DMU 50M retrofit 99% done, Z axis falling on brake release

Hey all,

For some time now I've been retrofitting a Deckel Maho DMU 50M to work with a Mesa 7i97t and 7i84u over LinuxCNC. I'm so close to being done with the machine, but there's just one major issue that I can't get past — the Z axis falls about 10mm on servo power on and brake release 3 out of 5 times on average, making it impossible to work with the machine. It falls no more than 10mm before what I assume is the emergency stop kicks in and catches it, leaving me with a bright red LED on the servo drive and no way to power the machine on without a hard restart.

Does anyone have any clue what I'm doing wrong? As I hope is obvious from the HAL (line 266), I am releasing the servo brake a short delay after the servo power on signal arrives (the AC power on and emergency circuit is separate from the Mesa entirely). I've tried connecting it with an and2 gate to what should be a signal for emergency circuit ok too, to no avail. What irks me is that this issue is inconsistent. I can't find a surefire way to get the axis to fall through or to work properly, it's different each time and every day.

I don't think it's the PID configuration that's causing this, as even on a stupidly incorrect calibration that makes the machine oscillate, whenever the axis falls through, it does so in the same way — as if it is merely doing so under its own weight.

I've exhaustively measured the analog out to the drive from the Mesa and don't see any relation to the issue. If I were to spam jog in a positive direction to get an analog out of 10V before AC power on and brake release, you'd expect the axis to shoot out at max speed upwards on release, but no. It falls through in the same way.

LinuxCNC itself, the Mesa and even the main servo controller seem to have no indication or idea a fault has occurred. The only indication is the red LED that I can't really find exact documentation for.

Coincidentally, I have the exact same machine with the original Heidenhain TNC 310 handy. Measured said signals there with an oscilloscope as well and I don't see the original controller handling it any differently than from what I am doing.

I'll attach the HAL, INI and some pictures of the servo drives and the relevant pages of the electrical documentation.

Many thanks for taking the time to read this.

By the way, if this post is in the wrong forum section, I apologize. Please move it to the correct one, if possible.

On my older MH400E, with Indramat drives, I made the mistake of adding a relay to give LinuxCNC co-control over the vertical axis brake. The original Maho design is that the Indramat axis driver solely controlled that brake. I had the opposite behaviour to you, where instead of dropping, the drive fought the brake during the period I had my finger on the machine start button.
Once I removed the 19K2 relay I added (second schematic in this post) the behavior worked correctly.
forum.linuxcnc.org/12-milling/33035-retr...0e?start=1740#340651

From the schematics you shared it I don't see the brake control circuit.Could it be that your brake control should be left to the axis driver alone?
Cheers,
Mark

I just had a very similar issue with a Yaskawa servopack drive on the Z axis of a Morbidelli. I tried adding `z-brake release delays via hal and had no luck. What eventually worked was finding a parameter in the servo drive that had the same purpose, something about keeping the control loop active so many milliseconds after the servo-ON signal or brake-ON signal was received. Perhaps your drive has similar? good luck!

I release z axis brake 0.2s after motion.motion-enabled and it works with my simodrive,
when does this servo power on signal becomes true? are you sure you are not releasing the brake before pid is working?

radom wrote:

I release z axis brake 0.2s after motion.motion-enabled and it works with my simodrive,
when does this servo power on signal becomes true? are you sure you are not releasing the brake before pid is working?
 

I don't think this is the issue. I checked just now to make sure. With me pressing F2 (which is also linked to the CNC ready signal going from the Mesa to the drive), I can clearly see the PID enabled and also motion.motion-enabled.

My AC power on button and circuit is separate from the Mesa entirely in an attempt not to have to reverse engineer the emergency stop, which is on the same board originally.

So my startup procedure is essentially:
LinuxCNC startup, LinuxCNC power on (F2), manually pushing the servo AC power on button, a signal returns to the Mesa board from the drive when AC power on is successful, that signal is linked in HAL to the brake release after a short timedelay.

In many cases, this works just fine. In many, Z falls through. Sometimes after a slight delay, so just when you think it's started up properly and the servo is holding the motor, it still falls through 10 seconds later.

Thanks for the suggestion, though!
How exactly is your servo AC power on handled? My Simodrive is sort of old, as you may see — the servos are 6SN1118-0AD11-0AA1 and the control board is 6SN1121-0BA11-0AA1. 

freemoore wrote:

I just had a very similar issue with a Yaskawa servopack drive on the Z axis of a Morbidelli. I tried adding `z-brake release delays via hal and had no luck. What eventually worked was finding a parameter in the servo drive that had the same purpose, something about keeping the control loop active so many milliseconds after the servo-ON signal or brake-ON signal was received. Perhaps your drive has similar? good luck!
 

Honestly, this sounds very reasonable and I've been trying to track such a parameter down for a while now. But my Simodrive is so old, it isn't supported by Simocom. I have to chug through manuals in the hopes of finding the correct one and the correct parameter. Been trying, but it's often conflicting information that I find, since Siemens don't publish specific manuals for a specific model, but general manuals for the entire product lineup... I will definitely keep searching, though.

It is probably worth noting that before we took out the original TNC 310 there was no such issue. So I would assume no parameters are corrupted or something of the sort.

RotarySMP wrote:

On my older MH400E, with Indramat drives, I made the mistake of adding a relay to give LinuxCNC co-control over the vertical axis brake. The original Maho design is that the Indramat axis driver solely controlled that brake. I had the opposite behaviour to you, where instead of dropping, the drive fought the brake during the period I had my finger on the machine start button.
Once I removed the 19K2 relay I added (second schematic in this post) the behavior worked correctly.
forum.linuxcnc.org/12-milling/33035-retr...0e?start=1740#340651

From the schematics you shared it I don't see the brake control circuit.Could it be that your brake control should be left to the axis driver alone?
Cheers,
Mark
 

Thanks for the feedback.
I haven't attempted to do anything like that. I'm not sure it is even possible for the brake control to be left to the axis driver itself. From what I see (on my untouched original DMU 50M too) the brake/release is nothing but an output signal from the controller to the drive. I don't think the drive itself is capable of handling such logic, or at least that wasn't the original machine manufacturer's intention. I will look into it, though.

Oh, also, my braking/release signal is a single one for all 3 XYZ drives. Is it possible to try and separate them electrically? Most likely. I'm not sure that's a good idea, though.