emi reduction brushed dc motor

[andrew_c]

I've just gone and failed compliance testing with the EMI of a small 12V DC peristaltic pump being way over the limit. The pump in question - https://www.adafruit.com/product/1150

The test result:




Without motors running:


The EMI from the motors is so random in both frequency and amplitude that this in itself didn't fail us. Instead, it was the motor feedback to the AC power adapter that really did a number on us. I nipped up to the testing facility yesterday with that I had on hand, some 1N4002 diodes (yeah not the best) and a few 0.1uF ceramic capacitors. No matter how many I put across the terminals/onto the chassis, I couldn't get the noise down.

What else I can do to eliminate some of this noise from reaching the power supply?

[Leon23]

You need to provide more information. Is the problem radiated emission or conducted emission?
How does the circuit looks like - schematic of the PCB. How are the cables placed? What motors are used? How are they being driven/loaded during the test? etc

[andrew_c]

Hi Leon,

It's radiated emissions we're dealing with. The peristaltic pump is on the end of 30cm 2-way unshielded ribbon wire, attached to our PCB. The PCB is rather simple with a darlington array switching the load.

The motor is a shenzen special. I've just had one apart and the brushes look completely messed up to me, like they've been machined/installed incorrectly. They look like this!


I'd expected to see something more like this:


The motor is loaded with the peristaltic pump head itself which takes the free-running current from 60mA to just over 200mA. It's a pretty constant resistive load - squashing silicone tube.

[Gyro]

Yes, those brushes are installed incorrectly (90' out)!  That's going to be contributing a lot to arcing on the comutator, it might improve with time, but it's going to take a long time for them to wear down to anything like the right profile - I'd reject it and try another one.

The normal suppression arrangement for low voltage brushed motors is a delta arrangement of capacitors, one across the terminals and one from each terminal to the case. If you look at your second picture you can see two (one broken) yellow capacitors running terminals to case. Adding ferrite beads to the supply leads might help too.

As I said, the first step is to get a correctly assembled motor.

[Siwastaja]

This might be a good fail, because that type of cheapo Shenzen special peristaltic pump will not work for long.

Have you tested it properly? Several samples, for long enough, to see if the pump rates keep up?

I have ordered several exactly similar looking things from Ebay, and they were either fully DOA, or they failed after less than 30 minutes of pumping. When complained, the replacements also were DOA or died soon. The issue is that the slippery motor axle directly contacts the slippery plastic rollers, and it will slip, or start slipping very soon.

Brushed installed in the wrong way is not surprising.

Get some proper peristaltic pumps.

I did solve the issue by moving to catering dishwasher pumps fitted with synchronous mains AC motors, around $50 each. You can get the housings without the motor (AFAIK around $25 each) so that you can fit a stepper, a BLDC or a proper brushed motor - maybe another $25 and have a solution that both works and also have chances of passing the EMI tests.

[martys]

This amount of noise you see  is probably normal, contrary to other replies here. What do you expect, brushes create low-power Marconi spark gap transmitters.

Also, the idea that these brushes are mis-installed is clearly  in error, the motor mfg knew what they were doing when they designed the shape and positioning of the brushes relative to the windings of the armature.

The most efficient way to silence the EMI is to kill it at the source..use a ferrite core(semi-flattened cylinder or donut torus core) with both power leads to the motor separated and each having its wires wound around it (at least a few turns or more  to form a common mode transformer.) Alternatively, add 100uH to 330uH powdered metal/ferrite chokes in series with the terminals to the brushes.

 At the same time a .01 to .1 uF ceramic capacitor from each motor terminal to the motor case and grounding the motor case to the nearest common ground point(Sugg: use>3mm flat copper desoldering wick for lowest inductance) will also greatly attenuate any RF EMI at the point of creation.

For the sake of experimentation, you can easily find common mode transformers and cores in any discarded piece of equipment like DVD players, VCRs, PC power supplies etc.

[Leon23]

This amount of noise you see  is probably normal, contrary to other replies here. What do you expect, brushes create low-power Marconi spark gap transmitters.

Also, the idea that these brushes are mis-installed is clearly  in error, the motor mfg knew what they were doing when they designed the shape and positioning of the brushes relative to the windings of the armature.

The most efficient way to silence the EMI is to use a ferrite core(semi-flattened cylinder or donut torus core) with both power leads to the motor separated and each having its wires wound around it (at least a few turns or more  to form a common mode transformer. Alternatively, add 100uH to 330uH ferrite chokes in series with the terminals to the brushes.

 At the same time a .01 to .1 uF ceramic capacitor from each motor terminal to the motor case and grounding the motor case to the nearest common ground point(Sugg: use>3mm flat copper desoldering wick for lowest inductance) will also greatly attenuate any RF EMI at the point of creation.

For the sake of experimentation, you can easily find common mode transformers and cores in any discarded piece of equipment like DVD players, VCRs, PC power supplie etc.

There is no such thing as a general guide to suggest "the most efficient way to silence the EMI". EMI issues are very dependant on your specific operating condition, how the cables are oriented, pcb layout, component selection, to name a few.

I do agree that using a ferrite core could reduce radiated emission levels. They could be tested to be put on either the +supply, the -negative supply, or both for common mode rejection. I will not give any suggestions on value but if you have any of these close to your lab a quick check might be able to provide information if its enough for sufficient reduction.

I also agree that capacitor(s) could help reducing EMI. I can't say however whats most efficient: Either place them between +- terminals of the motor or as suggested to the motor case.

Siwastaja also provided some valid content here.

[martys]

My suggestions were not EMI generalizations that speaks to all possibilities of bothersome EMI, but specifically applies to EMI of  brush motors. In any case, the idea of using L-C low pass filters, bypass capacitors and common mode transformers is extremely common practice in almost all equipment, every switcher P/S,  to kill EMI.

Connecting capacitors across the two terminals of a DC brush motor doesn't work as expected as neither terminal connects to ground and either or both of the brushes will just couple the EMI nose pulses down the other attached power lead and each wire acts as a  radiating antenna.

When I want to create a peristaltic pump in the lab, I tell the lab handyman to extract a brush motor from a used inkjet or laser printer. (High quality..low cost) to use as the motor for the  pump.

Attempting to overload a motor by continuously running near its stall speed is a sure road to rapid motor failure. Gearing-down  is essential in most cases where the load is non-trivial relative to the torque output curve of the motor at low speeds.

[Siwastaja]

In that design, the tiny (1.6mm?) motor axle directly touches the rollers (about 10mm), which doubles as gear-down, and the gear ratio itself is quite OKish. Of course, the issue here is that it slips.

[martys]

I know this is a MacGyver fix, but:
(1) apply(shrink) a short sleeve of electrical shrink tubing to the tiny diam. motor shaft.
(2)Apply rosin(available at any music store for string instr.) to the rotating surfaces. No slip! (for a long while)
 
else remove and remount the motor to allow fitting a tiny pulley to the  motor shaft and add a neoprene belt and modify the 1cm rotating driven element to create a grooved guide path for the belt. That's the way I build the ones I use in my lab.

[andrew_c]

Thank you guys for all the great replies!!

The pumps are mechanically perfect for our application, they may only be turned on for a few minutes a day. Accuracy is not too important either for this application. We've gone through over 30 pumps in our testing - not a single one DoA. We've had pumps running for a week or more in our early testing. I grabbed one of these this afternoon to see if it's brushes had bedded-in any - not to my naked eye.

I've tried capacitors from either pole to the chassis, this made the issue worse. I have no method to "earth" anything - we're using a double insulated AC power adapter. I would only be sending noise back to the power supply.

I've ordered a whole bunch of ferrites - many values/sizes to test. I'm only concerned that the motor attached to this pump is too-far-gone already. I could see visible sparks between the brushes earlier today. I'm hopefully this is going to be my saviour.

Is there such a thing as too many ferrites on the wires?

[martys]

I've tried capacitors from either pole to the chassis,

This wouldn't work. You need capacitors on both terminals to the metal case. If you don't have a chassis ground connection nearby use two-conductor shielded cable to connect to the motor after attaching two series inductors and ground the shield to the motor case and the other end to your metal case(if available)  or else to the common ground connection on your driver PCB.

You should only need two solenoid wound ferrite core inductors connected to the two terminals or else one small lossey-ferrite donut core for a low power motor like this.

[andrew_c]

That's what I meant, from both poles to the motor casing, with a third across the poles. I'll post back with results next week when I get another shot at the lab equipment.

[Leon23]

Thank you guys for all the great replies!!

The pumps are mechanically perfect for our application, they may only be turned on for a few minutes a day. Accuracy is not too important either for this application. We've gone through over 30 pumps in our testing - not a single one DoA. We've had pumps running for a week or more in our early testing. I grabbed one of these this afternoon to see if it's brushes had bedded-in any - not to my naked eye.

I've tried capacitors from either pole to the chassis, this made the issue worse. I have no method to "earth" anything - we're using a double insulated AC power adapter. I would only be sending noise back to the power supply.

I've ordered a whole bunch of ferrites - many values/sizes to test. I'm only concerned that the motor attached to this pump is too-far-gone already. I could see visible sparks between the brushes earlier today. I'm hopefully this is going to be my saviour.

Is there such a thing as too many ferrites on the wires?

Increased effect by using ferrites on the cable can simply be to turn the cable(s) multiple times around the ferrite bead.

[ace1903]

I once had problem with power supply unit. Motor was driving stabilizing loop in the power unit outside projected working range.
PU was amplifying noise from the motor. With resistive load PU worked great but with nonlinear load like motor was totally outside declared parameters.
Tried with SLA battery as source of power and then compared results from other switching type supplies against it.
Switching to other wall wart adapter type solved the issue. 

[andrew_c]

Thanks for the all the help guys; some progress was made yesterday! I took a mixed bag of ferrites with me to the test facility and this is the results:



Like the ferrites, a mixed bag of results - some good, some bad. Another half day of pre-compliance testing should hammer things home.

The best result we got used a large toroidal with a smaller ferrite core in series with one another. It's not pretty but it works!

[LabSpokane]

The solution is to use an X2Y capacitor on a PCB between the terminals.  I've already designed the PCB and used many dozens on the very same pump with excellent results.  If you search for X2Y, you should find my thread with my scope results.  The key is to use copper foil tape to seal up the holes in the motor and also connect the outer frame of the motor to the outer ground ring on the PCB. 

The capacitor spec is:

500X44W404MV4E JOHANSON DIELECTRICS  Cap Ceramic 0.4uF 50V X7R 20% Pad SMD 1410 X2Y

This is a not a cheap solution, but it is a good solution.  It works extremely well.

Link to my scope output:  https://www.eevblog.com/forum/projects/brushed-motor-emi-reduction-with-x2y-caps/msg621138/#msg621138

I don't have a SA yet, but I think this should be extremely indicative of the results you can expect. 

[Ghydda]

The solution is to use an X2Y capacitor on a PCB between the terminals.  I've already designed the PCB and used many dozens on the very same pump with excellent results.

This is a not a cheap solution, but it is a good solution.  It works extremely well.

I second that X2Y caps solution. It is very very effective. Applied correctly and the need for any ferrites is usually nil.

Sendt fra min Nexus 6 med Tapatalk

[T3sl4co1l]

The X2Y caps are an amusing gimmick.  None of the appnotes they (Johansen) give are particularly convincing, and certainly, they don't show any advantage in terms of filtering performance per dollar.   The layout and filtering concepts are still just as valid, though, whether you're using regular chips or X2Ys.  Indeed, single chips are probably better for the above board: they can be placed at each terminal, no need for stub traces.

Tim

[LabSpokane]

The X2Y caps are an amusing gimmick.  None of the appnotes they (Johansen) give are particularly convincing, and certainly, they don't show any advantage in terms of filtering performance per dollar.   The layout and filtering concepts are still just as valid, though, whether you're using regular chips or X2Ys.  Indeed, single chips are probably better for the above board: they can be placed at each terminal, no need for stub traces.

Tim

The results I get were extremely convincing. But it won't take anything to change the PCB and test the 2X MLCC solution. I doubt it will be as successful because the caps will not be as effective in cancelling broadband noise as the X2Y seems to be. I could only see the non-x2y solution being as good if used multiple values of capacitors per terminal.

But for a few dollars and a little time, I'm willing to settle the debate experimentally. 

What I do know is that ferrites are just a loser in this application.

[T3sl4co1l]

The X2Y caps are an amusing gimmick.  None of the appnotes they (Johansen) give are particularly convincing, and certainly, they don't show any advantage in terms of filtering performance per dollar.   The layout and filtering concepts are still just as valid, though, whether you're using regular chips or X2Ys.  Indeed, single chips are probably better for the above board: they can be placed at each terminal, no need for stub traces.

Tim

The results I get were extremely convincing. But it won't take anything to change the PCB and test the 2X MLCC solution. I doubt it will be as successful because the caps will not be as effective in cancelling broadband noise as the X2Y seems to be. I could only see the non-x2y solution being as good if used multiple values of capacitors per terminal.

But for a few dollars and a little time, I'm willing to settle the debate experimentally. 

What I do know is that ferrites are just a loser in this application.

If you have a speccy now, I'll donate to that cause 

Tim

[floobydust]

Johanson X2Y has smallest inductance possible and many graphs claiming better EMI performance.

I was leerie of using them due to sole-source (patent), unique PCB footprint and 5-10X the price of vanilla X7R's.

Appears there was massive patent litigation with the ITC and Intel, HP, Apple over these X2Y caps: http://www.itcblog.com/images/X2Y-Attenuators-v-ITC-and-Intel-Corp-CAFC-Decision-7Jul14.pdf

For EMI with brushed motors, I use three capacitors (one across motor and two to motor case; like an X and two Y caps) and two ferrite beads.

[T3sl4co1l]

Johanson X2Y has smallest inductance possible and many graphs claiming better EMI performance.

I was leerie of using them due to sole-source (patent), unique PCB footprint and 5-10X the price of vanilla X7R's.

That's the trick -- the inductance is fundamentally limited by the length between pads. If you assume the ceramic lump is an RF short circuit (which is true for all reasonable chips), length and width are the only remaining degrees of freedom.  If you use two pairs of 0603s instead of a single 1206 X2Y, you get the same overall dimensions, and the same performance.

But you're still way ahead on parts cost, plus you have the freedom of placing those four capacitors anywhere.
 You can use the trace length inbetween as additional filtering, or the FB (or preferably inductor*) between the caps for really high attenuation (3rd order), or you can position the caps straddling the trace for even lower stray inductance!

*Ferrite beads saturate at low currents: typically around 1/5th the rated current (which is a thermal limit only!).  In a power filtering application, there's real advantage to using "inductors".  What's the difference?  A "ferrite bead" is intended for AC filtering, an "inductor" is intended for DC filtering.  And yes -- you can get multilayer chips, that look identical, and have similar inductance (at low frequency), but one stores energy and the other saturates.  It's all about materials.

"Wideband chokes" are also quite good, usually the axial ferrite rod kind, or the multi-hole ferrite kind.  The latter is ungapped ferrite, but the multi-hole design makes them degrade more gracefully under DC bias, which is nice.  Rather large, and maybe not cheap, but nice to have on hand anyway.

In summary, you can always do worse -- of course, you can always do worse; but X2Y is limited by being one single part, whereas you have the freedom to do better when you have individual parts.

On a related subject: feedthrough capacitors.  If you take an X2Y capacitor and short the end caps together internally, you get this.  I don't know that they're any cheaper or more available, but the possible attenuation is even higher than the "two caps straddling a trace" example.

There are also combo bypass-and-ferrite-bead parts, usually in a "feedthrough" style design, some of which are usefully cheap.  Ferrite beads being what they are, these aren't the best possible solution -- but they can still take quite the edge off, even under bias.  Here's a good selection:
https://www.digikey.com/short/3vcm15

Tim

[LabSpokane]

OK Tim,

This work for you?  I'm using a 0.4uF X2Y.  The nearest practical MLCC equivalent would be a pair of 1uF or .68uF on each lead.  Do you agree with that?

I do need to say that for *my* needs, I actually prefer the "magic beans" solution.  My production volume will be measured in dozens, and the ease of the solution, which got me 85%+ of what's probably possible with these cheap motors.  If I was doing this for any real production volume, I'd pick a different motor. 

[T3sl4co1l]

OK Tim,

This work for you?  I'm using a 0.4uF X2Y.  The nearest practical MLCC equivalent would be a pair of 1uF or .68uF on each lead.  Do you agree with that?

Shouldn't matter as long as the reactance is small enough at RF, but to be fair, that should be fine.

I do need to say that for *my* needs, I actually prefer the "magic beans" solution.  My production volume will be measured in dozens, and the ease of the solution, which got me 85%+ of what's probably possible with these cheap motors.  If I was doing this for any real production volume, I'd pick a different motor.

Agreed

Tim