LitePlacer.com forums

Also, worth adding that I pushed the Z axis too, that one I can identify the bottom right bearing as whining near the limit
(Easier to localize audibly).

Very interested to see how your new servo steppers work out, I’ll switch too if you’re getting significantly better speeds.

I tried 30-toothed pulley at the X axis and concluded it is not worth using. The maximum speed is about the same, but maximum jerk is slower and i had to set higher current.

Jet wrote:At 18000, the steppers gave up. Tried at various acceleration speeds (250, 500, 1000) and tried reducing step count to 4.

Have you tried loosening the v-wheels slightly so that they still grip the rail but can be scrolled manually?

Jet wrote:Also tried increasing the current from just under half where I have it now to 100%.

Does your maximum speed decrease while increasing current? Mine does: https://liteplacer.com/phpBB/viewtopic.php?f=4&t=6127

Here are Juha's settings: https://youtu.be/T_v2VJKNZQ0?t=10

Here is my machine: https://www.youtube.com/watch?v=Kt-apUSUZnk

wormball wrote: Have you tried loosening the v-wheels slightly so that they still grip the rail but can be scrolled manually?

Yes, tried everything from tight (moving wheel manually moves the head) and loose (wheels move without moving head).

wormball wrote:Does your maximum speed decrease while increasing current? Mine does: https://liteplacer.com/phpBB/viewtopic.php?f=4&t=6127

Tough to say for sure as there isn't much of a change past mid point, but I can get a little bit more a full current. But the drivers on the TinyG aren't the best eiher.

wormball wrote:Here are Juha's settings: https://youtu.be/T_v2VJKNZQ0?t=10

Here is my machine: https://www.youtube.com/watch?v=Kt-apUSUZnk

What's interesting is that Juha's using 4 microstepping, I'm using 8, this would make stepper resonance worse (if it was that), but I don't hear any resonance/bearing issues in
his video. I did try 4, but it didn't increase the speed, but I'm going to revisit this and work through all the various current levels again as it'll provide a bit more torque.

On your video, I could hear the same bearing type whine starting, more on your Y than X (which is expected as there's more mass on Y). It's showing up on the longer moves at peak speeds.
Each move is Accelerate to Peak speed and Decelerate, so you may not be reaching the speed you think you are on the short moves. The real test is to run "Test X" and make
sure audibly that you hear a plateau speed indicating full speed was reached. I'd be interested to see what yours sounds like when you do this.

If the bearings are skipping it will wear down the V rails over time, another possibility is that it's just resonance of the V wheels themselves, or friction noise.
Also dust can accumulate and get compressed on the wheels in which case they need to be cleaned.

Juha: Was there any change in the V Wheel bearings or V Wheels between your machine and the ones shipped? Or are you lubing the rails? (I didn't)

Jet wrote:The real test is to run "Test X" and make
sure audibly that you hear a plateau speed indicating full speed was reached.

As you wish:



However my phone records sound almost as perfectly as it records video, so actually there is no such bathtub sound as in the video.

I loosened the X belt a bit and so i achieved 72 m/min with 30-toothed pulley (which is exactly 1.5 times the speed with 20-toothed pulley, so maybe i will try 40-toothed pulley which i also received) and also idle whistling sound decreased. However when i loosened the Y belts, the effect was not so pronounced. So maybe you should try loosening the X belt.

Jet wrote:What's interesting is that Juha's using 4 microstepping, I'm using 8

From my experience 1/8 microstep performs slightly better than 1/4 microstep. Maybe it would be significantly better if the drv8818 was able to give precise 1/8 microsteps, but it is obviously not able. https://liteplacer.com/phpBB/viewtopic. ... 9294#p9292

Jet wrote:as it'll provide a bit more torque.

I do not think 1/4 microstep will provide more torque.

Jet wrote:But the drivers on the TinyG aren't the best eiher.

Do you know better drivers? Maybe there exist drivers pin to pin compatible with the drv8818 so we can replace it with them?

Incremental torque decreases as you increase microstepping: https://hackaday.com/2016/08/29/how-acc ... ng-really/
Essentially this means the ability to hold or move to a microstepped position is less accurate. However microstepping also reduces noise and resonance, it's a tradeoff.
But there's also a sweet spot, so if you can get the accuracy at 4X microstepping required, then 8 will be slightly quieter, but it's not worth it as it's not real resolution.

Regarding loosing the belt, it's about as loose as I can have it and get repeatable 0.02-0.04mm accuracy on X/Y with backlash compensation. I do need really good accuracy.

I still hear that bearing noise in the video at high speed, it's worse on Y than X. It's like a grinding / whining noise. I'm probably just
tuned to it because of my previous experience with developing acceleration for 3D printer software, I ran a lot of tests for machine limits.

If I was to replace the drivers "properly" I'd probably go with one of the geckos, I've used em in 2 other projects, and they rock. That's assuming I wasn't going servo like you're doing.
But I don't see the point until I can get the bearing issue solved, as the geckos won't cure that.

It sounds like I'm complaining, but I'm not, I'm really quite happy with the machine, if I could double the speed it would be nice, but not a necessity as placement precision is more
important to me, and Up Camera will slow everything down anyway.

Jet wrote:Essentially this means the ability to hold or move to a microstepped position is less accurate.

It means the ability to hold or move to a microstepped position is less accurate than nominal microstep, but at least as accurate as more coarse microstep in the same conditions (if the driver has no severe bugs, of course).

Jet wrote:I still hear that bearing noise in the video at high speed, it's worse on Y than X.

But the X speed is much greater than the Y speed. I think this is cos the Y motor is attached directly to the table which acts as loudspeaker. And as i mentioned, after the resonance point this sound actually decreases with speed.

Jet wrote:It sounds like I'm complaining, but I'm not, I'm really quite happy with the machine

If i was not able to get speeds above 10 m/min, i would complain.

Jet wrote:Regarding loosing the belt, it's about as loose as I can have it and get repeatable 0.02-0.04mm accuracy on X/Y with backlash compensation. I do need really good accuracy.

My repeatability is about 0.02 - 0.04 mm even at the mentioned speeds and with 30-toothed pulley.

Jet wrote:If I was to replace the drivers "properly" I'd probably go with one of the geckos, I've used em in 2 other projects, and they rock.

I think $109 driver is a bit overkill. What speed had you achieved with them?

Jet wrote:That's assuming I wasn't going servo like you're doing.

Have you tried any servos and can you recommend them?

Jet wrote:But I don't see the point until I can get the bearing issue solved, as the geckos won't cure that.

I still think that it is not bearings and you should try 1.8 degree motors. Or you may remove the belt and try all the bearings by hand (e. g. wind 3 - 4 windings of nylon string on the idler and pull one end of the string imitating working load).

Also i found in my recordings that with 0.9 degree motors the maximum speed was higher with 2000 - 3000 jerk rather than with 1000 jerk.

And one more thing. I am using unshielded cables to drive motors. Maybe you are using too stiff cables?

I measured some numbers.

Homing dot X coordinate while moving the carriage by 0.035 mm (which is about 1 microstep):

36, 54, 79, 115, 145, 163, 224, 242, 278, 315, 351, 387, 424
36, 61, 85, 115, 151, 169, 230, 242, 284, 315, 351, 387, 430, 448, 448, 514, 526
696, 666, 629, 599, 569, 545, 496, 490, 460, 424, 387, 351, 321, 284, 266, 218, 194, 157, 121, 85, 48, 12
(Maybe the plateau at 13 - 14 steps is due to 0.035 mm is slightly less than the microstep , and no actual step was performed at this moment)

Winding voltages vs steps:

-1262, -1615, -1911, -2160, -2302, -2358, -2188, -2051, -1815, -1516, -1159, -731, -293, 0, 687, 868, 1284, 1645, 1947, 2195, 2348, 2405, 2227, 2090, 1847, 1544, 1181, 745, 299, 0, -676, -878, -1299, -1668, -1981
-1788, -1510, -1154, -730, -294, 0, 692, 866, 1274, 1632, 1930, 2169, 2359, 2380, 2188, 2052, 1813, 1521, 1165, 738, 303, 0, -688, -877, -1297, -1656, -1968, -2199, -2366, -2411, -2230, -2087, -1850, -1552, -1190
So i think the drv8818 is not the best possible driver.

Remove the X belt ant test the X motor loose. You will find that it stalls at 14 m/min. And if you unscrew this motor and hold it in your hand, it stalls at 35 m/min (do not forget to increase jerk and/or X axis length). At least i got these results before i replaced them with 1.8 degree motors.

So turns out we’re both wrong, it’s not Stepper Resonance or Bearing Skip.

I decided to do some more experimenting to prove to you why it can’t be Stepper Resonance, because:

1. Stepper Resonance is at a lower frequency then we’re seeing here (just do the calcs)

2. Stepper Resonance is under low load (we have high)

3. Stepper Resonance is under Full or Low Microsteps (we have high)

4. It’s at a specific frequency, not a large band (we have a large band)

The reason for this is that stepper resonance acts like a spring, it’s an oscillation at the resonant frequency around the full step detent location, caused by the sudden motion from step to step being a large amount to move and setting up an oscillation around that point, just like a spring.

Therefore the standard ways to solve it are to microstep (which reduces the force applied to the spring), and apply load (essentially damping the spring system). Both of which are already being done. STEPPERS ARE DESIGNED TO BE LOADED, running them unloaded is meaningless. Then the only other possibilty with respect to resonance that’s left is if the stepper is poorly made and there’s an additional resonance point at high frequency... Which there isn’t because it’s not in the Torque Curve for the 17HM19-2004S (check the datasheet and torque curve). Normal stepper resonance is generally below the torque curve range because you’re outside of the suggested operating range of the stepper motor anyway, so they don’t put it in the torque curve.

You said Gecko’s are expensive, but it’s for a reason, they’re extremely well made. If you want the best of the best, that’s what you get, they’re also works of art if you look at the PCB. They are typically optimized for 10 microsteps because that’s the best microstepping amount for steppers, anytime you see 8, 16, 32 etc., you’re dealing with less than optimal
stepper movement. Also, they switch to full stepping at higher speeds automatically (which gives you more torque).
And they also microstep more accurately. If you want to really understand steppers, look at some of their literature.

I’ve used Geckos in a number of commercial projects for high end equipment, so I’m lucky to have a number of different geckos on hand.

So I put a G251X Gecko on the X axis tonite, and no whining. So I decided to push it.. with the stock steppers.
I’m managing 28000mm/min and Acceleration: 40000 10^Gmm/min^3 reliably. This fits exactly with the torque curve for this 0.9 degree motor, (which ends at 30000mm/min). I tested both V-Wheels tighter and looser.

With a 1.8, due to the higher torque, I would expect considerably better.

Bottom line is the DRV drivers are crap. This isn’t something I didn’t already know, Allegro’s have always been better, but Geckos have always blown both of them out of water.

To reproduce with the stock steppers, wire up to the TinyG (don’t connect Enable), and put a 2K resistor for the 2A winding current.

G251X was the first one a grabbed as it’s small and fitted my short stepper wires better, but it is their most basic model, it’s also their cheapest... So there’s a couple of issues:

1. Enable/Disable is inverted, so I disconnected this line (I’ll look at changing the TinyG firmware if that feature doesn’t exist).

2. It’s only 10uStep, and TinyG only does multiples of 2, so I have it set to 8, and have to multiply the speeds by 0.8, however this is non-workable in practice because the movement distances will be off (again requires a firmware change).

Although I have other Geckos too, I’ll see if one of those will do 8 stepping (some include step multipliers).

But I’m going to have to do some case changes to mount the drivers. Then I’ll probably consider going to 1.8’s IFF the vibration isn’t too much.

So Geckos for my X/Y, Z and Rot I’ll leave as is.....
And I probably wouldn’t have even tried this if it wasn’t for your discussion about stepper resonance.