iPad / Procreate

I really can’t complain about the iPad drawing experience. The Apple Pencil Pro is incredibly responsive, and Procreate was very easy to pick up and get started with. I do have a few niggles with Procreate: layer handling, lacking non-destructive filters, and the copy paste process between canvases is insane. An example of this is that when you copy layers between canvases, if you’re not careful they get pasted in reverse order. Procreate does however (with a bit of input filtering) allow even me to draw smooth lines.

We have tried a few different styles, including some posterized images, some logo work, and some simple cartoon cats (which you will see later).

I picked up a textured “paperfeel” screen protector for the iPad. I didn’t notice much difference beyond the dust I trapped under it, but my partner says the pencil feel is much nicer.

Printing

So we have some drawings, but we need to be able to print these out. Handily, last year at some point I picked up a Canon Selphy CP1300 printer, a cute little dye-sublimation printer. It prints lovely high gloss high quality prints on 4x6-inch postcard paper (102×152mm)¹, which has caused us issues when trying to frame the pictures it prints, but for this project is not an issue.

Badge Machine

So my other half has had a “broken” badge machine in her craft room since we moved here… over five years ago. We decided to give it a go, and if we couldn’t get it working we could just throw it out instead of it taking up space. We grabbed a 1⅛ inch scalloped punch (which was a bit too small) and threw some blanks into the BlankBadge machine and it worked. A bit of scrambling to find a better fitting punch², and we were away to the races.

It took a bit of trial and error getting the images to print the right size. The content needs to be just under one inch to fit on the front of the badge, but needs about 1½ inches of total space so the design can be punched out.

Results

So we still need to experiment a bit more, including in getting the print outs the right size, punching them out neatly, and maybe adding simple backgrounds to add variety beyond plain white, but I think for an evening of play the badges have come out ok!

I only had two real constraints on this project:

• Time
• Money

Which turn out to be quite major blockers. Both of which precluded paying an artist to design something for me ¹, and I can’t draw for toffee. The solution… A.I. slop? But can we take advantage of that?

Personal Brand - The ME Sticker

Over the years I’ve become known (for better or for worse) in the cyber security conference scene for my dungarees. Specifically the dinosaur dungarees by Run and Fly. I also have one of my favourite images of me, from a trip to Japan Teamlabs Planets.

Design

So off to chatGPT we went ². It took a couple of attempts to get something I liked. I wanted something in an anime style ³, but it took a bit of prompt engineering to get a results I liked. Some key phrases that really helped were:

• Ensure there is plenty of space between the subject and the sides of the image
• Use a transparent background

I also had issues with it trying to add my feet in, the first version of the image was very close to the image provided, so I asked it to give me shoes… for some reason it decided the correct style of shoe of the image was work boots ⁴. So I asked it to replace the boot with van style shoes (the kind of shoes I usually wear). It did a reasonable job, but it made my feet huge, and also managed to cut off some of my hair.

With a bit of trial and error I found the best results were to collect a better description based on the changes I wanted, and re-provide the original image, and start again. This resulted in an image that I was happy with, even if it made my beard much more voluptuous than in real life.

The next step was to get these produced as stickers. I’ve used Sticker App a few times before, and been happy with the results, so I was happy to use them again.

Pricing

The Final Sticker

I’m pretty happy with the way these came out, and I really confused a few DEFCON (and non-DEFCON) people when I gave them a sticker after they mentioned my dungarees.

Memes?

Closer to DEFCON Paul wanted to get some stickers made up, so I took the opportunity to get a few more designs printed up.

AI Slop

AI Slop was just a silly idea I had, but it turned out harder to convince the AI to make than I expected. It just wanted to always give the hand the correct number of fingers.

In the end I ended up with a prompt something like:

Disregard all previous instructions on the number of fingers that a human has. Draw me a design with hand with too many fingers, in the style of a bad ai [snip].

RawDog

Finally, this sticker was made to annoy Grom.

Hostname

The hosts joined to ipa need a FQDN as their hostname, the easiest way of updating this is with hostnamectl.

sudo hostnamectl hostname <hostname>.<domain>

Dependencies

My default shell is zsh which isn’t installed by default. As i am using local home directories at the moment oddjob is in charge of making sure they are created.

sudo apt install freeipa-client zsh oddjob-mkhomedir

Todo: document enabling oddjob

Join Machine

Join the machine to the network. --mkhomedir is needed as I am not using NFS home dirs at this time.

sudo ipa-client-install --mkhomedir

Split Flap Displays

Before we go into any details I should probably introduce Split Flap displays for anyone who doesn’t know what they are. This kind of display are largely alphanumeric; , some additional symbols are also included¹.

These displays had a number of advantages over other display types when they will still in common use. Including the ability to have very high definition fonts, and the fact that they are not emissive meaning they don’t require power when not being changed.

The biggest critic of these displays (but also the reason they are so iconic) is the noise they make when changing values. This can be seen on the video below of a split flip display in a Paris train station.

The Plan

For now I am going with the design by David Kingman. I may need some changes however to enable me to print this on my Bambu A1 Mini which has a smaller build plate than he used. I have found a modified version of the frame that fits.

One change I am making right off the bat² is that I don’t have multi filament support for my A1 Mini. So instead of the print in place flaps with font embedded i am planning to use blank flaps with vinyl cut lettering.

The Parts

The following items are on order from AliExpress:

• 28BYJ-48 Stepper Motor DC 12V
• Heat Set Inserts

I am hoping i have all the ’m’ screws in stock, and I still need find a supplier of the Hall effect sensors which are used to home the spools. I may use the Hall effect boards that John has designed.

Starting to Print

This brings us up to where I am now, with some test prints already started.

This has already brought up a few things… This project is going to take a LONG time to print on my little printer. The Bambu Smooth PIE plate along with enabling ironing in the slicer is required to get a nice surface finish on the split flaps. On the other hand the split flaps are VERY hard to remove from the smooth plate.

Follow Along

If you want to follow along if before I write the next part of this blog post then I am keeping John Rogers up to date on mastodon.

John @John@chaos.social@mastodon.social

@Mexicals

Great to meet you. Let me know how you get on with the splitflaps.
I do have a few of the buddy boards I could send you, If you wanted!
I can also provide links to what I did if needed!

10:54 PM • November 23, 2024 (UTC)

Compared with last year this is going to be a much simpler design, with the “electronics” being only three components.

The only slightly odd component was the rear firing LED which will allow the eye of the horse to light up. I went with the APTR3216SURCK from Kingbright, based mostly on availiablity.

V1 - The Trouble with Silk Screens

This year I gave myself enough time to hopefully get a couple of revisions before taking the plunge and making the final order.

I am glad I did, as the first version revealed a couple of things I was not aware of when designing the SAO. Firstly, I did not realise that the PCB fab would remove all silk screen over bare copper. This makes sense for most PCBs are silk screen over copper doesn’t really make sense, but for “art” PCBs this is a bit of a weakness². I also realised that large areas of copper end up with a non-consistent surface due to the tinning that is applied.

V2 - Looking Good

Another order of three boards later, and another wait and the new revision arrived. This revision was mostly to fix the silkscreen, but i took the opportunity to tidy the vectors a little bit, and increase the cutout in the copper layer to allow more light to shine through the eye.

Happy with these changes it was time to order the boards!

PCBA

As ever, and in the hope for an easy life I called on JLCPCB’s PCBA capabilities to assemble the badges, and after some stress with JLCPCBs PCBA inspections tools not working for me³, the order was in, and then the waiting began. It wasn’t a long wait…

The Fuck-Up - Datasheets are Hard OK

Not long after going to production I got an email from JLCPCB saying there was an issue. “There is a problem with your design, and we won’t be able to solder the LEDs as is.”

I’d made a very silly assumption about the LEDS. I had assumed the height shown in the second diagram was “above” the PCB, but in fact these LEDs expect a hole in the PCB to be made for the LED to poke into. This would never have worked for my design, and with JLCPCBs manufacturing⁴ capabilities. Thankfully JLCPCB offered a solution (for a reasonable fee), and manufactured the boards with the LED installed “upside-down”, so I could still get the boards and components, but it would be up to me to fix the issue.

After a little waiting the boards arrived. JLCB did the panelisation, and they did a nice job.

And even with the LEDs firing away from the board the eye lights up a little, but it wasn’t good enough so the re-work started.

Re-working these boards required me to:

• De-solder the LEDs
• Partially Drill through every board
• Re-solder the LED the “correct” way up with the lens in the little cavity

Lucky I own a drill press, and some small flat bottomed drills/mills were cheap. Unfortunately my drill press is more designed for metal work than precision depth milling. The depth stop on the drill is very basic, what I found did work was setting the stop about level with the board, and applying some consistent pressure to the handle to achieve the depth required.

A little soldering later and we had working boards, and in the grand scheme of things this was less painful that last time…

Final Result & DEF CON

In the end of the 125 board I ordered, I managed to get about 70 reworked and out with me to DEF CON.

I have plenty left, including ones still need re-working so if you bump into me at a conference just ask there is a good chance I will have a spare on me.

We will get back to the details of the project later in this post, but as a quick spoiler the project was a success.

The Plan

The plan was “simple”; replace the static light/display from the controller with a screen and display some interesting alternatives to the old WAIT message.

Requirements and Stretch Goals

For the project I set myself the following requirements, in order of priority.

• Look (mostly) like an unmodified beg button when not interacted with
• Display random WAIT states based on UK and foreign beg button designs
• Display randomWALK states, based both on existing real displays as above, but also silly alternatives.
• be triggered from the existing button in the case.
• Play sounds
  • Random
  • Something appropriate for the WALK display being shown
• Effects instead of just static images
  • Rain

The Existing Hardware

The existing hardware, a module from Siemens (667/1/01478/001) is incredibly simple, the beg button is basically a very resilient case for a light bulb, a shadow mask and a push button. All the logic is handled via the road side traffic light controller, this part is essentially dumb.

Additional consideration include that the button is designed to be pole mounted, and has opening on the back where its wiring passed through.

The New Hardware

With the old light bulb and painted front “glass” removed, I could now measure everything up and find a suitable screen and computer to drive it. In the end I went with a screen from Waveshare and a pi4³.

• 10.1inch Capacitive Touch Display for Raspberry Pi, 1280×800, IPS, DSI Interface
• pi4 2GB
• gpio hat
• 5V relay module

The screen isn’t a perfect fit but it’s the closest option I could find that would fit and wasn’t crazy expensive. It has a very neat mounting system with the pi4 and is all powered by USBC, dramatically simplify power delivery.

The GPIO hat wasn’t technically needed, but allowed for an easy security way of wiring everything together without the use of custom cables, and it also eased debugging with it’s status LEDS.

This project also led to the purchase of a new 3D printer…

Mounting the Screen

To mount the screen I designed and 3D printed a bezel. This was a great opportunity to learn freeCAD and this was the first model I have designed in the package. The screen does not provide any mounting points, and as such is held in place with tabs I included in the 3D printed bezel.

The Software

For this product I chose to go with the pygame framework to draw the graphics. Getting started with this framework was very quick and I had a very simple working example within a day, loading font and SVGs and bliting the parts together to form the output image. It was however very quickly evident that my prototype based on the tutorials I found was very inefficient, and difficult to expand on ⁴.

I also wanted if time allowed to show effects as well as just static images.

Effects

The initial plan was to use pixel shaders to create the effects however pygame’s OpenGL support is provided primarily through third party libraries and these projects are either not well documented or are unsupported. Additionally the learning curve and time set aside for this project didn’t leave much room to get up to speed with how shader languages work. Finally this is also where the choice of a pi4 bit me a bit as the Pi only supports OpenGL 2.1 and the libraries for implementing pixel shaders I looked into all target OpenGL 3.0.

Rain

The first completed effect is of rain drops falling on the walk area of the display. The rain drops will fall until they hit the bottom of the surface or collide with the icon at which point they randomly respawn at the top of the screen.

import random
import pygame

def _random_drop(width, height):
    return  (
                random.randint(0, width-1),
                (random.randint(0, height // 10)),
            )

class Rain:
    def __init__(self) -> None:
        self._state = []

    def reset(self, surface: pygame.Surface) -> None:
        self._surface = surface
        width = self._surface.get_width()
        height = self._surface.get_height()
        self._state = [
            _random_drop(width, height)
            for _ in range(100)
        ]

    def _raindrop(self, xy):
        pygame.draw.rect(
            surface=self._surface,
            color="blue",
            rect=pygame.Rect(xy[0] - 1, xy[1] - 10, 3, 10),
        )

    def draw(self):
        width = self._surface.get_width()
        height = self._surface.get_height()
        for i in range(len(self._state)) :
            point = self._state[i]
            point_y = point[1]

            if point_y >= height or self._surface.get_at(point) != pygame.Color("BLACK"):
                self._state[i] = _random_drop(width, height)

        self._state = [(xy[0], ((xy[1] + 3))) for xy in self._state]
        

        for drop in self._state:
            self._raindrop(drop)
        return self._state

Sounder

The crossing box did come with the original sounder, this “pips” each time a voltage is applied, so needs to be pulsed to make the expected sound⁵. The unit is rated at 6v to 36v and after a bit of playing the volume looks to be directly related to the voltage applied. Around 12v was found to be a good volume but to do this I would need a 12v rail and some way of switching it as the pi and it’s screen run on 5v.

There are two options for this extra rail as I want to continue to use USB to power the device. Use a USB PD Decoy to get 12v from my USB power supply then down convert this to 5v for the pi. Or continue to run the pi at 5v and use a USB to 12v step up converter. Being unsure I bought components to do either way.

Driving the sounder was implemented though a relay controlled by a Pi GPIO pin which allowed me to easily switch the 12V supply.

The sounder work was somewhat waylaid by me ordering a 12v relay and trying to switch it with 5v… And the sounder being broken and not working reliably at 12v (well within it’s rated voltages) and in the end didn’t make it to EMF.

In the end I didn’t use the sounder as it wasn’t reliable at 12V even when run off my desktop power-supply. I expect if I want to include a sounder I will have to go hunting for a replacement module.

Speaker

My initial plan was to add a speaker to play crossing sounds from around the world however it’s been quite difficult to find high quality recordings of these without large amounts of road noise.

pygame’s audio support made playing a random sound file whilst the crossing walk state is active easy.

In the version of the project that made it to EMF Camp a battery powered portable speaker was glued inside the case and played the sound.

Final Thoughts

Overall I am happy with this a project, people interacted with it and asked questions. Some kids must have pushed the button 50 times. A few people even tooted about the project (OK they were replying to me asking for feedback but it was still nice).

PeterEbenezer @PeterEbenezer@mastodon.social

@Mexicals we had “hike” so went on a wander instead of going to bed 😅

9:58 AM • June 2, 2024 (UTC)

Tom Hodson @Tomhodson@tech.lgbt@mastodon.social

@Mexicals omg I pushed this button, highly recommend! Where did you get the housing?

11:15 AM • June 2, 2024 (UTC)

There were a few issues, I had made a code change just before I left for the field, which cause the app to not start on boot-up, and crazily no one I asked had a spare USB keyboard I was able to borrow. But once a keyboard was found and the crossing joined to the WIFI network is was nice and easy to remote in and fix the issues.

Immediate next steps will be to print a wall mount for the box so I can have it set up more permanently in my shed. Beyond this, I would like to add some extra features such as weather adaptive effects, and include some recording functionally to track the number of times the button is pushed. I would also like to find a more final solution for the speaker/sounder. I might even re-write everything in a more performant engine as an excuse to learn a new tool. I also now have some ebay searches set up for more street furniture… Traffic lights next time maybe?

BarCamps

I have written about BarCamps before¹, but for those new the the idea BarCamps are a style of unconference, linked by the name and ethos more than specifics of style or organisation. Any one can set up a local BarCamp, but they all follow a general style, and a general topic of technology. What’s key to understand however it that there is no call for talks at these events, or people gatekeeping who can talk about what, instead every attendee has the chance to talk about any subject they want.

BarCamp London vs BarCamp Canterbury

As these events were so close together and not to go over the same ground once again, I’m going to take this opportunity to compare and contrast these two great events.

The Grid

The “grid” is a key part of a BarCamp, this is how attendees schedule their talks and is fully self managed. BarCamp Canterbury and London use two very different systems for this. At Barcamp London a physical grid of record cards on a wall is used, as shown in the image bellow.

Whilst at BarCamp Canterbury a google spreadsheet with global read write access is used.

These two systems have their own advantages and disadvantages. The physical grid was very good at getting people talking, and served as a central meeting place, but did lead to people running up and down the corridor between talks. The online system on the other hand allowed efficient interaction with the schedule, and led to the schedule being adapted more over the day, but didn’t have the physical “space” in quite the same way.

The Spaces

The space is one of the most key aspects of events, though through experience I understand ths is often one of the aspects that the organisers have the least control of. As such I am going to make some brief points about the two spaces.

Canterbury

The Canterbury event this year was hosted in the University of Kent Library A block. There were a number of different sized rooms, including a full lecture theatre and seminar rooms. The space was not near full capacity and as such not every room available to the event was used. Two of the seminar rooms suffered from bad sound leakage which made hearing the presenter sometime hard.

Access to the site is easy if your visiting by car, with free parking right on the door. Whilst there are good bus links to the uni, the nearest train stations are in the city center which is a thirty minute walk away.

Wifi was provided through an hotspot without WPA security. Mobile signal within the venue was terrible

London

The London event was run at The London Academy of Excellence right next to the Tottenham Hotspur Stadium. Access from underground / overground was easy with the nearest station a couple of minute walk away. Unfortunately issue with the TFL network at large seem to have put some people off.

The talk rooms were nine similar sized class rooms, with TVs for AV. For the size rooms the AV was find, and whilst the rooms were not the largest there never seem to be too many for each space.

Catering (provided by the venue) went above and beyond what was needed.

Free secure WiFi was provided by the venue.

My Talks

This year I gave three talks across the two events:

• Canterbury
  • Is a SAO your perfect first step in PCB design
  • An intro to the Open Street Map Project (second time of giving this talk)
• London
  • An intro to the Open Street Map Project (third time of giving this talk)

The SAO talk was fine, but not good enough to want to advertise it, the OSM talk has gone down well every time I have given it, but I think it has now run it’s course and should be retired.

Final Thoughts

I am so glad that BarCamps are back, and seem to be back to stay. For anyone who just wants to meet some great people in tech they are amazing, but I also feel they are one of the best ways to practice presentation skills.

First up was updating Proxmox from 7.4 to 8.x, this went pretty smoothly, but would have gone even better if I had read the instruction manual more carefully. There isn’t actually anything in the release I need but it’s best to keep all these things up to date.

Next were the VM/Container host OSs, I run a mixture of Ubuntu and Alpine in my instances, Ubuntu just took a do-dist-upgrade, and Alpine was iterating¹ through the following steps:

apk update
apk upgrade
vim /etc/apk/repositories # replace all instances of 3.n with 3.n+1

ZZH to ZZHp

Next up was to make sure I was on the most recent Home-Assistant build, I use Home-Assistant core, so my upgrades are via pip (moving off core has been on my todo list for a long time).

Now onto the “fun” stuff, upgrading the stick. Proxmox doesn’t support hotplug for USB devices. So the machine was booted down and the container, and added a line to it’s config to mount the second USB serial device, and rebooted the host².

vim /etc/pve/lxc/100.conf

lxc.mount.entry: /dev/ttyUSB0 dev/ttyUSB0 none bind,optional,create=file
lxc.mount.entry: /dev/ttyUSB1 dev/ttyUSB1 none bind,optional,create=file

With both the old and new Zigbee sticks now visible to the VM it was possible to follow the Migrate Radio workflow from the ZHA setting page. I just accepted the defaults and removed the old stick when it was done with and all devices came up on the stick.

At this point I did a full Proxmox reboot, “just in case” and of course the new zigbee stick came up as a different dev node as it was now the only serial device. I could have done something smart here, but I just re-ran the Migrate radio workflow again to tell ZHA the correct dev node to use.

So… Was it all worth it? Everything is still stable, but I didn’t take any before and after signal statistics but the network is still stable.

• Festival Day
• OSHCAMP
• OSHCAMP workshops

Hebden Bridge is a lovely little town, and the venue was great with a onsite bar/cafe and decent AV setup. Moreover it’s great to have a technology event not in London or one of the other big cities for once.

Festival Day

The first day was full of talks all related to “technology”, but the range of talks was very broad, form a project to get your own chips FAB’d in silicon, to technology in Music and poetry.

One stand out talk, mostly because it was much more interesting than the title would lead you to think was on the History and future of the the UK’s telephone exchange buildings. I would recommend checking out Lisa Kinch’s work, and the twitter feed for a stream of Telephone exchange building pictures. Whilst many exchanges were cut and paste copies, some were very interesting architecturally and it would be shame if we lost nay more of these buildings because of the listing processes distaste for industrial buildings.

**Exhibition announcement**

I am really excited to exhibit some of my work to date at BT's London Headquarters.

The exhibition is open to the public 5-7pm on Thursday the 21 September.

Please RSVP via https://t.co/gElnaGycZL if you would like to join as spaces are limited. pic.twitter.com/oNngxTQHuu

— The Telephone Exchange Building (1945-1981) (@tele_exchange) August 23, 2023

There is still a chance to see this exhibit (depending on when your reading this) at 5-7pm on Thursday the 21 September.

OSHCAMP

The second day with the first part of OSHCAMP leveled up the geekery with focus on hardware instead of technology in general. Talks ranged from why you should consider four layer PCBs instead of two for your next board project, to custom guitars with 3D printed bodies.

We also get introduced to this years conference Badge by Omer and Oomlout.

The laser cut additions were a great idea, and it was really cool to see what everyone built and used them for.

OSHCAMP - Workshops

The final day was workshops, these included learning MicroPython, building your own keyboard (which I am sad I missed), and learning how to SMD solder by building up the conference badge. This was teh first time I had used a solder paste stencil and through I would need to buy some extra equipment did reduce the fear factor of trying this out myself some day.

Whilst there I got chatting with Ben about his awesome battle bots and arena, and it wasn’t long before I found myself with a remote in my hand driving a mean little spinner bot with all the power and none of the control. Somehow I managed to tame it well enough though and took first place destroying a bunch of robots, some of the arena, and a few kids hopes and dreams.

Being a maniac behind the controller would have been enough for me, but I also got a few goodies sent my way for winning, which is nice.

If you’re at a UK cyber security conference and see the Infosec BattleBots about I would recommend going to say hello and trying your hand at driving one of the little monsters. You can even bring your own bots (within the rules defined on their site) if you’re feeling brave enough.

The changes

The following changes have been made to the board since the last revision.

The reversed connector

Fixing this was pretty easy, the issue was the “front” of my board in KiCad, the part with the components on, is the back of the board when in use. Pins have been corrected, power should now be good ¹.

The power draw

All the LEDs on their board with their corresponding 100Ω current limiting resistors was pulling 0.26A. A bit of playing with my Digital DC Power Supply² and the LEDs were plenty bright enough at 0.15A so a new resistor value of 180Ω was picked.

MOAR LEDs

Four more LEDs were added around the center of the board to help smooth out the light toward the center of the board. This will bring the power budget back up a bit but I think it’s worth it, and will smooth out the light especially with the new higher value resistors meaning the current LEDs will be a bit dimmer.

Panalisation

For this revision with the help of Omer I had a go at panalising the board. In this case this was primarily to remove the tooling holes, but for other projects panalising can have a significant price benefit, and is the only way with JLCPCB to get more than 30 copies of a board in one order.

Experimenting with diffusers

I was planning to try some different techniques or materials for the diffuser, but the time crunch killed me so I went with the hot snot. I got a technique where I flooded the back with very hot hot-glue and then tapped the badges before it cooled to level it and help pop the bubbles.

This Time’s F*ckups

VCUTS

I panalised this for vcuts, but these are not available in JLPCBs economy option. Thankfully they converted the board to “mouse bites” for me. This won’t be quite the same edge finish but saved re-submitting the board and will be ok after a little hand finishing.

BOM

For this revision and as previously mentioned I intended to increase the value of the current limiting resistors, for the previous version Omer created my BOM file, but I made a quick change and sent this BOM to production…

When making the bom it turns out I can’t read JLCPCB part C284934 which is 62.5mW Thick Film Resistors ±100ppm/℃ ±1% 10kΩ 0402 Chip. It should be noted that 10kΩ is not the 150Ω that I should have selected.

This has led to hand soldering (at the point of writing) over 400 0402 chip resistors³. This was a very stupid and time consuming mistake. The pen marks in the photo above are my way of tracking which resistors I have successfully “fixed”. If I continue with making PCBs I think I will invest in some magnification, but for anyone else who needs to do the same, some decent flux and a nice sharp pair of tweezers are a must.

The Final Badge

You will have to wait until after DEFCON to see the final badge as they are already all packed up ready to go.

And here is the SAO living on CovertSwarms badge. The colour looked smoother in real life, googly eyes are not standard, but should be required.

Slides

Code Harbour meets up once a month currently in Canterbury, if your local and into the geeky lifestyle then you should pop along, entry is free, just book a seat via the Code Harbour website.

The Plan

The following criteria were set:

• Keep it simple
• Keep it cheap

So no clever tricks, just a window in the copper and solder mask layers to act as a light mask, and back-lit LEDs to illuminate the design.

After some searching of available components with the manufacturer I was planning to use (JLCPCB) it became clear that rear firing LEDs were going to be too expensive, but side firing LEDs were much cheaper. So a revised plan using these and “light-pipes” / diffuser was decided on (more on this later).

Schematic SAO3

As this schematic shows, there really isn’t much to this board. Just the connector the LEDs and their corresponding current limiting resistors.

The board

This is what KiCad thinks the boards should look like.

And here is a terrible 3D render of the board… sans texturing. It should rotate if you drag on it, but your mileage might vary. I might come back and fix this view later, but I have already spent too much time on getting the JavaScript for this view to work to not leave it in.

Production

The Bill of Materials (BoM), Gerbers and Centroid (CPL) were sent over to JLCPCB (thanks Omer for help with this step). A few tweaks were made with JLCPCB to correct the component alignment. Then money exchanged hands and the waiting commenced. This step was easier than I expected but having someone show me the ropes really was helpful. I decided to go with fully manufactured boards with JLCPCB doing all the work

The board arrive

Roughly one week later a box from FedEx arrived, and I got to testing the boards.

Here you can see the board all lit up, I am not super happy with the brightness around the centre, but there should be enough space to add four more LEDs here

Here you can see the “light-pipe” I previously mentioned. The technical solution of globs of hot snot.

The F*ck up

So there was one small, tiny issue, with this board. The power pins are reversed. ☹. It’s not the end of the world because I was going to run another spin of this board with the extra central LEDs. Testing has also shown that the LEDs are drawing quite a bit of current so slightly higher value resistors are going to be used in the next version.

I also didn’t plan around the manufacturing holes that JLCPCB require, they are not super obvious, but you can see them in the picture above. So in the next spin I will try and minimise these too.

What’s Next

I might update this post with more details after I finish the re-spin… Or I might forget.

I only went and actually made a second post!

One very handy library I did find was a peg parser that made consuming the text based input files much easier.

peg::parser! {
    grammar list_parser() for str {
        rule number() -> V
        = n:$(['0'..='9']+) {? V::new_value(n.parse().or(Err("u32"))) }

        pub rule list() -> V
        = "[" l:((number()/list()) ** ",") "]" {? V::new_list(Ok(l)) }
    }
}

As expected lifetimes and borrowing took some time to get used to, but after a while borrowing and cloning became second nature (though I am sure I have done some very un-optimal things with memory).

One area I did find very hard was containers such as RC and RefCell. Which I felt the examples and documentation for were not very helpful, and the API somewhat clunky to use.

Bottle Cutting

First step was to cut the bottles, I picked up a glass cutter from amazon (I now own two… but one has been lost to the hoard). Choice was made more on delivery time than anything so there may be better choices out there. The process is pretty simple, score the bottle with the tool then use heat/cold cycles to break the bottle. It took a few attempts to get the processes right, and few failed attempts, but the following worked out best for me:

• Not scoring the bottles too hard
• Heating gently (I used a craft hot air gun)
• Quickly cooling the bottle in ice water

Vinyl Sticker

It was also an opportunity to try out some new vinyl on the Brother Scan&Cut vinyl cutter. After good success with HVTRONT’s heat activated vinyl, I picked up a few roles of their permanent glue backed versions. Unfortunately I have been less impressed with these roles, which tend to become de-laminated whilst unrolling due to packaging.

Planter Kit

Finally it was just a case of collecting the following bits to complete the kits:

• Seeds
• Centrifuge Tubes (for storing seeds in)
• Clay Balls
• Coir
• Wicking cord
• Hydroponic food
• Eye dropper bottles

Everything was placed in the bottles, and then the bottle was tapped close along the cut line. If I was to make more of these I think I would hunt out some better tape than the bright green masking tape I used, but time ran out so Frog tape was used.

There is nothing clever with this project, but I am impressed with the heat transfer vinyl, which came form a brand called HTVRONT. The only issues I had were:

1. Setting the Scan and Cut cut-depth too deep and cutting through the vinyl and it’s backing
2. Not reading the instructions correctly and putting the film in the Scan and Cut upside-down
3. Forgetting to mirror the design

Thankfully all the above were spotted on the first test cut and everything went smoothly after this. The scan and cut software Brother CanvasWorkspace is pretty simple, and designed for hobbyists, but supports SVG import which means I can use a very similar workflow that I use on my laser cutter. The globe used is licensed under creative commons and is from Wikipedia¹.

Update: Now In Use

These are now being used by the nursery. If I remember I will post a further update in a few months to show how well they have held up to use.

This was the first time the even has run in person since lockdown, and whilst the turnout was a little lower than hoped, everyone made great use of event and jumped in with both hands. Allan did a great job organising and running the event.

This year I gave a talk on home automation with Home Assistant, which I have included here for posterity.

So a very quick and easy plan was devised to at least allow us to decorate our village. Due to time constraints things had to be kept simple, and whilst EMF Camp is one of the only festivals where power to the tent is almost a given being battery operated would allow for maximum flexibility. So a quick Aliexpress order later and the following was winging its way through the post.

• 12V LED “neon”
• 12V USB-PD Converter

As an aside these USB-PD boards are great, they do all the USB-PD communication and provide 12V² from any compatible power bank.

A little bit of soldering later, liberal quantities of hot glue and heat shrink for water proofing, and we had working strips.

The Install

Finally lots of clothes pegs to attach the strips to the event shelter and three neon ropes were sorted and installed, and they came out quite well if I do say myself, even if they were very simple³.

These were powered of a mix of USB-PD compatible battery banks, and USB-PD mains adapters. The strips were tested on a desktop power supply before the USB-PD adapters were soldered and found to draw around 2A at 12V, this means we got plenty of hours of life out of my 20,000mAh battery packs.

EMF Camp 2022 - The rest of the Camp

Finally here are a few images from the rest of the festival to show some of the amazing stuff there. Check out the food included with the screening of Fyre Festival, and of course the late night showing of Hackers was once again a massive laugh. Hack the Planet!

These are all shot with a Sony a6600, with a few different lenses. These photos are by no means good examples of Astrophotography, but show how quickly you can get up and running taking deep sky images.

References

• http://deepskystacker.free.fr/english/index.html
• https://siril.org/

A quick tidy-up and adding a tab for a base in Inkscape and it was over the laser cutter for a test.

I sprayed them with matt lacquer to help them survive outside, but I don’t expect them to last that long in the rain.

Out of my A4 sheets of ply I had some offcuts so a bit of scaling and cutting later and a set of fairy-lights, and some “pegs”² I had laying around also led to this.

Whilst the battery box frames made construction of the internals super easy, especially with the excellent KSpot Welder which I borrowed from a friend³, not everything went quite so easily.

Likewise soldering on the BMS which involved connecting the input and output terminals as well as balance leads to each set of parallel cells went well. The soldering was aided by the great ts80 and ts100 soldering irons from mini ware⁴.

The real issues arose when trying to solder on the output connector and close those box. The problem was mostly caused by the very small tolerances inside the box. Closing the box and fitting the assembled battery inside the case was eased by a liberal application of Kapton tape. In the end I resorted to ignoring the provided spade connectors and exposed the power leads via fly wires and an XT60 connector.

Shopping List

Whilst this is not an instructional post for those considering building there own battery I have included the components I used bellow. As previously stated I would not recommend the battery box I chose. Also ensure you understand the risks before taking on such a project, lithium cells need special handling, and are are fire hazard. The assembled (and partially assembled) battery contains a lot of energy and special care should be taken to prevent shorting and shocks.

• Battery Box
• Common Port BMS
• Nickle Strips (These should not be plated nickle)
• KSpot Welder
• Kapton Tape
• Silicone Wire
• 65x18650 Lithium Ion Cells

Thanks

A special thanks goes to Simon and Elliot for all their help and advice in this project, and for making sure I didn’t do anything too stupid.

Addendum

Soldering XT60 connectors can sometimes be hard…

More information about the Panellum library can be found in older post. If the view looks a bit strange it should be noted that camera height was set to around that of a child of the school to give a better idea of how they experience the space.

This was cut and etched on my k40 laser cutter. There is nothing clever here just a Creative Commons licensed photo grabbed of the net and a couple of simple shapes in Inkscape.

If you would like to make your own you can grab the SVG I used. The cut lines were used three times, for the front, back, and middle acrylic layer to write on. Only the front was etched.

Even better thanks to the ac

Cutting this box was a pain because of the very limited bed side of my k40 laser cutter (something I have been meaning to “fix” for a while). The only standard size ply I can readily fit in the cutter is A4, so each panel of the box used it’s own sheet of stock.

Since making this box I have found another box generator boxes.py which I think I will try for my next box project, it has a lot more box options, including segmented boxes.

Hyperlapse

This is a simple 16x speedup hyperlapse with flow stabilisation and blur enabled.

Smallworld 1

This small world centered on the ground at 4x speed.

Smallworld 2

This small world centered on the sky at 4x speed.

The 360 imagery was taken using a Insta360 One X.

Xiaomi WXKG11LM

This is a battery operation single button sensor. Pairing was super easy and quick, usual process of holding the reset button. Pairing distance of about 3m with the zig-a-zig-ah behind a closed wooden door.

I tested with a second identical device, but connected to ZOE, and found it to show a link quality of about 65, compared with closer to 90 under zig-a-zig-ah. Both coordinators are in the same cupboard.

Xiaomi QBKG12LM

This is a dual light switch with neutral support, this device appears as a router. Once again pairing worked first time (well second but first time was user error). This device is further away from the coordinator and upstairs.

With my old coordinator I was seeing link quality of about 100-110, however I had more routers installed in this configuration so I don’t think this is a fair comparison.

Link Quality?

I am seeing comparable signal quality to my old coordinator (zoe), however this may be just because of the difference in antenna positioning and routers added to the network. It should also be noted that Zoe has a range extender which zig-a-zig-ah does not, so it is not apples-for-apples comparison, however these results are better than I expected and both zoe and zig-a-zi-ah have better link quality than the cheap CC2531 board I started with.

So when Omer reached out about his new Zigbee USB stick, zig-a-zig-ah I jumped at the chance to help out, and be able to finally remove the reliance on the Pi, with the added bonus of jumping to Zigbee 3.0.

A few days ago the prototype arrived in the post, note this is not the final design or packaging, but everything arrived safely, and a quick test on my laptop showed everything was working. My release came pre-flashed with Z-Stack-Firmware from @Koennkk, but production devices will not come flashed.

Configuring Proxmox (LXC)

I use LXC containers under Proxmox for my home automation services, so I span up a new Alpine based LXC container and set up pass-through for the ttyUSB0 device that zig-a-zig-ah presents. The LXC ID of this new container on my setup was 103 and I will use this throughout this post.

First we need some details on the ttyUSB device. On my device the stick appeared as ttyUSB0, so from the Proxmox shell of the host (not the LXC container) we need to ls the ttyUSB0 dev node. Take note of the value to the right of the group, in my case this value is 118.

root@pve:~# ls -la /dev/ttyUSB0 
crw-rw---- 1 root dialout 188, 0 Mar 22 17:27 /dev/ttyUSB0

Next we will update the Proxmox LXC configuration at vim /etc/pve/lxc/103.conf, note the 118 value that matches which was found earlier.

lxc.cgroup.devices.allow: c 188:* rwm
lxc.mount.entry: /dev/ttyUSB0 dev/ttyUSB0 none bind,optional,create=file

Configuring Zigbee2MQTT

After this we can boot the LXC container, and install Zigbee2MQTT. The changed to the Zigbee2MQTT configuration I needed were:

serial:
  # Location of zig-a-zig-ah
  port: /dev/ttyUSB0

advanced:
  rtscts: false

Now all I needed was a npm start as in the Zigbee2MQTT, and everything seems to have started up OK. Next step will be rebinding all my devices to this new setup.