Ex-Cell-O Spindle Wizard to retrofit

[usfwalden]

Controller's PID tuning:

At first I tried a bunch of different tuning methods where you adjust one parameter at a time and wasn't  having much luck.  Then I tried a method where you set both p and d at the same time in a ratio determined by the pid loop sample rate.  The DSPMC/IP has a 5kHz sample rate which means the D should be 25 times the P--then if your amps are in current mode you probably have to turn up d but as you will see I didn't have to.  I have my scale set to 100.  This will allow me to have a small integer I--I will probably barely need any I at all.

First iteration I tried
P=100
D=2500




There was no motion on the first try to I multiplied by 10
Second iteration I tried
P=1000
D=25000
.....oops, I forgot to save that screenshot


There was very sluggish motion so I multiplied by 2
3rd iteration I tried
p=2000
d=50000









The motion was much better but still sluggish so I multiplied by 2 again
4th iteration I tried
p=4000
d=100,000






motion was excellent so I tested larger moves






Vital systems is still working on the axisworks software.  That's an alpha I'm working on now.  I asked Rufi/Abdul to add check boxes allow traces to be added for position error, commanded velocity, and commanded acceleration and to remove the commanded position and actual position.  When he has that done it will be easy to Dial in the feed forwards which I want to do before even bothering with throwing an I in the pid loop.

[usfwalden]

Oh--so nobody has to go back and search for what encoders I am running--they are 1000 line encoders so 4,000 counts per turn and I have 5 turn per inch ballscrews so that's 20,000 counts per inch so I was tuning with a velocity of 150 inches per minute.  My max is about 180 inches per minute so this was aggressive without any worries of clipping by either the amps or DACs.

[usfwalden]

I just bought a 10" troyke 4th axis to add to the mix.

[usfwalden]

I installed the one opto 22 g4pb16t board and wired up my io so now I have limits and I'm pretty much ready to run mach.  I left space for a second g4pb16t board in my control box but I don't forsee needing it.  I've got 8 ouput modules in the board for my 8 m codes (wired up and functioning) and I've got the first 8 of the 16 inputs from the 1st db25 io connector on the dspmc wired to it but I'm only using 6 of them so far--for my + and - limits.  I decided not to use the spindle wizard m boards.   There is no reason to output binary codes of m functions to m boards to be deciphered and then amplified and sent to the relays.  Each M function is activated by a single 24v signal that was connected on the terminal strips in the relay panel.  Since I can output 24v from my opto22 modules I pulled the m boards and wired straight through to the terminal strips so now I'm using a discrete ouput pin for each m code.

The amc amps have a "jumper" inside which is actually a soldered in surface mount jumper not a pull it off by hand type of jumper to switch the inhibits into enables (you can order them configured like this to being with).  I removed the jumpers so now the amps are inhibited until they are enabled by the control.  To enable the amps now the inhibit, + inhibit, and - inhibit pins must be grounded.  I wired these straight to the 2nd db25 io connector on the dspmc.  This prevents the servos from lurching when the control is powered down.

[usfwalden]

The control installation is done so here are some pictures.  The advanced motion control amps are installed where the bandit amps used to be and the control and io boards are installed where the bandit card file used to be.  The 4th axis stuff is all there but not plugged in until the 4th axis actually gets here.  I used pieces of one of my old tachometer cables for the +-10v analog control signals to the amps and a piece of one of my old resolver cables for the amp enable signals to the amps.  My US digital encoder cables were barely long enough to make it into the control section of the box since they had been cut to length for the galil amp back when I was going to use the galil stuff.  I also used the original io wires.  They go from the control box to spindle wizard side control box where the m boards used to be.  There I soldered and heat shrinked the wires from the control box to the wires to the relay box.  It's all 100% tinned, extremely well shielded, probably much better than I can buy today even for a small fortune cable.  I could make things much prettier with some ties but that would just be making things less serviceable so I'm not going to.

[Chaoticone]

Comeing right along. Looking good. This has been an interesting thread. Looking forward to a video of some chip slinging. 

Brett

[usfwalden]

With my wiring done I didn't want to wait for the next version of axis works to finish my tuning and start making chips so I didn't.  With a little trial and error I settled on quadrupling my gains and throwing in some I.  I dropped the scale down to 1 so all my gains have 2 less zeros than before.  This time I tuned at my full rapid speed of 180 ipm.  I tuned all the way down to doing 2 count moves (0.0001") and never have any instability or significant following error even with ridiculous accelerations for the size moves.

Here's a ten count move at 3000, 30000, and 300000 count accelerations.







Here's a 100 count move at 3000, 30000, and 300000 count accelerations.









Here's a thousand count move at 30000 and I seem to have a 100 count move over the one with 300000 acceleration and axis works doesn't like doing step moves that take too long so that's it for this one




And here's a ten thousand count move at 300,000 count acceleration



I think I'm set now and should be able to push my mill around to my satisfaction.

[budman68]

I'm in complete "awe". What a great project, thanks for sharing - 

[usfwalden]

Thanks guys,

I'm in the process of setting up Mach now.  I got my input and output pins mapped out easily enough and have decided I don't like the way mach handles +/- limits.  I'm going to leave them wired up just the way they are except I'm not going to enable the limit switches in mach and I'm going to wire the outputs from the opto 22 modules for them to the +/- enables on the a-m-c amps as well as the input pins on the dspmc.  The amps will only inhibit motion in the direction towards the triggered limit so you can still move away from the limit without having to dissable it.

[usfwalden]

where do you set the pin for m5?  I see you can define the pins for m3 and m4 when you enable relay spindle control but it doesn't say anything about m5 for stopping the spindle....