Thread Tools
Jul 03, 2018, 07:30 AM
Callused Thumbs
rafterrc's Avatar
Thread OP
Originally Posted by tom43004
I always cooled my g540. It ran in some pretty ridiculously warm conditions over the years. I had mine in an enclosure with power supply, E-stop, and a couple of 80mm biscuit fans.

Heat definitely is the enemy of good shop work. I would plan to at least have a fan on the 540 and power supply pretty much at all times.

FWIW, my G540 ran 3.1a Nema 23s (4) from a switching 48v power supply... and ran flawlessly for several years. I have since traded up to a larger machine with Nema 34s that have more torque than my Highlander.
If it weren't for my smartphone I would have no internet. A wildfire nipped the fiber optic serving our valley. More details when I don't have to type with my thumbs

The temp issue is with the 540. I placed it on an ice bag and it works flawlessly in the heat of the day.
Sign up now
to remove ads between posts
Aug 10, 2018, 06:52 PM
Callused Thumbs
rafterrc's Avatar
Thread OP

Looks Like a Faulty G540 Motherboard

You can watch this video:
Geckodrive G540 In-The-Field Troubleshooting (12 min 11 sec)
for guidance on Geckodrive G540 troubleshooting and G250xr3 replacement. I finally got some time to troubleshoot mine. I did exactly what the video shows on my apparently heat sensitive G540. The results: I get a green light when the motherboard is disconnected from the drivers (G250xr3s). There are no short circuits in the 4 drivers either. Yet the G540 kept failing at around 45C after ~10 minutes of operation.

I am running with Machinekit, so a .hal file is required to identify the configuration. I used XYYZ for the WorkBee gantry. Actually, it is X, YL, YR, and Z. It fails on the YR channel. The first test was to change the .hal file to use XYZY. This time the WorkBee failed in the Z axis, so the driver in the 3rd position was suspect. The next test was to swap the 3rd and 4th position drivers. The failure remained in the 3rd position. So the drivers are all good, and the motherboard is suspect.

As you can check above, the G540 was placed on a towel wrapped blue "ice" bag. At 10% humidity, there's not much condensation, but it was a precaution. It operated flawlessly for over an hour. So the problem is not upstream of the G540 motherboard, ie the MESA 5i25 or the Machinekit software.

I have contacted Geckodrive regarding this situation and given them a report on the troubleshooting. I'll report back on their response.
Sep 01, 2018, 05:15 PM
Callused Thumbs
rafterrc's Avatar
Thread OP

End of the Temp Sensitivity Saga

Or at least it seems to be.

I contacted Geckodrive via email with the observations and measurements above. Marcus at Geckodrive and I concluded to send the G540 in for testing. He ran a 4 hour stress test consisting of a four hour 3.5 A on the four axes simultaneously without any rest for any axis. My G540 passed.

So I sent that info on to Peter at Mesa Electronics. since the 5i25 is the next commonality. He asked what I was using for step time. The 5i25 provides the step generation for the machine. Step length is the length of time of high voltage in LinuxCNC and Machinekit, Step space is the length of time of low voltage. I said that I was using 1500 nanoseconds for step length /time and 2500 ns for step space. Peter said to double it. Marcus said that 1500 was 500 ns low. So it seems the G540 needs something in the range of 2000 - 3000 ns for step length/time.

I received the tested G540 back today and hooked it up. I edited the .ini input file to change the step length to 3000 ns. Then I ran the thing gcode programs that caused the Y right circuit to fail. In between, I jogged the Y axis at 8000mm/sec nearly full length 5 times. Previously, this would have caused a failure, but not with the longer step length.

It would be interesting to look at the wave form on an oscilliscope, I'm guessing that the G540 could not resolve the YR input as a recognizable signal at operating temp. Essentially, the square wave wan't square enough.
Sep 03, 2018, 03:58 PM
Callused Thumbs
rafterrc's Avatar
Thread OP

"Heat " Problem

Definite End!

I just successfully completed a 3hr project! Step time was increased to 3000 ns. Temperatures across the G540 were consistent with no local hotspots.

Cutting feed rate was set to 3048 mm/min (120 in/min). Motor temps were in the low 50's C. It looks like 3810 mm/min is achievable, but clamping may be the issue. This is for MDF and an 1/8" square end mill.
Nov 13, 2018, 10:09 PM
Callused Thumbs
rafterrc's Avatar
Thread OP


mack5, I cannot reply to your private message because your box is full. So here is my reply:

Originally Posted by mack5
what machine do you use? will it cue balsa and thin plywood? I am looking to buy one soon.

I built the mechanical WorkBee 1510 (Found here: For control, I use LinuxCNC, MESA Electronics 5i25, and Geckodrive G540. It is smooth. I've cut fuselage molds from MDF, white oak for drawer fronts, and chair leg backs with a 500w spindle.

I am pleased with the purchase. I will be cutting foam cores for wings soon.
Dec 15, 2018, 10:10 PM
Registered User
Here is a pic of the machine I built from scratch several years ago. Extremely versatile and built like a tank. NEMA 34 motors, Gecko 203V drives and the Mesa Electronics 5i25 interface. Use it for all kinds of things from boat hulls to carbon fiber frame pieces. Linux CNC controls it all and total cost was around $2k. All made from 3/4 MDO.

Uses chain drive for X and Y axis. The X axis has v bearings and aluminum angle for support. The Y and Z axis were upgraded to SBR16 linear rails a few years ago. I also added a 2:1 belt reduction on the X and Y axis which improves torque and resolution. Rapid moves are at 340ipm. Most cutting is done at 70ipm or less.

Amazingly, my spindle, a regular Bosch wood router is still running strong. Thought for sure the brushes or bearings would have worn out by now. I also added a precision collet which allows for easy bit swaps.
Jan 02, 2019, 09:42 PM
Ok that's high enough
FabFlight's Avatar
Have you cut foam cores yet?

How do you plan to proceed to machine both sides?
Jan 03, 2019, 12:41 AM
I can fix that ...
scottsdalejohn's Avatar
Attached is a photo I saved that shows the foam was cut on both sides. I did not bookmark the page, but I think that it was from the Ultima F5J.
Jan 03, 2019, 11:06 AM
Ok that's high enough
FabFlight's Avatar
I remember seeing this picture somentime ago, thanks for retreiving.

I presume all these holes are registration that have been machined on the first side, and each probably match a registration pin on a fenale mold used to hold the foam in place when doing the second pass. Perhaps the foam is also vacuum held onto the machine bed/female mold.

PS Even for foam, it must take a good while to machine to this finishing.
Last edited by FabFlight; Jan 03, 2019 at 11:13 AM.
Sep 05, 2019, 06:59 PM
Ok that's high enough
FabFlight's Avatar
I'm itching to buy one. Any update?

The published accuracy (0.1 to 0.2 mm) is a big warning to me. If it's not better than 0.1mm I'm afraid it wouldn't be good to prepare molds (Corian) or even Rohacell cores.
Sep 05, 2019, 10:01 PM
Callused Thumbs
rafterrc's Avatar
Thread OP
I tested the WorkBee with a 500W spindle capable of 12Krpm by cutting some hyperbolic shaped chair back legs from quartersawn white oak. The oak is nowhere near as homogenous as either Corian or Aluminum. I had to slow the feed rate to cope with vertical grain. I am confident that the combination will work for both Corian and Al.

Accuracy: The 40 inch legs' (8 laminated from 16 pieces), as cut, dimensional differences could not be detected by feel, but they are not sailplane wings.

Molds for full sized sailplane wings are completed by, yup, hand. To expect a cylindrical endmill to produce a result less than 0.1 mm across 4 meters is unrealistic. If you want to accomplish such then you will have to do it by hand, too. Get your measurement system ready, then figure out how to hand finish what the CNC router started. You will have to fill in too deep areas as well as how to sneak up on the correct dimensions. There are a handful of such artists in the full size world.
Sep 05, 2019, 10:09 PM
Callused Thumbs
rafterrc's Avatar
Thread OP
Still I recommend a.spindle of greater horsepower and higher rpms for the WorkBee. Say 1.5Kw and in the range of 18000 or more with the lower rpms nearer 5Krpm. You should be able to handle Al to balsa with that kind of range.
Sep 14, 2019, 10:11 AM
Callused Thumbs
rafterrc's Avatar
Thread OP
Yesterday, I dug out the profiles that I made by hand along with the MDF wing sections that I made by hand for the SynerJ. I put the corresponding ones in the cnc router cut wing tip molds. Some were very close fits for the full chord. Several had high spots around the max wing thickness position. All were less than adequate fits near the LE.

Is there an easier way? One laser 3D scanner claims accuracy to 0.1mm, and it costs approximarely $3K. Not precise enough. More precision needs more money. The organizations that possess this ability are reluctant to turn the untrained loose with their scanners. Not particularly good for hand finishing feedback.

For the Z4Davg fin, I am 3D printing a nylon version in white. The purpose is to place some marking ink on the mold and press the printed positive into it. If all goes like it does on gunstock work and dental tooth fittng, then I will have pictures. This approach is like that scrapers use to flatten steel surfaces by marking using reference surfaces.
Last edited by rafterrc; Sep 14, 2019 at 10:29 AM.
Sep 14, 2019, 12:36 PM
Registered User

Another way ....

WIP: first generation of DIY SLA 3D printers (for 405nm sensitive resin with laser).
What you can see here are the X and Y axis for the prototypes ( I hope to finish the build for Z axis next week).
Mechanics is very simple: MGN12 linear rails for all XYZ axis, movement by T8 driven by NEMA17 and NEMA14 steppers.
Printer is a duplex (for Y axis movement is mirrored so I can print the left and right segment in the same time ... in fact I can say it's a "quad" printer, on the same bracket I will mount 2 laser heads side-by-side to print 2 segments in the same time).
The vat allows to print 45 cm segments, one frame (like the one in the picture) has 2 printers (with a total of 4 vats). 2 frames like this and I can print the 8 x 45 cm segments for the whole wing, another frame smaller is for stabs. These are what I call "Mark1". Frame is made with 2020 aluminium profile, what you see in picture is 60x50x30 cm.
Printers "Mark 2" are made with 3030 aluminium profile with the same XY mechanics but with a 1 m height for Z axis. This will allow printing 95 cm height segments and also molds & plugs. 2 printers (with a total of 4 vats) will be on the same frame (116x116x33cm).
This frame will contain also several (4 or 6... I haven't made up my mind yet) smaller printers for the fuselage. I will try also "Mark 2bis" (same type of XY axis but Z axis made with MGN9 linear rail) and "Mark 3" (for XY laser beam moved with molibden mirrors driven by 0.9 degrees steppers with 1:256 micro stepping, Z axis also MGN9).
All vats will contain glycerine with a 5 cm layer of resin on top. No 3D printed parts, everything is done with cheap chinese components.
Software to generate G-code is also DIY (I've started a topic "Project ALTius" where I give dsome etails about it).
If you want to build a simple printer (not dual duplex printers like in the picture) probably the cost will be around 200 -250 USD ( 40 USD for X axis, 40 USD for Y axis, 50 USD for 50 cm or 60 USD for 1m Z axis, 20 USD for laser head, 40 USD for electronics , 20 USD for vat... I can post a BOM later in my topic).
With a printer 30 x 5 x 95 cm you can print in fact everything ("cores" & "molds" for wing and stabs and fuselage). In Europe you can buy 405nm resin with 40-50 USD / liter, probably it will take under 1 liter or resin for wing / stans / fuselage and 2-3-4-5 liters for molds.
Well, that's the plan anyway, a rigid structure capable to print with a decent compromise between speed and precision the complex 3D strucure of the wing (my plan is to print one wing per day).
And of couse some nice 80 gsm textreme to fit between the resin "core" and "mold"...
Last edited by ender67; Sep 14, 2019 at 12:44 PM.

Quick Reply

Thread Tools

Similar Threads
Category Thread Thread Starter Forum Replies Last Post
Question CNC tolerances for building molds Tim Harbour Hand Launch 18 Jul 24, 2017 05:35 PM
Mini-HowTo CNC Solid Molded Construction- Part Three - Cutting the top surface of the foam tom43004 Composites Fabrication 18 Oct 05, 2012 11:29 AM
Mini-HowTo CNC Solid Molded Construction- Part One - Creating the mold tom43004 Composites Fabrication 27 Jun 17, 2012 08:59 AM
Mini-HowTo CNC Solid Molded Construction- Part Two - Creating the vacuum hold down tool tom43004 Composites Fabrication 8 Apr 23, 2012 10:21 AM
Question Best mold construction method for Painting in the mold ScorpionRacing Composites Fabrication 3 Jan 14, 2011 03:28 PM