n get away with hand-eye coordination.