On May 21, 2012, at 10:36 AM, Jon Elson wrote: > Jeshua Lacock wrote: >> Putting these numbers in immediately made the drive stable - even in closed >> loop which is now what I am tuning. >> >> I was able to get the P up to 150. I can add some D, but it seems that it >> only makes the drive more jittery without decreasing the error. >> > Well, then lower the D until the loop is stable. At some point, too > LITTLE D will also > start to make it jittery, there should be a valley where it is good.
Thanks Jon! I am actually going to wait until I get the new 5:1 gear box installed before I tune things further, but there is one last thing I don't really have a grasp on. It felt like I wanted to get P set as high as I could. Is that correct? Or as long as P is high enough to command the machine to move is that adequate? >> Can you tell me what exactly is the goal? Do I want the P as high was >> possible while keeping it stable? In other words, why not just leave P at 50 >> or 100 and be done with it? >> > Yes, sure. Once you have good, stable response where the following > error is fairly low > in the cruise portion of a move, then apply FF1 to make the error nearly > zero there, > then add tiny amounts of FF2 to reduce the error in the accel/decel > ramps. You should > be able to get the error quite small, although with your encoder > resolution (I'm > guessing from some numbers you gave before that it is about 2000 > counts/inch) > you may not be able to get below .002" or so under all conditions. That's > probably OK for a router, and already less than the overall accuracy and > backlash > of the mechanical part. Got it, thanks! >>> And, it may be that you are having feedback from the G320 drive's PWM >>> outputs to >>> the encoder cables. This may be what is causing the "increasingly >>> jittery" effect. >>> Do you have shielded encoder cables? >> >> Yes. Do the shields need to be grounded to be useful? >> > Of course. You should ground the shield at the control end only, so > either the USC or the Gecko interface. Thanks! So I use the ground from either the USC or Gecko? >>> Are you using US Digital encoders that >>> are known to be noise sensitive? >> >> Not on this first axis I am trying to tune. It is a Renco. My other two >> axises are US Digital, but those will be replaced by CUI AMT102 encoders >> when I get around to installing the new much more powerful motors for the >> other 2 axises. >> > Hmmm, the AMT encoders have a lag in responding to acceleration, but it > isn't as bad as I originally thought. It seems to help to use less than the > maximum resolution the encoder can provide. Good to know! I guess I will set it for 1024 (instead of 2048) to start. >> I just went ahead and ordered some differential encoder/decoders just to be >> safe. Due to the size of the machine they are pretty long cables, I would >> guess around 15 feet. >> > OK, that's a good plan. >> Also, I routed my motor power lines as separately as I could. My encoders >> exit the machine on the left side and the power on the right. >> > That should also help, although shielded encoder cables can be routed > next to motor cables > in most cases. At least with my drives, that is fine, but they have the > filters on the output. Also good to know! >> What is the goal with I once P and D are good? >> > Well, that's the problem. I works great in a steady-state system, but > CNC never is > steady-state for long. It looks back in time at the average error over > some period > and applies a correction. But, as soon as the axis reverses, the > correction is > now in the wrong direction! There are a bunch of other places where I > also fails, > such as the change from accel to cruise to decel, or when the cutter is > cutting vs. > out of the cut. So, I would not work very hard with tuning I, and use it > sparingly. If you want, you can try setting I to 10% of P and see how > it works. > But, if the other params (P, D, FF1 and FF2) have reduced error to a good > minimum, then the I term really has no error to work on, and is benign. Thanks for the most understandable explanation of I that I have read yet! >> Interesting, I can't seem to get D anywhere near that high. With P at 150 or >> 200, it gets real jittery with a D of over 0.5. >> > Well, this all depends on INPUT_SCALE, motor response and inertia, and > all sorts > of other things, both software and mechanical. Don't think that your > numbers should match > mine, as the motor to encoder link is so critical to the whole > response. One thing is if > there is any backlash between the motor and the machine part to be > moved, it really > complicates the tuning. Whenever the gear, or whatever, taps from one > side to the > other of the backlash, it creates a big impulse to the system. I see. Currently, since I am only using pulleys with timing belts (no gears or lead screws per se), I was under the impression that backlash is nearly non-exisistent. Is that true? My z-axis is pulleys with a ballscrew. Of course as soon as I install the 5:1 gearbox on my X (hopefully here tomorrow) then it will have some backlash (see below). >> I see. So when I am tuning, I should start out with a modest acceleration >> factor, then bump it up until I see a spike? >> > Yes. If a big spike suddenly appears in a system that was previously > well-behaved, > that is a strong indication that some limit has just been reached, most > likely the > current limit if it appears on the accel/decel ramps. You want to back off > maybe 20% from there to allow cutting forces to be developed. You don't > want 100% of available torque to be used solely for acceleration. Got it. >> Yeah, I think the 5:1 gear box will get me somewhere in the same >> neighborhood as your Bridgeport. >> >> My motor is definitely not weak though it is a 1125 oz-in Peak 90V/40A. >> > But, with the Gecko 320 turned all the way up, you only get 562 Oz-In > since you can only feed it 20 A. But, you are less likely to fry such a > motor. Good points. >> Besides much more torque, I will be getting much better resolution. Just >> dithering I am currently at 0.002+ inch, and currently when I rapid I am >> getting an error of around 0.01. I think when I gear it down I could get >> that closer to 0.001... >> > Yes, a 5:1 reduction should automatically give a 5:1 improvement in > resolution. I hope those gear reducers are zero backlash, though. Actually the gearbox is rated as 6 arcmin backlash. I did the math and that seemed quite negligible to me, at least for my purposes. Thoughts? Do you recommend any tricks, like springs or semi-flexible drive shaft? >> Thanks again! You really got me feeling much better about tuning it went >> from scary to feeling like I am almost in command now! >> > Yeah, the FIRST time, it is TOUGH! Once you see how the various > parameters interact, you start to understand what each one does. Thanks again for all your help and patience Jon! I owe you a beer or two! Best, Jeshua Lacock Founder/Engineer 3DTOPO Incorporated <http://3DTOPO.com> Phone: 208.462.4171 ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
