Case review

[HuggeK]

Description

I´m creating this issue so we can track progress and a wish-list of features in the case.

My first thoughts:

• Build upon the IARC lessons learned.
• support the new IWR radar - how to mount it.
• Make mounting holes for a tilt sensor, such as discussed in the discord server, using the LIS3DH board.
• Make hole and mounting place for a potential potentiometer to adjust the sound trigger sensitivity. (Maybe switch out the sparkfun board in it´s interity to be able to set this via software, see issue Switch out the sound trigger board #189)
• Maybe explore multiple revisions - e.g. could we make it smaller to be printable on for example an Bambu A1 Mini-segmented?
• Make the case better to not crack that easy(I have gotten it to crack two tims at the same position in line with old kld-7 bottom "slot"
• Make more room for the usb-A adapters/micro-usb cables we are using.
• Document what length these cable needs to be in the case to cut down interference/voltage sag.

Intresting to note is that the rpi foundation launched a new 10" touchscreen, could maybe be intresting if we are going to revisit the case. https://www.raspberrypi.com/news/a-new-10-raspberry-pi-touch-display-2-available-now-at-80/ Seeing that our case still do not fit on a small print bed( e.g. the very common bambu A1 mini, some cm/inches to large to fit on that print bed)

I myself dont have a 3D printer and have only touched CAD software for 2h in my entire life so would grealty appreacate help.

Motivation

I´m creating this issue so we can track progress and a wish-list of features in the case.

Area

Hardware / wiring

Alternatives considered

No response

Contribution

• I'm willing to work on this

[HuggeK]

Also: We should try to avoid using M5 bolts becuase they are not that standard for electronic distrubutors so it would be better if we could find some nice mounting solution which uses just M.2.5 screws or something similair to just needing to buy one product instead of two and go hunting in the hardware store for M5 bolts and screws. As some other user noted in the discord, we could use those inserts that you heat up and push into your 3D print and you get a thread to screw into -but that raises complexity that not everyone can handle. e.g.: And having more metal is not good for the radar I would guess so nylon inserts then is the way:

https://www.mouser.se/en/c/electromechanical/hardware/mounting-hardware/?product=Inserts&thread%20size=M3&type=Threaded%20Insert
''

[HuggeK]

Measured some and threw claude on it:

Enclosure design notes: mounting, brackets, power, UPS and branding

Covers four related things that interact: fastener spec, radar bracket
geometry, planning for the optional UPS HAT + power connectors, and how the
logo is applied.
Assumes the current IWR6843ISK config, not the deprecated 2× K-LD7 setup.

---

1. Fastener spec (nylon)

All-nylon is preferable here: metal fasteners near the radar apertures perturb
the antenna pattern, and nylon can't short a pad if a standoff shifts.

Thread size per board

Board	Thread	Source
Raspberry Pi 5	M2.5	2.7 mm holes — M3 does not fit, do not drill out
Touch Display 2	M2.5	Raspberry Pi assembly docs
Adafruit LIS3DH (2809)	M2.5	two 2.5 mm holes
OPS243-A	M3	confirmed, 4 holes
IWR6843ISK	M3	TI SWRU546E — module mates with four M3 screws
SEN-14262	M3	confirmed, 2× ~3.3 mm holes (SparkFun 0.13" standard)

Quantities

M2.5 — 12 screws, 6 standoffs, 6 mm

Item	Holes	Screws	Standoffs
Pi 5 → bracket	4	8	4
LIS3DH	2	4	2

(Pi → display uses the 4× M2.5 screws supplied in the display box.)

M3 — 28 screws, 10 standoffs, 6 mm and 8 mm

Item	Holes	Screws	Standoffs
IWR6843ISK	4	8	4
OPS243-A	4	8	4
SEN-14262	2	4	2
Backplate → heat-set inserts	4	8	—

6 mm for board-to-standoff (~4 mm engagement through a 1.6 mm PCB).
8 mm only where passing through a printed wall into a heat-set insert.

Note: 3.3 mm holes against an M3 shank leave ~0.3 mm of float, so a board can
shift slightly before the screw clamps. No board should rely on hole position
for angular reference — use a locating feature instead. This matters
particularly for the SEN-14262, which has only two holes and can therefore
pivot about that axis: locate it against a printed shoulder or pocket rather
than trusting the screws alone.

What to buy

One M2.5 and one M3 nylon assortment kit (180–300 pc, ~€8–12 each) covers
everything with spares. Distributor route, filtering Material → Nylon
(Polyamide), Gender → Hex Female-Female:

• M2.5: https://www.mouser.com/c/electromechanical/hardware/standoffs-spacers/?thread+size=M2.5
• M3: https://www.mouser.com/en/c/electromechanical/hardware/standoffs-spacers/?thread+size=M3&type=Hex

Separately: ~10× M3 brass heat-set inserts (must be brass even where the screw
is nylon), plus a few M3×12 mm nylon screws in case the plate prints thicker
than planned.

Exception — IWR6843ISK thermal

TI labels the ISK a hot surface, and in the carrier configuration the four M3
screws are specified partly for improving thermal performance — they're a
heat path, not just fastening. All-nylon on all four corners means no
conduction out of the board, which matters in a sealed PETG enclosure.
Suggested compromise: nylon at the two corners nearest the antenna aperture,
metal at the two furthest, bonded to a panel if the case has one.

---

2. Radar brackets — should both radars be at the same angle?

They don't need the same angle, but they must be rigidly coupled at a
known relative angle. That's a stronger requirement than it sounds.

Why identical isn't the goal

• OPS243-A measures radial velocity, so it suffers cosine error:
  measured = true × cos θ. Narrow 20° × 24° beam, so aiming also affects
  whether the ball is detected at all.
• IWR6843 measures angle of arrival. Best near boresight, but ±60° in its
  fine axis makes it far more tolerant of aiming.

The cosine error is bigger than expected indoors

Outdoors, as the ball climbs and flies away, line-of-sight converges toward the
launch angle and the error shrinks. Indoors we only see the first 2–3 m, where
line-of-sight from a floor-level radar is roughly horizontal — so θ ≈ the
launch angle itself:

Launch angle	Uncorrected speed error
10° (driver)	−1.5%
15°	−3.4%
25° (mid iron)	−9.4%
30° (wedge)	−13.4%

Against the ±0.5% ball speed spec that's not a rounding error — it's a
systematic underread that grows up the bag, and since carry scales with ball
speed, it would show as wedges gapping far too short.

Implication for the bracket

If software corrects OPS243 speed using IWR launch angle, both must share one
coordinate frame. Any uncertainty in the angle between the boards feeds
straight into ball speed and therefore carry. Two independently adjustable
brackets make that offset a field measurement — error-prone, drifts if knocked,
and silently wrong rather than obviously broken.

How the LIS3DH fits in

The LIS3DH measures the gravity vector, so it reports the tilt of the body it's
bolted to — i.e. the assembly's orientation in the world frame. That is exactly
the number we want, but it only resolves the full geometry if the two radars
are fixed relative to each other:

• One fixed bracket: the relative angle between the radars is a CAD
  constant, and the LIS3DH's single measurement pins down the absolute
  orientation of both. Two unknowns collapse to one, and the remaining one is
  the one the sensor can actually see.
• Independent brackets: the LIS3DH measures whichever bracket it's bolted
  to and tells you nothing about the other. Half the geometry stays
  uncalibrated.

This is also why the LIS3DH must sit on the same bracket as the radars — a
reading taken from a flexy enclosure wall describes the wall's tilt, not the
radars'.

Proposal

• Single printed bracket carrying both radars at fixed, designed angles, so
  the relative orientation is a CAD constant, not a field measurement.
• Mount the LIS3DH on that same bracket, short standoffs, no grommets or
  compliant mounts in that path.
• Tilt the whole assembly up ~12–15° by default — centres the OPS243 in the
  launch-angle range and keeps uncorrected cosine error under ~2% for driver
  through mid-iron.
• Adjustability applies to the assembly (one DOF, one number the LIS3DH can
  verify), not per board. If per-board adjustment is genuinely wanted, use
  discrete printed detents rather than a friction joint.

IWR6843ISK orientation is now a real failure mode

The array is asymmetric: ±60° azimuth at 15° resolution vs ±15° elevation at
58°. Four symmetric holes let the board be mounted 90° out, which silently
degrades launch angle rather than failing. Suggest a keyed feature or alignment
rib in the bracket. (The K-LD7 pair each measured one plane, so this is new.)

SEN-14262 — the opposite requirement

Where the LIS3DH wants rigid coupling, the sound detector wants the reverse:
the electret capsule picks up structure-borne vibration as readily as airborne
sound, so it should be isolated from the chassis on foam or grommets.

---

3. Power, power button, and the optional UPS HAT

The UPS removes the 5 V/5 A problem entirely

The Geekworm X1209 takes 6–18 V DC via a DC5521 barrel jack (or USB-C) and
outputs 5.1 V / 6 A to the Pi. With it fitted, any generic 12 V brick powers
the build — no hard-to-source 5 V/5 A supply, no e-marked cable, and 6 A of
headroom rather than the Pi's 5 A. Charge current 2.1 A.

Minor correction to the framing: the Pi 5's 5.1 V/5 A is an optional USB-PD
fixed profile rather than proprietary on previous versions of usb PD which is not that common — third-party 27 W PD supplies with
5.1 V/5 A exist. But it's rare enough that the complaint stands: most PD
chargers give 5 V/3 A, the Pi then clamps total USB current to 600 mA, and with
two radars that's a real constraint.

Proposal: swappable flat backplate, not two case variants

Make the rear face a separate flat printed plate bolted on with 4× M3 into
heat-set inserts. Ship plate variants:

• Plate A — no UPS: oval aperture exposing the Pi's USB-C + momentary button
• Plate B — UPS: DC5521 panel jack cutout + momentary button
• Plate C — blank: for people routing power their own way

Same body, same internal mounts, one small part to reprint. Lets the community
contribute plate variants without touching the main model, and generalises to
future schemes (PoE etc.).

Keep the backplate flat, not curved. A flat plate prints face-down on the
bed with no supports, gives clean first-layer surfaces around the connector
cutouts, and is easy for anyone to remix in 2D. A curved rear face would need
supports or an awkward orientation, and would put layer lines across the
connector apertures where they're weakest. This is also the main practical
argument for the separate-plate approach over integrating connectors into the
shell.

Aperture sizing

Pi USB-C oval: size for the overmold, not the plug. The official Pi PSU
has a chunky moulded boot; ~14 × 9 mm is a bare minimum — go larger and add a
lead-in chamfer. Under-sizing this is the classic printed-case mistake.

Power button: momentary switch on GPIO3 (pin 5) + GND with
dtoverlay=gpio-shutdown. Gives soft shutdown and wake-from-halt, works with
or without the UPS, needs no extra hardware. Worth doing regardless — pulling
power mid-session is how SD cards die.

The X1209 handles this differently: a pogo pin against the Pi 5's onboard
button. So the HAT sits at exact standard GPIO spacing with nothing between
the boards, and the GPIO3 button stays the portable option for both configs.

Space budget — the part that will hurt

The X1209 takes 2× 21700 cells (also accepts 18650/4680, any 3.7 V nominal
/4.2 V full type). 21700s are 21 mm diameter, so:

Pi 5 with connectors     ~16 mm
+ HAT standoffs          ~15–20 mm
+ HAT PCB + cells        ~23 mm
------------------------------------
~55–60 mm before case walls

Geekworm note the H505/H509 coolers don't fit the X1207/X1208/X1209 due to
height — so the cooling solution and the UPS interact and should be resolved
together.

Layout constraint: 21700 cells are metal cylinders. Keep them behind the
radar boards, never in the beam path or near the apertures.

Two safety points for the docs, not just the CAD

1. Cold charging. Li-ion must not be charged below 0 °C — lithium plating,
   permanent capacity loss, dendrite/short risk. A launch monitor in an
   unheated garage in a Nordic winter will see sub-zero. If the UPS path is
   supported, the docs need a temperature warning, and ideally the software
   should read temperature and warn before charging.
2. Ventilation. Sealed PETG + Pi 5 + charging cells is a bad thermal
   combination. Venting must not be blocked by the battery pack.

Cells should stay user-supplied and out of any kit — Li-ion shipping rules make
that not worth anyone's trouble.

---

4. Branding — model the logo into the case, not as a separate part

Recommendation: emboss or deboss the logo directly into the shell geometry.
Drop the separate glue-on badge.

A separate logo part is the least reliable thing in the whole print. It's small
and thin, so it has minimal bed adhesion and warps or pops off mid-print; it
has a poor area-to-perimeter ratio, so fine features come out mushy; and it
then depends on glue that has to survive a garage temperature cycle next to a
warm Pi. It's an extra print, an extra assembly step, and an extra failure mode
for something purely cosmetic.

Modelled into the shell it costs nothing — no extra print time worth measuring,
no adhesive, nothing to fall off, and it survives being knocked.

Practical guidance

• Deboss (recessed) is more forgiving than emboss. Recessed detail is cut
  into solid material and doesn't rely on thin raised walls that can be knocked
  or print poorly.
• Depth: 0.6–0.8 mm — about 3–4 layers at 0.2 mm. Deeper adds nothing
  visually and can thin the wall.
• Minimum feature width: ≥ 0.8 mm (two 0.4 mm extrusion widths). Anything
  thinner will not resolve reliably. Test the logo at final size before
  committing — most logos need simplifying at 30–40 mm wide.
• Placement: on a vertical or top face, never on the bed-facing surface
  (elephant's foot smears fine detail) and never on an overhang.
• Avoid tiny islands — dots on an i, counters inside letters. Connect or
  remove them.
• Optional: a debossed logo can take a colour swap or a filament change at
  a set layer for anyone who wants two-tone, without changing the geometry.

The same reasoning applies to any text on the case (version markings, port
labels): model it in, don't add parts.

---

Open questions

1. Does the parts list specify the Adafruit LIS3DH (M2.5) or the SparkFun one
   (M3)? Not mechanically interchangeable.
2. Which cooler is the current case model assuming? That decision and the UPS
   height budget have to be made together.
3. Does the software already apply cosine correction to OPS243 speed using IWR
   launch angle? If not, that's arguably higher priority than any of the above.

[HuggeK]

Also: We would need to document that the build would need a raspberry pi active cooler too.

[jtemple967]

I would be happy to participate in the discussion about the case. I do have a 3D printer at home and have printed the enclosure although in my experience I had to make multiple modifications to make the case really usable:

• Ensured the holes for the M5/M2.5 bolts are all threaded to accept the bolts. This makes the case significantly easier to assemble
• Thickened the walls of the monitor housing. The case as designed causes the monitor housing to separate since on my printer it prints as a single thin layer
• Remove the post standoffs for the monitor bracket in the monitor housing and replace with a corner bracket. The standoff posts were too flimsy and would break easily
• Changed the back plate on the monitor housing to fit the new size of the housing. Also added holes so the plate can be attached to the back of the monitor housing instead of using the additional standoffs as supplied.
• Change the opening in the monitor housing as a hole, not a section removed. The section removed causes the housing to twist easily.
• Filled the hole where the logo plate goes. I had issues printing face down (the largest flat surface) with that small gap. Then the logo is just glued to the front of the sensor housing and it does stick out a bit.
• With the plates on the back of the unit, there was no place to pass through a USB connector for power. I added 2 holes on the side of the sensor housing and installed panel mount jacks for the ethernet, usb C and usb A connections (USB A so you can use a full size keyboard).
• The slot for the sound sensor was too thin and caused the sensor to come out easily. I pushed the sound sensor a bit further back and down then added 2 holes to accept M2.5 screws to hold the sensor in place with the 2 screw holes found on the sensor.
• I bought the K-KD7 radars before they were depreciated so we built ours with those. I redesigned the holders for the K-LD7 to be a 2 plate design where the radars are sandwiched between the two plates. The holders that came with the case were very thin and I broke them quickly while trying to assemble. The new design also does not put stress or flex on the radars.
• Not case related, but for the USB serial adapters that I bought they would not fit between the USB port and the monitor housing. I ordered 2 180 degree adapters to allow the USB serial adapters to fit in the monitor housing.

I would be happy to help with a design for the new radar but I probably won't get one for a while.

[HuggeK]

Great! @jtemple967 I bought the KLD-7 but haven´t mounted it so I think I will order the new one to build on that one too - but I will have the KLD-7 to test out with too. I saw that you what I assume is a keystone jack for the RJ45 jack on the side? I think we could easier get away with just putting a keystone cutout and then just use an off the shelf keystone jack you can wire up yourself to get a female RJ45 jack to the outer case!

[jtemple967]

@HuggeK yes, I put 2 jacks on the side, one for USB A and the other is a dual jack for Ethernet/USB C and both have a cover although the USB A cover doesn't work well. The Ethernet/USB C jack has a spring loaded cover that should do a nice job of keeping dust out of the connection. It would be nice to use a standard part for these and have the sensor housing just be a standard print so I'm all for that.

One other comment on the case...I'd like to propose putting a piece of thin plexiglass over the radars/sensor to protect the unit a bit better. I've read that the radar's functionality would not be compromised with a cover but I have not tested that out. Probably put a few holes where the sensor's microphone is so to make sure it picks up sound properly.

[jtemple967]

@HuggeK I would also be happy to model/print a mount for the new radar and send that to you to verify fit.

[HuggeK]

Plexiglas mounting in the case would be awesome too! I got a micro-ubs cable https://www.mouser.se/en/ProductDetail/Raspberry-Pi/SC0732?qs=T%252BzbugeAwjgDO%252BS%252B%252Bwhihw%3D%3D for the #101 (comment) and it wont fit either so having more space behind the usb-ports I think will help people too.

[HuggeK]

@jtemple967 Thank you for the offer, I think I will hit you up on that if you can send it all the way to Sweden(Write on discord and we can take it from there in a DM)
Could you post your files here in the thread so more people can start iterating on the design with the improvements?

[jtemple967]

@HuggeK I've been looking for the Discord channel on this site for a while and can't find it. Can you share please?

I will also upload my files here.

[HuggeK]

@HuggeK I've been looking for the Discord channel on this site for a while and can't find it. Can you share please?

I will also upload my files here.

Your so right: I created a PR adding it to this repo and not just colemans linktree on ig: #196

[jtemple967]

@HuggeK thanks! I might be one of the few out there that don't have IG...or TikTok...or X...

[JedS]

@HuggeK
Great idea!
I’ll add an entry about choosing a screen. The model Coleman originally recommended isn’t necessarily available in all regions, so I ended up going with the official Raspberry Pi screen (Touch Display 2). It would be nice to be able to choose the model on the future case.

[HarjotDhanota]

if it's not too much trouble, can we also add the option for these types of screen with the protruding mounting points? because a lot of them use this type of system. I can create a .step file for you after taking measurements from mine
https://a.co/d/02GhxZCA