My circles are always too small. Why?

Home built machine with 1605 screws. Nema 23 and Nema 24 closed loop steppers. Mach 3 with Fusion 360. 4.5kw 3P spindle. Machine is rigid and as tight as I can get. Using a 10mm 2 flute endmill that spits quite large chips out.

This is the 5th or 6th part I have made with alloy. This sample is out of a 76mm x 76mm x 130mm block of 6061. Im not looking for perfect, getting to where I want to be but always room to improve. I dont use it for business.

My problem is my circles are always, always, 0.2 - 0.4mm too small. Too tight for a fit. No matter how I machine.

I am thinking its backlash? Can only be mechanical? Although I have seen some settings in Mach 3 about inc and absolute with radiuses etc. Is there something else I should address too?

Pics attached of my last effort. Not perfect, will do the job, I just need to rehone out the round parts to 0.4mm more. Or - perhaps I run a climb cut around a circular boss, and then a conventional cut around the same boss. Maybe that will sort this small error.

Thoughts?

Tool undersize?

Can check by cutting a full tool width slot and measuring the slot.

Does the error depend on speed, that is does it get smaller with reduced feed rate?
Keep in mind that with a normal closed loop control (without feed forward compensation) the radius of a circle made by the machine is always smaller than the programmed radius.

That is the first time I've heard about forward compensation. Thanks for something else to look into. For small holes I have gone to using drill bits I found on ebay from kyocera. This way I know I'm getting the right size. My spindle is only able to hit 8k rpm at the lowest. When I peck and use sure with some cutting fluid, I get great results and the bits are cheap enough that if they wear prematurely, it's not a big deal.
The only issue I really have is that they do not have a long enough cutter for all applications. When this happens, I go as far as I can with the bit and finish with either going back to bore the rest or I bring it to the drill press after.

One day I may go with servos so I can take advantage of forward compensation as well as upgrading my cutting speed and depth. If the holes can come out correct all the time with the upgrade, that alone has me wanting to do it. One more thing to add to the reasons why I should, as long as it actually does work.

Sent from my SM-N960U using Tapatalk

It might be worth comparing two circles,one conventionally cut and the other climb cut.If they differ you may have some play in the mechanical components.Finding it and fixing it might be a challenge.

one other thing that it could be is deflection. depends on the depth of cut and tool stick out. but if the holes are bigger on the top and smaller on the bottom it could be deflection to some degree. the closer the cut is to the tool holder the better the size should be and at the bottom of the cut were the tool stick out is the most will be small. especially with a long small diameter tool. just some food for thought.

something else to check would be the tramming of the spindle too while i'm thinking about it :)

I have not heard of forward compensation. So on the sample piece attached to this string. The large hole was programmed at exactly 57mm. I aggressively pocketed it out with a 10mm endmill, and then finished the last with a contour, taking 2 very small 0.4mm stepover cuts at 400mm/min with a final finish at really slow 120mm/Min to get that polished look on the inside. I measured it properly this evening, its exactly 56.66mm to 56.58mm.

Is there a chance I am rubbing it instead of cutting it properly? You know, like on a lathe where you have to actually take a proper cut to get a dimension. Instead of trying to take 0.005mm off which doesn't work.

But I measured my tool, its on the money 10mm. Well, I am going to probe both of them out and cut too 57.5mm so I can fit my part together. And this part, is to accept a FK block/Nema 24 so I can install some new Y gantry 2510 ballscrews instead of the 1605s I have now. Infact, I reckon the 2100mm long 1605s at the length they are, when they wobble, will account for quite a lot of play. And thinking more, they in themselves will not be helping at all.

sometimes with deflection issues taking a spring pass will bring it in closer. aluminum is pretty forgiving so as long as your cuts are on the cutting edge of your end mills i don't think going slow will cause you any rubbing as long as the tool is sharp anyways.

So tonight I started thinking about my ballscrews. Being new about 8 years ago I bought the standard Chinese 1605 screws which I am still using to this day. Thinking more deeply into the issues I am having, my machine is only going to be as accurate as the screws themselves. I dont even know if the screws I bought were of any standard let alone C7 which is the common. So tonight I setup a digital vernier and took Mach 3 back to the original screenset where there is a setting to set your steps per measurement.

Inputting 120mm travel into my X axis, I got something (cant remember exactly what) about half a mm short. This I confirmed by moving the X in the positive direction again, and also against the vernier with a reset zero. to avoid any backlash. Got the same result. The settings I now have for my X axis is 400.23...something something. Hows that! I am going to have a bit more of a play tomorrow but feel this is a bit of a breakthrough. Like, my machine is tight, real tight. I have some new screws arriving in a month. But it would be nice to sort this.

Also, I am thinking about increasing the resolution to say, 800steps per mm - dunno - what does everyone else do with resolution. Gee, almost 10 years on and still playing with DIY CNC.

Hey - machined out a wicked knuckle hinge tonight for a device. Loving this alloy machining its COOOL....... just my work is not as accurate as it should be haha

And thanks you reminded me to tram just to be sure.

Sent from my SM-N970F using Tapatalk

Quote:

---

Originally Posted by boydage

And thanks you reminded me to tram just to be sure.

Sent from my SM-N970F using Tapatalk

---

Yes Traming is one of the most important thing to do if you want any accuracy

You can't just increase the resolution by adding more steps per mm, in your case the resolution is controlled by your stepper motors drive configuration and the Ballscrew you are using, so when you change to the 10mm pitch screws if everything stays the same as you have it now, you steps will be only 200 so the resolution will be less than what you have now

Thanks for pointing out the obvious. I was referring to the electrical resolution. Currently I think the steppers are 1600 steps rev. Or might be 2000. I was thinking about changing that. I do remember setting them really high once on the bench. The encoder is a 1000 line encoder. This is the first time I've considered this in years. I am only talking 0.3mm in a 57mm round pocket out of 45mm 6061.

But back to what I was talking about before, does anyone have input about the ballscrew and my finding its not a 5mm pitch its actually 4.9999 something something ( should work it out)

Sent from my SM-N970F using Tapatalk

...are you using G12/G13 in your program? Not sure how F360 post are..but if, you use a G12/G13 code than all you have to do is change the Diameter Offset (stored D value) to fine tune the bore(s) diameters.

https://www.cnczone.com/forums/g-cod...lp-needed.html

Thanks for that. Hey but hows this, I setup a 300mm vernier in my vice and spent an hour measuring and setting up my steps per mm using the Mach 3 original screenset. The screws I sourced from China I think about 8 years ago.

I got different results on the Y and X. The Z, perfect on account the ballscrew is only a few weeks old. I ended up with 400.45......something on the Y and 400.23.....something on the X.

Then I pocketed a piece of 20mm 6061 out to a 25mm round pocket. I measured it with a telescopic ball gage and got 24.99mm. Hows that!! And it's an almost perfect circle, with just 0.04mm out of round.

Happy with that result. Funny about the ****ty screws tho.

Sent from my SM-N970F using Tapatalk

...is it a Backlash problem or just bad quality ball screws?
https://www.cnczone.com/forums/bench...pensation.html

A couple years ago (remembered today) I was chasing what I thought was my steppers missing steps. Over (approx) 1400mm I would come up 1.5mm short. I've only really started with alloy this past year where I am manufacturing items that have to be accurate to +/-0.2mm.

I have a tiny amount of backlash, very small. This is definitely my screws. When I measured the travel, I moved in one direction first, then would measure from there forward. That would not check for backlash but was good and accurate for screw accuracy.

Should I be surprised my Chinese screws from 8 years ago are only as accurate as I have found? Started to work out just exactly how inaccurate they are, but need a pen. 400 steps rev 5mm to 400.45 steps rev 5mm. Ummmm...... can't work that one out in my head.

Sent from my SM-N970F using Tapatalk

Quote:

---

Originally Posted by boydage

Thanks for pointing out the obvious. I was referring to the electrical resolution. Currently I think the steppers are 1600 steps rev. Or might be 2000. I was thinking about changing that. I do remember setting them really high once on the bench. The encoder is a 1000 line encoder. This is the first time I've considered this in years. I am only talking 0.3mm in a 57mm round pocket out of 45mm 6061.

But back to what I was talking about before, does anyone have input about the ballscrew and my finding its not a 5mm pitch its actually 4.9999 something something ( should work it out)

Sent from my SM-N970F using Tapatalk

---

That is what I posted about for the resolution, ( Drive configuration :)

The Ballscrew could have any pitch being a low grade Ballscrew, any pitch number is possible, that number does not matter as long as it is consistent the length of the travel, you just use the setup for setting the steps per mm, this you want to test at no less than 150mm of travel, this will get you with accurate movements, you will need a precision gauge block to set this with if you want any accuracy

My work, is only going to be as, or more accurate as the device I measure my work with in the realms of my workshop and that's acceptable for what I am producing. So what I did was clamp one end of my 300mm vernier into my vice between two pieces of pine.

Then I moved in a positive direction against the vernier 10mm. Then I moved it 280mm and recorded the steps it took. This test I performed in 3 locations on my table to see if the results were the same throughout the length of the ballscrew(s). They were and I am happy with that.

It would have been a bummer to get different results throughout the table.

I am looking forward to the 2510 screws I have on order. The 16mm ones do wobble a tad. They have done their job. And I am sure I will use them one day for another project.


Sent from my SM-N970F using Tapatalk

Quote:

---

Originally Posted by CNCFr

Does the error depend on speed, that is does it get smaller with reduced feed rate?
Keep in mind that with a normal closed loop control (without feed forward compensation) the radius of a circle made by the machine is always smaller than the programmed radius.

---

I know this is an old thread but you have just hit on an issue I am struggling with. All my bores are coming out undersized at anything about 1000mm/min, at 500mm/min they are accurate. Is there any solution to this?