brainwagon

A couple of years ago, I decided to try to get into the entire 3D printing “revolution”. I liked the idea of being able to do some actual design of custom parts, and then be able to render them in plastic. I first bought an Anet A8 kit, and then later a Creality CR-10 which served as more of my workhorse.

But to be honest, the reliability of this printer was never above 80% in terms of my being able to design or download an STL, dice it with Cura, and then print it at reasonable quality without error.

The most common failure mode for prints was simply failure to get the first layer down evenly with sufficient adhesion to keep the print in place until completion. For PLA, the best means seemed to be to use a piece of mirror as the build plate, clean it every time I do a print with isopropyl alchohol, and then carefully preheat the plate, and do a good job of bed leveling.

The stock machine doesn’t include any kind of automatic bed leveling. You simply adjust the z limit switches and the corners of the build plate to try to get it even and calibrated, and then have at it.

That actually isn’t that difficult, but it’s kind of a pain, so I invested in some money to get an EZABL kit. This uses a capacitive proximity sensor to sense the bed position. You can then reflash the firmware of the printer to recognize the printer, and have it autoprobe the bed position before each print.

All that works great. In theory.

But over the last couple of months, I started to encounter a new failure mode that I hadn’t seen before. The proximity probe basically works to act like the previous Z limit switch: when the sensor gets to a certain position, it closes a little transistor circuit to ground the probe pins which used to be connected to a microswitch. When the sensor does this, it opens a red led (there is a similar red led on the sensor itself) to let you know that the switch is closed.

That all seems to work perfectly. But the new failure mode was that the firmware running on the board didn’t seem to see that switch closure, and it would run the stepper into the build plate repeatedly. The Z axis drive belt would then slip making an annoying sound which would continue forever, or at least longer than I could stand before switching the printer off and resetting.

Irritating.

It did it a few times, then mysteriously it stopped. I was making prints related to my telescope refurb project, and then suddenly it was back. Annoying. I did some brief testing. The sensor was detecting the position just fine, but it didn’t seem to register as a triggered event.

I finally got fed up, and sat back and thought. “If the cable connecting the sensor to the Z limit switch had a fault, that would be bad. How can I test it?”

Well, you need to disassemble the printer, pull the cable and verify its operation. So I disconnected the printer from its position in the corner of my dining room, carefully cleared my garage workbench so I would have plenty of work, and moved it there for disassembly.

It’s kind of annoying to disassemble the control box. The motherboard is mounted on the underside of the metal control box, with a 12V generic power supply module in the way. I had done it before, but I had forgotten how annoying it was. I eventually got all the parts pulled out, and extracted the cable in question. I set my multimeter into continuity check mode and…

The cable seems fine. Frown. Maybe it was just loose? I could take the half hour or so to reassemble it (annoying, since it feels like the kind of job that could just take a few minutes) but there would be no guarantee that I wouldn’t immediately have to take it apart again. Should I just spend $12 or so and order a new cable pack just to be sure?

While I had the motherboard out, I decided I should at least take the time to reflash the firmware (there were a couple of changes that I was thinking of making anyway). The motherboard is based upon the same sort of Arduino chipset that everything seems to use, but for some reason (ROM space?) it doesn’t include a boot loader, so needs to be reflashed via a six pin port. It would be easier to do this while the printer was apart.

But then I noticed something bad. The mini USB socket mounted on the board was loose. While examining it… it simply tore loose.

So… I could probably fix this. I could get a new socket, mount it in place with a little CA glue, find a smaller tip for my Hakko soldering station, and replace it. And in fact, I just went ahead and ordered some cheap USB connectors to do that.

But I wonder whether I shouldn’t just chuck this old 8 bit motherboard in the bin and move to something more advanced.

So I took what was a very small and limited project, and expanded it into something that will probably take me a couple of weeks to complete. I ordered a BIGTREETECH SKR 1.4 motherboard. https://www.amazon.com/gp/product/B082YTZJS2/ref=ppx_yo_dt_b_asin_title_o02_s00?ie=UTF8&psc=1 This has a number of interesting added features compared to the previous motherboard. It is based around a 120Mhz 32 ARM processor instead of the 20Mhz 8 bit Atmel chip that runs the old motherboard. It uses the stepper motor modules that are common, rather than having the steppers built into the original model. It can support multiple extruders and dual Z axis motors (nice expansions). It can also talk to the new stepper modules via serial connections, enabling you to set motor currents via software rather than tuning them with screwdrivers. And it also supports limit detection without limit switches: it detects the rise in current as the motor hits the end of motion, and uses that to trigger the limit events.

Oh, and I can compile the firmware and install it just by putting it on an flash card and rebooting the machine. Nice.

But it this does represent a significant set of new challenges as well:

• It’s complicated. Lots of new stuff to learn. New software options to reconfigure. New technologies (like no limit switch and serial motor drivers).
• It’s not actually mechanically compatible with my existing setup. If I were to try to put it back in the same box, the USB and microsd slots wouldn’t line up. That means I’m committing to building a custom case for it.
• It doesn’t actually address the sensor issue that originally started me on this journey in the first place.

But that’s the path I’m on. I think that I’m going to get some 1/4″ Baltic Birch ply and use some of my box building skills that I’ve been practicing and make a simple enclosure to hold the parts, but with plenty of room to make them all easily accessible. I think I’ll also include space for the Raspberry Pi 4 that I use to run Octoprint. I may make this the same size as the base of the printer, so I can actually stack the printer on top and reduce the overall footprint.

I’m trying to look at this as another skill building exercise, which is rather the point of my projects anyway.

If anybody has embarked upon a similar path, drop me a note (twitter is a good place @brainwagon) and let me know if you want to talk me through (or out of?) this line of endeavor.

Hope you all are well.

Part of my recent interest (obsession) with getting my shop equipped for wood working is that I want to return to an activity that I used to do a lot more of: telescope making. In particular, my six and eight inch Dobsonian telescopes that have been relegated to the garage, collecting dust could use some work to make their mount and tube assemblies as nice as the optics. The six inch was the first telescope that I ever did, and I began with my father back when I was only 10 years old. The eight inch was the first telescope that I completed back when I had first begun attending the Chabot telescope maker’s workshop in the early 1990s.

Currently I’m in the “analysis paralysis” mode. This mode is characterized by me sitting with my laptop and some of my reference books, and looking at what others have done and what I could do, and try to plan out an idea about what the design should look like and what it’s capabilities should do.

Here is where my thinking carries me…

I actually lack the patience to really do full on astrophotography, but would like to dabble, and at least be able to do video astronomy. There, the goal isn’t so much to be able to find or track dim fuzzies for long exposures, but merely to keep the telescope smoothly aimed at one of the brighter planets, sun or the moon.

This is actually an easier task than building a full on “GOTO” system. People have been building what are known as “equatorial tracking platforms” for years, and they seem practical and within my skill and tool set. It can be fabricated out of high quality plywood, some strip aluminum and skate bearings, and is not particularly challenging to design. Another good feature of this design is that it is basically a platform which can be added to a conventional Dobsonian design, which means that I can proceed with revamping my old telescope with the assurance that at some later point, I can embark upon this phase with some confidence that it will be useful.

I’m going to start collecting links and information for this project, so that other interested parties can follow along. I’m not certain that it will become a full on “how-to” guide, but I am going to document what my thinking is and log resources that I used along the way, which might be helpful to others.

You would think that with the shelter-in-place orders that we are all under, that I would have had a lot of time and energy to get back to blogging again. I mean for the last eight weeks, I’ve been having to find ways to entertain myself, and in the past blogging has been part of what I would entertain myself. And yet for the better part of the last eight weeks, I’ve basically blogged about nothing.

Which isn’t to say I’ve been doing nothing. Close to nothing, but not actually nothing. My last post was about experiments with AREDN, the Amateur Radio Emergency Digital Network. I’ve still been playing with that, and have got a new Raspberry Pi 4 to add to that network. I’m working toward creating a portable node that will be entirely solar powered. But that’s not today’s project.

I’ve been continuing to tinker with tools. Even in quarantine, I managed to find some tools for sale via Facebook Marketplace, and picked up four rusty planes that I spent a couple of evenings cleaning, de-rusting and sharpening. I have been working on converting a ratty 4×8 sheet of plywood into a rolling stand I can put my Work Sharp sharpening station on.

And, beginning last night, I started working on making a box. At this point, I’d normally insert a picture of the work in progress, but I doubt it would impress any of you. But for now, it will be a box about 12″ long x 5″ x 5″, and is in the style of Japanese toolboxes. If you just search for “Japanese toolbox” on Google, you’ll find hundreds of examples, many quite elaborate and fancy, some very basic. One of the more basic examples with instructions can be found from Make Magazine #34.

https://makezine.com/projects/make-34/japanese-toolbox/

And yet, the very first comment included this:

Sigh.

It really makes me sad (and faintly insecure) to read comments like this. Because I’ve probably spent eight or ten hours in the last week doing nothng but thinking about boxes. Not just boxes, but this particular style of box.

In a sense, this is just about one of the simplest forms of boxes you can make. There is no fancy joinery like dovetails or finger joints. They are not traditionally made from expensive hardwoods, since they were first of all a box to safely stash their saws and chisels. They were not even typically glued, and were instead held together with nails. If you want to see the basic idea, this video from “lemongrasspicker” on YouTube is a pretty good intro:

If you get to around 4:45 in the video, you’ll see him attach the feet (really just two battens) by nailing them on and then simply bending them over to clinch the bottom to the feet. This is actually also traditional, even though it probably offends our modern sensibilities.

You’ll also note that he manages to get a split in the board as he’s resawing it. Rather than chucking that piece of wood onto the scrap pile and trying again, he simply glues it back together and moves on. One of the things about this design is that it’s not supposed to generate a lot of ridiculous waste. I find that pleasant as well.

Woodworking is actually all about problem solving. What is the order of operations? How am I going to cut this accurately? How can I fix my mistake? How can I hide a mistake I can’t fix? How can I draw inspiration from other designs? How does my choice of material (or just this board) affect the overall design?

Someone who sarcastically says “you built a box, dude” has never built a box, or even given any thought to what it means to be build a box.

So, I had planed and cut some pieces of pine that I had that were around 5″ wide and 27″ long down to roughly 5/8″. I decided to make the overall box 12″ long. It probably won’t be for tools, but might be used to store some small items. I’m mostly doing it just to continue practicing my skills. Last night I glued the two sides to the bottom. Unlike the video above where the sides sit atop the bottom and are nailed in from the bottom. I attached the sides to the side of the bottom and attached it with glue. I also bored several small holes and pinned the bottom in place using bamboo skewers.

I’ve been pondering the rest of the project. The lid on these boxes are moderately clever, and can even be made to lock in place. I’m debating whether I should give that a try. I might make a simple lid, and then later try to do the beveling operations to make the locking wedge.

But in any case, boxes are interesting. In fact, a lot of woodworking is just different techniques for making boxes of different sorts. To dismiss it as trivial likely means that your understanding of the tools and techniques are only the most superficial.

I’ll get some photos of it when its a bit further along.

So, I’ve had a minimal AREDN network running at my house for the past few week or so, and it is working well. To recap where I stand today, I have three different pieces of RF hardware:

• a Mikrotik hAP ac lite, which looks like a small router or network switch. This is widely described as a “Swiss army knife”, since it is incredibly versatile. It can itself act as an AREDN node on either 2.4Ghz or 5.8Ghz, but it is even more useful when it is used in combination with a separate RF node. It has four LAN and one WAN port, and works really well as a bridge between the conventional internet and the AREDN network. It can also operate as a conventional wireless access point. Right now in fact, I am connected to both the AREDN network and the Internet through my Mikrotik . Works very well.
• a TP-LINK CPE-210, which is a node which is designed to be placed outdoors. The CPE-210 operates on the 2.4Ghz band, and includes an antenna with a 65 degree output. It is designed to be powered over the Ethernet cable using a power injector which comes with the unit. In theory, the Mikrotik can also serve power to POE devices, but I haven’t tested that yet. But I do have the CPE210 plugged into the Mikrotik device. They don’t communicate with each other over wireless, they use DTD (device to device) communication over the Ethernet cable.
• a GL.iNet GLUSB150, which is my “satellite” RF node. It’s currently in my garage, attached to an Ubuntu Linux box. It uses the AREDN mesh to talk to the TP-LINK (over a stunning 30 feet or so distance). The Linux box sees the GLUSB150 as an ethernet, and requests an IP address from the GLUSB150. Then, you can easily access services like HTTP and FTP from the attached machine.

Pretty nifty.

Over the last few days, I have begun to think about what it would take to turn this setup into a portable network node. The basic requirements that I had would be:

• Make the node solar powered. My experiments with radio powered beacons last year mean that I have a 25w and a 100w solar panel which I could use for the project, as well as solar charge controllers and a 7.2Ah SLA battery (not very big, but affordable).
• Include a small Raspberry-Pi like computer to provide additional network services.

I actually was set on using a Raspberry Pi, but as I dug around my box of development boards, I encountered a Beaglebone Black that I had not used in quite some time (when I booted it, I discovered that the Linux it was running bore a date of around 2013). I decided to revamp my knowledge of it, which could be a whole different story. In the end, I did something somewhat out of the normal mainstream: I loaded it with FreeBSD 12.1. I used to be an avid fan of FreeBSD, and found that one benefit was that FreeBSD had a much less bloated userland, and a minimal (but standard) installation didn’t include a ton of things I didn’t need. I used the binary pkg system to load just the few additional bits I wanted, and I was good to go.

I ended up attaching the Beaglebone Black to my Mikrotik as well, mimicking what I think the final portable node would need. And it works great. I created a very basic webpage which was served using the “one line Python webserver” command python3 -m http.server 8081 which puts a webserver on port 8081. I also enabled a simple FTP daemon (not sure why, but hey).

So now I’ve begun to think about how I want to package up the equipment into a tidy and reasonably portable setup. I created a small stand that would make it easier to place and aim the CPE210, and then started to wonder about what voltages I would need and what the power requirements would be.

The easiest thing would be if I could power everything from the 12V battery directly, but of course life is seldom that straightforward. To prevent cable losses, POE devices are typically powered by higher voltages, 24 or even 48 volts.

The Mikrotik hAP is actually pretty forgiving. It’s input jack can apparently take anything between 11 and 57 volts. It will also provide this voltage onto one of the Ethernet ports to power an additional POE device.

The TP-LINK CPE210 is a bit fussier, requiring power to come over the POE cable, and be between 16V and 27V. This means that we can’t just tap a 12V battery to power it. But we probably could simply use a boost regulator like this one from Banggood (don’t have one handy, but pondering it), use it to provide the input to the DC jack of the Mikrotik, and as they say “Bob’s your uncle.”

The Beaglebone Black specifies a good 5V, 2A power supply, although says that a 1.2A supply (6W) power supply is supposed to be adequate if you don’t have power hungry “hats” (add on boards) or USB peripherals, which I don’t.

I dug out a little Kill A Watt P3 power meter that I had lying around, and plugged in the three different AC power plugs into a power strip, and monitored the power overnight. In 14 hours, it drew about .11 KWh, which maps down to about 7.85 W, which matches the 8W that it also listed as its instantaneous power reading here.

This was somewhat better than I expected, considering losses from the various AC power plugs as well. But if we assume that value is roughly correct, it means a few things:

• In sunny conditions, using my much smaller and more portable 25W solar panel would be entirely reasonable for portable operations. The 100W solar panel would likely be reasonable in all but the cloudiest conditions.
• The 7.2Ah cell I have should easily provide 8 hours or more of operating time.

I think the next steps are:

• source an appropriate boost converter and manufacture or buy pigtails to get it hooked up on battery power.
• I have some small current sensors (INA219) that I used previously in my beacon project. They speak the I2C protocol, which the Beaglebone supports. It should be straightforward to use them to do voltage and current measurements. Ideally, we’d monitor both the current and voltage from the solar panel, as well as measure the input and output power to the boost converter (to measure efficiency).
• Start figuring out the physical layout to make a simple, easy to deploy portable node.

It’s fun to tinker with.

So, I have a spare computer out in my shop which I have used to host various small bits of networking. Previously I used it to host reports from my solar powered WSPR beacon. I thought it might be nice to make it available on the AREDN network as well.

This machine uses “conceptron” as its hostname, and because I didn’t have a convenient way to get a cable pulled out there, it connects to my home wireless network via a small TP-LINK USB wireless dongle. I decided to plug the dongle that I was using to provide a link to my AREDN nodes (the GL.iNet GLUSB150) into that node to also make it available on the via the AREDN network. I plugged it in, configured the network, and was easily able to access it via the AREDN network.

But I was no longer able to access the node via the conventional Internet.

The problem boiled down to a simple one: the GLUSB-150 looks like an ethernet device. When its plugged in, it is automatically recognized by my Linux box, and it configures a default path through it. Also, because it is a wired connection, it has a higher metric value than the wireless connection. This means that by default, traffic goes out over the AREDN network, rather than the conventional network.

There is a program that you can use to adjust the metric value a given interface. I lowered the new “Ethernet” interface, and things began to behave as expected.

I still have to figure out how to make sure this happens by default upon reboots.

Addendum: Actually, it doesn’t appear that this works flawlessly. It appears to work from within my home network, but not for links outside. I’ll have to think about it some more.

Okay, minor lesson learned about AREDN. I had tried creating a couple of additional services and reserved a couple of DHCP devices on my node K6HX-GLUSB150, but then decided to move the hardware to a new location and wire it to different hardware. So I did, and then tried to delete those services. On the the local node, the old services that I deleted were gone, but they continued to be listed on the pages of different node as accessible.

Apparently the issue is that while the node itself deletes the reference, there isn’t apparently a means by which the services push out deletion events to other nodes on the network. The only way that these cache entries are deleted is if the node itself disappears from the mesh network for a period of ten minutes. So, I disconnected the node from power, let the cache expire, and voila, it’s all back to normal.

Strange, but simple enough.

I didn’t have a lot of time to do further experiments with my tiny AREDN network, but I did figure out a few things.

To recap: I currently have three devices which have been flashed with AREDN firmware: a Microtik hAP ac lite (which serves as the indoor hub and wireless access point), a TP-LINK CPE-210 (which serves as a node which I can mount outdoors) and a GL.iNet GLUSB150 (a USB dongle).

I currently have the Mikrotik hardwired via Port #1 to my Google Wifi home network. It will acquire an IP address automatically on my home network. The TP-LINK is attached to the Mikrotik via Port #5. Communications between the TP-LINk and the Mikrotik go over this Device To Device (DTD) link, and not via Wifi. I’ve also configured the Mikrotik to create a wireless access point (with SSID K6HX-AREDN) that any Wifi Enabled device can use to access the AREDN network. One slightly unexpected (but welcome) benefit is that any device that is attached to this access point can not only access the AREDN network, but conventional Internet as well. In fact, I’m typing this post as I’m connected to my Mikrotik, and can access both my nodes and the regular Internet. Cool and convenient!

Freed from the need to use the GLUSB-150 to access the AREDN network from my laptop, I decided to move that dongle out to my garage where I maintain a small server. The GLUSB-150 communicates on the AREDN network by talking to the TP-LINK via RF, which then talks over Ethernet to the Mikrotik, which then tunnels through the conventional Internet to other RF nodes in the San Francisco Wireless Emergency Mesh. Neat!

Previously my garage server was serving up a small webpage that monitored the low power WSPR beacon that I setup. I took that beacon down when the weather turned bad last fall, but it should be possible to make that data available via the AREDN network as well. I experimented a bit with creating new services on my server machine (conceptron) that was linked off K6HX-GLUSB150, but had some difficulty and deleted them. Oddly, some dangling references to them (and legion, my laptop) seems to linger on the node. Not sure what’s going on there.

Oh well, I’ll learn more as I go on. Stay tuned.

For anyone reading this from some more distant point in the future, I’m in my fourth week of self-isolation to help halt the spread of the coronavirus. I’m exceptionally lucky: I’ve continued my job from home, working surprisingly effectively with a reasonably fast network connection as we try to put our next movie release in the can.

I haven’t been in a store or a restaurant in about a month. My wife has ventured out with mask and gloves to do some shopping on a couple of occasions, and we’ve gone out for brief walks, but the cabin fever is setting in a bit. I woke up at five o’clock this morning and couldn’t get back to sleep, so it looks like my work day will start early.

At least I have a short commute now, although my consumption of audio books is way down as a result.

Anyway, I’m looking for projects, and one that I selected was to experiment with something that I learned about at Pacficon, the Bay Area ham radio conference that I attended last October.

That something is AREDN, the Amateur Radio Emergency Data Network.

AREDN is a high speed digital wireless mesh network that uses inexpensive hardware to link sites into a dynamically reconfigurable mesh network capable of carrying video and audio data, providing chat and messaging, and generally just any other services you might support and provide on the conventional Internet, but using a completely separate and non-commercial network.

Bob Ross, N7RBP, goes over the basic setup of what it takes.

Anyway, I thought it might be fun to try to tinker together something like this. Last October it seemed like there wasn’t a ton of activity in my area, but since then the SFWEM or San Francisco Wireless Emergency Mesh seems to have gained some momentum, and there is considerable activity on both sides of the bay. Going to their website, you can see a map of activity in the Bay Area:

The purple nodes are operating on the 2.4Ghz band, whereas the orange ones are operating on the 5.8 Ghz band. (My own node is currently operating on the 2.4Ghz band, and is the node in the upper right of this map, but I am getting ahead of myself.)

From isolation, I decided to spend a few bucks and get some hardware. I chose 2.4Ghz hardware, for now other reason than it was slightly cheaper. For an external node, I decided to go with a TP-LINK CPE-210 which was about $40 shipped for free from Amazon. It’s basically a router, but with a more waterproof case that is meant to be mounted outdoors. It also has a POE (Power Over Ethernet) injector, which means that to power the node you only need to run a single Ethernet cable to the node. It carries (24v?) power along same cable that your data is carried, which makes installation simpler. I have not yet acquired outdoor-grade ethernet cable, but that’s on my list.

To support the AREDN network, you need new firmware which you can get from their website. My CPE210 turned out to be version 2.8, so I uploaded the v2 firmware from the AREDN site. For that device, it’s straightforward: just use the conventional interface to load new firmware, and you are good to go. The node will act like a conventional access point: you plug your laptop (or better, a switch) into the Ethernet that goes to the node, and it will act as a DHCP server and assign you a network address via Ethernet. You can then access the node via a web interface. Here’s what the mesh status page looks like from that node:

In the column on the left are the list of accessible nodes with their associated services. On the right is a list of the nodes that my node can see directly. There are two nodes you can see, both owned by me. The first is K6HX-GLUSB150, which is actually a GL.iNet microrouter which is a small USB gadget that costs about $30 and plugs into my laptop. It also is flashed to run the AREDN firmware, and looks to my Linux laptop like an ordinary USB Ethernet device. It is currently communicating via wireless to my TP-LINK device.

You can also see K6HX-ELSOBRANTE-HAP, which is a Mikrotik hAP ac lite device. which currently costs about $45 from Amazon. It is also running the AREDN firmware. It’s an interesting device, often called the “Swiss Army knife” of AREDN networking, because it can serve as a bridge between conventional Internet and the AREDN network. In particular, it is one of the simpler ways to provide an Internet “tunnel” to link AREDN networks which don’t have a good RF path to the network.

Like me.

My house is actually in a bit of a valley, without good RF paths to most repeaters. I used the Splat! RF prediction software to see just how bad it is. Assuming I have the ability to place an antenna 50 feet up (doubtful) this is my coverage area. If you’ll compare this to the node map I posted earlier, you can see it’s not got a ton of overlap with existing nodes.

Luckily, Chris, KJ6WEG agreed to give me a tunnel connection that allows my small (three node) network to connect via the Internet to the broader network. This is “cheating” in some sense, but given the current rather poor prospects for a strong link elsewhere, it at least allows me to gain experience.

Anyway, my current setup looks more or less like this (from the AREDN beginner’s guide):

In addition to providing a way to tunnel to the Internet, the Mikrotik device can actually create a wireless access point that other devices (such as cell phones) can use to access the network as well. I’ll eventually get a Raspberry Pi configured to hang off it and provide some network services (like webcam, file and HTTP service).

Anyway, I’m having fun from isolation. There is a possibility that the fledgling Pixar amateur radio club may be interested in doing a project (the K6PXR repeater is located in Emeryville, and should have good connectivity across the bay into SF). There is a ton of stuff I still don’t know, and didn’t write up here. Hopefully I’ll learn and write some of it down.

Stay safe out there.

I’ve got an opportunity to potentially loft a small camera module on a high altitude balloon launch next month. I had written some very simple code for an ESP32-CAM module which can be had for about $7 direct from China, or 2 for $10 from Amazon with Prime. The other day I decided to see how long I could expect the battery to last. I grabbed four fresh Alkalines from the Dollar Tree, and measured the voltage for the 4 pack as 6.45V.

I then did something really dumb: I hooked the voltage up to VCC and GND on the module. I should have checked the schematic. The appropriate way to hook it to 5V and GND. Net result, one module no longer responds. I might be able to repair it. Stay tuned for that.

But I then put it on the dashboard of my car and let it sit for the entire workday. After recording something like 1450 images (about 8 hours) the battery voltage had dropped to 5.784V (or about 1.44V per cell) which suggested that they actually had no problem at all. This suggests to me that we probably don’t really need to go to any kind of exotic cell: even at low temperatures, even cheap alkalines will have enough oomph to power the thing for the 4 hours or so that we estimate will fly. I haven’t done the math comparing their relative weights which is another metric we might use in evaluating them.

But in any case, here’s the video, mostly just showing clouds zipping by for about eight hours.

A few other things I learned:

The camera has a fairly narrow field of view. Similar modules are available with fisheye lenses: I ordered one via Amazon, as well as a couple more ESP32-CAM modules to replace the one I killed. I also learned that the AA battery holders I have do not really secure the batteries very well. In the real launch, I should wrap some tape around the modules to keep the batteries from dislodging in flight.

I’m also not sure that the auto-exposure stuff is working entirely correctly. I’ll have to read the datasheet for the camera module a little bit more carefully.

More work in progress.

Okay, I had some scrap 1/4″ plywood in the shop, so I decided to go ahead and see if I could make a T-Trak module. I went ahead and ordered some additional Kato Unitrack via Amazon Prime, as well as some really cheap budget scale trees, and set to work. I set the dimensions for the top to be 308mm, and cut the depth to be 14″ which appears to be the standard for single modules. I then cut some 2×2 into 2″ length to serve as corner blocks. I cut the panels to be 2 1/2″ wide, and while I got the front and back sides to be exactly right, the other sides were off in a way that made me believe that the module isn’t entirely square. This made me grumble a lot, and brought out some Wood Bondo to patch some really crufty joints. I haven’t sanded them down, but here’s the basic idea with a 3D printed house, one of my new cheap trees, and two lines of Kato Unitrack roughly placed in the right place.

Voila!

My woodworking skills are still pretty sketchy to be honest. I still have difficulty getting things cut accurately and square. It was especially important to get the width right: the modules are supposed to be 308mm wide. This is important, because the measurement is equal to the length of a 186mm combined with a 124mm makes 310mm. The modules are sized at 308mm with to provide a little tolerance (a little over 1/32nd of an inch). I’m concerned that my modules are just a little bit too large. I cut two additional tops before I ran out of 1/4″ ply, which I am going to double check against sometime soon. Since I am doing this all by myself, without any other modules made by other people to check against, I am not sure in practice what the tolerances will be.

I should note that the way I assembled this was just with clamps and some 1x2x3 blocks. I didn’t even have my brad nailer, which I loaned out to someone. This made assembly more error prone than I would like, with lots of clamping.

I then noticed that a couple of companies and eBay sellers had laser cut module kits for prices that total up to around $30 for a single module when you include shipping. This made me think to myself “what I could I get them made for?” Services like ponoko.com can do small runs of laser cut plywood to specification in 1/4″ ply, and the design itself is pretty straightforward. I was thinking of dusting off my Fusion 360 skills, but I am beginning to think that using something like Python to generate the DXF or SVG outlines of the pieces. I think I’ll try hacking something together later, and then see if I can use the online quoting system to see what it would cost.

The advantage of laser cutting (at least in my mind) are twofold. It means that I can generate identical modules easily, and the assembly should be easier, since instead of butt joints, we can use finger joints which aren’t really necessary for strength, but which make assembly more straightforward.

But in the meantime, I’ll probably try to plow ahead. Once I get a couple of single modules done, I’ll work on either making corner modules or end caps.

For the rest of today, I’m going to sand down the bondo edges and maybe get some primer or paint on it. If I was going to build some more modules, I think I need to go get some more 1/4″ ply.

Hope you all have a good weekend.

Okay, I’m super busy at work these days at work, but I do still have time to muse about projects that I’ll be getting to once my current project wraps up sometime in about 12 weeks. If you were following my blog, you saw that my intention was to get a couple of very small and inexpensive N scale locomotives, and then homebrew a throttle control/power supply using an Arduino and a DC motor control module.

I got that working!

This is just a simple oval of track running on a 2’x2′ chunk of pink foam I got from Home Depot. The locomotive is a little “Pocket Line” locomotive made by Kato. I also got a small trolley from the same line. I haven’t done much with them other than build them a nice little box to hold them:

Anyway, the other day I was in the workshop and was thinking of building a small table to hold the 2’x2′ section, and which I could start building some scenery and the like. But as I was musing the details, I began to think about the possibility of building it as a series of modules instead (or at least, make it compatible with modular standards).

I knew about NTRAK, a modular specification built around the idea of 2’x4′ modules which could be connected together to make bigger modules. I thought that was a pretty cool idea. But the reality is that I didn’t really want to build a large layout, at least not at first (the module above is actually only half the size of a standard N Track module) and I thought it might be overkill.

But while searching around, I found out about T-Trak, an even smaller standand.

It’s a pretty neat for someone who wants to get into model railroad modeling, but doesn’t want to invest a ton of time or money. A single width module is just 308mm wide, and up to 14″ deep, with a pair of tracks that run parallel to each 38mm from the front edge, and spaced 33mm apart. They specify using Kato Unitrack, which is what I used in my oval. Two packs of 4 segments is enough to make two modules, so the track isn’t expensive. You can also make corner segments, and then use them to assemble ovals or other more complex layouts.

It’s perhaps easiest to see how the basic construction looks by watching a video:

Because it’s a standard format, large sets can be chained together, such as the following at the National Train show in 2018:

It seems like a cool way to decentralize development of a large scale setup for a club, but also to enable people to get started without a large investment of time or space.

Neat.

The Unofficial T-Trak Handbook page has a lot of good information too.

So, I earlier this week I was working on making a wooden box as part of my skill building in woodworking. Actually I made two boxes, but that’s a different story. Here is the better of the two, made from some scrap cedar fence posts:

Rabbeted the sides into the front and the back, and then a small dado holds a plywood bottom in the box. I was all set to make a top which I wanted to round over the top when I discovered that something was wrong. The locking lever that was supposed to lock the router height simply wasn’t engaging anymore. Odd! Nothing that I knew. Oh well.

Tonight I disassembled the entire thing to see what was going on.

Here is the parts diagram from the manual:

So, on this diagram, the part 8A is the lever that is normally used to lock the router into position. It’s kind of a clunky mechanism, which perhaps is easiest to see demonstrated in this video:

https://youtu.be/1P5gUoJLDtQ

But oddly, the locking lever suddenly became entirely loose. I disassembled the router to see what the deal was. The lever is held against the bolt (7A) by two split ring washers (9A). The slightly odd thing is that the bolt is not actually long enough to directly push against the vertical guide rails. Apparently there is another part (6A, labeled LOCK PIECE) that.. well…

It appears to be entirely missing. I’m not sure how this little essential bit went missing, but I found no trace of it.

So, tomorrow it will be time to call Harbor Freight and see if I can find the little piece as a replacement. The router is less than a year old, so I would like to return it to working order.

But if you took the time to watch the video, you’ll see that this router table isn’t actually very good. There are tons of ergonomic issues and build issues that make it less than a complete joy to use, and I’m beginning to wonder whether a project in my future should be to take one of the two other routers I have in my shop which I got at estate sales, and build a true and proper router table (perhaps with a lift, and a better fence, and better dust extraction).

Addendum: I called Harbor Freight’s Technical Support line this morning. Because the router is beyond the 90 day warranty period (I’ve had it for something like six months, back when I built the table for it https://brainwagon.org/2019/04/13/a-rolling-stand-for-my-harbor-freight-router-table/) they didn’t have much to offer. “We don’t have any parts available”. Well, that’s… not great. I am thinking that perhaps I could replace the entire locking mechanism with a longer metric bolt and a 3D printed knob. I’ll have to have a closer look at it. The other possibility is to try to locate one on ebay to serve as a parts donor. Or give up on this cheap one entirely. Grump.

Addendum: Okay, I disassembled the thing, stripping it all the way down. I imagined that the failure mode was that the little LOCK PIECE 6A had slipped down inside one of the vertical posts. I had to pull basically every visible screw, and at each point I peered, poked and prodded. But finally, just as I was about to give up, I heard a faint rattle, and… voila! A small cylinder of brass dropped out. Huzzah! I wasn’t up to carefully reassembling it tonight, but I should be able to restore it to working condition this weekend. Did I mention that the mechanism is really dumb? It is. But I don’t need to immediately replace it.

Over the holiday I was watching my usual raft of videos on Youtube, and got interested in the circuitry that is inside the common garden light. I have a number of super cheap $1, and if you watch any teardown videos of them, you’ll find that there is almost nothing inside them: typically a solar cell, a battery, an LED, and an inductor and usually a small 4 pin chip. I was interested in running some experiments, so I disassembled one of mine which looks like this model. I hooked up the solar cell to my multimeter, and measured about 2.5v at 12ma or so, for about 30mw of output power. The chip inside was a YX8018, and it also had a AAA NiCad battery which tells me it’s capacity is 160maH. Tiny.

These lights seldom incapable of maintaining a light (even as feeble as it is) the entire night, but I was curious about the circuits, so I went ahead and decided to try to experiment with them. My ultimate goal is to characterize the cheap solar cells, and also monitor the charge cycle of the battery. In the next few days, I’ll setup an Arduino as a data logger, and measure how the current flows in the circuit over a couple of days.

I could have used the scavenged components, but I thought it might be fun to start from scratch, and be able to make several versions of the circuit to compare and contrast. So, I ordered a pack of 20 QX5252Fs (a similar chip to the YX8018, but cheaper) for a little under $2 shipped from China. I had a couple of 100uH inductors lying around, as well as some fairly bright white LEDS. So, I breadboarded the circuit up. Instead of using the tiny AAA battery, I thought it might be good to use a reasonably high quality ENELOOP NiMH battery. Using the simple curcuit from the QX5252F datasheet and some scraps of wire, I put it up on my workbench:

It works. So, over the next few evenings I’ll begin to experiment with it some more. I wanted to experiment with changing the inductor (which modifies the LED current) and perhaps driving more than one LED to make a brighter light. I’m also interested in potentially using this setup to make a very small solar powered microcontroller project, using some additional circuitry as described here. More experimentation is clearly possible.

Readers might recall that I’ve been interested in small solar energy projects. In the past, I’ve played with a 25w solar panel, a Chinese PWM solar charge controller and a 7Ah battery, and I used it as the power source for a WSPR radio beacon that I ran for a few months in my back yard.

You can read about my fun with that here.

But last month I got more interested in using very tiny solar cells to harvest power and power very small applications, like powering garden lights or maybe running very small microcontrollers in low energy applications. So, I ordered 10 small 5v, 60ma solar cells from Banggood for about $1 each.

https://usa-m.banggood.com/10pcs-5V-60mA-Micro-Solar-Panel-for-Solar-Power-Mini-Solar-Cells-DIY-Electric-Toy-Materials-p-1377649.html

They arrived yesterday. I was feeling kind of bored this morning, so I soldered some clips on one of them, and then went outside with my snazzy Sparkfun multimeter that I got as swag from the Hackaday Superconference and tried aiming it at the sun to see what I got. Mind you, it was only 9:00AM, and we are still in January so the sun isn’t particularly bright, although it was clear.

And the results were pretty good for voltage:

But less so for current. I only got about 36 milliamps instead of the specified 60ma. But again, middle of winter, early morning, not too bad.

I also decided to see what kind of voltage and current I could get if I oriented the cell unoptimally, roughly straight up.

The result was about 5.63v at just 12ma, which is perhaps better than I might have expected.

All in all, I’m fairly pleased (although I haven’t tested them all). I suspect I could have gotten similar performance by looting $1 garden lights from Dollar Tree, but I think they will still be useful and will find their way into an experiment later on. I also ordered some QX5252F chips from ebay which should arrive “any day now”. These are popular in homebrew (and commercial) solar garden light circuits. It will be fun to play with them.

Stay tuned for more experiments.

In my garage, I’m trying to do some simple skill building exercises to increase my precision and knowledge about how to construct useful objects. The recent kerfmaker and half lap projects that I’ve tweeted about were part of this attempt. Last night, after a full day of chasing pixel problems, I thought it would be fun to go into the workshop and use my kerfmaker to try to make a simple tool tote out of some scrap 1/4″ plywood that I had lying around.

I had gotten a tool tote for $1 or so at some garage sale, and had found it so useful that I had constructed a very similar tote myself a month later. But that tote was all butt joints and had the bottom merely cut to size and glued to the bottom, and I used brad nails to hold it together along with glue. I thought that maybe I make a similar (but slightly smaller) tote. I spent an hour and a half, and put together this:

Each of the sides is 8″ side. The front and back are rabetted and have a central dado to take center divider, and a dado to hold the bottom in a recess. The two sides just have the bottom dado. I dried assembled, and things looked “okay” (more on that later) so I glued the side joints, and assembled the four sides and bottom (the bottom just rested in the dado without any glue) and let it dry. I’m going to taper the middle and cut a slot in the center divider to serve as a handle. My experience with the similar tote I made before is that the properly glued joints will be plenty strong to handle whatever load I’m likely to put in it.

It looks pretty good, but there are a few niggly details, most notably that it doesn’t rest entirely flat on the desk. That is because the corners look like this:

The front panel here slightly protrudes. I believe this is because the dado I cut to receive the shot isn’t 100% parallel to the bottom, and so when I clamped the bottom in place, it pulled these sides out of flush/square by about 1/2 mm. Oh well, that’s what my sander is for.

Some other pictures below. If the mood suits me, I’ll finish the tote in another evening or so.