First Real Look At My Ez-b Controlled Full Size Lis B9 Robot

It looks like when you move the setpoint slider the command signal doesn't start to change until a half second later for moving down. And about a quarter second lag in going up. I would think that it should respond right away. Just a thought.

@Purple, I've watched that same thing very closely many times. I think what we'er seeing is the motor not moving till the slider gets out of the 15mv deadband setting. If you watch the control numbers it's hard to see but the reading is just about that or a little less before it starts to move. I'm not sure what your seeing with the 1/4 second lag. There is a responce setting in the control section and I've got it set as low as I can before it gets jittery.

I've taken a little time away from working on this problem. I need to clear my mind a little on this one and regenerate some burnt out brain cells. I'll sleep on this a few days and mull over all the great suggestions and support I've gotten on this. Hopefully Spectra Symbol or Demension Engineering well take a look at my support request soon and give me a hand.

In the mean time I'm working on a sound issue I'm having. The thread is here: Ez-b Soundboard

Just a short update (I hope). Edit: Nope, it a long one.

I have not heard back from Spectra Symbol about the Softpot or Dimension Engineering regarding the Kangaroo X2. It's been almost 2 weeks now. I did send a second request but nothing back. I'm very disappointed in them but not really surprised. DE puts out great products but myself and other people I personally know have found their customer support spotty and slim. Also, I do have a little history with Spectra Symbol's customer support. The one other time I asked for help from them I got a quick responce. Not this time. It would be nice if these companies don't know how to help to just send me a reply saying they don't know. This may not be the case but I'm left with the impression that they just ignored my request for assistance. mad tired

So, as usual I'm left with the feeling that I have to figure stuff like this out on my own. However, thank goodness I'm wrong about that! I've had many fellow builders jump in and lend a hand with support, ideas and advice. You guys are wonderful and have helped in more ways than you know. I really have to mention a certain builder by the name of Davis H. from Louisiana. He has spent hours personally studying this issue and more then two full hours on the phone with me going through the issue step by step. It doesn't seem enough to say "Thanks David"!

Anyway David H. has suggested many things that may be causing this issue of not being able to get the arm motor to stop at the same point depending on which side of the radius it's being approached from. I think that the total sum of this issue is a combination of a few of the problems David and others have identified. I've been working through most of these suggestions and have made some real progress and also eliminated a few things that were not the issue.

So far the things I've done that have really made a difference:

Reduce the speed of the arm over all and more so as it reached the set point. I found this to be the best way to reach a consistent set point.

Reduce all the deadband values as far as I could without getting a rocking action as it's trying to stay at the set point. Using the DeScribe software I found there are at least 3 deadband settings. Two for position and one for speed.

After I did some studying on PID adjustments and what they really are I tried to adjust the Coefficient values in the PID section. The Autotune filled these in for me but as I said before I had to redo them to get the motor to respond smoothly. With better understanding of each setting I was able to get better resolution of the set point (the point I'm trying to stop at). I was now able to stop with in a constant 2 or 3mv on each side of my set point. (I have to say here that the terms that DE uses for the different Gains are not the same as the rest of the world uses. They use terms like Fast and Slow Gain) This makes it even more difficult for the newbie to understand what needs to be set and changed). mad

Even though this resolution may have been enough to give me the tolerance I need on centering my arm my OCD kicked in. I thought I can do better then this. So I took another step David H. suggested and replaced the !0K resistors on the pos & neg pins between the Softpot and the Kangaroo X2. This would give me almost the full 5v (4.8 in think) that the Kangaroo is supplying the Softpot and expecting back to read (feedback) the variance as the arm moves. With the 10K resistors I was only getting 1.6 volts around the pot. Well, This change through off all my PID and limit settings in the Roo. None of my old setting would work and the arm was unusable.

I had to try to do a new autotune and I was not looking forward to it. In the past (only on this arm setup) I was never able to get a successful tune with the arm fully weighted down. After removing the arm from the robot and mounting it directly to my workbench I tried a few autotunes and failed. The software always reported back that it "couldn't move the system" even thought the thing was rocking up and down wildly. eek Obviously the Kangaroo is not able to auto tune such a heavy asymmetrical load as I have. So I took off the front claw section (about 4 pounds) and attempted another Autotune. This time I had a quick and successful tune. When I tested the tune I had nice smooth action with nice ramping effect. However when I reattached the 4 pound claw section the movement sucked.

At least now I had a good baseline tune to work with. These past two days I've spent about 10 hours resetting and testing the PID settings for the Roo with the DeScribe software. Late last night I've finally got the arm moving smoothly with nice ramping. I think I have the resolution very close to where I can get this thing stopped at within one MV of my set point. At least that's what I can see in the Live Test section of the DeScribe Software. Today I'll hook this monstrosity up the EZB and see if all this did any good.

Thanks again to all of yo that has lent a hand. Thank for nothing DE and SS! mad

Hi Dave,

Great to hear progress was made. I was really pleased to hear Dave H and you were able to home in on the main issues. It also sounded like the retune drove you crazy, but was fruitful. (PID loops can be a pain in the butt to set). Thanks for taking the time to post the steps and updates you did.

Ron R

PS I have a pocket full of Quarters.

Thanks Andy. No need for quarters. Everything is set to free play.

Yes the autotune was a failure because the Kangaroo can't figure out the 4 pound heavy load I have out on the end of the 2 foot arm. Once I removed the claw section it tuned just fine. However it was set up for the lighter load. I did give me a good base line to work with after I reattached the 4 pound claw. Man, setting up a properly tuned PID is a pain but I have it real close now. Maybe a little tweaking yet if I run out of things to do.

I attached the ezb last night and battle tested stopping at my set point from each direction. I'm going to call this issue RESOLVED! It's stopping either dead on or within 1mv of the set point. I couldn't be more happy, thankful and excited.

Here's a youtube video I took last night of the results. Enjoy :

NICE........ I like it.! Nice and smooth with good ramping, no oscillation or hunting. Repeats nice too.

Thanks for sharing..

Ron R

FREE PLAY ? MUST BE A PARTY NIGHT ! I'm playing Twilight Zone next..

Absolutely Incredible work Dave, you must be very proud. I imagine you go to bed at night mentally exhausted, so much hard work has gone into this arm, and the result, absolutely perfect.

Steve

Thanks Guys.

If any of you are ever in the Milwaukee, WI, USA area drop me a line and you can stop by to play some pinball and robots.

Steve, go to bed mentally exhausted? I found a couple shots of good Kentucky Whiskey about an hour before bedtime helps with that.