shaft synchronization

I'm currently working on a project which requires shaft synchronization. Anybody have any experience in this type of project? any help would be appreciated. z

It is easy, I do it all the time.

ok great. are you willing to discuss or was that your final answer?

????

zcut said:

????

Click to expand...

This isn't a chat, you can't expect an answer within minutes of posting.

I think what Peter was getting at was the lack of info in your first post. If you provided more info about your desired results and what you were using, more people can provide answers.

ok sorry, new to forum. what I have is a four axis indexed machined.I am using mitsubishi fx3cu, inc encoders. I set the least loaded axis as the master. I test for position relationship in a defined part of the rotation, currently between 0 and 180 degrees. I sample the error during this time, average it and pass to a PI loop. the old system I am trying to replace had a synchro and tach for feedback. It also had a very small range of operation +- 5% of rated speed. The customer required the machine to be slowed down to 1/6 of that. While it seems to "sync" up relatively quickly on initial startup, it does not upon stopping one axis "which they have to do for maintenance". Upon restart sometimes it locks in within 5 indexes, other times not at all. The original also had an independant master synchro signal. When looking at the encoder feedback it was in fact varying in frequency, altho it was slight enough that I felt the loop would overcome it. Also the allowable error between axis' is +- 2%. So it is somewhat forgiving. I have never had the opportunity to set one of these up before, so I know there is a better, more efficient, and tried and true method. Should I provide a separate, non-machine based master reference? again any suggestions are appreciated... z

So you are only looking to sync the axis once a rev?
If you are setting the master according to load I would think you would want the one with the most load as the master. Because it would have the worst response.

I assume this is a new system. If you want good motion control go to a motion controller.

CharlesM said:

I assume this is a new system. If you want good motion control go to a motion controller.

Click to expand...

Yep.
This has been covered many times.
PLCs & encoder cards are not designed to do tight motion control.
There are PLCs that have built add on motion controllers (ControlLogix for example) but typical PLCs are not meant for reasonably complex motion control.

You've got some challenges.

First of all, are you looking for position sync, or velocity sync.?..big question. Your original detailed post talked in 'speed' and 'velocity', so it sounds like you are living in the velocity world (VFD's), but thinking you need to be in the position world (servos).

If position sync, then, yes , you need a Motion Controller with sufficient firmware to position control axises, and drive them accordingly. Of course, you also need motors that can respond, not to mention decent position feedback (quad encoders are not sufficient, <generally>) You also want to start thinking in lines of 'virtual masters' versus 'actual master', if you want a smooth machine.

zcut said:

ok sorry, new to forum. what I have is a four axis indexed machined.I am using mitsubishi fx3cu, inc encoders. I set the least loaded axis as the master.

Click to expand...

It is best if all the axes follow the computer generated master. That way all the axes have the same master. The master is perfect because it is computer generated. Gearing to a physical master causes a lot of problems. The master position is quantized by the resolution of the encoder. This adversely affects the master velocity and acceleration calculations. The master will also suffer from wow and flutter. This will make synching to the master even harder. As stated above, a computer generated master position is perfect and so are the calculations for the velocity, acceleration and jerk etc. These higher order derivatives are important for calculating feed forwards. The four axes can use the same master or target psition by just issuing exactly the same command to all four axes at exactly the same time. This is easy. If all four axes are tuned well then they will follow their target or master position, velocity and accelerations with little error.

I test for position relationship in a defined part of the rotation, currently between 0 and 180 degrees. I sample the error during this time, average it and pass to a PI loop.

Click to expand...

I would consider using SSI encoders because they are absolute and don't need a homing routine. On start up all the axes should be commanded to a common point. That may be an average position as you stated. When all axes are in position then all the master are target positions will be at that same position. From then on you should give the same command at the same time to all the axes at exactly the same time. This way the masters or target positions will following exactly the same motion profile on a millisecond to millisecond basis as the targets move to the new position.

the old system I am trying to replace had a synchro and tach for feedback. It also had a very small range of operation +- 5% of rated speed. The customer required the machine to be slowed down to 1/6 of that. While it seems to "sync" up relatively quickly on initial startup, it does not upon stopping one axis "which they have to do for maintenance".

This is why I prefer SSI absolute encoder for these applications. Sick/Stegmann and Danaher make excellent absolute SSI rotary encoders.

Upon restart sometimes it locks in within 5 indexes, other times not at all.

Click to expand...

I thought you said you told all the axes to go the same average position on restart? In anycase good controller can handle that easily.

The original also had an independant master synchro signal. When looking at the encoder feedback it was in fact varying in frequency, altho it was slight enough that I felt the loop would overcome it. Also the allowable error between axis' is +- 2%. So it is somewhat forgiving.

Click to expand...

Again, this will not be a problem for a good motion controller because the master or target position for all the axes will match exactly at the end of the day.

I have never had the opportunity to set one of these up before, so I know there is a better, more efficient, and tried and true method. Should I provide a separate, non-machine based master reference? again any suggestions are appreciated... z

Click to expand...

A good motion controller can make your life much easier because the issues you bring up are not a problem.

You also mentioned incremental encoders. They will work but will require homing on start up and power loss. If there production on the line that must be cleared off during homing then I would stick to the SSI encoders. If power goes off during production then the four axes should drift to a stop. They should be close enough on start up where a go to the average postion will get them back in sync with little motion so the production that is on the system is saved and no time is wasted homing.

I have an idea........
I wonder what I can do with windows movie maker and a four axes motion controller. Hmmm.

Click to expand...

Click to expand...

position sync. On startup I run all axis at a fixed speed after the servos have ramped up for a period of time. Then I release control to the pi loops, which are only sampling if they are in the same quadrant. If after one rotation they are not in the same quadrant I pulse the function to sample them where ever they are. The loop responds and brings them closer.
I wanted to use absolute incoders (grey code) and my boss disallowed it.(some boss) I have work with what I have here. I do reset all encoders on startup when a comman shaft cam passes thru an opto. this is zero to everyone. all encoders are fed to software resettable counters. absolute would have made this much easier for me to resolve position error. I am using the servos in a speed loop only, and contolling the speed with pulse train input. 100,000 pps = 2000 rpm. I would create an internal fixed master reference but I don't have the equipment resourses to do it. All high speed outputs are in use.

You seem to have two independent but related issues.
The first is keeping the relative speeds of your axes as close to the same as possible. This is an electronic gearing issue. In a perfect world, if your gearing was exactly right, you could put the axes in the correct orientation once and never have to do anything else. This is where you want to start. If this is taken care of you will turn your problem into a static phase offset problem, which tends to make things easier.

As Peter said, use a virtual master. Keep in mind that a virtual master is completely a software construct. You need no specific hardware to implement this. However, you will want a relatively accurate time based interrupt in order to calculate the master values. Assuming you don't need to link the virtual master speed reference to something else you may be able to be a little sloppy with the interrupt time and still be OK. It depends on how your system deals with the dynamics that occur with inaccurate time interrupts. Again as Peter said, a virual master gets away from some very tough problems that real masters present, not the least of which is trying to follow another axis' dynamics.

Once you have all the axes tuned and following a common master you should have a simple phase offset issue. You should be able to correct this by making a 'simple' position offset move. Yes, I get real concerned too when people use the word 'simple' in this context. There is more to it than that. Keep in mind that even though you think of your pulse train in terms of velocity, it really is (??should be??)a position signal. 100000 pps = 2000 RPM is also 3000 pulses per motor revolution. By scaling motor revolutions to load motion you can figure out how many pulses you need to issue to correct for your phase offset. The trick is adding those pulses to the existing geared pulses in such a way as to not excite vibrations in your system or exceed your system's capabilities.

Having said all that I would tend to recommend going with a motion controller to do all this, especially if you are under a time constraint. If you want to play some and have the time then by all means give it a crack with what you have. But I can tell you that pulsding a PID controller to correct for phase may not be the way to go.

Get a hold of the guys at Delta Computer Systems. They should have just the thing for this.

Keith

zcut said:

position sync. On startup I run all axis at a fixed speed after the servos have ramped up for a period of time. Then I release control to the pi loops, which are only sampling if they are in the same quadrant. If after one rotation they are not in the same quadrant I pulse the function to sample them where ever they are. The loop responds and brings them closer.

Click to expand...

This is where a 'phasing' command is handy to get your axes in 'phase' in a controlled manner while moving on-the-fly. In this case you do pick one axis to be the master and then the phasing command is issued to each of the 'slave' axes. The first parameter is the offset from the slave position which is usually 0. The second parameter is how much distance the master will move before the slaves reach the desired offset. This parameter is usually calculated on-the-fly for optimal performance. If the axes are far out of synch then it is best to let the master move a farther distance before expecting the slaves to be in sync. If the slaves are just a little out of phase then the distance the master moves before getting into phase can be very short. This calculation is a trade off between minimizing the phasing time and distance without exceeding the speed of the slave axes.

go for the gold

Everything you could possibly need in multiaxis position sync can be done with an Indramat PPC, running SYNAX 200, and linked to IndraDrives equipped with SERCOS.

... or with Yaskawa MP series PLC. If we are going to be specific as far as our favorites are concerned, why not to chide in?

Kidding aside, there are quite a few excellent motion controllers on the market which are specifically designed to handle all the issues that have been mentioned here. I think, this would be the third or the fourth brand name mentioned in this topic. All these questions had been asked and answered before by many people; are you sure you really want to re-invent the wheel?