Limitations of old servos

Hi, Newbie to CNC here,

I have an old traditional BT40 CNC Bed Mill clone in reasonable mechanical condition with a defunct control I'm looking to replace. There are 3 working servos late 80's or early 90's vintage which would have to run in Velocity mode. While I understand there are limits on speeds and feeds on an old Way style mill, I wanted to ask opinions on the limitations of the old servos which have 5000 pluses per 5mm of axis movement which if I'm not mistaken equates to some where between 12 and 13 bit encoders. Max rpm is around 2000.

The servos are not really serviceable so that's a limit but what's the views in terms of machining quality like finish etc? Job speed is not a real issue for me. I would guess the servos can't be great if much higher bit systems are being built. So given the issues do I bite the bullet and upgrade with a clean system from day one rather than have them fail or change them due to poor performance.

Cheers

If the motors are in serviceable condition or just require a little care and new brushes, You could look at modern torque mode drives such as A-M-C, Copley, etc.
These drives no longer use tach feedback as many of the old velocity drives did.
The encoder may be quadrature, not 12/13bit versions.
What controller were you looking at?
Al.

If you were OK about using Linuxcnc, the Mesa 7i77 board has 6 axis of 0-10 volt servo control and 6 encoder input. There would be a couple of other things you'd need to put it together so if in doubt ask on the LCNC forum. Mesa have so many cards now its hard to keep up with them all. They now have a single board ethernet version of this card.

Thanks for the feedback rodw, I'm in Aussie too.

I've posted a repsonse to Al_the_Man but after hitting reply it frozen up. i'll wait a bit and see it materializes.

I did have a hardish look a LinuxCNC, very interesting in it's own right, and great people in the Linux world. And I got a Mint Preempt version running after some issues. And I looked at the 7i77 as a preferred option, there was no setup available but I saw some on the forum some folks got it running and that was to be my next step to try. A very cheap option as well. I guess my concern was I didn't really have the CNC knowledge to know if after putting the work into the 7i77 and learning the LinuxCNC setups (for a non-computing Windows man) if in the end I'd have a user friendly CNC system with good conversational etc. I had the impression it was a system that was for computer orientated people and may not be polished user wise, and that something like Centroid Oak would be (if you pay the money). so it's was to some extent a balancing of whether you wanted to spend more money or time. As I say I'm a beginner but have to make a decision eventually.

LCNC is very powerful but with great power comes great responsibility (or complexity!). There are some very nice GUIs emerging for it now and the conversational side is pretty good but as I use Plasma I don't get into that. Some people have done some amazing retrofits of big machiines efg routers and mills with tool changers and the works. I don't know the other systems, but from what I've seen the average Windows retrofit chops out a lot of the signals due to limited inputs and outputs available. With hardware from Mesa and others like pico systems, you can retain a lot more of the original hardware so it can cost less and you get a complete working system back thus preserving more of the machine's value on resale. The 7i77 is very popular and some people have had good luck with cheap 0-10V 750W 80ST servos from China. I think though with a new build on modern hardware the 7i76 or 7i76e ethernet step/direction card is a better way to go and add a 7i85 to manage the encoders or add linear scales to create a dual loop feedback system. I think there is a 7i97 which is ht eethernet version of the 7i77.

MInt is nice and I use it becasue when I started, I had to compile the Linux realtime to get prempt_rt as that is what you had to do. It was a small step to learn how to compile LCNC from source. But today, I take the least source of resistance and run with the official stretch ISO if I want to build a machine as my start was a pretty brutal journey.

Al the Man, looks like my previous typing has vanished. It had been saying autosave but I can't find anything, so I'll try and summaries what I said;
The servos are late 90's, not as I said before, and are quadrature, 2500 ppr geared down 2:1 to C5 30mm screws. They are 1000W 220V 3ph running of a star connected 2.9kVA tranny. A 4HP motor is fitted. The 12 bit story came from my attempt at binary math based on this info.

I had a look at the AMC site, quite a bit of info but i would be able to spec something but I guess given my current specs they could recommend something. Not sure about their costs. I had looked a Sigma V's before, seemed very good to me but not cheap, also looked at cheaper Chinese but don't really understand what I'd be trading off there.

On Control systems as stated above, while I have not given up, I had looked at LinuxCNC. I had not considered main industrial systems for cost reasons. I looked a GSK, Accurite, Protrack but thought Centroid Oak looked a polished system, with DIY well explained etc although still not cheap with add-ins. One theoretical thorn was the US control on full 5 axis that pushed cost option too far for me. If I'm going to the trouble of fitting a control it would be nice to have that option but reality I'm a long way from 5 axis CNC. The other Centroid thing is rumors of EtherCAT systems that may be better, easier etc but very little info is given. I'm sure they are normally under pressure and more so now but it would be nice to know what's what. EtherCAT would mean a clean start with new servos which has attractions although i don;t need to onward connectability .

Anyway any directional advice would be helpful.

Sorry Rod, I had been looking at the 7i97. It was a pretty steep leaning curve and to be honest I managed to side step or put off some of it because of the good forum posts around. But there was/is a long way to go to get running so I began to question the time input.

I don't know your machine but Ethercat is supported on Linuxcnc. They are both open source but there is a license incompatibility which means it can't be bundled with LCNC. But for smaller machines the significant cost of ethercat is probably not warranted. Yes the control box might look nice and clean but $200 or so for some Mesa cards in exchange for a few more wires in the controlbox is pretty compelling. I know that LCNC's flavour of Gcode is superior to Mach3, better probing and you can actually do things from gocde that don't break motion. and full 9 axis control....

I have not used LCNC, but recently helped one user that set up a DC T.M. motor and controller for synchronized tapping, which he succeeded in doing, using LCNC
Al.

IF the servos are 3ph AC servos then you could use DMM servo drives to run them. DMM can sell you a kit to do the conversion. you would need to install their encoders. I converted an old set of 1KW Sanyo Denki's motors with the DMM drives. Very easy to do.

(:-) TP

Quote:

---

Originally Posted by rodw

If you were OK about using Linuxcnc, the Mesa 7i77 board has 6 axis of 0-10 volt servo control and 6 encoder input. There would be a couple of other things you'd need to put it together so if in doubt ask on the LCNC forum. Mesa have so many cards now its hard to keep up with them all. They now have a single board ethernet version of this card.

---

The ethernet version 7i95 seems eto be perpetually out of stock/sold out.....

This is not the 90's or early 2000's where you had to build and compile everything. Linuxcnc is fairly user friendly, and comes in a prebuilt iso, so you won't need to worry about the inner workings of Linux, or the build files for Linuxcnc itself..

Disclaimer... I suspect my definition of user friendly may differ.. as I have been a linux/BSD user since the 90's.. That said.. if you install the prebuilt ISO, you should be up and running fairly easily. If you don't want to put out a ton of cash, you could go with an Raspberry Pi 4, and one of the Mesa RPI boards, worst case, if things don't work out, you can easily sell them.

Quote:

---

Originally Posted by ljvb

The ethernet version 7i95 seems eto be perpetually out of stock/sold out.....

---

Thats a newly designed board so maybe its caught up in the covid thing. Maybe check with Mesa by phone.

Adding the 7i85s to a 7i76e will get you to the same place more or less and not cost any more.

Hello.

To my understanding the limitation in old servos is mostly the way they are controlled. That is, in most of the newer controls the drivers are actually responsible for the response of the motors so that from the user point of view they act as if they were step motors giving them only the direction and pulse signals. The length of the step is given from the resolution of their encoders or their resolvers to digital converters and some other parameters within the drives. Some accept an analog voltage as an input but in the end the same process is involved as that voltage is used to set the pulse output rate. The controllers perform the PID loops or whatever other technique they could be using.

In the old controllers the drives were little more than smple amplifiers driving the motors. You had to have ways to read the position either from encoders or resolvers and then perform all the calculations requiered to follow the given geometry.

The following links show the very same machine. First it was retrofitted replacing only the computer and the main drive which were both damaged. Thanks to the fact that this machine performs continuous quilting it was possible to develop a simple angle fired speed control for the main motor. These are three DC servomotors. As a reference for the speed of the main motor a wired Digital to Analog converter with 16 discrete speeds was designed. The tachometer within the main motor was used for the PID speed loop. Two Digital to Analog converters with a +10/-10 Volt range were required to signal the drivers and circuiry to read the encoders on the axes motors as well as some microswitches and buttons. The program has to do all the calculations for the PID loop of each motor adjusting speeds as requiered for the geometries.

https://www.youtube.com/watch?v=SC3OrgzDfJU


The second link shows the same machine years later when the drives finally failed and the system had to be refitted using newer servos and drives. In this case the interface card has no more than a ULN2803 to signal the drivers and the aforementioned wired DAC.

https://www.youtube.com/watch?v=A488ULvflkQ


The third link was given to show one machine using step motors instead of servos. The program however is basically the same as the one used in the previous machine. The difference is due the fact that the latter has to work slower because of mechanical restraints and the main motor speed is fixed. The picture shows the actual interface (breakout board) used for this machine.

https://www.youtube.com/watch?v=pqYHGteyzf0

Having said all that, perhaps you would rather be considering using newer motors and drives. From my experience it can be rather difficult (as well as expensive) to get old software for similar drives and enduring the effort to write your own software might not be adviseble.

On the other hand, while probably not the idea you´d rather get, I´d suggest publishing your motors and drives in ebay or any other publishing page. They might be of help to some who still has equipment running with those parts and have the software.

I hope this can help you decide what is better for you.

Best regards.

Quote:

---

Originally Posted by BBMNet

Hello.

To my understanding the limitation in old servos is mostly the way they are controlled. That is, in most of the newer controls the drivers are actually responsible for the response of the motors so that from the user point of view they act as if they were step motors giving them only the direction and pulse signals. The length of the step is given from the resolution of their encoders or their resolvers to digital converters and some other parameters within the drives. Some accept an analog voltage as an input but in the end the same process is involved as that voltage is used to set the pulse output rate. The controllers perform the PID loops or whatever other technique they could be using.

In the old controllers the drives were little more than smple amplifiers driving the motors. You had to have ways to read the position either from encoders or resolvers and then perform all the calculations requiered to follow the given geometry.

.

---

There is nothing wrong with controllers that use non-intelligent drives, where the PID loop is closed back to the trajectory controller itself.
This has been done in many high end systems and the method that is presently used in high end motion cards use such as Galil etc, where the encoder feedback response is in the order of 12Mhz and up.
These drives are torque mode or simple trans-conductance amplifiers.
Al.

Quote:

---

Originally Posted by BBMNet

Hello.

To my understanding the limitation in old servos is mostly the way they are controlled. That is, in most of the newer controls the drivers are actually responsible for the response of the motors so that from the user point of view they act as if they were step motors giving them only the direction and pulse signals. The length of the step is given from the resolution of their encoders or their resolvers to digital converters and some other parameters within the drives. Some accept an analog voltage as an input but in the end the same process is involved as that voltage is used to set the pulse output rate. The controllers perform the PID loops or whatever other technique they could be using.

---

I think the reported limitation here is a motion control system that can't manage anything other than step and direction. I'd be very surprised if you had to build circuit boards with a Mesa/Linuxcnc environment. You can still buy brand new servos that only do 0-10 volt and Linuxcnc includes a software oscilloscope that allows you to tune them perfectly. Even the Mesa cards are dumb hardware. All off the smarts is in LinuxCNC itself. And it has the full power of a modern PC's CPU to do everything

Quote:

---

Originally Posted by Al_The_Man

There is nothing wrong with controllers that use non-intelligent drives, where the PID loop is closed back to the trajectory controller itself.
This has been done in many high end systems and the method that is presently used in high end motion cards use such as Galil etc, where the encoder feedback response is in the order of 12Mhz and up.
These drives are torque mode or simple trans-conductance amplifiers.
Al.

---

Hello again.

Just to clarify things a little.

I was not saying that DC servos are a bad choice or that he wouldn´t be able to find a suitable controller. I simply attempted to be in Hazron´s shoes. While it is certainly possible to get or even in some cases adapt current trajectory controllers to his actual drives I must say that I´d be concerned with the price of such systems as well as the support he can get. Browsing the internet I found very few suppliers for both DC servo motors and less for the controller.

Being in the business of retrofitting machinery for the last 34 years the constant I have always seen is that brand manufacturers always try to sell full new systems to replace any out of warranty controllers. The ones I have seen range from expensive to very expensive.

The link below shows a retrofitted router. As you mentioned the drives receive the tachometer feedback for speed stability purposes but the actual trajectory generation and hence the speed modulation is performed by the main computer. This machine uses brushless motors.

https://www.youtube.com/watch?v=B6FiInwcxh8

In any case I´m sure he will welcome any suggestion you can provide him.

By the way. Making an interface board is relatively simple if you can take the effort. The problem lies in the software to drive it. The first picture shows the actual card designed for a chain quilter controlling both the axes motors as well as the speed of the needles. It has two 12 bit DACS, one 8 bit DAC, 8 bits for digital inputs and 8 bits for digital outputs. The next link shows that machine. These are DC servomotors

https://www.youtube.com/watch?v=SC3OrgzDfJU

Now the following link shows that same machine years later retrofitted with the newer AC servo motors and drives.

https://www.youtube.com/watch?v=A488ULvflkQ

The second picture shows the interface card needed for this latter system. Please note the simplicity of the new card. One single ULN2803 and some resistors and capacitors are all that is required to communicate with the machine. And of course, no adjustments.

Software in this latter system as I mentioned before becomes a lot simpler than that of the previous two as the drivers as I already mentioned take the burden of keeping the motors in track.

For this option there are a number of programs even some for free (LinuxCNC among them) with a lot of support from forums like this one.

After all this I only hope that our friend Hazron does not get confused.

Best regards to all.

PS: if you need to be confused... I mean if you have any doubt or question please post a new thread.

Quote:

---

Originally Posted by rodw

Thats a newly designed board so maybe its caught up in the covid thing. Maybe check with Mesa by phone.

Adding the 7i85s to a 7i76e will get you to the same place more or less and not cost any more.

---

That's good to know. The main problem I have with the Mesa site, is that there are a ton of options, and no comparison chart that I could find to figure out what daughter boards to pair with main boards, in fact, the mesanet.com has a geocities look and feel to its website. And Mesaus.com is only slightly better.

I know the feeling! If in doubt, ask on the linuxcnc forum.

Mesaus is run by one of the Linuxcnc stalwarts (BigJohnT) and I believe it was to move retail sales away from Mesa themselves becasue they were struggling to get orders out the door and needed to concentrate on wholesale orders. I have bought from both and deal with Mesa direct now. They seem to have solved their delivery issues. John was quick too.

I m a bit confused I guess. Just to recap my servos are what was good quality in the day Japanese brush-less AC servos which can do position, velocity or torque control modes. I've been told they would have information overload if feed position from a modern DRO so Velocity mode is the way to go. They have PI control of motor position via the quadrature encoders in a local loop and feed that position signal back to the main control system which also has some algorithm for control. So I guess it's like a cascade loop where you tune the local loop reasonably tight and the main system feeds the set point. All pretty standard stuff I think. Severing is an issue if the break (the manual says replace all capacitors after 5 years!)

I didn't want to go open loop as in steppers and I don't really understand what can be done with them but I will stick with servos I think, and while I have some tech knowledge and can plod along if lead by the nose I don't really want to get into hybrid mods etc etc,. I'd prefer just a matched motor/drive. Having said that I was unaware you could convert old AC motors to run with servos like DMM. But is it worth it for the cost of a motor to move away form the manufactures matched driver/motor?

So I guess the question still is will I get a reasonable quality finish on curved surfaces (all other variables being equal) with 5000 pulses of position info feeding the system?? Logically the answer is probably not otherwise companies wouldn't have gone out and build servos with way way more resolution. So what would be the advice be...go cheap and try what I have and risk pulling apart again and redo all the servos later. Or start with a clean control build.

If it's the later then advice on new servos is welcome as some have already. DMM seems priced well and sound acceptable. Some have used GSK servos and said they were OK. The Yaskawa Sigma 5's (or 7's) have to be good stuff but pretty expensive for an old mill refit. Again I'm just not sure what you loose by going to the cheap brands, I assume a fair bit.

So the Mesa 7i77 should control those drives and read the encoders. Then as you say, you let the motion controller (linuxcnc) manage the feedback loop. Here is a tutorial on drive tuning by BigJohnT mentioned above.
https://gnipsel.com/linuxcnc/tuning/servo.html
Your best bet is to grab some photos of your drives and motors (part numbers too if you can find them) and post on the Linuxcnc forum. I don't really have the experience to guide you here but plenty of others wiull help.