[Andre' B]

Haas states that they do 1000 blocks per second block execution (same as ServoSoft). Haas states that they do "acceleration before interpolation" (I think the same, just worded differently). This all sound the same to me. What am I missing in all of this?

For smooth precise high speed control the servo loop rate is more important then the block processing rate. Servo loop rate should be over 10Khz. The next most important factor is acceleration, if the machine can not change the feed rate very fast it does not matter what the max feed rate is you will never see it on a lot of moves.

The higher pulse counts are for velocity control not position control. At slower speeds you need as many pulses per servo loop as you can get.
The other option is analog tachometers, but they do not fit the all digital world we live in today.

[Drassk]

So what are you trying to say, the largest cnc control builder Fanuc's top of the line control with nano interpolation, ai contour control, and nano smoothing isn't as good as a servosoft system?

So all you need is 10 or 20 pulses for the minimum programmable increment? Hmmm....I guess fanuc better just throw in the towel. I'd be willing to bet that if they have encoders with 16 million pulses per rev the rest of the hardware is on par with the encoders.

Someone posted a useful message while I was writing this one but:

Yes, 10 or 20 pulses is sufficient. Theoretically, 2 would be good enough but that would be pushing it and rotary motors are generally only good for +/- 10 encoder counts or so. More counts will give smoother motor motion to an extent.

So far as the SS VS Fanuc controller comparison (that I never made): you're confused. Learn what the words I'm saying mean before you try to put other ones in my mouth.

These things are all independent, and all important:
encoder count
servo loop refresh rate / latency
lookahead

Encoder count, ignoring mechanical inaccuracy, is the absolute precision with which you can read position. Servo loop refresh rate controls how fast you can move while being certain of maintaining a given accuracy. Since motors under load can't turn on a dime, lookahead is there to keep the inertia of the system within bounds.

[Andre' B]

Here is a resonable attempt to describe HSM.
http://www.geonicsinc.com/reference/Hot_Zone.pdf

[Edster]

Here is a link to the fanuc brouchure for the 31i control for anyone that's interested.

http://www.fanuc.co.jp/en/product/ca...A(E)_v06_s.pdf

Drassk, my original post stating the # of encoder counts was for informational purposes for some of the eariler posters that were discussing encoder counts and minimum programmable increments. Some builders don't advertise the encoder counts. I have owned haas and fadal machines for years and don't remember reading about encoder counts unitl I got my first okuma. In fact I'm not even sure how many encoder counts there are for the yaskawa servos haas uses. For your clarification I'm not under the impression the only important factor of a high speed servo system is the # of encoder counts. I took your statment "More/better information does you no good without the ability to apply it and change course." Immediatly after my post about the number of encoder counts fanuc has, as implying you think the fanuc control doesn't have the ability to process the data from it's encoders. Your previous posts boasting the speed of the servosoft control is where the comparison came in.

Hsm is about speed and accuracy. We can already move fast. It's about moving as fast as you can while keeping the tool positioned as accuratly as possible. The point of nano interpolation is to increase the resolution of the servo, which is important if you want to move the axes as accuratly as possible. The servo moves an axis in a linear motion. The resolution of the servo encoder controls how small of a linear move the axis is capable of. Which is independant of the mininimum programmable increment, btw. This determins how close to a curved path multiple servos can actually get.

Cam systems usually output linear moves. The tolerance of the toolpath in your cam software is what determines how small the moves are. Nano smoothing recognizes this and attempts to infer a smooth path between the line segments. This coupled with nano interpolation results in machine motion that is as close as possible to the surface/model the program was derived from.

There are pics in the link to the brochure above that show the differences between nano interpolation and regular interpolation, and that illiustrate the nano smoothing option.

Drassk, keeping all this in mind I don't understand how you could think 10-20 encoder counts per minimum programmable increment is sufficient.

There is a lot going on here, but the main goal is for the machine to produce a shape that is as close as possible to the surface/model it was programmed from and do it as quickly as possible. Ultimatly we all want to be in the tavern instead of at the polishing bench.

I stand behind my statement about servosoft. They have been around since 98 so if they are truely the best why aren't there controls on the best machines?

Keep them chips flying

[Drassk]

Here is a link to the fanuc brouchure for the 31i control for anyone that's interested.

http://www.fanuc.co.jp/en/product/ca...A(E)_v06_s.pdf

Drassk, my original post stating the # of encoder counts was for informational purposes f

...

Your previous posts boasting the speed of the servosoft control is where the comparison came in.

...

Drassk, keeping all this in mind I don't understand how you could think 10-20 encoder counts per minimum programmable increment is sufficient.

...

I stand behind my statement about servosoft. They have been around since 98 so if they are truely the best why aren't there controls on the best machines?

Keep them chips flying

Dude, you have to read the names above the posts before you start responding. I never said anything boastful about ServoSoft, that was some other guy at the start of the thread! If anything, I knocked them a bit by mentioning two other companies putting out superior products at a much lower cost!

The ten to twenty counts for a programmable increment is a combination of things. One is the accuracy of encoder-based motion on rotary motors: they're usually only on for a few counts so more counts means less likeliness they settle too far off the mark.

If using the encoder to commutate the motor then it's going to roll smoother if it's getting more continuous feedback. And, as someone mentioned, at lower speeds you still want to be getting enough counts coming through the drive to get good velocity control.

The last thing is just relating sampling aliasing to motion. Theoretically you'd be good with sampling a signal with two samples per oscillation, but in practice you only get a pretty good fit with ten times the sampling rate.

Having orders of magnitude more encoder counts won't give you any more accuracy if the software is thinking in tenths and the mechanical side is only accurate to tenths. In the case of massive count numbers they're usually interpolating which means that the count numbers are possibly useful for smoothness but not so much for increasing accuracy (they're analogous to microstepping a stepper). Sticking a sine/cosine interpolator on a 1024 count encoder and interpolating to ten bits doesn't give you rotary motion accurate to a half-millionth of a radian even though you get a million ticks out of it. To give the customer the true picture now, the company would have to say how many physical counts are on the encoder and the accuracy to which the disk itself was manufactured (so that you can figure out how much interpolation is useful). I'm sure Fanuc makes a fine product, but interpolated encoder counts are almost useless information without knowing how they reached that number.

[Edster]

Dude, you have to read the names above the posts before you start responding. I never said anything boastful about ServoSoft, that was some other guy at the start of the thread! If anything, I knocked them a bit by mentioning two other companies putting out superior products at a much lower cost!

The SoftServo system has a 2000Hz servo update rate over EtherCAT and most systems are even slower. Of course, it's rather rare that one would have to do 10,000 consecutive orthogonal moves!

I understand it was mactec54 that was claiming the servosoft was the best high speed control, but he wasn't the one that implied okuma or fanuc don't have enough processing power to use all the data from their encoders. Or the encoders they use are overkill for hsm.

Having orders of magnitude more encoder counts won't give you any more accuracy if the software is thinking in tenths and the mechanical side is only accurate to tenths.

What software? Your cam software? The machine's control's software? Just because you can't enter a command smaller that .0001 in the control doesn't mean the control isn't moving the servos in smaller increments. If your theory was correct and you wanted to move from 0,0 to .0001,.0001 would the servo jump .0001 in x or y first. It's going to take a bunch of small steps to get there and the resolution of the servo system is what determines how many steps. Hence the point of 16 million count encoders and nano interpolation. Nano smoothing and nurbs based interpolation address the issues of short line segments generated by cam software.

In a high quality ballscrew setup with pretensioned ballscrews and double ball nuts that are preloaded against each other, how much backlash do you think there is?

What about machines with linear motors with no physical backlash. Would nano interpolation be overkill there.

[Drassk]

Ed,

I don't know how much more clearly I can put this, but here goes: I'm a grown-up. When I talk in generalities about control equipment, you probably shouldn't assume that I'm trying to 'my dad can beat up your dad' you or your gear. I'll address you if I'm talking to you and I'll explicitly say if I think some piece of equipment is sub-par, because that's how grown-ups talk.

(Yes, I am explicitly questioning your maturity)

I noted the SS refresh as relatively slow but still faster than many controls...for example the giant percentage of CNC machinery made in the 1990s and still in service. Most of my equipment is from that era. Nothing about you in there.

How you would surmise that a minimum of ten counts per axis per move increment translates into one-shot motion jumps is a mystery to me.

I originally thought you were just a little underinformed and wanted to learn, but I was wrong about your motivation for participating. If anyone who wants to understand this stuff has any questions, I'm still reading the thread so ask away.

[Edster]

What you post a response about what someone posts, right after their post who do you think your addressing?

Your last post says a lot about your maturity level.

[Edster]

Btw, your fadal 4020 looks like a nice machine. Which control is on it? Couldn't tell from the pic on your website.

[Drassk]

88HS, with a -4 board in it. It's really a 386 on a custom motherboard. The -5 boards have a Pentium on them, and a higher RAM capacity, but odds are that I'll retro it rather than pay to to upgrade to a different 15-year-old Fadal controller. Definitely a machine handicapped by its brain rather than its body.

I found a sheet in the back of the machine with my ballscrew calibration surveys on it recently, which was the biggest wall in the way of retrofitting it (that and being able to be without my bread and butter for a week or two while I trace wires and debug).

[Edster]

How you would surmise that a minimum of ten counts per axis per move increment translates into one-shot motion jumps is a mystery to me.

Bad wording on my part. I was just trying to explain even though we want the cnc to move in say a straight line at an angle it is actually makes a stair step motion. The stair step motion is incredibly small, but it's still there. Nano interpolation makes the stair step smaller.

I have a 94 fadal vmc 15 with the same control on it, cnc88hs with the -4 system. I'm still happy with the machine and it runs parts for us every day. I upgraded the mem to 422k from the 38k it originally came with. I wanted to go bigger but that was the max for the -4 system. A service tech that was out here said I could go up to 16mb with the -5 system, but he said it wasn't worth the upgrade because their isn't much gain from going from the -4 to the -5 system.

Which control were you thinking about retroing to your fadal?

[Drassk]

Bad wording on my part. I was just trying to explain even though we want the cnc to move in say a straight line at an angle it is actually makes a stair step motion. The stair step motion is incredibly small, but it's still there. Nano interpolation makes the stair step smaller.

I have a 94 fadal vmc 15 with the same control on it, cnc88hs with the -4 system. I'm still happy with the machine and it runs parts for us every day. I upgraded the mem to 422k from the 38k it originally came with. I wanted to go bigger but that was the max for the -4 system. A service tech that was out here said I could go up to 16mb with the -5 system, but he said it wasn't worth the upgrade because their isn't much gain from going from the -4 to the -5 system.

Which control were you thinking about retroing to your fadal?

The 10-20 steps to a measurable increment is just a subdivision that tends to be the point past which returns are almost null. In other words, if you've got 20 diagonal steps and you can only see the 0.0001" square then the diagonal looks 'perfect' at that point. If the steps beyond 20 were 'accuracy' steps then your measurable motion wouldn't be 0.0001", it would be smaller, and the machine would be advertised as such. Mechanical limitations mean it's really hard to make a milling machine of any size good for better than a tenth.

My -4 has 180K of memory in it, and I have a bag of the memory chips that I bought a while back to upgrade it but I've been too busy to pull out the board and solder them in. My tech is actually big on the idea of the -5 over the -4; I think it does have a better refresh rate, and the extra memory would be handy in some cases. There are programs I filter more coarsely than I would like because there isn't enough memory on the control and DNC can't keep up.

I'd retro it with either a Dynomotion controller or a Galil. I'm going to build a machine with the Dyno, where I'm not in trouble if there are problems, and if that's solid then I'll go with them on the Fadal. Their platform is really beautiful if you're a nuts and bolts and algorithms kind of guy. 'My tech' is in California and I'm here, so when anything goes I have to fix it myself with some phone support. Crappy when things are down, but over time I've learned an awful lot about what's where and how everything fits together and that'll help a lot on the retro!

[Edster]

Good article on nanosmoothing.

http://www.imakenews.com/gefanuc/e_a....cfm?x=b11,0,w

Quote from a fanuc brouchure:

Whether used in turning or milling applications, 0i-D Series CNC controls can implement the "nano-interpolation" function to satisfy the most demanding finish requirements. This function generates positioning commands in nanometre units, which the downstream servo amplifiers convert into the corresponding currents. High-quality servo systems can follow the cutter path much more accurately because interpolation does not round off even the slightest movements, and they consequently achieve the best surfaces at high feed rates and speeds.

Link just in case:

https://www.fanuccnc-parts.eu/master...&yearonly=true