Unsteady plasma arc voltage Lotos LTP5500DCNC

[alangibson]

Hello all. I've just plotted the arc voltage on my Lotos LTP5500DCNC plasma cutter and I got an unexpected result. It swings rapidly between 160V and 110V, centered around ~135V. I expected the voltage to be much steadier based on how people normal talk about THC tuning. Does anyone know if these rapid voltage swings are normal for a plasma arc?

The Lotos LTP5500DCNC has a 1:1 arc voltage output. My oscilloscope is connected to that output. I also tried my multimeter. It reads 135 V. Lotos told me directly the expected arc voltage is centered around 120 V.

The work lead is connected to a star ground on the table. The plasma cutter chassis is connected to the same star ground. The star ground is then connected to an earth rod about 2 feet away. So interference shouldn't be an issue.

Any feedback is greatly appreciated.

PXL_20241009_200816078 (1).jpg

[weldguy]

I don't think that is abnormal, your multimeter will be smoothing it out and showing you a steadier voltage. Are you have torch height control issues?

[alangibson]

Hi weldguy,

The multimeter is definitely giving me RMS voltage. I've confirmed that with the RMS measurement on the scope. Both give ~135V RMS. What's thrown me off is range of actual voltages I'm seeing on the scope. I've never heard anyone mention seeing a 50V peak to trough difference in arc voltage. Of course that doesn't mean it's not normal.

The actual problem I'm having is that I can't find a target voltage setting in LinuxCNC (which has a built in THC) that doesn't either crash the torch or run away upward. Turning autovolt on, so that it decides the correct voltage for the THC itself, performs the same. My assumption is that it's because the input is so noisy.

LinuxCNC forums recommend smoothing the voltage with the lowpass component. I've tried that and it works to some degree. The issue again is that the input is so noisy that it's basically impossible to smooth it out without making the response time so slow that it's useless.

I think my next step is going to be to try to somehow calculate RMS voltage in a LinuxCNC component and see if that works better than the lowpass filter.

[tcaudle]

All plasma is noisy. The output is a PWM signal not DC . The normal PWM frequency in most plasma is int he 35KHZ to 50KHZ range so an LC filter with a 10KHZ roll off should work to give you stable DC. Where you get in trouble with plasma voltage is the workclamp is Positive voltage in relation to the electrode. You really need a filter on BOTH legs . Most of the "raw Volts " (1:1) are not direct, they are though 100K series resistors on both legs. Any type of external filter that does not take that into account will fail. They do that so if the output gets shorted fire and brimstone wont shoot out of the panel!

A beetler way is to connect at the work clamp and the Electrode connections inside and use a balanced LC (actually PI filter)

A 1000 uH coil in series and a .1 mfd cap to common on each leg gives a 10K hz 3DB roll off LPF . If you start with a 400V 1A full wave bridge then polarity gets taken care of and you filter the positive and the negative legs of the bridge, That will give you a nice clean varying DC at a frequency that is still plenty fast for THC response. Its important that you filter BOTH legs .


If you have your c AC grounds and the controller electronics to that same start ground you face an issue where that HF is getting into the controls. The table and workclamp needs to be connected to a LOCAL rod (within 10ft) to give it a noise anode. It will NOT get rid of all the HF noise stuff (depends on a lot of elements ) so if the workclamp gets connected to the control grounds you will have feed though problems. In short do not mix safety ground in your shop with plasma workclamp (which is positive voltage ) You CANNOT assume the plasma voltage will relate to you circuit ground so isolation is ususally required.

Even the divided voltage on CNC compatible Plasmas needs to be isolated .

MESA has a THC card that is isolated . It has to run on a floating voltage supply and it's signal pumped though an Opto isolation barrier.

[acourtjester]

What electronics are you using, you statement is LinuxCNC has a built in THC. I am using LinuxCNC with QTPlasmaC GUI with Mesa 7I76E card and a THCAD card which is external to the 7I76E card for the THC operations. I know there are other electronics that can be used but most have an external THC module . This deals with the arc voltage from the Plasma unit.
What you see on the scope may be ripple as the power out of the plasma may not be exactly a clean DC voltage.

[alangibson]

@tclaude

> The output is a PWM signal not DC

That perfectly explains what I'm seeing. I didn't realize the output was PWM. I'd like to think that would have occurred to me if it had been a cleaner square wave.

I'm going to add a filter to my board, taking what you wrote into account, and report back on results.

@acourtjester

> What electronics are you using,

I'm actually replacing a perfectly working Mesa 7I76E and THCAD with homebrew electronics on a new table I've built. I wanted to make one completely from scratch, table, electronics and all, as a learning exercise. One quick year later and I've almost got it done.

[acourtjester]

Slick choice, I have built 10 table from scratch using different electronics and software for the experiences, so have fun.

[tcaudle]

Outputs on most plasmas are IGBT modules and are PWm switched. Some units have some filter internally to make that less harsh (still not DC) Lower end units forgo any filtering or noise mitigation. I have a import combo Plasma/TIG/MIG that we use for noise studies . It can kill small animals at 50ft with the EMI and RFI it generates! Reminds me of the old MAX100 machines .....wow were they noisy! There are entire books devoted to EMI and RFI and sometimes wrong grounding can hurt you as much as no grounding. Plasmas operate in the lower end of the spectrum where EMI via a conducted path is more of an issue than RFI. Aany COMMON conductor can inject noise into another system .

[alangibson]

> Lower end units forgo any filtering or noise mitigation.

This definitely seems to describe my cutter. It's blowback start, but my monitor will occasionally go blank during piercing.

After I get this current problem fixed I'm planning to read up on how properly analyze what's going on with EMI and my grounding setup.

[tcaudle]

If your monitor does ANYTHING not normal when the torch fires that is a serious indication that plasma noise is getting into the PC . Its going to cause you problems. its possible ONLY the Monitor is affected (unlikely) but I bet that burst of noise is not coming through the air (RFI) . This where understanding grounds and how higher frequency signals travel though mutually connected devices comes into play. Get your plasma leads away form any control cables or switch wires and the unit as far from the controller as you can. Then unplug your monitor and controller from the AC and measure ohms between table and the PC chassis. Should be really high to infinity. If its not noise is going to be injected from the plasma workclamp to the table and into your PC and controls.

Consider that the plasma cutter is a Batter of 180 volts with a + and - terminals. Which one is ground? Answer: Neither. Its a separate power source and you can tie either terminal to an outside ground . In electronics we use the term "Ground" (GND: in several ways. You can have different "Grounds" in a system. They are separated to try and stop noise from getting into sensitive circuits. I prefer to call them "Common" or "Return" . Its not unusual to have Power GND, Analog GND and Digital GND all in the same control. Star grounds are great for things that need to be on the same common ground . The plasma voltage is like that battery and has its own RETURN in the unit ans it s NOT tied to the chassis ground in most cases.

Even two grounds in a building can be volts apart dependent on cable lengths, soil conditions and exposed surface area . If an AC safety ground is a few volts different than the main breaker ground rod its no big deal for AC stuff (safety) 5V wont kill you. 5 volts on logic level equipment can cause problems. SO you build a STAR ground to provide a LOCAL Common for all grounds that SHOULD be together . Introducing a bunch of noise into that star puts it on all of the connected devices.

While its important to ground the chassis of AC line operated equipment for safety, the plasma workclamp and torch electrode are not tied to the AC ground or plasma chassis . .

Anyway mull that over and think about the fact the workclamp lead is NOT tied to AC safety ground inside the unit. It is in fact POSITIVE voltage in relation to the Electrode leg. In a very noisy situation externally tying them together directly will buy you hours of frustration.

[alangibson]

I scoped my plasma cutter during a cut and it averages 500us peak to peak, so 2kHz. Resistance for each wire on the arc volt port is ~0 Ohm between port and either torch electrode or table. Makes me glad that I always turn the cutter off before messing with the arc volt cable.

Given
- input resistance of 47kOhm (voltage divider R1 resistor)
- output resistance of 10MOhm (for the LM358 op amp input)
- cutoff frequency of 1kHz
using this calculator
https://markimicrowave.com/technical-re ... sign-tool/
I get
- L1 = 8H
- C1 = 2nF
- C2 = 4nF

That value of 8H seems impossibly high. I must have misunderstood something with either the input or output resistance. It's also possible that dropping the filter between the voltage divider and unity gain op amp makes no sense for a reason I'm not aware of. You can see my current circuit in the attached screenshot.

The 24V_ISOL supply is a 500mA 24V isolated power supply that has its 0V_ISOL connected to the star ground. It is used for powering the op amp (since it's on the danger side of the opt isolator that comes later on) and for the Initial Height Sensing.

[alangibson]

> Then unplug your monitor and controller from the AC and measure ohms between table and the PC chassis.

I've got an unusual setup. I've got a Raspberry Pi inside of a plastic housing inside of a grounded metal control box. So from the RPi case to anything had better be infinity. The control box I'd expect to be short to the table because they are both connected to the same star ground for RFI.

> Anyway mull that over and think about the fact the workclamp lead is NOT tied to AC safety ground inside the unit. It is in fact POSITIVE voltage in relation to the Electrode leg.

I've been careful to avoid that. I got ahold of the Hypertherm recommendations a while ago and tried to follow them. I've attached a quick drawing of my various grounds.

[alangibson]

After some messing around, I came up with
L1 = 10mH
C1 = 2.2uF
C2 = 2.2uF

I'm using a function generator to create a square wave with settings of
f = 2kHz
Vhigh = 10
Vlow = 5

My scope shows steady, almost perfect 152mV output from the op amp. Almost perfect because the scope says the wave is 7.8 V RMS, and with a 50:1 voltage divider I'd expect to read 156mV out of the op amp. Plenty good enough though.