What's New


Workshop Stuff


Moving the mill

Spindle Noises

ShumaTech Digital Readout

ShumaTech DRO Continued

DRO-350 Repairs

South Bend 9" Lathe

South Bend 405 Lathe Bench

Grizzly Mill Revisited

Surface Grinder Rebuild

Surface Grinder Continued

Grinder April 6, 2008

Grinder April 20, 2008

Grinder August, 2008

Grinder September, 2008

Grinder November, 2008

Grizzly G3103 Mill

Grizzly G3103 Mill
Rebuild - Part 2

Moving the Shop

Moving the Shop 2

Bringing Home a Sheldon 12" Shaper

Sheldon 12" Shaper 2

Sheldon 12" Shaper 3

Sheldon 12" Shaper 4

Sheldon 12" Shaper 5

Sheldon 12" Shaper 6

Sheldon 12" Shaper 7

Sheldon 12" Shaper 8

Sheldon 12" Shaper 9

Way Alignment Tool



Email Jim


Grizzly G3103 Repairs
January 24, 2010 through March 30, 2010

Although I continue to improve the quality of the mill-work I produce, I got to be less and less satisfied with the tolerances I could hold with the Grizzly G3103 mill.  In early January, I finally decided to attempt to rescrape the mill. I've been working up to this project for a long time and think I have acquired the tools and the knowledge to be successful in improving its accuracy. 

After spending a year rescraping the DoAll surface grinder, I decided that I would afford myself a power scraper.  Hopefully this would speed up the next big job a bit.  I purchased a used Biax 7 ELM and got familiar with it by scraping a small surface plate and a couple angle blocks.  Using the power scraper is a whole different process and being able to adjust the stroke length and speed allows you to change the pattern you leave on the surface you're scraping. I'm impressed with what I am able to do so far and know that the more time I spend using the tool, the more comfortable I will become in using it.  I am a bit surprised at how sore I get from using it.  I figured it would be a lot less tiring than hand scraping.  That's only partially true.  It still tires the muscles to hold the scraper so that it makes good contact for hours on end.

One of the last items I wanted to have before starting the mill rescraping project was a cross check level.  A cross check level allows you to view the deviation from flat on two planes simultaneously.  This is a nice tool to have when you're leveling a machine tool, though I really wanted one to use with my shop built King-Way style way checking tool. Using the cross check level with the King-Way tool allows you to see the deviation between the two ways in both relative height (across the ways) and tilt (along each way) as you move the tool from one end of the ways to the other.  With a single vial, you can only read one plane or the other for each pass.

The job of making the level was pretty simple.  As I had done when I made my 6" precision level (for use where there was limited room), I purchased a couple level vials that are replacement parts for the Starrett 199A level.  I also bought some cast (ductile) iron for the body and brass tube and bar stock for the vial holders.  The T shaped body was cut from an eight inch diameter, two inch thick round.  I milled channels for the vial holders, then spent a week heating and cooling the body to relieve the internal stress.  After a number of heating and cooling cycles, I machined to top and bottom flat, then heated and cooled the base a couple more times.  After each cycle, I checked for warpage.  Finally, I ran the body through the surface grinder, then scraped the bottom surface.  It took about 10 scraping cycles with the Biax to get even coverage of bearing points and then another twenty-some cycles with a hand scraper to bring the count of bearing points up to 30 or more per square inch.  I've checked the flatness over a couple of months and there is no sign of warping.  I still need to build the adapter to attach it to the lathe tool, but for now it would help with leveling and scraping the mill.

Now that I had all of the tools that I thought I would need for this project, I had no excuse not to get started.

Crosscheck Level
Crosscheck top
Machined, ground, then scraped from a disk of ductile iron.  Heated and cooled a few times to help normalize the iron before scraping to 30+ per inch.
Two replacement vials from a Starrett 199A give 0.0005" resolution.  Brass screws didn't work well and were replaced by studs and round brass nuts

The first step in hopefully improving the accuracy of the mill was to measure the mill in its current state to see how far out of square it was and to give me an idea of what I needed to accomplish.  I started a log book and began to enter the measurements of the mill. Using my assortment of levels, straight edges and my K&E autocollimator, I leveled, then mapped the ways of the mill.  This would give me a starting point for making the mill square.

During the couple years that I have had this mill, I've complained about the Y axis the most.  With a level on the table and the saddle close to the column, the level's bubble would show one side of the table high.  When I cranked the saddle out and away from the column, the bubble would move, showing the other side high.  We're only talking about a few thousandths of an inch, but this was one of the things I wanted to fix.  I haven't heard of any other 6 X 26 mill owners complain of this, but I don't know how many people have checked for it either.  I came to the early conclusion that the twist I have found in every axis (X, Y, and Z) is probably due to the castings not being aged before and during the machining process.  After the ways were machined, the castings warped and twisted.  There's no other explanation I can think of that would cause my straight edge to touch on the right rear and front left of a single way.  Add to this that the other way of the same axis shows a similar pattern when printed with a blued straight edge.  So my project is to map out how much twist I have and try to scrape it out.  

My Levels
Small surface plate
In addition to the cross check level, I have a Moore and Wright (rear) at 0.0035" per foot, an import box level at 0.0002" per 10" and another shop made level with 199A vial at 0.0005" per foot.
To test out my "new to me" Biax 7ELM scraper, I scraped the back of a cast iron lapping plate.  To get over 30 points per square inch, I machine scraped 10 cycles and hand scraped another 25.

I started my measuring by leveling the mill using the mill's table as the reference plane.  I then measuring the displacement from level of each of the ways with the autocollimator. After that, I removed the table and checked the top saddle ways for level. I then re-leveled and took another set of measurements.  I then removed the saddle and took my measurements again. Unfortunately, there was a difference in level between each way in each axis.  This wasn't going to be an easy measuring or scraping job.

To measure the spindle plane, I added a 1" diameter, one foot long precision bar in a 1" end mill holder and put this into the R8 spindle taper.  Using a quarter wave, front surface mirror and my K&E autocollimator, I checked the alignment of the spindle to that of the Z ways in both the fore and aft and side to side planes.  The closest match was between the spindle and the right Z way.  These two planes were only out by a couple thousandths over a foot of distance.  The worst alignment I found was measuring the top ways on the knee where the left way sloped one direction from the column and the right sloped the opposite direction.  At this early stage, I thought that this was caused by warpage in both the column (Z ways) and the knee itself being a bit twisted.  I learned later that this was only partially true.

autocollimator 1
Autocollimator 2
Autocollimator view
The surface plate was leveled, then the autocollimator was checked for level
Shooting at a 1" front surface magnetic mirror on a piece of precision shafting held in a collet.
Once the crosshairs were lined up and readings taken, I compared the Z ways. Nothing matched up.

Having read "Machine Tool Reconditioning" by Edward Connelly, I knew that I needed to designate one surface as the master surface that all other surfaces would be aligned to.  I also knew that I wanted to scrape as little as possible while still achieving a square machine.  Since the mill had no true 'master surface' to start with, I would need to create it.  The position of the surface I would choose needed to be one that allowed me to change the existing angles of the mill's ways the least while allowing a precise 90° angle between the Z ways and the connection (mounting area) that holds the mill's head assembly..

I removed the mill head from the machine and checked the horizontal surface that the head swiveled on.  Again using both the autocollimator and levels to double check each surface, I leveled the mill on its adjustable feet using the horizontal head connector surface.  I then checked my Z ways for squareness.  They were off, but one of the Z ways was pretty close to square with surface on the top of the column.  I decided that I would start scraping the connector surface. My goal was to create a plane that was 90° to a plane somewhere between the current plane of the Z ways that now sloped in opposite directions.  Once I had the Z ways scraped flat and in a single plane, I would revisit the circular connection on top of the column and make sure it was as close to 90° from the Z as possible. 

Top of column
More checking
The top of the column showing how few areas (blue) made contact with my small surface plate.
I used the machine adjusters to level the column top so that the Z ways aligned to their best position.

Leveling the mill is not a trivial task when you don't have a single flat axis to work from.  I had to level to a position that was 90° from the imaginary plane between the two Z axes. I.E., one way titled toward me by a small amount and the other way tilted away from me.  Trial and error and some guessing was the only way to accomplish this.  However, I find that in leveling a 4 legged tool, it helps to adjust the leveling feet diagonally rather than side to side and front to rear.  I needed to end up with the fore and aft inclination of the Z axis vertical as well as having the left / right inclination vertical. I ended up with a plane where the difference from 90° vertical on the two column (Z) ways is about 0.06° in opposite directions.  In other words, one way is 89.04° and the other is 90.06°. 

The above angles are in the fore to aft position.  For the left to right inclination, I had it a little easier.  The small ledge on the outside of each of the two Z ways were milled pretty close to the same on either side.  Using the box level, then checking my angles with the autocollimator, I positioned the adjusting feet to make these two surfaces as close to vertical as I could. I split the difference from vertical so that I would be removing approximately the same amount of metal from each way as I scraped.

Scraping and checking
Scraped flat and level
Scrape and check, scrape and check.
Repeat, rinse, and repeat again until it levels.
Some period of time later, we have a scraped flat and level surface.  A good place to start from.

Once the mill was leveled, I started scraping the two cast iron rings that make up the mounting area for the head.  Between the inner and outer ring is the tee-slot used to capture the bolts that hold the milling head assembly to the column. There is a circular enlargement to the tee-slot in one section to allow the bolts to be removed.  As I scraped, I checked my progress by alternating between my hand-scraped small cast iron surface plate to check for flatness and testing with my 0.0002" resolution box level to keep the surface level in both planes.  Before I began scraping, I used a crosscut file to quickly remove stock from the front side of the column mount, then switched to a hand scraper.  The factory surface on this critical connection was pretty rough and my flexible blade kept catching on the large ruts, so I changed to a heavier, less flexible blade and powered through the rough areas.  This connection is critical to not only get flat, but also have a large enough number of bearing points to allow the mill's head to swivel smoothly.  It's important that the head stays in a single plane as it swivels from left to right so there needs to be enough bearing points to facilitate a good connection between the two pieces.  None of this was provided by the factory finish.

On completion of the top column mount, I now had my one flat surface to work from.  I decided to move on to the milling head assembly.  This assembly is made up of four sub-assemblies.  Just above the column mount is a "T" shaped connection that allows the head assembly to swivel left and right.  This piece is cast iron and has a flat base on the bottom with a protruding 3.150" cylinder that fits down into a similarly sized hole in the column mount.  To mount the milling head in the front and the motor plate to the rear, there are two more circular flats with protruding cylinders.  These circular flat mounts with cylinders are a bit smaller in diameter than the one on the bottom and allow both the milling head and motor plate to rotate about the horizontal axis.  Because each of these connections has the protruding cylinder in the center of a surface that needed to be scraped flat, it presented a problem.  I couldn't just lay the surface on a flat plate to check flatness without boring a hole in one of my surface plates.  I actually considered this as I have a couple import 9" X 12" granite plates that are still in their shipping boxes, but since all three cylinders are different sizes, I'd have to bore the hole three times to check and scrape all three surfaces.  I decided this would be more work that it was worth.  Instead, I used three 3" X 12" granite parallels I had match ground on my surface grinder.  I laid these in a star configuration with the cylinder from the "T" connection in the center of the three parallels.  While the three parallels didn't quite cover the entire circular flat, it covered most of it.  I covered the three parallels in Prussian blue and used them as my surface plate with a hole in it.  After each marking and scraping operation, I set the "T" so that the area that didn't get marked in the previous cycle got marked in this round.  I am not sure how they originally machined the connections of the T shaped piece. There was a slight depression around where the cylinder stuck out from the flat portion of the mounts.  I will say that when I set the connector up on my surface plate and checked the two side surfaces with my master square, they weren't too far from 90°.  No they weren't dead nuts on, but they were closer than I expected.

Test fitting the T
Before scraping
After scraping
Test fitting the "T" that holds the milling head.  You can see some evidence of Prussian blue between the "T" and the column top.
This is what the surface looked like before scraping.  Notice the slight depression between the center cylinder and the two bolt holes.
This face has been scraped enough to make sure that this surface is as close to 90° as I can measure with surface plate and Schmidt master angle plate.

I scraped the bottom surface flat, then blued up the top column connection and checked to make sure I had good contact between the "T" and the column mount.  Once I was satisfied with the contact while the "T" was in its usual straight ahead position, marked by zero degrees on the scale attached to the "T" and column, I cleaned the blue from the "T" and checked other settings.  At each increment of 10° in position, I checked for good contact and also used my box level to make sure that the surface where the milling head attached remained level. 

With all three of the "T" connections scraped, I still had two surfaces left to deal with on the head assembly.  The motor mounting plate and the milling head.  There wasn't a lot to do on the motor plate as it is not critical if the motor is not aligned to seconds of arc.  I scraped it to good bearing and made sure that it stayed close to level as I rotated the plate through its range of possible positions.  Setting up the milling head was a different story.  My goal in working on this mill was to improve its accuracy.  Getting the head to be dead on 90° to the table meant that this connection, as well as the column to "T" and the column to Z ways had to be as perfect as I could accomplish.  In order to scrape the proper angle into the mill head connection, I needed to find the center line of the R8 internal taper in the spindle.  

First, I removed the belt guard cap screws, then removed the spindle pulley.  The milling head can then be removed from the guard.  I inspected the taper and went over it with a slip stone to make sure there were no burrs.  I had purchased some new ETM brand end mill holders that checked out to be pretty concentric and mounted some precision shafting in place of an end mill.  This would, in theory, extend the center line of the R-8 taper out to where I could check it against the mounting flange on the surface plate.

head 1
head 2
head 3
The milling head has been removed and had a towel stuffed into the center hole to prevent chips from getting into the spindle bearings.  A 7/8"
precision shaft in a collet allowed the tilt of the head connection to be measured when the connection was placed on a surface plate.
The connection was scraped until the distance between the surface plate and shaft remained constant over the length of the shaft.

It worked pretty well.  I first set up with my new 3/8" end mill holder and a piece of 3/8" drill rod about a foot in length.  I cleaned up the flange a bit with a hand scraper as there was some ragged edges near the bottom edge of the flange.  I cleaned and polished my Starrett surface plate and carefully laid the head, flange down, on the plate.  I then used a 'half-tenths' indicator to check for parallel along the drill rod.  This would show me the relationship between the R-8 taper and the head's mounting flange.  As I measured from end to end of the drill rod, I also rechecked with the spindle turned in quarter turn increments.  The end of the drill rod was low by 0.012" at the best and 0.016" at the worst compared to the section of drill rod just outside the end mill holder.  That's a lot.  I was also showing 0.004" run out at about 12 inches from the end mill holder.  There was only 0.0005" run out just outside the end mill holder.

I swapped a 7/8" collet for the end mill holder and used a 12" piece of 7/8" precision shafting.  I got almost identical results.  This told me that there was a slope of about 0.001" per inch on the head mounting flange.  Since the flange has a diameter of about 5 inches, that meant that I needed to remove about 0.005" from the top of the flange and nothing from the bottom.  It took about 15 cycles of measuring, marking and scraping to get the flange parallel to the precision shaft. It took another 15 to 20 cycles to get a nice bearing surface on the circular area.

Column to Knee test
Mill Base
One of the many tests between the knee and column.  The bag helped keep the disconnected lines filled with way lube from getting everywhere.
A shot of the mill that few people get to see. I set the 4 connection fasteners to 75 foot pounds to make sure they were good and snug.

Once I was satisfied with the surface of the milling head and head connection, it was time to lay the mill on it's back.  I had considered scraping the Z ways with the mill in a vertical position, but this would make checking the plane of each way much more difficult.  It would be easier to lay a level on a horizontal surface than to hold a box level to a vertical surface. In order to position the mill on its back, I would need to support it so the Z ways would be held in an exact horizontal plane.  Again, the mill needed to be horizontal in both the front to rear and side to side planes. As it turned out, I had a couple of 4" cylindrical pieces of steel that were very close to the right height to prop up the end with the head connection.  I leveled the Z ways length-wise, then shimmed each corner to get the side to side plane level. 

Column 1
column 2
The mill's base and column have been laid on their back and supported so that the Z axis was level in both fore and aft and side to side planes.
With the Z ways now horizontal, I started with one way high toward the camera and one high at the opposite end. By this shot, they are almost level.

With the box level measuring the head connection plane, now vertical instead of horizontal, I made sure that the plane was parallel to the force of gravity. I then mapped out the Z ways using the box level. The left way was sloped up at the base end and a couple thousandths low in the center and at the head connection end it was higher than the center, but lower than the base end.  The right flat way was sloped up at each end and low in the center, though, as with the left, it was higher on the base end than the head end.  Each of the two ways were about 22" long and 2" wide, so I could use one of the granite angle straight edges that I had made as a scraping master.  Because my goal was to make the ways perpendicular to the head connection, as well as flat in both the length and width, I started out printing each way from the Prussian blue coated granite master and guessing at how much material to remove to achieve all my goals.  Because it was necessary that the two ways were in the same plane, I would need some sort of jig to keep the ways the same height and even with each other.  It could be done with a couple of parallels to lift the box level over the raised area between the ways, but using the parallels introduces more room for error as there are more pieces to deal with.  However, if I scraped the knee Z ways flat using my surface plate, I could use the knee as a template.  Between the two setups I would be able to scrape the plane of the Z ways pretty close to true.

Though the surface the indicator base sits on does not support or guide the knee, it is easily scraped flat and square with a surface plate.
Once flat and square, it makes for another way to ensure that the obstructed Z ways are flat, true and at the same height. They also look sharp.

I blued the master, printed both Z column ways and scraped about 12 cycles before I started getting sore from working on the concrete floor.  Working on ways about 8" off the floor wasn't going to allow me to produce the best finish.  I solved the issue by placing an old Formica counter on top of the mill's base, then hoisting the mill column casting to rest on it.  I used the same steel cylinders to block up the connection end and shim stock to level the base.  The ways were now about 30" from the floor.  Still a bit too low for my liking, but a whole lot better than working on the floor.  I had switched to the Biax 7ELM scraper after about the third round of scraping - just enough time to figure out where the problem areas of the ways were.

As it turned out, near the vertex of the dovetail ways, on the wide flat way, there was an edge that stuck up higher than the rest of the ways.  This was caused by the ways of the knee not contacting that section of the way as the knee was moved up and down the column.  Because it was close to the vertex, this area was not able to be scraped easily with the power scraper without the chance of putting dings or gashes in the angled way. To scrape this area I switched to a 1/16" thick carbide bladed hand scraper that had been further thinned by a couple years of sharpening the sides of the blade with diamond paste.  A couple cycles of scraping had the edge knocked down to the level of the rest of the way.  By about cycle 16, I was starting to get points in the mid-section of the ways.  I was still high on the base end, but it was coming down cycle by cycle.  I had stopped scraping the very end by the head connection side of the ways once I had about a two inch section of the end of the ways that were flat and showed points all across the width of each way.  Leaving this end unscraped would effectively lower the opposite end and bring the way into being perpendicular with the head connection.

After I had bearing points showing along the entire surface of each of the flat Z ways, it was time to make sure that the ways were in the same plane.  As I said before, I would accomplish this by using the knee as my template.  However, first I needed to check the ways that contacted the column on the knee.  Back to the surface plate. The knee was as bad as everything else I had checked. With the ways that contact the Z ways of the column sitting on the surface plate, I used a "L" square to check to see if the Y facing ways were at a 90° angle to the plate. They were close, but one was about 91° and the other was 89°.  More twist.  I then checked what amounts to the height of the Y ways by placing a piece of 1/2" drill rod in the vertex of the flat and angled ways. and using the "L" square to see if it was perpendicular to the surface plate.  This measurement was way out. (I knew this from using the mill.) The knee leaned to the right.  To visualize which plane this is; you're above the mill looking down on it.  The connection for the head assembly is in the center of an analog clock face.  12 o'clock is behind the back of the column and the mill's table runs from 9 o'clock to 3 o'clock. The knee would normally extend toward 6 o'clock, but in this case, the Y axis points to between 6 and 7.    It faced to the left of 6 o'clock by an amount equal to 0.0045" over 6".  To make the knee point at a right angle, I would need to remove metal from the knee's right Z way.  Since the crank end of the knee was not only facing a bit left, but also a couple thousandths lower than the column side of the knee, I would need to correct the Z ways on the knee in two planes.  Fortunately this isn't as hard to do as it is to describe.

more measuring
With the first print of the knee's Z ways, we find that there are only small contact patches in each end of the ways. Not too good for rigidity.
Checking for squareness of the tops of the knee's Y ways. In this check, the top of the L square should touch. Instead, the bottom does.

I made a jig to hold the knee with its Z ways facing up and clamped it to my cast iron saw table.  Out came the tools and I ground, filed and scraped for two evenings to remove enough metal to make the knee sit on it's Z ways and be close to perpendicular to the table in all directions. Or close enough for now. All I really needed at this point was for the knee to be perpendicular to the surface plate and for there to be enough bearing points that I could use it to check the plane of the Z ways on the column.  Removing about 4 thousandths from a way - even a short one like the Z on the knee is a lot of metal when scraping.  I didn't even try.  I started with a belt sander and some 60 grit, progressed to a bastard file, then the Biax, and finally hand scraped a couple cycles.  In between every few cycles of each tool, I cleaned up the knee with a shop vacuum and alcohol wetted towels and put it back on the surface plate.  I didn't want to remove too much material. Once that the mill is back with the column in the vertical position, I will add the gib and check the Y knee ways for level and being parallel. Eventually the crank end of the knee will be a small amount higher than the column side.  This will allow the knee to take the weight of items on the table and the forces subjected to the knee in machining without sagging.  The specs I've heard amount to about a half thousandth high at the crank end.

When the knee was square, I blued its Z ways, then lifted it on to the column which was still lying on its back.  I slid it along the ways; from the head end to the base end and looked at the pattern it left on the column's Z ways.  Amazingly, the pattern was very close to the same pattern I saw when I printed each way separately with the granite master.  If I had seen markings on both outside edges of the ways, or both insides, or one of each, I would have had to adjust the column ways to bring the two of them into a single plane.  As it was, I could continue my scraping on the column Z ways, confident that the ways were flat with each other.

I did encounter one issue with the knee to column fit.  I had removed enough material that the vertex of the non-gib side of the knee was touching the machined groove that is cut in the vertex of the column Z way.  There are three choices here that I can think of. Add a piece of Turcite B to the angled way of the knee ways to build them up, or cut the tip of the vertex of the knee ways down, or deepen the channel at the vertex of the column's non-gib side flat and angled Z way.  As I looked at the position of the knee on the column, I saw that the knee was closer to the gibless side of the ways, than to the gibbed side.  I could deepen the channel of the right column Z way and shorten the vertex of the knee ways without running out of width on the gib side of the column's Z ways.  I had read about this happening in Connelly's book and knew that it was a real possibility because of the amount of metal I needed to remove to straighten out the knee, but having the knee square with the column was more important than having the ways shifted a bit to the left. 

55° Template
55° template
My hand scraped 55° angled template has been "blued" with red marking fluid. I'm ready to print the non-gibbed side of the angled knee Z way.
The template is in place and is transferring color to the way. Using two templates, I can ensure that the way will be at the right angle as well as flat.

By the beginning of March, I was getting close to having the Z ways finished.  I was now able to measure the Z on the column with my 2 ten-thousandths sensitivity box level and the bubble never traveled further than 4 lines. Eight ten-thousandths difference between the lowest and highest spots on the ways.  The scraping goes a lot slower at this stage since there are many bearing points to scrape off in each cycle.  Also adding to the lack of speed is that the straight edge needs to be spotless before you can add the Prussian blue and make a print on the ways. If you pick up a fleck of metal in the blue, you won't get a true print or maybe no print at all.

I was also checking the column Z ways by bluing the knee and setting it in position, then sliding it along the column. I had my box level on the crank end of the knee and in addition to watching the bubble in the level, I watched the markings left on the column ways.  In the early scraping cycles, the bubble moved from the left, to the right as I traversed the ways. It moved much less now, but I still had about a half ten-thousandth to go.  I had decided I would stop when I could keep the bubble within 1 line to either side of the box level and I had full coverage of bearing points. This would be about a half of a graduation on the Starrett vial. Trying to chase after no movement of the level bubble was not something I was planning on doing - even if it was possible.  With 2 ten-thousandths per 10 inches sensitivity on the box level, almost everything I did changed the bubble position.  Whether I touched the level with a warm finger, breathed on it, didn't position it in the exact same spot, you name it and the bubble moved.  I had already improved the Z ways to be much more accurate than they were when the mill arrived a few years ago. The knee now moves with the touch of a finger, rather than a stiff push, so I know it is going to be a great improvement.  My biggest question is whether the mill has finally settled in and if I can expect the new accuracy to last.  I tend to think that if the casting does continue to settle in, it will do so to a much lesser degree than it has done up to this point.  I am hoping so, anyway. I have a whole lot of time invested in this mill now and I'd hate to think that it would twist like a pretzel after all the scraping I've done.

level 1
level 2
level 3
With a 0.012" feeler gage under the left side, the level bubble is centered.  Scraping the bottom right angled knee Z way, the left side will rise.
A day later, the feeler gage has now been removed and we have 5 increment lines to go before the knee will level out.
By the end of the next day, we're showing within a ten-thousandth of having the knee level.  Getting the knee level here will allow the table to be level.

March 13, 2010
This afternoon I figured I'd scraped the column enough.  With the box level set up on 2 pieces of 7/8" diameter shafting, I could move the level from the base to the top and the most the bubble moved was less than 1 division.  I'd been at this point for 2 days and the only thing I was accomplishing by scraping was more bearing points.  Not to say this is bad, because it isn't.  It's just that this is about as good as I am going to get it without driving myself into boredom.  I'm ready to move to the next axis.

Using my engine hoist, I lifted the base and column assembly and while it hung in the air, I checked and cleaned the base and mounting points.  I set it on the stand and bolted it down, being careful to torque the bolts incrementally and to about 75 foot pounds.  I then leveled the column using the box level on the Z ways.  The difference between trying to level it before and after scraping the ways was drastic. One half hour later and I was close to dead level. The connection for the head was right on level in the front to back plane and about 1/2 thousandth high on the left.  This was due to having to make an adjustment to the left Z was dovetail.  I scraped the column to head connection for a few hours after dinner and called it a successful day.

The saddle is pretty rough. This is the Z ways side. Not a lot of contact as it came from the factory.  I'm not impressed with the quality.
This is the X axis (table ways) side of the saddle. I've been using this mill for a couple years and there is still very little contact area showing.

March 18, 2010
I started working on fitting the knee a couple days ago. When I first tested it on the Z ways, I found that my scraping on the Z ways side of the knee had been a success.  The knee now is aligned 90° to the face of the column. That is to say that if you looked down on the mill from above the knee is in the 6 o'clock position. As I've said, it used to be somewhere between 6 and 7 o'clock. Not exactly square with the column.  The next step was to measure for tilt between the two top ways of the knee. This would be the Y axis ways and I am measuring the height difference between the left and the right. The left is low. It started very low. About 12 thousandths.  That is a lot for ways that only have 5 or 6 inches between them.  I also saw this same measurement while checking the knee on the surface plate.  It's nice to see that my measuring was accurate.  On the other hand, I am pretty disappointed that the knee is tilted with the right side higher than the left.  One would have thought that this would have been caught by some sort of QA process at the factory.  If the saddle above the knee and table above that are all perfectly symmetrical, the left end of the 26" long table would be 20 to 30 thousandths low.  I've measured it and it isn't, so there is something else messed up that I haven't gotten to yet.  Lovely.

On the surface plate, I also saw that the tilt between the Y ways close to where they meet the Z ways, to the end where the crank is was a couple thousandths low at the crank end.  As I said before, this measurement should actually be a fraction of an inch (couple ten-thousandths) high. Now that I measure them with the knee attached to the mill, I find that one side is 3 thousandths lower at the crank and one side is 1 thousandth lower.  This (I'm guessing) is because the knee is twisted.  As I remove metal from the bottom of the angled right knee Z way, the height of the left Y way will increase.  Once I get the left to right tilt out of the knee, I'll get back to scraping additional metal from the top of the knee Z ways, which will raise the crank end of the knee in the Y axis.

Grinder marks?
Grinder Marks?
Pondering the difference between the pattern left by a Biax scraper and the marks left in the center, it looks that a grinding wheel was used.
After a few cycles with the Biax scraper, I am beginning to get some contact. I hope I can work out the deep gouges in the center of the saddle.

In a couple of days scraping, I have decreased the left to right twist from 12 to 2 thousandths.  To keep the slope of the angled way constant, I am using two templates to print from. The first is my 12" granite angle template.  This template has a 45° angle with a very sharp vertex.  This allows me to blue along the slot at the vertex of the two ways that make up the dovetail way.  I am also using one of my first scraping projects, made about 4 plus years ago, when I first got the bright idea to rescrape the mill.  The template has a 55° angle and is scraped on 3 sides.  I used the Y axis dovetail ways of the Grizzly mill to create the cast iron template and it turned out very nice.  It also took the better part of a month to scrape, but I was a lot slower then than I am now.  By using the pair of templates, I'm confident that the angled way will turn out flat and at the correct angle to fit nicely with the column Z ways.  Because I am removing metal from the angled way, the distance between the two angled ways is increasing.  It is now getting tough to measure the progress accurately as the gib no longer fits with just some shim stock behind it.  By scraping the angled way on the right side bottom of the knee, I am purposely changing its angle in order to raise the left side of the knee.  This in turn changes the angle of the gibbed side clearance.  To help me get the knee snug on the Z ways, I've needed to add more thickness to the bottom of the gib.  This is trial and error work, measuring with my shop set of gage blocks, then shimming the gib with different thicknesses of shim stock to match the measurements.  This is turning out to be very time consuming, but it's the only way to try and ensure that I don't scrape too much off of the angled way. Even so, from the top left of the knee being 12 thousandths low to being only 2 thousandths low went pretty quickly. Even quicker once I got a handle on how the gib measurement was being changed. However, from 2 thousandths to zero tilt is going to require many test fittings to make sure I don't go past zero.  Once I get within a few ten-thousandths, I'll start scraping for removing some of the "drop" from the crank end of the knee at the same time.  This should prove to be an interesting task, but once this is done, the rest of the scraping - the saddle and the table - will be pretty straight forward - or so I am hoping.

Well, so much for hoping that the saddle scraping would be straight forward! The saddle turned out to be the worst piece of the mill.  When I went to measure the saddle so I could use it as a template to keep the Y axis ways of the knee parallel and level, I found out a couple things.  First, the saddle had been ground with what appears to be a hand held grinder to adjust its fit. Second, the fitting - if you want to call it that - had been done so that the saddle was a different height on the left than it was on the right.  Not just a little bit, but about 20 thousandths. It appears that someone knew that the knee was not sitting straight and instead of fixing it, they ground the saddle at an angle so that the table would appear level.  The gouges in both sides of the saddle ways are so deep that I doubt that I will be able to get them all out without making the saddle too thin where the angled ways meet the flat ways. However, seeing as how the saddle ways were only contacting the knee and table ways in a very small area, even with a couple deep gouges left after I finish scraping, the accuracy and rigidity are going to be much better than it was.  Here I have been trying my best to blame the poor fit on warpage, but that's not the case with the saddle. The only way to describe it is shoddy workmanship.

measuring the saddle
measuring the saddle
Next to the cat, we see the indicator reading a little less than 0.002" - and signs that a grinder was used to 'precision' adjust these ways.
The reading is now over 0.004" That's 2 thousandths in about an inch. That's a lot of tilt for one inch of linear distance; and there's more...
From one side of the saddle ways to the other, we have ~0.018" height difference. This is on the Y ways.  The Z ways are further apart, so worse.
It doesn't appear that much of an attempt was made to finish these ways with any sort of accuracy in mind. Very ugly workmanship.

It could be that this mill is an isolated case of a "bad unit" making it through the manufacturer's usually exemplary quality control process.  However, based on what I have seen with this mill, it appears that rather than fix the problem with the knee not fitting as it should have, they did whatever it took to get this one out the door.  I just happened to be the lucky guy who got it.

So, my task becomes reworking this saddle so that the following conditions are met:

Fortunately, the side of the saddle where the Y way stands "taller" is on the (left) gibbed side.  This is good because as I remove metal from the flat way, it will effectively increase the distance from the gibbed side angled way to the opposite side angled way.  Since it is the gibbed side, I will just need to make a new gib to take up any additional clearance I create and it will not change the position of the saddle relative to the knee.  When I went through this same situation on the knee, I was forced to remove metal from the non-gibbed side.  This did shift the position of the knee on the column toward the gibbed side, but I don't think that it changed enough to cause any problems.  I'll know that for sure once I reinstall all of the parts and see how they fit.

DoAll Grinder
DoAll Grinder
The chuck's been cleaned and the wheel dressed.
The last passes on the Y ways have been ground.

I scraped a few cycles on the saddle Y ways with the Biax. Once I got enough bearing points that the Y ways would sit flat against my surface plate, I turned the saddle over and began to scrape the X ways.  My aim here was to try to figure out whether it would be better to remove all of the metal on the Y way side, or to remove some from the Y and some from the X ways.  I also did a check of the mill's table to see whether the table ways were parallel to the table top.  These are very close to perfect. (Hoorah!)  There is less than 0.001" between the highest and lowest spots on the table's X ways.  This is good news and means that if I can pull the saddle into specification, I should have a square mill. -- Or at least to my best ability to measure and scrape it square.

I decided to remove most of the metal from the left saddle Y way.  With this decision made,  I realized that there was way too much metal to remove by hand, so no more scraping tonight.  I needed some time to think the process through.  I had the thought that by removing more metal from the gibbed side, I stood a good chance that the angled ways would not be at the proper 55° angle once I was done.  I got out my dial protractor and some scratch paper.  After measuring and calculating where the ways needed to be when finished, I found that the angled ways would, in fact, be back at 55°. This leads me to believe that the part was initially machined properly, but then the non-gibbed way was ground down to make the table level while compensating for the twisted or poorly machined knee Z ways.  If I removed enough stock from the gibbed side to match the non-gibbed side, then flattened them both out, all angles would be close to where they needed to be.  Once I figured this out and got an idea of how much metal needed to be removed, I wished that I had a working mill to make the job go quicker. I had a whole lot more metal to remove than I wanted to do with hand tools, but I did have my surface grinder.  A mill would be preferable as I needed to remove a total of about 22 thousandths off the gibbed side and about 9 thousandths from the non-gibbed side.  This would almost take out all of the deep gouges in the surface of the flat ways while bringing the flat ways back into proper geometry.  I considered the possible issues with taking that much metal off of the saddle.  The Y axis leadscrew nut needs to fit into the bottom of the saddle so that the lead screw is kept exactly parallel to the Y ways.  By effectively lowering the saddle, this would lower the place where the leadscrew nut attaches to the saddle. This could cause binding or uneven wear on the nut's threads.  Fortunately, the angled Y ways of the knee rise 0.750" from the flat ways and the bottom of the saddle flat Y ways are 0.860" to the point where the nut attaches.  (These measurements were taken before my work began.) That's a total of 0.110" that theoretically could be removed before I would start having interference problems.  I'm now at 0.855" (0.005" from 0.860") after filing and scraping to get more level ways to measure from and have another 0.022" that I think still needs to be removed.  That's still enough room to work with.  To get the nut in the correct position, I can either surface grind to "raise" the mounting area or chuck the nut into a 4 jaw on my lathe and "lower" the top of the nut. There is more than enough thickness on both the saddle and nut to remove the amount I need without interfering in the operation of the Y axis.
X side of saddle
Y side of saddle
A view of the X ways after surface grinding. The two holes between the ways are for the X and Y leadscrew nuts. The X hole is near, Y is behind it.
The Y ways were ground again to make sure that they were parallel with the X ways. You can see that it's pretty thin above the way adjusting screw.

With a dial indicator, protractor and shim stock, I set up the saddle on the surface grinder.  With the X ways against the table, I shimmed until I was level in the lengthwise direction of the Y ways and the angled ways were as close to a 55° angle as I could measure with some 1-2-3 blocks as a reference.  About an hour and a half later, I had two Y ways that were pretty much flat. Grinding five ten-thousandths of an inch per cycle meant 44 cycles of multiple passes to grind first one, then both ways.   At the end of the 44 cycles and a couple spark-out passes, the non-gibbed side with the worst gouges still had some rough spots along one side.  However, I had at least 90% of the way flat and smooth.  Coming from a saddle where I had less than 5 to 10% contact area total for both ways, this would be quite an improvement. I ended taking off 0.0221" and the last couple thousandths made a big difference in the contact area of the non-gibbed way.  To get the remaining gouges out, I'd need to go another 5 to 7 thousandths and I just didn't think it was worth thinning out the saddle further for the little gain I would get.  As it was, I totally removed the lubrication channels in a few places and will need to put those grooves back in.

March 27, 2010
I test fit the saddle on the knee.  There was enough clearance between the area the that the Y leadscrew nut mounts and the top of the Y angle ways on the knee.  Good.  I didn't go any farther with the leadscrew nut mounting as I still had the saddle X ways to grind and probably another grind of the Y ways.  I decided that I should probably clean up the grinder's magnetic chuck.  I hadn't done that recently and the garage gets some pretty wide temperature extremes.  I ended up taking about 0.0012" off the chuck before it was again level and free from nicks.
measuring the ways
measuring the ways
With the saddle back on the surface plate after being ground, I measure all 4 corners of both sides. There was less than 0.0002" difference between the highest and lowest measurements.
I have a bit of a pattern showing in the ground surfaces.  I hadn't seen this in other pieces I've ground, so I did some looking.  The spindle belt has some cracks in it. Looks like I need a new one.

I cleaned up the Y ways on the saddle and laid it on the mag chuck.  I took almost 0.010" off the X ways before I had all of the gouges out.  I cleaned the chuck and the X way side of the saddle, dressed the wheel and set the saddle up for the last bit of grinding on the Y ways.  I took another thousandth or so off of the Y ways, then headed for the basement and the surface plate.

Perfect!  Or as close to perfect as I was likely to get. There's a maximum of about 0.0002" between the highest and lowest points on either the X or Y ways. 

I think I'm going to just leave the saddle with the ground finish.  This configuration would actually be 180 degrees backward from normal grinding and scraping practice, but the size of my surface grinder dictates that I can't follow usual practice.  Most mills and lathes come with the longest ways ground and the mating ways fitted by scraping.  In my case, it would be the column Z ways ground and the knee Z ways scraped.  However, the column of the mill is too wide to fit into my 6 X 24" grinder.  The grinder will actually grind out to about 7 1/4" across and almost 26" in length. I think I can even go a little longer if I disconnect the hydraulic cylinder from the table and run the X axis with the manual crank.  I may give this a try and resurface the mill's table.  I haven't yet decided as the table is in pretty good condition, but I'll know more when I map the X ways of the table.  So far, I've just made a quick measurement to affirm that the ways are very close to being parallel to the table top.

Any way, the saddle came out nicely and since I need to scrape the Y ways of the knee, I'll have scraped on one side and ground on the other.  It is preferable to have the surface that faces up being the ground surface, as it is easier to design wipers to clean the smooth surface of swarf.  In my case the up facing ways on the knee Y axis will be scraped which will make cleaning them a bit tougher, but I have some ideas for building some new wipers that should do a better job than the stock ones.  In fact, I took off the stock ones immediately after getting the mill and building some felt wipers.  I also made covers for both the Y and Z ways to keep the swarf off of them.

Well that's all I have for the moment.  I took a day off today to pick up our son's new car and as usual, what should have been a couple hours in the dealership took half the day.  Oh well.  He should be pleased when he returns from his year in Afghanistan next week.  It will be good to have him state-side.

Tomorrow I'll get back to work on the mill and start on scraping in the knee Y ways to their final fit and the knee's Z ways to bring up the crank end of the knee.  After that, I'll do the table and reassemble it so I can machine 3 new gibs.  Once those are fitted, I need to decide if I'm going to repaint the whole mill or just clean it and do some touch-up.  Then I get to remount the DRO scales and call this job finished.

Adios for now.