Hello Guest it is March 28, 2024, 02:16:05 PM

Author Topic: CS Labs CSMIO/P-A  (Read 51061 times)

0 Members and 1 Guest are viewing this topic.

Re: CS Labs CSMIO/P-A
« Reply #70 on: July 18, 2018, 06:25:49 PM »
Mike

Is the analogue drive to the motor you are using mapped to the same  channel as the encoder inputs? It sounds like your motor is hooked up to the output of channel 1 but you are taking encoder feedback to channel 0 on the CSMIO. This doesn't matter per se, but it would make things more logical if the two used the same CSMIO channel. Have a look at the plugin settings to see if they are correct.

Allan
Re: CS Labs CSMIO/P-A
« Reply #71 on: July 18, 2018, 06:29:06 PM »
Hi Allan,

Quote
the IP-A is capable of excellent results.
That the CSMIO/P-A is capable of great results is not in question, its just that its
not necessary to spend 700 pound to do it.

Mick is struggling to tune his servos, and he is still going up the garden path.....wouldn't happen with good
AC servos.

To Mick:
you should not throttle your servo speed back with RV2. Limiting the axis speed should be done in Machs
motor tuning page. The facility to do it in the  drive is for Velocity Never Exceeded imposed by physical plant,
eg lead screw whipping or reduction of over voltage spike on deceleration. Your servo and drive should be tuned for maximium
safe speed and acceleration.
Craig
'I enjoy sex at 73.....I live at 71 so its not too far to walk.'
Re: CS Labs CSMIO/P-A
« Reply #72 on: July 18, 2018, 07:10:45 PM »
Hi Allan,
one situation, which as AC servo technology advances is becoming increasingly common, is serial encoders.

Delta's entry level encoder is 17 bit absolute and 20 bit absolute encoders are an extra. You may have noted that
Seimiens for instance have 23 bit muliturn absolute. Serial encoders will stonker the CSMIO.

The majority of modern AC servo drives have their own motion control built in as befits the emerging
'distributed motion control' paradigm sweeping through industrial CNC with EtherCat be the predominant
communication protocol. Ethercat is not common with Mach users....yet. Make no mistake as time goes on
it will. When that ocurrs our current motion controllers will lose the 'motion control' component and become an
EtherCat master.

Craig

'I enjoy sex at 73.....I live at 71 so its not too far to walk.'
Re: CS Labs CSMIO/P-A
« Reply #73 on: July 19, 2018, 03:34:09 AM »
Hi,
when you think about it servo drives have been getting smarter for many years.

Back in the 70's and 80's all servos were like Mick's. You needed a controller that could close the loop and apply PID dynamics.
In the late 80's digital servo drives became common where the drive could read an encoder and apply PID dynamics. That meant the controller
was not required to close the loop. Gecko 320 is an example of this idea. They work OK but are not particularly sophisticated, certainly not in comparison
to the CSMIO/P-A of our current debate, nor, in comparison to modern AC servo drives either.

In the 90's microprocessor technology had improved to the point that Field Oriented Control was feasible and has advanced rapidly ever since with increasingly
sophisticated control options. More recently still they are in themselves a single axis motion controller, able to apply limits and home autonomously. Strictly speaking
a  PC with realtime ethernet you could have a CNC machine without a motion controller at all, just a trajectory planner in the PC and motion control that is distributed amongst
the servo drives.

All of this points to the trend that servos and their matching drives are becoming evermore capable. I guess you might say that manufacturers are taking advantage
of the processing power of microprocessors/FPGAs/DSPs. I think the medium to long term prospects of highly sophisticated and expensive motion control boards
is poor as servo drives take the challenge.

Craig
'I enjoy sex at 73.....I live at 71 so its not too far to walk.'
Re: CS Labs CSMIO/P-A
« Reply #74 on: July 19, 2018, 04:43:17 AM »
Hi Allan

Pins 1 and 14 go to 10v as soon as you power up. I mentioned this before in an earlier post but I think we didn't see it as a problem at the time. I swapped The Command Hi-Low over to Pins 2 and 15 but at the same time changed the DAC output within the CSMIO to channel 1. I think there is a fault with channel 0 in the I/O/? I have done some more tests and everything seems to be working right and very accurate too.

I haven't managed to tune the CSMIO yet. I am struggling to achieve the right KP. If I set the KP at 2000 for instance then autotune, it goes through the process which takes a while. but then resets my KP way up into 40,000's. Run a small program and the machine jumps around. If I reset everything to 0 and leave the KP on around 500 the program runs smoothly. Motors are quiet no jerking etc.

Craig

Was just following the Norwin manual with regards RV2.

Cheers
Mick   
Re: CS Labs CSMIO/P-A
« Reply #75 on: July 19, 2018, 04:59:52 AM »
Hi,

Quote
Was just following the Norwin manual with regards RV2.
I presume you mean this piece of the manual:
Quote
5.3.1 COMMAND GAIN
In some situations it may be necessary to scale down the command signal in order to reduce the
top speed of the motor. This is done by adjusting the potentiometer RV2 on the personality
module.
I contend that limiting the axis speed, as opposed to setting the axis speed in Machs tuning is not what is meant by 'In some situations it may.... '

Craig
'I enjoy sex at 73.....I live at 71 so its not too far to walk.'
Re: CS Labs CSMIO/P-A
« Reply #76 on: July 19, 2018, 07:38:46 AM »
Hi Craig

I will speak to Norwin about it and feedback to you there comments.

Allan/Craig

The motor seems to be working fine now. I have had to manually tune the motor in the CSMIO as the auto-tune feature was giving mad results. Following the manual tune prodecure in the manual didn't really seem to work either apart from adjusting the KP. The kI and KVFF I adjusted by choosing the setting which gave the lowest errors. 
The motors are running smoothly and quietly with no jumping at all. Getting accuracy of around 0.02mm on a random program just going back and forth at various speeds.

Thanks for your time and patience on this. Hope I can return the favour at some point. ;)

Cheers
Mick
Re: CS Labs CSMIO/P-A
« Reply #77 on: July 19, 2018, 12:01:11 PM »
Hi all

Good news that the system is now operating, at least at some level.

I  am a bit concerned that we are still unsure that the amp and motor are working to specification. The forward gain does seem rather low suggesting the amp or motor may be at fault in some way. The motor can possibly be verified by substitution. We can try adjusting RV4 but ideally need a signal generator and scope to do so. But I think it could be done via the CSMIO scope and I think we could generate the velocity steps via Mach or CSMIO if we use Vff and set all other PID coefficients to zero to eliminate any feedback within the position loop, as this would detract from optimising the amp's velocity loop. Does anyone out there agree with this or have any alternative ideas; or does the system seem  responsive enough as it is?

If the velocity loop is too limited in bandwidh, or is Poorly damped, we may have difficulty tuning the position loop, or the overall dynamics may be rather limited. The Kv Mick has achieved seems low to me - when I was using v2.910, admittedly  with Mach4, Kv was typically over 100000.

Mick

Thanks for the more detailed explanation of how you have reconnected the encoder and analogue output. I was struggling to see how it could work with the encoder on one axis and the analogue output on another. Haven't personally had any hardware faults but I suppose it is possible. Which axis is the motor on?

Craig

Sure, technology continues to advance, but it will be some time before these changes become common practice in the home shop. Heck, many users still cling to Mach3 despite the superiority of Mach4. They have a system that does what they ask, and that is all they need or desire. Anyway, we need to get back on topic.

Allan
Re: CS Labs CSMIO/P-A
« Reply #78 on: July 19, 2018, 01:12:59 PM »
Hi Allan

The motor is on the X-axis.

Are you still thinking the KP is set at 550 as stated in my last but one post?

Setting the KV in the CSMIO and pressing auto-tune ignores the KP preset and starts again using its own KP gradually increasing it. Then going to the other parameters. I've done it twice now but renders the axis unusable once it's finished.

This is the process I have used to tune the CSMIO.

KP with everything else set to 0: Started with a setting of 500 jogging the motor forward and back. Kept increasing the KP by 500 each time doing the same jogging movements until the motor started jumping around.   This started happening at around 28000 KP. So then knocked off around 15% like the manual says and set the KP to 23800.

Ki. Same as KP but increasing in smaller amounts of 10 each time. I got up to around 240 before the motor started jumping around. I knocked off 15% and set the KI to 200.

KVff: I couldn't really see any huge difference with motor movement but there was a large difference in the error counts. I found a point when adjusting it either way, started to increase the error rate.
This was found at 172 KVff

So the settings in the CSMIO are as follows.
Kp: 23800
KI: 200
KD: 0
KVff: 171

The motor seems to be running fine on rapid and slow moves Allan but if these setting seem odd or not right I am happy to try and improve them.

Cheers
Mick
Re: CS Labs CSMIO/P-A
« Reply #79 on: July 19, 2018, 03:13:13 PM »
Hi,

Quote
The forward gain does seem rather low suggesting the amp or motor may be at fault in some way
I agree, I'm of the opinion that the velocity loop is still a long way short of where it could and should be.
Consequently the CSMIO tuning just makes the best of what is rather than the best of which it is capable.

Describing the tuning process has not been effective, and a signal generator/scope is indicated. In absence of that then
the performance of OPs current settings will have to be deemed adequate.

Craig
'I enjoy sex at 73.....I live at 71 so its not too far to walk.'