I/O noise

G

Thread Starter

Guy H. Looney

I'm using a non-linear 24VDC, 10 amp power supply by SOLA (spelling?). It's used to power the following:

A) 2 motion controllers
B) 24 relays (24VDC relays that switch 110VAC to hydraulic & pneumatic cylinders)
C) 1 operator interface
D) 20 or so Bimba reed switches (3 wire switch configured as PNP)
E) 50 or so proxes from Turk (PNP, inductive & capacitive)
F) 40 or so proxes from P&F (PNP, inductive)
G) Control power to 2 servos (maintain encoder position during E-Stop)
H) Drive status to 4 servos (NO relay determining whether or not drive is
enabled)
I) 4 brakes (130mA
J) Monitor circuits for MCR & 8 motor contactors (230 single & three phase)

Everything in the system seems to working perfectly except the Bimba reed switches & the P&F proxes. The Bimbas are wired into a PLC & the P&F proxes are wired into the motion controllers as overtravels on ballscrew tables. The PLC has no problem seeing the input signal, but the LED on the switch itself is going nuts. It's on, it's off, it's flashing, it's not
flashing, etc. The P&F proxes also have LEDS, but they don't go nuts.....they just quit working. All of a sudden they'll just quit working (they're NC switches that "freeze" open). There are quick disconnects on all switches.

I've put a scope on the power supply in the panel & it looks good. All the switches come straight into a small junction box. At the box, we tie all the 24VDC & -24VDC power connections together via terminal blocks. We then bring only the signal (load) wire back to the main panel. The 24VDC power signal in the junction box looks good & clean. The distance between the main panel & the junction box is about 15 - 20 feet. The
distance between the junction box & the devices is about 10 - 15 feet. I took a look at the power at the switch itself & saw some 5 - 10 volt spikes at divisions of 1 microsecond and faster. The fact that I didn't see these spikes at a slower division level explains why the PLC & motion controllers aren't seeing the fluctuations (both have a debounce time of at least 1 millisecond).

We contacted the manufacturers & really didn't get an answer. So far we've lost 3 P&F sensors in less than 3 weeks. The Bimbas haven't failed,
but the disco light show from the LED's is very annoying (sometimes the LED will stay out for 5 or 10 minutes, even though the switch is actually
made).

Thoughts, suggestions, comments? By the way, I am using fly-back diodes on the relays & brakes.

Thanks,
Guy

PS. I've posted to this list in the past as a distributor for automation products. Thanks to words of encouragement & advise from many of you, I've left distribution & joined 2 other guys to form an integration company. It's much more fun "playing" with the equipment than selling it.


Guy H. Looney
Motion Control Engineer

A.C.E. Systems, LLC
170 Medearis Drive
Old Hickory, TN 37138
email: guy.looney[email protected]
website: www.acesystemsllc.com
work: (615) 754-2378
fax: (425) 944-5017
cell: (615) 330-0044
 
Guy,

Wow, you sure got a lot of devices connected to that power supply ! But I'm sure you've done your math (over and over and over by now).

You mentioned use of a non linear power supply. Is it a switcher ?

You may find that because your load is so heavy, that ripple on the output of the power supply is significantly high, say in the order of 2% to 5%. If you are using a switcher, the ripple voltage will be at a fairly high frequency (40Khz ?), not at 60hz as in linear supplies.

The proximity sensors may also operate at frequencies in this range, thus you may be getting some sort of interaction going on between internal oscillators and power supply ripple..

Scope the output of your power supply, with a static load, so that you are drawing 9 amps or so. Measure the amplitude of the ripple, and of course the frequency.

A simple filter made with a resistor and capacitor connected to the +24 lead of each prox may do the trick. Connect resistor in series between 24 V power supply +ve, and + 24V wire of prox. Connect capacitor between +24V wire of prox and common wire of prox.

Calculate resistor and capacitor t=1.1RC where t is the period of one cycle of the ripple, and R is a resistor about 10 ohms or so. For 40Khz ripple, you can use a 10 ohm resistor, and a 2.2 microFarad capacitor.

Try it on one or two sensors, and let me know how it worked out.

Best regards. Bob
 
R
I ran into a similar situation one time using some capacitve proxes. Every time one of the motors would start up, the LED on the prox would blink. We ended up chaning to inductive proxes and the problem disappeared.
 
E

Eric M. Klintworth

Guy,
I would not be able to resist trying the choke from a Radio Shack #270-030 noise eliminator kit (for car stereos) in series with the power lead at the junction box. The lit on their web site says it's high potted a 100VDC, and is rated 4 amps. You said the power looks clean at the j-box, but I still have to suspect the noise you are seeing originates upstream of the j-box. And I prefer the choke (inductor) to a capacitor because it doesn't charge (pulling mucho amps), and doesn't age (like electrolitic caps).

This isn't guaranteed, but it works wonders for whiney car stereos, and at $3.99 it's probably worth a try.

Good luck,
Eric M. Klintworth, PE
Sharp and Associates, Inc.
Columbus, Ohio USA
---------------------------------
 
M

Michael Griffin

At 09:11 27/12/00 -0500, Guy H. Looney wrote:
<clip>
>Everything in the system seems to working perfectly except the Bimba reed
>switches & the P&F proxes.
<clip>
>I've put a scope on the power supply in the panel & it looks good. All
>the switches come straight into a small junction box. At the box, we tie
>all the 24VDC & -24VDC power connections together via terminal blocks. We
>then bring only the signal (load) wire back to the main panel. The 24VDC
>power signal in the junction box looks good & clean. The distance
>between the main panel & the junction box is about 15 - 20 feet. The
>distance between the junction box & the devices is about 10 - 15 feet. I
>took a look at the power at the switch itself & saw some 5 - 10 volt spikes
>at divisions of 1 microsecond and faster. The fact that I didn't see
>these spikes at a slower division level explains why the PLC & motion
>controllers aren't seeing the fluctuations (both have a debounce time of at
>least 1 millisecond).
<clip>

The above paragraph would seem to be the meat of the issue. You said the voltage at the power supply "looks good", while at the sensor end you are seeing "5 - 10 volt spikes". In other words, the power supply end of the wire has a fairly clean signal (no spikes), while the sensor end does not. This does not sound like a problem with the power supply.
If this is in fact the case (and it probably wouldn't hurt to double check this precise point), then I don't see what it could be other than inductive coupling from some source, possibly from the servo cables. I would suggest checking cable routing and shielding. Since this can potentially involve a lot of work, I would very strongly recommend checking precisely where in the circuits you are seeing this noise if you have not already done so.

Your description seems to indicate this is a continuous problem, rather than an occasional transient one. If so, it should be easier to find than one which comes and goes. May I also assume that the problem occurs when the machine is stationary as well as when it is operating? What happens when you remove power from the servo systems?
If the noise is being induced through the cable which runs between the junction box and the main panel, you could try either moving this cable
(temporarily at first) or if this is too difficult just for testing, then to put in a temporary cable for test purposes. To test if it is being induced via one or more of the individual sensor cables, try disconnecting the
sensor cables at the junction box one at a time.
You may of course have more than one source of noise. Since you know the layout of the machine you should be able to test the most likely
culprits first. If one or more of the drives is the noise source then this might also possibly be tested by turning off all the drives and any other
possible noise makers and then turning them back on one at a time.

If this does turn out to be coupled noise, then I will say that I have seen probably more servo systems installed using "bad" practices
(routing, shielding, etc.) than I have seen using "good" ones. Most of the time people seem to get away with it, but occasionally a mysterious problem occurs. Previous experience is not a good guide in this as it is probably
impossible to accurately predict this sort of problem.
As a general note, I believe that when a machine contains wiring (servos, instrumentation, etc.) which may be sensitive to the physical
installation, the design drawings should indicate how the wiring is to be physically routed. Cable routing in cases such as this usually seems to fall into a grey area between mechanical and electrical design and often ends up being at the disgression of the electrician installing it.


**********************
Michael Griffin
London, Ont. Canada
[email protected]
**********************
 
W

William L. Mostia, Jr

I tend to agree with Michael in regards to the possibility that interference is being coupled into the circuit. Since it effects a number of instruments and is continuous, the point of coupling must be common to these instruments and the offending signal must be there all the time.

Any large electric fields in the area may effect the instruments.

Two other possible suggestions are: 1) wire up one of the sensors external to everything and see if the problem goes away. 2) put ferrite beads at the inputs to one of the victim instruments and see if the problem goes away.

Also, you might check the power supply and system grounding. I have found that when I have strange problems, it is always good to check the grounding.

Bill Mostia
=========================================================
William(Bill) L. Mostia, Jr. PE
Principal Engineer
WLM Engineering Co
Independent I&E Consultant
P.O. Box 1129
Kemah, TX 77565 USA
281-334-3169
E-Mail: [email protected]

These opinions are my own and are offered on the basis of Caveat Emptor.
 
G

Guy H. Looney

Michael,

I appreciate your response. Unfortunately it is not the servos. I won't bore you w/ one of those "...in my vast experience of working w/ servos...." statements. Bottom line, the problem is present (lights flickering on the Bimbas & measured spikes) during an E-Stop condition. During E-Stop, the drives have absolutely no power, thus the motor cables have no power.

Guy



Guy H. Looney
Motion Control Engineer

A.C.E. Systems, LLC
170 Medearis Drive
Old Hickory, TN 37138
email: [email protected]
website: www.acesystemsllc.com
work: (615) 754-2378
fax: (425) 944-5017
cell: (615) 330-0044
 
J

Johan Bengtsson

A note about capacitors and inductors in a filter.

An inductor generally does a better job when the current is high
A capacitor generally does a better job when the current is low

Why?
If you put in a capacitor parallel to the voltage it tries to limit the changes in voltage, that is what you want so it is ok. The lower resistance you have in the circuit (between voltage source and capacitor) the harder do the capacitor have to work in order to get its work done. Putting in a series resistor will make it easier, but if you do that you won't be able to use very much
current since you get a high voltage drop.

If you put in a inductor in series with the load it will try to limit the changes in current. If you don't have any changes in current the inductor won't do anything at all.

The best way is to combine those two, an inductor in series and after that a capacitor in parallel. All voltage changes need to charge/dischange the capacitor, thus requiring a current change. The inductor don't want current changes and
therefore they will combine their work to stop the noise.

A normal car stereo already contains capacitors, that is why the inductor is enough in this application.

For calculations: sqrt(L*C)>1/noice_frequency
the more difference the better filtering


/Johan Bengtsson

----------------------------------------
P&L, Innovation in training
Box 252, S-281 23 H{ssleholm SWEDEN
Tel: +46 451 49 460, Fax: +46 451 89 833
E-mail: [email protected]
Internet: http://www.pol.se/
----------------------------------------
 
D
Guy,

I am sure you have checked this, but I would check it again.

It sounds as though the ground circuit or DC common are not completed properly. What is the voltage between the machine ground and DC common?
I normally ground the DC common to panel ground at the panel.

Is there a ground wire between the machine & electrical junction box back to the panel ground?
Possibly the DC common is really open and making a path on the machine ground.

Dave Kane
[email protected]
 
M

Michael Griffin

At 09:14 28/12/00 -0500, Guy H. Looney wrote:
<clip>
>I appreciate your response. Unfortunately it is not the servos. I won't
>bore you w/ one of those "...in my vast experience of working w/ servos...."
>statements. Bottom line, the problem is present (lights flickering on the
>Bimbas & measured spikes) during an E-Stop condition. During E-Stop, the
>drives have absolutely no power, thus the motor cables have no power.
<clip>
I imagined that you had already investigated inteference from the servo system, but you didn't state what happened when the servo were off so I thought it bore mentioning. What would be particularly interesting to know is if the proxy lights continue to flicker when everything except the DC power supply and its loads are powered down.
I'm not sure if our sitting here saying "check this" and "check that" is helping much, but this does sound like quite a challenging problem. I think we would all be very interested in hearing what the final resolution is.

I would also like to mention that the problem, although it sounds like induced noise may be from another source. I have seen some rather difficult to explain phenomina involving what looked like "noise" but wasn't. Also, if your scope is of a conventional design it can sometimes be difficult to take a measurement without affecting the circuit or picking up
phantom "noise" (noise which is introduced by the process of attempting to measure the circuit). A completely isolated one (like a Fluke Scopemeter or something similar) is sometimes necessary.

If induced noise begins to look unlikely as the source of your problem, a few other possibilities to consider are:
1) It might actually be your power supply, even though the voltage at the power supply end looks "ok". At higher frequencies a wire may no longer behave like a simple wire and it is possible to see effects at one end from something at the other. The sensors would have a fairly high input impedance in their own internal circuits which may make the problem show up
there more readily. If you have a different suitable power supply (preferably a simple linear one) which you can use for test purposes, it
wouldn't hurt to try it.

2) One of the other active devices may be introducing the noise, but it is only showing up at the sensors. The unfortunate part of this type of reasoning is that it is so vague and indeterminate that it introduces many possible factors. You may try disconnecting various circuit elements to see if the problem goes away.

3) I have seen switching power supplies have odd effects on certain types of 24VDC powered signal conditioners. The problem was resolved by replacing them with a linear supply. I don't really see how something like this could be affecting your sensors, but at this point since we are grasping at straws anyway, one more straw won't hurt.

4) Is your DC circuit properly grounded at the proper point? I have seen some odd things occuring from failures here as well. The sensors
themselves shouldn't care if they float, but I think you may have more than one power supply involved in this machine (I/O, PLC, etc.). There may be some non-obvious connection between them. Are all the terminal blocks tight, etc. (e.g. do you have a loose terminal screw on a DC power or ground feeding you terminal box)?

5) One last thing, which you may perhaps check, if you have not already done so. I hesitate to mention it because it may be too obvious. However, you mentioned you are having problems with the Bimba 3 wire MRS switches and also with the P&F inductive proxies, but not the Turck PNP inductive and capacitive proxies. If I were dealing with this problem, I think at this point I would personally check all these switches to see if they are indeed the correct ones and were correctly installed and connected. I have seen problems with incorrect type or incorrectly connected sensors which still somehow functioned in some fashion, when they theoretically should not have worked at all. One of the more interesting problems occurred when 2 wire MRS sensors were used on a PLC connection instead of the proper 3 wire sensors - the switches would work most of the time when according to the specs they shouldn't have worked at all.

There have been several suggestions to put some sort of filter in the wiring. It wouldn't be the approach I would attempt on a machine I was working on because:
i) When I try to imagine what your machine looks like, I cannot think of any practical place to put these filters.
ii) You know the problem is visible at the sensors, but you don't know what the cause is or how it is getting into the circuit. In this
situation, putting filters everywhere the problem is visible isn't solving the problem, it is just sweeping it under the carpet. The underlying cause
could surface again in a different form at a later date.

**********************
Michael Griffin
London, Ont. Canada
[email protected]
**********************
 
W

William L. Mostia, Jr

Mr. Looney,

You need to be more clear about your problem statement. It would have been helpful to have had this information in your original problem statement as it is a key bit of information.

This brings me to ask some more questions. Is the problem only during a E-Stop or some other operating condition? Is the problem different at any time? For example, is there any difference between the operating and the E-Stop condition? Is the problem with all the P-F prox and Bimba switches? Is there anything different about the P-F and Bimba installation than the
rest of your instruments? What is common about the physical or electrical installation of all the effected instruments? Are all the effected
instruments acting exactly the same? Is there any AC voltage (not necessarily any current flowing) in the cables around the P-F and Bimba
cables even during the E-Stop condition? If so, it is possible to couple electrostatically(capacitively) on to your instrument cables. Are there any RF sources near by? Are there any other equipment in the area that might generate large electric fields? Have you had any other unusual or repetitive failures on the system? Have you always had this problem?

Some additional troubleshooting techniques:

Try drawing an imaginary "spherical" shaped surface around all the affected instruments as close as you can to all the instruments then expand it outward and looking for the common interfaces to all the affected instruments at the "sphere's" surface including non-wired ones. Then remove the common interface or determine a test to see if the interface is causing the problem.

Try turning off the power one at a time to anything else in the area within about 10-15 feet during a E-Stop. This includes any other instruments or circuits on the system that are powered during E-Stop, lighting, other equipment in the area, etc.

Make sure that you have only one ground on your power supply and that none of your shields are grounded at both ends.

Bill Mostia
=========================================================
William(Bill) L. Mostia, Jr. PE
Principal Engineer
WLM Engineering Co
Independent I&E Consultant
P.O. Box 1129
Kemah, TX 77565 USA
281-334-3169
E-Mail: [email protected]

These opinions are my own and are offered on the basis of Caveat Emptor.
 
M

Mike Johnson

I thought noise covers the entire frequency spectrum, sort of like the 3 degree Kelvin background radiation noise from space and since the Fourier transform of the ideal Guassian noise function is a constant function over all frequencies, hence no matter how much filtering you have there will be some noise present.

However, using a nth-order low pass filter (like a
4th-order low pass Butterworth filter) will help but not much. You might want to look at all possible sources that may lead to noise induction on your DC ground plane.
 
J

Johan Bengtsson

Of course there will be some noise left, but it could be small enough to not disturb the circuit.

The amount that could disturb the circuit is probably different for different frequencies.

"Ideal" noise covers all frequencies, but where do you actually find ideal noice? Ok, there is probably some very small amount of that, but most of it is probably generated from something quite close and probably not even close to "ideal" noice. In fact I would beleve some frequencies can be found to be very much higher than the rest.

But actually it doesn't matter much if it is ideal or not as long as you can bring the frequencies that disturbs the circuit down enough.

/Johan Bengtsson
----------------------------------------
P&L, Innovation in training
Box 252, S-281 23 H{ssleholm SWEDEN
Tel: +46 451 49 460, Fax: +46 451 89 833
E-mail: [email protected]
Internet: http://www.pol.se/
 
W

William L. Mostia, Jr

Every electrical circuit has noise in it. This is a matter of imprecise definition. Noise can be defined as any voltage or current in a circuit other than the desired one. Interference(undesirable noise, if you wish) on the other hand can be defined as noise that causes undesirable effects in a circuit. Noise is many times used synonymously with interference indicating all noise is bad when if fact, below a certain signal to noise ratio for a given circuit, noise is irrelevant. The goal of filtering or other noise reduction techniques is to reduce the interference down to a level so that it longer causes the undesirable effects. While white or Gaussian noise exists (resistor thermal noise for example), it is generally not a problem, except possibly for low, low level measurements. On the other hand, noise/interference generally of interest in the industrial environment is man-made (lightning being one of the exceptions) and as a result can be defined as a particular range frequencies which can be reduced by various noise reduction techniques. Bill Mostia ========================================================= William(Bill) L. Mostia, Jr. PE Principal Engineer WLM Engineering Co Independent I&E Consultant P.O. Box 1129 Kemah, TX 77565 USA 281-334-3169 E-Mail: [email protected] These opinions are my own and are offered on the basis of Caveat Emptor.
 
Top