Belt Drive Transmission Plans

[rb99]

Are there any plans available? I have been reading some older threads and they might be my best anti-backlash solution...

[bleeth]

I have some-Send me an e-mail or PM.
Looking at your other threads you should note that the belt drive solution is intended for direct drive motors and not geared.One of the first SB conversions using a gearbox was done to a first generation Alpha which had 1:1 motors and a serious chatter issue and others have used it with new 1:1 motors on Standards.
The design I did is for new direct drive motors and gears them at 3.6:1. If you are using geared motors then you are into a whole new set of calculations. My design was done to maintain use of SB Standard 25 tooth pinions on the x and y and the unit values built into SB software.

[gerryv]

I don't believe that belt drives can eliminate any backlash that already exists in your motors, they can only eliminate any additional backlash that a gearbox might add. Check with others but you might be wasting your money in doing this.

[rb99]

Thanks guys. Yes I am talking about gearbox backlash... I would have to remove the gearboxes. My motors (gearboxes) are not the TH type. I am not sure if I would build them, but threads from 2-3 years ago make a good case for them.

[gc3]

http://www.cncpartskit.com/belt_reduction.html

[Steve M]

I have been using a belt drive on my Y axis for the past 4 years with a tiny NEMA 23 motor. There is still some backlash due to belt stretch, but it's not as much as the original geared motors. Ratio is 6:1.

[bleeth]

Steve:
Misumi carries belts that are Kevlar re-inforced and shouldn't stretch like unreinforced ones. You might see if they have your size. Price difference is minimal.
Gene: I've looked at those boxes on-line before and there are a couple things that bother me about them.
1. The shaft is totally round making it problematic to prevent spur gear slippage.
2. The gearing ration means all new Unit Values to use with SB software (although they can be determined).
3. The input side is designed for their motors only and I would rather stick with Orientals.
4. The plates are kind of thin.

The ones I designed the other year use 1/2" plates which just won't bend, have a machined shaft with keyway slots for all components, and are designed for mounting Oriental motors. The belt is easy to change out or adjust if/when needed and the shaft rotates on 2 bearings (one on the end and one between the gear pulley and the spur gear and the bearings are press fit into the inner and outer plates so the shaft remains dead true.

Richard: All new motors from SB have come with Tabered Hobb gears for a long time now so gear backlash is greatly reduced. The Alpha motors are off the shelf but the Standards are strictly made for SB and can only be bought from them. You will note that the entire threads about gearboxes are related either to older Alphas with 1:1 motors or standards that get new motors (PK 99's) as well as the gearboxes, resulting in NEAR Alpha like performance (High Torque and high speed as well as improved resolution). The thing to remember though is that the older design of the PR and PRT chassis' can't handle a whole lot of increased speed and torque without the gantrys twisting. This is what led to the PRS chassis design in the first place.

Let's face it guys-If you drop a hemi into a 32 ford without redoing the chassis the thing will self destruct the first time you punch it!!

As some of you may recall when I put the first Gecko controller on my PRT I could only jog faster, but cutting speed was only marginally improved. With the PRS upper chassis rebuild my cutting speed doubled.

[gerryv]

Very well said Dave, it's very much about balance and matching isn't it. No magic bullet. It sounds to me that between some of the excellent advise Richard has received from the real Guru's here and some of the good buys he's already made, that the already has the makings of a pretty danged good shot at squeezing the most out of his machine and stands a very good chance of being pretty thrilled with it's improvement. I base that opinion on a number of years of studying others knowledge and experience like yours, having had the time to do that. I don't think he needs to spend any more money yet, just put together his major upgrade, fine tune a little and SMILE.

I should also add that from my metal days with lathes and mills, backlash was really more of an issue when there was a great deal of "reversing" on typically much smaller and less forgiving parts, such as when cutting tight threads. I'm not so sure it's that much of an issue here.

[rb99]

Dave I bought a 3.6 from OM for $257 shipping included. The SB motors are $400 plus shipping. Last week when I ordered I had forgotten about the TH from SB. But for me to buy 4 new motors at $400 each plus shipping etc. would be close to 2k. I have spent my upgrade budget for now. The new controller, computer, Z axis and motor, dust foot, e chain, pinions... and I will still have to change out the rails...

I am not looking to go so fast I twist the rails, rather I am looking for the ability to cut small circles, letters and detail without deformity from the backlash or play coming from the steppers.

[gc3]

here's a shot of reduction box on my second machine

[danhamm]

More than one problem is solved with belt drives, Motor gear vibration marks reduced also I have seen no backlash using belt drives unless they were really loose, it also adds a operational cushion to the drive system.

[rb99]

here's a shot of reduction box on my second machine


How big are the motors on that machine compared to your SB? That is belt drive? The rack looks huge!

[gc3]

nema 34 listed as 906 oz/in with 3.6/1 box

[bleeth]

Cutting very small shaped parts accurately and smoothly with a Shopbot is one of the most challenging things to ask it to do. Larger machines that do it well are very very expensive. Among smaller machines the SB desktop is designed for it using servo's and ball/screw drive.
Be careful about thinking that a machine running Mach is the answer. Like different models and configurations of SB's, what a machine can and can't do varies greatly from manufacturer to manufacturer and model to model. Based on my researc, to have a full size flatbed cutting very small parts (1/2-2" each) with a very close size tolerance as well as really great edge resolution you will be paying some very big bucks.

[rb99]

OK I submit that the software backlash is not that great... an email from Ted about it today:
__________________________________________________ __________________________________________________


Hi Richard,

Happy New Year!

And thanks for the suggestion. In fact, we used to do backlash compensation in ShopBot – there is even an obsolete command for it [VK].

Unfortunately, it almost invariably created more trouble than it was worth. Indeed, as the guy did in the video, you can tune a particular correction to work in one situation within a simple piece of geometry. Unfortunately, in any complex cutting scenario, we found it was producing distortions as often as it was correcting, and that it could contribute to a tool losing position. In any case, the larger problem with excessive backlash is that it can contribute to poor cut quality – and this is not affected at all by backlash comp.

What we’ve tried to do instead is to keep backlash to a minimum. That usually means good/new pinions and grease. As well as trying to get rid of any source of flex or looseness. Generally speaking, in most applications that ShopBots are put to, there should not be enough backlash to produce significant errors or obvious flats in circles.

To test for backlash, drive an axis to a position from one direction, put a dial indicator up against it, and see how large a move it takes to move the tool away in the opposite direction (how much “take-up” is required). Try it in a couple different places. This is typically < .004” for a new gearbox.

As tools get older and gearboxes wear, they can develop greater backlash. We shifted to low-backlash, tapered-hob gearboxes a number of years ago to help minimize this. But even these boxes will wear over time. To a degree, they can often be brought back to minimal backlash conditions by adjusting the gear relief … there is some info on the ShopBot forum about doing it – but this does take some fussing with the gearbox.

The biggest issue that I have come across related to backlash is that with very small cutters, backlash can become a contributor to bit breakage. A solution to this for fine cutting that does not involve a large cutter load, such as doing small inlays for a guitar neck, is to preload the X and Y axis in one direction (I have seen it done creatively with bungee cords). You don’t need to worry about the Z because the weight usually creates a preload, if the springs are not too strong.

Regards,

=Ted
__________________________________________________ __________________________________________________

Does anyone know where the info on adjusting the gears is, and if there is a way to adjust the regular gears?

[richards]

Before spending a lot of time and expense fixing backlash, it might be a good idea to look at any cumulative errors that will still exist after correcting the backlash. If those "standard" errors would cause parts to be rejected, then fixing backlash would not give the results desired.

Years ago, when I owned and operated a photolab servicing professional photographers, I bought a $45,000 Video Color Negative Analyzer that allowed me to look at color negatives and adjust settings so that those negatives would print correctly without making test prints. When I complained to the Kodak tech representative that I was still wasting too much paper and chemicals, he reviewed the "circle of confusion" with me. That circle of confusion included all possible variances that could occur in the printing cycle. It included chemical temperature, variation in motor speed in the processor, variation in the voltage to the print lamp, variation in the time that the shutter and color filters took to open and close. In effect, he showed me that the "circle of confusion" was twice as large as my desired "target". No matter what I did and how closely I monitored the entire process, there would be a very good chance that the "gremlins" would act together to cause an error greater than what I considered acceptable. That was with a system that cost very close to $200,000 total. It was the best system available to independent photolabs at the time.

Make a chart and compare each major component on the Shopbot with components that other machines use. For instance, compare the V-rollers and rails with linear rails. Compare rack and pinion with ball screws. Compare stepper motors and gear boxes with stepper motors and belt drives. Use an Oscilloscope and compare the pulse train generated by a Windows computer to the pulse train generated by a $20,000 "real-time" computer. You'll find the "circle of confusion" for your machine. As long as you cut parts whose quality fit within your "circle of confusion", you'll get excellent results. If you try to cut parts that require resolution finer than your "circle of confusion", you'll never be satisfied.

I did a lot of complaining when I first used my PRT-Alpha. Even after I finally learned how to use the machine and reduced the error rate substantially, I still had too much "slop". Adding a belt-drive reduced that "slop" to acceptable limits for 95% of the work that I hoped to do. I simply stopped trying to cut parts that required quality (resolution and repeatability) finer than I could repeatedly produce.

The Shopbot is a wonderful machine, but it can't be expected to cut all types of material with accuracy of 0.001". It wasn't designed or built to do that. Very few machines, at any price, could do that, and if they could, yearly maintenance would cost more than buying a new Shopbot PRS-Alpha.

[rb99]

I agree with what you said... However!

The one thing that fixed you to 95% satisfaction was getting rid of the backlash in the motors.

I really see the gearboxes as a glaring flaw in the whole thing.

I can stiffen my rails, get new pinions, make sure the motors are tight against the rack, square the machine, buy a 4g controller, buy and install a retro Z, move to a new PC and 3.x windows control software etc. But the backlash in the gearboxes is there before they ship me the motors and it won't go away unless I go to belt drive.

I notice many people put a caveat at the end of their posts about how the SB won't get 30 inches per second or accuracies to .00001 etc. All I want is decent small circles and letters but the inherent gearbox backlash makes that very difficult if not impossible.

You made your own gearboxes? Did you get new motors or did you adapt the gearbox ones?

[rb99]

I may have spoke too soon about using Mach's backlash compensation. Here is a post about backlash from the MM forum:

pblackburn (http://www.mechmate.com/forums/member.php?u=19759)
Just call me: Pete #98
Pennsylvania
United States of America



I know this is an older thread but I to have seen the backlash in the PK296A2A-SG7.2. All of mine have between 9.5 and 10.3 thousandths. They probably have less than 90 hours runtime on them with no crashes. I have not noticed any problems with circles or carvings but on complex carvings with 3d models inset in a dish you can tell there is backlash. I measured it, and used Mach's backlash comp to eliminate it. I think the PK296A2A-SG7.2 are a great motor for the x and y but not for the Z. Just my 2 cents.


So it just might work?

[myxpykalix]

Richard,
I've been following this thread loosely and i get lost in some of the mechanical details but i'm wondering what you are cutting that you need such accuracy for?
In theory, you're right in wanting the best resolution you can get and maybe you can see the differences in tolerances but do your customers have the same requirements for such tolerances?

I had a fellow shopbotter come here once and we made some test cuts on my machine and compared them to another machine (I think running mach3) and you could tell the difference in resolution, but i'm not sure the average woodworking customer would see it or care:confused:

[rb99]

Hello Jack,

It's all in the details! It is just not cool to have play/backlash before even getting off the ground. If I want to make parts for people, do I say they will be OK, but any slots or 15/32" mounting holes won't be the right shape, they might be deformed so your part won't work. Or if I want to make signs or plaques any small letters may be deformed... If I need to plane a background flat it will not be flat... Or I carve something and it takes hours of sanding and fiddling to get rid of erroneous ridges and tool marks.

[steve_g]

I came across this tonight... it's from a furniture factory with "big iron" machinery

CNC panel (beam) saw: 0.4mm (+/-0.2mm)
CNC router: 0.2mm (+/-0.1mm)
Wall saw: 1mm (+/-0.5mm)
Table (panel) saw: 0.5mm (+/-0.25mm)
Spindle moulder*: 0.5mm (+/-0.25mm)
Shaper*: 0.4mm (+/-0.2mm)
Edgebander* 0.4mm (+/-0.2mm)
"Festool" rail-guided saw: 1mm (+/-0.5mm)
SCMS: 1mm (+/-0.5mm

I thought it interesting that they only expect about +- .004" from their CNC equipment. As a High School shop teacher before CNC was even a "word" we expected +- 1/32" precision but were excited when they got 1/16"! Even then we still expected +-.001" from the metal shops... Later in life as a contractor, I was excited when summer help knew what the "little marks" on a tape measure meant!

SG

[richards]

My PRT-Alpha came with direct drive motors and 20-tooth pinion gears. The stepper driver was set to 1,000 steps per revolution. Doing the math, I should have gotten 0.0031416" per step, which is roughly equivalent to the thickness of a piece of copy paper. I did not get that resolution on my machine.

A 4G controller has 2,000 steps per revolution, which should give 2X the quality of the Alpha controller. It does not (on any machine that I've seen in operation).

Adding a 3:1 belt-drive tansmission to my PRT-Alpha reduced "chatter" substantially. Cut quality was acceptable. Life was good.

I bought the 7.2:1 motor/gearbox that Shopbot offered as an upgrade. There was no visible difference in quality of cut from my 3:1 belt-drive transmissions, but the motor/gearbox was substantially smaller.

Looking at the cause of excessive "looseness" in the machine, it's easy to see some of the culprits.

1. The machine flexes. Solution: add bracing or weld the joints to eliminate as much flex as possible. (Big iron machine have massive amounts of metal to reduce flex. The Shopbot has less metal and more flex.)

2. The V-drive system allows some movement in more than one direction. Solution: replace the V-drive system with linear rails - at a substantial cost.

3. The stepper drivers are less than optimal. Solution: buy the 4G upgrade or build your own Mach 3 based controller using Geckodrive components, but be prepared to re-write your software if you use Mach 3.

4. Replace the router with a spindle to reduce "run out".

5. Use a cutter that is sufficiently "stiff" to handle the load. (I consider a 3/8" cutter to be the minimum size to reduce "cutter flex".)

6. Maintain the machine properly. Keep all moving parts cleaned and greased. Replace all worn components. Follow recommended warm-up times. Securely fasten all material to the table before cutting.

Sometimes poor cut quality is the result of a number of additive factors. Be sure that YOU have done your job before blaming the machine.