Hypersensing with THCAD - better way to do ohmic sensing

Would you show the lifter on your torch that moves at 12-15m / min

snowgoer540 wrote:

I think rods .comp is set to ignore the ohmic sensing while cutting though.

Yes, it has to do that because the THCAD "sees" the arc voltage when cutting and depends on the overvoltage isolation of the power supply and the THCAD to prevent any damage from the arc voltage. Because we have the THCAD calibrated as a 24.5 volt full scale, this cutting shows up as a full scale reading instead of say 130 volt arc voltage. Well actually we know the THCAD's can always read a bit more than full scale so when the component sees > 24.5 volts it turns on an output to say its cutting. I can't see its possible to use Ohmic sensing to test for plate contact. It would be nice if it could be done. So we are restricted to using the float switch.

stivemaster13 wrote:

However, can you still explain exactly how you solved the problem with the sharp rise in voltage when going through a hole. As far as I understand you have worked on this. Hypertherm as you can see used for a limit of 12 volts (unfortunately it is not written for what period of time).
We used 15 volts for 300 ms (the same could be done with a sharp drop in voltage in a collision. Only here we need an active action - raising the torch)
In addition, in my opinion, an analysis of the curvature of the part is needed in order to be able to reduce the reaction rate of the axis. Otherwise, if we assume that our THC regulates by more than 5000 mm / min, then the speed becomes dangerous and means that a collision may occur after passing the hole. Simply because it will not be able to stop immediately.

Plasmac does something similar to what you suggest with the 12 volt threshold but in my opinion it is the totally wrong algorithm.

Instead one should use dv/dt (the change in voltage over the change in time). This becomes fairly complex to code. The first step is to be calculating a moving average in a way that does not require to use a loop in the interrupt service routine (our servo thread). If the number of readings averaged is large enough, the average is barely affected by a single high reading so it forms a good (straight line?) benchmark to test the increase in voltage against during a void (hole) crossing event. The dv/dt over a void becomes orders of magnitude higher than the dv/dt oscillation during normal cutting due to warpage or thermal distortion and the migration of the anode spot from the top of the sheet to the bottom of the sheet (that increases and decreases the arc length). So the dv/dt might be around 3000 during normal cutting and rise to 30000 or more just when crossing a 1.5mm kerf. (eg. an order of magnitude of 10 times).

So in my testing, it is possible to sense a void within 3 to 5 milliseconds and lock out the THC easily but the hard part is to enable it again. This requires studying a lot of oscilloscope plots to understand dv/dt behaviour. I got good results with this approach but just needed to tweak the final result when say doing a sever cut of a sheet and you run off the edge on the far side.

I don't experience voids when I lay out a sheet and Plasmac is pretty hard to alter its default behaviour in a way that is easily maintainable for future upgrades so I have not looked at it for a long time. I have asked Phill to add a seperate THC hold pin to Plasmac for third party use but so far he has been unwilling to do so. That would make extensions like this much easier to implement.

stivemaster13 wrote:

And is it possible to add a record of the cut distance from the consumable? In this way, it is very easy to keep track of the time to change. Is this a very valuable feature of branded Plasma CNCs?

Yes, Plasmac supports this and has a Statistics tab that records this and the number of pierces and other data for both the job and cumulative until it is reset.

stivemaster13 wrote:

Hey Rod, I'm constructive all the time, I hope you understand. Despite the strange English, I'm not picky.

I know that now.

tommylight, maybe I don't understand something. You mean that you regulate with PID a simple stepper motor Nema17 with 1500-2000 rpm?

Rod, you may not have read that, unlike Hypertherm, we monitor a 15-volt threshold for a time that changes according to the speed of operation!
In addition, since we rely on SheetСam to start and stop THC for the time being, it can happen that the voltage increases and that the THC is stopped too late. So I expect LinuxCNC to solve this problem.

rodw wrote:

The hypersensing method we've devised (which thefabricator03 is not yet using) seems to have solved the problem and its in use by quite a few now on several different brands of machines.

I find it really hard to upgrade a machine used everyday for production. Its the same reason that mines and power stations have shutdowns which are week blocks were the whole place stops.

In the past I have had little things that stuff me up from not reading the instructions correctly or just wiring something wrong that takes time to fix. At the end of the year we shut down for six weeks so I will get the Hypersensing (hopefully) sorted then.

tommylight wrote:

BTW i did push it to 68m/m but it started tearing the belts.

If you were to use rack and pinion and a quality planetary gearbox, do you think it would handle those speeds well and be reliable?

The difficulty comes from the fact that depending on the speed of work, the passage through the hole can take different lengths of time.

As for the lifter, I use far cheaper components that do not have to work in a mode beyond their capabilities. Motor with feedback that does not lose position. Tomorrow I will upload photos.