[Sandy H.]

This project is not specifically for my DIY router, but it is leveraging the information I've learned so far in trying to build my own from this forum subsection, so I thought this would be a good place to post it. If there is a better place, please let me know.

This 'machine' is a simple 1 axis traversing 'do-dad' for my company. After researching DIY routers here, I thought it might help solve a problem at work and also let me experiment with someone else's money! Its purpose is just to get to positions along 1 axis. The load is less than 2 pounds.

I constructed it using an 8020 frame, 3/4" linear rail/ball guides and a 16mm rolled lead screw with a 5mm pitch (spec says 0.002"/12" accuracy) with an anti-backlash ballnut. The electronics are from Keling and include the C10 BOB, a 4030 stepper driver and a 270 in/oz stepper motor. The PC is an old Pentium 4 running XP 32 bit, using the parallel port. The software is Mach3 and the parallel port passed the driver test with 'excellent' at 25khz. It also passed at 65khz with an 'excellent' but I'm running Mach3 at 25khz still.

The issue I seem to be having is that I'm not moving as far as the PC thinks I am. For example, if I make a move from 0 to 40", I only move 39.95". I would assume that if everything is working properly together, it should be much closer than 0.050" off. This was measured with a 72" vernier caliper, so I believe the measurement is accurate.

In Mach3, my motor tuning specs are 8128 steps/unit (inches; The driver is set for 1/8 microstepping). I have the max velocity set to 50IPM. The results seem to be very similar even if I use a feed rate of 10IPM as well.

The machine does repeat very well (i.e. when I go back to 0, it is there within 0.001" every time), so I doubt the coupler is slipping. I'm not sure if I'm losing steps, the screw is way out, my configuration is off or what. The stepper motor is very hot to the touch (guess 120-130 deg F).

Any help troubleshooting is greatly appreciated. I'm sure I've left out a few key pieces of data, but I'll be glad to fill in what I can with more information as needed. I do suspect it is user error on my part, but I'm not sure what additional steps to take to resolve the issue.

Thanks for the excellent forum and any help that can be provided. It was pretty cool for sure to see this thing move the first time. I can't wait to finish my own router!

Sandy.

[ger21]

The act that it's repeatable would point towards the steps/inch.

What do you get when you try to move 10"? 20"?

Try 8138.16 step/inch and see if it's accurate, but make sure it's consistent along it's length, unless you only need the two end positions to be accurate.

Since 50ipm is fairly low, you can probably lower the power supply voltage a bit and the motor will run cooler. What voltage are you using?

[Sandy H.]

The data is still at work, but it did 'seem' to be losing between 0.004"-0.02" every 10". I'm not sure its fair to say that exactly, but it didn't seem to be linear at the time.

The supply voltage is 36V. I would say it got hot within the first 30 minutes and stayed roughly the same temperature for the next 8 hours.

I'll try 8138.16 step/unit tomorrow, but I'm not sure where that is derived. I used the following to get 8128 s/u.

25.4 (mm/in) / 5 (mm/rev) * 200 (step/rev) * 8 (microsteps/step) to get 8128 (microsteps/in).

We are going to use this apparatus as a go-no go to verify another linear position sensor, so we need multiple data points to be accurate along the way, not just the end points. It doesn't have to be perfect for sure, but almost 1/16" over 40" seemed pretty poor to me.

Thanks for the help!

Sandy.

[TiagoSantos]

Can't help much with the problem, but I can tell you where the 8138.16 number came from.. Assuming the error is repeatable (it always ends up 0.05" short, every time), then your machine might be accurate but the ballscrew is messed up. Anyway, if the intended travel is 40 and you're getting 39.95 - multiplying 40 by the number of steps you're currently using and dividing by 39.95 = 8138.17.. So you could use 8138.17 (.16 is close enough!) to compensate for whatever error the ballscrew is introducing.

[ger21]

I took .05 and divided by 40, then multiplied by 8128. You are off .00125/inch, so 1.00125 x 8128 = 8138.16

[louieatienza]

I took .05 and divided by 40, then multiplied by 8128. You are off .00125/inch, so 1.00125 x 8128 = 8138.16

I don't know if .01 will matter, but I get 8138.172... it's a simple ratio:

8128/39.95 = x/40

or

x = 8128*40/39.95

So over 40" you'd be off 5/8 of a step, or <.001" hehehe

[ger21]

You're right, my method was incorrect. Close, though.

[Sandy H.]

Thanks for the explanation. I guess I should have gotten there myself, but I didn't.

I brought home some data, after tweaking the step/unit value and improving the measurement technique.

(left is stepoint, right is measured; all in inches)
0=0
10=9.997
20=20.000
30=30.001
40=40.013
30-30.001
20=20.002
10=9.998
0=-0.001

These were measured from zero, out to 40 an back to zero without compensating for backlash.

The next move was from 0 to 40 without stops.
0=0.001
40=40.010

I left the setup sitting for 2 hours to let the motor come up to temperature, but it was only warm, not hot, so I assume the heat was from actual movements, not just sitting there as I assumed before.

0=0.000
40=40.011

We then played around a bit. I made a mark on the coupler (between the stepper and ballscrew) and told it to move 39.37 (i.e. an even number of rotations based on a 5mm pitch). It came up very close (i.e. my opinion is exactly) to the same position. I feel that it is safe to say that we're not losing any steps at this point, definately not 0.010" worth of steps which would be over 10 degrees of rotation, which should be easily viewed with the mark.

I am somewhat comfortable that the issue is with the screw at this point. I want to make mroe measurements tomorrow, but any other opinions/suggestions are appreciated. If I can repeat tonight's measurements tomorrow, we'll try to map the screw (just measurements, not the software function) to see if there is something else interesting happening.

Thanks for all the help!

Sandy.

[louieatienza]

I'd say if you get the similar readings, it's most likely a tiny lead error; though, .013 is a heckuva lot better than .05!

I'd be inclined to take at least 5 readings at each interval, and see if it's consistant. If so, then backlash is not your issue, and maybe mapping the screw might help. The largest discrepancy seems to be in the 30-40 region.

Have you tried starting from 40, and measuring back every 10, to see if the readings are the same? Or even make 40, 0 and reverse the axis direction in Mach3 and see if the differences are consistant...

[rowbare]

Hoss made a video on Mach 3 calibration: [ame=http://www.youtube.com/watch?v=XkO5tc-jSxw]Improved Mach 3 Axis Calibration - YouTube[/ame]

If you need to measure at multiple points, you may want to map your lead screw to compensate for lead error.

bob

[Sandy H.]

Well, I'm now satisfied with the accuracy that it is currently calibrated for. It might not be good enough for a 3 axis machining center, but it is fine for the test apparatus we're going to use it for.

There were (of course) a combination of things going on. The main thing appears to have been user measurement error. As the vernier caliper is 6 feet long, it is somewhat of a 2 man job to measure with it and we weren't working well together. That, along with old age eyesight, meant that the precision of the measurements were poor at best. We tweaked our set-up and used a 5X eyepiece to read the vernier and all of the sudden, measurements were at least more consistent along the screw.

Then, we calibrated it against the more properly measured value and have +/-0.004" over the 40" of measurement. We verified it at 5 points along the span to make sure we weren't just getting the start and finish right with the middle being way off. As I said, +/- 0.004" is where it is, somewhat evenly distributed along the way. I think the screw is far from 'perfect' and I don't really think it was exhibiting +/-0.002"/ft per spec, but overall it isn't very far off from spec and there could still be some accuracy issues with my design and the stepper motor control.

Overall, I am satisfied with the current status and I think it could be improved if necessary by trying to map the screw and tweak the kernel frequency etc. I imagine I'll do more of that on my router, though, as it will be my home project, where time doesn't really cost much. . .

Thanks for all the help. I both appreciate the forum and the quality vendors involved in this hobby. Its pretty refreshing to say the least.

Have a good weekend, folks!

Sandy.

[PaulRowntree]

Good news that t is working for you. Accurate measurments over long runs is always tough. JerryBurke ('Big Bamboo') showed a neat technique with a metal ruler and a USB microscope on the gantry. You might want to check post #145 of http://www.cnczone.com/forums/cnc_wo...w_machine.html

Cheers!

[RomanLini]

...
Then, we calibrated it against the more properly measured value and have +/-0.004" over the 40" of measurement.
...

That's not too bad at all.

Just be aware that the thermal expansion of steel leadscrews is about 12 microns per metre per degree C (based on my memory, but you can check your leadscrew specs).

So on a 40" length it might typically rise 15'C in temp from movement or workshop heat, which will give an increase in length of about +0.18mm or +0.007".

If the machine moves fast, or the leadnut is tight etc the screw heating might easily be double that giving a length increase of maybe 0.014".

[Sandy H.]

Very good points. This is going to be used as a test machine for verifying the fuction of a sensor. It will be used in a climate controlled lab and probably only only cycled 3 full strokes in about 20 minutes a few times a week.

I think it should be pretty stable. I plan on trying to verify the measurement a few times a year as well. Overall, I am very pleased with how its turned out.

Sandy.

[stevespo]

I have really enjoyed this thread. I always wondered how accurate my rolled ballscrews are without having an easy way to measure/verify them. I can take small measurements with my 6" calipers, but after that - no luck.

It seems to me that a 24" vernier caliper would allow me to verify my setup with a high degree of accuracy. My question is - should I hunt around for a used Starrett model #123, or pick up a cheaper import (vernier or dial)?

My assumption is that the Starrett (Fowler, B&S, etc) will be factory certified to some higher standard. The imported models, not so much. Thoughts?

According to the Starrett website, the vernier caliper is more accurate regardless, and perhaps that already moves the decision towards a used vernier.

"The best electronic and dial slide calipers, regardless of resolution, are
accurate to within .001" or 0.03mm, every 6" or 150mm. The best vernier
calipers are accurate to .0005" or 0.013mm per foot or 300mm."

[Sandy H.]

Glad you've enjoyed the thread. Me too!

I can only give you an impression of what I've seen so far. I really expect that I will get more and better data as time goes on, but at this moment, I've performed the following tests:

Use a 60" vernier (annual certification of accuracy) to measure an arbitrary zero and a 40" move. Use that data with the automatic step size button to establish the steps per unit. In my case, it was similar, but not the same.

Following that test, I checked back from 0 in 10" increments to see if it was a mile off. It was somewhere around +/-0.004" at the 20 and 30" marks, and 0.000" at 0, 10 and 40.

I also mounted a 1" dial indicator and tested it over 0.1" increments at various locations along the screw. In general, it was an average of around +/-0.001", with the result sometimes going +0.0005/-0.0015 and the opposite. I never saw a result off more that 0.0015" in either direction over the 1" of the indicator testing at 0.1" increments.

For me, the real kicker was the measurement method. When using a 6" caliper, my results were more along the line of +/-0.005", as it was not easy to get a consistent alignment of the caliper. With the big vernier, we had to make a temporary bracket and use alignment marks etc to get the measurement accurate. No kidding, when trying to eyeball it, we could be off 0.040" every now and then. Once we made a more repeatable set-up and used an eye loupe to verify the readings on the vernier, the numbers dropped to what I'm reporting now. In actuality, I was getting closer to +/- 0.002" over the length, but my helper was getting +/-0.004 (at the same places, in the same direction) so I'm stating the worst measurements we saw.

When I get more time, I want to re-do the measurements in 1" increments, not 10" increments just to see what's going on. I have a few more measurement devices to implement just as an experiment, so if I find something of interest, I'll report back.

I am a bit mind-boggled, though, at how well written the Mach3 software is.

Sandy.

[stevespo]

Thanks Sandy, that is helpful. I will be on the lookout for a quality vernier and I do understand how important a repeatable setup and test procedure is for accuracy. I'm not expecting to find anything "wrong" with my screws - they are good quality rolled screws - I just want to know that when I drill a hole at 20.000", it really is +/- some reasonable number.

Steve