[SCzEngrgGroup]

I received the new gear-motor for the ATC carousel yesterday, and today I took a few minutes to try it out. It looks fantastic! Exactly what I wanted - MORE than fast enough (faster than what I showed in the video last week), with MORE than enough torque (>7 ft-lbs). And it's REALLY quiet to boot!

First thing I had to do was graft an encoder onto the motor, as there was not even a provision for one. It turned out to be quite easy to add a 3.2mm back-shaft to the motor, and adapt a Hall Effect quadrature encoder. I tested it just holding it by hand, so the motion was "sloppy", but it looks very, very promising. With luck, I should be able to make the mounting adaptor tomorrow, and test it for real. I also need to make a new hub for the carousel, as the broached hex shaft bore on the current one is slightly crooked, so the motor wants to wobble as it turns.

Over the last few days, I've also gotten about half of the transfer arm drive assembly made, now have almost all of the parts on-hand, and the rest are in-transit due to arrive within a few days. The big thing I'm missing is the metal stock for the two largest parts. The metal finally shipped out yesterday, so I'm hoping it will show up on Monday or Tuesday. It won't take long at all to make the remaining parts once the metal shows up. The transfer arm will use the same gear-motor and encoder as the carousel.

I'm starting to get pretty jazzed about this ATC. So far, the development is going incredibly smoothly, and I'm really pleased with how well it's all working. Once the transfer arm drive is done, the remaining work is looking not at all daunting, and should go very quickly. I see no reason I shouldn't have a fully-functional prototype running on one of my machines by sometime in May at the latest.

Regards,
Ray L.

[SCzEngrgGroup]

Most of the transfer drive is done - just a couple of small parts to go, and I need to re-make one part, due to a minor dimensional error. I also have both servo motors mounted to the carousel.

Attachment 229980 Attachment 229982 Attachment 229984

With a little luck, I might just manage to get the bracketry and other bits done so I can show a nearly complete Torus Pro ATC at Cabin Fever!

Regards,
Ray L.

[rlockwood]

Seems like a whole lot of machined aluminum bits.. What's this thing supposed to cost when its done?

Sent from my Nexus 5 using Tapatalk

[SCzEngrgGroup]

Seems like a whole lot of machined aluminum bits.. What's this thing supposed to cost when its done?

Sent from my Nexus 5 using Tapatalk

Kinda hard to do a decent toolchanger without a fair number of parts. But they're all quite simple, and don't take terribly long to make. Price is not fixed yet, but it will be "competitive". There is no such thing as a "cheap" ATC.

Regards,
Ray L.

[hjl4]

[QUOTE] Quote Originally Posted by rlockwood View Post
Seems like a whole lot of machined aluminum bits.. What's this thing supposed to cost when its done?
/QUOTE]


It's a prototype. Everything is usually machined like that to prove the design. I say 90% of those parts can be cast in aluminum. We spin cast aluminum in moulds everyday.
This would result in an even more rigid and lighter design. And being that you have alot of experience with the 'Pro', making the moulds wouldn't be much of a problem either, and it would bring your costs down significantly making it affordable for the average hobbiest..

[rlockwood]

[QUOTE=hjl4;1463144]

Quote Originally Posted by rlockwood View Post
Seems like a whole lot of machined aluminum bits.. What's this thing supposed to cost when its done?
/QUOTE]


It's a prototype. Everything is usually machined like that to prove the design. I say 90% of those parts can be cast in aluminum. We spin cast aluminum in moulds everyday.
This would result in an even more rigid and lighter design. And being that you have alot of experience with the 'Pro', making the moulds wouldn't be much of a problem either, and it would bring your costs down significantly making it affordable for the average hobbiest..

Not to put any words in Rays mouth, but I'm 98% confident the intention is to produce the production parts through very similar methods to the prototypes, in much the same way the current PDB is being produced. I don't necessarily think there's anything wrong with that, just hopeful it doesn't result in poor margins or an overpriced ATC. I sincerely prefer the design as shown to anything else in the class, but sometimes simplicity is king..

Sent from my Nexus 5 using Tapatalk

[hjl4]

I agree.

[SCzEngrgGroup]

You guys are focusing on all the wrong things.... The fabricated parts represent a relatively small portion of the overall system cost. All of the fabricated parts are small, very simple, and don't take long at all to make, and can be easily fixtured for fast production in quantity. Casting would make no sense at all in the relatively small quantities I would expect for these, and the castings would still need to be machined to get the required precision for many parts.

It's all the actuators, sensors, and other electrical/electronic parts that account for the vast majority of the cost. But, those are also the things that will ensure make this ATC is very high performance, very reliable, robust, and expandable.

Regards,
Ray L.

[hjl4]

Wow. I think I will no longer post any more replies to these Novakon threads. Seems everybody knows better. FYI, Spin and die casting will hold .0005 tolerance all day provided you have rigid temperature controls. Try to beat that Ray. Electronic components are cheap, no one will pull the wool over my eyes on that one.. Motors bought in lots are also cheap. Now go ahead and tell me how I'm wrong. Your not the only engineer on these forums. As for small quantities? Thats a choice you made.

[keithmcelhinney]

Hey hjl4 I don't think Ray meant anything against you here. I am sure you are very knowledgeable about casting/etc and what you say is true. Maybe Ray either doesn't have the same knowledge or want to do a different process. The thing that I see here is that in casting you are putting a lot more of your costs on the front end. Maybe that has something to do with it. The small quantities is probably dead on as of right now. Maybe a couple of years down the road he could be selling a bunch but I wouldn't assume anything. I have dealt with Ray a bunch and I can tell you he is a straight up great guy.

-Keith

[SCzEngrgGroup]

Woo Hool! Just tried the transfer arm "lift" under power for the first time - it looks very promising. Plenty fast, plenty of "thrust", and it even has a really cool sound to it! Tomorrow I'll make up some rubber buffers to cushion it at the ends of travel, then I can do some real testing. I also hope tomorrow to get a high-resolution encoder mounted on the carousel drive motor. The Hall Effect encoders work ok, but because they are relatively low resolution (10 PPR), the servo "dithering" when stopped is rather annoying. Higher resolution should resolve that. Fortunately, some good encoders have recently gone done in price significantly, so it's not too much of a cost hit.

Regards,
Ray L.

[SCzEngrgGroup]

Since the rest of the hardware did not show up as expected today, I made more incremental progress on the firmware side. I pretty much completed implementation of the full -blown servo control loop, including acceleration ramping, which makes for much smoother operation. It all seems to be working nicely, but I'm at the point where the performance of the bodged Hall Effect encoders is limiting performance, so the next step will be to adapt the high-resolution capacitive encoders to the carousel and transfer arm gearmotors. That will no doubt necessitate adding a small CPLD or FPGA as a hardware encoder interface, as the poor little Arduino is just not quite up to handling the high interrupt rate the encoders will impose. Not a big deal, given that I've spent much of the last 30+ years designing chips, quite a few of which ran into seven digit gate counts. Designing a few thousand gates worth of simple logic will be a rather nostalgic exercise.

Regards,
Ray L.

[Hirudin]

Will the weight of the tools have much affect on the servo tuning? Will a carousel with 12 light-weight tools (or possibly with only a couple tools loaded) behave differently than another with 24 heavy tools? I only ask because I've heard servo damping is dependent on weight/load. Maybe a pot could be added so people could adjust the damping if they ever experience resonance.

PS
The progress is very fun to watch! I love the videos!

[SCzEngrgGroup]

Will the weight of the tools have much affect on the servo tuning? Will a carousel with 12 light-weight tools (or possibly with only a couple tools loaded) behave differently than another with 24 heavy tools? I only ask because I've heard servo damping is dependent on weight/load. Maybe a pot could be added so people could adjust the damping if they ever experience resonance.

PS
The progress is very fun to watch! I love the videos!

That should be completely unnecessary.

Regards,
Ray L.

[SCzEngrgGroup]

MUCH better! I got the high-res encoder on the carousel gearmotor, and things are working really nicely now. So far no issues with the higher resolution - the servo code seems to be perfectly happy despite the higher interrupt rate. A very quick tuning of the servo gets nice, "crisp" response, with fast settling to final position, and minimal over-shoot (just a few counts) The biggest issue will be preventing overshoot due to the high inertia of a fully-loaded carousel. To help with that. I've added separate acceleration and deceleration rates, so it gets up to speed very quickly, but is a bit slower coming to a stop. I will have to come up with a tighter coupling between the motor and carousel, as there is a little backlash in the current one, which wreaks havoc on the servo when it stops. But, overall, this is looking really good now. Here's a short video:

http://youtu.be/FGsd8K8k-dg

Regards,
Ray L.

[mmprestine]

I see problems arising with the servo sizing. Currently, without any empirical data, the servo is way undersized. I have been doing motion control on embedded industrial systems for 20 years and based on the what I see in the video your moment of inertia is only going to get worse once the system is loaded with tools. The inertia mismatch of the system has to be well over the 3:1 and that is no good, well at least if you want it to last. If you would be willing to share a few details I could run the calculations to give some examples of where it is and should be. Things like diameter (or radius/length of moment arm) of carousel, estimated weight of each arm segment with and with out tool, number of arm segments, gear ratio, motor rpm, and desired time of moves.
Matt

[SCzEngrgGroup]

I see problems arising with the servo sizing. Currently, without any empirical data, the servo is way undersized. I have been doing motion control on embedded industrial systems for 20 years and based on the what I see in the video your moment of inertia is only going to get worse once the system is loaded with tools. The inertia mismatch of the system has to be well over the 3:1 and that is no good, well at least if you want it to last. If you would be willing to share a few details I could run the calculations to give some examples of where it is and should be. Things like diameter (or radius/length of moment arm) of carousel, estimated weight of each arm segment with and with out tool, number of arm segments, gear ratio, motor rpm, and desired time of moves.
Matt

Perhaps you have in mind a level of performance beyond what I'm expecting. This is not a difficult application - speeds are relatively low, and even positional accuracy is not critical. It is now very consistently positioning quite nicely within about 2 counts in under 1/2 second. The relatively high inertia of the load is really only a problem when stopping, and for that I've added separately programmable acceleration and deceleration, so it stops nicely with almost no over-shoot. You may be mis-interpreting some of what you see in the video, as there is considerable backlash in the coupling between the motor and carousel, making it look like there's considerable overhsoot and oscillation at the end of each move. In fact now, with a better coupling that is all but gone. What little backlash remains, is there for a reason.

That said, I am still looking at alternative motors. I just received on today that looks promising, and is both cheaper, and in a much more desirable form-factor.

Regards,
Ray L.

[neilw20]

Woo Hool! Just tried the transfer arm "lift" under power for the first time - it looks very promising. Plenty fast, plenty of "thrust", and it even has a really cool sound to it! Tomorrow I'll make up some rubber buffers to cushion it at the ends of travel, then I can do some real testing. I also hope tomorrow to get a high-resolution encoder mounted on the carousel drive motor. The Hall Effect encoders work ok, but because they are relatively low resolution (10 PPR), the servo "dithering" when stopped is rather annoying. Higher resolution should resolve that. Fortunately, some good encoders have recently gone done in price significantly, so it's not too much of a cost hit.

Regards,
Ray L.

With the dithering, as you have full control of your servo system(s), after in position for a short time and no more movement, turn off the dithering.
You were mentioning you need for a little backlash, so it will most likely work in nicely.

Neil

[SCzEngrgGroup]

The latest candidate for the carousel drive motor is looking very promising. It's a little bit slower than the others, but still within acceptable limits. It is very compact, which makes for much nicer packaging. It is a worm drive, rather than planetary, which means effectively no back driving, which reduces the effects of load inertia. With backlash compensation enabled (the production code will have self-calibrating backlash compensation that is updated on each power-up), it seems to provide good positioning accuracy, even without the second "layer" of positioning (aligning directly to each tool receiver, rather than just to carousel position) enabled. And, it's nice and quiet.

Regards,
Ray L.

[SCzEngrgGroup]

Here is the latest candidate motor, which is looking very good. It's not quite as fast as some of the others, but it is close, and certainly well within an acceptable range (i.e. - it's not slow enough to significantly affect overall toolchange time). It's very smooth, fairly quiet, has gobs of torque, and is very nicely made. I just need to add a simple friction damper to damp out the oscillation that sometimes occurs as a result of the backlash - you can see this at one point in the video. Damping that out will also allow me to tune for better performance. Other than that one minor issue, I think this one is the new "leader of the pack".

http://youtu.be/XfFa9ITjKsc

Regards,
Ray L.