[rcheli]

Backlash would not cause the tip to be .080 low after running your program. Unless you had .080 in backlash and you would feel that if you grabbed the spindle and lifted up on it. After all of your checking and assuming that no settings were accidentally changed then I would have to believe that either the stepper is getting weak and missing steps or there is an issue with the BOB and all the steps that the software thinks have been sent have not.

John

[Mr.Chips]

Thanks John,
I agree with you about the backlash.
Do you have any suggestions for testing the stepper and BOB?

The entire file that a ran was generated from a .JPG file, I have done that hundreds of times before without any issues.

Is there a Gcode oR Mcode that could change the Z position, if there is one I could scan the file and see if there were any codes like that in the file. Kinda shooting in the dark but don’t know where to go from here.

Thanks

[GME]

Yes Gary all measurements were made with a dial indicator. I just used the 2x4 method to see if I could feel any play.

I checked the X, Y and Z axis while I was at it all were less than 0.001” using a starret dial indicator with a resolution of 0.001” per division.

Yes, Steps per inch has been set.

Backlash was determined by zeroing ind at start position. This method was used on all axis
Use Mach to move + 0.050”
Then Mach back to zero.
Use Mach to move -0.050”
Then Mach back to zero, if it does not return to zero, that is the amount of backlash.

I don’t think you can get an accurate backlash reading on the Z axis like this, because all the weight of the router and z assembly are pulled down by gravity, and are always resting on top of the thread. Both the X and Y axis are not effected by gravity in this way, they move to the right and touch the right side of the thread then to the left and the distance they move till they hit the left side is backlash. Or am I all wrong on my Z backlash theory?
I’m guessing the only way to measure backlash on the Z axis is to mount the dial indicator on the Z assembly that doesn’t move and indicator tip on a part that does move, and then use an outside force (2x4) and gently push up on the router mount, any movement observed would be backlash. I don’t think you could use Mach to measure backlash. But again I could be wrong.

Rethinking this:
The only time that the underside of the thread is touched is when the router is cutting in a downward direction. Then backlash will be when the Z is at rest, and the distance the Z axis moves up when cutting in a down direction. I think this is correct.



Thanks for going through this with me.

HH

Backlash would not cause the tip to be .080 low after running your program. Unless you had .080 in backlash and you would feel that if you grabbed the spindle and lifted up on it. After all of your checking and assuming that no settings were accidentally changed then I would have to believe that either the stepper is getting weak and missing steps or there is an issue with the BOB and all the steps that the software thinks have been sent have not.

John

John, I agree that .080 low would not be backlash. Apparently tightening the coupling got it down from .080 to .010. .010 would still be a lot of backlash though. I might be able to feel .080", and I emphasize "might", but I'm absolutely positive that I could not feel .010" of play in a weighted Z axis.

HH, I don't favor the 2x4 method, because there is too much of a chance you would induce structural flex into the measurement. You're dealing with aluminum after all. You would never know for sure whether you were reading backlash or flex on your indicator

I don't necessarily agree with your concern about being able to measure Z axis backlash. When I was using an acme screw, I had .001" of backlash. I had no problem measuring it repeatedly in both the + and - directions, and when I zeroed to my spoilboard, I got dead on zero every time. If you were correct about the weight of the router causing the axis to always rest on the top thread, then your stepper would not be doing its job of holding the axis in position. FWIW, I have a 2.2 KW water cooled spindle which weighs more than your router. I was able to measure that .001" of backlash consistently. Maybe my Nema 34 holds better than your Nema 23, but for just holding, it shouldn't - even if it is wearing out.

Anytime I am making any sort of through or profile cut, I zero to my spoilboard. I never cut into it - even as much as .001. There is a technique for doing this. Look up CNCnutz on YouTube and watch his video, Don't Spoil Your Spoilboard. As long as everything to do with Z is configured properly, it works like a charm 100% of the time.

One way to check for missed steps is to hook your dial indicator back up, zero at the Z axis's highest position, and then use Mach to command longer runs down and back up. I'd go to as near to the spoilboard as possible to get the most travel length up and down. I would repeat the up and down motion several times, checking the up measurement on the indicator each time. I am making the assumption that lost steps would not be entirely uniform, if it's something to do with a stepper wearing out. My assumption could be wrong, but in my experience, lost steps can be fairly random, assuming there isn't some sort of obstruction/bind causing them. Where you end up after maybe ten cycles may reveal a problem, assuming there is one, or rule out the stepper and/or BOB altogether. If you end up pretty much where you started, I'd be inclined to call both the stepper and BOB good.

Something else you might try: Attach your dial indicator to the router and zero the indicator to the spoilboard. Then do a commanded maximum distance up and down several times. Do you still end up .010 too deep? if it's the stepper or BOB, the amount of increased depth should keep getting larger and larger. If it doesn't, I'd take another shot at calibrating Z's steps per and rerun the tests. BTW, how do you go about measuring for steps per? I use 1 2 3 blocks (I have about 16, I think) and find they work very well measuring all axes. (Sorry to be a pest on the step per issue, but that .010" you are now left with sounds a lot like a steps per issue. On this issue, I ask that you be patient with me)

Sorry, but I have no suggestions about how to test the BOB, other than buying a new one and seeing if the problem goes away. Kind of an expensive and time consuming option.

Gary

[rcheli]

If it were a steps per inch issue the error would only show if you moved in one direction. When commanded to move back to the starting point it would move back the same incorrect number of steps and end up exactly where is started.
To check the steps per inch you just need to command the longer move you can and accurately measure the amount it actually moved compared to what the software thinks it moved. This test would best be done moving the Z in the downward direction to minimize possible lost steps and start the test after having moved in the downward direction to take up and backlash that may be present. The larger the steps per inch error is the shorter the distance you need to move to see it. My mill was off by a little bit and It would always return to the same spot but if I commanded the bit to plunge a certain distance it would always cut a few thousandths more than I command. This is not what yours is doing. My error did not accumulate over time.

[GME]

If it were a steps per inch issue the error would only show if you moved in one direction. When commanded to move back to the starting point it would move back the same incorrect number of steps and end up exactly where is started.
To check the steps per inch you just need to command the longer move you can and accurately measure the amount it actually moved compared to what the software thinks it moved. This test would best be done moving the Z in the downward direction to minimize possible lost steps and start the test after having moved in the downward direction to take up and backlash that may be present. The larger the steps per inch error is the shorter the distance you need to move to see it. My mill was off by a little bit and It would always return to the same spot but if I commanded the bit to plunge a certain distance it would always cut a few thousandths more than I command. This is not what yours is doing. My error did not accumulate over time.

Maybe I misunderstood something. I thought HH indicated that he believed tightening the set screw solved most of the problem, (from .080 to .010). I didn't get the impression that the depth of cut was necessarily continuing to increase during cutting. I gather from your comments that you read HH's posts a different way. HH, what's the scoop? Is the .010 a static number, or is depth of cut increasing over time?

At this point, we are looking at Z going .010" further than commanded. If .010" is constant, it would strongly suggest there are no lost steps or a BOB issue. Steps per would be the next thing to check. As you pointed out, when the steps per are off, Z will plunge more or less than the amount commanded, depending upon whether the steps are set too low or high. This applies to all axes. The longer the distance commanded to travel, the more the overall error you'll see. That's why it's best to calibrate steps per over the longest distance possible. Errors from short distance calibrations can greatly magnify over long distances. It you calibrate at the longest distances possible, you won't have errors adding up and you will be as accurate as you possible. .005" off at 48" (more than I tolerate) is only off .00125" at 12"

I agree with you that steps per are done in one direction only, and you move slightly in that direction to take up any backlash before making the measurement run. I use 1 2 3 blocks, because they are way more accurate than using a tape measure, or even a precision rule. On Z, I run the axis to the top and then down a few thousands. Then, I stack up 1 2 3 blocks and zero off the top block. Then, I remove all but the last block, and command a move the distance represented by the blocks removed. The dial indicator will tell tell me the actual distance moved (the distance of the blocks removed + or - the indicator reading). On X and Y, I just lay them out with the first and last blocks standing up on edge (3" high). On X & Y, I just have to remove the first block, which makes the confirming reruns a snap.

Gary

[Mr.Chips]

Maybe I misunderstood something. I thought HH indicated that he believed tightening the set screw solved most of the problem, (from .080 to .010). I didn't get the impression that the depth of cut was necessarily continuing to increase during cutting. I gather from your comments that you read HH's posts a different way. HH, what's the scoop? Is the .010 a static number, or is depth of cut increasing over time?

At this point, we are looking at Z going .010" further than commanded. If .010" is constant, it would strongly suggest there are no lost steps or a BOB issue. Steps per would be the next thing to check. As you pointed out, when the steps per are off, Z will plunge more or less than the amount commanded, depending upon whether the steps are set too low or high. This applies to all axes. The longer the distance commanded to travel, the more the overall error you'll see. That's why it's best to calibrate steps per over the longest distance possible. Errors from short distance calibrations can greatly magnify over long distances. It you calibrate at the longest distances possible, you won't have errors adding up and you will be as accurate as you possible. .005" off at 48" (more than I tolerate) is only off .00125" at 12"

I agree with you that steps per are done in one direction only, and you move slightly in that direction to take up any backlash before making the measurement run. I use 1 2 3 blocks, because they are way more accurate than using a tape measure, or even a precision rule. On Z, I run the axis to the top and then down a few thousands. Then, I stack up 1 2 3 blocks and zero off the top block. Then, I remove all but the last block, and command a move the distance represented by the blocks removed. The dial indicator will tell tell me the actual distance moved (the distance of the blocks removed + or - the indicator reading). On X and Y, I just lay them out with the first and last blocks standing up on edge (3" high). On X & Y, I just have to remove the first block, which makes the confirming reruns a snap.

Gary

Here’s the scoop!
I scratch built this machine ten years ago and haven’t had any problem with the VcarvePro, 60deg V bit tool depth getting deeper over time on long cuts, and this file ran over two and three quarter hours, I frequently ran longer ones in the past without this proble,.
I posted the problem on the forum, and got suggestions that set screws could be loose. I hand tested for tightnes a little harder than normal feeling a “tick” as the screws turned a little more, but nothing had to be turned more than this.

My guess, I used a flange bearing to support the X axis acme thread, tightening the two set screws on the threads of the acme screw. I think that tightening on the threads was not a good idea, and that this allowed the acme thread to drop down 0.080” and because the set screws just found a new home it was not detected when I tested for tightness. I plan to remove the Z acme rod and cut the threads off in the area and add two flats for the new flange bearing.

I’m going to rerun the file today and I will cut a 1/4” pocket before and after the run to see if there is any change in the depth.

I like your method of using 123 blocks however that method would require sensing the blocks electronically and I’m still using the paper method. I have wanted to add electronic Z zeroing for a long time, CNCNUTZ and many others have posted very detailed directions, the problem is I used CandCNC BOB and electronics and that required using their home screen, and I don’t know if their screen could be modified to add auto Z zeroing. I have frequently tried calling them but they never answer or return calls. And am afraid it try and edit it using the Mach screen editor. As you can tell I’m more of a mechanical person and not so much electronics and programming, but I’m 79 so that’s my excuse, LOL

Thanks Gary for you detailed comments I appreciate it.
HH

[GME]

Here’s the scoop!
I scratch built this machine ten years ago and haven’t had any problem with the VcarvePro, 60deg V bit tool depth getting deeper over time on long cuts, and this file ran over two and three quarter hours, I frequently ran longer ones in the past without this proble,.
I posted the problem on the forum, and got suggestions that set screws could be loose. I hand tested for tightnes a little harder than normal feeling a “tick” as the screws turned a little more, but nothing had to be turned more than this.

My guess, I used a flange bearing to support the X axis acme thread, tightening the two set screws on the threads of the acme screw. I think that tightening on the threads was not a good idea, and that this allowed the acme thread to drop down 0.080” and because the set screws just found a new home it was not detected when I tested for tightness. I plan to remove the Z acme rod and cut the threads off in the area and add two flats for the new flange bearing.

I’m going to rerun the file today and I will cut a 1/4” pocket before and after the run to see if there is any change in the depth.

I like your method of using 123 blocks however that method would require sensing the blocks electronically and I’m still using the paper method. I have wanted to add electronic Z zeroing for a long time, CNCNUTZ and many others have posted very detailed directions, the problem is I used CandCNC BOB and electronics and that required using their home screen, and I don’t know if their screen could be modified to add auto Z zeroing. I have frequently tried calling them but they never answer or return calls. And am afraid it try and edit it using the Mach screen editor. As you can tell I’m more of a mechanical person and not so much electronics and programming, but I’m 79 so that’s my excuse, LOL

Thanks Gary for you detailed comments I appreciate it.
HH

HH, I believe you misunderstood my use of 1 2 3 blocks. I use them with a dial indicator. Nothing electronic involved. It is a purely mechanical measurement, except commanding your Z to move up and down with your software. Simple G-code commands, e.g., G91 Z-6 (Z down 6 inches), read indicator. Then G91 Z6, (up 6 inches) read indicator. My suggestion was to use to the blocks as an accurate way to test for consistency, i.e., looking for lost steps or perhaps something else. As far as setting steps per. it is also done with 1 2 3 blocks and a dial indicator. Again, nothing electronic involved, other than commanding moves. I use an aluminum plate with a milled hole to not only set zero, but to edge find. With it, I can set height and at one corner perfectly. I use the Maker's Guide Triple Edge Finder, The Makers Guide - Triple Edge Finder - THE MAKERS GUIDE, but cheaper ones are available on Ebay. Anyway, my use of an plate to zero has nothing to do with what I suggested to test and measure. Note: I wouldn't use a piece of steel to zero to. Too great a chance of breaking a cutting edge on the point of a bit.

Okay, I'm a little confused. Is your Z still going progressively deeper as you cut, or is it cutting too deep at a consistent distance - .010, or .080 or whatever? I thought you indicated that tightening the screws reduced the error from .080 to .010. From your current comments, it sounds like it returned to .080.

Cutting a pocket may or may not tell you anything, and it will probably be far less accurate than using the techniques I suggested. Obviously your call, but using a dial indicator and blocks, IMO, would give you better results. Even zeroing high and using the spoilboard to run down to (in place of running down to a 1 2 3 block) would be more accurate than cutting a pocket.

Nothing wrong with changing out the screw, but it would be a lot more work that may ultimately be unnecessary. Your call.

I get the impression I may not have been clear enough in describing the testing process. When you concluded it involved something electronic, it told me that I did a bad job of delivering the message. My apologies. I would photograph the steps on my router. but I have torn down the one I had been using and am in the process of building a new, much stouter one (using 3030 extrusions 3" x 3" from 80/20 throughout. except for the gantry). Even the leg are 3030. Right now, I'm milling the ends perfectly square and getting the lengths consistent. 80/20 has a .002" per inch tolerance for square, and .015" tolerance for length. They are pretty consistently at or near the max tolerance error, which isn't good enough for me on a 4 X 4 router. Anyway, I'll give some thought about how I might construct some photo documentation that would be more useful to you..

BTW, you have a few years on me. I'm pushing 70. I love being retired. I come from a pressure cooker professional life. Since I retired, I've never been so relaxed.

Gary

[GME]

HH

Here are a few photos to illustrate the testing I described. I used some of my 80/20 to make a mock up of a CNC.



Step 1. This roughly where I mounted the indicator. I zeroed it off the top of the gantry.






Step 2. Here, the mock up shows the gantry lowered by the G91 distance.






Step 3. This photo shows the gantry raised by the same G91 distance as it was lowered in Step 2.




Step 4. Note the reading in Step 3. In this example, I mocked up a reading of -.1035, which would indicate that the screw was slipping, or the stepper was losing steps.

Repeat a total of 10 times.

Here is the other suggestion was to mount the indicator to Z and zero it to the spoilboard. After zeroing it, it's the same steps as above, only your first move is up, and the second is back down.

Does this make what I described a little clearer? I hope so.

Gary