Anyone Actually Using External Offsets Branch for THC?

Yeah its a V-F converter so the noise will be +-1 count/sample so about +-.1% at 1 KHz servo thread
you can of course use the lowpass filter to reduce the noise (and this is pretty much necessary to reduce the plasma arc hopping noise)

John,

Something to look out for based on my experiences. If you are using the Hypertherms divided voltage, The THCAD-10 may be seeing your plasma cutter's voltage divider as part of its external scaling resistor so your arc voltage in hal may not be what you think it is. This caused my published divider to out by a factor of 1.5

I would apply your 168 volts to the Hypertherm's raw arc input to their CNC interface so the voltage divider is in the circuit and see what you get with a multimeter at the THCAD. Then measure the divided voltage with the input disconnected from the THCAD. If the voltage changes, then the THCAD is counting the Hypertherm's resistor network.

Alternatively, just use the voltage sampling component I've shared so the voltage reading at the start of the cut becomes the target and any scaling errors are no longer in play.

Rod

Thanks, I have actually got this handled. My old TD has no internal voltage divider or series resistor. My voltage divider mounted inside the TD unit is directly connected to the work and torch electrode terminals on the BIG PC board. My scaled arc voltage tracks the arc voltage or test voltage perfectly.

John

Hello

I'm back from the trade show and only have a couple of days for experimenting before I need to put the table back into production ( at my present rate of progress it will most likely be using the old Torchmate (Flashcut) controller).

Thinking about the apparent built in high gain my system has (sorry it's the best way I can describe what is happening). I am wondering what the cutting parameters any of you are using? I have a Hypertherm Duramax torch and mostly use two different sets of consumables for the cutting I do.

So for the Finecuts on 14 ga (1.905 mm) hot rolled A-36 steel:
45 A cut current;
220 ipm (5588 mm/min) cutting speed;
.090" (2.286 mm) pierce height;
.073" (1.854 mm) cutting height (74 V arc voltage) for best cut quality on my system ( book specs are 45 A, 200 ipm, .090" pierce, .060" cut (78 V arc voltage)). I know that the inches of height per volt of arc voltage is non linear but if we assume it is, .073" (1.854 mm) / 74 V = .000986"/V or .025 mm/V of torch height. Hypertherm recommends cutting within +/-.005" (.127 mm) of target cutting height. This is +/- 5.071 V on my system.

Now for normal 45A consumables on 1/4" (6.35 mm) hot rolled A-36 steel:
45 A cutting current;
48 ipm ( 1219 mm/min) cut speed;
.150" (3.81 mm) pierce height;
.060" (1.524 mm) cutting height (137 V arc voltage) for best cuts, same as book settings. Assumed linear arc voltage, .060" (1.524 mm) / 137 V = .000438"/V or .011 mm/V of torch height. The cutting range is +/- 11.42 V on my system.

The actual trial cuts I have made show that with low gain settings my system is cutting within the height specs required even with the oscillation of torch height (which will wear out the mechanics in short order) . Sorry for the rambling, I'm just getting back to working on this problem.

John

Hi Islander,

I read your story, but what is the probem?

I can sent you a Proma thc. It has a arc on,,, up and down signal. This works always.

If you want no problem's you can play on safe with your work. I integrated it a few weeks ago with the THCud signal in linuxcnc.
This was for me one step. But many step's to make.
The Z dro code is not updated, but is only on screen input as variable input value possible.
So that has to be changed.

If i can help you to do safe things, you can mail me.

You know that 45 amp value's can be adapted for 2mm to 10mm in a range of 5 volt's difference with Hypertherm inverters.
Your start delay time up to 1.5 sec. is mostly good. For 20mm thickness up to 2 sec.
You know that 65 amp's has the same character's. So if you device it in classes, it's easy and not the problem.

Grotius

Thanks for the offer but I have a working external THC. My only interest in LinuxCNC is the elimination of an external THC box.

I know it was a bit of a ramble but my post is about this thread using the external offsets branch for torch height control. I am curious about the operating parameters used by others when using the external offsets branch for THC. I don't have a Hypertherm power supply, only the torch. 99% of my cutting is at the two conditions listed.

John

Hello

I know this is a late reply, in an earlier post in this thread PCW mentioned plasma arc hopping. Attached is a copy of an older paper about this topic I found on the web some time ago. I don't believe this is causing the problems I am having because the sheet I am cutting is so thin and the frequency of oscillation I am having doesn't match.

John

@Islander

For my englisch :
Is hopping defined by going up and down during the cutting path?

Then is the solution quite simple. Slow down the reaction speed of the z - axis of the thc external software.

In the AD torch height system i think you have 2 input parameter's they are written by JT. So i know. It's a file you must compile if
you want to use your own one.

If it's an other problem, let it know. We find out what the problem is.

islander261 wrote:

Grotius

I know it was a bit of a ramble but my post is about this thread using the external offsets branch for torch height control. I am curious about the operating parameters used by others when using the external offsets branch for THC.

John

John, I think you and I are the only ones who have tried it. I found a bit of time last weekend to try an experiment but got disillusioned by the results (but no cutting and did not even turn the plasma on). Its on the agenda still though. on thin material (I did a lot with 2mm plate), I don't think the anode spot cycling from the top to the bottom of the plate is an issue. The signal was a bit noiser on 8mm cuts.

Just to clarify what this thread is all about it is based on this branch of Linuxcnc
github.com/LinuxCNC/linuxcnc/tree/dgarr/external_offsets
more particularly live systems built on the hpid simulation which is accessible from the LInuxCNC browser once this branch has been installed. This does not require any external THC hardware, just a way of reading the torch voltage (we use the Mesa THCAD card)

The eoffset_pid component's contribution to open loop dc gain is the pgain pin.

Changing other system variables may alter gain but it is simpler and less confusing to characterize a pid implementation's dc gain by the pgain term alone.

The k pin is made available to set the error quantization -- excessively small values would be expected to be destabilizing,
excessively large values could lead to numerical instabilities.Changing the k pin value does not directly change the component's dc gain provided the expected connections are made for axis.z.kcounts. and axis.z.kreciprocal as handled by LIB:hpid.hal.

Please refer to the documentation
github.com/LinuxCNC/linuxcnc/blob/dgarr/...external-offsets.txt
github.com/LinuxCNC/linuxcnc/blob/dgarr/...eoffset_pid/hpid.txt
and $ man eoffset_pid
As an evolving experimental branch, this can change so keep your system current

If anyone does decide to play with this, please migrate your working config to the HPID example per the included instructions. If we are all using the same base code, fixing things will be easier.