Excellent example of Servo vs. Stepper vs. Feed Speeds. This is by John Dammeyer, developer of the Electronic Lead Screw Project: Electronic Lead Screw Main Page
…now I did some experiments as a result of this posting subject. I found if I turned the handle for moving the carriage at what was a pretty normal speed it went 10” in about 4 seconds. That means 10 inches in 4 seconds is 2.5 ips or 150 ipm.
If your goal is to replace the lead screw with a ball screw and still be able to traverse as quickly as the rack and pinion while maintaining knowledge of the position to the nearest 0.0005” you need to consider how you drive this lead screw and the pitch of the screw.
Let’s look at a stepper based solution first.
1. Max RPM with a stepper and still decent torque is about 700 RPM.
a. Use 600 RPM to be conservative (10 RPS).
b. 10 uSteps per step results in 2000 steps per rev x 10 RPS is 20,000 steps per second. (Gecko Driver)
c. Resolution is 2000 steps per rev but accuracy is still only to the nearest ½ step so 400 steps per rev.
2. With 400 steps per rev and a target of 0.0005” accuracy a pitch of 0.2” or 5 TPI is required.
3. With 10 RPS we’re moving 0.2”/Rev x 10 Revs/Sec = 2”/second (2 IPS). However, your target is 2.5 IPS or better.
4. This system won’t move the carriage quite as fast as you normally would with the handle.
Now a Servo.
1. Max RPM is usually 3000 RPM and torque is good all the way to the top.
2. Aiming for 750 RPM means a 4:1 belt reduction.
a. 4:1 on the belt though results in 4x the encoder resolution.
b. With a 100 line encoder in quadrature you get 400 lines per rev x 4:1 or 16,000 steps per rev.
c. Accuracy is still 16,000 steps per rev.
3. Stay with a 4 TPI ball screw and 0.2” pitch we resolve (and position) down to 0.0000125” per step.
a. But at 20,000 steps per second from the ELS you can really only turn the lead screw 1.25 turns per second
b. That’s 0.2”/rev x 1.25 revs/sec = 0.25”/second (0.25 IPS). A factor of 10 too slow.
4. I think Gecko has a step multiplier that turns the servo 10 lines for every step in.
a. Now we’re at 2.5 IPS which was the goal.
So: A DC servo Motor with 100 line encoder on a 4 TPI ball screw results in a position resolution and accuracy of 0.000125” with a top speed of 2.5 IPS and a step rate input of 20,000 steps per second.
Please correct my math if I’ve made a mistake.
From: E-LeadScrew@yahoogroups.com [mailto:E-LeadScrew@yahoogroups.com]
Sent: March-19-16 8:29 AM
Subject: [E-LeadScrew] Re: Ball screw question
Though I am not an engineer it looks to me that the stock design of the
leadscrew on this particular lathe looks to be at fault. The leadscrew has a
slot that runs down the entire length of the screw itself that is used to
power the cross feed. The screw was turned and then the slot was milled.
This left many “teeth” that in a relatively short amount of time “cut” away
the half nut. The half nut itself just pushes forward on a lever. There is
no “backing ” at all to support the half nut and in order for it to track a
considerable about of force is applied to keep the half nut engaged.
I doubt it is the design’s fault. Thousands of lathes use this exact
system to drive the carriage and cross slide without and wear issues. The
half nuts stay engage without any extra pressure also. I am wondering if
someone ground the bed and didn’t take the extra time to align the lead
screw with the carriage? I can assure you something else is going on with
your lathe other than the keyway down the center of the lead screw.
Rick in WA State
[Non-text portions of this message have been removed]
Posted by: “John Dammeyer” <firstname.lastname@example.org>
|Reply via web post||•||Reply to sender||•||Reply to group||•||Start a New Topic||•||Messages in this topic (12)|
EXACTLY ! I use a 10.000 count encoder on the servo, and 1:2 belt drive, and 5 mm pitch screw. => 20.000 counts/rev at screw. => 4000 steps/mm. = 0.25 microns. This is pretty much the ideal setup. However, 3000 rpm = 50 revs/sec. 50 revs x 10.000 counts = 500.000 kHz. So, I must have a hardware timing engine able to pulse at 500 kHz (Cslabs CSMIO-IP-S). End result. The servo will spin upto 3000 rpm in approx 0.03 secs (20 ms no-load, and 20-30 ms with the 100 kg carriage). This is far too fast, huge wear on belts and screw and ways, and very much more than needed. So, I limit it in sw to something same. Speed. At 1500 rpm at screw, or 25 revs/sec => 25 x 5 = 125 mm (==5") per sec. Movement on z axis is approx 400 mm max. So, end to end is 3.2 secs. This is far too fast, so I software limit it to something sane. Generally, the tools are only about 100 mm from the workpiece, and in less than a second the lathe has the potential to crash. Thats why I normally keep the speeds way down. The 0.25 micron step size results in real-worl, actual, resolution of approx 1 micron, or a bit better. Hope is to get 0.5 microns, and may (or may not) need ground ballscrews. Engineering theory and lots of real world examples says 0.5 microns is routine, given a sufficiently rigid system. Thats why I use a very think, 32 mm, ballscrew, because it has very high rigidity of 54 kgf/um. (High precision optronics screws with 0.25 mm rise can position to 0.5 micron accuracy and several manufacturers quote similar accuracies. Available from thorlabs, about 70€ each. Its the kind of thing I will be making). Accuracy is not, at the moment, 0.5 microns, but in theory and practice I will probably get there. Plan is to add glass scales, and secondary feedback to the CSMIO controller. At the moment, its possible to do incremental movements to 1 micron. This means I can make gages, with steps, of 1 micron. By then measuring the gage, one of the steps will be 49.999 mm. This will then be what I make bore for, to mount high precision bearings. (7210AC-DUP-P2. Yes P2 or ABEC 9). The z axis started working last tuesday, after 300 work hours (this time round), and 12 years of development and tries. Alignment was really hard. Last TS bracket (not really needed) will hopefully be finished today. Very Big Grin. Gonna go make bracket. On 19/03/2016 17:18, 'John Dammeyer' email@example.com [E-LeadScrew] wrote:
Now a Servo. 1. Max RPM is usually 3000 RPM and torque is good all the way to the top. 2. Aiming for 750 RPM means a 4:1 belt reduction. a. 4:1 on the belt though results in 4x the encoder resolution. b. With a 100 line encoder in quadrature you get 400 lines per rev x 4:1 or 16,000 steps per rev. c. Accuracy is still 16,000 steps per rev. 3. Stay with a 4 TPI ball screw and 0.2” pitch we resolve (and position) down to 0.0000125” per step. a. But at 20,000 steps per second from the ELS you can really only turn the lead screw 1.25 turns per second b. That’s 0.2”/rev x 1.25 revs/sec = 0.25”/second (0.25 IPS). A factor of 10 too slow. 4. I think Gecko has a step multiplier that turns the servo 10 lines for every step in. a. Now we’re at 2.5 IPS which was the goal. So: A DC servo Motor with 100 line encoder on a 4 TPI ball screw results in a position resolution and accuracy of 0.000125” with a top speed of 2.5 IPS and a step rate input of 20,000 steps per second. Please correct my math if I’ve made a mistake. John Dammeyer
I think its a fact that microstepping, where you are actually stopping at the microsteps (i.e. using them for resolution) definitely is less torque, simply because the motor cannot hold that position accurately against load. In a dynamic situation however, lots of factors come into play. For example, if the axis is in constant motion the only torque required is to overcome friction and cutting loads, there is little torque required to provide acceleration. Unless the motor is being operated close to its torque limit (at that speed/volt/current combo) then microstepping should have little impact. The general rule I have used is in the spreadsheet is that the motor should provide 3x required dynamic torque at the maximum speed.
Large motors have high inductance so the torque drops off very fast with speed – the corner speed of those motors is 240rpm. I don’t know how big your axis are, but I’m guessing its going to be around 1 – 1.2m? With your 10mm pitch screws 2.5m/min = 250rpm, so that is close to optimal (and 1/8 stepping = 6664steps/sec) and it looks OK at cutting speeds, but its very marginal at 10m/min rapids and that is where you may have lost steps (=27000steps/sec). You need to reduce rapids to 7m/min but it should be OK at 1/4 or 1/8 stepping.