Making a METAR Map

The METAR (METeorological Aerodrome Report) is a staple for pilots wondering what the weather is like at a given airport. While a full report looks like incomprehensible output from a 1960’s teletype (because, well, that’s pretty much exactly what it is), pilots are lazy and sometimes just like having a “yeah/nah” appraisal of weather rather than something like
KOSH 290304Z 10012KT 3SM BR OVC004 02/01 A2990 OMG WTF LOL

Laziness aside, it’s pretty handy to get an idea of flying conditions across a region by looking at a weather map such as the one provided at aviationweather.gov or via various forms of flight planning software. Green, blue, red, and purple indicators give a quick indication of current cloud ceiling/visibility for reporting airports. In the past couple of years, making a live lighted wall map to provide this information has become a popular craft project for flyers, and commercial versions have become available as well.

I’ve been wanting to make a METAR map for my flight instructor ever since first seeing examples online, and finally got around to building one. My primary guide in this was Philip Rueker’s blog post as well as his GitHub project files (though there were several other sites that also provided information and inspiration). Rather than rehash those excellent instructions, this post is regarding my own approach to building.

First off, shadow boxes are much more expensive than regular picture frames. I decided to try to keep the depth of lights and electronics to a minimum so that I could use a standard picture frame. I wound up getting a 20×28 frame from Michael’s for the purpose (and made sure to wait for a good coupon or sale, as $57 is a ridiculous price for a few pieces of MDF and a sheet of glass).

While pretty much every other METAR map build I’ve seen just has through holes for the LEDs, I didn’t want to use this approach for two reasons. First, I wanted to keep the map intact so that it would just be backlit (and thus allowing replacement in the future should there be a big airspace change – sectional maps are only valid for 6 months, after all). Second, I needed to keep the components as low profile as possible. Thus, I decided to forego the WS2811 LED strings used by others and instead opted for the far more compact WS2812B PCB ‘chip’ LEDs.

The picture frame used a fiberboard backing, which I figured would be a fine substrate to mount the LEDs to. I used an outdated sectional map cut just larger than the frame backing, and gave the back of the map a light mist of 3M spray adhesive. After letting the adhesive dry off for about 10 minutes (I didn’t want it to be sticky, just tacky), I carefully applied it to the fiberboard backing, aligning a longitudinal line with a corner of the backing (ensuring that I could remove the map and replace it with a new one).

After trimming the map to the backing, I then zoomed in on the online weather map at aviationweather.gov and marked off every airport on the paper map that showed up online. 38 airports in total would need to be represented. Once I had them all tagged, I used a 1/4″ brad point drill bit and a section of clear acrylic sheet on top of the map/backing to drill a hole cleanly through each airport location.

After that, I marked the backside of the backing board with each airport identifier and name. Then, I removed the old map and made sure each hole was cleared of swarf. I prepared a current sectional chart with spray adhesive, let it tack up, and applied it to the backing board in the same fashion as before, leaving an unscathed map on top (after each application, I used a razor knife to trim the map to the size of the backing board – would have been much more difficult to try getting it sized right beforehand).

Then it was time for the drudgery. Namely, hot-gluing 38 tri-color LED PCBs to the rear of the backing board. And then soldering them all together in a chain. Six solder pads per LED. Honestly, the process wasn’t that bad and I did it over the course of a few evenings. With that completed, I could finally think about powering the whole project. While Philip Rueker chose to run his directly from the RasPi’s GPIO port, he was only dealing with 22 LEDs. With this map being 38 LEDs deep, I opted to have them powered directly from a separate 5V supply. So, how to power the RasPi and LED chain effectively? Well, the RasPi has a micro-USB connector for power. And USB itself runs at 5 volts… A bit of digging through boxes of old hardware yielded a powered USB hub that didn’t need an active host in order for the ports to be powered.

I 3D printed mounting brackets for the RasPi and USB hub. With yet more hot glue to secure them to the backing board, the hardware side was complete. Data and 0V ran from the RasPi GPIO to the LED chain, while 5V and 0V ran from another port on the USB hub to the chain (I didn’t need a diode or level shifter to bring the signal line voltage closer to the LED supply voltage).

The USB hub was a little bit taller than hoped, so I used self-adhesive wall hooks on each corner of the frame to give everything a little more wall clearance (and serve as a way to hold up the power cord).

Much to my shock and amazement, everything seemed to be working great. It only took a little bit of tweaking the metar.py file to complete the project (the LEDs I used have an RGB addressing scheme rather than the WS2811’s GRB order).

Not what I wanted to see weather-wise this weekend, but 100% what I wanted to see project-wise!

Edit: The system wasn’t as stable as hoped, and a strange reboot loop would result within an hour or so of startup. A bit of googling pointed to a feeble power supply as a culprit. I figured a big beefy linear supply wall wart coupled with a substantially warm voltage regulator on the USB hub PCB would equate to an inefficient yet thoroughly reliable voltage source. Yet replacing it with a cheapo dual-port USB wall outlet cube eliminated all issues. Huh, sometimes “they don’t make ’em like they used to” turns out to be a compliment regarding the newcomer!

I also made a few final tweaks to the software side of things. My home airport (where the map will live at) doesn’t have any sort of automated weather reporting system, but I still installed an LED for it. I set rc.local to illuminate that LED in white right away as an indication that the system had booted. After that, I added a “sleep 10” line before running the startup.sh script that initially fires up metar.py (I found that the RasPi wasn’t necessarily yet connected to WiFi, and you might have to wait for cron to kick in before any LEDs would light up as a result of the API lookup failing). Finally, I modified metar.py to show our airport as white (I have it as an entry in the airports file, but since it doesn’t have weather reporting, the LED is normally off). I probably know as much about python in a zoological respect as I do in a programming respect, so it took me a little bit of trial and error to finally figure something out. In the end, inserting a line of “pixels[27] = (85,85,85)” just before the final “pixels.show()” command did the trick.

Jeppesen TechStar Flight Computer Manual

Recently a friend gave me their duffel bag of old student pilot training materials (they kept their log book – hopefully they might pick up training again at some point!) and among the various old books, binders, and brick-a-brac was this gem:

It’s an old handheld flight computer, and not the ‘whiz wheel’ E6B circular slide rule we’re accustomed to, but an actual electronic flight calculator. I had been pondering getting one for my knowledge test just because there’s a handful of cross-country planning questions that require you to work backwards on the E6B from the standard “set wind direction, mark wind speed, set course, etc.” process, and it’s pretty confusing. Whereas with an electronic flight computer, you key in what you know and the magic silicon pixies do the rest for you.

For a 26-year-old device, it still does handily, but it seems like nobody out on the webernets has a manual for it (searches turn up a load of spam links). Fortunately, I found the manual tucked into a pocket, so I scanned it and am posting it here for others to hopefully find:


Joystick Repair

I intended to post this a few months ago, but the old server had simply gotten too rickety to cope with new posts, much less OS or WordPress updates.  So I opted to roll an entirely new VM with an OS version published this decade.  Hopefully previous posts imported properly, but the site may be a work-in-progress for a while.

A friend’s all-in-one Namco joystick had broken, and they wondered if I might be able to fix it.

The issue was that the bushing that the joystick shaft is pressed into had split.  Since it looked to be Nylon, my guess was that just trying to glue it back into place (even with epoxy) would not last long as a fix.

While the manufacturer is still in business (though on a different URL), their site doesn’t appear to have any mention of this particular unit.  I wasn’t really expecting to find an exploded diagram and parts ordering form anyhow, but once in a while pleasant surprises pop up.  Not so in this case, so time to wield Phillips screwdrivers with careless abandon.

The joystick mechanism looks to be secured under that square base plate.

This turned out to be a far more complex mechanism than I had anticipated.  It wasn’t until I gave the unit back that I learned that this is due to the Pole Position game using joystick twist for steering instead of normal left/right movement.

My plan was to find some thinwall stainless or brass tubing to sleeve the outside of the bushing and thus collapse down the split section.  I didn’t have much luck in finding an appropriate size tube in the McMaster-Carr catalog, however.

No big deal when you have machine tools, though!  I found a piece of scrap stainless rod, faced the end, and drilled a shallow hole with a W size drill bit (0.386″, which is just the size of the bushing’s outside diameter).  I then thinned down the outside until I reached a point of ‘that looks about right’.  This turned out to be a wall thickness of around 0.017″ which I hoped would be thin enough to retain full movement of the stick.

After parting off the ring, I did a little light deburring.

The ring pressed snugly into place on the bushing, and the crack is practically invisible now.

I carefully hammered the joystick shaft back into the bushing when done.  Interestingly, the end of the shaft has both splines and ridges to keep it from pulling out and from twisting – that should have been a clue to me regarding the twist operation.

A few games of Galaga verified that operation is back to normal!

Aviation pt. 2

It’s been close to 2 years since my previous post on the subject of aviation.  I’ve done a great deal since then – I obtained my sport pilot license, I’m now working towards my private pilot license, my main instructor moved away so she could become a test pilot for an aircraft manufacturer (which I brag about at any given opportunity), and I stumbled into helping out with two different aircraft restorations via her father, who has since become a good friend and source of inspiration.  That summary glosses over a lot of details of little importance to anyone who is not, well, me.

Last night I drove to the field to drop off some drawings and get some dimensions for one of the restorations we’re working on.  I stopped by my flight school’s hangar to see if anything interesting was going on.  It was a perfect evening, with almost zero wind, endless visibility, and a beautiful setting sun in a sky dotted with only a handful of clouds – the sort of sky pilots dream of.  As it turned out, the school’s owner and chief instructor was doing a little flying with one of the kids who had been helping out at the hangar since I started my own training, and after they landed, he hopped out of the Cub and gave some instruction and encouragement to her.  Yes, this was to be her first solo.  She absolutely nailed her landings, leaving those of us on the ground wishing we could make our own approaches as cleanly.

One of my other instructors who was trained and then actually taught at the school (and is now a fresh new FO at a regional airline) even drove 2 hours to show up for the event.  After my own first solo, I’d still tell people that I was just a ‘student pilot’.  However, one pilot that I admire chastised me for prepending the ‘student’ moniker, admonishing that “there is no more binary distinction than those who have flown a plane by themselves, and those who have not”.

When she finally taxied back to the hangar, pictures were taken and scissors were deployed to remove her shirttails as per tradition – the rafters of the hangar are filled with the autographed fabric (mine included) of those who have made that leap across the divide.  There’s a lot more training and studying to be sure – but I hope she now calls herself  a ‘pilot’ instead of ‘student pilot’.  We’re all student pilots, after all – even my old instructor, the regional FO who now flies airline jets, will need to get checked out in the school’s Warrior before he can give an IFR refresher to one of the new instructors.

We all recounted our own first solo, how the plane becomes a rocketship without your instructor in front, and the fear and exhilaration that comes with flying an airplane all by yourself.  It’s a really special occasion, and I think it’s only properly appreciated by others who have gone through that same experience.  Flying is a very small brotherhood, but last night, we added one more to our ranks.

Stratasys EEPROM hack revisted

Five years ago, I published what is probably the single most popular post on this blog – namely, how to reset a Stratasys material cartridge EEPROM so that it can be refilled with much cheaper third party filament.  Dan at gnurds.com took things a step further and came up with some great step-by-step tutorials on how to accomplish this, and even had a post featured in hackaday.  I received a lot of questions and feedback on the hack (and even helped hack a few machines running in the field), but the fact that the process requires modifying contents on the printer’s hard drive put it out of reach for users who were barred from doing any sort of ‘invasive surgery’ on the machine (generally students stymied by school officials).  What was needed was a truly ‘touchless’ hack that didn’t require any modification of the base machine itself – we needed a way to actually generate our own EEPROM data from scratch rather than simply re-using the EEPROM data as it had come on the cartridge from the factory.  The problem is that the EEPROM data is encrypted.  Worse, my understanding of how DES crypto is done on a practical level is pretty much zero – I’ve done embedded programming before, but this sort of stuff is far outside my area of expertise.

Fortunately, in 2013 a wizard cracked the EEPROM crypto and posted the code to github!  Benjamin’s sorcery is just what’s needed to build your own EEPROM image, and he’s put a great deal of time and effort into it for zero financial gain.  As a result, it really bugs me when I see people using his code to sell programmers for hundreds of dollars, with zero attribution for his work.  This post, then, shows start-to-finish how you can refill your own P-class Stratasys cartridge using only a Raspberry Pi.  It’s admittedly long and detailed, but I think it’s important to understand the whole process before trying to use any shortcuts.

The reason that it’s taken me so long to document this hack is that I previously had no real access to a P-class machine.  Fortunately, my friends at Into3D LLC have one in their shop and were more than willing to let me attempt refilling one of their empty cartridges.  It’s a Dimension BST 768, which fortunately is still supported for 2 more years.  This is what the front panel looks like with a near empty (1% material remaining) model cartridge:

Here’s that very cartridge – the label on top lists some applicable patents (6776602, 7063285, 7341214, D436111, and 7754807):

The side has a recycling information label and identifying information for the cartridge itself – we’ll see how that matches up with the information on the EEPROM itself.

Use a 7/32″ hex wrench to remove the 4 screws on the underside of the cartridge (I used a fold-up set to break them free first, then a standard L-wrench to remove them).

Then, flip the cartridge over and give the shell a few raps with your knuckles to shift the internal desiccant packs into the bottom half.  Carefully lift off the top half of the cartridge, exposing the 1% of material remaining on the spool:

Despite my dislike of using cartridges for 3D printers (it’s a lot of excess material to house the consumable), these are actually quite well designed – there’s a pair of very simple drive wheels at the corner exit, and the other 3 corners get desiccant packs.  The orange-brown circumferential seal has a spot for the filament to exit through, and the screws actually thread into brass inserts, not into the raw plastic.  A single cartridge can be reused many times, potentially lasting a fair portion of the life of the actual machine.  Parked right next to the drive wheels in a slot is the object of our interest, the EEPROM board itself:

Here’s what the front and back of the EEPROM PCB looks like:

In case anyone is wondering, the text on the chip itself is:

DS2433
1226B1
586AC

As noted in the original blog post, the chip is a DS2433 (originally a Dallas Semiconductor product, hence the ‘DS’, now owned by Maxim).  Importantly, it is a 1-wire device, hence requiring only 2 contacts to the chip.  As an aside, this nomenclature has always annoyed me – power and signal may be carried over a single wire, but you still need a ground connection.

Now, to extract those pesky bits from the EEPROM, all 4096 of them!  Technically, there’s a few more bits that we’ll need as well – a 48-bit serial number, 8 bits for CRC, and an 8-bit family code (0x23).  I’ve used a Bus Pirate before, and you can use an Arduino as well, but for this post I’ll be showing how to use a Raspberry Pi, since we can do everything on a single, standardized platform.  In this case, I’m using a Raspberry Pi model B with an Adafruit breakout board, and I started with a clean NOOBS image (v1.7.0 to be specific) on a freshly formatted SD card.

I booted the RasPi and was greeted with the installer – since I had the RasPi connected to my network, it provided me with network installation options, but I selected only the first option to install the Raspbian OS:

While that was installing, I prepped the breakout board.

I connected the black alligator clip to a GND (0vdc) terminal and the red alligator clip to IO4 on the breakout board (which is pin 7 of the RasPi header).  Note that pinouts on the RasPi can be very confusing – IO4 on the breakout board is not GPIO.4 on the RasPi as I had first thought, but pin 4 of the Broadcom BCM2835 processor at the heart of of the RasPi (also known as GPIO.7 on the RasPi).  You can read more about this confusion at wiringpi.com.  I connected a 2.2k resistor to a 5vdc terminal and the IO4 terminal – this acts as a pullup.

By this time, Raspbian had finished installing and I was looking at a fresh new desktop.  I find it easier to just SSH into the RasPi, so once I determined its IP address (just hover your mouse over the network icon in the upper right of the desktop), I could fire up PuTTY and connect right to a shell (you can certainly do everything via a terminal window on the desktop, though).

On a default install of Raspbian, the login is pi and the password is raspberry.  While Raspbian includes just about all the software we’ll need by default, the Python crypto library will be required later on, so run sudo apt-get install python-crypto right away to install that package (the RasPi will need to be network connected with internet access for this to work).  I then attached the breakout board to the RasPi and proceeded to do some testing with gpio to make sure that my wiring was correct.

I hooked up an oscilloscope to the red and black leads (you can use a multimeter, but a scope will let us see when EEPROM reads are occurring, which is handy when debugging), and saw that I had nearly 5vdc, which is just what I was expecting due to the 2.2k pullup resistor.

Now, let’s have a look at actually controlling the pin that the red alligator clip is connected to.  The gpio readall command gives us a snapshot of what the status is of all the header pins (again, see wiringpi.com for details).  By default, all the I/O pins are set to be inputs (note pin 7 is set as ‘IN’):

We can change the mode from input to output on that pin with gpio mode 7 out

And as soon as that command is issued, the voltage drops to zero:

Issuing gpio write 7 1 will bring that pin high:

Resulting in 3.3v output (RasPi digital I/O is 3.3v, not 5v):

That’s all for verifying that the wiring is correct.  There’s one final change to make before we can actually try reading in an EEPROM, and that’s disabling Device Tree (I understand it’s possible to get things working with DT enabled by means of some other configuration changes, but disabling it altogether is the route I went with).  Run sudo raspi-config to bring up the configuration menu, and select Advanced Options:

Then, select Device Tree:

Select ‘No’:

And DT will then be set as disabled:

Back at the main menu, select Finish and you’ll be prompted to reboot (select ‘Yes’):

With the RasPi rebooted, we’re finally ready to read in an EEPROM.  Recite the magic incantations sudo modprobe w1-gpio gpiopin=4 and sudo modprobe w1-ds2433 (note that you’ll need to run those commands again if you reboot the RasPi, so it may be worthwhile to add them to a startup script):

Now, connect the clips to the EEPROM – black clip to the ground pad, red clip to the data pad (you can tell which is the ground pad because it has a trace on all 4 sides connecting it to the ground plane):

If you have a scope hooked up, you’ll see that the voltage drops to 3.3 volts and then there will be a data read every 10 seconds or so:

We can cd /sys/bus/w1/devices/w1_busmaster1 to have a look at the connected 1-wire devices seen by the RasPi, and therein is a specific directory created for that EEPROM (23-0000014d4762 in this case – your EEPROM will be different!):

You can also use the xxd command to hex dump the EEPROM’s full UID as shown (0x2362474d0100006b).  This particular UID consists of the family code (0x23, which should be the same on all Stratasys P-class EEPROMs, except for those used on uPrint cartridges), the device serial number (0x62474d010000 – note the endianness) and finally the checksum (0x6b).  The screenshot also shows a dump of the 512 bytes of EEPROM data itself.  In order to actually do anything with this EEPROM data, though, we’ll need Benjamin’s code.

We’ll cd back to our user directory and then grab a .zip archive of the code via wget https://github.com/bvanheu/stratasys/archive/master.zip, after which we can extract it into a directory via unzip master.zip.

We’ll cd into that directory and try running the main program by executing ./stratasys-cli.py -h

This is just what we want to see!  If instead you get a number of errors mentioning crypto, make sure that you have the python-crypto library installed.  At this point, we don’t need to do any further configuration on the RasPi, and we can actually dive into the EEPROM data itself.  If you’ve managed to get this far, you’re probably capable of basic command line Linux work, so I’m going to gloss over those details from this point onward and let screenshots do most of the talking.  First, we’ll copy the EEPROM data out to a file that we can actually work with.

A directory listing confirmed that the resulting file is exactly 512 bytes in length, and all the formatting looks just like all the other Stratasys EEPROM dumps I’ve seen.  Let’s take one more look at the EEPROM UID, as we’ll need to format it correctly to feed into the stratasys-cli program:

We need to reverse the byte order of the shown UID, so instead of 23 62 47 4d 01 00 00 6b we’ll use 6b 00 00 01 4d 47 62 23 (remember, the family code of 0x23 is at one end, and the checksum is at the other end).  Here we finally feed the EEPROM UID and the EEPROM data through the program, using ‘prodigy’ as the machine type (any Dimension series machine should be ‘prodigy’, but a different machine type (Titan, Maxum, etc.) will have a different family name):

If you don’t have the UID formatted properly, the program will fail with a checksum error:

If we take a quick look at the data, everything agrees with the label that was on the cartridge – the color, serial number, manufacturing lot, and manufacturing date all match perfectly (I’m guessing that the timestamp on the EEPROM data is actually GMT).  Note that while the last use date is a separate entry in the EEPROM data, I’ve never seen it differ from the manufacturing date (maybe Stratasys intended to write this data back to the cartridge, but couldn’t be certain that the printer itself would actually have the correct day/time set).  A service manual I saw indicated that this timestamp was intended to be the date/time that the cartridge was actually first inserted into a machine, so perhaps Stratasys intended to have cartridges ‘expire’ after a certain amount of time.

The EEPROM stores material quantity in terms of cubic inches.  A brand new cartridge contains 56.3 cubic inches worth of filament, and the current level as shown on the EEPROM is under 0.75 cubic inches, so the 1% filament remaining message on the printer’s front panel was right on (there was actually a little more than that remaining on the reel, but the overage allows for nozzle purges, waste due to swapping cartridges, etc.).  The original hack worked by simply setting the current material quantity back to 56.3, but the printer would remember the serial number of the cartridge (and refuse to work with that cartridge serial again), so files needed to be deleted on the printer itself.  Since we can now create our own EEPROM images from scratch, we can simply change the serial number in addition to the material quantity, and the printer will be none the wiser.

In addition to decrypting the EEPROM data and displaying it in a human readable format, Benjamin’s program can provide all the parameters needed to generate that very EEPROM data, which is extremely handy.  We’ll simply tack on the -r option to the command we just used:

We only need to change the highlighted options to create our desired EEPROM image:

Reading the generated neweeprom.bin file back through the program shows that the serial number is now different (I incremented the value by 1000 rather than 1 just in case there were consecutively serialized cartridges the machine had already seen), and the material quantity is back to ‘full’.  Now, we can finally write that file back to the EEPROM itself (remember to use sudo when writing to the EEPROM due to permissions):

A quick hexdump directly from the EEPROM verifies that we’ve modified the contents successfully.  Note that the new image we generated is only 113 bytes, while the original EEPROM data is a full 512 bytes.  This is because everything after the ‘STRATASYS’ at 0x69 is random garbage (early on, Stratasys simply padded out 0x71 onwards with zeros), so that portion of the EEPROM can be left unchanged.  That’s all that we need to do with the RasPi – the EEPROM can be disconnected and set aside.  All that remains is to reload the cartridge with fresh material.

My favorite current filament is the ‘High Performance ABS’ from Coex3D – I’m partial to them because they’re local, their product quality is excellent, and Chris gave an amazing presentation on polymer processing and filament extrusion to our 3D printing user group a few years ago.  He listened to my pleas for MG94 filament and thankfully started producing it.

While Coex3D reels fit right into my FDM series machines, they don’t fit into P-class cartridges (the bore is just a little too small, and there is a rubber piece attached to the top shell of the cartridge that would interfere anyhow).  So I needed to respool the new filament onto the old reel.

In fact, respooling the filament itself is probably the hardest part of the entire process!  I admittedly got a bit hackish with the use of electrical tape as a makeshift drive dog, but it worked.

Make sure that you have the desiccant packets in place, the EEPROM PCB in its slot, and the drive wheels in position, then very carefully feed the end of the filament through the hole in the disc portion of the gasket.  Hold the filament in its guide slot, and carefully place the top shell half back in place (this is pretty tricky without having the gasket jump out of place).  Flip it over, assemble the cartridge with the 4 screws, and our work is complete.  Now, for the moment of truth…

After loading the refilled cartridge into the printer, it showed a 100% full model spool!

Of course, the proof is in the prints, and the machine is now happily churning out parts far more inexpensively than before (and with better material properties than the OEM P400 filament had).

Notes:

1) Benjamin’s code works for other Stratasys printer families as well, including the big T-class and Maxum/Quantum machines.  It will even decrypt and generate binary images for uPrint cartridge EEPROMs.  However, uPrint cartridges use a different EEPROM model that unfortunately utilizes an HMAC authentication scheme for writes.  So while you can read EEPROMs and generate new .bin files, actually writing them back to the EEPROM is impossible without knowledge of the secret key used for the HMAC authentication.  If you only have access to a uPrint, you’re kind of stuck. [12FEB2017 addendum]  It turns out that an HMAC protected EEPROM can be re-written by using the DIAG port on the printer.  ‘Odin’ discovered this workaround and authored these concise instructions for the hack.  This isn’t quite ‘touchless’, but until someone breaks or sidesteps the HMAC protection, it’s the best option.

2) 56.3 cubic inches isn’t actually the maximum material quantity you can set on the EEPROM.  It’s a floating point number, so the sky is the limit!  However, on a P-class machine, 60.0 cubic inches is the maximum that the printer will accept as valid, as anything larger will instead be considered zero. [22FEB2017 addendum] Reports indicate that larger volumes can be used, but the ‘initial material quantity’ field needs to be similarly large.  However, others (Odin, for one) find that this doesn’t work on their machine, so this could be a difference between firmware revisions.

3) I mentioned using an Arduino to read/write the EEPROM.  There’s actually a fork of Benjamin’s code available on github called CartridgeWriter that runs in Windows and interfaces with an Arduino for EEPROM reads/writes.  I haven’t tried it myself, but others have used it successfully.  I prefer running Benjamin’s code directly on a RasPi due to the amount of control it provides.  Specifically, it makes some automation possible, such as…

4) …this clever little EEPROM rewriter made by Sneaks Hacks.  He’s posted the wiring schematic and STL files for the housing, and is currently working on an updated version as well as documentation for people wanting to build their own.

5) The Stratasys EEPROM PCB has a surface mount 4.7k pulldown resistor right next to the DS2433.  While most wiring schematics for reading/writing a Stratasys EEPROM show to use a 4.7k pullup between 5v and the data pad, I used a 2.2k to help ensure that voltage to the DS2433 remains in an acceptable range.  You can certainly start with a 4.7k pullup and reduce the resistance if you’re having flaky results, but I don’t think I’d go below 2.2k.

6) Benjamin notes that use of the –output-file option when generating EEPROM images isn’t actually correct in a traditional Unix sense, and that I/O redirection is really the proper Unix way to get binary output from stratasys-cli.py (that is, using something like ./stratasys-cli.py –foo –bar > my_file.bin).  I think using –output-file is a little easier to see and understand what’s going on, but if you’re building an automated system like Sneaks Hacks did, using stdout can help streamline the processing chain.  For example, you could write directly to the EEPROM and save a backup of the image all in one go by using something like ./stratasys-cli.py –foo –bar | tee backup_image.bin /sys/bus/w1/devices/w1_bus_master1/23-0000014d4762/eeprom.

7) Everything in this post is simply the compilation and distillation of other people’s hard work – I simply wanted to bring it to a wider audience.  These are the folks who deserve all the credit:

  • Huge thanks goes to Mjolinor, who I believe was the first to pioneer using a RasPi as an all-in-one solution for rewriting Stratasys EEPROMs.
  • Thanks to the anonymous and unnamed people (you know who you are) who have assisted in developing and testing these techniques over the past few years and helped review this post before publishing.
  • Steve, Joe, and John at Into3D LLC for letting me use their Dimension BST 768 as the guinea pig for this project.
  • First, last, again, and finally, Benjamin Vanheuverzwijn.  This man is an absolute electronic wizard and none of this would have been possible without him.  Send him coffee.

Gunsmithing the RAS-12 – Part 1

It’s no secret that I have a penchant for oddball guns, be they paintball marker or firearm.  Seeing a design that’s off the beaten path is always enjoyable, be it for mechanical ingenuity or sheer impractical novelty (though usually a peculiar blend of both).  During a recent visit to my friendly local FFL (always good for a rousing discussion and perhaps lightening of my wallet) a friend and I perused a distributor’s sale flyer and immediately spotted an intriguing item – a 12ga shotgun upper for AR-10 lowers.

For those unfamiliar with the AR-10, it is the bigger, older brother to the AR-15 rifle.  In fact, much of what was considered new or novel at the time of the AR-15’s introduction is properly credited to the earlier AR-10 design.  Unfortunately, there is far less standardization on the AR-10 platform than there is on the AR-15.  Eugene Stoner actually updated his AR-10 design decades later to have much greater commonality with AR-15 parts (resulting in the KAC SR-25), while Armalite (not the original Armalite that actually developed the AR-10, just somebody who bought the name and rights) developed the AR-10B using an upper from an SR-25, and somehow DPMS came up with a mashup of these designs, and then…  …yeah, I don’t really understand it all either.  We’re left with saying ‘AR-10’ as a generic terminology for something that looks like an AR-15 but fires a .308 round – as someone once wryly observed, “there’s an XKCD for everything“.

The important part is that the modularity of the AR-15 and AR-10 allows different upper receivers to be mounted to a common lower receiver (which is, as I’ve noted in previous posts, the one part that is itself considered the ‘firearm’ under US law).  While the AR-15 is far more standardized than the various AR-10 incarnations, the magazine well of the AR-15 limits what ammunition can easily be fed through it.  In fact, this limitation has been the underlying factor in the development of various new cartridges such as the .50 Beowulf, 6.8 SPC, .458 SOCOM, etc.  Although a 12ga shotshell by itself will just barely slip through the magwell of an AR-15 lower receiver, designing a practical magazine to feed that ammunition through said magwell is out of the question.  So, the next best thing is to scale up to the larger AR-10 lower receiver.

The ever-popular Magpul .308 magazines will happily accept a standard 12 gauge shotgun shell (extracting said shell is another matter, though).  The rimmed base of the venerable shotshell does not lend itself well to use in a box style magazine (as opposed to the tubular magazine that most traditional shotguns use).  As an aside, this issue isn’t unique to shotshells.  The rimmed base of the famous .44 Magnum round (feel free to insert your favorite Dirty Harry quote here) has limited its use in semiautomatic handguns to only 2 models in history, as far as I am aware – the iconic Desert Eagle, and Emilio Ghisoni’s masterpiece, the Model 6 Unica.

The designers of Kalashnikov derived shotguns (such as the Saiga-12 and Vepr-12) use special magazines with a fairly severe feed angle to improve reliability when stripping and chambering a round.  Unfortunately, that doesn’t translate terribly well to using a straight magwell and magazines designed for rimless ammunition.  The designers of the RAS-12 opted for a pretty radical approach to this problem – they designed their own ammunition.  Which is probably why I managed to snag this very interesting upper for less than a sixth of its original retail price – not much more than 2 years after announcing the product, the company doesn’t seem to be in business anymore.  I may write more on this later, but I’ll limit myself to covering just the ammunition in this post.

The ammunition comes in boxes of 5, with 20 boxes to a case.

The cartridges look very little like a traditional shotshell, and very much like a modern rebated rim pistol cartridge.  In this manner, it is reminiscent of the .50 Beowulf cartridge designed for the AR-15 platform.  The rounded nose and rebated rim makes feeding far simpler than with a standard 12ga shotshell, and allows for easy adaptation of existing magazines.

The most significant feature of the cartridges is obvious – they are not of metal construction, but polymer (US patent 9109850 calls out polycarbonate and nylon as suitable materials, though various online sources specifically note polycarbonate as the hull).  This feature alone is what made me take notice of the system, given that I’ve done a bit of tinkering with 3D printed polymers in gunsmithing applications.  Even if supplies of the original ammunition dry up *cough* Gyrojet *cough* Dardick *cough* EtronX *cough* it should still be possible to recreate the cartridges in a reverse-engineered fashion.  I’m somewhat surprised that the RAS-12 designers didn’t opt to ‘open source’ the design, as SAAMI standardization is precisely what has allowed previously proprietary cartridges to survive in the market if not outright flourish.

I carefully disassembled a cartridge to determine the weights (in grains) for all of the components:

  • Projectile half: 512.4 gr
    • Nosecone: 24.8 gr
    • Nine pellets of 00 buckshot: 481.2 gr
    • Wadding: 6.4 gr
  • Propellant half: 191.0 gr
    • Nitro card: 13.0 gr
    • Gas seal: 17.0 gr
    • Powder: 29.0 gr
    • Hull: 117.2 gr
    • Primer: 14.8 gr

…for a grand total of 703.4 grains for a fully assembled cartridge.  There’s certainly a bit of tolerance to these measurements, but they should serve as a suitable starting point for weights.  Now, to start measuring the hull and nosecone to draw up in CAD…

4-jaw chuck for the Keiyo Seiki

I recently had another Keiyo Seiki/Homach lathe owner contact me, and I mentioned that I had been meaning to get around to posting a copy of the manual that came with my machine.  It turns out that other owners have been looking for one, so hopefully this helps a few people: Keiyo Seiki Homach KM-1800C lathe manual

When I purchased my lathe, it came with both a 3-jaw and a 4-jaw chuck – the 3-jaw was mounted to the headstock and the 4-jaw was in a box.  However, the first time I wanted to try using the 4-jaw, I discovered that the chuck uses a D-6 camlock mount, not the A-6 mount that the lathe spindle actually has.  Apparently I had been given the incorrect chuck, but it had been long enough that I didn’t have the seller’s information anymore.

My first thought was to see if I could somehow mount the 4-jaw with an adapter plate of some sort – the jaw body itself mounts to a ‘spider’ that has the camlock lugs on it, so all I would need to do is to replace the spider with a backplate.  Well, in an ideal world, that would be the case.  The screws that attach the spider to the chuck body are attached from the spider side, not from the front face of the chuck.  Thus, using an adapter plate would be a mechanical impossibility with no way to reach the attaching screws.  So now I have a D-6 4-jaw that needs a new home, and I needed to start from scratch.  Off to Enco!

Enco did in fact have the sort of 4-jaw chuck I was looking for (I selected it based on the size of object it could pass – anything smaller than the headstock ID would be wasteful), so I purchased it during one of their free shipping promotions.  All I needed then was a backplate, which I procured from “Industry Recycles” on Ebay.  I think it was actually an Enco offering but I managed to snag it for about half price.  Score!

Unlike the backplate for the 5C collet chuck, this backplate mounted up just fine.  The trick, then, was mounting the chuck to the backplate rather than mounting the backplate to the lathe.  Mounting hardware was easily purchased from McMaster-Carr, and then it was time to start making holes in things.

Given my experience with the 5C collet chuck mounting, I decided to be a little more precision oriented this time around.  I started by measuring the hole locations on the chuck itself.  Using calipers and screws inserted into the mounting holes, I measured the distance across each pair of screws (both ‘inside’ and ‘outside’ measurements).

I found that the locations of the four mounting bolt centerpoints differed by about 5 thousandths of an inch.  Instead of just basing my cuts on a single measurement, I opted to use the average instead, which came out to be a bolt circle of 5.10425″, or a radius of 2.552″.  Easy work – put the backplate in the vise in an orientation that will clear the existing holes, indicate the center bore, and drill 4 holes at +X, -X, +Y, and -Y locations of 2.552″.

Wait, the Y handwheel stopped….

Noooooooooo!  I ran out of travel little more than an eighth of an inch from the lowest hole location.  I started to resign myself to moving the vise to a new location on the table (I hate having to re-indicate in a vise almost as much as tramming the head), when I realized “duh, just rotate the hole pattern by 45 degrees”.  So each hole would be at X/Y +/- 1.8045″.

After drilling, I ran a M14x2.0 tap through the holes.

The backplate mounts just fine to the lathe spindle….

And the 4-jaw mounts just fine to the backplate!  I started right in on drilling some Delrin for a new product offering.

PIR foam for Stratasys build trays

Filament is the primary consumable on Stratasys machines, but not the only one – the build trays are also considered consumable.  On newer machines that use plastic trays, the official recommendation is to use them for only a single print, but that starts getting expensive fast, so users will wipe them down with acetone or try sandblasting them to freshen up the surface for more use.  The old FDM series machines that I have don’t use plastic trays as a build base, but rather a foam square.  I had already investigated alternate materials with poor results, and John Branlund had started to look into craft store foam instead, as it seemed to be the material to beat.  We eventually concluded that the foam is most likely General Plastics FR-7104 modeling foam based on the sample kits that the company provides.  I haven’t pursued this further, however, as I can’t imagine that shipping on a 4’x8′ sheet of rigid foam could be anything close to ‘low cost’, and I have about half a dozen extra foam bases anyhow.  However, I’ll be turning the FDM 1600 loose to be used by other Makerspace members very soon, and I’d like to give them the ability to simply replace the foam bases as they deem necessary without burning through what I have left.

Some time back, fellow Makerspace member the_digital_dentist was experimenting with various slurries and sprays to improve part adhesion on his very large scratchbuilt 3D printer.  I gave him a Stratasys foam base to try, and it worked like a champ (almost too well, since the ABS support doesn’t break away nearly as easily as HIPS support).  I mentioned my woes in trying to source more of the foam, and he suggested trying the foil-sided polyisocyanurate (PIR) foam that you can find at some home improvement stores, and gave me a few pieces to try.  I cut a 12″ square, peeled the foil from one side, and sprayed the other side with 3M 77 adhesive before pressing it onto my material testing plate.  I put the tray into the machine and let it heat up before testing it.

Unfortunately, I failed to take differential thermal expansion into consideration.  With the chamber heated to 70 C, the foam expanded ever so slightly, buckling upwards due to the foil still being attached on the underside.

I removed the foil from the underside, and lacking anything better, used a bunch of paperclips around the perimeter to secure the foam.

While the PIR foam surface isn’t as smooth as the Stratasys bases, it’s ‘good enough’ for my needs and a test print went fine.

The part and support removed from the PIR foam very easily with just a little tearing of the substrate.  Overall, it’s still not quite as good as the FR-7104 foam (tears more easily and doesn’t grip as well), but for a cheap, readily available material, it’s the best thing I’ve found so far.  The only drawback is that locally available sheets are only 1″ thick and I need 1.25″ of thickness, but shimming up the base from the bottom shouldn’t be much of an issue.

FCG pocketing with CNC

A long-running project I’ve been tinkering with occasionally over the years has been machining an AR-15 lower receiver from scratch (or more precisely, a raw 0% forging).  I’ve made a fair amount of progress thanks to the excellent video done by Frank Roderus (which is a great machining tutorial even if you have no interest in firearms) as well as the Ray-Vin machining tutorial.  However, it’s really slow going, and I’ve made a few small mistakes.  Completing an 80% lower seemed like a good way to feel like I’ve accomplished something and would be a nice CNC project.

Drilling the holes for the takedown pins, selector, hammer and trigger pins, etc. is pretty straightforward after planing the top surface of the forging.  For an 80% lower, the takedown pin holes are already drilled and you can just use a drilling jig to drill out the holes on a drill press.  Call me a purist, but I still prefer just putting the lower in the mill vise as above, indicating from the front takedown pin hole, and drilling the rest of the holes with positions called out with the DRO.  The jig is invaluable for the various milling operations, though.

While the jig works great on the big manual mill, I can’t exactly clamp it into the diminutive vise on the CNC Taig, so I needed to determine a way to secure the fixture to the Taig table.

The 10-32 tapped holes on the A2Z tooling plate I use are spaced at 1.16″ intervals, and it turns out this spacing could be accommodated by the CNC Gunsmithing jig I use.  I drilled a line of holes for 10-32 clearance on the right jig half, and three 1/4″ holes on the left jig half (I used larger holes on one side to allow for slight differences in receiver thicknesses).  The magwell and trigger guard still stick out from the bottom of the jig, so I use a par of 1″ square bars on either side to raise everything up.

Generating G-code was pretty straightforward – I used HSMExpress to generate a path for the upper ‘shelf’ and a second path for the rest of the pocket.  Yes, I tend to make Z-steps very shallow, but hogging out material with a 3/8″ endmill with a fairly long length of cut is quite a bit for the Taig to handle, so I try to keep all the cutting lightweight, even if it means really long cycle times.  This was my first attempt to use HSMExpress, and I had no idea how well the toolpaths would work, or if the mill would suddenly take off on its own and carve a giant trough through the piece, turning a $70 part into 62 cents of scrap metal.  If only I had some sort of cheap plastic version for proving out the setup…  Quick, to the 3D printer!

I took a standard AR lower CAD model and filled in the FCG area to simulate an 80% lower.  I made sure to print the part with sparse infill to save plastic and time, but I still needed a way to have a ‘solid’ FCG area to provide material for the mill to actually cut.  I cut an opening over the FCG area and mixed up a batch of West System epoxy (I wish every hardware stored carried bulk epoxy supplies – luckily I live near one) loaded with a whole bunch of talcum powder.  I could have used some other filler, but I went with talc specifically to give minimal wear on the cutting tool – many common epoxy fillers can be quite abrasive (to say nothing of the fiberglass, carbon fiber, etc. that generally constitutes the matrix of a composite material layup).  I poured the epoxy/talc mix into the FCG area and let it sit overnight to cure.  In retrospect, I should have tried doing a vapor treatment on the part first, as the epoxy seeped out the sides a bit, but it wasn’t enough to affect the part’s purpose.

After reaming out the takedown pins, I was able to mount the printed 80% lower in the jig bolted to the Taig’s table.  I ran my generated program, and it worked great!  Right up until the very end, when the mill plunged right through the rear of the bolt catch area.  Precisely the sort of thing I wanted to test for, so it was all worthwhile.  I made a few adjustments to the G-code and ran the program through again to make sure that nothing else looked awry.

After that, I switched out the printed plastic lower for my partially completed 0% lower.  This showed me some other issues I hadn’t anticipated, specifically that the endmill liked to bog down in the aluminum.  I hit STOP right away, turned off the spindle power and took a step back to ponder.  The 4-flute endmill was probably not the optimal tool for this pocketing operation, but I had gone with it to increase rigidity due to the long tool length required.  The default toolpaths were also using climb cutting rather than conventional paths – this should result in less power required, but I’ve had bad luck before with climb cutting on the Taig and wanted this pocketing program to ‘just work’ even if sub-optimal.

With those changes made to the program and the purchase of a 2-flute endmill, I felt confident enough to subject a purchased 80% lower to the gauntlet.  One thing that proved very helpful were a pair of 1/4″ ground drill rods from McMaster-Carr that I could slide through the takedown pin holes in the left jig plate all the way through the holes in the lower.  This helped keep everything perfectly aligned while tightening down all the screws, and then I also used the rod to indicate in my X-axis position with an edge finder.  Make sure to remove the rear rod before starting the machining operation, though, as it sits right in the area to be machined.

Gripping a WD-40 sprayer in one hand, and Shop-Vac hose in the other, I exhaled deeply, turned on the spindle, and hit the green Start button in Mach3.  This operation makes a lot of noise, and I worried that something would seize up eventually, but my revised toolpath (I used a slightly different step-down path between Z-levels as well) seemed to do the trick.  When the program finished and the spindle fully retracted, I had no unintended cuts on the part and a nice, bright, shiny FCG pocket.

After washing it off, I installed the FCG components and did a safety check (make sure it will not fire if the safety is engaged, make sure the disconnector is working properly, etc).  Everything functions perfectly!  I’ll do a bit of identification engraving on the lower, and then it’s time to start picking out anodizing colors…

A pair of quick machining projects

It’s been a busy summer/fall/early winter, but I’ve managed to make a few chips in the shop over the past few months.  First up, dad had a front tractor axle that needed a little work.  The hole drilled through it wasn’t quite at the needed 5 degree angle to allow for proper assembly, so he ground out the sleeve that had been welded in and I gave the boring operation a try.

Clamping the part was a little easier than I thought it would be.  It’s a heavy beast, and is as long as the entire mill table.  There’s a pair of blocks near either end that worked great for aligning it on a a horizontal plane, then I just had to clamp the center section in the vise.  I tipped the head forward by 5 degrees to complete the setup.

I seem to have lost the feeler clamp screw for my trusty Blake Co-Ax indicator, so I used an edge finder to pick up the hole center instead.

I didn’t have a boring bar long enough to plunge all the way through, so I bought one just for this project.  I selected a bar tipped with C2 carbide since the cut would be interrupted due to being a slightly different angle than the original hole.  I started by setting the boring head to a small enough diameter that it would just start cutting at the bottom of the hole (where the offset between the existing hole and my new cutting axis is greatest).  I then adjusted the boring head to make the hole a little larger, set the powerfeed, and bored again.  It was long, slow work, but I finally got to the point where I had better than 50% circumference all the way through.  Finally, I turned a matching sleeve on the lathe that dad could weld into the axle hole for the whole assembly to ride on an axis pin.

A friend is assembling a new upper for his AR, and one of the components is a Sentry 7 adjustable gas block.  Unfortunately, the freefloat handguard he’s using makes adjustment of the metering setscrew almost impossible.  Dremeling out a simple slot would fix things just fine, but would look pretty ugly – milling the slot would look much cleaner.

The problem is, how do you clamp a round object while also properly indicating it so the cut is right over the central axis?  After a bit of pondering, I came up with the above solution.  I first drilled and tapped holes in a 1″ square aluminum bar and clamped that into the mill vise.  I used strips of tape on the vise jaws to protect the anodizing on the handguard, and then used a pair of 1/2″ bolts to secure the handguard – tightening the bolts works to wedge the handguard against the top edges of the jaw faces, perfectly centering and aligning the handguard.

At that point, milling the slot (really just removing the web from between two holes) was a piece of cake.