Troubleshooting and Fixing Backlash in the G0704 CNC Mill

[@JasbirSingh-iw9oz]

This is absolutely amazing. 3 years in community college and they didn't even mention backlash leaving alone teaching me to troubleshoot it. Thank you

[@beachboardfan9544]

Great vid! This is the kinda stuff you never see cnc channels discuss, its always, 'look at this new machine that has zero issues, see how awesome it is, learn nothing'! Trouble shooting cnc issues is such great knowledge, thanks for this!

[@TurbineTed]

I designed my CNC conversion software (Arduino based system) to use feedback from the DRO linear encoders such that the stepper motors keep driving until the error between the desired position, from a GRBL CNC controller, and the actual position, from the DRO, is zero. Therefore no backlash. This means that I use the original leadscrews and nuts supplied with the mill and no expensive ball screws. This method has been working really well for the last 5 years with no fancy setting of gib's etc. Maybe worth thinking about with your programming skills (ELS for example) as opposed to my very amateurish inefficient software.

[@philiprisser938]

I was thinking about doing the exact same, do you have your code and interface documented anywhere?

[@TurbineTed]

@@philiprisser938 Hi Philip, I'm afraid I don't have the code etc. on the internet anywhere, just on my pc. I don't have any problem in giving you my stuff but not sure how to do it. Any suggestions?

[@livewire2404]

Hello, interesting implementation. That sounds like a closed loop system. Did you modify the original GRBL code?

[@TurbineTed]

Hi, Yes you are correct it is a closed loop system. The GRBL code was not modified (to be honest I don't understand a lot of it) I just use the step and direction pulses from the GRBL controller and feed them into an Arduino Mega. The 'step' pulse of the GRBL is set to be the same as the resolution of my DRO i.e. 0.005mm/step and it drives a 'set point' in the Arduino, this is compared to the DRO position which produces an 'error'. The Arduino then drives the stepper motors to reduce the error to zero. It is a bit slow, my max table speed is approx. 250mm/min before the steppers stall but I think that's down to a combination of too small stepper motors (NEMA 23) and a relatively slow Arduino. Still it does everything I want it to do with very good accuracy, + or - 0.005mm, and it's been doing it perfectly for at least 5 years now. One of the demo's I like to do is to physically push the table away from it's position with the backlash (approx. 0.1mm) and it immediately bursts into life and returns the table to its previous position, very satisfying. Thanks for the interest.

[@livewire2404]

@@TurbineTed I didn't quite get it. You have a second arduino which does the closed loop and it reads the DRO? From where the stepper driver gets it's step input? Maybe you can do a video or a schematic.

[@taranson3057]

James, I really like your approach for checking your machine for backlash. Chasing backlash issues can be time consuming and can also be rather frustrating especially for folks like myself.

[@routercnc9517]

When I built my current CNC machine I put double ball nuts on X and Y with a fixed preload. I measured no backlash when I did that. On the Z axis I have linear rails so they move smoothly and the weight of the Z axis is enough to effectively have zero backlash. So either a linear rail conversion on Z or try fitting another second ball nut with a bit of preload. Those ballscrews typically have 50 microns backlash so if the Z axis is getting held up on the gibs you will be fighting that. Good run through anyway. Regards from the UK 🇬🇧

[@IBWatchinUrVids]

I fought with this as well, in my case, the aluminum ball nut mount was rocking on the slide plate. Solved by pressing hardened pins into the ball nut mount so it had a more stable surface. Funny to watch someone going through the exact same struggles as I have.

[@44mod]

Nice video!! Great closeups on indicators. Thanks James!!!

[@DougsMessyGarage]

Good that you were able to adjust things back to where you want them

[@Clough42]

I can live with what I have, but I'd like it to be better. The slip-stick in the dovetail ways on Z is tough to overcome on a machine like this.

[@stephenjohnson6841]

Antler great video! Very well done and explained as usual! Thanks again James!

[@TheAyrCaveShop]

Man that was a lot of detective work...glad you got it sorted !!

[@GraphicManInnovations]

The weight of the spindle virtually eliminates any backlash (zero on my machine) surprised that is not the case with you

[@airgunningyup]

went through this once when my cnc started cutting ovals.. Went through everthing , never solved , and eventually decided to build another machine .. it can be a nightmare hunting these things.

[@taranson3057]

If everything appears to be tight you may want to look at your lead screw nut or ball screw nut specs there could be back lash in those. A high quality system will provide the specs that you need, Chinese systems very rarely if ever provide you with the information, and the quality is rarely what they claim it to be. I’m not saying they can’t make decent product, I am leaning more towards you get what you pay for.

[@dadoVRC]

Those issues are why I decided to convert mine with linear rails!

[@dwightcarlson7136]

Nice to see you made progress. Great instructional video showing logical/step by step analysis. 👍👍

[@nickw226]

perfect time to do linear rail/way conversion on the z axis

[@Clough42]

It's only time and money. When I did this conversion, I had more of the former. At this moment in my life, it's the other way around. Not sure I'm up for building completely from scratch.

[@visionforge1254]

@@Clough42 After a while I got tired of chasing backlash on my G0704 conversion and tightening loose items on the ATC and mill. Ended up buying a used 5 axis Robodrill in good condition for almost the same price I put into the conversion. I don't think I'll buy another machine without an ATC or do a conversion again if I'm serious about making good parts.

[@Deezle_Gaming]

Our CNC converted Bridgeport at work has some Z axis backlash, the yoke that attaches to the quill is not a great design and there's not much you can do about it. I ended up dialing in some backlash compensation and it's good enough now.

[@chrisarmstrong8198]

It would have been interesting to also measure the Z backlash obtained by manually pushing the head up and down (i.e. without turning the motor) to see if gravity and stiction are affecting the tests.

[@Clough42]

I'm sure both are. It's better with the 28lb spindle than it was with the original 70lb head, but it's still an issue.

[@AmateurRedneckWorkshop]

I think that slack is inevitable just like taxes. Thanks for the video.

[@taranson3057]

Now, that there is funny ARW

[@my1956effie]

I pick up a new tip with every one of your videos I watch.👍

[@glenntrewitt]

[hypotheticals] 1) Suppose that (a) one or both nuts are a little loose and (b) the leadscrew isn't exactly parallel with the travel of the carriage. At one point, the nuts will be as loose as possible - maximum backlash. As you move away from there, the side force on the nuts will increase, possibly(?) forcing out some of the slop. 2) Or that the gibs are unevenly worn, leading to the same sort of effect. ?

[@cncart]

Hi how do you change your test indicator contact screw angle? Thanks.

[@chrisadesigns]

Since you have the other nice big mill, it might be time to dissamble the G0704 and do a linear rail conversion .

[@Clough42]

Yeah. Maybe not a bad idea. Also servos. And better ballscrews. And maybe a better mill. You know, a Tormach 1100MX would fit in basically the same space in my shop...

[@chrisadesigns]

@@Clough42 yeah I guess but tormach is still just a dovetail machine and I personally am not a fan. And for the price of tormach, there are better options but it's your choice

[@MakeTechPtyLtd]

@@chrisadesigns yeah I can't get over the dovetail. Obviously it works but I keep thinking of it as WW1 technology.

[@michaelalcocer6778]

@@Clough42 Haas has a TM-0 coming out in Q3. Same size and they are shooting for the tormach price target.

[@m3h5l5]

Random question, but how do you go about oiling your ways and greasing your ballnuts? I'm in the process of converting my mill and wanted to know if it was necessary to plum oil or grease lines. Thanks

[@ericsandberg3167]

James....is the leadscrew (ballscrew and nut) metric....if that is the case, does the S/W compensate for this so when you command a move of .001 the stepper or servo moves the exact number of degrees or micro steps...????

[@paulmanhart4481]

Hi James. I like your Grizzly mill. Do you have a video on how you converted it to a 3 axis CNC? Would I be better off with a PM 728 VT and starting with that? Thanks for your time and videos. Paul.

[@susurrosjoe6817]

Hi James, what are these types of couplings called and where could they be bought? Very good video and explanation.

[@joels4208]

Hello James. Great Video! Very informative. Quick enclosure question. Do you ever think you need to put something between you and the part for safety. Like Lexan?

[@stargatefred]

I am currently trying to convert my G0704 to CNC. I am very slow at it. Maybe by the end of the year. Bill from Seattle.

[@tcratius1748]

Ok, just for the fun of it, I'm going to guess before you figure it out, and my guess is you have a worn washer or spacer. Otherwise, it could be the ACME nut. And I'm already wrong because you have a ball screw and nut. Well played both interesting and educational.

[@CraigVanSickleAK]

Great video, thanks for the info.

[@durangotang1681]

I found that if I helped mine along after I finally got my rpi down, that the stick slip from the gibs was the problem. If I gave the table or head a little help In whatever direction I was jogging it, it went right to where it was supposed to. And if I forced it past where it was supposed to go, it would spring back. Didn’t seem to effect it when it was actually cutting. But I was pretty stoked when I got hit hitting to the tenth. Still spring tension for several tenths when initiating a move. Almost drove me crazy. Took a month of whatever free time I had to get it perfect.

[@thebotformalityknownasdale2564]

James would it be worst case be a combination of all the posabilitys ? so a sloppy line in the chain for the z axis . And would removing as close to 100% of back lash not induce a premature ware if so can you program in compensation and if so I'm thinking it's best to spread it out over the hole line of the z axis ? Or just at the ballscrews ? Sory I know I was a pain in the a*s to my professors too .

[@SecurityisaMyth]

This may be a foolish question...is there any possibility that the canted spindle impacted the accuracy of the measurement?

[@woodscreekworkshop9939]

Many books and movies have been written about that last thousandth

[@TheAyrCaveShop]

and loose nuts....

[@stargatefred]

Hey Yuchol are you home now?

[@michaelalcocer6778]

Are you 100% sure the z ball it mount fits well in the pocket on the backside of the slide? Mine never did. Also make a real bearing block for the z ac bearings.

[@AlmostMachining]

Nice go through on this James! Do you have control backlash compensation turned on if it exists in that software?

[@forgeperformanceand4x4]

Most of the backlash comp changes the way the mill with move and can cause more issues.

[@ADBBuild]

Mach 3 does have backlash compensation. I believe all it does is add the amount you specify to the amount of movement, but only when changing directions. That way the backlash is taken up on the first move.

[@thetruthyouneedtohear]

You should put a L Hex Wrench on those nuts. You will see that with a T Handle it feels crazy tight, but that is due to poor leverage. Put a L Wrench on it, and you will be able to get it tighter. Also, loctite the screws.

[@forgeperformanceand4x4]

Can you get rid of the dove tail slides and mount hgr rail onto the slide ways and go that route?

[@Clough42]

It's all just metal, so yes. Is that practical? Don't know. Maybe.

[@forgeperformanceand4x4]

@@Clough42 I would think milling off the dove tail and using the main base could make for a really interesting video. Would eliminate all head nod too. Pending results could open possiblitys for doing the X and y axis though you may loose z travel due to heights. But could you then make up the travel with extended z liner rails. Into the deep end fast.

[@Clough42]

@@forgeperformanceand4x4 yeah. I'll have to think about it. That does sound interesting.

[@ikbendusan]

this is exactly what i'm doing to my machine

[@gn6034]

Why don't you use the Mach3 backlash compensation feature

[@SteelCogs]

You're supposed to eliminate as much mechanical backlash as possible before using backlash compensation.

[@ikbendusan]

i've read that it doesn't work for milling operations in mach3

[@jonathanlausberg5834]

I think that the backlash problem in the Z axis may lie in the friction of the dovetail slide.

[@mrechbreger]

The nut of my one was also loose .. I just fixed it last week. The entire head was jumping a little bit from time to time.. that's completely eliminated now.

[@artmckay6704]

Thanks for sharing! :)

[@Engineerd3d]

James any plans for scraping the ways of that mill?

[@terryallemann2654]

Just to help shorten time working on chasing backlash first mount test indicator on nut or as close to nut put indicator tip on screw pitch move back and forth any movement in indicator will be lost motion in ball screw over .0015 replace ball screw or have reground if found good then work from there out nut mount screw mount bearings, gibs and so on. Next you need to make more of a movement you will suffer a lot from slip stick move down and zero out on a 1 inch gage block then move up 1 inch and slide a 2 inch gage block under indicator should read zero what it is off is actual lost motion.

[@Clough42]

Good suggestions. In my case, I think most of the backlash is slip-stick causing the nut mount to flex. Not a lot of easy solutions, unfortunately.

[@dannirr]

4 axis?

[@Hossimo]

I bet this video represents 90% of hobby machining.

[@weaves7]

What do you use for way oil on your g0704 and g0602? I have both of these machines as well and use iso 32 and iso 68 in accordance with the manuals but it is just tractor hydraulic oil and I feel like it doesn't stick to the ways very well.

[@michaelalcocer6778]

Mobile vactra no2 is the way to go

[@tjh44961]

Sorry, I'm running a bit behind on video watching so this is a little late... Z-axis errors are inevitable. The lighter the machine, the more difficult they are to compensate for. I have a Taiwanese-made, round column mill/drill, and a Sherline mill. Of course the heavier machine can take heavier cuts before it starts to flex the Z-axis, but you can never dial in a depth of cut on either machine and expect to get that depth correctly in one pass, even with a brand-new, razor sharp end mill. They just aren't rigid enough, massive enough, heavy enough, whatever you want to call it. You still have to sneak up on it. And at about 300# your G0704 has about 40% less mass than my mill/drill, with a corresponding reduction in rigidity. I know it's not linear, but it's likely that the mass of the 200# head hanging on the 5" diameter column deflects less than your G0704 when applying downforce for milling or engraving.

[@Steve_R]

Thank you James. I'm not into CNC, but all the info is still good for any kind of backlash trouble shooting. I've been meaning to ask. With a 42 on the end of your name, are you into the "Hitchhikers guide to the galaxy" books? I have all 5 books in the trilogy.

[@patrickbeaumier8616]

Nice job 👍

[@deemstyle]

On the top if your spindle, you have a large-ish wire coming in labeled VFD. Do you know what the name of the connector is called? I've been looking for some reasonable sized multi-pin connectors like this. Thanks!

[@Clough42]

No. You could contact contact CNC Depot. They manufacturer it.

[@ikbendusan]

it's an amphenol brand connector. cnc depot has close-up pictures of the connector on their website

[@richardcurtis556]

Another example of why precision counts. Ignore it at your peril because you may need it someday.

[@pudster2237]

James, I was just wondering. I noticed that sometimes you had quite a bit of pre load on the dial indicator and sometimes you didn't and I was wondering about the quality of the dial indicator also. Was it a good one and could these have caused you to get some differences in your readings? Just wondering. Love the content of all your videos. Thanks for your hard work.

[@Clough42]

For a large, heavy system like this, the dial preload will have little effect. I have used several indicators, including a $500 Haimer on this setup and seen similar results. Questioning the tool is a good thing though--especially when results are inconsistent.

[@tobiasripper4124]

11:06 *waves hand* "these are not the 5 thou you are looking for"

[@jjbode1]

I think he said "two tenths is not the five tenths we're looking for".

[@tobiasripper4124]

@@jjbode1 i think you missed a very obvious and well known movie reference.

[@TomChame]

Interesting, thanks.

[@Stefan_Boerjesson]

Interesting video. I've got a mill issue. If I adjust the Z gib "softly" the column will be "hanging" on the lead screw but then the resistance from the milling bit cutting pushes the column up and I don't get the depth of cut I want. Adjusting the Z gib "hard", making the lead screw "force the column down" loks like working but is likely causing wear on the gib. Any advice from You?

[@ikbendusan]

the gib is a wear item; adjust it so there is no play and also no binding. you can't eliminate stiction altogether

[@Stefan_Boerjesson]

@@ikbendusan Thanks for Your reply! Your fully right. I was thinking about the wear on the column, caused by the gib. Is it Your opinion that I should tightened the gib to make the column "hang" on the gib tension and some force from the vertical handle, including fine tuning, will "force" the column down? Now I place my hand on the column and force it to the "down position" in order to avoid "an unexpected free fall". My old dictionary don't know the word "stiction".....

[@stevenrichardson7882]

Got the can just about right

[@Cromwell648]

Excellent detective work, Holmes👍

[@platin2148]

Was this mill always cnc or is this a conversion almost looks like a opti mh20

[@Clough42]

It's a Grizzly G0704 conversion.

[@donmittlestaedt1117]

Well done.

[@scrout]

You should wear THE BELT

[@dannywilsher4165]

Champion!!!!!

[@greener121]

Very nice, step by step, I think it was @watchweswork that he discusses all of the ways to evaluate machine accuracy, but I can't remember the book referenced or the video off the top of my head. I'm loosing steps in z, but i think it is the stepper motor just cant move the mass fast enough. Also, Mach 3 backlash comp is total crap if you don't know where the back lash is coming from.

[@GameBacardi]

...screw worn out from height where head mostly working. :P

[@HansFormerlyTraffer]

The easiest way to deal with backlash is preload. If you can implement it you have magic.

[@Clough42]

If the backlash is in the nut.

[@Tupeutla]

i am at 7 minutes but commented already :) : why not software fix the 5 thou if this backlash is constant ?

[@ikbendusan]

mach3 supposedly does not support backlash compensation during milling operations. this excerpt is from the manual: 5.6.3 Configure Backlash: ''The Config>Backlash dialog shown in Figure 5-24 lets you provide an estimate of the distance the axis must back up to ensure any backlash is taken up when the final “forward” movement is made. You can also specify the speed at which this movement is to be made. Mach3 will attempt to compensate for backlash in axis drive mechanisms by attempting to approach each required coordinate from the same direction. While this is useful in applications like drilling or boring, *it cannot overcome problems with the machine in continuous cutting where changes in direction occur.* you would need another controller (i have an edingcnc which does do proper backlash compensation)

[@silden11]

It looks like you are fighting stick/slip. and you can’t really fix this without moving to linear rails

[@Clough42]

Yup. Very likely.

[@michaelalcocer6778]

@@Clough42 what are you using for way lube? Try loosening the gib a lot and see how much backlash that removes. When setting the gib tightness on my G0704 I have a DTI on and check back lash as I tighten. I tighten until I start to introduce backlash into the system. My screws are double nuts that I custom shimmed, they account for about .0005 of my back lash.

[@Clough42]

@@michaelalcocer6778 Quick Slide 68. Essentially the same as Vactra 2. I did basically what you suggest, and found a happy .001" setting. This seems to be pretty temperature dependent, though. It was worse again the next morning when the shop was 20 degrees cooler.

[@michaelalcocer6778]

@@Clough42 yeah I should have waited till the end of the video to comment. The gib will grow and shrink with the outside environment. Set it for when you would most likely be using it. I have spent hours chasing my G0704. I am over it unless you scrape it, it will have one issue after another. Linear rails if I keep it. Just trying to finish up my TC22 retrofit.

[@henmich]

Ball..... screw..... will change your machine forever...

[@court2379]

Firmer disapline is required for fixing backlash. Set that machine straight.

[@gn6034]

Yeah I'm the first to comment

[@ikbendusan]

nice

[@hotfuzz1913]

Congratulations

[@MakeTechPtyLtd]

Well done