[Emc-users] Return of the wrong-way DIR signal

[Ed <ed....@po...>]

I fired up my notorious cam-cutting code and, after two 
seemingly successful air-cutting runs, chucked up a failed 
cam from last year to see what happened. Everything went 
well until the last Z-axis move that should have cleared 
the workpiece.

At which point the Z-axis motor gave off a -thunk- and 
proceeded to burrow straight down, through the fixture and 
into the tooling plate. Buried the cutter right up to its 
shoulder. Fortunately, Sherlines pretty much stall out 
rather than killing themselves (or the operator), so no 
damage was done.

I ran another air cutting session, which failed after the 
first few passes when both the Y and Z axes took off in the 
wrong direction. In this case, they were both making short 
moves, but departed in the wrong direction much faster than 
they should have, making rude noises all the way.

I'd reported this behavior on 28 November last year and we 
kicked things around in the ensuring days. To summarize 
that discussion:

-----------

I have observed incorrect DIR signals at the parallel port 
pins with both an external ("real") oscilloscope and 
HALscope, so the problem is not in the motor or controller 
and -is- in the EMC output. It occurs with nothing 
connected to the port, the mill spindle turned off, and no 
external hardware running.

The direction signal changes take place early in the motion, 
while the pulses are ramping up to traverse speed. The 
overall timings appear to be correct, but the direction 
signal flips for one or more step pulses.

The INI file sets lethargic acceleration and top speeds, so 
the timing parameters are correct and work OK in normal 
operation. Top speed is about 25 in/sec, accelerations are 
on the order of 2 in/sec/sec. I can -watch- the knobs spin 
up at the start and end of long moves.

I added step & direction pulse timings to the HAL file that 
set the pulse duration & rate well under the controller's 
ability. The direction setup and hold times are far longer 
than required (20 x 20 us each, in this case).

Other folks mentioned the behavior of steppers confronted 
with incorrect step inputs. In short, stepper motors can 
reverse direction and / or run at multiples of the correct 
stepping rate, due to resonances and other transient 
peculiarities that aren't faults of the motors or drives. 
That is exactly what I'm seeing here.

It is true that many folks are using CNC Sherlines with no 
trouble. It is possible that my setup is more sensitive to 
the incorrect direction signals than others. It is likely 
that anyone else confronted with this would simply give up 
on EMC and use something (-anything-) else that worked.

-----------

Back to the present...

I contend that EMC should not be flopping the direction 
signal around like that.

Evidently some code had been found that explains this 
behavior, but the change logs don't indicate it's been 
updated. Of course, I may have missed something, but, 
independent of any changes, the problem seems to still 
happen in 2.1.1, the Latest & Greatest for mere mortals.

I have not hitched up the oscilloscope to capture the 
symptoms again, but I can if that'll help figure out what's 
going wrong and get it fixed. I sent some scope shots in 
the previous discussion and I presume any new ones will 
look much the same.

Thanks...

-- 
Ed