Troubleshooting electrical interference on the Schaublin 125 || RotarySMP

[@mrtnsnp]

"…that I can move on with more important aspects of this lathe restoration". Like cleaning the type plate 😁

[@RotarySMP]

You got it :)

[@vincei4252]

I keep telling Mark to toss the entire lathe in the ultrasonic cleaner but he won't take me seriously! It would save thousands of hours of nameplate cleaning. SMH

[@RotarySMP]

@@vincei4252 That would rob me of content for the next 55 videos. :)

[@vincei4252]

@@RotarySMP Damnit :)

[@whatevernamegoeshere3644]

Man I love to watch these videos more and more. I recently got a job massaging machines back to life and I see the same sort of stuff going on at your place lol

[@RotarySMP]

Yeah, for me this sort of machine is never "done"! :)

[@whatevernamegoeshere3644]

I mean if it makes that better there are machines actually manufacturing products that are still missing 30% of their components lol

[@RotarySMP]

@@whatevernamegoeshere3644 Good to know.

[@tinygriffy]

Fixing bad hardware in software .. a classic. Well done !

[@RotarySMP]

Still need to improve this, but at least I am no longer stuck.

[@Paddington60]

Good to see all the trys that did not fix things, I feel sure they may help someone else. Good work and thank you!

[@RotarySMP]

I am good at not fixing problems :)

[@ViniciusMiguel1988]

To solve this in the electrical level you need shielded cables up to the card, also low pass filters ( i usually use 40hz cutoff ) to remove noise from the inputs.

[@RotarySMP]

I thought someone would say this :)

[@UnreasonableSteve]

Whether the low pass filter is in software or hardware how big of a difference is it really?

[@tapiorautio457]

@@UnreasonableSteve In some cases, if system is interrupt drivem this kind of spurious signal may cause some issues and generate lag in response times, due to high amount of input transitions, which all have still to be handled. It is always better to solve issues electrically, even if software solution is available, it keeps code/configuration much more clean and less hw dependent.

[@thomasnixon4440]

@@tapiorautio457 If you can eliminate the root cause then sure, but a hardware low-pass filter is not always better than software. It's easier to get exact timing with software debouncing, and it's a lot easier to change when required. Plus, slow-moving signals into digital inputs can have their own issues (metastability, noise sensitivity around the threshold). Normally a schmidt trigger will eliminate those, but not always..

[@LesNewell]

@@tapiorautio457 The inputs aren't interrupt driven. LinuxCNC polls them every 1ms so rapidly changing inputs won't affect software timing. I'm not a great fan of software debouncing but if it works, that is the most important thing.

[@michaelhaardt5988]

Great summary and structured trouble shooting! Given your lack of ferrites, a variety pack of clap on ferrites sounds like a useful purchase. If 5 us debounce suffice, the pulse must be 200 kHz or higher. You don't suck at using oscilloscopes, but your scope does. :) If you do not want to spend much on a tool rarely used, something like the DSO1511G (small, handheld, battery operated, video output for larger screen and USB for image copy, way higher promised than actual but still sufficient bandwidth) would be a good replacement. LinuxCNC also has builtin scope functions if you are curious to see what it saw, but a real scope tells so much more. Ask me how I know: "Let's quickly mill this part" ... and two hours later the mill was cluttered with cables, scopes, diving deep into electrical trouble shooting, the part to be manufactured long forgotten.

[@RotarySMP]

One of my YT Members, Alexander, is sending me his old Osci, so that should be a major improvement.

[@vincei4252]

Morning, Mark! Just started watching! [edit] Good to see you quieted at least some of the gremlins.

[@RotarySMP]

Yeah, not a complete solution, but enough to be able to move on with other aspects of it.

[@TheKnacklersWorkshop]

Hello Mark, It's so great that so many people are happy to assist with the problem, even if it became a heated debate... Always a pleasure to see the name plate. Cheers. Paul,,

[@RotarySMP]

Thanks Paul. It is nice to be able to move on with this project, even if it is not the perfect solution.

[@danstiurca7963]

Denouncing is mandatory magic. Thumbs up to the hero who suggested it.

[@RotarySMP]

Very helpful. I'd still like to measure that spike though.

[@danstiurca7963]

​@@RotarySMPI add firmware debounce as SOP to all switchable inputs. It seems like cheating, but many microcontrollers have this feature built-in because it's so useful. To actually lower the noise: 1. Try and trigger your scope on 0.5V instead of 2V. Make sure you use the ground clip, clip near the sensor input. 2. It's not clear from the video, but I'm not sure you twisted your wires. Twist all the wires that belong together, like +/- sensor lines or +/- power. It makes a huge difference. 3. It's not clear where your grounds are, if you have enough ground, and if you're grounding the right things. It would be cool to do a quick video on just your wiring setup. 4. Ferries. Add a ferrite next to the input of the sensor, and your problem will likely disappear. If it doesn't work, loop the wires 2,3 times through the ferrite to get more filtering.

[@chrisrhodes5464]

I enjoy watching you work on these machines I learn stuff doing so

[@RotarySMP]

Thanks Chris.

[@MikelNaUsaCom]

GREAT glad it is working. Fun times ahead.

[@RotarySMP]

Me too, Mike.

[@UnreasonableSteve]

Would be cool to see the transient on the oscope but the methodology for probing can be surprisingly finicky even for someone relatively good with an oscope

[@RotarySMP]

What would you suggest Steve? What trigger voltage, and sample rate?

[@tapiorautio457]

Yes those kind of signals are really hard to get good picture of even with very good equipment. And here, since the 5us filter implemented fixed the issue tells that the spur is really short, most likely in sub microsecond class, which don't help the issue. :)

[@RotarySMP]

@@tapiorautio457 Alex is sending me his old (40MHz) Osci. Maybe I can image it then?

[@LunaticCharade]

​@@tapiorautio457it is? I'm an EE so I should have more experience than most but as long as you have a scope, and you can set it to single shot, then you should catch it. But scopes are a bit black magic to many and plenty of them have bad UI so that might make it difficult. And I'm a bit unsure if what's shown in the video would qualify as 'a scope'. But since the poll rate is 1ms and it was tripping so consistent, you should be able to see something even if you only have khz bandwidth!

[@vonSaufenberg]

That Siemens module plugs in to the bottom terminal of the Schütz, but is only compatible with the newer Sirius devices as far as I know. There is the same module for the older ones (3RT1916-1PA2 or something like that ) but they are ridiculously expensive by now. Don't take my word though, I'm not an electrician, I just read the parts catalog when I'm bored.

[@RotarySMP]

It looks like that new grey one should fit on the old one.

[@E1nsty]

What about electroplating over the schmoo on the nameplate?

[@RotarySMP]

That would require nameplate removal, and I dont want to mess up the little rivets.

[@TheAppelsiini123]

Considering how modest you were regarding your background in electronics this is quite impressive!

[@RotarySMP]

I know very little, but am just summarising the wisdom of the crowd here.

[@soothcoder]

Could try scoping the power supply on the module that handles the Hall effect module. The three phase switching maybe is browning something out for a sec? Just a guess. Never setup one of these.

[@RotarySMP]

Thanks. I can check that.

[@tapiorautio457]

Wunderbar! As I mentioned last week comments, the digital debounce and/or filtering is most straightforward solution to this issue, but since I did not have familiarity with LinuxCNC coud not give any specific instructions. ;) As of the length of the filtering, I have used folllowing approach with quite good results: First somehow determine maximum allowed reaction time input has to have, since there is no reason to have nanosecond reaction times/edges in system, where needed time frame is in class of tens of milliseconds. Then divide said requirement by 10 to achieve very good amount of headroom and use this time value as your starting point for your debounce length and filter time constant. For example if you want 10ms reaction time, set the RC filter to 1ms and debounce to same value, and you still get ~2ms reaction time (filtered signal rises from 0 to 60% of voltage in 1ms and then debounce delays input for another 1ms) but are already very much more immune to external interference to signals, and if there is still issues with some interference, the values can be tightened even more. As said this approach has served me well for years. Nevertheless, this is more of a nitpicking, if the 5us debounce you implemented was enough and everything works nice, that is excellent. Happy to see the issue solved, and hope we as a community were helpful aloing the way, even my long winded descriptions of these things. (That is an issue for us engineers, we like to babble endlessly about thing we like to do :P)

[@RotarySMP]

I found the advice given to be really helpful. It seems I should still try snubbers on the contactors and RC on the Hall sensor inputs, to try and make the system more noise resistant though.

[@LetsJeep]

So nice to see the response you received on this request. Most excellent progress! I have also used software debounce timing on microcontroller 3-5v inputs to ignore high frequency pulse signals. Mostly on dry contact microswitches and tactile pushbuttons where 5v=1 and/or anywhere microseconds detection frame wasn't necessary in a noisy industrial environment. I had used a 200mhz recording O-scope to capture that bugger triggering the program logic during troubleshooting phase. Then set the debounce duration accordingly. Again, another option if the input board doesn't already have them, is to install Pull-up or Pull-down resistors nearest the inputs to keep those inputs stable(not floating) until intentionally triggered. Then program conditional statements as required to match that input logic. The NO/NC applicable Hall-effect sensors(P or N channel) will pull the signal high/low to mitigate false pulse issues. Never had as much luck with filter capacitors on inputs. While this particular scenario isn't a potential crash situation, it may be some peace of mind insurance to use both of these hardware/software approaches on all your low voltage inputs.

[@RotarySMP]

Thanks for your detailed advice. I am still working through options.

[@Nicoya]

When it comes to interference coupling between cables, what you want to look at is loop area. Picture a basic electronic circuit with a lamp on the right, a battery on the left, and a wire connecting across the top and the bottom. This big circle (or square) looking thing outlines the loop area, and if you overlay another loop on top of it, they'll couple like a transformer. To reduce this coupling, you can either move the loops away from each other, or make the loops smaller in area, by re-routing the top and bottom wires closer to each other. Since the area is (in effect) signed, you can also cancel out the coupling by flipping over half the area (making it look like a figure-eight, or one of those twisted pastries if you add a bunch of flips). These two techniques combined form the basis for how twisted pairs work: the two wires twisted together have only a very small area between them, and because they're twisted that area between them alternates facing towards or away from the source of the interference, cancelling out any remaining coupling. So for any of your sensor inputs, you should run a twisted pair with the signal and ground (or a twisted triple of signal, ground and power if you need that). You should terminate both the signal and ground (and preferably power too, if you need the third wire) together on the board doing the sensing, so as to avoid opening up a loop area inside your wiring cabinet by having the ground take a detour through narnia. As a bonus, you can wrap the twisted pair through a ferrite toroid (aka ferrite ring) to form what's known as a common-mode choke, which greatly increases the impedance for interference signals which look the same (are 'common') between the pairs, while leaving the differential signals (what you're actually trying to measure) alone. If you can't terminate both the signal and ground/power lines on the sensor board, then do your best to bundle them together with the respective power and ground lines for that sensor board back to the terminal blocks supplying them. As another bonus, you can look at installing an RC snubber over at the motor itself, so that the back-EMF you get when you disconnect the power stays more contained at the motor rather than bouncing down the lines and into your control box.

[@RotarySMP]

THanks for your very clear summary of the practices available. That is very helpful.

[@mr.reflect1027]

The Siemens RC-Snubbers get clicked into place on the outgoing side of the 3RTx contactor series (Terminals 2/4/6). Contact to the input is made via some pins sticking into the contactor. If you choose to use them make shure to have a good look into Siemens’s catalog, there is a lot of variety that looks like it fits but it is for a different „frame-size“ or has the wrong electrical characteristics. Usually there’s a legend to the Partnumbers somewhere in their catalog. Sticking to best practices in cabinet building, like running shielded cabling for the sensors as close to the interface as possible and adding in some filtering to the inputs could, like many others suggested as well, be the easiest solution in my opinion. Crimping pigtails onto or twisting the cables screen to a single strand is out of favor nowadays, like the clamps that you made for the vfd‘s filter your supposed to connect to screens with as much undisturbed area as possible to get low impedance against higher frequency interference due to skin effect. Very much enjoying the project, and how it’s coming along. Your videos are something to look forward to on the weekend.

[@RotarySMP]

Thanks for your tips and kind feedback.

[@SeanBZA]

@@RotarySMP Yes the grounds to the shields should be made with larger cable, at least the same wire area as the conductors. Simplest is to take thick stranded wire and use that, and the connection to the star point should be through a braided wire, short as possible, and close to the VFD output. The install manuals for the VFD's normally include quite good instructions, at least for the EU manufacturers, like Omron and Siemens, as to how to ground the cables and install the input and output filters, plus if you use line reactors as well. The contactors the supressor modules are connected across the contacts, and often they unfortunately also cannot be used with auxilary blocks, though as you have 4 pole relays, and only need one feedback connection, you easily can use the spare contact position on the main contact plane to provide this function, allowing use of the supression block. I needed some, but they were not available for that particular relay, so I instead made my own, using separate capacitors and resistors instead, wired across the contacts. 47nF 1500 polypropylene capacitors, and a series 10R 1W resistor, which worked to keep the contact arc down to a lower level, and not burn out the contacts every 3 months, as they had a very frequent operation. I was tired of changing that contactor, so went with a plug in 11 pin relay instead, as it is fast to replace, but the second one with the supressors lasted till the machine was sold more than a decade later.

[@RotarySMP]

Did you just run those resisters and capacitors over across the top of the contactors from input to output on each line?

[@SeanBZA]

@@RotarySMP Yes resistor in series with capacitor, and run over the top of the contactor.

[@wizrom3046]

You could bump that 5uS debounce up to 500uS, it is just for end stops and even 500uS would be no significant mechanical delay. My suggestion of adding 0.1 uF caps to the hall sensor input terminals has been field proven on a lot of equipment and does basically the same thing as increasing the software debounce. In my opinion it is better to be in the habit of suppressing the actual noise on the signal input, rather than just telling the software to ignore more noise. 🙂

[@RotarySMP]

Thanks for the tips. What type of capacitors would you use there? Given it is a 5V signal, what voltage rating would the Caps need?

[@zaprodk]

@@RotarySMP A small ceramic or foil type, 50V rating is all you need.

[@RotarySMP]

@@zaprodk Thanks for that.

[@wizrom3046]

@@RotarySMP ... ceramic tend to be brittle and can physically crack when wires are tugged around, I prefer greencaps (polyester) or monolithic polyester called mono caps. Voltage rating not important they are all 16v or 25v at least. Physicallymlarger ones are better just for physical integrity. They are not polarised, so you don't need to identify a positive lead or anything like that. Generally I install them with a bit of lead showing (8mm ?) It gives some strain relief when tightening the terminals, and of course with the writing on the cap visible so you dont have to bend it around to see what value it is later.

[@LesNewell]

By default LinuxCNC runs a servo loop interval of 1ms so a debounce of 5 will be 5ms, not 5us. That's a bit long for homing but fine for estop. Maybe only debounce estop and have the home inputs before debouncing.

[@philip_fletcher]

Another enjoyable, informative Sunday coffee break video - thank you. Good to see the software squirrel coming to the rescue - in the absence of ferrets, what would we do without it?

[@RotarySMP]

Thanks Philip. I am glad to be able to at least move on with programming the control logic.

[@mariopukac1]

Hy from what I read vfd not supposed to be shut it off by contactor because the spindle can turn into generator and probably create those spikes if you think about it make sense

[@RotarySMP]

Mine isn't turned off by the contactor when the motor is running. The VFD contactor is only used at machine start and shut down.

[@Bob_Jones_]

A bit of debounce in the code saves the day. Nice

[@RotarySMP]

Glad about that as well.

[@swatcat7928]

I think the problem is the Limit shitch... when is not activated, the input for the switch has high impedance.. meaning it can receive radio impulses.... MESA card has optocoupler?

[@RotarySMP]

I am not sure.

[@swatcat7928]

@@RotarySMP if it jus a logic input means a 1u amp of interference cand activate that input.. try to pus a resistor in parallel as a dumy load.. 2.2kohms for a 24v input should be ok.

[@RotarySMP]

@@swatcat7928 This is only a 5V signal. It is a Hall effect sensor.

[@swatcat7928]

@@RotarySMP ok.. 100R 0.25w resistor.. try it without debounce....

[@steveggca]

Software fixing hardware for win!😁 nice troubleshooting, the debounce function does exist for a reason. there is still the underlying problem of the noise. every machine cabinet that I have ever been in uses the spark suppressor module when controlling motors, even if entirely analog ,mostly to protect the contacts themselves from erosion and welding , but it also stops the noise .

[@RotarySMP]

Thanks for your Tips on them Steve. I now know how to install them as well. :)

[@steveggca]

no problem . I noticed that there are 2 pairs of 3 wires leading away from the magnetic contactors Is the motor wired with all 6 wires or is the the phase rotation crossover done at a terminal block? If it is the second ,there is a module that connects to the MC pair that does the rotation crossover and then a single output for 3 motor wires direct or the spark suppressor can be used. adds simplicity and eliminates the need for 2 spark suppressers. By the way if you are the neurotic sort, there is an available mechanical interlock between the 2 MCs that will physically prevent the second MC closing while the first one is.

[@RotarySMP]

@@steveggca That Siemens electrical Lego is pretty cool. Yes the two sets of wire rejoin at the terminal block. I have the interlock down through the auxiliary contacts, opening the circuit to the offside coil.

[@steveggca]

@@RotarySMP (A brief interlude from the regularly scheduled comments 😉) WAY back in the late 90's, all of the machine tool builder introduced direct drive spindles in their smaller lathes and machining centers (ie Mori seiki sl-250 lathe or MS sh-40 horizontal machining center) These spindle motors had high/low motor windings controlled with 2 magnetic contactors. this is where I first encountered mechanically interlocked MC's . I'm sure that they existed long before that. I don't recall much of the shifting logic, but I do recall that both the NC and NO contacts were monitored, and that because the MC's were located between the Invertor/VFD/spindle drive, the Inverter output had to be paused, one MC opened and confirmed open and then the other was closed and confirmed closed. finally the inverter output would start again. the spindle didn't have to stop turning to shift, and the shifting took about 2 seconds in total ,mostly the mechanical time for the 2 MC's

[@thomasrahm]

Big shout out to the linuxcnc wizard who came with the suggestion! The KISS principal always apply. 🤓

[@RotarySMP]

Thanks Thomas. It was a nice checklist he made.

[@HansWurst12345]

Software layer debounce is perfectly acceptable. all PLCs use it, the factory setting Siemens sets for all standard input modules is a whopping 3,2ms. Sounds a lot but if one factors in cycle time of the main loop it becomes a non issue

[@RotarySMP]

Thanks for the feedback Hans.

[@bostedtap8399]

Most likely, this fault may have existed when built, but the original controls sample time, was not fast enough to pick it up.😮 Many sensors have built-in filters, ref advance in microelectronics, but you definitely need to invest in an Oscilloscope, why not 🧐 lol. Having both Clean Earth, and Ground is good practice. Thanks for sharing.

[@RotarySMP]

Thanks. A generous YT Member is sending me his old Osci.

[@bostedtap8399]

@@RotarySMP Most magnanimous

[@graealex]

Just fixed it in software, what a madlad. I spent two days searching for the OWON original box, as it crucially contains the power supply, because why would the oscilloscope charge with 5V USB when you can have a 8.8V specialty charger with a deeply recessed DC jack that no adapter would fit. I'll send it over with some other stuff next week. Maybe at some point you'll revisit measuring the low-voltage lines for the EMF spikes and fix it in hardware.

[@RotarySMP]

Hi Alexander, No worries. If I was sure that the Siemens snubbers could be attached to my contactors, I would use them.

[@graealex]

@@RotarySMP In the end, you would probably need to keep the shielding intact until very close to the contactors. I'm also not sure about the low-voltage signaling lines, and how they are terminated in your Mesa card. I assume the input is a FET, so very high impedance. That means it is very sensitive to interference from other lines. In addition, even for normal signals, it can create reflections. Basically, when a voltage changes, it actually travels along a line, as obviously transmissions aren't instant, like a sound wave in a tin can telephone. That impulse travels, and at very high impedance ends, it simply reflects and travels the other way round. A simple but hacky solution is to load it down with a simple resistor in parallel to the transmission line in the screw terminals. People suggesting a cap also aren't wrong, because at high frequencies, capacitors act like a resistor.

[@RotarySMP]

@@graealex I didn't pull out the ground circuit. The noise may be induced in there. That is a much more complicated web of runs.

[@julias-shed]

Glad you got it working 😀 think the Hal counts as software debounce

[@RotarySMP]

Yes, it does. I am also glad to be able to move on with the control programming for the spindle.

[@TheDistur]

I love looking in those tidy panels. It would be absolute spaghetti if I had done it. 😂

[@RotarySMP]

I had to name every wire and make a a wiring diagram, as I have a head like a sieve, and would not be able to trouble shoot it a week after I did it :)

[@TrPrecisionMachining]

good video RotarySMP

[@RotarySMP]

Thanks for watching.

[@tuberroot1112]

So you smothered the interference with a software low-pass filter instead of eliminating the cause. You mention Hall effect sensors but we don't see you circuit or wiring. The unshielded cable run from the sensors to you controller board is a massive aerial antenna to pick up all the EM interference in the workshop. You should use some kind of shielding on those cable runs. More to the point , find the *source* of the EMI and shield that.

[@RotarySMP]

I am not finished, I just have a fix that allows me to continue with the programming. Most of the cable is shielded. It is only that last 1m within the race way which is unshielded.

[@Bluecoldcaffe]

Have you measured the difference in home position between the filtered and the unfiltered input?

[@RotarySMP]

Should be the same, as the homing is first a bit faster to the home switch, then back off, and a slow approach, and finally, it turns past the home switch to the next encoder index.

[@TheDIMONART]

Try to put VFD from other electronics 1-2meter away, in "Faraday's cage"☺️

[@RotarySMP]

With the filters I installed, I haven't had any issues from the VFD.

[@jnorris8649]

Haven’t finished the video yet, but have you considered using shielded cable?

[@RotarySMP]

In the cabinet? I may have to.

[@jimsvideos7201]

Very good knowledge to tuck in a grey cell.

[@RotarySMP]

Thanks Jim.

[@DreitTheDarkDragon]

8:06 - it goes in between three phase wires and your contactor, pins on top of module plugged instead of wires into bottom screw terminals of contactor. Input power then should go from above if I'm not mistaken. There is actually option to buy even "motor starter" (your E1 and E2 in cabinet) like that, but of course any addon means additional height and it might not fit in your cabinet. I had to replace few contactors with motor starters in cabinets few years ago and it was real pain, everything was so tightly packed! Actually open google and try "siemens combined contactor motor starter", that gives some pictures how it's supposed to look like :) Last time I was too busy and forgot to type comment but **what kind of hall sensors are you exactly using for homing?** Outputs can either be "push-pull" or other one I think it called "single ended". I'm honestly surprised that push-pull aren't very common. 1) **Push-pull** means there are two transistors which force output between two states, it can be either strong 1 or strong 0. If some EMF appears around, transistor will keep try its best to keep line in desired state. 2) **Single ended** sensors use one transistor to push (or pull) output strongly into one state BUT other state is "floating" and only some high value resistor is keeping it in desired state. Of course higher value of resistor means less wasted heat but higher sensitivity to EMF. If your sensors for example can provide up to 100 mA current, I'd calculate resistors to use around let's say 30 mA and place them closest to MESA card inputs. Is your sensor PNP type, meaning there is positive voltage when it's switched on? Then connect resistor in parallel between MESA card input and ground. If any EMF appears on wire, it will need to be much stronger to overcome that resistor. Of course if you plan to keep machine standing on limit switches for some time (some emergency error), consider heat dissipation! But for short clicks in normal mode even some lower power resistor might be fine, at least for test if that helps or not. For 24V and 30 mA it would be around 820 ohm 1W resistor. For 5V and 30 mA it would be 150 ohm 0.2W resistor. I'm not sure which voltage you use, I think you mentioned 5V (so probably even more sensitive to EMF) but I'm not sure. Resistors mentioned above are from E24 series so should be easy to get.

[@RotarySMP]

The Hall sensors seem to be a Schaublin built part. The sensors are cast into epoxy tabs, with the leads potted in. There is no part number or other marking. They are 5V devices, Active high.

[@DreitTheDarkDragon]

@@RotarySMP Ah, so we can only guess what is inside :( But I guess original board for sure loaded sensors more than modern MESA board, so some pull-down by few miliamps might actually make difference. But it's just suggestion you might look into :) Or some RC circuit, which will filter out quick weak pulses but can be easily overtaken by strong signal from sensor. You might also try to cover that short section in cable tray in tinfoil (or just use shielded cable) and connect it to PE :)

[@freddylejoly5781]

Bonjour, J'ai déjà connu ce problème au boulot. Usine SPA MONOPOLE en Belgique Il s'agissait du blindage, il "doit" être raccordé à la masse du coté où le signal est émis mais pas à l'autre bout du câble." UN" seul point de masse, cela peut être au passage dans le coffret électrique.. Bonne continuation dans votre développement. Freddy

[@RotarySMP]

Merci d'avoir partagé votre expérience.

[@TheDIMONART]

Well, if you using VFD, it can cause very heavy EMI interference.. It's not a problem for industrial machines, because they use servomotors with 50Hz impulse frequency..But EMI- real headache for DIY CNC machines/routers/lathes...

[@RotarySMP]

That is why I used the filters both in front of the VFD and in the motor lines.

[@LCalleja]

Haven’t had time to watch the video yet but will get too it soon as I am on 10 of 46 :)

[@RotarySMP]

Hi Luke. Thanks for doing this.

[@LCalleja]

grate video glad to see you found a way to sort it out.@@RotarySMP

[@RotarySMP]

@LCalleja Yeah, even if it is only a software band-aid, for now I can move on to programming it.

[@LCalleja]

its a step forward at least that the important part @@RotarySMP

[@RotarySMP]

@@LCalleja It was.

[@besrizhurnalov8485]

Just to satisfy my curiosity. What current flows when the contact is closed to the Mesa card (24V goes to the mesa input.) Thanks for the video!

[@RotarySMP]

I haven't measured that.

[@LongnoseRob]

Good work!

[@RotarySMP]

Thanks Rob.

[@MakarovFox]

nice

[@RotarySMP]

Thanks.

[@raindeergames6104]

I have a,similar issue. I split open my loom and removed all the wires for spindle and steppers away from each other and the problem is gone. My machine came without shielded wires on any of the steppers or spindle.

[@RotarySMP]

I considered that, but the way my I/O card is aligned, I will have some parallel runs regardless which way I run things.

[@raindeergames6104]

@RotarySMP I'm upgrading ny machine to centroid acorn now and changing out all the wiring for steppers en spindle with high quality shielded wires. I'm keen to see if the problem resolves with the new wires. It sucks to have the whole loom split open like a octopus just to get the machine to run.

[@RotarySMP]

@@raindeergames6104 I'll be keeping my fingers crossed that you dont have any EMI problems.

[@624Dudley]

I, too, suck at oscilloscopy. For proof I need only watch CuriousMarc at work with modern Keysight equipment, repairing classic HP oscilloscopes. And Apollo comm equipment. And many others. 😢

[@RotarySMP]

Hi Doric, Any progress on that DoAll pick up? Cheers, Mark

[@soucevit]

Hi, the Siemens module should be in line with the contactor output in my opinion. But be careful. The white is mostly the old obsolete type of the contactor and gray is the new one. The accessories do not always fit together. So you might need to buy new contactors for this solution. It looks like it will slide instead of removable contacts, but I am not sure. I never saw these suppression modules in use myself. I am not sure what is actually inside. What I would do instead is to use a shielded cable from the terminals to the IO card. I have seen all delicate signals connected this way. Shielded as far as possible. Debouncing is a good idea but it might be covering stuff up and the spikes can in the end fry you IOs. I once had a servo system with roughly 100m between VFD and the drive (terrible design) We needed to do this on every spot and then it somewhat worked. But it was not reliable anyway with switching a couple hundred KW loads around it. It was on the big stretching press they make aviation grade aluminium on and servos were supposed to move some measuring arms. The project that nearly made me quit...

[@RotarySMP]

THanks for that. That module has RC pairs in it. So you just plug it into the output with the wires?

[@FullSpeed_only]

Very good Points in the Comment. Debouncing is not a bad Thing, but you might want to do it in Hardware to keep your Inputs healthy. And the Spike might also occur in other Places that are not throwing Errors at the Moment but still can be problematic.

[@axelweimann5274]

@@RotarySMP @soucevit the Addon Modules Clip in the Bottom part of the contractor and the pins align withe the screw terminals so clip in and fasten the screws they should be a version for the light gray ones too .... this is model number for the light gray Version 3RT2916-1PA2 3RT2916-1PA1 but i think the main problem is that the controller card picks up any emv in that box so try to shield the cable and the card

[@RotarySMP]

@@axelweimann5274 Thanks Axel. That really helps.

[@steveggca]

Hi the module shown by Mark was listed as the replacement for the obsolete part. I found the original model listed on ebay and amazon here in canada, so finding it in europe should be even easier. Those suppression modules exist primarily to stop the contacts from arcing ,eroding ,and welding but have the bonus benefit to also stop noise in digital systems.

[@jesseservice7828]

Muito bom ! Obrigado.

[@RotarySMP]

Thanks for watching.

[@DPTech_workroom]

Signal wires and power wires should go far away from each other. If needed, they can cross each other at 90°, but never be parallel. Sometimes helps ferrite rings, or capacitors (like the debounce)

[@RotarySMP]

Thanks. I should have laid out that control cabinet differently.

[@DPTech_workroom]

​@@RotarySMP The shielding must be correct too (PE). No loops in the hole system. If the main motor has PE connection, the servo motors too so you will have 2 loops already. It will work like an antenna for catching interferences.

[@RotarySMP]

@@DPTech_workroom I dont really understand what that means in practice for this installation.

[@DPTech_workroom]

@@RotarySMP 3 phase motors use only 3 wires. PE connect only to the lathe. The same thing with the drivers. If the mounting plate connected to the PE, so the most drivers heatsink also is the PE (but need to ask others too about this. I'm not sure)

[@RotarySMP]

@@DPTech_workroom The driver cables were factory manufactured. I just hooked them up were the manual.

[@bogenjaeger98]

Der Drehstrom Überspannungsschutz wird unten an des schütz gebaut, anstatt eines Bi Metall Motorschutzrelais.

[@RotarySMP]

Hast du ein Link zu ein Photos bitte?

[@bogenjaeger98]

@@RotarySMP suche auf Wikipedia nach Motorschutz (Elektrotechnik) Und unter Überlastrelais findest du ein Bild wie es zusammengebaut wird. Siemens zeigt leider immer nur bilder der einzelnen Komponenten.

[@RotarySMP]

​ @bogenjaeger98 Prima Danke. Jetzt ist es mir klar.

[@samrodian919]

So the "quick and dirty" fix worked! Lol

[@RotarySMP]

It is a start :)

[@r0311]

What I noticed during watching, that you don't have a separation between low voltage control signals and high/mains voltage for the motors. This is for sure a safety issue but proper wiring here would also prevent the surges you have on your control signals.

[@RotarySMP]

Yep, I learnt a lot about wiring doing this project, but some of the early ignorance let to design decision (=mistakes) I can no longer easily change.

[@7duser10]

Interesting debugging video.

[@RotarySMP]

Glad you enjoyed it.

[@7duser10]

@@RotarySMP Would have been good to see a root cause electrically, but I suspect it would have taken a while to diagnose. Glad the debounce worked for you.

[@RotarySMP]

@@7duser10 It would probably need a better Osci to isolate the source.

[@loydsa]

I am not sure how useful (or otherwise) what I am about to say (type) will be but here goes 🙂 As a general guideline never coil wire in loops whilst it satisfies our need for neatness its a bu**er for causing EMI. Also when you zip tie cable bundles, don't zip them tighter than is functionally needed, the closer we force wires together the greater the risk of crosstalk. The longer unshielded wires run next to each other the greater the risk of cross talk, if you know which circuit is picking up the interference try routing its wire along a different run for the majority of its length if you can. If all else fails ten or so turns through a ferrite toroid should sort you out.

[@RotarySMP]

Thanks for your Tips. I have some ferrites coming in.

[@SeanBZA]

@@RotarySMP Plus with signal wires twist them into a pair, signal wire and the associated ground, so that interference picked up is going to be mostly cancelled out. For an active sensor with 3 wires, all 3 twisted together tightly in the loom till at the controller works best, and if possible make all the wire, even in the raceways after the connections, shielded, with the shield connected at only the one end, probably the termination block. Separate shield connection than the motor shields, and then each one connector shield set with a separate braided wire to the chassis star ground point. Yes can be a pain to do, but gets rid of a lot of noise, and then a ferrite toroid over each separate shield as they come into the cabinet, before the connection area, helps as well.

[@RotarySMP]

Thanks for the best practice. Sean.

[@cornelisvanrensburg1826]

Your wiring screen cable should be connected to -24v dc only on one end. Connecting to earth does not provide screening. I had a gas detector trigger in a room, everytime the 3phase motor brake would activate. Connecting screen cable to earth did notthing. Connected to - supply of detector and all issues went away

[@RotarySMP]

Oh, good point. I thought screens always wen to common earth.

[@x_ph1l]

It's already too late, but generally it's a bad thing to run signal wiring along the high voltage/high current wiring. It's a good thing that this issue was easy to reproduce. Much worse if it would've been an intermittent issue that you found while running some real parts.

[@RotarySMP]

Good point Phil. Intermittent faults can be a nightmare to isolate.

Using zero-switching SSRs instead of mechanical contactors would very likely avoid any current spikes on the 3-phase wiring entirely, eliminating interference to the low-voltage lines.

[@RotarySMP]

Thanks for the tip.

[@theafro]

Fixing a problem in software is great, until the next time it bites you! It's also not as satisfying and doing it the hard way. It bloody works though!

[@RotarySMP]

:)

[@engineerwrecker8153]

Why would a voltage spike trigger the E-Stop? The E-Stop is voltage high during normal operation and goes voltage low during an E-Stop event. It is wired that way so if something happens the system fails safe. It sounds like your E-Stop is backwards. 🤨

[@RotarySMP]

Schaublin used Hal sensors for X axis end stops. So these are Active high. They trigger an E-Stop, from LinuxCNC but are separate to the hard wired, NC E-Stop loop.

[@engineerwrecker8153]

@@RotarySMP Ah ok, I understand now. Even though Linxcnc is calling it an E-Stop, it’s a limit fault not an E-Stop. In the USA if it’s not part of the hard wired E-Stop chain, or wired through an E-Stop safety PLC it’s not an E-Stop.

[@deanworrall4567]

is your switch cable shield / screen grounded both ends? its only supposed to be one end.. you say "single star ground"... i assume your panel is properly earthed and that's just whatever your local area calls earth.. ?

[@RotarySMP]

Yes, correct. I dont think Schaublin terminated the shield at the Hall sensor end.

[@deanworrall4567]

@@RotarySMP I’m a control panel builder by trade so I’m just going back through the rest of the series ( this was the first one I saw on my suggestions ) to see if there’s anything obvious in the electrical wiring. The only thing I can see so far is maybe the pcb boards would benefit from some form of emf shielding as they don’t appear to be “industrial” components.

[@RotarySMP]

@@deanworrall4567 I am using Mesa boards. Not sure if they would considered industrial quality. store.mesanet.com/index.php?route=product/product&product_id=306&search=7i92 They have a good reputation.

[@deanworrall4567]

@@RotarySMP I’m not sure which I/O card you’re using ( I’ll check in a while when I’m back home ) but it looks like the 7184? It says the threshold and filters can be programmed on the card itself so might be worth a look into the setting of the daughter board unless you’re happy with the software debounce you set?

[@deanworrall4567]

ok I've gotten through a few more of the panel building episodes and so far there's a few possible things I've picked up on.. Your power supply 0V lines don't appear to be grounded so there is not a common reference point for your 5V and 24V. Your IO cards are on plastic mounts and are not themselves grounded.. the little metal bits round the screw holes usually ground them to the chassis when mounting them on metal standoffs ( like those that were in the drive you blew up and repaired.. )

[@bradthayer6782]

Sorry the bounce in your step is gone. But well done!

[@RotarySMP]

Thanks Brad.

[@cnc_workbench]

seems like your spindle was acting like a generator during deceleration, pushing some extra back voltage into the system, the debounce seems to act like a filter to cut that spike.

[@RotarySMP]

Thanks for the explanation.

[@TheLukemcdaniel]

It's a U not an O, linUx(ucks, or ooks depending on accent) not oxe....

[@RotarySMP]

Since I learned and speak mostly German, I screw up my english vowels.

[@robertfontaine3650]

Glad you are able to move past this for now. Electrical gremlins are evil

[@RotarySMP]

I agree.

[@ulwur]

You dont suck st oscilloskopy, that device sucks. I have the same, useless for anything but repetetive low frequencies.

[@RotarySMP]

That is good to know. I thought it was me :)

[@der.kleine.pyromane]

Erster 😂

[@RotarySMP]

Thanks for watching.

[@leckmiamoarsch3222]

The additional rc module adds below the contactor, wher the wires enter

[@RotarySMP]

Thanks for that.