Would anyone consider there to be a speed or accuracy advantage to using some form of encoded feedback for the linear movements of CNC machines like there often is for the rotational aspects? I know pulsed laser rangefinders aren't accurate enough at short distances, but perhaps ultrasonic transducers could be used. Alternatively some kind of laser/linear CCD triangulation setup. I've seen some people use the resistive tracks on digital calipers/rulers, though that seemed to be as a safeguard in development rather than an intrinsic part of their machine. Given that brushless motors operate rather like AC motors and not entirely unlike stepper motors, but deliver far more power at lower prices, are we all going about this the wrong way by trying to assume the driver's behaviour rather than actually measuring the driven's? Can we massively increase the precision of off-the-shelf parts by implementing feedback systems? Like the micromachining thread, I'm not necessarily coming down on either side, just looking for discussion and opinions.
Hi Rob, I think you are absolutely right - servo driven machines incorporating optical scale feedback can be emulated using steppers and linear glass scales to complete the feedback loop. The overall cost would be considerably higher than just a stepper driven system but it may be worth it for some applications. However, as long as a stepper driven system doesn't loose or gain steps then there would be little or no advantage. Tweakie.
This is true, and what I'm wondering about. For a screw-drive system, does it serve any purpose whatsoever? Could I go from 1/100mm to 1/1000mm accuracy? Or cut times in half (perhaps through faster positioning rather than any cutting speed increase)? Or is it just insurance on belt-driven systems? Do servos have any intrinsic advantages over steppers beyond convenient short-distance/rotation positioning? I can't think of any reasons why to do any of this, but the strips are sold (even at Sparkfun!) and are used in industry, so there must be some kind of reasoning. Perhaps per-run calibration? If the potentiometer strips are a known quantity, the system could self-calibrate. Though you could do the same thing much more easily with lasers or contact switches. It's rather odd.
Hi Rob, Servo driven systems can be operated much faster than stepper driven systems and of course, do not suffer with the lost / gained steps issue. They do have their own specific issues such as lag etc. and tuning is an important part of their setup. On my system, for some applications, I drive my steppers at 900 rpm giving a feed-rate of 4500 mm/min (via ball-screws) with 100% reliability but at this speed they have very little torque. Just fine for laser work where there is no tool loading but if the system was servo-driven it would still have useful torque at that speed (assuming I had a need for it). I can, however, foresee future applications where a feed-rate of 9000 to 14000 mm/min would be useful for me so who knows, one day I may change over to using servo drives. Linear scales used in a feed-back loop certainly overcome some main issues regarding precision. The aforementioned lack of torque at high speed, variations in lead-screw (or rack and pinion) pitch and also belt stretch. Back-lash could still remain an issue as the motors may ‘hunt’ to find position but this is where the tuning comes into play. So overall I would say yes you can expect to improve accuracy but only to the extent of the least accurate component of the machine (a chain is only as strong as it’s weakest link). Go for it, it’s only money. Tweakie.
I'm looking at ball screws, too, because I don't want to deal with ACME/trapezoidal screw drive wear, which apparently happens all too quickly, even with brass nuts. Zero-backlash (or should that be "zero"-backlash?) ball screws aren't actually too expensive for what they are, it seems. I'm looking at SFU2005s in 500 and 200mm lengths, which seem to be a good compromise between accuracy and price, though I'm planning on doing some more research along those lines. There may be other factors like holding force, screw angle or load values/diameters that I've missed. Hmm, interesting. I'm having difficulty finding continuous servos other than the "high torque" ones which just look like converted Futaba 148s. I can't see that being sufficient for my purposes, which- with a 1.5kW spindle- may include some small-scale (sub 50mm square?) mild steel machining. Apparently just a few kilos of force is sufficient for a 1mm carbide mill to cut into steel, so I guess I need servos that can both turn a screw drive with sufficient torque to convert into around 50N or so of linear thrust. That, I'll need to look into (maybe multiple lower-torque servos, who knows). I guess the screw diameter will be a tradeoff between force conversion and accuracy too. Know of anywhere that sells reasonably-priced (under say, $80?) low-end industrial servos? Trying to eliminate accuracy errors is what I'm looking at. It may simply be a case of sticking with steppers, but creating some kind of per-run self-calibration with lasers and photodiodes and some careful measuring- I'm not sure I trust the regular limit switches over the 500mm/200mm that I'm looking at. Perhaps on a laser machine which is naturally much larger the error is acceptably less per unit length. Variations in pitch? How much is that a thing? The others shouldn't be so much an issue. (This is a fascinating thread, btw, both what you're putting in and what I'm having to research to continue the discussion. Thanks!)
Hi Rob, My apologies if I gave the wrong impression, by "servo's" I did not mean RC servo's but rather the CNC machine type servo's http://www.goodwinelectronics.com/servos.htm Tweakie.
No, I've seen them before, but when searching for "continuous high torque servos" mostly RC servos come up. I'm just trying to find an industrial servo site which doesn't require me to call, email, request, quote, or write a short novel in order to just get a price list though! Probably eBay's my best bet. I'm gonna have to look into that screw pitch variability thing though, that may put a dent in my plans, and I don't want to pay $800 per screw (saw one of those last night! ) Edit: Looks like servos are impractical for the moment, I think I'll be sticking with steppers! Linear calibration may still be a concern though, we'll see. I may upgrade to NEMA 34s, too, I'm not sure yet. Depends on the drivers, probably. I'm trying to optimise price, power and accuracy when I should really just be picking two. But compromise was never my strong suit.
I did just have another idea though, but it might result in too-jerky movement: Using a laser that can provide very consistent power (+/-0.01%) and a highly linear coefficient (99.9%+) photodiode, mounting one on the frame and one on a gantry should produce a near-perfect inverse-square (does that apply to lasers? Anyway, some calibrate-able graph) signal from the photodiode which could provide excellent feedback for a highly geared DC motor instead of a servo. The issue would, I suppose, be stopping the movement in a timely fashion, though an algorithm that slows the motor in proportional steps to how close it is to the target position could possibly work. Random idea out of the blue just now, but I think it could potentially be interesting- at least for some applications. Edit: answers to laser power with equations: https://answers.yahoo.com/question/index?qid=20110214073235AAxrBgk (TL;DR: collimation over a sufficiently short distance will produce no discernable difference in laser power, so no go with lasers unless you have an appropriate power divergent lens- might be easier to stick to LEDs!)
Another option I've been looking into: Hall Effect sensors and linear arrays of tiny (~2mm) magnets. A simple relative encoding with end-stops would potentially be fairly cheap to create.
Hi Rob. Try this site for Servos. https://www.motioncontrolproducts.co.uk/sections/4/servo-motors/ Gray
Still ludicrously expensive for more than one axis, but useful for future reference. Thanks! Definitely worth looking at along with NEMA 34s, I think.
Yea. Proper Servos don't come cheap, that's for sure! I got my Nema top end 23's from there. Good price, and just down the road, so no postage. Those 34's look real meaty don't they. Gray
Going with magnets might be best, if not working with metals ? Laser/LED route could get false triggers with the chips/dust/smoke generated while machine is at work ? Would having dual shaft steppers help ? The 'rear' shaft could be shielded from any interference (chips/dust/smoke) and spin optical encoded disk to determine speed, direction, expected position, etc. Skipping steps would still be a problem since we are not looking at actual position of tool tip. Going to a fixed reference point (datum) from time to time would probably be required in any case.
I got dual shafts because you can then add heavy rubber dampers, which reduce resonance and vibration, to a certain degree. They also have a quietening effect as well. Failing that you can always put little flags on them and watch them go round. This Guy, http://www.neo7cnc.com/ who initially got me into the idea of building my own CNC, although he is into the older, much heavier type of CNC, covers lots of items like the dual shaft thing as well as many other bits and bobs. Watch some of his videos, there are lots to choose from. Gray
Or one of those disks with spirals and get hypnotized by your own machine Did watch some of his videos early on, and again as you reminded me of them. They are quite long with lots of chatter ... too much like some of my posts, eh ? A video gets kind of boring after a few minutes of talking and no action, like the old lectures in class. Maybe it's just my ADS.
From a old guy, To make a servo system fully run to it's potential you must have a dual trace 0-scope.. one channel on the "signal" and one on the response out to tune it. Steppers are mucho easier to tune in.. I got a trio of Sanyo servos, three gecko servo drives I bought about ten years ago, I keep saying I am going to bolt them onto something again.. but..
A scope is something I've really wanted for a decade, and actually needed on a few occasions over the last year or so. Steppers are definitely easier, for sure, but positioning seems to be the key to unlocking higher power, speed and accuracies. I have relatively little interest in the whole "gantry router" thing. Not to say I may not pursue that as an early option, but it just doesn't seem interesting to me. My ultimate goal is to make $50-100k machinery possible on a $2k DIYer's budget. To be able to machine really useful stuff, not just signage, decorations and flat pack furniture. Their cost is all in the precision and repeatability of the drive systems, the accuracy of the feedback mechanisms and the quality of the path-to-GCode-to-steps translation system. For the former I just started looking at roller pinions since they seem to be a pretty good alternative to ball screws unless you need crazy accuracy, but 20um seems to be pretty high on a hobbyist level and roller pinions'll do that if you have a finely graduated enough motor to drive them. What I've been looking at up until recently is in my Big Thread O' Linear Motion. Feedback is obviously what this thread's all about, and the GCode conversion is my other thread in Interfaces. I'm content to do the research for a while since I don't have the funding to put any of it into practice yet. I was actually looking at feedback this morning, discovering that even a regular mass-market CMOS sensor wouldn't be enough pixelage and physical width to really justifiably do a good implementation of trignonometric positioning. Practically the only way to go with lasers would be my very first idea when I created this thread, some kind of phase-shift system with a split prism, and that's a lot of effort to get a way over the top level of accuracy and no actual positioning, just a relative motion "clicker". Not that I'm intrinsically opposed to that, but it's not on the cards for now. TL;DR- The relevant part to those only looking for workable positioning ideas: What I did discover this morning though was something called a linear variable differential transformer, where the motion of the machine controls the position of an ferrite rod relative to three coils. The primary is in the centre, and there are two secondaries wired in such a way as to either cancel each other out, or provide two separate voltage outputs to a comparator. As the iron core moves back and forth, the flux linkage between the primary and any one of the secondaries improves and worsens in opposition to the other secondary. Hence the "differential" part of the name. LVDTs are accurate over several inches, apparently, but I imagine with the right shape, consistency of winding and ferrite rod could be accurate over a couple of feet. They can be sealed if necessary, and are apparently already used in machine tools. They're immune to power loss, provide easy absolute positioning, etc etc. There only appear to be positives, other than the fact that they apparently only come in relatively short lengths. I haven't looked into them much, only Wiki and one random linear positioning article, but they seem like an ideal option for machines under, I guess/hope, about two feet. The length they can handle may depend on other factors like voltage and stuff, I don't know. But I'm pretty excited, anyway.
