[WesM]

Hi,

Thus even a 400W servo would easily hold the axis aloft using only a small proportion of it's available torque to do so. It would be a good idea to have a servo with a brake though,
so when the power is cut off the weight of the Z axis does not back drive the screw and the Z axis fall towards the table. So 400W should be ample, but if buying two 75W servos for
X and Y then one more for the Z is no trouble. Often there is little price difference between a 400W and 750W servo, so I would go bigger. If at some stage you increase the spindle mass
say then you will still have the Z axis authority to accommodate it.



I do not like Clearpath servos, they are too expensive for what they are. Over $500 for a 400w servo. They have only one digital output, a sub par encoder and just too bloody expensive.
Compare that to a 750W Delta B2 for $438, with a 160,000 count/rev encoder, 6 digital outputs, 8 digital inputs,2 analog outputs, auxiliary encoder output, 230VAC line input, ie no
power supply required.......twice the power and vastly better IO for less money. Tecknics do have an excellent reputation for quality and support, of that there is no question.
They also follow the forums and hound me every time I post this opinion.

Craig

Hah thats funny about the clearpath folks. I wanted to give them a shot though, since a buddy of mine likes them. Still, like you I had noted they seemed very expensive for what you get.

The reason I suggested the 400w servo for my Z is the 400w Delta and Dmm servos have 14mm shafts, which would be a clean swap for my existing timing pulley and it matches up with the existing mounting holes too. The 750w Delta servos have 19mm shafts, so I will have to get new timing pulleys for them and also modify the mounts. My X and Y axis mounts are much easier to modify though, the Z would be a pain. Like you said, the cost difference is not really the issue, its mounting that I am thinking about.

[joeavaerage]

Hi,
I should not be harsh on Clearpaths, many people use them to complete satisfaction, and really that's what counts. I think I can do better for my money, and that's what counts in my book.

I do not believe Clearpath have any models with electromagnetic brakes. You can fit an external brake but I seem to recall someone posting the the external brake was $375.

Clearpath do have 900W models, but they are the next tier up and closer to $1000 each.

Craig

[burs]

Maybe I am missing something, but it seems like I cant edit posts I have made? odd.

Hello @WesM, posts can be edited by members within 60 minutes. After this time window, only moderators and admins can edit them.

[RaderSidetrack]

Maybe I am missing something, but it seems like I cant edit posts I have made? odd.

Editing posts depends on the time frame:

Posts can only be edited within 60 minutes. After this time window only admins and mods can edit them..

Otherwise, we would probably lose control over spam posts.

Its unfortunate that spammers engage in bad behavior, but if editing is not time-limited, spammers create an "innocent" post, then sometime later go back and insert spam links. Since it is no longer a "new" post that kind of spam is very difficult to detect. Allowing unlimited editing would result in spam all over the forum.

You can send a PM to a moderator asking them to make an edit -- I am happy to edit older posts. Or make you can make a new post with a correction.

[Teknic_Servo]

Hi WesM,

It sounds like you may have already reached to Teknic and spoken to someone, but if you haven’t, we’re happy to help. Please feel free to contact us through https://teknic.com/contact/, or call us directly at +1-585-784-7454.

Hi Joeavaerage:

[Teknic] hound me every time I post this opinion.

Teknic follows these forums (and many others) to respond to users and provide support, we’re sincerely not trying to hound anyone. Teknic’s two reasons for responding to any forum posts are to answer questions and correct factual errors or misleading information.

I’m sorry if you feel that Teknic is hounding you in any way. That is definitely not our intent. A few of my colleagues have tried to explain Teknic's goals in responding to forums on some of your previous posts, but I'll restate them here for others as well. Teknic follows these forums to respond to technical questions and to correct any factual errors or misleading information about our products and service (and many times other topics related to general motion control).

If you state your opinions about Teknic or our products without any incorrect or misleading/incomplete information, regardless of how negative your opinion is, we will rarely post a response. And, if you post a misstatement of fact that puts a Teknic product in a better light than it deserves to be based on the facts, we'll correct that too (you can see many examples of this on YouTube or various forums). Our goal is to help others understand motion control technology and apply it in a way that makes them successful. If this forum’s members (and the general market) are well-educated, they will choose the best products for their needs.

With that in mind, I have the following comments and corrections:

I do not believe Clearpath have any models with electromagnetic brakes. You can fit an external brake but I seem to recall someone posting the the external brake was $375.

Teknic offers NEMA 23 brakes for $219, and a NEMA 34 brakes for $296 (single-piece pricing).

17 bit encoders are nice, but not critical, a well set up servo with a 10,000 count encoder will be perfect for CNC in every way.

[Teknic has]…a sub par encoder…

Teknic’s NEMA 23 and NEMA 34 ClearPath motors contain a 12,800 count optical encoder, and the NEMA 56/143/D100 models have a 64,000 count per rev optical encoder.

A few of my colleagues have made mention elsewhere on what makes for a high-quality encoder, but it bears repeating here:

It would be fair (although incomplete) to say that the Delta encoder has more resolution based on its specs, but it isn’t accurate to say that the Teknic encoder is subpar. As you stated, “a 10,000 count encoder will be perfect for CNC in every way.” Moreover, resolution is only one figure of merit for an encoder, and the number of counts per revolution does not convey any information about the other important metrics that characterize a high-quality encoder.

In order to get extremely high resolution out of a magnetic encoder (or even a high-quality optical encoder, for that matter) you have to do an extreme amount of interpolation. An encoder on a motor at the price point you mention will typically have noise, drift and other variability that makes the accuracy about 15 times (!) worse than the resolution. And if that isn’t enough, the filtering required to pull that fine a signal from the noise means that the servo gets its information delayed (and if it’s a serial encoder, the communication delays make it even worse). This all has a significant negative impact on servo performance.

Finally, even with a hypothetical “perfect” encoder, the mechanical imperfections in the motor and bearings will turn that extreme resolution into random noise, so even if you’re not bothered by the poor accuracy, the noise means your servo loop gets feedback that impairs its smoothness and dynamic performance.

We know all this because our engineers evaluated these encoders for ClearPath (and again recently for some yet unreleased products). It’s tempting to use these encoders because they’re so cheap, and because some people do mistakenly assume that more resolution must always be better. But we’ll be sticking with a high-quality, more expensive optical encoder and reasonable interpolation until there’s a magnetic encoder that can give us better performance.

Thanks,
Bridgette O. – Teknic Servo Systems Engineer

[joeavaerage]

Hi,
I say you hound me because every time I post that your servos are too expensive for what they are you come along and try to correct me....that is by any definition being
hounded.

You have argued that your encoders a good quality, and I have no dispute. Also I have posted that many a servo with a 10000 count encoder (of good linearity) will be in every way adequate
and were for many years considered state of the art.

What you do not seem to accept however is that a higher resolution encoder allows a higher resolution numeric data stream that approximates an analog control where the only limitations
on performance are linearity and noise. Linearity is and always will be a critical performance limitation. The degree to which a digital system can approach perfection is determined by its resolution.
I argue that a higher resolution encoder allows a 'smoother' numeric enactment of the Field Control algorithm. My Control Engineer Professor called it 'discrimination'.

The Nyquist Sampling theorem tells us the max theoretical frequency that can be detected is half that of the sampling rate. A Control Engineer will tell you that any reasonable discrimination
cannot be achieved by a sampling rate of less than five times that of the highest frequency of interest, and preferably more like ten times.

All the leading servo manufacturers of the world are cheerfully touting 20bit, 24bit and more resolutions....and yet you expect to sell the notion that 12800 is good.......well I say is adequate but adequate only.

Teknic offers NEMA 23 brakes for $219, and a NEMA 34 brakes for $296 (single-piece pricing).

I stand corrected, but still they are expensive.

Craig

[RCaffin]

I am not sure I understand the interest in very high resolution encoders. They seem pointless to me. I had better explain why.

My CNC has DC motors with 256 line encoders, a 3:1 reduction via backlash-free GT2 belts to ball screws with double nuts.
This combination gives me a single-step resolution of 0.8 microns. I have checked that this is so with a good touch-probe.

However
I doubt my ball screws have good enough linearity to justify a claim of 0.8 microns linearity over more than 5 mm.
I doubt my bearings can really claim to have less backlash than 0.8 microns.
I even doubt that the encoders, good though they may be, are sufficiently well centred on the motor shafts to have a wobble of better than 0.8 microns effective.
I was able to show that my mill table is NOT flat to that accuracy. Maybe to 10 microns end to end.
I doubt my ability to set a zero on any axis to 0.8 microns.

But Mach3 + ESS + Geckos will drive the CNC to within 1 step on all axes. Can I justify anything better?

Cheers
Roger

[joeavaerage]

Hi,
an AC servo runs on a Field Oriented Control algorithm first proposed and published in the 60's. The control is essentially mathematical. There are some superb videos made
by Texas Instruments about Field Oriented Control and it explains it superbly if you are a newcomer to Field Oriented Control.

A stream of digital data, a string of numbers is processed. The closer those numbers are to each other (resolution) the smoother numeric representation of a real variable say,
armature angle. The smoother the numeric data the more and more the calculation resembles a continuous or analog value.

The respondent form Tecknics argues that beyond a given point the non-linearity of an encoder renders the finest resolution moot, and I do agree. She makes the argument that if the
non-linearity is greater than the resolution then you might as well give up.

What the same respondent does not consider is the 'smoothness' of the numeric calculation that happen inside the drive.

I would agree with you that increased resolution of your encoder beyond it linearity is not likely to result in any greater positioning accuracy, but internally the smoothness or discrimination is such that highest bandwidths
REQUIRE highest possible resolution.

If you had to measure a quantity, lets say current or position, but some real and varying quantity, with only modest measuring equipment. You might measure the quantity more than once and average over
several measurements, hoping to achieve greater accuracy. Its not that the measuring equipment has improved but the averaging has reduced the uncertainty. High resolution encoders offer a similar
kind of thing. The encoder measures an angle. Tecknics would have you believe that its only necessary to measure the angle up to the limit of the linearity of the encoder. I argue that you measure
closer, much, MUCH closer than that if you can. There is the distinct possibility that the non-linearity will skew your measurement but if you don't do so then you preclude to opportunity to increase the
bandwidth by the averaging effect.

I repeat the encoder resolution is not particularly about positioning accuracy but the numeric stability and smoothness of the mathematical algorithm going on inside the servo drive.

Craig

[joeavaerage]

Hi,
I recall first year Chemistry while at University, like all budding chemists we had repeatedly to titrate different solutions with best possible accuracy. You were graded on the accuracy of your answer.

On a burette there are markings of 10th of a ml. So when doing a titration you might record 54.6ml plus/ minus 0.1ml. The Chemistry Professor blasted us, he said that yes the uncertainty is 0.1ml
but we should attempt to read much much better than that. In the first instance he insisted we practice ourselves to measure within 0.025ml, and then latter, now we 'had our eye in' to 0.01ml.
He was of course correct. When you do the calculation you had to record an uncertainty of 0.1ml but you measured better than that, to 0.01ml....and that was your only hope to achieve the sort
of accuracy that the Chemistry gurus were looking for.

The lesson has lasted a lifetime. If an approximate answer is good enough then there is no point in measuring beyond the linearity or uncertainty.....but in situations where you need the best result
then you extract every last digit of precision you can irrespective of linearity or uncertainty, because you want the best you can get.

Craig

[RCaffin]

Hi Craig

The thing which has concerned me in the past about those encoders with a calculated output (as opposed to the simple design which simply counts edges in quadrature) is the delay inherent in the calculation and also in the serial transmission of the data. The delays limited the servo bandwidth. It may be that today the processing speed and the data Tx speed are high enough to make those concerns obsolete.

Your comments about accuracy are important of course. However, I can machine a round stub from above with a good cutter and have it come out very satisfactorily 'round' and smooth. That is enough for my needs. And the gear is a lot cheaper, which matters for a hobby.

I agree with your chem prof about accuracy, but in context. I am a physicist specialising in measurement, albeit now retired. It was always a rule of thumb that the measuring instrument should be 10x better than the required accuracy, which made life interesting for the people at National Standards! I once had a servo-controlled laser system which had a frequency accuracy beyond what our NatStand lab had, so when I finished with it they begged it off me. The design was very cute.

For my needs a machine with 0.8 microns resolution is quite enough to give me parts made to 10 microns. I dare say the machine could be better than that IF I wanted to push it. But beyond that I would have to have air conditioning for the workshop - which is an old barn built around 1900. And 10 microns is 10x better than I could do with a manual machine!

Cheers
Roger

[joeavaerage]

Hi,
when I was studying physics there was a much more rigorous approach to uncertainty.

Chemists use the notion of a fixed uncertainty, say +-0.1 like my previous post. That assumes that there is an equal probability that the actual measurement was actually in one of
a number of equal bands within the uncertainty window. If the measurement is 54.6 then there is a 10% probability of the actual measurement being in the band 54.55 +_0.01 and a 10% probability
that the actual measurement being in the band 54.65+-0.01

Physicists had a rather different view that uncertainty is normally distributed. Thus a measurement would be 54.6 +-0.1 with a 95% probability, ie three standard deviations. But that also suggested that
the actual measurement is 54.6 +- 0.033 with a 67% probability, ie within one standard deviation. This suggested that you should make a measurement with the best possible accuracy you can for despite uncertainty
it is still probabilistically the best estimate you have of the true measurement.

This is the argument I have with Tecknics, they seem to think that if the linearity is such that the uncertainty exceeds the resolution that you should not bother. I say you should bother if you want the best measurement
you can get irrespective of the uncertainty. It would appear also that other servo manufacturers have come to the same conclusion.

The argument for better resolution is less about final positining accuracy but the smoothness of motion that the servo can achieve with increaed resolution.

As an example I have my Delta servos direct connected to 5mm pitch ballscrews. For my purposes I have settled on a 1um resolution, ie the servo is required to assume one of 5000 distinct
positions within one rotation....but the encoder has 160,000 counts per revolution. This means that each of the 5000 positions has 32 encoder counts between locations and can therefore position
itself with a smoother action than if there were just one count for each of the 5000 distinct positions. Just because I demand or require a resolution of 5000 internally the servo operates at 160,000
resolution. I hasten to add my B2 series servos are old school now, nearly all the manufacturers current models are at least 20 bit and more commonly 24bit. Again this results in smoother movement
rather than ludicrously high positioning resolution.

Craig

[jaguar36]

The argument for better resolution is less about final positining accuracy but the smoothness of motion that the servo can achieve with increaed resolution.

As an example I have my Delta servos direct connected to 5mm pitch ballscrews. For my purposes I have settled on a 1um resolution, ie the servo is required to assume one of 5000 distinct
positions within one rotation....but the encoder has 160,000 counts per revolution. This means that each of the 5000 positions has 32 encoder counts between locations and can therefore position
itself with a smoother action than if there were just one count for each of the 5000 distinct positions. Just because I demand or require a resolution of 5000 internally the servo operates at 160,000
resolution. I hasten to add my B2 series servos are old school now, nearly all the manufacturers current models are at least 20 bit and more commonly 24bit. Again this results in smoother movement
rather than ludicrously high positioning resolution.

Craig

You're assuming though that those encoders have a similar level of noise, accuracy and consistency of the clearpath ones. You can throw more encoder resolution on a servo but if it comes with increased noise, decreased accuracy and less consistent results it could easily be worse. I'm not saying the Clearpath encoders are better, I have no idea if they are, but just looking at how many bits the encoder has is only one piece of the puzzle.

Going back to your chemistry analogy if the table is bouncing around like crazy while you're trying to do your titration, it doesn't matter how fine the markings on your buret are.

[joeavaerage]

Hi,

but just looking at how many bits the encoder has is only one piece of the puzzle.

I rather think you've missed the point of the argument. Lets say the encoder is not quite linear and that position 3541 (of our 5000) is off by 10 arc sec. The servo will eventually arrive at position 3541
but will perforce be off by the same 10 arc sec. The high resolution encoder means there are 16 encoder pulse either side of location 3541 and those extra pulses allow the servo to more smoothly
approach the location EVEN it is still off by 10 arc sec.

The high encoder resolution allows smoother motion, but the ultimate positioning accuracy is still determined by the linearity of the encoder.

As to which, Delta or Clearpath encoders are better, or at least more linear, than the other I cannot say, but that does not describe the smoothness of movement. By my argument I expect the Delta servo to move
more smoothly in the vicinity of its commanded position. Whether that position is more or less accurate.......neither company publish the data to make that determination.

Going back to your chemistry analogy if the table is bouncing around like crazy while you're trying to do your titration, it doesn't matter how fine the markings on your buret are.

100% agreed. Still the best estimate you have of the true measurement is to read the burette as closely as you can anyway. Probabilistically its the best you've got, whether it be in error or not.
The probability approach used to measurement as employed by physicists leads and lends you to make the best measurement you can whereas the chemists approach suggests that you
need not bother with that last digit of precision because its in the error window and therefore has no significance.

Craig

[joeavaerage]

Hi,
actually I do a dis-service to chemists. The measurements I am referring to were made by ham-fisted first year University Chemistry students.....at more advanced levels then there is no lack
of sophistication or mathematical rigour applied to their measurements.

Craig

[jaguar36]

Hi,



I rather think you've missed the point of the argument. Lets say the encoder is not quite linear and that position 3541 (of our 5000) is off by 10 arc sec. The servo will eventually arrive at position 3541
but will perforce be off by the same 10 arc sec. The high resolution encoder means there are 16 encoder pulse either side of location 3541 and those extra pulses allow the servo to more smoothly
approach the location EVEN it is still off by 10 arc sec.

You're assuming that position 3541 is always off by 10arc sec. If its off by 10 one msec and then 20 the next, its not going to be smooth at all, its going to be jerking all over the place. Adding extra digits to the position reading doesn't matter if that reading is noisy and inaccurate, you're just putting lipstick on a pig.

[joeavaerage]

Hi jaguar,

You're assuming that position 3541 is always off by 10arc sec. If its off by 10 one msec and then 20 the next, its not going to be smooth at all, its going to be jerking all over the place. Adding extra digits to the position reading doesn't matter if that reading is noisy and inaccurate, you're just putting lipstick on a pig.

Then just ignore the whole idea. Yaskawa and Seimens have not.

Craig

[jaguar36]

Then just ignore the whole idea. Yaskawa and Seimens have not.

Presumably they use high quality encoders where the higher resolution is actually beneficial too. Perhaps Delta does as well, I'm merely expanding on Teknic's point that having a higher resolution encoder does not nesscarily make a better servo. I am not insulting your choice in servo.

Similar to the op, I'm trying to decide on a set of servos and have been looking at Delta, Yaskawa and Clearpath. There seem to be alot of folks who are very happy with Clearpath, and of course Yaskawa is used in lots of 'real' CNCs.

[joeavaerage]

Hi,

There seem to be alot of folks who are very happy with Clearpath

That is certainly correct, they work well, are simple, well at least simpler to set up than most, but by my reckoning have a sub par encoder, one and one only digital output and they are expensive for
the power output. Take from that what you will. Its your money and your choice.

Craig

[ardenum2]

Hi WesM,

It sounds like you may have already reached to Teknic and spoken to someone, but if you haven’t, we’re happy to help. Please feel free to contact us through https://teknic.com/contact/, or call us directly at +1-585-784-7454.

Hi Joeavaerage:



Teknic follows these forums (and many others) to respond to users and provide support, we’re sincerely not trying to hound anyone. Teknic’s two reasons for responding to any forum posts are to answer questions and correct factual errors or misleading information.

I’m sorry if you feel that Teknic is hounding you in any way. That is definitely not our intent. A few of my colleagues have tried to explain Teknic's goals in responding to forums on some of your previous posts, but I'll restate them here for others as well. Teknic follows these forums to respond to technical questions and to correct any factual errors or misleading information about our products and service (and many times other topics related to general motion control).

If you state your opinions about Teknic or our products without any incorrect or misleading/incomplete information, regardless of how negative your opinion is, we will rarely post a response. And, if you post a misstatement of fact that puts a Teknic product in a better light than it deserves to be based on the facts, we'll correct that too (you can see many examples of this on YouTube or various forums). Our goal is to help others understand motion control technology and apply it in a way that makes them successful. If this forum’s members (and the general market) are well-educated, they will choose the best products for their needs.

With that in mind, I have the following comments and corrections:



Teknic offers NEMA 23 brakes for $219, and a NEMA 34 brakes for $296 (single-piece pricing).





Teknic’s NEMA 23 and NEMA 34 ClearPath motors contain a 12,800 count optical encoder, and the NEMA 56/143/D100 models have a 64,000 count per rev optical encoder.

A few of my colleagues have made mention elsewhere on what makes for a high-quality encoder, but it bears repeating here:

It would be fair (although incomplete) to say that the Delta encoder has more resolution based on its specs, but it isn’t accurate to say that the Teknic encoder is subpar. As you stated, “a 10,000 count encoder will be perfect for CNC in every way.” Moreover, resolution is only one figure of merit for an encoder, and the number of counts per revolution does not convey any information about the other important metrics that characterize a high-quality encoder.

In order to get extremely high resolution out of a magnetic encoder (or even a high-quality optical encoder, for that matter) you have to do an extreme amount of interpolation. An encoder on a motor at the price point you mention will typically have noise, drift and other variability that makes the accuracy about 15 times (!) worse than the resolution. And if that isn’t enough, the filtering required to pull that fine a signal from the noise means that the servo gets its information delayed (and if it’s a serial encoder, the communication delays make it even worse). This all has a significant negative impact on servo performance.

Finally, even with a hypothetical “perfect” encoder, the mechanical imperfections in the motor and bearings will turn that extreme resolution into random noise, so even if you’re not bothered by the poor accuracy, the noise means your servo loop gets feedback that impairs its smoothness and dynamic performance.

We know all this because our engineers evaluated these encoders for ClearPath (and again recently for some yet unreleased products). It’s tempting to use these encoders because they’re so cheap, and because some people do mistakenly assume that more resolution must always be better. But we’ll be sticking with a high-quality, more expensive optical encoder and reasonable interpolation until there’s a magnetic encoder that can give us better performance.

Thanks,
Bridgette O. – Teknic Servo Systems Engineer

what kind of motor tech are teknic? pmsms?

[joeavaerage]

Hi,

what kind of motor tech are teknic? pmsms?

Yes, while the most common monniker applied is an 'AC servo', but in truth AC servos are permanent magnet synchronous motors. One distinguishing feature that AC servos have from some
other PMSM's is that they use sinusoidal modulation whereas many, say RC motors for example, use trapezoidal modulation. The control theory is called Field Oriented Control and was first published
by Hasse in 1968 and later by Blaschke in 1970. Many very good articles have been published since but I'd have to recommend some videos on the subject made by Texas Instruments. Great introduction
to a potentially complicated subject.

Through the 70's and 80's Field Oriented Control relied on analog techniques and devices such as resolvers and synchros, and were expensive and limited to the very best
of equipment. The military used the idea extensively in the control of missiles and rockets, and to hell with the cost. From the 90's however microcontrollers became cheap enough and powerful enough to
enact the algorithm and cheaper and reliable optical encoders were coming available and has all but taken over the market for servos, and at a price that mere mortals can afford.

Craig