Monarch 10EE electrical noise suppression

[purplepicker]

Help!!
I recently completed the CNC conversion of my 1953 10EE (Ward Leonard drive unmodified with 7.5 hp RPC) see [h=2]http://www.hobby-machinist.com/showthread.php/17006-Monarch-10EE-CNC-conversion-complete[/h]Everything worked fine except for the occasional split thread when running a thread program. I hooked up a scope and found that as I suspected there was an occasional noise spike making its way into the spindle index input of the controlling computer which made the controller think it was time to start the thread.....analogous to closing the half nuts at random.
I was able to greatly reduce but not completely eliminate the problem by covering the spindle encoder cable with braided sheath. By sheathing the cable I am basically treating the symptom rather than the disease. I really need to reduce the motor and or generator brush noise at the source.

Has anyone here attempted, successfully or not, to suppress the arcing EMI on their 10EE? I have obtained a half dozen Sprague .05 mf 600v "bathtub" caps that I intend to put across the brushes on the motor and generator. I think it will be tough to mount them very close to the brushes so they will probably have to go in the DC control box.

I was unsure where to post the question here or on the CNC forum but since I got so much help here when getting the machine up and running in its manual mode I thought there might be a source of similarly vast knowledge in the electrical area of this machine.

Any help greatly appreciated.

Cecil

 

[Ray C]

I personally would continue on the path of extra vigilant grounding and shielding before putting caps across the brushes.

I'm not familiar with the 10EE motor/generator setup but, I gather it's a variable DC system. What's the range of frequency? I think you'll be chasing your tail finding the right cap values with high Z across all frequencies.

I would first try line and/or load reactors to clean things up. Also, shielding things is a perfectly valid way of solving a problem provided you're not trying pass some form of regulatory testing where radiated emissions is important.


Ray

Help!!
I recently completed the CNC conversion of my 1953 10EE (Ward Leonard drive unmodified with 7.5 hp RPC) see http://www.hobby-machinist.com/showthread.php/17006-Monarch-10EE-CNC-conversion-complete

Everything worked fine except for the occasional split thread when running a thread program. I hooked up a scope and found that as I suspected there was an occasional noise spike making its way into the spindle index input of the controlling computer which made the controller think it was time to start the thread.....analogous to closing the half nuts at random.
I was able to greatly reduce but not completely eliminate the problem by covering the spindle encoder cable with braided sheath. By sheathing the cable I am basically treating the symptom rather than the disease. I really need to reduce the motor and or generator brush noise at the source.

Has anyone here attempted, successfully or not, to suppress the arcing EMI on their 10EE? I have obtained a half dozen Sprague .05 mf 600v "bathtub" caps that I intend to put across the brushes on the motor and generator. I think it will be tough to mount them very close to the brushes so they will probably have to go in the DC control box.

I was unsure where to post the question here or on the CNC forum but since I got so much help here when getting the machine up and running in its manual mode I thought there might be a source of similarly vast knowledge in the electrical area of this machine.

Any help greatly appreciated.

Cecil

 

[Tony Wells]

As I understand it, the problem begins with the arcs created by the brushes. Being rich in EMR, it is easy to pick up with any nearby conductor, just like a radio picks up its signal. Capacitors directly across the brushes is thought to be effective in reducing the strength of the arc, and hence the radiated noise, so that is a popular approach. Another method is the same thing seen in CAT wiring, and that is twisted pairs. The magnetic field surrounding conductors tends to cancel itself out if the leads are twisted in pairs. The twists should not be even and consistent. CAT 5 and 6 wires have a varying twist, and do pretty well to protect from picking up stray radiated energy. With higher current involved, I'm not suggesting you used CAT wiring, just the principle of it. If you have used something like THHN in flex, for example, pull it out and make it twisted. That along with caps, which should be rated for at least twice, better three times the expected voltage should help.

When you used the braid for the shield, where did you "ground" it? One end? Both ends? Sometimes it takes a little experimenting to get this to work. It's an inexact science in application, though the theory is well established.

 

[DMS]

I agree with Ray; even if you manage to reduce emissions, you should still look at protecting your encoder signal. One question I had is, how are you grounding the shield you added? If you don't have a good ground on that, then it is going to be pretty ineffective. Also, if things are still an issue, you could look at using an encoder with a "differential" output. These use 2 wires to transmit each signal, but are very resistant to noise.

I don't think the cap will hurt on the motor/generator. If it were a alter model with active control like one of the WIAD models, I would not risk it. If you can't get the cap really close to the motor, it may make things worse. The inductance of the leads, and the added capacitance you are adding will form what is called a "tank circuit". It's the basis for an oscillator. If you get the frequency right (er, maybe "wrong" is the right word in this case), then you will actually _amplify_ the noise.

 

[purplepicker]

Thanks for the input.
The most common shield grounding rule in my experience is to ground at as near a "star point" or "mecca point" as possible so that all shields and ground leads come to one point so that all the ground points either stay at ground potential or if not at least they all change together. Most commonly the star point is at the "power source" because if the ground is at the load end of the run the ground potential might move around due to currents flowing in the power circuits. I tried most possible combinations including grounding the shield at the control box (power source), at the load (on the machine) and at both ends. Sure enough grounding the shield at the control box only seemed to leave the lowest amplitude of noise on the scope.
The encoder cables are all cat 5 (now shielded cat5).

The shielding has reduced the problem to "almost" non-existant. Out of maybe a dozen threading runs it only split a thread once. That was of course while I was showing off to a friend and bragging about how I had fixed the problem.

What is interesting is that the noise spikes as seen on the scope are not steady in nature. They only show up maybe once in every 3 to 10 sweeps and they only seem to show up "between" the legitimate index pulses. That might be because they are two short in duration or have two short a rise time to trigger the scope so I only see them when the scope has been triggered by a real index pulse. It is also possible that the noise from the brushes is not "steady" but has some random feature where it is occasionally worse.

If you have ever looked at the DC end of a Ward Leonard type 10EE you will realize that just about everything that can be in conduit already is but the wires to the brushes are pretty much exposed.
The spindle encoder is homebrew and is not differential although I could redesign it and make it so. Another improvement would have been to make the index pulse correspond to half a revolution of the encoder wheel instead of the same width as the quadrature slots. I could then use an RC filter to lowpass the index pulse and discriminate against the very short duration noise spikes.
If I ever do this again I will definitely use a long index pulse and lowpass filter it.

Tomorrow I am going to try a heavy braided copper bonding strap from the contol box to the lathe frame and also try the caps across the brushes. I will also replace my standard cat 5 with braided shield with a real industrial strength shielded cat 5 with foil wrap and braid. I'll monitor the noise as best I can and see if any or all improve the noise situation.

I should note that this is not a fault in the lathe as when it was built in 1953 EMI was not weighing heavily on anyone's mind so noise suppression was not an issue. By marrying a great old machine to current technology I am just looking for trouble but I love a challenge.

Thanks again for the input.
Cecil

 

[DMS]

Can you post or describe the encoder circuit? If the outputs are not driven low (IE, they are floating in the "off" state), then that will make them much more susceptible to noise. If this is the case, adding a small pull-down resistor should help greatly.

- - - Updated - - -

Oh, and regarding grounding shields, it is best to tie the shield into a good earth ground (not the power supply ground). Tying it into the power supply ground will just direct the noise into your power supply circuit.

 

[purplepicker]

The spindle encoder consists of a slotted opaque disk with 20 equally spaced shallow notches (approx .15 inches deep) with one notch cut deeper(approx .3 inches deep). The 20 shallow notches work with two carefully located photointerupters to produce an 80 edge per rotation quadrature signal with the A and B pulses 90 degrees out of phase. The index signal (the bad guy) comes from a photointerupter which reads the deeper notch one time per rev.
The photointerupter is a sharp GP1A58HRJ00F. See link for data sheet.

http://www.sharpsma.com/webfm_send/1554

It incorporates and amp and a voltage regulator and creates full clean TTL level output. I have a 1 microfarad cap from VCC + to VCC - directly on the chip socket. I have a 250 ohm resistor in the LED path to drive the LED at about 15 ma. I'm not sure what a pull down resistor would do but it looks like I could add a pull-up of maybe a few K to lower the impedance when the output goes high.

Probably one of the best solutions would be to make the index notch half the circumference of the disk. That would create a VERY long index pulse at least a unit of magnitude longer than any noise spike. I could then lowpass the index pulse and discriminate against any noise on the index line.

I could mount two index interrupters and generate a differential signal but this index is getting fed into the parallel port of the computer where it is processed by LinuxCNC. I'll have to do a little research to see if LCNC can take advantage of the differential signal.

The problem is that now the noise signal is low enough most of the time as to not be detectable. Unfortunately it rarely, occasionally, and cursedly appears just at the wrong time.

I think the closest I could get to a "real" earth ground would be a ground rod driven maybe 20 feet away from the machine. I have an unused ground rod so it might be worth driving it just for a hoot. It might even help.


Thanks again for the input.
Cecil

 

[DMS]

That part its driving active low, and has a pull-up. That should be fine (actually, its quite good). Given that, and the fact that you are shielding, I am a little surprised you are seeing any issues; perhaps I am underestimating the noise on those brushes. I still think improving grounding would help.

Out of curiosity, what is the supply voltage on the interrupter's "output" side, and what are you feeding the signal into?

When I did the spindle encoder for my milling machine, I built my own out of a reflective sensor, and designed/built my own conditioning circuit. My problems early on were mostly related to getting too much reflection from the "black" part of the encoder disk. Is it possible that your problems could be caused by the sensor itself mis-triggering, rather than a noise issue?

 

[purplepicker]

The LED is supplied from a 5v 800ma regulated supply with a 250 ohm dropping resistor. The chip VCC is supplied from the same source. There is a 1 microfarad 25 volt bypass cap directly across pins 3 and 5 (VCC to common). The output from pin 4 goes by way of the shielded cat5 cable to the parallel port of a PC running LinuxCNC where the index is processed and used to trigger the z axis so that it follows the spindle position. The parallel port is the onboard port on the computer and is running at standard TTL logic levels.
The first thing I did was to blame my homebuilt encoder (light leaks, scratches or pinholes allowing light to go through when it wasn't supposed to etc.) but I laid all those thoughts to rest when I discovered that the noise spikes occurred even when I slipped a piece of foil over the receptor(no legitimate pulse possible) and when the noise was still there even when the spindle was at rest).

It should be noted that the noise spikes appear on the scope occasionally even when the lathe spindle is not moving (scope in free run triggering). That probably indicates that there is noise from the generator brushes which is either instead of or in addition to the drive motor brushes. (no noise spikes when the lathe drive unit is completely off)

Last night I replaced the spindle encoder cable with a "factory" cat 5 shielded cable which has an aluminized mylar foil shield under the braided shield. This is of course much better E-field shielding than just the braid I had previously used to cover the "standard" cat 5 cable. At this point I cannot see any noise spikes and after several threading runs have had no false starts. It would seem that in light of the corrective action working that I must be dealing with "radiated" fields with my cable acting like an antenna as opposed to conductive or ground loop problems. The better shield attenuates the E-field and the 5 turns per inch of the cat 5 attenuates the M-field.

For the record I shall make it a rule on any future retrofits to :
1. Use "real" foil shielded under copper braid cat 5 cable and ground at power source.

2. use differential encoder technology.

3. If I build the encoder disk I will make the index notch VERY long even half a revolution so that I can lowpass it if necessary.

4. Try to avoid retrofitting 70 year old noise generators.

Thanks for the input.

Cecil

 

[Tony Wells]

Sounds very much as though you have correctly identified and isolated the problem. Time will tell for sure. SO after making the cabling change, did you see any noise on the scope at all?