tallest build so far

70mm tall

10mm tall

It's official

                              Mr. welding robot does a better weld then PaPa Steve

Mr. welding robot is gaining on me !

Mr. welding robot is welding better then me ?!?!
Today's machine improvement was to update the wire feeder adjustment pots to 10 turn precession with counting knobs.  

Quite pleasing to see a firm step forward towards acceptable weld quality. 
A thinner then normal build plate is used for this test. Speeds and amp setting are reduced a bit from previous samples. Beginning point for the run is the bottom right corner progressing to the top right.
During the run the three knobs were adjusted by hand / eye as well as adjusting the remote amp control knob for the welding machine. Note the quality of the final portion of the weld !
(weld ran over a random hole on the scrap metal)

Model building coming soon ???

Samples 23-24-25-26

Probably not visible in the pictures ... but ...
Pulsing the wire has upped the quality and control of the weld by a noticeable amount.
Up to this point the welding current has been run high on purpose so there was a nice large molten puddle to hit with the wire. With better wire control the welding amperage can be lowered to the appropriate amount for the material size.
Small changes in amp, wire speed and path speed are defining the end result within 10% of whats expected.
When the end results are within 3% it will be time for the first real model build.

Human Powered Speed Championships

Photo by Tom Amick shows the dozen speedbikes that showed up to race last week.
A new trike record and several personal bests were set.
Rik Houwers becomes the third person to exceed 80 MPH posting a 82.18

Velox IV @ start

Aero Velo @ start

High Desert Sunset

Wire Feeder Updates

A significant motor control update happened this week.
Chatting with one of the members at our local 3D printer Meet Up, Terry Slocum, about the feeder control design goals has yielded fantastic results.
Terry has much experience coding for motion control so he offered to have a look.
A typical, cold wire feeder found it the welding industry would have three control knobs.
One knob for : overall speed, another for time between pulses, and the third for percentage of "ON" time.
Terry added 2 more potentiometers to the circuit and rewrote the code so it does all these things.
Plus he added some lines in the code to facilitate a retraction option.
Thank you Terry !
I'm very much looking forward to trying to weld again. In a week to 10 days from now ...

Next week is the annual Human Powered Speed Challenge @ Battle Mountain Nevada.
The current crop of top machines and athletes are pushing 83 MPH on a flat stretch of roadway.
Truly amazing seeing these self propelled silent speedster wizz by.
Designing, building, racing speed bikes was a 25 year passion I had.
The best sprint speed I could pedal in competition on the same road was 62 MPH.
This year I'm doing track worker duties. Chucking hay bales, manning radios, general grunt work.

Always nice to hear kind words

After seeing a spike in the blog's viewing traffic I tracked back to find where it was coming from and found this article :

Extreme Tech 

Published 4July ... written by Graham Templeton

THX  Graham !

Tool to set tungsten length

This tool sets the "stick out" distance between the end of the electrode and the gas cup.
A simple, quick gauge is made from a coupling nut, bolt and regular nut. I sanded the end of the bolt and coupling nut  to make a good flat surface and then cut a slot in the side of the coupling nut for visibility. Dial calipers are used to set the distance of 6mm.
Before starting a build cycle a fresh tungsten electrode is installed. This ensures a stable current flow directly off the tip. The electrode is a consumable and should be inspected and cleaned often.
The wizardry in getting this welding process to function is the metal feed wire needs to hit perfectly in the center of a molten puddle of metal ... for many, many hours !
A pending design improvement is a vernier adjuster for the feed wire angle.

It's getting better all the time

Weld sample # 17
Resolution is starting to improve. With better control of the process it's possible to turn the heat and wire speed down which is improving the weld quality and constancy.
This is 5 layers.
Finish height measures 7.4mm ... yet ... should be 6mm.
Some tweaking is needed.

Finally ! ! ! Welding again ! ! !

16 May was the last major welding session. Back then I was trying to knock out a decent piece to display at the Makers Faire.
Today is going much better. This is my second attempt after 3 months of mods and updates to the machine.
Definitely in the ballpark now. The stack height measures 20% too high after five layers. The weld is decent and complete.

Wire feeder is working

A few different brands of stepper drivers were purchased for evaluation. All have the same general output ratings. I decided to try the tb6560 first simply because it was the largest physically sized board thus has the appearance of delivering the most power.
It hooks up easy enough.
The first good sign is the board and driver run in the proper temperature range and within a few degrees of each other. Pulling power on the wire is much stronger now. With the original configuration I could squeeze my fingers pinching the wire coming off the spool and stop the motor. Now I can't.
A special THX to John Saunders @ nyccnc for his youtube video ... a big help in figuring out the UNO to tb6500 schematic and code.

Here's the code and driver switch settings for this round of testing.

/*
Code: 10K ohm Potentiometer speed control
arduino UNO to microstep driver tb6560
CLK+ / step / pulse+  wire to arduino pin 9
CLW- / step / pulse-  wire to ground
CW+ / direction+  wire to arduino pin 8
CW- / direction-  wire to ground
potentiometer end wire to arduino +5v
potentiometer end wire to arduino ground
Potentiometer center wire to arduino A0
enable+ to nothing
enable- to ground
*/

int sensorPin = A0;
int sensorValue =0;

void setup() {
  pinMode(8, OUTPUT);  //direction pin
  pinMode(9, OUTPUT);  //step pin
  digitalWrite(8, LOW);
  digitalWrite(9, LOW);
}

void loop() {
  sensorValue =analogRead(sensorPin);
  sensorValue= map(sensorValue,0,1023,3600,1);
  digitalWrite(9,HIGH);
  delayMicroseconds(sensorValue);
  digitalWrite(9,LOW);
  delayMicroseconds(sensorValue);
}

Switch settings for tb6560 stepper driver
SW 1 = ON
SW2 = OFF
SW3 = ON
S1 = OFF
S2 = ON
S3 = ON
S4 = OFF
S5 = ON
S6 = ON

Tidying up a bit

While waiting for the drive board to arrive the build plate was improved by addition of leveling screws and hold down clamps.
Even though this is a welding machine having a wood base has proven to be a good idea.
The wood is a means to easily attach a thin sheet of metal to complete the weld circuit. Long wood screws are used as the temporary hold down.
This metal sheet can be removed from the model after the build is completed.
A section of a recycled oak table top is cut to size and then is covered with an aerogel blanket to prevent fire. Next up is a layer of fire bricks with the metal sheet topping the stack.

Testy Testing

Having a speed control for the wire feeder is looking promising.
Although ...
The motor driver is not quite sorted out yet. The chip is overheating while the motor is stalling trying to push the wire and staying at ambient temperature. i.e. the system is not balanced. The same motor had enough power and ran at proper temp when driven by the TinyG board.
A tb6560 chip is the next experiment. It is a common part in many stepper drivers. There is a plethora of pre-made drivers with this chip so there is no reason for me to build a board from scratch.
The biggest cost is the time delay waiting for parts to arrive.

Using CAM software to create an additive tool path is working well for simple objects. The lines of code is about one tenth of a typical model slicer output. This is due to the fact this CAM software can write code for G03 output (circular interpretation). I haven't seen any of the 3D model slicers use this shortcut. A typical slicer cuts the model into a series of lines so something like a small circle can have a big bunch of line segments which can often be seen in how smooth the tool moves.

"d'oh!" ... A Homer moment

O​h man am I embarrassed.
Before hooking up the wire feeder to the main control board I thought it prudent to check for voltage output at the terminal. I started checking for voltage by probing the top of the terminal screw with no wire attached and found a dead circuit ... or so it seemed. Many hours were were spent looking over the firmware code, investigating the schematics and tracking the circuits on the board to no avail. Everything looked correct but still no voltage to the top of the screw.
While out for a walk it occurred to me to install a short length of wire into the terminal and then clip a volt meter to the wire.
Suddenly there is proper voltage at the coolant pin ???
It turns out that when the screw is backed out and no wire is in place the top of the screw is not conductive. A "gotcha" that wasted at least 10 hours.
 It seems odd that the top of the metal screw has no conductivity without pinching a wire but that was exactly the problem.
The wire feeder is now hooked up and turns on / off as it should.

Now back to trying to weld.

No cigar ...

After all the effort of creating a separate speed control for the wire feeder ...
Now the TinyG motion control board is not activating a command to start the motor.
The strategy was to use the coolant on / off pin to run this circuit.
Commands M7 / M9 should be switching on / off  3 volts at the pin. 
G code commands M 3 / M5 are used to switch the welder on / off  with the spindle pin on the board.
The welder turns on / off as expected.
When the wire feeder is connected to the spindle pin all is fine.
Hooking two circuits to one pin out didn't seem like a good idea ... but ... I tried it anyways.
Both circuits switch on but won't switch off.
Operating both circuits from one pin isn't going to work.

A solution ? My first guess ... a firmware change or upgrade will need to happen.

Happy to say the feeder drive board has evolved away from bread board testing to a soldered connection prototype board. Also changed the potentiometer to a 10 turn precision version for finer  speed tuning. 

Bench Test Power Supply

As the control electronics evolve so do the voltage requirements supplying each board.
TinyG is 24V, Arduino is 12V and the H-bridge chip is wanting 5V.
The work bench is getting messy with all the various hook up wires and alligator clips that keep slipping off while doing a test.
so ...
I've seen PC power supplies being being converted to a bench test power supply. I hacked this one with a little guidance from wikiHow.
http://www.wikihow.com/Convert-a-Computer-ATX-Power-Supply-to-a-Lab-Power-Supply

The gears are turning ...

The current design task is to adjust wire deposition rate with the turn of a knob.
After getting stuck trying to figure out Arduino stuff I was able to bounce questions off  Jonathan Ward at OtherMill an Zach Radding at zachraddingdesigns to good results. The motor runs and is speed adjustable.
Yesterday the motor was bench tested with no load for several hours and no issues. Microchips and motor temperatures were on target.
Today I'm trying to test the full load of the wire feeder and push some wire. Unfortunately something in the system is not happy and lays down for a nap after running just a few minutes. It appears the H-bridge chip taking a temperature spike causing a shut down. Allowed to cool a few minutes and away we go again. First impression for the cause of the problem is the input power to the H-bridge chip is not regulated.
More testing needed ...
Update
The motor RPM has been slowed to the expected inch per minute wire speed which has lowered the chip temperature.
With this change the system has run for 3 hours with no problems.
This chip is a little too close to heat failure so when doing the final soldering a second chip will be stacked parallel to add a measure of reliability. On future builds a larger chip would be better ...

The Arduino code that works :

#include <Stepper.h>

// change this to the number of steps on your motor
#define STEPS 200

// create an instance of the stepper class, specifying
// the number of steps of the motor and the pins it's
// attached to
Stepper stepper(STEPS, 8, 9, 10, 11);

//Create starting position for motor / pot
int previous = 0;

void setup()
{
 Serial.begin(9600);
 //How fast will we try to move the motor
 //If your motor stutters, its too fast so just lower the value
 stepper.setSpeed(30);
}

void loop()
{
 //Read analog 0 (pot) and map it to the range of the motor
 int val = map(analogRead(0), 0, 1023, 30, 0);
 Serial.print(val, DEC);
 stepper.setSpeed(val);
 stepper.step(5);

}

Solidoodle calibration

This early model Solidoodle V2 was purchased from a Kickstarter campaign a couple years ago. It has been attached to several different computers but has never produced exceptional quality prints.
I decided that with this latest computer set-up to spend the time to sort the parameters all the way thru to achieve the best possible results. 
When purchased the Solidoodle came with PLA filament so that set the template of use. The general defaults were used to produce OK prints.
The last time we tried to use the printer it was switched to ABS filament with very little success. Mostly because we couldn't run the temperature the material wanted. Scouring the internet I learned this early drive board has overheating protection written into the firmware that prevents it from running the temps ABS filament needs.
SO, then, to run ABS the drive board will need a cooling fan and an firmware change.
OK, then
How about refining the PLA settings ?
Easy enough ... oh sure ...
With PLA the machine prints the perimeters well but the infill and support just didn't work no matter what speed or heat setting is tried. The infill is a mess and this effects the top layer quality.
After about 20 tries of printing calibration cubes I finally hit the internet to understand what's going wrong. Turns out PLA needs a cooling fan which is absent from this machine.
So a small fan was attached to the print head by using some hardware wire. Power to the fan comes from an open 12V port on the board. Not the best fix as the fan runs continuously. This band-aid approach help print quality tremendously.
Done properly the slicing software can output different cooling speeds depending on what pat of the model the machine is printing. A different speed is used for perimeters, infill, support, bridges etc.
This Solidoodle drive board doesn't have the capability to turn the fan ON, OFF or change speed.
The board can be modified with some soldering and firmware changes.
Gosh ... what a couple days of lessons that was ...
Now I know all the major shortcomings of the early Solidoodle printers.

Projects with wood

Lately  I’ve been feeling compelled to expand my artistic horizons with a few wood projects.

This is one of them.

The cutting board backstory.

Start of the idea ...

On the drive home from visiting mom I stopped at an intriguing wood fantasy castle in Kirby Oregon called the Burl gallery. Actually a gathering of wood artisans who have made an awesome roadside attraction.

One artist is doing fine cabinet work featuring dovetail joints as the cornerstone of his artist expression. Mostly a build to order craftsman he also had a collection of cutting boards made from the scraps generated from the cabinet work. The tourist trinkets for the walk-ins like me.

The artist sales pitch for his superior cutting board was how the dovetail piece keeps the board flat and solid making a longer lasting product. Something the engineer side of me can cherish.

His dovetails were done as “legs” that protruded off the bottom so the boards are “single sided”.

Second part of the idea ...

At home, my spouse, Chaney has a bamboo cutting board that she wrote “meat” on one side and “vegi” on the other with a black permanent marker. This made hers “double sided”.

Third part of the idea, materials …

While out on a morning walk with the dogs we came upon an old dining room table busted down to fit into a dumpster. Being in decent condition and solid oak construction it looked too nice to discard. This re-purposed wood comprises the main board. The materials used for dovetail come from the front lobby of the 140 year old Chronicle building in San Francisco. I only saved a dozen boards from a large pile. The grain is very tight, strong and straight.

My part of the idea …

A drip rail that morphs into a deeper well to capture the spilled liquids. 

And some engraving. One side is engraved "carnivore" the other side is "herbivore"

The finish coating is olive oil which should be reapplied every now and then.

 The CAD software I like for a project like this also has a powerful CAM package. By taking the time to do this project I discovered a way to reconfigure subtractive machining software to output additive control. A new lesson to be put to use soon.

Wire feeder control

The current task for the welder project is to separate the control of the extruder (wire feeder) from the software so that the speed can be controlled from an analog potentiometer.
The idea is to have the slicer software output on or off commands to the wire feeder and then I can adjust the speed of the wire by eye and feel. A little bit of an "old school" hands on fix. The long view advantage to this strategy will be the ability to pulse the wire.
I'm trying to use an Arduino for the task. The programming language is not intuitive to me yet so there's a bunch of head scratching going on. 

Image courtesy of Arduino Knowledge Base

Remote Amperage Control

My schedule has opened enough to make some progress on the robot welder. Yeah !
This remote amperage control will allow subtle power adjustments during the welding.
It plugs into the port normally used for the foot pedal remote.
Hopefully it will be one of those "small details" that makes a difference in weld quality.
Finding a suitable large enough potentiometer to match the needs of this old welding machine turned into a bit a treasure hunt. A modern welding machine won't need one so large.

Other chores beckon

The northern California drought conditions has put me in the tree trimming and landscape clean up mode. Several days of removing all the kindling around the house and shop.
Also took some time to make a wooden Porsche for my brother's birthday present.
Porsche released this Cayman model as a .stl file earlier this year. It would have been easy to have simply made it with one of plastic printers in the shop ... Chris is house builder so he will appreciate the wood. An interesting plus is the wood came from his property. He has a collection of about 50 Porsche models. They are all plastic or metal ... no wood ones in his collection. This will be something new.
It also provides a fun project on the CNC machine. Machining a 4 sided model on a 3 axis tool takes a bit of planning.

The 3D welder project starts up again next week. 

Welding Tests

The Maker Faire fire marshal had approved this project to run welds during the weekend fair. So in the days leading up to the event a focused effort was made to get the details sorted out. Unfortunately the effort spotlighted all the items that need to be sorted out. With such poor results I decided not to run the machine in public. The set up time would have far exceeded any useable demonstration time.

The numbers on the sample correspond with the attached spread sheet.
In most cases the run was stopped before the completion of the second layer so a measure could be taken of the stack height of layer one and layer two and then compared to the software request. 

The machine change list :

• Bed levelers
• move “Y” pulleys out 1mm
• temp probe motors
• volt / amp meter at welder
• micro adjuster for “Z” axis height
• micro adjuster for nozzle angle
• ceramic nozzle
• try different drive board
• slicing software support
• change torch body
• electronics enclosure mount
• temp probe model
• change electronics enclosure
• remote amp adjustment

Fifth time is the charm

This is the fifth configuration change for the "Z" axis limit switch.
During the last round of weld testing it was discovered that the distance of the starting gap between the tungsten electrode and the build plate is crucial.
The initial size of the "air gap" was vital to how the first layer formed.
As is well known with all other types of 3D printing if the first layer goes wrong the rest of the print goes wrong.
As weld testing progressed it was found the "sweet spot" for the air gap to be plus / minus .25mm (.010")
The coarse adjustment was revision 4
Fine adjustment is revision 5

Internet media exposure from the Maker Faire - 2014 San Matteo

a nicely done short article
http://build-beta.slashdot.org/story/14/06/18/1853208/3-d-printing-with-molten-steel-video
the video that goes with the article
http://tv.slashdot.org/video/?embed=01b3Jjbjq_5YhB6ipoWGYtORDcJ7w5OO

Popular Science picked up from Slashdot to publish here :
Popular Science

Summing up the methods it took to solve the transient voltage problem 

1. All wires are shielded and terminated to common ground. This combines the ground between welder and robot.
2. All electric lines follow a parallel path. An example would be having the welding return follow the torch input line. In all my years of welding this is the first time this technique has been needed. 
3. To be used in industrial environment the current generation of low voltage motion control drive boards for 3D printers will need some form of protection circuitry between the board and motors and/or limit switches. Preferably as close to the board as possible. In my case there was a direct path from the limit switches to the processor chip. The surge at welder start up created a voltage spike in the switch circuit and caused the processor to reset. 

Shielding the torch lines.
The idea was originally to try and control the transient voltage leaking from the torch input and use the shield to ground the machine movements to the welder chassis.

This photo shows how to use a screw driver to open an entry into the braid.

Torch with shield added.

When my electrical engineer friend Geoff Dolan saw the machine he suggested using the shield as return path from the build plate to the welder. This strategy keeps the return path parallel to the input and as short as possible. A great suggestion although I had to resize the shield to handle the amp load. 

The transient voltage problem seems to be resolved.
The 2 major updates to the machine are :
Implemented the circuit protection scheme David McQuate suggested.

A grounding buss is added so it is easy to implement all components to a common ground.

One of the motivators
driving the creation of this project is to create an affordable welding robot that can be taught to weld better then me. A machine that is happy to weld hour upon hour and only asking to be fed in small quantities of electricity and production consumables. A machine that won't be distracted the myriad of fun stuff a human would rather engage in every second of every day.

This belt buckle was recently rediscovered in the closet. It is something I hand welded at a NASA launch facility 35 years ago. The "Scout" missile was the smallest and most reliable satellite launching system making it the perfect delivery vehicle for the original constellation of GPS satellites.

Maico is a German off-road motorcycle. The base material is .062" thick 304 Stainless steel with .062" filler rod.
If I were to create another today I'm confident the craftsmanship and welds would look similar.

A transient voltage problem
created by the welder arc start cycle has been slowing the machine build progress.  I know there is a solution but not being strong in electrical engineering has left me poking in the dark for answers. One of my Linkedin connections, Michael Fitzpatrick suggested contacting his friend David McQuate at Clear Stream Technologies for assistance.
David rode across town yesterday on his Easy Racer recumbent bike to examine the machine.
After studying the driver board  schematic and the machine he made several suggestions.
He suggested adding this circuit to the limit switches to stop the voltage intrusion.

Each limit switch will need the addition.
Hopefully this will turn a corner in the design and the pace will pick up.

 

Electrical Enclosure Change

When showing the machine at Makers Faire and discussing the electrical interference problems a common suggestion is the original electrical enclosure didn't look correct. Too many large holes in the sides. Being made from aluminum probably isn't helping either. So this steel junction box was sourced with a trip to the local Habitat for Humanity ReStore.  $10 for a decent used box for further testing. First test shows it does seem to solve the start voltage jumping across the limit switch wires but not the start voltage jumping across the drive motor wires. Next experiment is to add shields to the drive board power feed lines and also to the USB cable connecting the CAM software on the laptop.

Maker Faire 17 ~ 18 May

This past weekend Maker Faire graciously offered space to show the molten3D printer.
What a crowd ! Fortunately we had a space in the far south east corner of the fair grounds where the flow  of people was slower and more intimate. I was able to discuss with several people the successes and future challenges of the machine. These valuable conversations will help solve a few of the bugs still lingering in the design.
Thanks to all those who stopped by.
Thanks to the Maker Faire for the experience.

   

Molten 3D welding printer

Beginnings of the CAD model
August 2013 

General look of the machine

3 D printed plastic test parts for the movements

Checking the fit and accuracy

tacking the main frame

movement parts done in steel

capstan pulleys for the X,Y,Z drives

drive assembly

first movements test

electronics enclosure flattened for cutting

CAD for wire feeder assembly

wire feeder fit check

electronics assembly

torch and wire nozzle attachments

first sliced model movement tests

First welding tests 1 April 2014

Looking to solve a high frequency cross over problem

Starting to add shielding on all motion control cabling