-
10:30 in the morning here in Hamburg.
Tim, thx a lot again for your time and your very informative and educating comments e.g. on what we can see in the charts. You're closing a gap within the information I found online so far. Hope there are and will be some others following your explanations.Oh, I take it you won't be reversing the one motor while running? The SPDT will be open in the middle (well, unless you get a make-before-break type; most devices are break-before) so should be snubbed if it will be switched under power.
The manual switch was a 3 postion one: dir1-off-dir2.
I swapped in two Relais on-on in series like in the schematics: 1st is on-off, 2nd is dir1-dir2. Before change of direction I switch off the motor and allow for spin down:
digitalWrite(PIN_PLATTER_POWER, HIGH); //switch platter off
delay (2000);
digitalWrite(PIN_PLATTER_DIRECTION, HIGH); //switch direction
delay (500);
digitalWrite(PIN_PLATTER_POWER, LOW); //switch platter on again
I will do some more repetitive measuremens later to doublecheck for the delayed surge with the arm motor 1.
And some tests with smaller caps as suggested and come back.
I wonder, if I should work into some simulation. What would be the appropriate no or low cost SW? LTspice?
Would it be worth the effort?
-
Ah nice, then that handles direction.
Modeling is hard, because sparks are hard to model -- but basic models can be found, and you can easily set up the EFT scenario for example. Likewise, the motor is hard to model unless you have complete measurements of it -- also its minimum spinning voltage/current, to know what threshold to set for leakage. But there too, a basic RLC model will do the job, just to show off how it works, if not necessarily the particulars.
Which in turn gets into matters of RLC equivalents, damping factor, stuff like that -- the voltage and current aren't so important, as the ratios between them (impedance), and other things in turn (e.g., note that sqrt(L/C) has units of impedance).
Tim -
Here are some more results with
Motor 1
This is the line synchronous clock drive motor that drags the suction arm across the record with L=9.1H.
In the original setup of the machine it was this one that had the 47R / 0.1µF RC snubber across the switch.
The resistor seemed damaged. Here’s a better picture, and damaged it is indeed.
Here‘s this 47R/0.1µF snubber with new R and C components.
Channels/Colors are:
• C4, green: Voltage over the switch and snubber
• C1, yellow: relais control line of the ATmega (+5V off, GND on) for trigger and noise pulses
• C2, magenta: Voltage over R
Switching on …
… there’s 400V over the resistor at first, which would be 160.000/47 = 3.4kW.
I found a Vishay paper on that topic.
https://www.vishay.com/docs/48516/_ms9702509-2003-vishaychecklistpulseload.pdf
We see an avarage of about 1000W over 15µs. Vishay suggest to average for 3ms, which would be 5W only. That would be OK acc. to the paper but was obviously not acc. to the picture above.
Switching off:
Generally the snubber seems to work fine:
But the capacitor ist too big and the motor is rattling against the gear in off state.
Rattling does not vanish with 47nF, but with 22nF it finally does.
Here’s switching on with 47R/22nF:
There’s 300V over R or about 2kW.
Switching off shows, the snubbing works OK. Here are two different switch off cycles:
To limit the power pulse into R I’d settle with 200R/22nF:
Switching on:
Voltage over R is 300V, initial Power 450W.
The sharp drop down period is strange. It’s more visible here, but we can see it for 47R/22nF above too.
Snubbing the switch off is not as good as with 47nF but still OK. Here are two different cycles:
Overall 200 Ohm / 22nF to me seems to be a good compromise for this motor:
No 'off'-state rattling, acceptable snubbing and acceptable 'on'-pulse into the resistor.
-
Nice!
Got a nice capture of contact bounce too...
The pulse width BTW is about the RC time constant, so take peak power over that time scale and that's about what rating you need. Probably a metal oxide or wirewound resistor would be ideal. Peak power doesn't depend on capacitance, but gives a shorter duration and less energy --> more pulse capacity for the same resistor. And yep, increasing the resistance looks to be a good idea, had some ringing with the lower value and reduced C, more R dampens that out nicely.
As for the common mode -- that's what you're seeing on the ATmega trace. Most likely it's a probing artifact, because your other probes are connected to the line that's doing the transient. This will be reduced with better isolation, and then for whatever's left that still reaches the board -- well, the board really ought to be designed to withstand such things, with adequate shielding (ground plane construction), filtering at all inputs/outputs, etc. (One downside: programming cables are rarely shielded, so can be difficult to program/debug in the presence of noise. Probably, your mains isn't noisy enough for this to be much of a problem.)
Tim -
Tim, thx a lot again for your comments and ideas!
Meanwhile I did some more tests with the remaining
Motor 3:
This is the two directional permanent 10W split capacitor motor that drives the platter of the record cleaning machine.
All tests are done with the same motor direction, only the on/off relais is used.
In the original setup there was no snubber installed.
I showed some shot already, but anyhow here are two more shots switching off without any snubber:
C1 yellow: relais signal (active low)
C4 green: Voltage across the switch
C2 magenta useless (forgot to deactivate):
Following the calculation done above, the snubber should be R 2k2, C 1.5µF.
But with 2k2R/470nF, 4k3R/270nF and 100R/270nF the Motor runs in full action with the relais open.
With 200R / 47nF the motor hums but does not rotate, with 200R / 22nF it hums not that loud, even with 4k3 R / 22nF there’s some hum left. (I doublechecked to switch life not neutral.)
This is ‚off‘ with 4k3R/22nF:
C1 yellow: relais signal (active low)
C4 green: Voltage across the switch or RC
C2 magenta: Voltage across R
And 200R: / 22nF:
Here, there are near to no spikes on c1.
So this might be the RC to go with, if we ignore the humming of the motor.
To address the humming it might be an option to e.g. half a second later switch off neutral too with a spare relays I have on the relais module.
I also had a look on the ‚on‘ switching, which shows some contact bounce, but maybe tolarable …
… compared with 4k3R / 22nF ‚on‘:
As a next step I will implment the switching of neutral, which shouldn’d have any effect beside stopping the hum. But maybe the hum is a live to earth ground effect. An earth ground of the motor was missing, but I had added an earth connection again.
And I will have some tests with an china clone Omron G3MB-202P SSR module that just arrived.
https://www.codrey.com/electronic-circuits/solid-state-relay-ssr-primer/
As always, any comment is highly appreciated.
-
Here’s the test with the same
Motor 3
… of the last post but with the china clone Omron G3MB-202P SSR module instead of the mechanical relay module. The G3MB-202P is designed to switch at zero crossing. Which it seems to do.
SSR, no snubber, switching off:
Of course there is no bouncing any more and no noise on the relay control wire (yellow), but still a huge induction voltage peak of -800V over the SSR. (I needed to know and was lucky: The SSR survived.)
With 200R/22nF snubber, switching off:
And to get the complete picture:
With 200R/22nF snubber, switching on:
Finally everything is fine. 😊
…except that the motor ist humming with the snubber in place.
In case I go for the SSRs I will stick with an mechanical switch für L and N for the whole machine for security reason. Or better go with an contactor that could be swiched off completely by the Atmega and a mechanical relay after some idle time.
To prevent the motors from humming I had the idea to switch off N for all motors e.g. 500ms after all motors are L-switched off, and switch on N before switching on the first motor.
Now I’m thinking about two points:
1
I did some research and all sources agree that it is not OK to switch neutral alone. You shall always swich N together with L. Would it be OK to switch N alone, when you know L is switched off before N, or is it important to have the L and N switch ‚mechanicly‘ coupled?
2
Is it possible to use an SSR to cut N after L is ‚off‘ already, or does it need a voltage difference at the switch contacts?
-
Going back to an earlier idea, you can put the snubber across the load as well, and this will not cause leakage current to flow. Prerequisites are having a bypass cap across the input (check) and short enough lead length between cap, switch and snubber (some feet will be okay).
SSRs are much slower than sparking contacts, with risetime in the 100ns range, and turnoff delayed until load current drops to zero (so the amount of induction remaining in the motor will be small).
Note there are two kinds of SSR: AC-only and AC/DC. The former uses a TRIAC, which latches on until load current drops to zero. The voltage drop is modest (1-2V), not varying much with load current. The latter uses back-to-back MOSFETs, with a stack of solar cells opposite the LED to generate gate drive voltage (typically something like 6-8V at a few µA); due to the tiny current into the large gate capacitance, switching is very slow (a few ms), and capacity is not usually very much, especially at line voltages; they do however excel at low voltages, where 10s of A ratings are easy enough, and of course being suitable for DC.
The G3MB is a TRIAC type.
TRIACs also have the property that, when exposed to overvoltage (breakdown), they latch on. Or they usually do. I'm surprised to see what looks like a flat-topped spike there, and 800V is plausibly the breakdown voltage of that device (the actual breakdown is always somewhat more than the rating (output PIV >= 600V) so this makes sense). I wonder if it just wasn't enough current to make it turn on? Certain types of TRIACs may be more robust to this stimulus, though I haven't seen any white papers documenting it yet.
It also claims to be a snubber-integrated type, though clearly not too much (value) compared to your load, and they don't document what that value is, unfortunately.
TRIACs can be destroyed under certain conditions, which I'm not real clear on, honestly; because they tend to latch on under breakdown, they tend to be quite robust. The most common failures I think, are overheating (too much current flow, die overheats, melts, whatever), and dI/dt (too fast turn-on causes current flow to be uneven across the die, causing local hot-spots and rapid failure). The waveform shown, should be okay: if it's not enough avalanche (breakdown) to cause damage outright (and I think they would latch on before that point?), it's fine. Most semiconductors act like a zener/avalanche diode in breakdown, reliably handling some current flow that way; typically just not a whole lot more than that, unless rated specifically so (like common rectifiers failing at a few to 10s of mA, or avalanche-rated MOSFETs handling up to some amperes but not more than that).
Overall, an SSR will do a fine job here, free from the sparking of mechanical contacts, and (optionally, it would seem?) with a snubber to dampen the load.
Tim -
Tim, thx a lot again.
The question what exactly happens inisde the SSR exists somewhat cloudy in the back of my mind and you've already anwered it.
Overall:Overall, an SSR will do a fine job here, free from the sparking of mechanical contacts, and (optionally, it would seem?) with a snubber to dampen the load.
After these first measurements I plan to go with SSRs. I will check, if these 2A SSRs are able to switch the 0.5A pump motor without cooling, otherwise I have some chinese "Fotec 40A" block ssrs somewhere, overkill, but definitely able to handle the motor's start current.
Even if not strictly necessary, I'd like to stick with snubbers, now that I've spent some time and beginning to understand what's going on.
That spike inspired me to think about a MOV again to get some security in addition to the snubbers.
Put snubbers on the load, not on the swich:Going back to an earlier idea, you can put the snubber across the load as well, and this will not cause leakage current to flow. Prerequisites are having a bypass cap across the input (check) and short enough lead length between cap, switch and snubber (some feet will be okay).
I'm happy to follow your advice to put the snubber on the load. So I can stop thinking about weird N switching.
Distance to the small motors is less than 1 ft, to the 80W pump motor it's 10 feet. It would be nice to have that snubber with the others, but it's possible to mount it on the pump motor also.
From my basic understanding of physics I could imagine, that design and calculation might stay the same for R and C, so that the same values would be right?
How do I size a bypass cap? Are the 0.1µF sufficient that are in place now?
MOV:
Thx for your ideas regarding the 800V spike.
There was no stutter or a motor running at that time, nevertheless I like the idea to cut spikes with a MOV.
I found a nice design paper, aplication note and datasheet:
https://https://www.tdk-electronics.tdk.com/download/531266/5897f2786eaab0be1e043f85ba4ca55d/siov-selection.pdf
https://www.tdk-electronics.tdk.com/download/185762/8cd06030e8ddf20bd381c1e8917f0e9c/siov-applications.pdf
https://cdn-reichelt.de/documents/datenblatt/B300/YVLEADED-VARISTORS_EN.pdf
I thougth about oversizing and choosing EPC B72220-S051 510V type, that covers up to 190J over 2ms. 510V would give some leeway above 325V for snbber overshoot to avoid stressing the MOV every cycle. Some useless cent spent due to oversizing would cause no harm, it's no mass production.
When I put the snubber to the load, would you clamp the MOV on the load too or would it be better on the switch?
Ah, and when I order that MOV I can spent some cent on oversized 5W resistors now that I know the values required. -
I'd be worried actually that a 40A SSR might not even stay turned on, with as little load as you have...
MOVs are kind of dubious, because the clamping voltage is fairly high compared to nominal. For example, the 275Vrms family in the first link (Fig.7 p.11) has a nominal rating around 350VDC, 430V breakdown (at 1mA), and 580V at 1A. Which is acceptable for a 600V switch (that's been observed to break down at 800V, even), but it's not much of a margin -- so this solution isn't always possible/reasonable. So, it does seem to be reasonable in this case, which is nice.
Evidently -- knowing that your switch breaks down where it does -- a 385Vrms MOV might even do. A 510V definitely will not however.
You don't need much MOV here. A little SMT chip type would even be fine, but a 5 or 7mm disc will probably be easier to wire in, heh.
Mind that a smaller MOV will have a higher clamp voltage at the same current; checking the curve at the worst-case point, say 1A, should be fine.
You can put the MOV across the switch. If you find that causes too much leakage current, then I would recommend putting it across the load, with some ohms (10 say?) in series with the MOV -- this limits surge current through the switch so it's not just hammering the poor thing when a line transient does come along. The resistor should be a fusible / flameproof type.
Regarding the motor on a long wire -- note I didn't say anything about the load, only the switch, snubber and bypass! This is exactly such a situation where this arrangement is ideal: the stray wiring inductance acts in series with the motor, so we don't care about how much length is out there. Just that the switch and snubber are nearby.
Tim -
Tim, it must be in the middle of the night over there in Wisconsin?! 04:30, something like that.
That a fasts answer was not expected at all.
I will take your comments on MOVs and read the papers in more detail. Obviously I've got RMS and cut voltages confused.
If I may ask again:
- Is a bypass cap of 100nF sized fine?
- Is it correct to just swap the same RCs across the loads, or is some recalculation needed?
Any tip for a nice textbook to study the basics on snubbing etc.? I found some sources online to calculate R C, but nothing really comprehensive for educated beginners. :-)
I'm starting to feel somewhat guilty to take advantage of your generous explanations.
I mean, it's just great to get this one to one coaching. Appreciated! -
It's a popular misconception that I don't sleep; rather, I just sleep when I want to.
Yes, 100nF is fine; basically, it should just be bigger than the largest snubber cap. That might not even matter (the exact answer depends on mains impedance, which depends on frequency; so, at the load ringdown frequency, if the mains impedance in parallel with the cap is less than the snubber impedance, it's fine..) but should be pretty safe. Anyway, you don't need much snub C here so that checks out regardless.
Same RCs, yes.
Theory comes down to: you're making a damped series resonant circuit.
Tim -
Finally I’ve settled with a ‚preliminary final‘ setup.
It integrates all three approaches:
- snubber across the motors,
- MOVs,
- and SSR zero crossing switching
which might be some overkill.
There is one china clone module with 4 Omron G3MB-202P SSRs switching the motors at zero crossing, one for each direction of the platter motor. They stay cool with the ‚bigger‘ 0.5A pump motor and seem to have no problems switching the motors.
I put all of the snubbers and MOVs in parallel with the loads now as described below in more detail. Because place is no issue, I put all of them on a breadboard pcb.
Unfortunately, taking the snubbers from the switches does not fix the hum issue. Because the SSRs do not switch off completely but leave 7V RMS on the load, especially the platter motor is humming and you can feel that hum mechanically when touching the motor.
Voltage across the platter motor when SSR is switched ‚off‘:
So I stuck to the idea to have an additional mechanical relay switching off L for all motors with 1s delay after all SSRs are ‚off‘ (see schematics) and switching on with the start of the first motor. This with a delay to allow the mechanical contact to settle in ‚on‘ position before the SSR starts the motor. It works fine.
For security reasons the complete deivce is switched (L and N) with a small but serious major brand contactor (see schematics). A second mechanical relay (not in schematics) sits normally connected in the contactor‘s coil circuit and engages to switch off the machine completely after an idle time of 15 minutes.
Also not in the schematics are four fuses, one in the power connection for the complete machine as well as one for the 12 and 5V PSUs, one (1A slow) for the motors, MOVs and snubbers, and one for the contactor coil circuit.
In more detail:
All snubber capacitors are Wima X2 rated ones. The snubber resistors are 5W metal oxide resistors now.
The MOVs are the 275V ones discussed above. I did some measurements to verify the effect but was not able to get any significant differences with versus without MOV. As Tim stated above, because of the zero Volt switching of the SSRs there seems to be no overvoltage to engage the MOVs. I decided to leave them in place anyhow. But - as they are oversized and seem not to be triggered into action normally, I did not put a e.g. 10 Ohm resistor in series. In case they end shorted, there is the 1A motor circuit fuse in place.
Motor 1
This is the line synchronous clock drive motor that drags the suction arm across the record with L=9.1H.
It's the motor that had killed the original snubber resistor. From the last test setup, the capacitor now is increased to 100nF, R decreased to 100 Ohm, because the snubber no longer bridges the switch, which in the last setup led to motor rattling.
Switching off:
C2/magenta: SSR input signal (active high now with the new four SSR module)
C4/green: voltage across the SSR switch/load side.
Motor 2
This is the 80W vacuum pump motor, dragging 0.5A.
The snubber here is 470nF/330 Ohm now.
Switching off:
There is some +/-450V at the beginning. I assume it’s somehow caused by the motor inertia, maybe modulated by pressure inside the three pistons. But honestly I have no idea. Anyhow, the 450V are not enough to trigger the MOV into action.
Motor 3:
This is the two directional permanent 10W split capacitor motor that drives the platter of the record cleaning machine.
I did not chage the values. Originally the calculated ones were higher, but as 22nF/200 Ohm worked in the last setup, it’s 22nF/220 Ohm (new 5W resistor) now.
Switching off for both directions:
There is a small overshoot of 370V, but even less critcal than the 450V at the pump motor.
One question left: Does the mechanical relay see any kickback when switching off the contactor’s coil?
Obviously not. There is some bouncing, I’d say, but nothing to worry about. Even when switching exactly at max voltage as it is the case here.
During switching, the ATmega shows no problems at all which was the major objective of this endeavor. I had it counting tenth of seconds on a I2C-display and there were no hickups at all.
Tim and all, thanks for your input.
It’s far from being a professional solution now, but for me it’s a status to work with, and I do now understand parts of what’s going on at least.
I will now print a case for the display unit and start to write some more software to make the appropriate LEDs blink. 😊
As always, any comment appreciated.