[joeavaerage]

Hi,
more good thought has gone in there. Yes, any and all gains in torque are appreciated if torque is the limitation, and realistically with hobby spindles
it is the limitation. The other reality is that going up one or two or even three sizes of spindle gets you more torque than any fiddling around with 4P/8P.

I've considered a two spindle set up, with a servo driven low rpm main spindle like yours, and then a 40-60000 rpm ER11 tiny little guy for all the contour milling. Which would be pretty sweet but a fair amount of hassle

I agree. It takes about ten minutes to swap from the high-speed to the low-speed spindle, which is a drag, but not insurmountable. That is why I had the 'steel' spindle lying inside my mill enclosure, rather than disconnecting it and removing
it completely...and in turn that created the circumstance for the failure. The real reason I made it in the first place was because it was the most cost effective way to get decent steel capability.....at that time.

In the years since I've gotten a bit more income, and have, just barely, scraped together enough for a better high-speed spindle with a modest but useful increase in steel capability. The other glaring fact is that my business
relies on the output of my little high-speed spindle and has done for a while. It is highly appropriate that I at least replace it to ensure that I'm not left high and dry should it fail. It was not really in my scheme of 'things to do',
but when I saw this Ebay listing I couldn't help myself! I have replaced my high-speed spindle and a significant upgrade in addition. The upgrade in power and rpm is not strictly required, but I'm hoping
the ATC feature will pay for itself. Time will tell.

Craig

[joeavaerage]

Hi,
well the 'proof is in the pudding' as they say.

I have the trunnion table in place with the vice atop most. I clamped a small piece of mild steel and have been cutting it with my 'steel' spindle.
The machine is too flexible to be considered good, but it is cutting.

The pic of the cut is 15mm deep, 1mm width with a TiN coated 12mm four flute tool doing 3500rpm, with coolant at 300mm/min feed rate.
Have gone in both directions and climb milling is smoother, but not a lot in it.

As you can see the finish is only fair at best. The tool has a few chips in it and is not helping the finish, but its machine flexure that is largely degrading the finish.

I will repeat this experiment but with the vice clamped direct to the table. This is how I've used this machine before with this spindle and similar if not identical
cutting conditions.

I would estimate that the machine is perhaps 70% 'as rigid' with the trunnion table in place verses not in place. Naturally I was hoping it would be super stiff, but it's not.
Having said that its not too bad either, this cut is about the capacity of my machine. That the flexure has increased with the trunnion table is to be expected.....but by how much?
I have done a few small jobs with my little spindle, rather than this high(er) torque spindle, and any extra flexure due to the trunnion table is hardly noticeable. The extra torque and cutting loads
does show up the difference.

I would guess that the extra flexure is all about the trunnion, however the headstock is also quite wobbly. This has been the case for a long while now.
Indeed, I've been making a pattern to have a new headstock cast. After having spent so much on a new spindle it might just have to wait a while!.

Craig

[peteeng]

Morning Craig - Did the cut sound normal? or did it have a chatter? Looks like a chatter? Peter

FS wiz says up your feed to 1130mm/min so your CL goes up to 0.082mm currently its 0.021mm which is thin so maybe not a big enough bite and its skidding due to lower machine stiffness? Try 0.05mm chip load... double your current then try triple...

[peteeng]

Hi Craig - A small change in speed may do the trick. Herse some info. I think I build a tap tester plus input shaping (IS). IS uses the same info to modify the commands as the program runs. So IS and tap testing is the research project for 2024.





enjoy - Peter

[joeavaerage]

Hi peteeng,
definitely sounded like chatter.

The reason I chose this combination is because I have used the same parameters on another job prior to my 'steel spindle incident'. I am sort of trying to compare
the two cuts. The profile cut I was doing at that time also showed chatter, and I was trying to determine if it is worse (with the trunnion in place), and my guess is yes, it is worse, and by about 25%.
I'm not really trying to find the ideal cutting parameters, but rather just trying to get a broad view on the rigidity or lack thereof of my new trunnion table.

One thing that I noticed for instance, is that when I was torquing the bolts retaining the vice to the platter that it was possible to back drive the platter by several degrees before the servo
would kick in and resist any further movement. I would guess I was applying 20Nm to the platter. That it should rotate should not be a surprise I suppose. The gear reducer is 6.75:1 and the rated servo
torque is 2.4Nm. I would expect therefore the combination would resist 6.75 x 2.4 =16.2Nm applied to the platter. That is about what I observed. As soon as you release the load off the platter it returns
to its neutral position. It sort of feels like a torsion spring, if you lean on it, it will yield but it will too spring back.

I presume the trunnion would back drive also. The trunnion gear reduction is 19.5:1 and if there is movement you might expect it to be less, and to date I have not detected any unintended rotational displacement,
but I presume there must be some.

Judging by the sound of the spindle I would guess the cut I was taking resulted in about 4Nm-5Nm. That in turn caused flexure in the fourth and fifth axes; which should come as no surprise.
The question in my mind is 'is what I have observed reasonable?' To be honest I think yes, it is reasonable. Naturally I hoped for no flexure whatever, but that's unrealistic. What I have observed is an
increase in machine compliance but not devastatingly so. My small spindle, having only 0.3Nm of torque did not cause any observable flexure...so somewhere in that continuum should be a 'sweet spot', I'm thinking
and hoping that about 1Nm to 2Nm will be a nice compromise. That comports with the torque of my new spindle.

Given that my 'steel' spindle is servo driven and the servo drive has an analogue output I feel that I might hook up a voltmeter to reflect the servo current, which is in turn proportional to motor torque.
This would give me a very accurate real-world measure of the applied torque. This may allow my assessment to be somewhat more rigorous.

I would estimate that the most flexible parts of my machine are:
1) The headstock.
2) The trunnion table.
3) The steel frame.

I would like to confirm that assessment and if possible ascribe some numbers to each component and its contribution to the overall machine compliance.

When adding a new piece of kit to an installation there is a time when you have to explore the useful limits of that extra kit...and the contribution or otherwise it makes to
the overall installation. I am far from being definitive about that yet, but my early observation is that when used at low torque (1Nm?)/cutting forces the trunnion table performs well.
At higher torques (5Nm) and cutting loads the compliance of the trunnion table adds to the underlying compliance of the machine with poorer results. To date I am delighted with what I've seen.

Craig

[joeavaerage]

Hi peteeng,
I can't help but think the required torque assessment of FS Wizard is a mile off.

I suggests that 0.34Nm is all that is required to perform the cut that I actually did, and I reckon its total BS. Feeling the heat buildup in the servo motor alone would tell me that its delivering
1.5 to 2hp. At 3500rpm that works out to 3Nm to 4Nm.

This decides it, I will fit a voltmeter to the servo drive and thereby get an accurate servo current measurement. Then I'll know the torque, the actual torque, not some computer model of it,
within 5%.

Craig

[peteeng]

Hi Craig - real data is always good.... The FS torque and force agree via the radius T=Fr so it comes down to how they determine the cutting force... Soon you will know. One of the videos shows the tool forces chatter vs no chatter and the chattering tool requires considerably more force to drive it. That maybe what's happening with your case... Peter

what steel are you cutting? and what max torque can be applied to the tool via the spindle/pulley system (if it has a pulley) this then can be calculated to the max force the tool can apply via its radius. I think its 6.1Nm so a 12mm tool can apply 1016N max to the cut so you have lots of grunt for the cut. F=T/r 6.1/0.006= 1016N

[joeavaerage]

Hi peteeng,
according to FS Wizard the cut that I was taking would require 124W, that's 1/8th hp....BS. It out by a factor of 10, 1.24kW I'd believe, 124W complete rubbish.

The tool holder is direct coupled to the servo. 6.1Nm rated from the servo means 6.1Nm at the tool, and yes it has plenty of grunt. The steel is just an offcut of hot rolled mild steel.

One of the videos shows the tool forces chatter vs no chatter and the chattering tool requires considerably more force to drive it. That maybe what's happening with your case... Peter

I've been milling bits of steel on and off for decades, on drill mills right through to a huge Cincinatti that could spin a six-inch face mill in steel no trouble. I know what sort of power it takes to make steel chips,
and the cut that I was taking would be 1.5hp at least.

Craig

[joeavaerage]

Hi,
some more progress. As it turns out I don't need a voltmeter to measure servo current, the set-up and tuning software displays it live.

The scale is programmable. I set the scale to 1V per 1A servo current. At 3500rpm but under no load it showed 950mV, ie its idling current is just a little under 1A.
I set an offset of 950mV, thus the reading is the load current (over and above the idling current).

With the same parameters as yesterday,12mm four flute, TiN, DOC=15mm, WOC=1mm, 3500rpm and 300mm/min in mild steel the average (average of the climb milling pass and the conventional milling pass)
is 1.5V.

The published torque constant for this servo is 0.59 Nm/A. Thus at 1.5V the torque applied to the tool is 0.9Nm. This is rather less than I expected, but still three
times what FS Wizard had calculated, 0.32Nm.

I then increased the feed rate in line with peteeng's suggestion. At 1300mm/min I measured 5.1V in the conventional direction and 3.9V in the climb direction for
an average of 4.5V which is equivalent to 2.65Nm. This again is still less than I expected but still double what FS Wizard predicts, 1.38Nm.

I advanced the feed rate in increments right up to 5000mm/min, but the chatter was too scary, the brief glimpse at the current measurement was 8.7 V or 5.22Nm. The analog output
has a +10VDC to -10VDC range, and rather suspect that if I were prepared to continue this experiment I would find that the current was actually somewhat higher than my brief glance.

Either way the machine is chattering quite badly, and getting worse as I increased the cutting speed.

The FS Wizard estimate of the required torque does not reflect real world cutting, being in error by some double or triple. Unless I can find some logical reason
for this discrepancy I am tempted to largely disregard the FS Wizard prediction. You are expecting a cut to take 2Nm but in practice takes 6Nm....that's just the
sort of thing that breaks tools and wrecks parts.

I certainly need to repeat this set of tests but with the trunnion table removed and mount the vice direct on the table and observe the difference. I rather suspect the headstock
is the most compliant part of my machine and is therefore largely implicated in the chatter. Removing the trunnion table would confirm/deny that suspicion.

So I have worked out what I would call an upper limit, so now I decided to try to find parameters that result in a better finish. I have one 8mm TiN coated Winstar four flute
left, brand new. I fitted it and tried to work out some reasonable parameters. Much like the previous experiments with a 12mm tool the surface finish was marred by chatter.
I used climb milling exclusively as that resulted in the best cut with the lowest torque. I did find however that even if I were being reasonably aggressive with material removal
(DOC8mm, WOC1.5mm,Feedrate=1500mm/min, 3500rpm)passes that a light and slow pass, sometimes called a spring pass resulted in at least fair finish.

My overall impression is that cutting in the region of 0.5Nm to 1.5Nm with small to moderate sized tools (6mm thru 10mm) will be acceptable provided the toolpath has roughing passes and
then a finishing pass. This is in mild steel. I had hoped for better, but am satisfied with what I've got.

This is largely the first use of the trunnion table, and so trying to get an impression of the overall performance is the order of the day. Rather more refined and accurate measurements
are still very much required. I have to get back into paying work today so those experiments and measurements will just have to wait a while.

Craig

[joeavaerage]

Hi,
just reviewing what I've written and realize there is a small change I could make.

The analog output of the servo drive is a varying +10VDC to -10VDC signal. In this instance it is monitoring average servo current. I set it to display 1V per 1A.
The torque constant for this motor is 0.59Nm/A, so I could set the scale of the display to 590mV per 1A, and thereafter the voltage of the analogue output is a 1:1 scaled
reflection of motor torque. 1V=1Nm, 2V=2Nm. Easy!!!

Then I could hook up a voltmeter to the output. Except for taking measurements is it really of that much value?. When I'm operating my machine I tend to look at the workpiece and the tool,
not looking at some other gauge. I mean I can tell usually at a glance whether the machine is doing it easy or doing it tough. Still I may get around to fitting a voltmeter, after all its only
a couple of wires d finding a place to put it that it does get drenched in coolant.

Craig

[catahoula]

Great report, thanks. I was watching a Stefan gotteswinter video last night and doing FS Wizard calcs as he put a 4flute 6mm endmill to a 10mm deep full slotting cut in tool steel at 2000 rpm. FS wizard (plus the power-torque-rpm equation) puts that at 0.22kw/0.96 Nm, and while it was relatively slow feed with about .01 mm/tooth chipload, and while I'm no expert, it's still hard to believe that's just 1 Nm.

[peteeng]

Evening All - I have been digging around trying to find equations that predict the tool load. There are references saying these exist but they do not say what they are or point at them. There are cutting co-efficients for equations but no equations.... Looking at some of Craigs numbers - DOC=15mm tool dia 12mm feed 300mm/min 3500rpm and radial cut of 1mm FS Wiz says 5.3kgf and 0.31Nm. These are disputed numbers. Milling is an incremental process , it peels the metal away , it does not shear the material in one go like a press punches a hole for instance. If it did the 15mmx3.5mm cut would take around 1.5 tonnes to shear the chip off in mild steel. Obviously this is not our situation. This equation is used with the co-efficients to knock down this number to the "real" number. So if FS is "correct" the knock down would be 1493/5~300x Still looking for a better estimating process. May have to email FS Wiz and ask what algorithm they are using... Peter

[joeavaerage]

Hi peteeng,
its not immediately clear to me about how you would estimate the cutting forces, my training is electrical/electronic rather than mechanical.

The calculation you have outlined seems perfectly reasonable....except its out by an order of magnitude!! Whats an order of magnitude among friends?

I too would be interested in the algorithm that FS Wizard uses. It may well have an 'experience based' conversion factor, and if that's the case then a correction could be made
to the numeric values.

Either way, at this time I suspect that FS Wizard is under reporting the required torque by a factor of two or three. That is my visual observation and that corresponds with
what I believe is an accurate measurement. My intention is to regard the FS Wizard estimate say 1/3 the actual figure, and try that as a starting point. Almost inevitably you end
up tweaking the cutting parameters to find the best possible result. I'm thinking that Fusion FEA is somewhat similar, do I trust it to give accurate real world stress/strain numbers in anything
other than an extremely basic model....no I don't. What I do expect is to give me a starting point, and give some sort of rational analysis to progress to a real world solution, one that can be
experimentally verified.

Craig

[peteeng]

Hi Craig - et al - I have sent a question to the developer but found this in the FSWiz forum. The force is back calculated from an industry table that quotes HP per MRR. This is then converted to a torque as the rpm is speced. once torque is known we know the force from the tool radius. The snip below is from C2013 perhaps its time for a review. Peter

[joeavaerage]

Hi peteeng,
the only problem is that the energy per cubic mm or whatever is that it is an experimental measurement. There is no doubt some theory that underpins it but the number of variables
renders any theoretical analysis a wild guess at best. If I know the energy/unit volume of making chips, then all those calculations I can do no sweat. The problem is the
experimental measurement that underpins it. It can vary so greatly with the subtlest of changes to tool geometry alone.

I have a little more to report on my trunnion/fifth. As you know I make circuit boards with my machine. Mechanical etching is very demanding of Z axis repeatability. A few um too low and the
tool cuts too deeply into the fiberglass, a few um too high and it fails to cut through the copper. I reckon about +-8um is about as much as you can tolerate. I have a software utility that probes the PCB blank
and 'flattens' it in software. Very useful. What it cannot accommodate is any flexure in the machine especially in the Z direction. To date my vice has been bolted to the table and the only flexure in the
Z direction is the basic stiffness of the machine. For the cutting forces involved in making PCB's that flexure is near enough to zero.

My concern is that now I have the trunnion in place and the vice is now bolts to the fifth axis, which is in turn bolted atop the fourth axis (trunnion) that any flexure in those two axes would
compromise my ability to make PCBs with the vice sitting atop-most. My concern is misplaced. I made a board this morning, and the flexure in the Z direction as a result of the (very modest) cutting forces
is still nearly zero. This is good news, and naturally what I was hoping. So I can still use my mill to do all the little low cutting force jobs that I am accustomed to with the vice on the platter without undue flexure
over and above what I would expect were the vice on the table.

I have encountered one wrinkle I did not think of before. If I wish to use the vice (sitting atop the platter) and do regular three axis toolpaths then the trunnion table must be dead flat.
Therein lies the trouble. Lets say you tweak the trunnion by 0.1 degree or six arc min, over a 100mm PCB that amounts to a variation in Z of 175um !!! Yikes!! I would have thought six arc minutes
would have been near enough to perfect. Even setting the trunnion to 0.01 degree or 36 arc second is still 17um over 100mm. I really need to home the trunnion to within 0.001 degree to
0.005 degree ( 3.6 arc second to 18 arc second). The resolution of the trunnion is 250 steps per degree or 14.4 arc second per step. So the resolution of my trunnion is adequate, but little did
I expect to have to get so close.

That's going to make the trunnion table Homing quite a challenge. I think a Home switch plus Index Homing is going to be required.

Craig

[peteeng]

Hi Craig - doesn't your Z probe fix that? or is it too big for the correction? Peter

[joeavaerage]

Hi peteeng,
yes I suppose the probe would fix a PCB blank that is sloped in the Z direction. I have always strived to keep the PCB blank as flat as possible and have not
therefore explored how well or otherwise it performs if the blank is sloped. Even if the software can cope with that what it cannot cope with is flexure in the Z axis.
The probe tip exerts merest milligrams of effective force when probing the surface. The engraving tool exerts tens of grams of effective force downward when it is engaged in
cutting the tool path. If the Z axis flexed even as little as a few um then the tool path will fail for the purpose of a PCB. So, while the software utility is very good at accommodating the somewhat wavy
surface of a PCB blank, the machine must still be very rigid or the tool path will fail to produce useable results.

Many of the small parts I make for instruments are brass, aluminum and plastic. Probing the top of the vice say to establish that its flat would be very time consuming, even if there were
some sort of software correction it would be a real drag. No, I should be able to hit the Home button and the trunnion table should move to its 'flat' position...but the surprise to me
is that I want or need to determine that 'flat' position and have the trunnion achieve it with a few seconds or arc. I had imagined that a few minutes of arc would have been adequate, but no at least
on order of magnitude better precision is required.

I had chosen the resolution of the fourth axis servo such that it was fine (250 step per degree) and yet still be able to do its rated speed of 3000 rpm. When the fourth axis were a lathe chuck having it rotate fast
is a definite bonus for coordinated fourth axis paths where the part may rotate many thousands of revolutions. In order for the tool path to execute quickly in turn requires the fourth axis rotate quickly.
So it is with that in mind that I set the balance of resolution verses speed. In current programming trim the fourth axis can do 150pm.

Now of course the fourth axis has morphed to a trunnion table and it will not be rotating many revolutions, but be restricted to +90 degrees and -90 degrees. Currently the trunnion can move that 90 degrees maximum very VERY quickly indeed,
mere milliseconds, much MUCH faster than I will ever require. Thus I can reprogram the servo to offer much finer resolution but slower max speed. I have not made a decision yet but think that I will change the
resolution to 1 arc second per Step. To stick within the max signaling speed (500kStpe/sec) then I would restrict the speed to 138.88 degrees per second! That's still blindingly fast!!!! Man, you've just got to love
servos!!!!

I have yet to decide how the platter should be programmed. When I bought the gear reducer I was delighted that I found a reducer with lower reduction, namely 6.75:1. This meant that with a
servo at 3000rpm the platter (and attached chuck) could rotate at 444rpm, a real boon for long continuous rotation tool paths. This necessitated a somewhat lower resolution of 100 Steps/degree.
Given that I now want or need to position the rotary axis within seconds of arc the 100Steps/degree no longer looks adequate. If however I increase the resolution to say 1000Steps/degree then the top
speed will be reduced to say 50rpm.

So the C axis will be a trade off, fine resolution (desirable) resulting in lower speed (undesirable), OR high speed (desirable for continuous rotating tool paths) with consequent loss of resolution
(loss of positioning accuracy).

Being able to program the servos with 'Electronic Gearing' is very VERY seductive. If I were using steppers the speed/resolution balance remains pretty much 'as designed' with little or no flexibility.

I know you are considering a trunnion fifth axis so here are a few things that I would recommend that you consider:
1) If you intend to have continuous rotation tool paths then you must choose the gear reduction and resolution to give an acceptable max rotation speed of the axis
or risk tool paths taking an age to run.
2) The resolution required to accurately position a rotary axis is measured in seconds or arc not minutes of arc.

These two requirements are contrary to each other. That brings me to my last recommendation: use a servo as the electronic gearing will allow you to tailor the resolution/speed balance
where steppers do not.

Craig

[joeavaerage]

Hi peteeng,
try this for a calculation to do your head in, you may want to take your shoes off, you'll need fingers and toes for this one!

My Delta servo has a rotary encoder of 160,000 counts per rev. The gear reducer of the trunnion table is 19.5:1.

According to my rationale above I want the resolution of my trunnion table to be 1 second of arc, or 3600Steps/degree, where one degree is one unit.

So the number of encoder counts that corresponds to 1 second of arc rotation of the trunnion is
EncoderCounts= (1/(360 *3600)) *19.5 *160000
=2.407407407407407

I need to use the electronic gearing of the servo to do this. I wish my motion controller apply one step and have the servo rotate 2.407407 encoder counts (out of 160000).
For this I need to find whole number numerator and denominator such that:

N/D=2.407404407407

Being first thing in the morning here I could not devise an analytical calculation that gave me an answer so I cheated and used Excel

As you can see the first whole number ratio that satisfies the requirement is 27 and 65. So If I program these two number into the servo drive, a one minute exercise given that I have PC software do it.
Then if I apply one step pulse to the drive it will result in the trunnion rotating by precisely 1 second of arc. That is then the Step/Unit value I will put into Mach4, namely 3600 Steps per degree. Then I will
limit the speed to 90 units (degrees) per second. That will in turn mean the maximum signalling rate is 90 x 3600=324000 Step/second or 324kHz. My ESS/BoB/servo drive handle 500kHz in a canter....so
this combination will be fine.

As I say, you've got to love servos, the flexibility to match the machine is a delight.

Craig

[joeavaerage]

Hi peteeng,
have just implemented the scheme, so now one Step is one arc second of the trunnion.

At the smallest step size I have programmed for jogging is 0.001 unit per MPG detent, or 3.6 arc seconds per MPG detent. Very nice and smooth, and you can dial in very nicely
indeed and still go 90 degrees per second, which is still way, WAY, WAY fast for a trunnion table.
The Mach settings are:
3600 Steps/Unit
5400 Units/minute max velocity
540 Units/sec² max acceleration

The servo drive setting are
Numerator = 65
Denominator= 27

Craig

[joeavaerage]

Hi,
suitably encouraged by the trunnion axis I'll repeat the same process for the C axis or platter. I do want to retain the highest possible rotational speed as I reasonably may, so that should I use the Caxis for
continuous rotation tool paths that its as fast as I can.

For this reason I have decided that the resolution of the C axis be 6 arc seconds per Step, or 600 Steps/unit.

The servo has a 160,000 count/rev encoder. The C axis platter gear reducer has a ratio of 6.75:1.
EncoderCounts(per Step)=(1/(600 * 360)) * 6.75 *160000
=5

Well that's easy, just 5 encoder counts per 6arc seconds, my target C axis resolution, don't even need Excel for that!!

So:
N=5
D=1

If I specify that 500kHz is my max signaling rate then the max speed is:

500000* (6/(3600 *60))
=2.31 revs/sec or 138.888rpm. Slower than I would like, but I think the balance is about right. I think I'll try it out and evaluate whether its good or not. Lets say I restrict the speed to 100rpm,
a good human rounded number, then the max speed in units/min = (360 *100)=360,000 degrees per minute. That also corresponds to a signaling rate of 360kHz, well within the capability of the ESS/BoB/servo drive.

Have programmed it no trouble:
Mach settings:
600Step/unit (ie 6arc second per step)
360,000 units/min max velocity ie 100rpm
36000 units/s² max acceleration

The servo drive settings are:
N=5
D=1

Seems very smooth. Will have to wait to see whether 100rpm is satisfying with longer continuous rotation tool paths.

Another nice thing about programming servos in this way is that you can save all the parameters to a file, either to clone a new drive or send to a friend.

Craig