Sunday, February 14, 2021

Quick flight to qualify new probe

Today I went out in N291DR for a quick jaunt to verify that the new probe works like the other ones. It does. Basically, out of the box, the results are indistinguishable. Slap in a new probe and things work exactly as expected. This is a Good Thing [tm]!



Saturday, February 13, 2021

Just like that -- a new probe is born!

This one incorporates the latest aerospace science regarding how to plumb hoses around without getting them kinked or snapping off various hose barb fittings. More work to come.

At this point I'm suspecting that, for best air tightness, the probe "ball nose" and the shroud over the on-board barometer should probably be made at 100% infill, or maybe even farmed out to Shapeways or a similar operation.

I'm also thinking carefully about whether we should maintain the polycarbonate tube, which is strong and light and pretty and "open", versus making a 3D printed clamshell body that would be easier to "pack" with components before screwing it shut.






Wednesday, February 10, 2021

N291DR is re-balled!

There has been a bit of a hiatus when we were in between old and new designs for the probe and display, and our test ship N291DR was devoid of an Airball.

One of the issues was electrical interference. We were powering our Raspberry Pi display from a USB adapter plugged into the 12V "cigarette lighter" port in the plane. We discovered that the Waveshare LCD panel in our display for some reason backwashed enough noise into the airplane that it defeated the squelch on our radio!! So we were trying to think of solutions.

Now meanwhile we have a new display being built by PCBWay in China as we speak, but this is not here yet.

What we now have is the display wired into the same breaker as the power to our removable GPS. These are now both non-critical devices sharing a 1 amp breaker. All seems well:

The probe is now mounted as you would expect, GoPro style:



So how did it all work out? So far so good. The new probe, with the bogo-calibration we did a while ago using the car on the highway, actually displays a much more "reasonable" IAS than the old one, and I suspect it's more accurate than the ASI in the plane.

There are two main problems we need to work on:

The first is that the Wi-Fi connection from display to probe takes a long time to be established. This is just a need for wpa_supplicant tweaking on the Raspberry Pi. This is super minor.

The second is more annoying. For some reason, our probe "bottoms out" in its AoA measurement before our plane gets to stall AoA. That's super annoying, and also sad because it could be solved with proper calibration had it not been for COVID. But these are the breaks.

We had more "straightforward" calibration and performance with the old probe geometry that had the static pressure "tube" sticking out of it. If we need to go back to that, we will. The actual sensors and electronics will be exactly the same. We'd make a slightly different 3D printed blue part and stick the same hoses into that.

Wednesday, February 3, 2021

Displays -- good news and bad

In the midst of all this talk about the probe nose and the plumbing thereof, it's worth a note about our displays. You might recall that we've been limping along with a commodity "Waveshare" LCD and a Raspberry Pi. This is bulky, and whaddya know, Waveshare discontinued the model we were using! These are described in this post from November 2019.

I designed a new display board based on Jeremy Cole's design and the latest Raspberry Pi Compute Module 4. From the outside, it looks like this:


It's super small, and I've decided to eschew the knob to shrink the form factor even more. I designed a circuit board for it, which looks kind of like this:


At the top right of the leftmost photo, you can see two 0.1" pitch headers. One is for wired installation (accepts+5VDC in and UART signals) and one is an "expansion" header (provides +3VDC out and breaks out several of the GPIO signals). Both are sized so that a screw terminal block can be used.

Note also the two Hirose DF40C-100DS-0.4V(51) connectors to mount the CM4. We shall return to these....

When assembled, the insides are arranged something like this:
So with optimism -- and a whistle on my lips and a spring in my step -- I ordered some PCBs from PCBWay and some components from the Interwebs, and set myself the task of soldering all this stuff together.

Under the best of circumstances, this work is dastardly difficult. The 40-pin LCD connector and the TFP401 HDMI decoder chip are a pain in the neck to get right. But then, I have unsteady hands.

The Hirose connectors, though -- these are another ball of wax of a different color entirely. The connector pitch is 0.4mm, and instead of regular pins, they are actually made up of a whole stackup of vertical "plates" that suck up solder bridges in between them like crazy.

The results looked horrible. I even bought a cheap USB microscope and 3D printed a stand for it, just so I could examine my boards and prove to myself just how horrible they were:



Finally, I gave up and ordered a full PCB build from PCBWay. The cost, compared to buying my own components, is actually amazingly reasonable. When you add in the cost of my hassle and my loss of faith in myself and humanity at large, the extra cost is cheap indeed. And the folks at PCBWay are so incredibly helpful and nice!

But of course, this turns out to be the Chinese spring festival from February 6 through 18 -- who knew? I wish my PCB manufacturing friends a happy time. But meanwhile, I must wait for my boards. Do stay tuned, though. If they work, they ought to be super cool!

Partial triumph with probe nose

Last night I started a print of the following design, which splits the nose between the threaded and un-threaded portions. This can be printed with zero supports for the whole thing:

This morning the printer had a MINTEMP BED error and had stopped partly through the print. It's possible it just got too cold in my garage! But the general hypothesis is confirmed: This is a good way to get all the parts done properly. The two halves are joined with four #2 self-tapping screws, per our previous practice. If we desire better sealing, we can smear a bit of glue on the mating surfaces prior to joining them.

I still need to add a 140 degree countersink to the threaded holes to better accept the geometry of the Eldon James barb adapters, but that's a detail. It works! Yay!






Tuesday, February 2, 2021

Another probe nose 3D print

Our next design for the probe nose has it in 2 pieces, to avoid messy supports. This is what it looked like, just finished:

The idea is for these parts to be glued together after cleaning up and finishing. The following are the steps involved in cleaning the 10-32 threaded bits. In general, it's hard to do that without damaging the threads. One of the holes had bad threads, which I tried to chase with an AN3 aircraft bolt (which is 10-32 threaded) and this improved things a bit, but still. It may or may not be related to the horrible de-laminated layer indicated on the part, which in turn may be related to our attic maintenance people slamming some huge heavy thing onto the ground just outside the garage where my 3D printing was taking place.

The universe conspires against us at every turn. :)

The take-home lessons from today's episode of melting petroleum products and squooging them into weird shapes are:

  • The overall plan of a 2-piece nose with threaded holes for the 10-32 barb adapters is sound. There are no show stoppers at this point.
  • Next time, keep trying more ideas:
    • One possibility is to print without supports, and design the threaded holes to have "domed" or "conical" tops so there is not a lot of overhang when filling them in.
    • Another idea is to change the cutting plane where the two pieces are "split" -- so that the threaded portion is in one half, and requires no supports, and the un-threaded portion is in another half, and contains just the teeny tiny 1/16" holes and channels.
  • Longer term, the actual nose portion is a good candidate for making in SLA or SLS, perhaps at Shapeways. This probably applies to the "barometer shroud" as well.
  • Having gotten a reasonable thing to work for the probe nose, we should proceed to print another copy that does not have the horrible de-laminated layer, and then move on to another test "kit build" of the probe to verify that all the bits work as intended.











Probe nose 3D printing update

 My current design for the probe nose looks something like this:




I've pretty much decided that 10-32 threaded barb adapters are the way to go. Now on to the 3D printing.

The idea is for the threads to be 3D printed into the part. I'm trying this out using standard hobby FDM printing (with my Prusa and ABS). I'm also experimenting with Prusa's version of Slic3r to see what it can do. Here are the results, including the process of cleaning out the part and trying to get a threaded adapter into it:









Here are a few conclusions:

  1. Believe it or not, 3D printing 10-32 threads using my commodity Prusa actually works.
  2. The Slic3r generated supports "stick" to the part far more than the Ultimaker Cura ones.
  3. The 1/16" holes going through the part were completely plugged up. Could also be a Silc3r versus Cura issue.
  4. There are too many fancy "cavities" in my part that are a pain in the neck to clean.
The next design is going to be in 2 pieces -- the front part resting on the bed, and an "extension". I imagine I will make it so you just glue the parts together permanently after they are made. And I'm going to eliminate the "cutout" for the temperature probe -- there's a cutout on the other side of it that should be adequate.

Monday, February 1, 2021

Hoses for Honeywell TruStability pressure sensors

The Honeywell TruStability brand of pressure sensors is what we have settled on for our probes -- for the time being. Followers of this blog will recall painstaking tests with different brands of sensors and our conclusion that "nobody got fired for choosing Honeywell". It's not a bold or ambitious place to find oneself, but sometimes discretion is the better part of valor.

The Honeywell sensors, for their indisputable quality, have one problem. The barbed connectors they have are nominally sized for a 1/16" ID hose, but the barbed fittings are kind of skinny:

This means that if you put a soft (low Durometer hardness) 1/16" ID hose over them, it will tend to slip off. If you use a hard hose of the same size, it will be very secure. The manufacturer, to their (dubious?) credit, in a technical note entitled Technical Note: Pneumatic Interface Recommendations for TruStability® Board Mount Pressure Sensors, HSC, SSC, TSC, NSC Series, recommend two hoses, with 85 and 95 Shore hardness respectively. The harder one is Frelin-Wade Fre-Thane® 95a-157.

So let's just be clear. A 95 Shore hose is super stiff. Stiff enough to beat a dead horse with. And that, of course, means that routing this hose around the inside of a cramped probe is super difficult. So much so in fact that, in the process of wrenching the hose this way and that, I'm worried I'm going to snap off the delicate barbed connectors on the pressure sensors.

Meanwhile, we in Airball Headquarters are pretty enamored of this other tubing. Tygon® S3™ B-44-3 has a hardness of 66 Shore, and it's a delight to work with. It doesn't kink, doesn't yank on connectors, and can be threaded around as you please.

Now of course you probably would ask: But doesn't our dream tubing also pop off every barb fitting in the universe? Well no. In fact, when mated to this Eldon James adapter or just about any other 1/16" barb fitting in the planet, it seems really quite secure. It's just the Honeywell sensors that have this trouble.

I have tried everything, including using upholstery thread and a surgeon's knot to tie a "suture" around the tubing where it fits into the Honeywell sensor and then securing the knot with cyanoacrylate glue (verdict: works, but not super well). I have tried 3D printed "clamps". I have looked for clamps on the Internet; many of these are too bulbous and would not fit between the pressure sensor and the PCB. It has seemed hopeless.

Finally, my solution as of the moment is a combination. Use a very short length of hard tubing to go from the pressure sensor to an elbow connector. Then from there, use soft tubing to go everywhere else in the probe. Whether this holds up in the field or not I don't know, but at least it seems reasonable for now.


This sort of highlights how in many ways the problems we're solving now are ones of scaling: we are looking for ways to make it easy for folks to reproduce our results, not just to have results of our own.

But in some ways, apart from the visualizations (which to our knowledge are novel), this whole project has been about scaling. Spherical alpha/beta probes have been around since time immemorial. This is not rocket surgery. But making it cheap and accessible is the challenge we have taken on from the beginning.

Stay tuned for more details of our probe builds.

Adding hoses to 3D printed parts

One of the challenges of this project has been how to attach hoses to holes in 3D printed parts.

This might seem like such a mundane problem -- surely we can get something to work. And indeed we did. We got many things to work, including:

  • Gluing hoses directly into the parts
  • Gluing pieces of brass tubing into the parts and pushing the hoses onto these
  • Tapping the parts and screwing on threaded hose barb fittings
But the challenge is this. Given a 3D printed part that came off the machine, can you figure out an easy way to attach a hose that is not dependent on whacky techniques and skills, so the stuff can be sent to people as a user-friendly kit?

Today I'm trying the technique of 3D printing 10-32 internal threads directly into the parts, and then screwing in plastic barbed fittings. This is an example of one of the tiny parts of the probe, with a threaded hole, printed using a 0.4mm nozzle at 0.1mm layers in ABS on my Prusa i3 MK3S:


This is an Eldon James A1032-1-209BN fitting, that goes from a 10-32 UNF thread to a 1/16" hose barb:


And these are the parts screwed together:



You could imagine putting a tiny smear of silicone or glue or whatever to make the seal super tight but, for aerodynamic pressure measurement, a microscopic leak will not change the pressure reading much if at all.