I agree Donalder and just enough flaming to lighten it up
I don't remember seeing a picture in your post, if you are sure it went through then they must have deleted it pretty quickly. Sometimes these forums get buggy and everything does not go through correctly.Originally Posted by 109jb
I would hate to see you stop offering help as your posts have taught me a lot and I am sure others are grateful for them as well.
Post up the picture again in a new post or edit the old one to include it, it might help better explain what you are talking about.
I more than agree with you about the completely arbitrary moderation here. However, I believe your edited graph never made it to the forum. I saw your post very shortly after it first appeared, and there was no graph there. Can't imagine why an admin would have deleted it, without at least deleting the entire post.
Regards,
Ray L.
Well I guess we will have to chalk it up to an upload error. I know I did the upload attachment thingy because I had to wait for it, but to be honest I didn't check the post for completeness after submitting itbecause I had other things to do. If you guys say it wasn't there from the get-go then it had to have been an upload problem. Stranger things have happened on forums before. I think I wil set up a make-believe mapping that I can use to illustrate a few things about both "true screw mapping" and the linear typ compensation.
OK. I've had time to think more about the different ways to compensate for lead screw errors. I took Ryan's graph and backed out the mapping of his screw and created a spreadsheet to play with. As I see it, if you are going to compensate for the screw at all you basically have 3 choices:
1. What I call "true screw mapping"
2. A steps/distance calibration using a best line fit to a mapped screw
3. A steps/distance calibration using 2 arbitrary points along the travel.
The differences in these 3 schemes is that "true screw mapping" attempts to take care of positioning errors by applying linear adjustments between individual calibration points in relatively short increments and applying that to the machine coordinates. Both of the steps/distance schemes apply a linear adjustment but do it for the entire table travel and the results vary depending on where you zero your part as you will see.
I have attached a graph to illustrate the differences in outcomes. As I said, I used one of the curves from Ryan's graph of his screw and backed out his mapping of the screw. I then used this data and applied what the different schemes would do to the mapping of an actual screw. In the graph you can see that the basic error of the screw was not linear and the magnitude of the error kept increasing. This indicates that the average screw pitch is slightly off with other short distance errors thrown in.
I applied what "true" mapping to the screw in 25mm increments. I didn't pick and choose the best points to use, I just started at 0 and went up in 25mm increments because that would be close to a 1" interval for us in the USA. What would happen in the machine control software is that you tell the machine to go to a position of say 50mm from home and it looks at a table and turns the screw the appropriate amount to get exactly 50mm from home because that position is discreetly defined in the compensation table. The table tells it the adjustment values for every 25mm. How about for a position in-between those 25mm positions. Lets say you told it to go to a position 35mm from the home position. there isn't a discreet value for 35mm, but there is for 25mm and for 50mm and 35mm is between those values. The software then interpolates a value for the 35mm position based on the adjustment values it has for the 25 and 50mm positions. I did just that in my spreadsheet and graphed what the errors you could expect from this sceme to be. You see that the screw itself had a maximum error of 0.1mm (0.0040") and after appling the true screw mapping it is down to 0.01mm (0.0004"). Now true screw mapping is dependent on having an accurate home position, but it isn't dependent on the part zero position as the control software is smart enough to do all of the conversions between part zero, machine zero, etc. At least it should be. So, one of the benefits of true mapping is that the software is making adjustments based on machine coordinates, not based on part coordinates.
The second scheme involves doing measurements at short intervals along the screw travel just like true screw mapping, but then finding the best straight line fit for that data. A spreadsheet makes it easy because it will calculate the best line fit for you. I did that for the data and applied it and plotted it and it shows a large reduction in error. However, this scheme is affected by whatever zero position you have selected for your part. I placed 2 curves to show this one shows the errors based on part zero being at the home position and the other shows it based on the part zero at 45mm from the home position. It basically slides the curve vertically. Still pretty good results though.
The final scheme is just using 2 arbitrary points and adjusting the steps/distance that the control software uses based on those 2 discreet points. I used a 75mm spacing for the 2 points because many people use the 3" length of a 1-2-3 block as their reference and 75mm is close to 3". This sceme puts you ate the mercy or your own luck. As an example, I deliberately picked spots to show what can happen just due to where along the screw travel you happen to place your 1-2-3 block. The first one uses a 75mm spacing from 10 to 85mm. As you can see this doesn't come out so good and basically just flips the curve over and now you have errors in the other direction. The second curve used values at 70 and 145mm. This one looks good once you get past about 50mm from the left end of the screw, and shows the error oscillating around 0, but I hand picked this from the mapped data to get the best result for case #2 of this scheme. Unless you actually map the screw and do this you are at the mercy of where the errors on the screw happen to be. I can actually envision this scheme being able to make positioning worse rather than better if the screw happens to have a spot that reverses the direction of the error slightly and you have the bad luck to set your 1-2-3 block right on that spot. You also have to remember that the curve can shift because of the same part zero problems that scheme 2 has.
The true mapping yields the best results but requires the most work. It is easiest with a linear scale and the shorter the interval the better.
The second scheme is better and can give give decent results, but is dependent on where you zero your part, and on just how bad the screw is. However it requires taking measurements along the whole length of the screw. At this point, you may as well just create the compensation table and do true screw mapping.
In my opinion, the third scheme is a gamble. It can pay off and lots of people use it, but I would not.
In my opinion the bottom line is that you either buy a screw with a guaranteed accuracy to fit your needs without any kind of mapping or calibration, or you use true screw mapping. The other schemes are just too iffy to me and I personally would not use them.
Wow...
I just learned so much.
This is what makes this forum amazing.
I am definitely going to apply a linear scale to my setup and I am looking forward to some good results!
Thank you!
After waiting days to get a quote, I called one of the distributors again and he was waiting for prices from Nook.
I then found a source, reidsupply.com, for online purchase for Nook and Thomson ball screws & nuts with pricing.
Then finally, late this afternoon, I got a quote on the Nook ball screws & nuts.
Let's start with the Nook quote from Motion Industries
Nook 5/8" ball screws & nuts
SRT9987 5/8" x .200 x 48" x .004"/ft accuracy ball screw (18.38/ft) $ 73.52
SRT7540 5/8" x .200 x 72" x .004"/ft accuracy ball screw (18.40/ft) $110.41
XPR6320R72 5/8" x .200 x 72" x .001"/ft accuracy ball screw (26.58/ft) $159.49
SBN10325 SRT series 5/8" single ball nut $ 56.19
Nook 3/4" ball screws & nuts
SRT7296 3/4" x .200 x 96" x .004"/ft accuracy ball screw ($34.04/ft) $272.35
SRT7248 3/4" x .200 x 48" x .004"/ft accuracy ball screw ($34.05/ft) $136.18
XPR7520R96 3/4" x .200 x 96" x .001"/ft accuracy ball screw ($50.53/ft) $404.25
XPR7520R48 3/4" x .200 x 48" x .001"/ft accuracy ball screw ($50.53/ft) $202.13
SBN7201 SRT series 3/4" x .200 single ball nut $125.17
SBN7202 SRT series 3/4" x .200 double ball nut $248.36
SEL10057 SRT series 3/4" x .200 adj. preloaded dbl ball nut $625.02
PRN10109 XPR series 3/4" x .200 preloaded dbl ball nut $775.84
SSN0390 XPR series 3/4" x .200 preloaded flanged dbl ball nut with wipers $909.14
The prices for the Nook 3/4" SRT series from reid supply were just a bit higher, but they don't carry the XPR line. For 5/8", they offer Thomson products with a .004"/ft lead accuracy at a lower cost.
Reid Supply
AR-315 Nook SRT7296 3/4" x .200 x 96" x .004"/ft ball screw $277.07
AR-305 Nook SRT7248 3/4" x .200 x 48" x .004"/ft ball screw $138.54
AR-315 Nook SBN7201 3/4" x .200 single ball nut $127.34
TBS-20 Thomson 5707540 CTL 36 5/8" x .200 x 36" x .004"/ft ball screw $ 47.43
TBS-29 Thomson 5707540 CTL 72 5/8" x .200 x 72" x .004"/ft ball screw $ 91.16
TBS-412 Thomson 7820827 5/8" x .200 single ball nut $ 42.91
To convert my mill, I will need about 3' for the Z axis and 3' for the X and less than 2' for the Y. This will allow me to mess up once on each axis when machining the ends. Of course if I don't then I will have some left over, perhaps a saleable scrap.
My original plan was to use Roton 3/4" ball screws for all axes.
Roton option - 3/4" on X Y & Z with doubled single ball nuts $438.34
Screw, 3/4 X .200 x 96" .003-.009" 24.56/ft 8 @ 24.56 = 196.48
Round Ball Nut, 3/4 X .200 single ball nut 6 @ 40.31 = 241.86
Option 1 - Nook 3/4" on X Y & Z with doubled single ball nuts $1,023.37
Upgrade to XPR screw additional (16.49/ft) + $ 131.92 = $1,155.29
Nook SRT7296 3/4" x .200 x 96" ball screw 1 @ 272.35 = 272.35
Nook SBN7201 SRT series single ball nut 6 @ 125.17 = 751.02
Option 2 - Nook 3/4"-Z & Nook 5/8"-X Y w/ doubled single ball nuts $721.69
Upgrade to XPR screw additional 65.96 + 49.08 = + $115.04 = $836.73
Nook SRT7248 3/4" x .200 x 48" ball screw 1 @ 136.18 = 136.18
Nook SBN7201 SRT series single ball nut 2 @ 125.17 = 250.34
Nook SRT7540 5/8" x .200 x 72" ball screw 1 @ 110.41 = 110.41
Nook SBN10325 SRT series 5/8" single ball nut 4 @ 56.19 = 224.76
Option 3 - Nook 3/4"-Z & Thomson 5/8"-X Y w/ dbl single ball nuts $649.32
Upgrade to XPR screw on Z axisadditional (16.49/ft) + $ 65.96 = $715.28
Nook SRT7248 3/4" x .200 x 48" ball screw 1 @ 136.18 = 136.18
Nook SBN7201 SRT series single ball nut 2 @ 125.17 = 250.34
Thomson 5707540 CTL 72 5/8" x .200 x 72" ball screw 1 @ 91.16 = 91.16
Thomson 7820827 5/8" x .200 single ball nut 4 @ 42.91 = 171.64
Option 4 - Nook 5/8" all (counterbalance head) w/ dbl single nuts $521.07
Upgrade to XPR screw additional (8.18/ft) + $ 81.80 = $602.95
Nook SRT7540 5/8" x .200 x 72" ball screw 1 @ 110.41 = 110.41
Nook SRT9987 5/8" x .200 x 48" ball screw 1 @ 73.52 = 73.52
Nook SBN10325 SRT series 5/8" single nut 6 @ 56.19 = 337.14
Option 5 - Thomson 5/8" all (counterbalance head) dbl single nuts $396.05
Thomson 5707540 CTL 72 5/8" x .200 x 72" ball screw 1 @ 91.16 = 91.16
Thomson 5707540 CTL 36 5/8" x .200 x 36" ball screw 1 @ 47.43 = 47.43
Thomson 7820827 5/8" x .200 single ball nut 6 @ 42.91 = 257.46
Conclusions:
The Nook ball nuts get really pricey fast!
Option 1 is just too expensive for my budget and more than the mill requires
Option 3 is more expensive than option 4 with the XPR upgrade with much lower accuracy
Option 5 is about the same cost as the Roton plan, but with better accuracy
Option 2 w/XPR is twice what I budgeted, but the best choice for accuracy & life but is about $230 more than option 4
I choose ...
Option 4 w/XPR ball screws at about 50% more than I planned. I will lose some accuracy with the SRT series Ball nuts but save a ton of money. Compare the 3/4" ball nut prices above.
Wow! Nook prices have really increased a lot since I bought mine in Jan. 08!
BTW - You lose no significant accuracy using SRT nuts with XPR screws. The screws are the primary determiner of accuracy, not the nuts.
Regards,
Ray L.
Absolutely fantastic! I have been waiting from nook,Rockford,thk and motion all for quotes they are all taking their sweet time. Thank you very much for sharing this. Couple questions for you:
These prices do not include any type of end machining or end support bearings, correct? How will you get that done?
When you double up on the ball nuts how exactly do you do this? Is one rh and the other lh?
Thanks in advance for all of your help
Rob
Absolutely fantastic! I have been waiting from nook,Rockford,thk and motion all for quotes they are all taking their sweet time. Thank you very much for sharing this. Couple questions for you:
These prices do not include any type of end machining or end support bearings, correct? How will you get that done?
When you double up on the ball nuts how exactly do you do this? Is one rh and the other lh?
Thanks in advance for all of your help
Rob
I will do the end machining myself. I have a lathe with a tool post grinder and a torch for annealing the ends.
By putting one ball nut on each side of the mounting plate with belleville washers they can be preloaded and in this way backlash is reduced/removed. At least that is my understanding.
End machining, Is this something that most tool shops should be able to do?
Thanks
Rob
Thanks, Bigspike! That information is golden. I see that the 5/8 .001 accuracy ballscrews are really a no-brainer in terms of cost vs. accuracy. I will likely be getting those!
I have four things to decide before I move on getting my ballscrews:
1. I'm guessing you can get any custom length, like 54 inches instead of 48? I will have about 49" travel on my x.
2. I have to decide on single with preload vs. double ball nuts. I will study your post some more.
3. I wonder if .200 lead is good? Will max speed be fast enough? I think that the 5-start acme screws I was initially looking at were .5 lead.
4. I am also wondering about end machining costs. That will complete the picture for me. I will then re-design my machine around these ballscrews.
Thanks for starting this thread. I hope you don't mind my jumping in.
-Steve
Some people on here have posted that they had some trouble getting the ends machined due to the hardness of the steel. Check with your local Nook distributor for machining quotes, they can also provide end blocks and talk to other machine shops. Be sure to tell them the hardness level of the material.
Shops that specialize in machining ball screws will generally use an induction heater to anneal the ends. If using a torch, the ball screw needs to be wrapped in water soaked rags to protect the screw that will not be machined from overheating. Premature failure can result if the heating is allowed to be conducted down the shaft.
Some people just use a tool post grinder to grind off the hardened threads.
1. From what I understand, the 5/8" XPR ball screws (.631") are only sold in 4' & 6' lengths.
2. I am working on a design for pre-loading 2 single nuts, the adjustable preloaded double nut is just toooooooo expensive.
I have searched the forums and if anyone has a design for an adjustable pre-loading mount with 2 single nuts, that will fit an RF45 mill, please post a picture or link here.
3. I think the .200 pitch is what most people use for accuracy on these mills.
4. End machining can add significantly to the total cost. I have plenty of time but little money so I will have to make it work. End blocks at reid supply are $308.70 ea. WOW!
Another thing to consider are flanges. They are $36.17 each at reid supply.
If you plan to use 6 ball nuts, then flanges would add $217.02. If you make your own, a 15/16-16 tap will cost much less. Amazon has one for $25 - requires 7/8" drilled hole.
In terms of preloading a pair of ballnuts, you basically want to fix one to the mount and use belleville washers when mounting the second. You can see what I did on my g0704 Z axis here:
In this case, the right ballnut is bolted to the block of steel and the left ballnut is being held by about 160lb of spring washers.
This way even if the preload fails you still have one ballnut to fall back on.
One advantage to proper ballnut mounts is that they almost certainly use matched pair AC bearings with heavy preload. Not worth the money for most DIY applications though.
Here is what I am using on my Harbor freight conversion. About the size of a Grizzly G1005. A little explanation. This is my x-axis screw and the red arrow is pointing to the block that is bolted to the saddle. The first ball nut to the right is threaded into that block with locktite. There is a gap between that one and the second ballnut. To the right is another aluminum block. Between that block and the right ballnut is a stack of 2 wave washers that have a fully compressed force of 100 pounds each. I have the bolts tightened to provide about 175 pounds of load. This had worked well. The aluminum block on the right has a recess milled into it to prevent the ballnut from turning. I only have about 2 months on this setup but I see no reason it won't continue to work well. I am planning a similar system for my G0704, and when/if I ever get a RF45 type mill I will use the same type setup.
As for prices, I should point out that I bought my 5/8" Thompson ball nuts were $30 each and the screws were $18 per foot in any length you want. No flanges needed since I just threaded the aluminum blocks. My tap was an e-bay find for $8. Accuracy seems to be good but I haven't measured them yet, but parts come out accurate. I will be measuring my screws and inputting a mapping table into LinuxCNC if necessary on my G0704.
Last edited by 109jb; 07-29-2012 at 02:19 AM.
The Nook screws are hardened pretty much all the way through. On mine, even after we got well past the threads, they were very difficult to machine.
The Nook pricing for end machining was astronomical - on the order of $500+ per screw. If that has gone up as much as the screw prices, it would be closer to $800 now.
Regards,
Ray L.
Don't know if anyone else would like to do it, but I machined an end out of a piece of steel stock and then welded it to the end of the screw on one end and then I just had to turn the other end rather than having to thread it too. I kept the screw and the machined end in alignment by chucking the ballscrew in my lathe chuck and holding the machined end in the tailstock. Worked ok. I had to do some mild straightening afterward.
Roton double ball nuts... Made by milling 1/8" slots into the ballnuts then making a carrier with tangs to go in between them.
Reply with Quote
