Juki LU-562: Servo install, tune-up, and a few questions

[brongle]

As I mentioned a few weeks ago in my member gallery thread, I've been on a bit of a machine buying streak lately (which has not stopped, by the way - more on that soon), and picked up a LU-562 at a price I couldn't pass up. I've been going through it over the past week or so now that my servo is here, and the machine is under power now! 

Per usual, I can't leave well enough alone, and was immediately annoyed by the servo I bought. The included bushing that mounts to the balance wheel was total junk - visible runout between the OD and ID, crappy fit in the sensor bore. No idea why they decided to make this a casting, that's definitely slower to load and probably more expensive than just running them from solid bar...

Anyway, I made my own. The belt guard on my machine interfered with the needle positioner if I used the stock bushing, so I added a ~8mm long shoulder to space it out. 

Yes, I've already realized my mistake in mounting the needle positioner 😅 

Managed to get the bushing dialed in to about .005" of TIR and called it good enough. Drives me crazy when I see videos of needle positioners wobbling around on the end of a balance wheel... I'd probably add a journal to locate it more positively on the balance wheel if I make another for a future machine.

Now, this awful, terrible stupid servo drive. It's one of the "four button" units as described thoroughly here, but it's missing a bunch of the parameters people have documented. Some of the parameters listed in the manual seem to make no difference at all - notably, the needle down offset. It seems to be about as stripped down in terms of functionality as they come, which I suppose is what I get for buying the cheapest option on Amazon. The needle positioner works, kind of. About 1 out of 5 stops in needle down mode has the needle end up somewhere other than BDC, whether ahead or behind it, and when the sensor is timed so that it hits BDC, switching it to needle up mode or kicking the pedal back will not bring the needle all the way up to TDC. Instead it's about 4mm above the needle plate and has the take up arm fully down, and changing the offset parameter does not change this behavior. It does repeat to that position more reliably than it repeats to BDC, which is interesting. 

The included needle positioner lacks separate "up" and "down" outputs, and isn't even optical. Instead, it's got a plastic wheel with a magnet embedded in one of two holes (presumably the second is for a two position model), which it reads with a hall effect sensor. This has the unfortunate side effect of creating a ~15 degree deadband where it thinks it's in position. The PCB inside with the hall sensor has a pad for a separate down position output and LED, so that makes me wonder whether populating the missing passive components would allow me to add that functionality. I've read through several threads about these servos and NP sensors now, and it seems like NPs are a divisive subject around here. I absolutely love the one on my Durkopp-Adler 272, and the fact that it works so reliably with an electromechanical clutch motor and 90s era electronics leads me to believe that this is not a very difficult control systems problem. It also seems like there are a really wide range of visually identical servo units, each with their own undisclosed set of parameters exposed. If I can't get this one performing the way I want with the needle positioner, I'm going to grab one of the alternate 4 button designs (square grid of buttons, not a plus shape), which seem to have more consistent parameters with quite a lot more exposed, including PID constants. 

The other solution I've seen mentioned is decreasing the motor pulley diameter. I imagine the increased torque helps the servo more authoritatively position the needle, particularly when I can't change any of the parameters that would help it work better. So, I made one today, while the lathe was still set up for aluminum:

(Pardon the black streaks on the front there, I had some chip wrapping happen on the sub side boring ops - I promise that counterbore doesn't feel as rough as it looks there 😅)

I went with a 40mm pulley here, and didn't key it yet. Gonna see if the tapered shaft mount is sufficient to transmit the torque involved here, and I suspect it should be. Worst case, I'll pop it off and key it if it becomes a problem. Of course, my belt is now about 5" too long, and the tensioner was already at the end of its travel to begin with, so further testing will have to wait until a shorter belt arrives in a day or two. 

 

As for tuning it up, I have mostly been working on the reverse mechanism. As is typical, the forward and reverse stitch lengths started out very different. However, when I got to the timestamped point in Uwe's video below, I found that the arm he adjusts is cross-pinned through the shaft on mine! Short of extracting that pin, which seems very unpleasant, I'm not really sure how to proceed, as the other arm on the same shaft uses a setscrew that clamps in a straight groove, so it also can't be adjusted. In the meantime, I modified the bottom side reverse endstop bracket that Uwe removes entirely in that same video to allow a little extra travel while keeping the forward and reverse lengths pretty close. That's better, but I run out of adjustment at 6 stitches per inch, and it seems like every other 562 I've found online can go up to at least 5, if not even longer lengths. With both limiter brackets removed, the stitch length wound all the way out, and the reverse lever sitting right at the point before the mechanism binds up, I'm able to squeeze out roughly 4.5 spi, but it was very unreliable and the stitch length changed if the reverse lever drooped at all.

I've seen "floating" reverse levers mentioned on here before, and I got a little bit of that behavior with the limiter brackets removed entirely. It feels like the lever relies on inertia to make it to the very top of its travel, and if you let off too gently, it'll hang up maybe 3mm from the top and yield a different stitch length. Sometimes, it'll droop down on its own, usually when set to longer stitch lengths, gradually decreasing as you sew. At first, I thought it was just a sticky shaft somewhere - everything was a little sticky inside from years of oil turned to varnish with little care paid to it, but kerosene made quick work of that and it seems like things are moving much more freely now. Even so, the reverse lever doesn't always return to the same point unless I deliberately stop it before the end of its full travel with one or both of the limiter brackets. 

Have any of you seen other 562s or variants with that arm pinned to the shaft? I'll grab pics of mine tomorrow.

Has anyone else had a reverse lever that doesn't fully return like mine? 

 

 

 

[friquant]

15 hours ago, brongle said:

when I got to the timestamped point in Uwe's video below, I found that the arm he adjusts is cross-pinned through the shaft on mine!

Here's at least one reference to removing that pin: https://leatherworker.net/forum/topic/118024-cobra-class-26-forward-and-reverse-stitch-lengths-adjustment-questions/#findComment-748568

Also 

https://leatherworker.net/forum/topic/95183-matching-forward-and-reverse-on-juki-lu-563/#findComment-647927

and

https://leatherworker.net/forum/topic/114613-machine-recommendation-for-best-fwdrev-stitch-matching/#findComment-742608

 

There's another one out there too but I'm not finding it at the moment.

 

15 hours ago, brongle said:

it seems like NPs are a divisive subject around here

For me it depends on how slow the machine can go. With minimum machine speed 2 rev/sec I loved the needle positioner. With minimum machine speed 0.5 rev/sec I can stop the needle pretty close to where I want it by timing the release of the treadle, so I go without the positioner.

15 hours ago, brongle said:

my belt is now about 5" too long, and the tensioner was already at the end of its travel to begin with, so further testing will have to wait until a shorter belt arrives in a day or two. 

polyurethane belts have the convenience that you can resize them. 10mm round polyurethane fits my machine that used a 3/8" V-belt before.

[MikeG]

I tapped the pin out on my LU-563.  It has been working fine this way.  

[brongle]

3 hours ago, friquant said:

Here's at least one reference to removing that pin: https://leatherworker.net/forum/topic/118024-cobra-class-26-forward-and-reverse-stitch-lengths-adjustment-questions/#findComment-748568

Also 

https://leatherworker.net/forum/topic/95183-matching-forward-and-reverse-on-juki-lu-563/#findComment-647927

and

https://leatherworker.net/forum/topic/114613-machine-recommendation-for-best-fwdrev-stitch-matching/#findComment-742608

 

46 minutes ago, MikeG said:

I tapped the pin out on my LU-563.  It has been working fine this way.  

Glad to hear it - I'll see if I can get that out. 

 

3 hours ago, friquant said:

polyurethane belts have the convenience that you can resize them. 10mm round polyurethane fits my machine that used a 3/8" V-belt before.

Totally forgot that those were an option, thanks for the rec! I just ordered some, we'll see whether it arrives before the V belt. 

[friquant]

20 hours ago, brongle said:

If I can't get this one performing the way I want with the needle positioner, I'm going to grab one of the alternate 4 button designs (square grid of buttons, not a plus shape), which seem to have more consistent parameters with quite a lot more exposed, including PID constants. 

I've got a digital servo motor that exposes PID constants. I call it the "ALZILLA 100 RPM MOTOR" because @AlZilla recommended it to me, and because the label on the side indicates that it will go down to 100 rpm. 

At its lowest speed (mine actually measures at 125rpm, not 100) it has a lot of jitter. That is, a lot of speed up / slow down from frame to frame. The PID constants are there for experimenting, but I didn't find a setting that improved the jitter at the lowest speed setting. So the belt shakes at lowest speed but smooths out once it gets off of idle. I still prefer it over other low-cost digital servo motors I've used (apparently low speed is more important to me than jitter-free.)

@Gymnast has shed some insight into the challenge of building an inexpensive digital servo controller that can go very slow in this post.

 

[brongle]

8 minutes ago, friquant said:

I've got a digital servo motor that exposes PID constants. I call it the "ALZILLA 100 RPM MOTOR" because @AlZilla recommended it to me, and because the label on the side indicates that it will go down to 100 rpm. 

At its lowest speed (mine actually measures at 125rpm, not 100) it has a lot of jitter. That is, a lot of speed up / slow down from frame to frame. The PID constants are there for experimenting, but I didn't find a setting that improved the jitter at the lowest speed setting. So the belt shakes at lowest speed but smooths out once it gets off of idle. I still prefer it over other low-cost digital servo motors I've used (apparently low speed is more important to me than jitter-free.)

@Gymnast has shed some insight into the challenge of building an inexpensive digital servo controller that can go very slow in this post.

 

Ah! That's the exact ebay listing I found, funny. Glad you mentioned the low speed jitter, because I was about to pull the trigger on it - that's definitely not going to satisfy me. On to plan C!

I found some schematics of various flavors of these drives that folks have shared on other forums, along with photos of the PCBs, and it's a very primitive design as far as 3 phase BLDC drives go. That makes sense, given the price point these are sold at, but thanks to the ever-dropping cost of electronics, you can now buy an industrial 600W AC servo with the drive and an integrated 17 bit encoder for about $150 on Amazon. That gives you 2^17 counts per revolution on the motor, which is actually pretty low for the intended application of these servos (machine tools, mostly), but is far more than we could ever make use of here. 

The servo drives we use on sewing machines are sort of open-loop in the sense that they don't have their own encoder built into the motor, so there is no possibility of repeatable position control without additional external sensors. They instead use hall effect sensors to measure the magnetic field as it rotates with the armature in order to produce more torque at lower RPMs as a result of more accurate rotor position estimation. Once the motor is up to speed, it's common for these drives to stop using the hall sensors because it's just as easy to measure back EMF in the form of each phase's current in order to calculate RPM and keep the rotor synced with the magnetic field. This is how most R/C BLDC motors are commutated as well. Small motors for drones and airplanes are entirely open-loop and do a little "startup dance" every time they start from 0 RPM, which helps the ESC find the rotor's position when no back EMF is present. You'll see some, mostly for R/C cars, which include hall sensors just like on our servos for that extra torque at low speed and more predictable startup control. Those brushless ESCs are dramatically more capable than the drives that are included with these servo systems at this point, and they cost about $30. Sadly, they can't easily be directly adapted to the motors we use, but it shows that the technology exists at a low cost at the very least. Some of these tiny R/C motors can fit in a closed fist and output multiple times more mechanical power than any commercially available sewing servo - hell, I have a few custom wound 2207 outrunners (22mm dia x 7mm thick stator) that can output about 1kW if you can keep them cool.

Anyway, I think I'm going to run my current setup as-is until I get one of those industrial servos I mentioned above. Those are not a plug and play solution, so it's going to involve quite a lot of integration work on my end. The drives they include typically communicate over either RS485 or EtherCAT (on nicer units), so a separate controller will be necessary. It appears that the drive can accept a +/-10V analog input for simple speed or torque mode control as well, but building a custom foot pedal sensor that can provide that output is about as much work as just building a full featured controller. I could see that as an option for folks who don't care about needle positioning and just want smooth, reliable low speed control, though.

The good news is that the servo itself is almost all of the expense here, and the rest of the system, including a control panel along the lines of an Efka V860, can be put together for maybe $50 in parts these days. This system should allow me to maintain smooth operation and full torque all the way down to 0 RPM with very accurate needle positioning, fast acceleration and deceleration, exponential throttle control, etc. Aside from the servo kit and my controller, it'll need a foot pedal sensor of some kind and another encoder for the needle position, if the machine isn't driven via a timing belt. I might add an actuator for the reverse lever as well, for auto backtacking and more complex programmed stitch patterns. 

Is this all overkill? Yeah, definitely. I like my equipment to do exactly what I want when I use it, though, and that's exactly why I got into industrial sewing machines in the first place. Domestic machines didn't do what I wanted reliably enough and I was fighting the machine more than I was able to actually get any work done with it. I also don't rely on these machines for my primary business operations, so I have the luxury of being able to tinker. If this goes well, I'd like to duplicate the setup across all of my machines so that they all share a common interface and as much of the same functionality as possible. 

[AlZilla]

56 minutes ago, brongle said:

Ah! That's the exact ebay listing I found, funny. Glad you mentioned the low speed jitter, because I was about to pull the trigger on it - that's definitely not going to satisfy me. On to plan C!

Hang on a second. If you're using  speed reducer, you might never run the motor at 100rpm. My setup at 100rpm produces 11 stitches/minute. 250 to 300rpms produces 25 to 35 spm. Plenty slow, but I can still drop down to super slow if I'm in a tight spot.

I'll have to pay attention next time. I'm not sure I'm getting jitter ...

[brongle]

15 hours ago, AlZilla said:

Hang on a second. If you're using  speed reducer, you might never run the motor at 100rpm. My setup at 100rpm produces 11 stitches/minute. 250 to 300rpms produces 25 to 35 spm. Plenty slow, but I can still drop down to super slow if I'm in a tight spot.

I'll have to pay attention next time. I'm not sure I'm getting jitter ...

I'm not planning on using a speed reducer beyond the smaller motor pulley that I machined. I don't plan on sewing super heavy leather, and I do like having some top end speed available for longer straight seams, particularly when I'm working with textiles (most of the time for me). I also have 7 more machines en route to me right now, 4 of which will require motor retrofits, so it seems worthwhile to figure out a system that solves the laundry list of small gripes I've got with the current servo setup. Short pedal travel, the needle position issues above, low speed operation, control panel located all the way at the back of the table, and jerky/unpredictable behavior when starting/stopping are all issues that can be solved, with some compromises, while sticking with the standard servo units, but I think I'm going to end up pretty unhappy if I have to duplicate those compromises across half a dozen machines or so. 

What I'm proposing is not really all that different from @friquant's 3 phase gearmotor solution on his blog - my servo system just replaces the gearmotor and VFD, while being capable of far more torque output at low speeds than a typical sewing servo thanks to the integrated encoder and modern control scheme. It actually costs much less than the gearmotor solution too, and while it can't quite match the torque you get from a gear reduction, the ~6Nm peak torque output should be more than sufficient for my needs, particularly with a ~2:1 final drive ratio via a small motor pulley. 

The other reason I'm interested in this system is that it opens up a lot of options for interesting software-defined machine behaviors. One that I'm particularly interested in is a sort of "digital treadle," where the needle follows the foot pedal position in real time, sort of like Uwe's project here: 

I do like the thumbwheel control too, and I may add something similar to my implementation. I definitely want a button panel on the head like Efka includes with their drive systems. I'm not interested in using a stepper for most of the reasons he described in that thread (noise being a big one), and fortunately, industrial automation parts keep dropping in price as their scale of production increases. At the time he shared that project, the type of servo system I'm integrating would've cost 5-10x as much as it does today. If you're buying more than a couple of them at once, they come in closer to $100 each if you order direct from the manufacturer, even after shipping and duties. The software ends up being a little simpler than stepper control, as everything is handled through the Modbus protocol over RS-485, for which there are plenty of existing libraries available for most modern microcontroller platforms.

Originally, I'd wanted to build something like this for my Chandsew back when I first picked it up, using an old Clearpath servo I had laying around, but that machine was a backburner project since I needed to build a table for it first. Now that I have a complete machine with table that needs a good drive system installed before I bring it home, this has become a higher priority, and I've decided to use a more appropriate, somewhat more powerful, and much less expensive servo selected for this specific application instead of spare parts.

 

21 hours ago, brongle said:

Glad to hear it - I'll see if I can get that out. 

I did get the pin out by the way - much easier than I'd feared. Glad I have a bunch of surplus/worn out gage pins floating around, they make great drift punches for small pins. I also found out that the upper shoulder screw pivot for the reverse lever linkage had backed itself out far enough that the linkage bar was no longer constrained at all, which gave me a few mm of play in the lever. Kind of a pain to get that return spring hooked back up after fixing it, but it's back together now, and with no play at all! Seems to have solved a lot of my reverse issues already, including the stickiness as it returns to the forward position, and I'll see about balancing the stitch lengths between forward and reverse this evening. I added a small dab of blue loctite to hopefully prevent this from happening again, being careful to keep it off of the linkage surfaces.

Before and after: 

[AlZilla]

4 hours ago, brongle said:

seems worthwhile to figure out a system that solves the laundry list of small gripes I've got

Absolutely. Most of the time these are one off situations. 

[friquant]

6 hours ago, brongle said:

What I'm proposing is not really all that different from @friquant's 3 phase gearmotor solution on his blog - my servo system just replaces the gearmotor and VFD, while being capable of far more torque output at low speeds than a typical sewing servo thanks to the integrated encoder and modern control scheme. It actually costs much less than the gearmotor solution too, and while it can't quite match the torque you get from a gear reduction, the ~6Nm peak torque output should be more than sufficient for my needs, particularly with a ~2:1 final drive ratio via a small motor pulley. 

Your AC servo project sounds quite exciting 😀

The downside of VFD + 3-phase AC induction gearmotor, in my opinion, is finding a low-cost 3-phase 4-pole AC induction gearmotor that is readily available. To do a VFD again for a machine for sewing textiles (which is most of what I sew) I'd aim for a plain (non geared) 3-phase 4-pole AC induction motor which would bring down both the price and the noise. (My gearmotors make some noise, suspect it's old bearings.)

6 hours ago, brongle said:

the ~6Nm peak torque output should be more than sufficient for my needs, particularly with a ~2:1 final drive ratio via a small motor pulley. 

What is the stall torque of a typical digital servo motor? I've measured stall torque on some domestic machines but haven't measured anything bigger.

What do you think about a direct drive option for your AC servo? You wouldn't need a second encoder for the machine head to do needle positioning sorts of things.