Grizzly G3103 Repairs - Part 2
April 03, 2010 through June 11, 2010

When we left off, the saddle had been surface ground to get rid of the somewhat less than marginal job of fitting that had been done at the factory.  It looked like it was performed with jack-hammers and angle grinders. Now I was able to use the saddle as one scraping template, along with a 24" granite angle template.  The granite angle template was cut from a 18" X 24" import surface plate on a wet saw modified to run a 12" blade and lengthened to allow a 24" slab of granite to fit on the rolling table.  The 3" X 3" X 24" piece of granite was then cut to a 45° angle along one 24" face and then flattened on the surface grinder using first a 150 grit, then an 800 grit diamond wheel.  This brought the surfaces within a couple ten-thousandths of an inch of being flat.  Final finishing was done on a dedicated - for lapping only - surface plate sprinkled with diamond dust.  The templates indicate to about a tenth variance between adjacent areas and a couple tenths over the entire length of the template with checked with my shop made planecator copy, but this is in a non-clean room, non-temperature controlled environment.  What the actual tolerance works out to is a mystery, but they seem to be as flat as my 36" Microflat granite straight edge.  I've made a few for friends and a few for myself and am pretty pleased with both the flatness and the finish I was able to get on these angle templates.   Having an angled surface template is the only way to reach the hidden area beneath the dovetail section of the mill's ways.  Having the angle at 45° means that the templates will work on most any dovetail way and having a pretty sharp point on the vertex allows the template to reach into smaller dovetail ways, like those of my South Bend 9" lathe.

Working with both the saddle and the angle template, I continued finishing the Y ways on the top of the knee.  At this stage, I had the angled knee Z ways finished so that the knee now runs parallel with the column Z ways and the knee flat ways were scraped to a point where the knee flat Y ways were almost exactly 90° from the column flat Z ways.  Well, at least the right side was about 90°.  The left knee Y way tilted down on the crank end of the knee by about 0.0025". The Y ways also needed a bit of adjustment to their height.  The left way was higher then the right near the column, but lower than the right near the crank end due to that 0.0025" drop in the left knee Y way.

My plan was to dial in the right knee Y way so that it was level with the left way near the column and sloped up toward the crank end by some amount greater than zero and less than or equal to 0.0005".  I also needed to making sure that the left way was as close to being an exact duplicate of the right as I can get it. This meant scraping the right way to the proper angle, then "scraping straight down" until the crank end was the same height as the left. This was due to the crank end of the left knee Y way being the lowest spot on that way.  Once the right way was correct, I could use the saddle with the box level on it as one means of helping me make the right way match the left.  Are you totally confused?  I must admit that it's hard to describe without pointing to the spots I'm talking about.

Scraping the top Y ways of the knee using the saddle as a template to keep both flat Y ways in the same plane. This is the first step.	The second step is using my 24" granite 45° angle template to show the high spots that need to be scraped in this cycle. Both templates are required.

With the knee attached to the column, I used the box level with feeler gauges to determine how much metal needed to be removed, then pulled the knee back off and set it up on my work table.  With the knee clamped down to keep it steady, I rolled Prussian blue on to the saddle and slid it along the Y ways.  The tricky part is that while this shows contact, it doesn't show what needs to be scraped.  The reason this is true is that the saddle is shorter than the run of the Y ways.  As the short saddle is slid along the long knee way, it can only leave a print in the area it touches.  If the long saddle way was high on the ends and low in the center, you would only get two marks - one at each end of the saddle.  If you pick the saddle up and place it back on the long knee way one inch to the right, you would have two different marks.  It doesn't take much imagination to see that you could have solid color for the entire knee way and still not know where the high and low spots were.  However, if we use a longer template to mark a shorter way, we see the true picture of the ways. (As long as our template is flat.)  I found out through trial and error that using a template shorter than the ways, even by an inch or two, makes it more difficult to scrape the way flat.  On my surface grinder, I used a 36" template on a 40" way.  It can be done, but it makes it a lot harder to get the way truly flat. It is tough enough to get a way to "tenths" of flatness without handicapping yourself with a too-short template.

So, after marking the pattern of contact with the saddle with blue, I then used Dapra water soluble red marking fluid on my 24" granite angle template and printed over the top of the blue prints.  I would scrape all of the red marks and try to get more and more blue coverage on each consecutive cycle of printing and scraping.  Since I was again trying to change the angle of the ways in two planes, I used a progressive method of scraping.  You start by dividing the length of the ways into equal one to two inch sections.  Next you print the entire way with color and scrape only the section on the end you want to be the lowest.  When you finish that section, you clean the way and print it again.  This time you scrape sections 1 and 2 of the low end.  This is repeated until all sections have been scraped.  If I feel I am close enough to the finished angle to take some measurements, I then print and scrape the whole way in one pass a couple times to get rid of the steps I've made by scraping individual sections.  Then I take my measurements.  If I know that I am not close to being ready to measure, I repeat the step scraping technique until I am ready to measure.  Using the Biax power scraper, the step scraping goes quickly and each complete cycle of steps may take off 7 times as much metal from the low end as from the end that will be higher, so the angle of the way changes fairly fast.  To increase the angle of the way more quickly, I go over each step a couple times, scraping in different directions.  Scraping in different directions not only removes more material due to repeated scraping of the same area, but also makes it easier to see what was scraped on the last pass versus the current pass.  Last, but not least, scraping in different directions helps prevent the forming of ridges/ripples in the grain of the metal that happen if a piece is scraped repeatedly in one direction.

Last night's progress was pretty typical.  In about two hours, I lowered the column side of one way by about 0.002" while I continued to scrape for bearing points on the other way.  Scraping for bearing (picking the tops off of the high spots) takes a lot more time than running the Biax over areas of color to change the angle of the way. And yes, even after scraping almost every day since January 24, I still get sore arms, a sore back, too.

I'm getting closer to having the Y ways of the knee level in all planes. A couple night's work brought me within a half thousandth on an inch side to side "height" of the Y ways.  The right way is about a quarter of a thousandth high on the crank end (perfect), but the whole way is still a half thousandth high.  So, I have the angle right, but need to scrape down a couple more passes to make the box level sit level when straddling the two Y ways.  I check this measurement by placing the box level on two 7/8" precision shafts; one on each way.

Using a short nap roller, I rolled out some water soluble marking fluid.  Rolling diagonally, I was almost able to mark the whole table	The corners of the table are off the surface plate, but there's enough contact to give me a good print of the majority of the table. .
This is the print made by my largest Starret surface plate.  The table is a little high in the center and has a slight warp. Scraping will even out the surface.	By alternating between the 24" surface plate and the 36" long template, I can scrape the 26" long table flat.  Next I'll scrape the table ways.

Another couple of night's work had the saddle sitting plumb on the knee. I was from 0 to 0.0005" from flat in every direction according to my box level. For the time being, I stopped scraping the top of the knee and started on the table.  I blued (redded?) up my largest surface plate on the diagonal and printed the top side of the table.  There was a little twist in it, but not terrible.  I considered surface grinding the table. but at 26" long, it is as long as the travel on my grinder. I could have disconnected the hydraulics and manually fed the table to get enough room for a 1/4" run-off in either direction, but the more I thought about it, the more I thought that a scraped surface on the top of the table would look cool.  I ended up scraping. It took me two 4 hour sessions to get the table to 20 or so points per square inch and I stopped there.  Going for surface plate quality scraping was over-kill.  The table is always accessible, so if I feel the need to produce a finer surface at some time in the future, all I have to do was scrape. No disassembly required.

Using my B&S half-tenth indicator, I measured the height of the ways from the surface plate.  The ways were out by about 1 and 1/2 thousandths on opposite ends.  The front way was high on the right and the rear was high on the left. I found the lowest point of the two ways and set to scrape both ways to that height.  Scraping dovetails with a Biax is an art.  One that I haven't completely mastered yet.  You need to have pretty near perfect control of the tip or you run into the opposing angled way.  So far, I've found the easiest way to scrape the last bit before the vertex is to sneak up on it.  I start scraping with light pressure well before the crucial section and slowly advance at an angle until part of the carbide tip hangs over the channel that's cut into where the two ways join.  With the blade half on and half off the way, I travel down the edge applying more pressure when I encounter marking fluid.

After a few cycles on both of the tables X ways, it was back to the surface plate to check the height.  This went on for about a week of evenings until I was within tenths of having both ways parallel to the table top. About once an evening, I would blue the X ways of the saddle, then lift the table on to the saddle and install the gib.  I'd test for proper tightness of the gib, then place the box level on the table and push the table from end to end while noting the bubble in the level.  Each night it got a little better and the bubble stayed centered over more of the table's travel.  Each time I removed the table and checked the blue marking, it was covering more and more of the ways.  This was satisfying, as I could see my progress improving and the better the markings got, the more I wanted to scrape to get them even better.  I also began making marks on the table indicating where the gib was tightening up and where it was loose as I slid the table from side to side.  This would help when I started scraping the angled ways of the table.

I again checked and made a couple adjustments to make sure the mill was level.  I did a few more test fittings and a little more scraping.  I was about finished with the major scraping.  I had never really liked the cream colored paint on the mill.  It shows every dirty fingerprint.  I had also chipped the paint in a number of places while taking it apart for the many fixes I'd performed over its short life.  I wiped down the base and column with acetone, then sanded them with 500 grit wet and dry paper. I brushed on a few coats of Rustoleum gray. I decided that I liked the color on the mill, so I got to work cleaning masking and painting each of the parts.  I figured since I had it in pieces anyway, now was the time to give the mill a cosmetic update.

April 21, 2010
I've gotten the painting under way.  While various pieces are drying, it's time to do the final way scraping. This involves cleaning up the 55° angled ways so that column, saddle and table intersect at as close to perfect 90° angles as I can get.  On this particular mill, this is easier said than done.  The present goal is lining up the table slots with the table's rear angled way.  In a perfect world, the outside perimeter that makes up the table would be parallel/square with the T-slots and the rear angled way.  The front angled way is not meant to be parallel with the rear as the angled way and its gib are wedge shaped in order to adjust for wear.  In the real world, on this mill, none of the surfaces; rear of table, angled way and T-slot are parallel.  Close, but not close enough.  So the task becomes aligning the T-slot with the rear angled way.  Forget the rear surface of the table altogether as it really has no bearing on clamping work to the table and milling it square.  As long as the T-slot and rear way match and end up at a 90° angle to the Y axis and parallel with the surfaces of the Z flat ways, we will be set.

To check the table for parallelism, I placed the table on the surface plate with the front surface down and the rear way on top.  I shimmed the table until I could measure a constant distance between the surface plate and the T-slots.  I will say that all three T-slots were all very close to parallel with each other. I then clamped a 1/2" piece of precision round stock against the rear angled way and measured between the round stock and the surface plate.  The left end was high by a little under a thousandth.  Not bad.  I scraped the angled way until the measurement was the same across the entire way.  To ensure that I kept the way angle at 55°, I used my 13" cast iron 55° angled template.  To make sure that the way remained in one plane, I alternated with my 24" granite angle template.

With the T-slots aligned to the table, the next step was making sure that  the T-slots were at a 90° angle to the saddle, the saddle is parallel to the knee and the knee is 90° to the column flat ways.  The knee to column fit should be darn close, as I had already scraped the flat Z ways of the knee to align at 90° in both the horizontal and vertical planes.  The adjustments I'd be making now to the angled ways on each of the axes not only ensure that the flat ways contact tightly, but are also a constant distance apart from each other. The angled ways need to be scraped so that when the gibs are installed and the cranks are turned there is no binding or looseness. However, since my current gibs were an amalgamation of cast iron, sheet aluminum and steel feeler gauges, I figured I ought to make some proper gibs first.

Reassembly begins	I like the Rustoleum gray on the mill.
The temporary gib is installed.	The "T" connection is bolted up.
The milling head gets attached.	Two of my buddies watch my progress.

I reassembled enough of the mill that I could actually make some test cuts. I was most interested in whether the X and Y axes were square - I.E., the table square with the saddle. I chucked up a 2" by 8" piece of cast iron and machined the 8" length along the X axis. I machined both ends at a 90° angle. I then removed the work and placed it on my surface plate. I used my 8" square to check how close I was able to machine the angles. Using the shimmed gibs on both the saddle and table, the block came out pretty square. Certainly a lot better than it had been. I measured about a half a thousandth between my square and the top of the cast iron. So that's a half thousandth with a 2" by 8" piece with shimmed gibs. I wanted to take another couple of cuts on a larger piece of steel to get a better opinion of the squareness, but in setting up the steel to be cut, I noticed a bit of looseness in the table. I set a DTI in a collet and tried pulling and pushing on one end of the table. I had a couple thousandths of play.

I decided i should go ahead and make a proper gib to tighten up the table as well as one for the saddle. The shims I had added to the gibs amounted to a shim of one thickness at one end and a different size of feeler gage at the other end. This meant that there was not much contact between the two ends and this would have some effect on the final fit of the saddle and table.

I used my shop set of gage blocks to take the measurement of the space where the gib would fit. I did this by shimming the gib a little too thick to check the small end. This wedged the ways against the gib without allowing the gib to travel all of the way to the end of the space where it normally sat. I then used the gage block to measure the distance between the two ways. I added 1 thousandth to my measurement to make sure that the gib would be a hair too fat. I would scrape off the excess.

I then shimmed the gib a thousandth or so too thin and reinstalled the gib. This would make the gib run too deep in its home and free up the large end for measuring. Again I used a combination of gage blocks to find the thickness that the gib should be. Again I added a thousandth for scraping.

The slope of the gib has been machined and the gib surface, both front and back, is pretty flat.	Both sides get scraped.  The back side doesn't need or get as many points per inch.
The sine table is is set to allow a 55° cut.	The print is beginning to fill in nicely.

The saddle gib is 6.125" long and the thick end measured .375". the small end was 0.090" smaller at 0.285". I divided 0.090 by 6.125 to get the taper per inch and multiplied it by 5 to get the measurement to set the slope of my sine plate. I used some 1.125" X 0.50" by 7" cast iron stock and cut the slope angle. I then cut the two angles to turn the end view of the gib from a rectangle to a parallelogram. The Grizzly mill uses a 55° angle on the angled ways and the gib obviously should match this.

Once I had an overly long copy of the gib, I used some black permanent marker to color the side of the gib that touches the sliding way. I oiled up the ways and installed the gib. I moved the axis with the new gib enough to wear away some of the black marker. I removed the gib to check it. Since it appeared that the gib was contacting a bit more on the small end than the large end, I did some scraping on the small end and tried again. After a couple attempts, I had pretty even contact on the gib, though it was still a little thick and did not seat all of the way into its home. Some more scraping fixed that. I now had one gib that fit well, but was a little long on the large end. Since that extra length was my adjustment material, I decided to cut the gib a little on the long side - just in case. The new gib ended up at 6.625" instead of 6.125". Having the extra material on the large end caused no problems and would possibly save me from having to machine another gib if I needed to scrape a bit off the ways to adjust for squareness before I was finished "tuning up" the mill's ways.

All three gibs were made the same way, but the third one was charmed and came out perfect.	A lot of overhang and the parallel lifted a bit, but the 55° cuts matched the ways perfectly.
It was nice to use the mill again after many months of being without it.	The finished Z gib after being fitted.   The marking fluid was rubbed off evenly showing that it fit well.

I moved on to the long table gib. It's a little over 11 inches long and I had a piece of cast iron already cut to 12" inches and scraped flat on one side. I had cut this to size and scraped one side flat a couple of years before as I was going through the long process of acquiring the necessary tools and experience to strip and scrape the mill.

I made my measurements in between interruptions of folks coming to our house to adopt some kittens. This turned out to be a mistake. Over the next couple nights, I set up and machined the slope and two 55° angles of the gib. I finished it and gave it a try. It only went about half way into the slot before it stopped. After some head scratching and remeasuring, I discovered I had machined the thick end wrong. It was exactly 0.050" too thick. Lovely.

Five hundredths is too thick to scrape. Especially when only one end is that much too thick and the other end is dead on spec. I'm sure it could be done, but it is something that would take a lot longer than just making a new gib. Since I had already machined the 55° angles into the gib, it would be very tough to hold in my large milling vise. I decided that the repair technique that made the most sense would be to use the magnetic chuck on the surface grinder. I would put the old gib, shims and all on the table, then the new gib on top of it in the opposite orientation. I'd then build a wall around the two stacked gibs so that they wouldn't wander off. The magnet in the 24" chuck is pretty strong and should hold everything pretty firmly. The built-up wall would prevent the shim from launching itself through the shop or me. When the gib was within a couple thousandths of where it should be, I'd finish it by scraping.

As it ended up, I was too concerned about grinding the gib too small and left it a few thousandths too thick.  It took a few nights to scrape it to the desired thickness.  The most important consideration I had was that I get the thickness right on the small end.  The small end of the gib fits into the right side of the table and this was where I had a couple thousandths play.  When I finally had even wear marks in the marking medium I used on the gib, I cut the gib about a half inch too long on the thick end.  Again, this would allow me some adjustment if I needed to do any scraping on the angled ways to align the table.

Even though I still had some "tune-up" work to perform on the angled ways to get the mill as square as I could get it, I was impressed with the feel of the three axes.  The X axis of the table was especially nice.  Prior to scraping, the only way I could get a smooth feel when cranking the table from end to end was by loosening up the gibs to the point of introducing too much play.  Now with a DTI in the spindle resting against a block on the table, I could give a strong push and pull near the table ends and see a total deflection of a little less than 0.001".  Before scraping the mill, I couldn't even get close to a thousandth deflection with both table locks cranked tight.  I was now getting anxious to finish the mill and work on some projects I have been wanting to do.

I was now coming to the final stretch of this project.  All of the ways were scraped close to where they need to be.  I could finally start checking and setting the last adjustments.  The first thing to check was making sure that the end mill was being held square with the table.  The usual method for this is called tramming the mill.  This is nothing more than using a dial test indicator to check for squareness between the spindle and the table.  Since the G3103 mill has a table width of only 6 inches, I use a extension that is 3 inches in length.  My extension is a 1" diameter piece of drill rod with a 1/4" by 1" piece of flat stock attached at the bottom of the round rod to form an L shape.  At the end of the flat bar is a provision to hold my half-tenths-reading indicator.

The procedure is to eyeball the milling head to align 90° to the table, then install the indicator and holder.  Using a known flat half inch wide parallel between the indicator tip and the table, I bring the table up to find zero on the indicator.  Starting with the indicator pointing at me, I slowly hand-turn the mill's spindle in a full revolution. I move the parallel so that it stays centered under the indicator tip as I move the indicator in a full revolution around the spindle.  I note the relative height of the indicator a each quarter revolution.  The readings at the 3 and 9 o'clock positions should be the same if the table is flat and square in the longitudinal axis.  If the 9 o'clock reading is 0.003" high and the 3 o'clock is 0.003" low, then the milling head will need to be revolved counter-clockwise to make both measurements zero.

To tram the mill, I started by bringing the table up until the DTI read zero and locked the knee. I then took measurements every 1/4 turn of the spindle.	Another shot of using the parallel as a spacer between the table and DTI.  Using the parallel also helps to bridge the T slots in the table.

I set the mill head to be level along the length of the table - 3 and 9 o'clock positions.  Unfortunately, the G3103 doesn't have an adjustment to change the tilt in the 12 and 6 o'clock positions, like the Bridgeport "nod" feature.  If this plane is out, all that can be done is to scrape one of the junctions between the milling head and the column.  If all else is equal, my choice would be the junction at the top of the column that allows the head to rotate around the column. However, I had started this project by scraping this junction square with the column ways and then tried to line up the spindle axis with the flange on the milling head, but fully expected that some adjustment would be needed.  It was. I had about one and one half thousandths difference between the front of the table and the rear as I revolved the indicator.  As the rear (column side) of the table was indicating high, I scraped a small amount from the bottom of the vertical front flange junction to lower the front of the milling head a bit.  By lowering the front, I was also raising the rear where the indicator was reading that the table was high.  This was a real pain to do as it required me to disassemble the milling head and pulley covers to gain access to the flange.  I got lucky and only had to test fit a couple of times before I had less than a half thousandth difference between the front and rear of the table.  Seeing as the measurement I was taking was three inches from the spindle and my indicator showed less than 0.0005" between the front measurement and the rear, I think that most of my end mills and face mills, which only go to three inches in size, will cut true.

With the milling head trammed, I ran the table to the ends of its travel and checked the measurements.  Since this is a lightly built mill, the weight of the table pulling down on one side of the saddle causes a little drop when the table is near the end of travel.  This is amplified because I have extended the travel by about an inch on each side.  The drop at the end of travel amounted to a little over 0.001".  The drop isn't something that I will encounter in most of my projects since I generally work in the center of the table, but when taking a cut of a foot or more, I will need to remember that it will cut a bit deep at the table ends as the other end of the table will be drooping a bit.

I keep thinking about how long I have been working on this mill.  In the early stages, I tried to follow the plan I had set, but it was hard to judge what kind of progress was being made.  Now as I get closer to finishing the project, I'm able to make some observations of what has changed.  Cranking the handles on all of the axes is much easier now.  The ways slide very smoothly.  The knee and saddle locks now clamp down on the gibs and lock up more positively.  The settings don't wander as much when locking the axis as they had.  Best of all, I can place my box level on the table and crank the mill on all axes and have the bubble stay fairly well centered.  Not perfect by any means, but before rescraping I couldn't even get usable readings on the 0.0002" per increment box level.  I had to use the much less sensitive level and the bubble moved all over depending on which way I cranked.  All in all, a great improvement in the accuracy and feel of the mill.

Now that everything is back together, I took a test cut on a small piece of cast iron.  I side milled the length along the X axis, then carefully re-clamped the right end without moving the setup and made a cut along the 4" side at a 90° angle using the Y axis.  I removed the cast iron and set it on the surface plate along with my L square and master square.  The 90° angle between the short side and long side are darn near perfect - or at least as close as I can measure to perfect.  I repeated the test with a 12" X 4" X 1.25" thick piece of cast iron.  Again, the results were as I had hoped.  The end cut was square with the side cut.

The last piece to do was to make a new Z gib.  I have been using a gib that is made up of a number of shims and the old gib and it's time to replace it with a proper gib.  The 12" X 4" piece of cast iron will be sliced and a new gib made.

I cut a 0.450" slice off the bar of cast iron, then machine the cut side flat.  I scraped the 12" X 1.25" surface enough so that there were no obvious high spots.  Since this axis had been modified a bit by cutting the knee right side angled way a bit deeper at the bottom, I remeasure the shimmed gib a couple times and measured the space it fit extra carefully.  I had 0.077" taper over 8.125" inches of gib.  I divided by 8 and multiplied by 5 to get the settings for my 5" sine bar.  I then used my 6" machinist's vice to hold the sine bar and 12" gib.  The gib was left unsupported, which I suppose is a bad thing, but I planned on only taking 0.005" cuts.  I figured that deflection would be minimal.  Once I got the slope of the gib machined, I set up my 6" sine table to machine the 55° bevels on the top and bottom of the gib.  These end up being a compound angle cut to keep the face of the gib a constant size.  I did the math for the depth of cut then checked the face for size as I machined the 55° angle.  Once the gib was cut, I did a real quick bluing and some preliminary scraping.  I really wanted to try it and see how close I had come.  I was concerned that I might have made an error in my calculations.  No error.  This gib fit almost perfectly without any more scraping.  Of course I scraped it anyway as the surface I got from milling it wasn't the best, but I sure hit the milling measurements right.

With a properly adjusted gib in the Z axis, I once again checked the level of the table.  I was a small amount high on the right side. Less than a half thousandth.  I figured that this might happen when I finally got a proper gib fit.  Unfortunately, this meant that the table and saddle had to come off again so that I could scrape the top of the knee back flat and level.  I've had the table and saddle on and off so many times that I could do it blindfolded.

I lowered the right way of the knee and adjusted the left way so that the saddle made good contact on both ways.  I checked my progress with the box level and stopped when the bubble moved less than 1 increment in any direction (0.0002" per 10").  I am still pretty amazed that I can crank the saddle forward and back and have the bubble not move more than 1 division.  Yes, it's supposed to be like that, but until now it's never been even close to level and this is a real treat.

One of the last pieces to install was the guide for the power feed stops.  The DRO scale is hidden behind it and covered with some aluminum sheet.	The first project on the rebuilt mill was taking some skim cuts off the bottom of an 18" camel-back straightedge casting.