THCAD Noise / THC comp movement

Hi there,

First i'll want to thank you in advance, not only for comments here but for the support work you are doing on forums, documentation and everywhere.

I've been working with 5i25-7i76 + THCAD10 for a while, and i have a "problem". I have seen comments and searched for a time, but still have doubts and need to improve the performance of my systems.

As recommended by PCW, i'm using /32 on THCAD Board, and i'm applying lowpass filter with a gain of 0.02 To reduce noise.

I have now configured THC-velocity on 0.01

The problem is that Noise is still affecting performance. I have tried few things:
1.- Reduce lowpass filter gain to 0.01 or lower, it helps a lot, but any voltaje change takes too long, and system can't react on time.
2.- Reduce THC-velocity, to 0.004 or even lower, and it stops bouncing (going up and low all the way when THC is enabled), but speed seems to be too low for adjusting on irregular material.

What i would like to achieve is to cut Diamond steel plates, and/or corrugated steel sheets with the system, as it does CandCNC and other manufacturers with their own hardware.

Maybe i'm looking for too high performance for this system, in wich case i'll like to know also.

One thing i have noticed is that even when Power source is off, there is a little noise (that can be effectively filtered with a lowpass of 0.1, so nothing crazy), and when Powermax system is ON, even without cutting, the noise increases greatly. What i have noticed also is that this noise seems to have some kind of regularity, it goes up and down with a more or less stable values, and it seems that same noise is affecting reading when system is cutting. This seems to happen on all the Powermax systems i've tested (3 by now)

- It'll be possible to create some kind of "online" filter, so on start noise can be read and then deducted from readings ?
- Any other type of filter, wich doesnt affect or reduce voltaje reading precision/speed (as it does the lowpass)
- I found also BigJohnT THC with Pid (Where i only see a P gain value). It'll help, if i'm right, on THC adjusting speed because of offset, but not on noise.

I would like to read from THCAD on 0.2V precision (if it gives 10 bit precision, with 20:1 voltaje ratio, reading from 0 to 200V) every 1 ms (as its on servo thread on regular timing), and move torch up and down as fast as 0,08 mm every ms, so i can go through slopes of up to 60% at max speed (8250mm/min on thin material with Finecut)

I'm using Yaskawa servo motors with 1000mm/s/s acceleration on all axes.

Any comment will be welcom.

Best Regards,
Azeri.

The THCAD will have about 10 bits of resolution at a 1 KHz sample rate

if you low pass filter this with a 0.1 setting, you will have
approximately 11-12 bits of resolution so I dont think resolution is your issue.

If you only see excess noise when the plasma power is enabled, this suggests that
the plasma power source has significant noise and or ripple ( for three phase power
the ripple frequency would be 300 Hz on 50 Hz mains and 360 Hz on 60 Hz mains )


Can you plot the noise with halscope so (and indicate the scale)?

How much Z axis motion does the noise cause (even a few volts should not cause much motion)

Thanks for fast reply PCW.

Attached 2 Halscope logs from encoder velocity on /32.
The OFF one is with powermax OFF, the other one is with power source ON.

Hal scale 1
With a thc.vel-scale of 0.007523216 its an up/down of little bit above 3V of noise on real data.

Best Regards,
Azeri.

What DC voltages do these represent? (3V ripple is certainly possible on 200V but would not be expected on a much lower voltage)

Is it possible that you have noise introduced into the frequency data?

That is, is the encoder filter enabled and are you using differential signaling on the encoder signal?

A Halscope screenshot would probably be more useful and more helpful to other forum users as well

Looking at the numbers again this seems to be at 0V, so a couple of things may be at play

1. The plasma voltage may be noisy when under no load (the noise looks like a square wave)

2. You need to make sure the 0->10V leads are shielded and kept well away from noisy signals
3V of readout noise is only 150 mV of pickup on the 10V leads

Dear PCW,

its 3V variation on 0V reading, and same/similar variation when cutting.
Yes, its at 0V i mean, just with system ON and powermax ON, you can see the differene when powermax is OFF. When power source is off, noise is really small i can handle this variation on readings, but when power source is ON, the noise goes up a lot. When system is cutting, noise has same pattern as when power source is ON at 0V (torch off), but instead of going 0-3V it goes 78-81V when cutting finecut on thin material, so torch bounces up and down.

Video con Gauge 16: sendvid.com/w0jxai4a

Encoder is differential mode on 7i76 with W4,W5 and W6 on right hand position.

Sorry but i cant see anything on Halscope to take screenshot of it, just see values on Log, i don't really know how to use it, this is my first time with it.

1.- Yes, i also supposed that it was only on 0V, but same pattern applies when cutting.
2.- All cables are shielded. If i put thoose cables with AC noisi cables near problem got bigger, i learned that on hard way . If it where about cables, noise should be there also when power source is off, what happend when cable is not shielded and/or near noisy cables.

This could also be a grounding issue (where the low side THC sense voltage comes from)

Are you using the plasma power supplies built in divider or your own?

The noise looks like a ~4 Hz square wave, can you measure this with a standard voltmeter?
This could be a beat between a higher frequency noise source and the frequency sampling

Also make sure the filter bit is enabled on the 7I76's encoder input (and with the /32 mode you can also set the
encoder sample frequency down to 2MHz or so giving a 7.5 usec noise filter on the encoder inputs)

One thing to try would be re-routing cables while measuring the noise with halscope

Dear PCW,

Hypertherms divider and reading from CPC connector.

I don't know how to measure a square wave with voltmeter.

I moved, changed cables and even tryed with external 7i76 and THCAD, this last one was the reading i sent here, the nearest cable is the 24V DC of the 7i76 and its separated as much as possible (Encoder and VField connector distance on 7i76 is the nearest).

If the variation is really 4 Hz, it should be visible as variations on a voltmeter reading

Did you enable the encoder filter?

Hi Tommylight,

THCAD is on electric panel, near de 7i76, more or les 2 to 3 meters from plasma source.
All shielded cables grounded on one point on electronic side only. Cables are apart from AC source and torch/table cables.

Maybe i'll need to try different DC power source. Computer is like 20cm apart from Mesa electronics (except 5i25 wich is inside). DB25 cable goes alone without any other cable touching it.

I have tested with Multimeter power source, and you are right, measures change, so problem is not THCAD, but plasma power source.

I made some tests, with THC component disabled, at fixed height cutting 3/8" and 1/2", and what i have seen is an stable voltage reading (around 0.2-0.4 real V variation, or even less), so it seems that bouncing that can be seen on machine is not caused neither by THCAD problem, nor power source variation.

I was focusing only on seeing that variation was "similar" on power source ON (0V) and cutting, but it seems different. I'm thinking that maybe, too fast movement on THC component is creating some kind of overshoot to Z axis and this is what creates Voltage variations...

THC Comp just adds an offset (that can be negative if voltage is too high), and creates new Z position, wich is linked, in my case as:

net z-pos-cmd thc.z-pos-in <= axis.2.motor-pos-cmd
net thc-pos-cmd thc.z-pos-out => pid.z.command
net z-pos-fb <= thc.z-fb-out

Maybe if THC velocity compensation is too high, offsent sent to PID.Z is doing it slower than expected and its overshooted in wich case, the Lowpass filter at 0.02 gain makes it worse, as because of lowpassing, voltage change goes latter to THC component...

Is there any way to log lots of variables for a long time ? i have seen that halscope has his limits, 1 pin can be logged for long time, but as you add pins or signals, time goes down. It would be helpfull to see how Voltage, THC com offset, and PID pins change over the time...

Best Regards,
Azeri.