This page is currently in blog format as it is just my research notes for the project. Spindles are hard to build properly, whichis why they are so expensive.
Usethis information at your own risk!
Running a Three Phase VFD on Single Phase Power by Derating
You can use a 3 phase input VFD, and power it off of single phase. You just need to derate the VFD by a factor of 1.73 to account for only a single phase of power.
Or to put it in more useful terms, you need to buy a more powerful VFD to accommodate your 5 HP motor, using the formula of (5 HP)* 1.73 = 8.7 HP.
This doesn’t mean you are drawing 8.7HP worth of current. The current max current draw off of 230 V single phase for this inverter is approx I=(5 HP)*(746W/HP)/(230V * 0.85 * 0.85)= 22A (last two 0.85 factors account for the inefficiency of the motor and the VFD). Powering 5 HP off a single phase 230V circuit is doable, and since the VFD is acting as a soft-starter, and limiting the starting surge, it is much less problematic than powering a 5HP cap start-run straight off the same circuit.
http://www.dartcontrols.com/wp-conte…hase-Power.pdf
This is a common practice, and the VFD manufacturers usually provide instructions about doing this.
—————-
My newer inverter welding machine is connedted like that too, you just use the two 230 single phase leads and leave the third lead unconnected that would be used for three phase, it is derated some though.
—————-
ATCSpindle Cut Away
GD.Marsh’s RF-45 Belt Drive
Some pix and comments from his CNCZone thread:
Exploded view of the mill head. This conversion involves the part with the red arrow…
The sleeve contains the splines that engage the spindle. A gear had to be turned off the original…
Components…
Bearings will be used to support the sleeve…
Here the stack is assembled with the bearings. Motor pulley to the right…
And here it is fitted to the top of the spindle case…
Bottom view…
Ready to receive the motor…
Ready to run, save bolting rather than clamping the motor mounts. The gizmo on top is so the power drawbar can squeeze instead of push and avoid damaging spindle bearings...
Here are the Nachi angular contact bearings he ordered from VXB:
Kit10017 7207 Nachi A/C 35x72x17 – $20.46
Kit10016 7206 Nachi A/C 30x62x16 – $16.11
He later upgraded to these:
7207CYP4 Abec-7
7206CYP4 Abec-7
Great Source for V-Belts
Bearing Greasing Factoid
There’s a formula to get the cc’s of lube, it is the last to digits of the bearing number x 5 x .06. This will give you the amount of cc’s per bearing.
ex for bearing 7008:
08 x 5 x .06 = 2.4 cc’s of lube.
Teco Sensorless Vector VFD
Saw the Teco 7300CV written up in a CNCZone thread. Several of the long-time posters like the drive and it is a sensorless vector which is what I’d prefer (broad torque band to reduce the need for belt changing).
Kelling 3HP Brushless DC Motor Setup
This seemed like a nice setup from Kelling if 3HP is enough. Unfortunately, the later threads I’ve read indicate the controller is not reliable and there are some questions about whether it really produces 3 HP. I’ll leave this in here so people can find it, but I’ve gone back to thinking about a more traditional VFD setup.
It’s $229 for 3HP and 6000 rpm as I write this. Here is a video from cjdavis618 who also wrote a CNCZone thread on the controller:
The rpm range is 400 – 6000 rpm. The controller is PWM-ready, so is easily interfaced to Mach3. There is also a charge pump/E-Stop input. The motor is a brushless DC, so it should last a good long time. The power supply is integral so you can plug it into either 110 or 220, and there is even a nifty LCD tachometer that shows the RPM.
Judging from the video, it isn’t real fast on acceleration, but it isn’t terrible either–about 6 seconds to ramp the entire range from S400 to S6000. You’ll definitely want to think about using a G4 Dwell in your g-code though.
3HP is not too bad. I had to cannibalize my 3HP VFD I had bought for the mill so I could run my big bandsaw, so I will need to be in the market for something else. This would definitely be another setup to consider instead of a VFD + AC motor. It’d be cool to have one with both an “Aluminum” and a “Steel” speed range. The Steel would gear down the 400-6000 rpm range to maybe 100-1500 rpm, while the Aluminum would go the other way for a range of maybe 550-8000 rpm. That Steel range might be too low, so a 1:2 ratio would give 200-3000 rpm. I need to play with it on G-Wizard to see what ranges I’d favor.
There are some issues if you read on in the CNCZone thread. Also, it appears that $229 will not buy you both the motor and controller. Maybe not as godo a setup as it seemed. Back to sensorless vector drives for me!
Another Preload Measuring Setup
Given a dial indicator to measure deflection, and a known preload that is to be established, here is a fixture to measure the deflection and hence the spacer required to create a given preload:
A Hydraulic Scale for Measuring Bearing Preload
Here is a hydraulic scale for measuring preload on bearing pairs similar to what I describe near the bottom of this page:
The idea is to use the scale with a test cell to measure the displacement of the bearing inner race when the appropriate preload is attained. Once that has been determined, spacers can be machined that deliver the same displacement and hence the same preload. There is some good discussion about this on the PM boards.
The Value of Spindle Wall Thickness
The stiffness of the spindle goes up as the fourth power of the wall thickness according to some information I saw in one of the bearing catalogs. That means you want the largest possible bearings you can afford because a small increase in the thickness makes for a much stiffer spindle.
How thick is thick enough?
A Mazak QT-20 lathe runs 120mm ID bearings on the spindle and has a spindle bore of 2.44″. That works out to a wall thickness of about 1.142″ or about 29mm.
Now let’s say we want to build a lathe spindle with a spindle bore of 2″. Let’s give it a 1″ wall thickness, so we need bearings with 4″ ID which is 101mm. Since bearings come in 5mm ID increments, we can look at a 100mm or 105mm ID, which would be a 7220 or a 7221 angular contact bearing.
Brand new Nachi 7220 CYDUP4 bearings are $1351 each and you need 2 of them. Ouch!
I can see some others on eBay such as an SKF 7221 BECCM for $249.99. Better, but still painful for an experimental spindle project.
The Holy Grail: Preload via Bearing Spacers
I finally tracked down some real information on how to set angular contact bearing preload via spacers on the SKF web site.
Consider two bearings with spacers connecting their respective inner and outer races. By making these spacers slightly different lengths we displace the bearings in such a way as to preload them. The configuration looks like this:
As we grind away at the spacer faces, we achieve the following:
| Necessary spacer ring width reduction | |||
| Preload change | Width reduction | ||
| Value | Spacer ring between bearings arranged | ||
| back-to-back | face-to-face | ||
 | |||
| Increasing the preload | |||
| A to B | a | inside | outside |
| B to C | b | inside | outside |
| A to C | a + b | inside | outside |
| Decreasing the preload | |||
| B to A | a | outside | inside |
| C to B | b | outside | inside |
| C to A | a + b | outside | inside |
So, for example, suppose we have a pair of AC bearings arranged in a back-to-back configuration. If we want to increase the preload we need to reduce the width of the inner race spacer. Note that A, B, and C refer to bearing preload classes. So if we wanted to increase the preload from class A to class B, we would reduce the width of the inner spacer by amount a. Here are the a and b amounts:
Okay, let’s work an example. Let’s say I have a pair of 7207 CD bearings that I want to preload. The minimum spacer size difference I might use is a, which is 9 micrometers. That’s 0.009 mm or 0.00035″. The largest displacement I might use would be a+b = 21 micrometers = 0.00083″. Pretty small spacer size difference!
Bear in mind that these differences are used to step of from preload class A to B or C. Those preloads for the 7207 CD series are 60, 120, and 240N for class A, B, and C respectively. Let’s compare:
– Going from A to B increases preload 60N, and requires a 0.00035″ spacer.
– Going from B to C increases preload 120N, and requires a 0.00047″ spacer.
– Going from A to C increases preload 180N, and requires a 0.00083″ spacer.
We can extrapolate that to go from no-preload (i.e. we bought raw bearings that are not ground for preload) to the class C preload (i.e. 240N preload) might take something on the order of 0.00118″ or a tiny bit more.
In terms of the spacer rings themselves, SKF had this to say:
The spacer rings should be made of highgrade steel that can be hardened to between 45 and 60HRC, depending on the application. Particular importance must be given to the plane parallelism of the face surfaces, where the permissible shape deviation must not exceed 1 to 2µm. The overall width of the inner and outer spacer rings should be identical.
The most accurate way to do this is to process the width of the concentric inner and outer spacer rings in one operation.
Making the rings to the requisite accuracy is definitely a grinding operation since we’re measuring progress in tenths. These rings are not going to be very wide either. Let’s say not more than the width of the bearing itself.
We will mount the pair of bearings on the spindle nose against a shoulder on the spindle. The inner race spacer will go between the bearings as will the outer race spacer. A precision nut will thread onto the spindle nose and down onto the inner race of the bearing stack to hold it in place. The nut is precision simply because it has been faced so that the surface against the bearings is flat.
Getting the Good Grease
If you scan way down this page you’ll see my write up about oil lubrication of spindle bearings and the chart that shows bearing rpms versus the dN. Really good bearing grease can achieve a much higher dN than the wheel bearing grease you were tempted to use from the auto part store around the corner. The best spindle greases come from a company called Kluber. The trick is finding the stuff in small quantities without breaking the bank. There are a variety of grades available.
One good one is Kluber Isoflex NBU 15 available from:
– Woodtechtooling: $34.50 for 50g tube
– Kluber NA: $22.45 (You have to register to access the online store)
Update: Ordered a tube of my own from Kluber NA on April 27.
Delta Inverters
The inverter is the thing I’m researching quite a bit at the moment. I want one I can use to effectively make the spindle a big servo drive. At the very least, it needs to do spindle positioning for tool change, and at the most, it should accomodate rigid tapping. The Hitachi mentioned below would work, but it’s expensive. Looking around at what other machine tool builders have used, a couple using Delta drives. The drive that Dave DeCaussin (the “D” in “Fadal”, so he knows a thing or two about CNC) uses in his video below is almost certainly a Delta VFD-M:
Interesting, if true, because that’s a pretty cheap sensorless vector drive. Looking through the manual, it does have provision for a PID feedback loop and an index sensor.
Atrump is using Delta drives too, although they use the more expensive closed loop VFD-VE series. A nice 3 HP is going to be $553.80 from one supplier I found. Looks pretty easy to set one up to index for a toolchanger. According to the Atrump site, you just enter the encoder count in the VFD when things are lined up right.
Hitachi SJ300 + SJ-FD Board = Spindle Servo Drive?
Suppose you want to build a toolchanger that requires spindle orientation. While we’re add it, let’s do rigid tapping. And let’s require an extremely broad torque range for the spindle to minimize the need for any gear changes. That’s a hard task, but I’m starting to think the Hitachi SJ300 with the optional SJ-FD encoder board might do the trick. The SJ-FD, in particular, looks like it provides something called “pulse train mode”, which looks suspiciously like step/dir control.
The whole thing is not cheap, and requires 3 phase power input to boot.
Spindle Orientationfor Drive Tangs
Suppose you wantto run a 30 taper. It has drive tangs, so you need to be able to line up your spindle properly with the tangs to build an automatic tool changer.Most of the commercial VMC’s are using a servo to drive the spindle (even
if it is a fairly ordinary motor with a fancy drive). That’s one approach,but not cheap. Try looking for a 2 or 3 HP servo drive and motor!
Another thoughtis to use a smaller stepper that can be disengaged to index the spindlefor tool changes. You want to disengage it because if it’s driven allthe time it wears out the stepper and it generates back EMF that willfry the servo drive. At the least you need to electronically disengageit.
But there is yetanotherway that I just saw on CNCZone. NEATMan suggests using a cam action.Imagine a cam that looks like this and is clamped to the top of the spindle:
Now let’s saywe have an impact wrench style powered drawbar or an air cylinder thatcompresses a Belleville spring stack. Either way, there is a downwardmotion. We use that downward motion to push a roller equipped assemblyagainst the cam. The roller will then turn the spindle until the rolleris at its lower point, which should line up the indexing for the drivetangs.
Clever!
Vector DriveMagic
I’ve been keepingan eye on vector drives for a little while now. Essentially, they area “better VFD” to drive an AC motor with variable speed. Whybetter? Because they enable a wider speed range without loss of torque.I read a white
paper by Reliance Electric that indicated you could expect a 2:1 rangefor a regular VFD, and a 4:1 range with a vector drive. In other words,run your motor at up to half speed without losing torque with a normalVFD, and 1/4 speed with a vector drive. Why does this matter? Becausethe speed range you need to cover the gamut of machining tough steelsall the way up to aluminum and soft materials is huge. To span a rangeof 100 rpm all the way to 8000 rpm (still nowhere close to what a lotof CNC’s run today) takes an 80:1 range!
That’s why youneed gear changes, back gears, step pulleys, variable pitch pulleys, ora host of other mechanical transmissions used on different machine tools.But those transmissions are a hassle to deal with too, especially if youwant to build your own machine, or radicallyalter the performance envelope of your spindle.
So it was withsome interest that I came across a SumitomoVector Drive spec. I guess a good vector drive these days has morerange than Reliance gives them credit for. A sensorless drive is one youcould just hook up like any VFD, and Sumitomo claims a range of 120:1for their HF-430 unit in sensorless mode and that with a speed accuracyof 0.5%. But it gets even better. Add the encoder board and put a suitableencoder on the motor and now they’re claiming a range of 1000:1 and speedaccuracy of 0.05%! That’s starting to be quite a range, and these driveshave gotten a lot cheaper over time. I came across this particular onein an eBay buy it now for $249 for a 3 HP unit.
The only thingmissing from it seems to be true servo operation, or at least a meansof parking the spindle at a known location. True servo operation wouldallow the spindle to be indexed to any arbitrary position. You can imaginethat would be handy if you wanted to set up your lathe spindle with a3 HP motor (or more) and treat it as a C-axis. Index to a location, putan air powered spindle on your CNC gang slide, and you can suddenly drilla bolt circle for a flange under CNC power. A parking position would makeimplementing a tool changer for a mill a lot more feasible. Park the spindlein a known location and the “dogs” on the toolholder are properlylined up.
Principles of a Spindle Oil Mister
Commerical VMC’s running athigher speeds (10K and up) use an air-oil mist to keep the bearings lubricatedand cool. It works along these lines:
Snowmobiles Use Variable Pitch Sheaves
Could be a source of cheapparts. Check thisGates article.
Fadal’s Spindle Transmission Using Ribbed Belts andIdler Pulleys
Downloaded a setof manuals for a Fadal 4020. It’s amazing what you can learn fromthese. There are exploded diagrams and detailed maintenance procedures.
The two things that interestedme most were the details of how the toolchanger works, and the detailsof how the 2-speed transmission for the spindle works. The latter is particularlyintriguing for this belt driven spindle project.I can’t quite fathom all the details from the drawings and pictures, soI posteda question on CNCZone on a likely Fadal thread I found. If what Isuspect is true, it would be a slick way of building a 2-speed transmissionto extend the speed range beyond what a VFD will comfortably deliver.
Hopefully the gang on the ‘Zone will fill in my blanks.
Here is a picture of the idler assembly, $800 to buy one!
Fadal Idler Pulleys…
The system has 2:1 pulleys on the motor andthe spindle. There is a ribbed belt on each, a 315K10 if that matters.
To engage one of the speeds, the idler pulley is triggered to apply tensionto the belt being engaged by an air cylinder, visible in the picture above. A belt guide keeps the belt that isn’t engaged from getting tangled up:
Fadal Spindle Belt Guide…
Node the ridges, and the way the top belt has a shelf to support it. The ridges allow a little bit of belt slippage without too much surface area for the belt that is disengaged. Not too hard to imagine the bottom beltjust sitting there quietly. If tension is removed, it will flex out andnot be driven. In addition, it is held slightly out of the way by gravity.
Given the ratios available, the “hi”gear will drive the spindle at 2x motor speed and the “lo” gearwill drive it at 1/2 motor speed. Couple that with a 3600 rpm motor andyou should have a range that extends from a hundred rpm all the way to7,200 rpm. If you can overspeed the motor safely, you can get to the magic10,000 rpm number. The Fadal-style transmission is also relatively compact.It seems an ideal addition to the project if I can figure it out. I mighteven build a manually operated prototype and look into air cylinders later.
More views of the Fadal Hi/Lo Drive system:
Good view of one engaged and one disengaged drive ratio…
Another Belt Drive for the IH Mill is in the Offing!
What ho! The game is afoot!
Cruiser is documenting his belt drive conversion experienceover onthe CNCZone as we speak. In the same thread he has also gotten jiggywith the idea of attaching some linear slides to the sides of the IH columnto help facilitate the Z-axis and overcome some of the stick slip fromthe poor marriage of a heavy head and very average dovetail ways. That’salso quite an interesting mod:
Cruiser uses a router tied to the
existing dovetail slide to cut a bed for his linear slides. Cool!
Here is one “teaser” photo:
Go check out the CNCZone thread for fulldetails!
5/12/07: I thought I’d adda little progress update. Cruiser has completed his belt conversion. Onething Cruiser mentioned is that his belt drive is not working very wellat slow speeds–torque falls off rapidly. I believe he’s running a littleunder a 2:1 ratio on the motor to get it to not quite 6000 rpm. The motorseems to be good to half speed and then torque falls off rapidly. Thismay be a problem for cutting steel, although he has tried some passesat 3000 rpm with limited success (currently fighting a noise problem of
some sort).
My conclusion on this is that a back gearof some kind is needed for low speed operation. There’s just too muchloss of torque at low speeds if you try to run belts directly with noability to change gear ratios. One could use something as simple as conepulleys to provide different rations, or something more complex like thevariable pitch pulleys I have discussed below. Another idea is to usea much bigger motor, so that the torque drop off is less noticeable. Forexample, these mills start out at 2 HP. A 5HP motor should notionallybe able to run at 1/4 rpm and have the torque of a 2 HP.
To get an idea, I did some rough back ofenvelope calculations. Let’s say your motor is great down to 1/2 rpm,so a 3400 rpm motor is good at 1700 rpm. Let’s further say you want topreserve torque to run at 100 rpm. I’ve worked out some possibilitiesin this spreadsheet:
The spreadsheet estimates horsepower availableon the assumption that:
– The motor will deliver full horsepowerdown to 1/2 speed.
– Below 1/2 speed, HP falls off in directproportion to how far the rpm is from rated speed.
– You can’t run the motor over its ratedspeed.
I’ve experimented with drive ratios, whichare listed along the top 2 rows, as well as motor horsepower, listed rightbelow. The numbers in the sheet are “estimated horsepower”.They’re likely bogus as exact horsepower figures, but helpful as a guideline.If I had less than 0.2 horsepower, I marked that off in pink as beingunusable. If I had less than 0.5 horsepower, it is marked yellow–probablyusable if you go slow with a cutter that won’t take much horsepower. Togive an idea of some high horsepower operations, I played with a Feeds and Speeds Calculatorand came up with big twist drills (always use lots of power) and a4″ indexable face mill. When cutting tool steel, they need 1.2HP.This just to give some “worst case” checks.
My conclusion from all of this:
– Bumping up to 5 horsepower doesn’t buyyou a lot more speed range. It’s pretty similar (not surprisingly) toa 2:1 back gear.
– A 2:1 back gear is essential for lowerhorsepower motors just to survive with steel. 5 HP with a 2:1 back geargets you all the way down to 100 rpm.
– 5:1 back gear is too wide a spread witha 2 HP motor. You’d have a dead zone from 800 – 1500 rpm.
– 3:1 back gear 2HP looks about right tome.
I like the use of a variable pitch pulleyfor system for back gear. Hey, if it was good enough for Bridgeport, itsgood enough for me!
You can use the variable pitch as your hi/loshift mechanism. That means the vari pulley needs enough total ratio tocover the back gear + high speed. So, using the numbers above, for a 6000rpm spindle that’s 2 * 1.8 = at least a 3.5:1 vari pulley setup. For an
8K spindle (which I’m aiming for), you’d need 4.8:1. I’d run the systemwith preset stops. Probably 3 stops on the range is good, and use theVFD to fine control speeds within a stop.
A 2-stage system could use a 2:1 back gearand a 3:1 varispeed to get up to a 6:1 advantage, for real power downlow. Some kind of tumbler system like the lathe guys use could work forthe back gear shift.
An alternate gear train option would be aplanetary gear set for low speed operation.
1964 Patent on Grinding Duplex Bearings
Available on Google, search for patent #3330634.
Another Thought on Making Up Duplex Bearing Sets
It seems most of the time that preload is going to bespecified as pounds. For example, NCCams claims that 150lbs for a pair(75 lbs each) is the proper minimum loading for angular contact bearingsholding a ballscrew. I keep trying to envision how to go about gettingthe bearings ground for that sort of preload. I presented my load cellidea below, but here is another idea.
Let’s build a little fixture that uses some Bellevillewasher springs to come up with 75 lbs of preload. I can get 40lb Belleville springs from McMaster-Carr verycheaply. These springs are 0.312″ in diameter, 0.025″ thickuncompressed, and they reach their desired load after 0.004″ of travel.So let’s design a preload fixture:
The preloader works by applying compressingthe Belleville washers (red) with a 1/4-20 socket head cap screw (blue)from the bottom. This compression is transferred by an anvil (green) thatmeets up with the bearing’s inner races (bearing is purple) and forces
them up by whatever amount the 75 lbs manages. Before use, the fixtureis calibrated by placing the desired preload weight atop the anvil andscrewing in the cap screw until the weight is just lifted up.
Once calibrated, a bearing is placed on theanvil, and the slotted top piece is bolted down onto the fixture to holdthe outer races flat against the surface of the fixture. A small sectionof the bearing’s inner races will protrude. The fixture with bearing needsto be run through a surface grinder until it “sparks out” againstthe outer races, indicating that the inner races are now flush despitethe preload.
If the fixture were to be used a lot, varioussizes of cap and top anvil could be made up to suit different bearingsizes. The only requirement is that the anvil not contact the area verymuch outside the inner races, and the cap be made to only hold down theouter races and provide access to the inners.
I designed this fixture to be largely madeon the lathe because that’s where I think most will get their greatestaccuracy. For best precision it seems to me the top and bottom of thecap as well as the top and bottom of the fixture’s main barrel shouldalso be ground flat and parallel. The slot in the cap need not be precision,and can simply be milled out.
Addendum (12/9/08)
Slocum’s Precision Machine Design book describesa method very similar to thisfor grinding angular contact bearings for preload.
Here you can create the content that will be used within the module.
