Banding in glowforge engraves

I’m also thinking this is the case. Belt drive with a cutting head of virtually no load (unlike the load forces of a CNC router) not going to create any level of measurable backlash.

From my limited understanding and use of bi-polar stepper motors, if the motor were to “lose” a step or two, the rotor would be out of phase with the pulses that drive it and the motor would just sit there and jitter (very unsettling to see your motor doing that). I’ve experienced this when I drive a bi-polar stepper motor past its max speed capability as an example.

Different issue though if the software is somehow missing a step or two between it sending a command to increment a step(s) and the controller sending the command to the stepper motor. The software has to check the controller to see if it has issued pulses to move the requested number of steps. On the firmware on the bipolar steppers motor controllers I have used, the controller updates its “current position” register every certain number of milliseconds (in the case of my controllers, 8 ms). This can be an issue for directional changes as the app software can do a lot of things in that 8ms between when the controller reaches the destination step and when it updates its “current position” register for the app software to obtain. Well, it was a problem for me until I figured out what was happening and added a few lines of code to compensate for it.

There is no “bio feedback” between the controller and the motor to know where the motor actually is, the controller only knows where it told the stepper to go.

Geeze that sound confusing to me as I read it. . .

1 Like

All you need to produce backlash with a stepper motor is friction. The torque it produces is proportional to the displacement from its target position. I.e. at the target position it is zero, so it can stop short or overshoot with a small displacement giving a torque equal to the static friction. Since you cant have zero friction there will be some finite backlash, hopefully very small in a well designed system.

Also it is easy to get backlash with belts and pulleys if the tooth profile is wrong.

1 Like

Can someone with a K40 and someone with an Epilogue post close ups of raster engraving at different resolutions to demo some of the things we are discussing here? Something that demonstrates the state of the art so to speak.

3 Likes

Hopefully they are doing some documentation videos as they work. Even if they don’t release them until shipping starts, or until the first chinese clone hits the market. The ad revenue from everyone pouring over the details we weren’t privy to after the fact cannot be insubstantial.

The lowest my Trotec will let me go is 500 LPI. If I enter “1” into the field it automatically sets it to 500, but it seems like it’ll take anything higher than 500. For instance, 501 is apparently fine. I can do the test at 500 if you want.

Here’s something I engraved at 1000 LPI a while ago. I’m just posting this since I already had the image uploaded to the forum…



The coin in the second picture is a US dime (diameter: 17.91mm).

6 Likes

Regarding what is going on in low-res mode, here is a sample of what the top of a “2” looks like at various DPI (72, 300, 600, 1000) when magnified (with no anti-aliasing). This could probably be another thread in itself - but the first thing that comes to mind is CICO - crap in, crap out. Don’t expect magic from files that are low-res to start with.

72DPI

300DPI

600DPI

1000DPI

8 Likes

The more I think about it the more convinced I become that the banding is from backlash.

(as long as it isn’t a crap in crap out situation like @jbmanning5 showed)

The alternating overshoot direction is pretty easy to see in the name Allen in the photo posted above by @davidhknowles. Backlash will make it look like the laser is firing slightly earlier than it’s supposed to. That being the case, I’m guessing the top line of the “g” in “glowforge” was engraved moving from left to right. Is there a video of that engrave somewhere?

There is evidence of backlash in the wavy lines around the outside of the Celtic knot ornament(?) as well. As @GrooveStranger pointed out, the spacing is not even. If you watch the video you’ll see that the first wavy line that is scored the the outermost one and they work their way into the center. The first line is scored counterclockwise, then it changes direction and scores the second clockwise. It changes direction again, back to counterclockwise, to score the third line but then it scores the fourth line without changing direction. You’ll note that the spacing between the third and fourth lines is basically perfectly uniform.

I can’t quite tell, but I would bet that the spacing between the first and third lines is also pretty perfect, since those two lines were also scored in the same counterclockwise direction. So, if it wasn’t for line number two being scored clockwise, I think the wavy lines would have looked totally uniform.

Thankfully, mitigating backlash in hardware is pretty straightforward.

According to the specs: “Positioning precision to 0.001” (0.025mm)”, which is about an order of magnitude less than the kerf. I took that to mean positioning accuracy but I am beginning to think it is just the resolution if backlash is visible. So what is the positioning accuracy? This may be a show stopper for me.

I took that to mean positional accuracy down to 0.025mm as well. Personally, I think they’ll fix it before release.

Fix the spec or fix the mechanics?

I don’t think 0.025mm accuracy is achievable with belts and V rollers unless it has closed loop control.

.001" is both the inches-per-line and the positional accuracy.

–dan

4 Likes

Really? So you have a closed loop control system? Open loop stepper motors and belts would be nowhere near that accurate. Even the thermal expansion of steel would give a bigger error than that over 500mm with a 3C temperature change.

You certainly don’t need closed loop to have 1-thousandth accuracy. How big are you imagining the pulleys to be and how many pulses per revolution do you think they’ll be limited to?

I think you are confusing resolution and accuracy. Yes it is easy to get that resolution, for example with 20 tooth GT2 pulleys you get 40mm per revolution. A 200 step motor with 8x microstepping would give a resolution of 0.025mm. However the accuracy of a typical stepper motor is +/-5% of a full step. That would make any particular position vary by +/- 0.01mm from its correct place to start with. Any friction would cause some backlash reducing the accuracy and so would thermal expansion of the belts as the distance travelled depends on the belt pitch which would change if they were stretched on a steel chassis.

1 Like

I don’t feel like having this conversation. Sure, any system that does not compensate for thermal expansion will lose accuracy when the temperature changes. The motors (steppers or servos), linear bearings (V groove, rail, rod, ways, etc.), and method of motion transmission used (belts, screws, rack and pinion, anything else) is completely irrelevant to this fact.

I’ll concede that I do not think parts will be coming off of Glowforges with an accuracy down to 0.001". How would you feel if the claimed positional accuracy was ±0.002"?

Comparing the top of the cut to the bottom will probably reveal a discrepancy of a thousandth. In fact, a taper of just 1º in 1/8" material will be off by ~2 thousandths. No need to look any further than that.

Yes ±0.002" would be OK but not what I paid for. However the photo @takitus posted looks like the machine has alignment errors about 3 times the kerf when engraving.

@dan, are you saying the engraving errors are not down to positioning inaccuracy?

1 Like

While it is an internal linear actuator, one of the bipolar steppers I use in my CNC coil winder has a resolution of .00012" per step. So on the mechanical side of a stepper resolution of 1 thou’ shouldn’t be out of the range of possibilities. Can’t comment on the belt drive aspect.

1 Like

All I know is, the apparent positioning issues in the image above are concerning. At about 2 months before shipping I would expect fundamental issues like cut positioning to be put to bed. However, I do not operate a manufacturing company so I will have to trust that everything is on track, or that the gap I noted is a problem with the source artwork.

1 Like

Camera accuracy sets positional accuracy. Since the macro lens picks up the Glowforge logo on the top of the cutting head to know where it is.

So, regardless of thermal expansion or any other factor… the machine will know exactly where the cutting head is, and then can move it based on maximum resolution.

Building a machine with enough resolution to achieve the stated accuracy is definitely possible, you just can’t achieve that much accuracy in practice unless you can measure and compensate for things like backlash and thermal expansion.

They could use the macro camera to implement closed loop control (i.e. take the number of steps the controller thinks is necessary, use camera to check position, adjust as necessary) but since it only has a resolution of 0.05mm they will struggle to achieve an absolute positioning accuracy of 0.025mm without doing something quite clever or making use of an additional, more precise position feedback system.

1 Like