Proma THC SD or 150 discussion

[bLouChipBlacksmith]

I'm looking for / hoping to exchange usage experience and ideas with folks that are using the Proma Compact THC SD or 150 models. I'm using the SD model on my table with a grbl v1.1i 4 axis controller. The DANI THC7SDP model is clone of the SD I believe, so let's include experience with that brand/model also.

cross linking to details in Sheetcam forum here: https://forum.sheetcam.com/viewtopic.php?p=31408#p31408

thanks,
Lou

[DieselFumes]

I hope this discussion takes off, I have been interested in trying one of these but have read of so many difficulties getting them up and running properly. They are affordable enough that cost isn't the issue, I just don't have much free time if any to mess around with things and sometimes things like this can consume a significant amount of time when the project goes sideways.

[bLouChipBlacksmith]

Since there is no interest yet over in the Sheetcam forum I'll report some findings here.

I spent a few hours testing at the plasma table today, attempting to fine tune some of the THC settings. I was able to make some headway into mitigating 'THC induced Z error' (THC-IZE) as defined in my OP. I certainly discovered trends and had predictable behavior as I made minor THC setting adjustments.

Bottom line: I'm getting 0.223mm deviation on average of gap adjustment per Volt setting adjustment on the THC. The behavior seems to be more stable and consistent than it has been in recent past. I'm happy with the settings now, and I'm happy with two active mitigation actions that I can apply during a cutting job. These actions are largely independent of each other:

action 1) Zero Z axis at the first XY pierce coordinate before starting the job, and record the machine coordinate of the Z retract position (Sheetcam's safeZ) at that location. The change in Z machine coordinate at safeZ can be tracked shape to shape while cutting, I just pause the job during a rapid motion at safeZ between shape cuts (using a pushbutton on my console), and look at the Z machine coordinate in my gcode sender (UGS) digital readout (DRO). The direction and magnitude of change in machine Z indicates correctness (or not) of THC V setting. If machine Z coord is increasing (getting less negative) then THC V is too low, raise it. If machine Z is decreasing, THC V is too high, lower it. In my case, 1 volt of adjustment will change the machine Z trend by 0.223mm on average per shape cut. If 1 volt of adjustment doesn't arrest the trend, don't be afraid to bump it another volt. This can be done with the machine/job paused or not.

action 2) As mentioned in my OP, I have recently begun planting Z axis homing cycle commands in my gcode jobs. How often this is done depends on how fast THC-IZE is accumulating per shape cut. I have 25mm + and - from my WCS Z zero at start of job before it will throw a Z soft limit alarm due to THC-IZE. So if I don't manage THC-IZE at all per action 1 above, then worst case I have 0.3mm (rounded up) of error accumulating per shape or about 75 shapes before my machine throws an alarm. I already have a machine pause in my gcode jobs set for every 4 minutes of runtime, to allow for my undersized compressed air supply to catch up on tank capacity, so I just added the Z axis homing cycle command ($HZ) at this event. It's working out very well, so well that I really don't need to take any action at all per #1 above.

action 3) This doesn't apply to me because I'm stubborn... if I were to change my grbl configuration to run with soft limits disabled ($20=0), then there would be no soft limit alarms to interrupt my jobs. However, I prefer to run with the crash protection. But know that if you run soft limits disabled, then THC-IZE is a non issue, your machine Z coordinates are, well, limitless. But I've had bad things happen under this circumstance, like Z running off the rails due to diving into open water, so live and learn.

Here are the raw test results and test case I used today.

Screenshot 2024-02-25 232600.jpg

Screenshot 2024-02-25 231230.jpg

The red X in the table photo is where I take the probe readings before and after the test cut sequence of 7 open shape rectangles. I WCS zero Z at this point via torch probe cycle before the cut, no flex in sheet metal during measurements and cutting, as the surface is held down with a 8 lb weight. Then after the 7 cut sequence I clean off any top dross, clear the torch tip of any blowback dross (90% of the time there is none as the PMX nozzles are awesome), and then take another torch probe cycle measurement on the red X. The "Accum Z error, mm" in the spreadsheet is the difference in these two measurements; a negative value means the machine Z reading got lower during the 7 cut sequence, positive means it got higher. This is THC-IZE accumulated over these 7 cuts; divide by 7 to get the average THC-IZE / cut.

The 2 right angle corners in the cut are there on purpose, to make the THC Z Anti-Dive do its job, and as I watch the Z axis motion during the traverse through the corners the ADC does it's job perfectly, block the THC Z signals as the XY comes to a stop and starts motion again at 90 deg. The blocking time is a blip, perhaps a 50-100 milliseconds, but its visible, and I also have an LED on the circuit to indicate blocking is active or not.

The sliver of metal left inside the rectangle is by design also, I want to minimize the test cut area so I can run many cuts in a sequence, and many sequences, but also give the plasma jet enough metal to eat 100% in the kerf but NOT enough metal to heat warp up or down and influence the arc voltage.

IMG_2676.JPEG

IMG_2674.JPEG

[bLouChipBlacksmith]

Plasma THC Z error diagram.jpg

[acourtjester]

Looks like you put some time into figuring things out, great you posted the results.

[bLouChipBlacksmith]

Thanks Tom. Well, some of these discoveries I stumbled into over a year ago, and was able to get the THC SD dialed in pretty good for what I was doing with plasma at that time, which was mostly cutting < 40 shapes / job, mainly because I was spending a lot of time double checking too many parameters and conditions than I care to keep track of. Now that I have my scpost complete, tuned up, and very automated, I take on arts-crafty designs and jobs with 2x and 3x the number of shape cuts, and I don't put much time at all into checking or double checking anything but bridging between kerfs, its mostly draw, turn, and burn I like it like that SO revisiting the THC-IZE is one of the last cruxes of the plasma process for me, at this time. It was worth a few days of deep dive to get my head around it. I have piece of mind now. If I can help others with it, I glad to do that. These standalone THC units have their purpose, but also their limitations. Hopefully I'm able to live with the limitations now without putting much ongoing thought into it and being concerned that it may louse up a good cut 80% into the job. It's full on automatic again, I like it.

If there are better ways to work with these THC units, that's what I want to discover and why I started this thread. For the bottom line though, the quality of the plasma cut with these units is very acceptable to me with the process I'm using.

take care,
Lou

[acourtjester]

Interesting that you are using GRBL for your controller, does SheetCam have a post processor for the GRBL. I think some of the MPCNC guy ae using GRBL. I had used it for a laser setup with an Arduino hooked to my table, I was just doing simple things turning the laser on/off. I see some are doing some advanced things like etching images.

[bLouChipBlacksmith]

Sheetcam ships with several grbl scposts, each is specialized for a cutting process (milling, laser, marking, or plasma) and then again some or all of those have rotary axis flavors, and the plasma types have THC or not. I'm generalizing, but that's the gist. I picked one that best fit my basic plasma and milling needs and customized both. The most customized flavor I have is for plasma, "GRBL plasma LDC w Rotary.scpost". The latest version of it is always available at the bottom of this thread: https://forum.sheetcam.com/viewtopic.php?p=31407#p31407
For my laser process on the same machine, I use LightBurn CAD/CAM/ and sender.

Code: Select all

OnAbout()
===GRBL Plasma XYZA post processor, for 3 or 4 axis GRBL; .scpost Version: 26.6
 -Operates WITH or WITHOUT the Rotary-plasma plugin.
 -Operates WITH or WITHOUT a THC.
 -Rotary Z0 at object surface, NOT rotary axis center of rotation.
 -Assign XA, YA, or YB as linear/rotary axes for gcode motion.
 -Machine Pause for both air supply catchup and parts removal.
 -OnDrill() support for pecking or piercing WITHOUT tool change.
 -Supports Marker Tool / Scriber.
 -Arc as Moves (G1) or Incremental IJ (G2/G3) gcode motion.
===== External Post Variables available:
LdcTorchProbeTravel -- floating torch mount float distance to activate probe sw
LdcThcPresent -- Is THC present on machine? Y=1, N=0
LdcThcEnable  -- Enable THC dynamic control? Y=1, N=0
LdcDrillPdPct  -- Plasma Drill peck tuning, % of pierce delay, 0-100
LdcDrillChPct  -- Plasma Drill peck tuning, % of PH to CH distance, 0-100
LdcAirSupplyTime  -- Continuous cutting air supply time, in minutes
LdcPauseOnNewPart  -- Machine pause to retrieve finished part? Y=1, N=0
LdcArcAsMoves -- 0 for G2/G3 arcs, else is line segment length for G1 motion
LdcBenchTest  -- Y=1, N=0, used for testing on grbl without a physical machine, default OFF

or use the 'Set custom post options' button after loading the post and re-entering menu Options->Machine->Post processor

[bLouChipBlacksmith]

Great news !! I made an adjustment and the THC-IZE got smaller !!

After running a few more test cases, same as those described in previous posts above, for the purpose of observing and studying closer the Z axis motion during XY cornering where the ADC is blocking THC STEPs, it was definitely noticeable that more Z axis rotation occurred on the way into the corner (during deceleration) than on the way out of the corner (during acceleration). That stands to reason due to deceleration STEP blocking engaging at < 85% of feedrate speed, and acceleration disengaging blocking STEPs at > 95% of feedrate. That is by design, to have a reasonable gap between the two thresholds in order for the circuit to work reliably, else the timers would race each other for the correct output state. So I did the math and found that the decel time to begin blocking from full speed of 1600mm/m was 10ms, while the accel time to unblock state was 3.3ms before full speed attained. Still, I would not normally think that would be a problem because the job of the THC SD is to feed STEPs to the Z motor in the correct DIRection until the arc voltage = the monitor voltage, and 48mm in the shape cut is plenty of distance to make that adjustment after the last XY corner is traversed. Its a closed loop system, move Z until voltage compare is satisfied ! And the THC DRO showing arc voltage agreed with that theory during all test runs.

So I wrote to Proma to inquire if there was something in the THC MCU program (assuming its using a micorprocessor to manage all this) to explain this and I'm still waiting to hear from them. In the meantime, I thought 'what the hell', just close the gap between < 85% and > 95% decel and accel thresholds and try it. So I set the decel threshold to < 94% and the accel remained at 95%, HOLY COW ! MAJOR IMPROVEMENT !!

I'm still perplexed why there is such an improvement given it's a closed loop system, but I'll take it ! Perhaps Proma will explain.

So bottom line: THC-IZE is now measured at 0.1mm / v best case, 0.15mm worst case, per shape cut ! That's a factor of 2 improvement.
And the over and under per volt correlation to THC-IZE distance is spot on equal ! I can't ask for better than that.

Screenshot 2024-02-27 184524.jpg

I almost forgot, 1 more picture to post...
btw- this was after the testing, but while writing ...
Salute and Enjoy !

IMG_2693.jpeg

[kicktillmonday]

Wow! Dialed in! Very nice work.

[bLouChipBlacksmith]

Thank you. It was fun. Hopefully there are some takeaways for others using this style of THC unit.

No word yet from Proma (in Poland) on the seemingly odd behavior of the THC not being able to normalize back to the correct cut height (to better than 0.25m) in a straight run of 48mm. Very odd.

If I had to guess, the THC microprocessor firmware has some sort of anti-dive algorithm that kicks in when it detects a positive arc voltage spike, such as what occurs in sharp XY cornering, and in that case it tries to keep the STEP count symmetrical in and out of the corner, but my digital ADC was originally hijacking STEPs in an asymmetrical in vs out count. By closing the asymmetrical (<85% in, > 95% out) thresholds of the ADC to < 94% in and > 95% out, that remains within the margin of STEP count symmetry that the THC is programmed to handle. I'm just guessing. I know that before I even built and put the ADC in play, very short straight line segments, such as those seen while cutting text letters, were causing a lot of THC-IZE, and that is why I built the ADC in the first place, not to mention the quality of those text shape cuts were terrible. They have been much less terrible since using the ADC, even with the former 85-95 gap. I will cut a job later today with text shapes in it and will have something to compare soon.

I updated the ADC schematic and reposted to the original link:

[bLouChipBlacksmith]

update:
I've been corresponding with the Proma THC SD developer in Poland for the past several days. He clarified that there is no THC built in anti-dive algorithm in his unit, nor is there an algorithm to detect XY cornering and thus no attempt to equalize STEPs into and out of a corner. So I can forget that speculation. Although he doesn't explain the anomaly I experienced in the testing posted above, I just sent him a video of the tests in several plasma cut demos, so there may be more to come. However he did seem to frown on my ADC acting to block THC STEPs during XY deceleration and until acceleration > 95% of feedrate. But it works. I was able to glean from him that the unit can be problematic at times, and disconnecting the arc voltage sensing was his recommendation. That's just doesn't seem practical, I've asked him for clarification on that.

In the meantime, it occurred to me that since he confirms the unit is microprocessor based and as such its running program logic, then when I perceive it to be behaving oddly and illogically (as I have many times in the past 2 years), then my course of corrective action will now be to Pause the CNC plasma job during a rapid motion between shape cuts, and reboot the THC; this is accomplished simply by power cycling the unit OFF/ON, it takes 3 seconds. It has no power switch built in, so I installed a simple toggle switch on the 24vDC power input line and mounted it right next to the unit the other day, so bingo, reboot as needed. Not a problem. So far so good. I haven't had to reboot while running a plasma job yet, but I have rebooted between jobs and in special one-off situations. One of the use cases that I believe throws it into an illogical behavior is when I block the STEP signal for an entire plasma cut, which can be 20 to 40 seconds in duration, in order to get an arc voltage measurement from it to use as a reference setting during THC active control use later. So the past couple of days I've simply power cycled (rebooted) the unit after I made those measurements. I'm making them a lot now since I dialed down the arc voltage calibration setting just prior to the recent testing, I have to recalibrate for all my sheet metal gauges as a result.

Two years ago I discovered a bug in the THC SD unit that concerns its DIRection setting for Z up vs. down motion, the setting take the values 1 or -1 depending on how the Z motor coils (A and B) are wired to the driver. My machine was wired for -1 setting, which apparently is not typical. The -1 setting had a little side effect (aka bug) in that on the first arc detection after a power on cycle, the THC would move Z in the opposite (wrong) direction to correct arc voltage/gap while cutting. Yeah! Drove me nuts until I figured it out by video recording. I couldn't believe it. When I contacted Proma, he confirmed it had been reported before; the fix was to reverse motor driver polarity on one of the 2 coils in the Z motor, that reverses its direction, then change grbl config for direction of Z, then use DIR = 1 in the THC SD, thus avoid using -1. So what I just went through was not the first time a power on cycle sent the unit into la la land.

I read somewhere in this forum a post by Tom (@acourtjester) that there are reports from some Proma THC SD users that it can be troublesome to setup or dialin. I concur. But, it seems like it can be done with some persistence. I'm sticking with it for now, recall that I'm self admitted to being stubborn

[bLouChipBlacksmith]

here's a video of the latest test cases to demo correct operation of the THC SD, and the need to monitor the accumulating THC-IZE from shape cut to shape cut in a job, then adjust the THC monitor/arc voltage setting to reverse the THC-IZE trending direction. This video uses a test case pattern different than what I use a week ago and posted above, this test case has many zig zag motions, very short line segments limiting straight cut length in order to make it difficult for the THC to correct torch height after a corner, so intentionally stressing the system.
Net: I'm happy with the test results and with the improvements added by the ADC.
Caution- watch with a strong cup of coffee else risk nodding off...zzzzz...zzz...
Enjoy.

[cuttinparts]

Very in-depth and interesting testing you’ve done there. So at the end of it all is this THC one that you would recommend to others?

[bLouChipBlacksmith]

@cuttin- Good question. I've discovered enough at this point that, with limited exception and IMHO, I do not recommend a THC that operates autonomous to the CNC controller. This is somewhat concluded in my original post "... if I build another plasma table, I will use a THC that is either integrated into or well behaved with the CNC controller."

The exception being: IF one has a CNC controller they really love, or they don't have the means to replace it, then adding an autonomous THC to the system is a reasonable and low cost option, and this unit is likely as good as any other. And as noted in the 2/25 post above, there are actions you can take to mitigate THC-IZE, even without an ADC. Perhaps adding to this exception is that it is best if the part profiles being plasma cut are consisting of shapes that have at least 25mm or greater straight line segments as opposed to zig zag short line segments, and where cuts begin and end on those longer straight line segments. This gives these THCs time to recover Z to correct cut height after diving Z while XY traverses a corner.

Ideally though, if the means exist, you'd want a CNC controller with an integrated THC, so the controller is always managing the Z motion and tracks its position relative to Z home, machine Z0. Then there is no THC-IZE. The Anti-Dive in a system like this is managed in part by the controller firmware or can be embedded in the gcode or both. Sheetcam for example has Path Rules to disable and enable a THC on corners and on other conditions of the tool path.

My ADC uses THC STEP blocking to 'disable' it, and works in conjunction with gcode and Path Rules, but Proma THC SD is fighting against it. I have more testing and tweaking to do using the more realistic zig zag test pattern, and I'm now corresponding with the Proma developer. I'm hopeful that the THC-ADC combo can be configured to work nearly as well as a CNC controller with integrated THC. I've now got gcode to measure the THC-IZE after each shape is cut in the zig zag test pattern, while I fiddle with the THC settings.

[cuttinparts]

Appreciate the feedback Lou. This is a great topic, very helpful.

[bLouChipBlacksmith]

update: I believe I'm ready to close the book on this one. After stepping back for a few days and giving some thought to the prior results and open questions, I decided to take a more methodical and automated approach to the testing process and data collection for analysis away from the shop, rather than trying to analyze real time while also running the machine. The new approach has worked well.

I made some further tuning adjustments to both the THC settings and to the ADC timers, which I may have crossed up mistakenly in the prior round of testing that yielded some crazy results at times. The ADC timers are now adjusted to block THC STEPS at XY speed decelerating when <93% of feedrate, and UNblocking during XY acceleration when speed is > 95% of feedrate. This is for automatic blocking/UNblocking. The circuit also allows for on demand blocking/Unblocking via gcode stmts. All of this should be explained in the diagrams and annotated photos in the zip file. And the .nc files have the gcode in practice as well.

bottom line: the Proma THC SD is responding just fine to the ADC STEP blocking to prevent diving while XY traverses a corner, and when XY comes to a stop such as at end of cut. There certainly are some fine lines of adjustment via THC settings though, the most critical being SPD (or frequency of STEPs), which would stand to reason that it may change over the full range of XY feedrates on my machine. I ran these tests at a mid range feedrate of 1440mm/m, cutting 16gaHR at 30A with a PrimeWeld Cut60 using a UPM105 torch and PMX30XP unshielded tip. But I'm setup to easily dial in another SPD value if/when needed.

Following are results and two short 3min videos that demonstrate the testing process and test patterns.

zz pattern v2 R5 measurements log.jpeg

[bLouChipBlacksmith]

Correction to the process description panel in the 2 short videos.

Zerror measurement process annotated.png