开云体育

ctrl + shift + ? for shortcuts
© 2025 开云体育

Re: strange stopping - dealing with noise issues


 

开云体育

?
Almost all PCs with a three prong plug have the DC signal ground connected internally to the AC mains safety ground. This makes it important to consider the potential for ground loops. USB signal cables will unavoidably have their?signal ground reference and shield connected to the AC mains safety ground because of this.
?
You do not need to provide an external shield over a USB cable in most situations. The data signals are differential and carried by a twisted pair. Twisted pair differential signals inherently resist common mode induced noise which couples into both conductors at about the same amplitude. Sometimes the induced signal can exceed the "common mode range" of the receiving device. This is?when a ferrite choke over the cable can help. PMDX has put a tiny common mode choke on all of our USB based boards, but even so some host computers suffer from this problem at the host end of the cable and a choke there can sometimes help.
?
All USB cables have a shield which may also serve as the negative side of the power delivery and the signal ground reference. More commonly there is a separate ground wire to serve as the power and reference connection. Even so, it is common for the shield of a USB cable to be connected to?signal ground at both ends. This is why I recommend a redundant larger wire be connected between the computer chassis and the signal ground at the controller so that most of the current caused by a ground loop flows through this wire instead of the shield.
?
It is possible to defeat ground loops by the use?od USB isolators or to design USB interfaces that are electrically isolated (example the PMDX-424 SmartBOB), though this is not common because USB was originally intended for human interface devices such as keyboards and mice that do not suffer from ground loops. USB isolators can be purchase on eBay for less that US$20 each. Twisted pair Ethernet also provides ground loop isolation because the signals are?usually transformer?coupled.
?
The PMDX-126 is designed to allow isolation between the PCs logic (DC) ground and the machine frame ground. Success of this isolation requires proper attention to ground points and avoiding things such as encoders whose body connects their signal ground to the machine frame.
?
Coaxial shielding of a cable protects only against electrostatic fields. It does nothing to block magnetic fields. That requires a magnetically permable material such as iron, nickle, ferrite, or mu-metal. Mu-metal alternates layers of copper and nickle and has properties that shield both electrostatic and magnetic fields. Nice as this sounds, it is not very flexible mechanically and it tends to empty your wallet quickly.? 8-)
?
Electrostatic shielding is provided by a coaxial shield that is grounded at either end or both ends.?For signal cables, grounding at only the end that is receiving the signal is preferred.
?
Grounding both ends of a coaxial shield that covers signal conductors risks allowing current to flow through the shield. When this happens, the cable will act as a one turn transformer and couple the magnetic field generated by the shield current into the signal conductors.
?
If the cable is delivering power (stepper motor drive current or spindle motor power) then it is desirable to ground both ends as there is no worry about corrupting the signals and the grounding provides safety features.
?
USB, twisted pair Ethernet, and many encoder signals are differential mode signals which help to cancel common mode noise induced either electrostatically or magnetically. The operative word is HELP, not eliminate.
?
Motor drive current to bipolar stepper motors is essentially differential and this helps reduce the interferance they generate. If ground loops are a concern, then these cables, if shielded, should have the shield grounded at the driver end. The best ground for this purpose is the negative terminal of the power source feeding the stepper drivers.
?
Power cables between a VFD and a spindle motor are a potent generator of noise due to the high voltages, and fast rise time of the PWM modulation of the simulated sine waves generated by the VFD. Ferrite chokes and/or "reactors" can be used to reduce the noise level somewhat. Because this is a power cable and there is no signal information to be compromised, you can and should ground the shield at both ends for safety reasons.
?
As was mentioned by John D. separation distance is your friend when trying to reduce the amount of noise induced by one conductor into another. It is also helpful to have noise generating cables cross signal carrying cables at right angles. Lastly, if cables are longer than needed, shorten them or fold them in a Z pattern. Coiling them up results in a better target for noise induced by magnetic fields.
?
With all of this being said, sometimes eliminating noise faults is more art than science. Overthinking and analyzing things does not always trump simply trying things in the real world.
?
Regards,
Steve Stallings
PMDX



From: [email protected] [mailto:[email protected]] On Behalf Of John Dammeyer
Sent: Friday, March 27, 2020 6:59 PM
To: [email protected]
Subject: Re: [MachCNC] strange stopping

Electrical noise can be a bear.? Steve Spallings mentioned PC grounds and system or motor grounds.? I have his PMDX-126 and unfortunately the PC I'm using internally somewhere has the DC ground connected to the system frame which is the third earth prong on the power cord which then connects to the motor frames through the machine metal. The CUI encoders on the DC servos serve to further muck things up where their DC ground is connected to the frame.? Which is of course connected to the 3 prong and so DC ground earth ground are connected in several places. ?Shielding has no rhyme or reason here but it's one place where a single ended connection for a shield is a bit more logical.

?

My EE book on electrical noise states that the signal grounds should never be connected to the DC power grounds or frame. ?The whole point of shielding is that it isolates the two.

?

For your system also look at magnetic coupling.? The way a transformer works is a changing magnetic field introduces current in windings that are also in the magnetic field.? So the spindle motor current flowing through the wires can generate a field picked up encoder or limit switch wires. ?Short of what is called mu metal shielding the only solution to electromagnetic interference is distance.? Route them far apart.

?

Electrostatic shielding is different.? Twisted pairs help to null out electric field noise. Shielding can be grounded at one end or both of the frame (third prong) earth.? The point is you only want the noise current flowing back through the shield into the frame.? Not through your signal wires which includes the DC 0V wires.? People tend to forget what goes out a signal wire is still a signal as it comes back through what is commonly called the ground or common wire.? (Some manufacturers have called the +5V to +24V 'common' for limit switches)

?

John Dammeyer.

?

From: [email protected] [mailto:[email protected]] On Behalf Of Todor Boshnakov
Sent: March-27-20 2:39 PM
To: [email protected]
Subject: Re: [MachCNC] strange stopping

?

You are trying to fix a wrong schematic with cosmetic changes. You sure have a ground loop somewhere. Check your shields, they should be connected one side only with heavy short wire to good ground. Preferably star connection.



On 27 Mar 2020, at 18:32, "spencer@..." <spencer@...> wrote:

?

i forgot to mention that, i have ferrites everywhere. that is one of the first things i do.? i have them on the signal leads to each stepper and even on the usb cable to the pokeys although the mfg suggested it would not do anything. i tried a usb signal conditioner thing but it did not work, meaning that most things plugged into it stopped working.

?

anyone know of a signal conditioned for usb that does work?

?

?

?

Best regards, Spencer Chase
67550 Bell Springs Rd.
Garberville, CA 95542 Postal service only.
Laytonville, CA 95454 UPS only.
Spencer@...
Spencer@...
Spencer@...


(425) 791-0309
(707) 223-8212

?

------ Original Message ------

From: "gayhurst56 via Groups.Io" <robin.howie@...>

Sent: 3/27/2020 4:03:12 AM

Subject: Re: [MachCNC] strange stopping

?

I would probably try adding ferrite beads to the leads.


On 27 Mar 2020, at 00:20, "spencer@..." <spencer@...> wrote:

I have my router working really well with the following configuration. Solid state tiny computer running win 10. Mach 3 latest version Pokeys USB CNC thing connected to a PDMX.

?

I am finally getting smooth motion no lost steps etc etc. Everything works except.

?

Occasionally Mach3 just stops and is totally unresponsive. Can not reset, jog etc. The diag screen shows all ins and outs off. I have to kill Mach 3 and start again.

?

I am pretty sure that every time it does this, when I restart the 4th axis DRO shows 4099.xxx? the 4th axix is angular.?

?

I am running a gcode with a subroutine that repeats over and over with a tool change to allow inserting a new part for machining. i am also using the subroutine to make it possible to count parts.?

?

when this failure happens is always close to the spindle motor going on or off. it is a three phase servo run by a high quality US inverter. Forgot the name right now but it is a good inverter not Chinese no name stuff.

?

I am pretty sure the spindle motor is causing noise that kills the Pokeys. I have proper grounding and shielding.?

?

I changed the inverter carrier frequency from the default of 5Khz to 2khz and have not had the problem yet but it some times does not happen for a considerable time.?

?

any ideas? does the 4099 signify some overflow?? does it sound like it is inverter noise? might the lower carrier frequency be the fix?

?

?

?

Best regards, Spencer Chase
67550 Bell Springs Rd.
Garberville, CA 95542 Postal service only.
Laytonville, CA 95454 UPS only.
Spencer@...
Spencer@...
Spencer@...


(425) 791-0309
(707) 223-8212

?

Join [email protected] to automatically receive all group messages.