LitePlacer.com forums

_So there’s a 50KHz pulse rate limit on the TinyG output, which equates to 30000mm/min with stock Liteplacer hardware.

The G251X can handle up to 300Khz, if you’re running 1.8deg, the TinyG will be able to handle 60000mm/min.

Other options for 0.9 deg is to 5 uStep and use one of the other Geckos, although as is at the end of the torque curve anyway, there’s no point.

and TinyG can handle Geckos apparently.

So I’ll go G251X / TinyG for X/Y and then maybe switch out X to a 1.8 if I have one here suitable.

1.8 @ 10uStep = 0.02mm accuracy which is around the limit of accuracy of the machine anyway (although I have a hunch
that the limit maybe related to the poor DRV drivers as I’ve been seeing some anomalies as I’ve been improving calibration in OpenPNP)

So I just tried a 1.8 (not a particularly great one) the only one I have that’s this small. Kysan 112040, a bit worse than your 42BYGHW609.

Standard spec (stock gear/belt) you were achieving 45000mm/min

With the Gecko I’m getting 60000mm/min @ 4000 Acceleration and no whining.

I can’t get past that point as 60000mm/min is the max the TinyG can do, and when that happens, the pulse
rate can’t keep up, you loose steps, and then that causes your stepper to stall due to inertia, because the missing steps
create sudden speed changes.

This 33% improvement over yours, I attribute to the Gecko....

So if you really want to go faster than you are now without compromising your resolution, get a Gecko, do 10uStep
and shop around for a more decent motor, wouldn’t be surprised if you could get to 90000, if you drop to 5uStep, but then you’ll need to go with one of the higher model steps that can do 5uSteps using the built in step multiplier. At 5uStep, your speed limit for the TinyG would be 120000.

For my use, I’ll just stick with the 0.9’s for now with Gecko G251X for X and Y, and I’ll build an inverter for the Enables.
I don’t believe in motion systems pushed to the limit, you always get degredation over time due to dust, stepper heating etc. plus then there’s the vibration / instantaneous jerk consideration, for 3D printer design, I’ve generally gone with 1/2 the mechanical limits. Also with the Liteplacer, there’s the possibility for damage... So for this, I’ll probably settle on 20000mm/min, because of accuracy considerations and it’s a lot of mass to fling around. I need reliability and accuracy, speed comes second for me.

I’ll get the G251X properly mounted in the case, and put in one for Y, then do some repeatability tests. I wouldn’t be surprised if the repeatability was better with the Gecko, certainly I was seeing some things I think may have been caused by the DRV.

If you follow what I do, don’t forget to turn your current down on the TinyG drivers on the axis you’re using the Geckos, it can cause issues if you leave the drivers running at current and no load. Also, Geckos can be driven at higher voltages too if desired.

Thank you all for putting up your knowledge, measurements and experiences. I'm learning a lot on this thread.

Very cool! I do not think i will spend $140 for drivers, but i will try cheap chinese drivers i bought without any particular purpose.

wormball wrote:Very cool! I do not think i will spend $140 for drivers, but i will try cheap chinese drivers i bought without any particular purpose.

Seriously, don't do that, the cheap chinese external drivers are generally junk, infact if you need those I have bunch you can buy off me real cheap, I stopped using them a while back. Even the DRV's are likely better than the cheap chinese externals and with the Chinese ones you'll get less torque.

The G251X's are only $65 USD, and if that's too expensive for you, get the G250X (which has no heatsink and header pins instead of screw terminals, and
supply your own heatsink for the FETs), those are $45, and only 2 are needed... they are functionally equivalent to the G251X.

I decided to dispense with the inverter in the end and will use a toggle switch for enable/disable. That fits my usage better, I can disable X/Y when I need to manually move, but most of the time I need it powered anyway to maintain position / accuracy.

So i’ve permantly added Gecko’s to the install for X and Y, an enable/disable switch and done some additional testing after Nulling the drivers with the trim pot. This is all the stock 0.9deg steppers, belts and gears.

Max speeds:

X: 30000 mm/min and Accel: 18000
Y: 30000 mm/min and Accel: 4000

Note, 30000mm/min is 50KHz pulse rate @ 10uStep the maximum the TinyG can handle. To get faster than this, you would need 1.8deg or geckos with the stepper multiplier and 5uSteps.

The acceleration maximum difference between X and Y is due to the extra mass on Y.

For repeatability error, I have some minor improvements (with backlash compensation), approaching from 2 directions:

X: 0.01-0.02mm (was 0.02mm all the time)
Y: 0.01-0.03mm (was 0.03 - 0.04mm)

One noticeable improvement was with the TinyG drivers, every 0.01mm move would not always move due
to poor microstepping, it used to move stop and skip as it caught up with microsteps.

Now every 0.01mm move, moves, and it’s fairly smooth too, so it should equate to a positioning accuracy improvement.

With the steppers stopped, heating is minimal, steppers are just warm to the touch (despite driving them at the rated 2A).

All this is extreme though, I could probably eak out a bit more acceleration if I mess about with belt tensions and v-wheel tightness, but in practice I’ll be running slower, to the point where the table isn’t vibrating due to the extreme movements.

Also, probably should keep my head away from it

Here’s a picture of the install, geckos are top left (with the red led):

Jet wrote:X: 30000 mm/min and Accel: 18000
Y: 30000 mm/min and Accel: 4000

Why so low jerk by the Y axis?

Jet wrote:Seriously, don't do that

No! I must try it myself.

What chinese drivers do you have? I have one A4988 and two DRV8825s.

Jet wrote:I don’t believe in motion systems pushed to the limit

It's not nearly the limit. Had you seen this?



This mod is officially blessed (and collaborated) by Juha. https://liteplacer.com/phpBB/viewtopic.php?f=4&t=5972

Jet wrote:I need reliability and accuracy, speed comes second for me.

I think motor accuracy within 0.1 mm is ok for current liteplacer and 0.5 mm pitch parts. And i noticed no accuracy impact at any achieved speed. But i noticed some creep of the nozzle to camera offset with time, even more mysterious than the coordinate dependency that i defeated. I replaced stock camera mount springs with rubber washers, so the camera became more steady, but it seems it had no effect on the creep. I think if we defeat that, the liteplacer will be able to place 0.4 mm pitch parts at 60 m/min (at least with vision).

Jet wrote:Also with the Liteplacer, there’s the possibility for damage

And what with the openpnp? I think it is possible to invent some rubber bumpers to prevent serious damage.

wormball wrote:Why so low jerk by the Y axis?

Google Mass, Inertia, Acceleration and Torque.

wormball wrote: No! I must try it myself.

I assumed you meant the higher end Chinese drivers, not the cookie cutter step stick type A4988 / DRVs. The DRV8825's you will find will have poorer
micro stepping accuracy than the A4988's, and the A4988's won't be anything like the quality of the Geckos. Also, they don't do morphing,
Gecko does.

wormball wrote:It's not nearly the limit. Had you seen this?

It's awesome, but it's not the same hardware. Servo's will always perform better than steppers, although typically there is some inaccuracy due
to the encoder being on the motor, not the axis. Gecko also make Servo Controllers.
if you wanted to do this correctly, you'd put a positional encoder on the axes, have servos and re-engineer the machine, but then it's no longer a LitePlacer.

The point I'm making is, sure you can always go faster, but it impacts accuracy due to vibration, overshoot, belt stretch and ringing. It's
a fun exercise to go faster (and I've had fun doing it), but there's practical limits if you want accuracy.

wormball wrote:I think motor accuracy within 0.1 mm is ok for current liteplacer and 0.5 mm pitch parts.

Except it's not and that's unachievable with the stock Liteplacer Software on 0402's anyway.
BTW, I'm placing down to 0.3mm pitch in OpenPNP. which is why I need an accurate machine.

0.1mm motor/positional accuracy is no where near enough to place 0402's. With 0.1mm positional accuracy, a 0402 can be placed completely off it's pads, and a 0603 with only 12.5% pad contact.

wormball wrote: And i noticed no accuracy impact at any achieved speed. But i noticed some creep of the nozzle to camera offset with time, even more mysterious than ....I replaced stock camera mount springs with rubber washers, so the camera became more steady, but it seems it had no effect on the creep. I think if we defeat that, the liteplacer will be able to place 0.4 mm pitch parts at 60 m/min (at least with vision).

Sounds like the high speed is moving things, try slower. One thing I would say if the cameras has fairly poor lenses, and minor lens rotation will equal a significant offset difference.

The faster you move, the longer you will have to wait for the camera to settle before it's used due to Frame Rate and Vibration, it could be as you're pushing the speed you're seeing more creep and it's due to this.

wormball wrote:

Jet wrote:Also with the Liteplacer, there’s the possibility for damage

And what with the openpnp? I think it is possible to invent some rubber bumpers to prevent serious damage.

What with openpnp? Bumpers?... Um, still a lot of force.

wormball wrote:What chinese drivers do you have?

Way too many.

Left to right, beefier steppers, encoders, servo motors, higher end Chinese drivers, A4988’s (13 of them) and a variety of Geckos.

I don’t bother with the DRV’s, because I know from previous experience that they are poorer than the A4988s. The A4988’s are my go-tos for lower end stuff, Geckos where I need really good performance.

Also have a bunch more stuff in storage (but couldn’t tell you what’s in there, likely more of the same).

If you want to buy anything off me at a reduced cost for your “need for speed”, you’re welcome to... except the Geckos and Servo Motors, I’m keeping them.

Jet wrote:Seriously, don't do that

You were right! I tried drv8825, it worked in full step mode, but then i decided to try microstepping, so i connected one of the microstepping pins to the positive voltage. But this voltage was +24V instead of +3.3V. So the drv8825 burned spectaculously, and now the Y axis is not working. So i will replace the processor and the drv8818 when i receive them.