[slow-poke]

My control theory (PID loops etc.) is long forgotten it was 40 years ago. My understanding is that ac servos are a bit more finicky to tune than dc servos partially because they have wider bandwidth (they maintain their torque at higher speeds better than dc servos)

Hopefully this thread can help guide future ac servo tuners. I realize there are a lot of clearpath fans out there (they keep reminding me of how effortless it was to get their servos tuned), that's great but is probably not relevant to this discussion.

Application is CNC conversion of a 8x30 knee mill with 5mm pitch ball screws installed (16mm for X axis and 20mm for Y axis).
Rough guess on weight is:
X: 125lbs unloaded and say max of 200lbs loaded
Y: 175lbs unloaded and say max of 250lbs loaded

Everything is mounted, connected and working with very preliminary tune. I'm focusing on fine tuning the X-axis at the moment

I'm using StepperOnline T6 750W ac servos with a direct drive (1:1) configuration using compliant couplers. With very preliminary tuning they move the table silently and effortlessly at 40-80 ipm, I will likely limit rapid to 40 or 50 ipm fast enough for me.

The only setting I have tweaked so far is Inertia ratio, the default setting was 200 and this caused the move to oscillate at the destination, I kept tweaking the value up and the end point oscillation kept diminishing. with the value set to 800 it seems quite stable now.

The next critical setting is stiffness, and after that there are about a dozen other fine tuning parameters.

If you have any experience to share, I'm all ears and eyes and if you have any good links on the subject please pass them on. If someone has actually tuned either the Leadshine or StepperOnline servos that would be most helpful

Thanks in advance

[slow-poke]

Response curve as is:

[joeavaerage]

Hi,

My control theory (PID loops etc.) is long forgotten it was 40 years ago. My understanding is that ac servos are a bit more finicky to tune than dc servos partially because they have wider bandwidth (they maintain their torque at higher speeds better than dc servos)

I learned my control theory about the same time as you, and at that time is was all DC servos, with resolvers, synchros, tachos and all that fun stuff.
If you've not played with a modern AC servo then you are in for a treat, they are far better than the DC servos of yesteryear. They are not more finicky than DC servos
and they can be tuned to much better performance than DC servos due to that extra bandwidth.

First thing to note is that most of the better quality servos, other than the cheapest Chinese made ones, have setup and tuning software, which is a true Godsend. They also have Auto-tune which for inertia ratios
'in the zone' is very good indeed. If Autotune works well, and it usually does then you'd be very VERY hard pressed to manually tune it any better. Beyond the 'zone' then manual tuning comes into its own.
Most of them have notch filters which are distinctly uncommon in DC servo PID loops. Some, the cheaper Chinese ones typically, have just one notch filter, while many have two filters and the very best may have
more. These notch filters are just the ticket when you need to tune a servo with a 20:1 inertia ratio.

For inertia ratios of 10:1 and less just let Autotune do its thing. Its fun to experiment with the tuning afterwards but I doubt you'll get any improvement.

The first thing you need to do it calculate the inertia ratio. The inertia ratio is a reflection of the machine, you can guess at it if you wish, but then you might expect Autotune to fail.
Calculate the actual ratio and use that and that alone.

The inertia ratio is the ratio of the first moment of inertia of the armature of the servo to the effective first moment of the axis mass and ballscrew. You might be of the opinion that the ballscrew inertia is insignificant,
and in some cases you can neglect it but in others, and I will show you an example the ballscrew inertia dominates, and can never be neglected.

The inertia of the armature is published by the manufacturer. My 750W Delta B2 series servo has 1.16 x10^-4kg.m² first moment.
I use 32mm diameter ballscrews of 700mm length with 5.8 x 10^-4kg.m²
The pitch of the screws is 5mm and I have an assumed axis mass of 150kg for 0.94 x 10^-4kg.m²

The total effective first moment of inertia is 7.9 x 10^-4kg.m²

The percentage make up is highly illustrative:
12% linear momentum of 150kg of cast iron, vice and workpiece
15% is the rotational momentum of the servo armature
73% is the rotational inertia of the ballscrew.

You read that correctly, the inertia of a 150kg axis is only 11% of the total inertia, and the the inertia is dominated by the ballscrew.
The ballscrew inertia is extremely sensitive to diameter, to the fourth power. In your case with only 16mm and 20mm ballscrews then I expect the contribution to the
overall inertia to be low, but it MUST be calculated or you risk 'going up the garden path'.

Just to round out my example:
Inertia Ratio = (5.8 + 0.94)/1.16
=5.8

As a consequence I used Delta's Autotune and found it to be at least as good as I could do manually.

Lets try calculating the inertia numbers for your machine, starting with the Y axis as that has the bigger diameter screw. You don't mention a length, so I'll guess at 1m. The calculation
is linearly dependent on length, so relative to other factors a guess at length introduces less error than either screw diameter or pitch

J_ballscrew= 0.5( pi. r².l .r². rho) where r is the radius of the screw and rho is the density of steel, 8000kg/m³ approx and l is the length of the screw

J_ballscrew= 1.26 x 10^-4kg.m²

J_linear= (m. p²)/4.pi² where m is the linear axis mass,and p is the pitch of the screw
J_linear=0.625 x 10^-4kg.m²

I guess your servos being similar in size to mine will be 1 x 10^-4kg.m²

total effective first moment is 2.885 x 10^-4kg.m²

As expected with the smaller diameter ballscrew (20mm) the contribution to the overall momentum is reduced, but still significant at 44%. The linear load is 22%,
and the armature rounds it out at 34%.

The inertia ratio is 1.885:1. This should be 'quids in' for Autotune.

I have no idea how you get an inertia ratio of 200. Do you mean to tell me that the ballscrew and linear axis have 200 times the momentum of the servo armature??

Craig

[WesM]

Have you tried doing the auto tune for inertia and stiffness in motion studio? The quick start PDF on stepper online has a basic procedure for this written out. Curious what the results were. I will be going through this myself pretty soon with the same servos, so I will be following this thread with interest.

[slow-poke]

Have you tried doing the auto tune for inertia and stiffness in motion studio? The quick start PDF on stepper online has a basic procedure for this written out. Curious what the results were. I will be going through this myself pretty soon with the same servos, so I will be following this thread with interest.

Perhaps I'm getting old and slow but I find the documentation for MotionStudio to appear at first glance to be simple and straightforward, however when actually using it it's not clear what the proper calibration sequence is from start to end and at what point I should be using autotune. My gut feeling is that MotionStudio in the right hands can quickly allow yo to tune your servos because you have access to all the control settings, and can easily generate response graphs etc. I was able to within minutes and simple fudging my way through get the X axis to move initially slowly and in an unstable way to very quickly without any oscillations. Even though the current settings are in all likelihood out to lunch the X axis moves would be totally acceptable (for me) except the last segment of the move where the servo slows down to a crawl as it reaches the target destination. From my limited knowledge I'm interpreting this as too little "I" in the PID loop or not enough "P" I'm going to take a step back and focus on the inertia ratio. joeavaerage has posted a really good response and I'm going to respond to that next.

[slow-poke]

Hi,

The first thing you need to do it calculate the inertia ratio.

I have no idea how you get an inertia ratio of 200. Do you mean to tell me that the ballscrew and linear axis have 200 times the momentum of the servo armature??

Craig

Craig,

Your thoughtful response is much appreciated. The inertia ratio was just a pie in the sky value that I entered to see what would happen. Picking up from your example, I’m going to focus on the X axis for the moment.(16mm diameter, 5mm pitch, 1.04m long)

J_ballscrew= 0.5( pi. r².l .r². rho) where r is the radius of the screw and rho is the density of steel, 8000kg/m³ approx and l is the length of the screw
J_ballscrew(x) = 0.5( pi * 0.008² * 1.04 * 0.008² * 8000) = 107.061 x 10^-6kg.m²

I'm missing something with the J_linear calculation, I will use your 150kg for now, and p = 5/1000. Please explain where I'm going wrong...

J_linear= (m. p²)/4.pi² where m is the linear axis mass, and p is the pitch of the screw
J_linear= (150 * 0.005²) / (4 * 3.14² ) = 94.99 x 10^-6kg.m^{2 ????}

My T6 inertia of the armature as published by the manufacturer. 1.5 x10^-4kg.m²

So total effective first moment is: 107.061 x 10^-6 + 94.99 x 10^-6 = 202 x 10^-6kg.m² (if I use my J_linear)
or : 107.061 x 10^-6 + =0.625 x 10^-4= 169.6 x 10^-6kg.m² (if I use your 150kg Jlinear)

Resulting in a inertia ratio of: 1.347 (if I use my Jlinear)
or: 1.13 (if I use your 150kg Jlinear)

So to your point miles away from my randomly picked 200 value.

I will try again today with an inertia ratio closer to 1.2 and see what happens.

One of the settings in MotionStudio that had a really significant impact was the "Load" setting, the default was Rigid Structure and I later switched to High Inertia, The result was the response went from underdamped to overdamped, so with the much lower inertia ratio I'm guessing I will need to go back to Rigid Structure for load type.

Jeff

[joeavaerage]

Hi,
I calculate the same as you.

My preference is to convert all inertia numbers to 10^-4kg.m². I find that if I do so then a clearer picture emerges

J_armature= 1.5 (units)
J_ballscrew=1.07 (units)
J_linear=0.94 (units)


Total =3.51 (units)
J_armature=43%
J_ballscrew=30%
J_linear=27%

The important takeaway is that the momentum of the axis, despite being heavy at 150kg is moving slowly and is still less than 1/3 the total momentum. The realisation is that while the ballscrew and armature are
light they are spinning and hundreds if not thousands of rpm, and its that speed that determines that they should have that much momentum.

The inertia ratio is (1.07 +0.94)/1.5= 1.32:1

Just as an aside I derived the expression for the equivalent of first moment for a linear axis by the following argument.

The kinetic energy of a mass m moving at velocity v is E=1/2 m.v²
Now imagine some rotating body of first moment J_effective rotating at w rad/s. Its kinetic energy is E=1/2 J_effective.w²

Equating the two:
1/2 m.v²=1/2 J_effective.w²

But we know also that the velocity of the linear axis is related to the angular velocity of the ballscrew by v= w.p/(2.pi) where p is the pitch.

Simplifying and solving for J_effective:
J_effective= (m. p²)/40 where (2.pi)²=40 and m is the linear axis mass.

This allows you to quantify the momentum as a result of the linear axis AND the rotating components. First little bit of primary physics I've done in years.

Craig

[slow-poke]

Craig, Wes

Craig, thanks for the confirmation on J_linear. Your point about the inertia contribution makes complete sense. I vaguely remember doing these calculations in first year (1980's)

I'm happy to report that the MotionStudio auto tune appears to work quite well. I have the X-axis moving quickly, near silently and every bit a good as I could hope for.

I was somewhat concerned that the MS software might be flaky or possibly useless being a beta version. I initially had an issue because none of my laptops have a DB9 serial port and the supplied tune cable is a RJ45 yo DB9 serial. Initially I tried a USB to serial converter cable and it would connect but then stop communicating after a click or two and often crash the program. Initial assumption was the software was flaky as hell. I changed the USB to serial converter cable to a brand name one (original was an aliExpress probably with counterfeit FTDI chip) and it now connects and communicates flawlessly. I swear those blue ended transparent over shielded cable converters are not worth 2c for all the trouble they have caused.

With the communication issues out of the way, and a starting point for the inertia ratio the auto tune page seems to work quite well, with a very simple one button tune that you can increment or decrement, besides that It shows your current settings beside the recommended settings with a button to try the new settings or revert back to the old. A few clicks and runs later the servo was behaving very nicely. From the graph of the response, it appears to be slightly over-damped but I'm going to live with it because its plenty responsive enough for my needs I can always tweak in the future if I feel the need.

Thanks again for your help.

[WesM]

Thats excellent! yeah one of the youtube guys (clough42) had issues with the db9 converter as well. For anyone reading this thread in the future, stepperonline sells a DB9-USB converter thats compatible with the leadshine/T6 software. stepperonline mentions it in their listing, Its cheap and worth getting along with the servos.

https://www.omc-stepperonline.com/us...converter-c006

Glad to hear you figured out motion studio, I was getting a little nervous that I would have to go through a complex tuning procedure if autotune was not up to the task.

[slow-poke]

Thats excellent! yeah one of the youtube guys (clough42) had issues with the db9 converter as well. For anyone reading this thread in the future, stepperonline sells a DB9-USB converter thats compatible with the leadshine/T6 software. stepperonline mentions it in their listing, Its cheap and worth getting along with the servos.

https://www.omc-stepperonline.com/us...converter-c006

Glad to hear you figured out motion studio, I was getting a little nervous that I would have to go through a complex tuning procedure if autotune was not up to the task.

I would actually enjoy digging into manual tune for educational purposes if I could get useful answers to my questions from tech support. So far the responses have been polite but rarely answer the actual question asked.

[WesM]

I would actually enjoy digging into manual tune for educational purposes if I could get useful answers to my questions from tech support. So far the responses have been polite but rarely answer the actual question asked.

When I was researching the T6 servos, the Leadshine EL6 servos/drives looked identical, they seem to have updated the EL6 drives since then though. But I had started just asking leadshine questions about the EL6 servos rather then going through stepperonline for technical questions. My impression was much like yours, stepperonline was polite, but could not (or took a very long time) answer my questions. When I went direct through leadshine however, they were able to answer some of my more technical questions. Since motion studio is a leadshine software product, maybe you could get better responses going directly through them?

[joeavaerage]

Hi,
Steppersonline are a seller, not an engineering company. to get technical support you need to talk to the people who designed an build the things....Leadshine.

The good news is that setup and tuning software is very similar across different brands, if you can use one, then excepting different terminology then you can use another.

When I learned system and control theory all those years ago we had a lab jammed to the gills with gear, servos, synchros, resolvers, chart recorders, analog storage scopes, and the list goes on.
To characterise and tune a servo system was quite an undertaking. All of this had to be documented, you had to do all the tedious and drawn out calculations by hand and submit the whole shebang to the
professor of the department for marking, and he was tough, you might even say a martinet!

Modern AC servos setup and tuning software is a breeze by comparison. All the calculation is done for you and displayed on a PC screen as a 'scope'. There is no cause to whinge here, you have to roll up your
sleeves and learn the software, just like ANY AND ALL software you have to get familiar with it......before you can use it to advantage.

What I found that the little dual core Atom single board PC that runs my machine was not a good PC for tuning. It does a fine job of setting up and saving firmware to the drive but it did not have the sort
of bandwidth to do Autotuning nor use the Tuningscope. For that purpose I set up another (i5) PC along side my machine which does have the power to Autotune and Tuningscope.

Craig

[slow-poke]

When I was researching the T6 servos, the Leadshine EL6 servos/drives looked identical, they seem to have updated the EL6 drives since then though. But I had started just asking leadshine questions about the EL6 servos rather then going through stepperonline for technical questions. My impression was much like yours, stepperonline was polite, but could not (or took a very long time) answer my questions. When I went direct through leadshine however, they were able to answer some of my more technical questions. Since motion studio is a leadshine software product, maybe you could get better responses going directly through them?

I'm not sure what's up with the leadshine website, but when I tried to access to download the EL6 manual etc. MalwareBytes warns me that it's not safe and is attempting to place a trojan on my PC. I enquired via email to Leashine requesting the datasheet be sent via email noting the trojan warning and including an image of the warning I was receiving. Leadshine response was that unlike their competitors they provide excellent tech support from the USA and then included a hyperlink to the same address that MalwareBytes is warning me to NOT go to, and no comment about the trojan.

Hmmmm... your house is on fire, so you call the fire department and get an automated message telling you to call the fire department if your house is on fire. LOL

From my really foggy memory of PID tuning,( warning this may be incorrect) your supposed to temporarily disable the I and possibly the D and introduce a step change and the response should be a damped sinewave with only three zero crosses. If you have more than three the system is under damped and if less than three it's over damped. More P gets you there faster with the risk of excessive overshoot and instability. The D counters the overshoot, and the I contributes to the final settled value.

Thinking about this if the PID system was for say velocity overshoot might be okay, obviously for a positioning system overshoot would only be okay if the overshoot was completely damped out before the destination point is reached. I need a PID refresher crash course.

[joeavaerage]

Hi,
I have had some software that triggers that kind of response from Firewall/AntiVirus software. I ignore it if I have confidence in the software.
I usually regard the Firewall/AntiVirus software as malware itself. Its forever updating itself, interrogating your PC, slowing it down or outright preventing
you from doing what it is you want to do. A PITA.

I presume if you are going to pay many hundreds or even thousands of dollars for Leadshine servos that you have faith in the company and by extension the setup and tuning software.

The mathematical best is a damping ratio of 0.7071 which results in the modest overshoot and with but a few cycles of ringing. This results in the best RMS error....but that may not be suitable for CNC.

For say a a missile aileron if it were to overshoot a little....that's not a problem provided it settles to its commanded position as quickly as possible. With a CNC machine that overshoot may in fact remove material
from the stock which cannot ever be returned to the stock and therefore you may be best off demanding a dynamic solution that has NO overshoot but perforce accept the somewhat longer time to achieve its commanded
position. The difference between the two dynamic solutions is very close, say 0.7071 verses 0.95.......and would be hard to see in practice. In the case of a CNC machine with a slightly overdamped dynamic if the
toolpath is slowed a little then the extra time taken to settle to commanded position is irrelevant. With a missile control you require the fastest settling which corresponds to the highest possible bandwidth but with a CNC
you trade bandwidth, or equivalently machining speed, for accuracy. More often than not the machining speed is limited by the spindle/tool/cutting forces rather than the motion ergo you tune for a mildly overdamped motion
dynamic for best accuracy knowing full well that the machine motion will be in almost all instances slower than the motion dynamic anyway.

Servos are fun to play with, but unless you have an inertia ratio well outside the norm, you can spend a lot of time fiddling only to find either no improvement, or worse a degradation of response from
Autotune.

Craig

[RaderSidetrack]

I'm not sure what's up with the leadshine website, but when I tried to access to download the EL6 manual etc. MalwareBytes warns me that it's not safe and is attempting to place a trojan on my PC.

My guess is that MalwareBytes is objecting because the Leadshine page was using "http:" rather than the newer "https:". For instance, see tthe attached screenshop.

https://www.cnczone.com/forums/attac...d=500782&stc=1

In that screenshot, note that in the first listing, the URL/address starts with "https:",and the second listing is "http:". There isn't really anything wrong with "http", but it is an older protocol and these days is considered less secure. But its not like you are trying to do say, banking transactions, on that page -- you are simply trying to download a file that the site is offering.

Virus scanners like to make a big deal about "http", because that makes you think they are great stuff!
But if you are not engaging in something like a financial transaction, and the site is otherwise 'normal', I would not be too concerned and just download the PDF anyway.

[KlausDeno]

If you're considering giving jeans as a gift, you might be interested in men's selvedge jeans. They're well-known for their top-notch quality and unique self-finished edges that prevent fraying. My buddies love them for their durability and the personalized wear patterns. They might be a tad costlier than regular jeans, but you're paying for the superior craftsmanship and fabric. Maybe give it a shot, here is where I found them https://jeans4you.shop/collections/mens-selvedge-jeans. Hope it helps.

[NeeqOne]

Has anyone successfully tuned their servo fully using the quick start guide. It contains relevant information but it is not that useful when doing a complete tuning of the T6 servo. I have tuned the inertia ratio but still struggling to get the other parameters tuned.