Input Analog Jittering

[Tinine]

'Scuse me but masking the actual problem is "engineering" to you guys?

"Hey, I hear a weird knocking noise from this car engine"
"Yeah we know about it...the fix is to increase the radio volume and you'll be good to go"

 

[drbitboy]

Did find out that the stable North and South Tempo is a SSI. I tried the logic don't know if I actually did it correctly?

1) What value in what tag are you trying to stabilize? The filter code you posted (post #11) seems to be filtering the raw value of tag out_r and putting the filtered value into tag out_f, while it looks like the actual raw value is in tag east_tempo_whatever and is unaffected by, and has no effect on, the filtering algorithm. Replacing tag name "out_r" in the CPT instruction with tag name "East_tempo_whatever" would make the value of tag out_f be the filtered value of East_tempo_whatever.

2) what is the character of the "jumping around" mentioned in post #1? Is it process noise, measurement noise, or something else? Post #1 used the phrase "rested state," which would seem to eliminate process noise. Any measurement will have noise/error, so investigating, and hopefully eliminating the source of excessive noise as suggested by others (wiring, power supply, grounding, emr induction, etc.) should be the first line of attack.

2.1) if the source of the noise cannot be eliminated, then if the noise is stochastic/gaussian/random, a first order filter only attenuates the noise but does not eliminate it, as can be seen by looking at the formula/algorithm. A periodic or moving average will probably provide a better, and more accurate filtered result. Caveat: the size of the average sample set affects the response time of the filtered value when the raw value is actually changing (increasing our decreasing) to a new mean level.

3) why is the "jumping around" not acceptable? Is it the perception of people looking at a display of the value? Is a PID or other loopback control using the fluctuating value as its PV, and the noise is amplified by the controller chasing the noise?

 

[Peter Nachtwey]

I have a lot of experience with Temposonic rods. Temposonic usually has Delta Motion test their rods out before releasing them to the public.
I don't like analog Temposonic or Balluff rods for a number of reasons, but they shouldn't have the amount of noise that that OP has indicated.
i agree with Tinine that filtering is masking and not fixing the real problem.

What is worse is that no one has described the noise. Is it due to 60 Hz hum? Is it due to other noisy loads like drives? Does the noise disappear if other devices are turned off? Why do I even have to ask these questions after all these years? This isn't the first thread about noise. I have often had to recommend that all analog sensors have their own power supply and not be connected to any switching type loads.

Another sad thing is that after 25+ years on this forum is that no one has figured out how to specify the coefficients for a low pass filter for a certain bandwidth let alone that there are two pole Butterworth filters that are significantly more effective at reducing noise and take only a little more effort.

Is this going to be yet another thread that drags on and on without resolution?

 

[parky]

Noise is usually due to a couple of things, poor screening, noise from external sources like drives, especially If using remote I/O Had this on a system, the screens were connected at both ends on the remote I/O also a seperate earth cable to the remote I/O & bonding to the machine with RIO from the main panel, just removed the screen from the RIO cured the problem. I think manipulating the analogue value is not the real answer.

 

[MaxK]

This post focused on closed loop filters only. Open-loop filters are not interesting because this is not a question of mathematics, but of delirium in the head of the user/manager... - we can draw any nonsense.

Another sad thing is that after 25+ years on this forum is that no one has figured out how to specify the coefficients for a low pass filter for a certain bandwidth

In my case because I don't have an answer to the following question:
How, for a real process with real noise, or rather, given the actual data about the process and noise, can one determine the required filter bandwidth for its use in a closed loop?

let alone that there are two pole Butterworth filters that are significantly more effective at reducing noise and take only a little more effort.

And why is the Butterworth filter so great (except for the fact that you have a video about calculating the Butterworth filter)? In application it is obvious to closed loops.
If you still decide to answer this question, then please attach the Magnitude response and, most importantly, the Phase response of the Butterworth filter

Is this going to be yet another thread that drags on and on without resolution?

What an intriguing question... Let's think about it
OP don't even realize how does 1-st LPF (FOL) "formula" looks like. What “resolution” you talking about?

 

[robertmee]

Not suggesting an LPF is the proper application to the OP, but why use timers and ladder. CLX has a built in LPF function block. I've used it on agitated level control.

 

[APS_Programming]

Hey,

Before implementing software filter, I would try to eliminate any electrical noise.

1. Make sure the grounding tab at back of tempo sonic is grounded.
2. Cable is shielded
3. Then I would implement a software filter.

Having said that, recently I had an issue with temposonic Ethernet/IP version and after sending it back for firmware upgrade all noise issues disappeared

 

[Peter Nachtwey]

This post focused on closed loop filters only. Open-loop filters are not interesting because this is not a question of mathematics, but of delirium in the head of the user/manager... - we can draw any nonsense.

What do you think is the difference between "closed" and "open" loop filters.?

In my case because I don't have an answer to the following question:
How, for a real process with real noise, or rather, given the actual data about the process and noise, can one determine the required filter bandwidth for its use in a closed loop?

This is easy. That is why I wanted the OP to use a scope. 60Hz hum be a problem for many applications but right now we have no clue about the type of noise. First I would tell the user to fix the noise. If it can't be fixed, then I usually tell the user to set the low pass filter at about the geometric mean between the frequency of noise and the frequency of motion. Since most industrial applications only have a motion frequency between 5-10 HZ the geometric mean would be sqrt(10*60) which is about 24 to 25 Hz. Motion systems often sample at 1000 Hz. There is quantizing/sampling error. A filter at about 100 Hz reduces the noise significantly without interfering with the motion. Also, for motion systems I don't recommend filtering the feedback. Filter the output when used in a closed loop.

And why is the Butterworth filter so great (except for the fact that you have a video about calculating the Butterworth filter)? In application it is obvious to closed loops.
If you still decide to answer this question, then please attach the Magnitude response and, most importantly, the Phase response of the Butterworth filter

The Butterworth filter just has a sharper response at the corner frequency or bandwidth. Also the drop off in the response is much faster than a single pole low pass filter. It is marginally better than a two pole low pass filter. I deally you want no higher frequencies to get through than what you want.

https://deltamotion.com/peter/Mathcad/Mathcad%20-%20Butterworth%20NG.pdf

https://deltamotion.com/peter/Mathcad/Mathcad%20-%20Butterworth%20PLC.pdf

Notice these are old, but the math doesn't change. I just have better methods now.
I have a directory of filters too.

What an intriguing question... Let's think about it
OP don't even realize how does 1-st LPF (FOL) "formula" looks like. What “resolution” you talking about?

Resolution, it has multiple meanings. I should have use final solution or similar.
The OP hasn't told us what the application is..........yet

I NEVER recommend using analog MDT rods. I always recommend SSI rods. Some have resolutions, increments, of 0.1 microns.
Do you know how MDT rods work?

 

[drbitboy]

OP said the the grounding appears to be done properly; that still leaves other sources of noise (power supply, emr, etc.) as a possibility.

[...] recently I had an issue with temposonic Ethernet/IP version and after sending it back for firmware upgrade all noise issues disappeared

I'll put my chips on this as the final answer, although for all we know the "firmware upgrade" is applying a filter to a fundamentally noisy system.

We still do not know why a reduction in "jumping around" is desired and/or needed.

 

[padees]

Is this going to be yet another thread that drags on and on without resolution?

It appears so. Has anyone even considered a mechanical issue? Yep, that can happen.

 

[MaxK]

I started to respond to #23.
I put a lot of effort into staying calm.
But you once again showed your skills.
Well, if you please:
Why do you publish something for which students are kicked out of exams?
Why are you sending me links to the writings of a third-grader?
I had a simple request for you to demonstrate Magnitude response and, most importantly, the Phase response (I tried my best to be tactful)

I believed that an educated person, looking at the curves, would wonder about the phase shift (for example, at the cutoff frequency), and then ask the question about the influence of such a phase shift on the behavior of the closed circuit into which you absolutely recklessly and irresponsibly propose to introduce this same phase shift.
Well, I wasn’t too lazy and took the curves from your note.
The cutoff frequency is 5. The question is what is the phase shift at the cutoff frequency and how will this affect the closed loop?
For you this is "This is easy."
For me, this is a problem that should be approached very carefully and requires research in each specific case.

 

[Peter Nachtwey]

I started to respond to #23.
I put a lot of effort into staying calm.
But you once again showed your skills.
Well, if you please:
Why do you publish something for which students are kicked out of exams?

Why would studdents get kicked out of their exams?

Why are you sending me links to the writings of a third-grader?

WTF?

I had a simple request for you to demonstrate Magnitude response and, most importantly, the Phase response (I tried my best to be tactful)

i plot the magnitude and response and gave examples of filtering noisy data.

I believed that an educated person, looking at the curves, would wonder about the phase shift (for example, at the cutoff frequency), and then ask the question about the influence of such a phase shift on the behavior of the closed circuit into which you absolutely recklessly and irresponsibly propose to introduce this same phase shift.
Well, I wasn’t too lazy and took the curves from your note.
The cutoff frequency is 5. The question is what is the phase shift at the cutoff frequency and how will this affect the closed loop?
For you this is "This is easy."

Yes now but it took many years to get here.

For me, this is a problem that should be approached very carefully and requires research in each specific case.

The Delta Motion RMCs incorporate input and output filters. They have for years. Filters on the feedback hurt or limit performance. Filters on the output are OK if the frequency is chosen right. For motion control I prefer Luenberger Observers. For filtering noise from feed chains I prefer the alpha-beta-gamma filter.

 

[I_Automation]

I had the same thing in a ControlLogix with 2 FAST analog cards. Even with a calibrator monitoring the mA signal, and using the calibrator as the power, it jumped the same as yours. All channels whether only 1 in use or all on a card.

Every channel bounced from about 1000 to what the reading should be every 1/2 second. Card settings did nothing, Grounding out unused inputs did nothing. Dampening the value always gave a lower than actual result.
Moving the analog cards to other slots did nothing.

Recording the highest value in the past 90 seconds wasn't a good idea because then a lowering of the signal would be ignored those 90 seconds and I wanted immediate readings for PIDE and the HMI display.

I finally replaced the FAST modules with regular modules and the problem went away without any settings, grounding or dampening needed.

 

[MaxK]

WTF?

How right you are! Such a fitting comment for many of your posts!!!
You do not consider my questions worthy of your answers. Accordingly, I consider myself to have the right not to answer your questions.
Let nothing distract you from enjoying yourself.

For those who may be interested or may find what I write useful:
Below I have attached a picture in which I will try to illustrate my point.
Top chart “Closed-loop” - closed loop (plant, feedback PI-controller) step response
Lower chart “Closed-loop + filter” - the same closed loop with 1-st LPF (FOL) step response.
As you can see, the behavior of the system has changed significantly. This is due to the fact that the filter has the effect of phase shifting the signal. The red curve is the output value of the system, and the blue curve is the filtered value. As you can see, the blue curve is shifted relative to the red one, and it is this value that goes to the input of the PI controller, which leads to ineffective control.

 

[Peter Nachtwey]

Your response with the filter is due to the putting the filter on the feedback which I said above, DON'T DO!.