G95 feed per revolution

[cncdrive]

I now can't find the forum thread where I posted about this.
It is a motion controller with breakout board combination which can output both step and direction signals and also +-10V signals for analog servo drives.
The encoder signal has to be connected to the board and a PID controller (per axis) has to be setup to control the speed loop.
Then the system is fully closed loop.
But you can also mix step/dir stepper and servo drives (in open loop) and analog drives (closed loop).

I found only one picture now of the board and it was only half assembled at the time.
I'm attaching that picture.

The board looks so "clean" only because all small SMD parts are on the bottom side.

[cncdrive]

And this very short video which my collegue made when the Xilinx FPGA on the board first started working.

[Robertspark]

I have just talked to my collegue about the G95 and he said that in mach3 it is planned with at least 100msec delay, so I think the mentioned few tens of milliseconds should be fine if it is OK how it works in mach3?!
If so then it is not that complicated to implement this, because then everything can be planned on the PC side, no need to modify the path calculations in the motion controller.

Some delay will be inevitable / expected, the key element is to get some benefit in utilizing the spindle speed feedback signal to vary the feedrate so surface finish can be improved

I may be wrong here as I've not done any calculations, but to my mind all we are talking about is a compensation in feedrate relative to "minor" spindle speed variation, it will be minor because spindle speed varioation that it is trying to correct for will be small within the sample period, but at least it uses some amount of feedback.

[Robertspark]

The problem with teh Mach3 version is The function with using the spindle index signal does not work properly

Why does it not work properly?

I expect that uccnc will use the encoder feedback signal from the spindle.

so to gain the benefit of feedrate correction relative to actual spindle speed you need to install a spindle encoder

[Robertspark]

The feedback to teh G95 is broken in that Mach3 looses control over teh changes and it does NOT obey teh Acceleration profiles. Z gets in a runaway condition at times and either looses step or trips an error on servo drives.

Well we should not have that problem with UCCNC because it does obey the acceleration profiles

[cncdrive]

I don't know if it works properly or not in Mach3 since I have never used it.
What my collegue said is that how Mach3 internally works it does have a 100msec delay with handling the spindle speed feedback and acting on it.
So, it simply can't be faster than that, it must have a 100msec delay and this is the best case scenario. Again the reason is the internal loop speed of Mach3 which handles the feedback signal.
What I don't know is if this is fast enough or not.

Ofcourse the UCCNC will also not be able to handle this in realtime unless we will handle it directly in the motion controller which I'm not sure if it is even possible.
So, if it will be OK to have upto as much delay with this as it is with mach3 then we can implement that on the PC side like how it is in mach3.
We can set our loops faster so it will be probably faster than in mach3, but still there will be some delay.

And Terry is right the overal delay will be lower if we use encoder feedback and not index only even if the loop is the same speed as in mach3, because if only index is used then the system is always blind for a full 360°and that is again a delay in the signal...

[Robertspark]

How does it work in mach4?

[cncdrive]

I have no idea unfortunately, it may be the same slow or a bit faster.
There can't be a miracle, because there has to be a loop which handles the signal. The loop can be faster or slower, but it will be never realtime.
The only way to make this realtime is if the motion controller calculates the path or if it has a way to modify the precalculated data, but it is probably not even possible, because it would shift the time slices of the precalculated data which would cause a problem and probably that can't even be resolved.

[cncdrive]

If the controller handles it then there is no need for Mach4, because then the motion controller plans the motion.
So, I doubt that if someone handles this in the motion controller then they will implement into Mach4, because then they will just have to make a simple user interface and they are good without mach4.

[cncdrive]

No, I don't think it can be done. It is not done in realtime in mach3 and in mach4 too.
But there is no such thing as "realtime" though since electrons need time to flow. Even light needs time to travel.
And so because you cannot measure rotational speed in zero time and you cannot transmit that data in zero time it is impossible to make it realtime.