EMI reduction with X2Y caps (Brushed DC Motor)

[LabSpokane]

I need to do some EMI reduction on some 12VDC brushed motors for design integrity as much as EMC compliance.

Does anyone have experience with X2Y caps in this application?  I will obviously be building a PCB that solders to the motor terminals.

The manufacturer's charts look so good compared to a multi component solution, I have to wonder what liberties they took or if the X2Y is just that much better if a mousetrap.

[LabSpokane]

Buhler...?

This is what I'm looking at:
http://www.johansondielectrics.com/images/stories/surface-mount/x2y/JDI_X2Y_2013-04.pdf

I hate the price, but I really like the one component solution.

Anyone have any experience to share with EMI reduction / EMC with brushed electric motors?  I'm looking at the data sheet below and it all seems quite straightforward, but I'd really appreciate any feedback before I send out PCBs for fab.

http://www.x2y.com/appnotes/dcmotors/4001.pdf

Thanks!

[T3sl4co1l]

Actually looked at those before... not impressive at all.  If they had made them with inverse construction (ground side connections, end-to-end passthrough connection), you could use them like feed-thru caps.  I mean, yeah, you can ground the two ends and use the sides that way, but that doesn't appear to be how they measured it.  It's not obvious that it's any better than a pair of equal caps, and certainly not better than two caps and a ferrite bead (now, one cap and one bead, harder to say, but probably also not).

The "high speed" test case is actually rather amusing, because if you look at the (poor resolution) photos of their test board, it's clear that there are many footprints around.  They could've used any technology (from 0201 to 1206, single or multi), and with enough futzing with placement, get even better performance.  Not to mention the invisible, and far more important, internal planes in the PCB!

For your motor, I'd say, throw down a few ceramic caps, followed by power inductors on the motor side of the wires.  Keep the leads up to the motor close together, and mind that both leads will generate RF relative to the enclosure.

Tim

[LabSpokane]

Thanks, Tim. And I just read that this guy agrees with you:
http://speedingedge.com/PDF-Files/X2Y%20vs%200402%20081605.pdf

I really wanted a miracle produc, though. What a letdown. 

[ConKbot]

Thanks, Tim. And I just read that this guy agrees with you:
http://speedingedge.com/PDF-Files/X2Y%20vs%200402%20081605.pdf

I really wanted a miracle produc, though. What a letdown. 

Hmm, about what I always suspected. It always seemed fishy when even in johansens documentation that they compared an x2y cap with 6 vias to a cap with 2, when it would be quite easy to stitch a capacitor to the planes with 4 vias if you were so concerned about inductance. 

[T3sl4co1l]

There's no substitute for good old fashioned design.

[Richard Head]

I would have expected the vast majority of brush noise to be below a Mhz or two, if that. I find it difficult to believe that it extends up to 40Mhz.
Have you actually measured the high frequency noise?

[Siwastaja]

I would have expected the vast majority of brush noise to be below a Mhz or two, if that. I find it difficult to believe that it extends up to 40Mhz.
Have you actually measured the high frequency noise?

Brushes by definition are arcing radio transmitters with unlimited bandwidth. It extends up to UV light and beyond, or hundreds of THz.

With brushed motors, you do what you can to filter the noise, and accept the rest. Very much depends on the actual brush design. Some motors are surprisingly "quiet" and some cheap-ass ones cannot be fixed to acceptable EMI with any filtration.

[Richard Head]

Have you measured the noise spectrum?

[itdontgo]

Their product page shows an odd example

For filtering GSM TDMA noise they use:

1    No input filter, 2 discrete MLC 100nF power bypass caps.
2    2 discrete MLC 1nF input filter, 2 discrete MLC 100nF power bypass caps.
3    A single X2Y 1nF input filter, a single X2Y 100nF power bypass cap.

1nF/100nF caps won't do sod all at 900MHz.  You typically need 80-90pF caps @900MHz and about 33pF at 1800MHz.  Strange comparison to make that.  I was going to try them once but normal RF caps are fine.

[LabSpokane]

Have you measured the noise spectrum?

I don't have a spectrum analyzer (yet), but the brush noise results in approximately 1 MHz of ringing plus some other random crap. The problem isn't the frequency as much the amplitude. I'm getting up to 1V peak of ringing in my 5V bus from the EMI. thus far it is an issue, not a problem, but it's more noise than I want.

[SeanB]

You get small motors with built in EMI filtering on the commutator, which does a good job of reducing RFI. Most common method I see is to use 2 disc ceramic capacitors soldered to the terminals and the common soldered to the case. Then a ferrite bead on each supply line right by the motor is used to tame the remaining RFI to low levels. This works well, so if you do a PCB you should use ceramic caps ( pretty much any value from 100n to 1uF with a voltage higher than the motor by a good margin) and some SMD ferrite chokes, and make a pi filter as well for the input stage.

[sweir]

Buhler...?

This is what I'm looking at:
http://www.johansondielectrics.com/images/stories/surface-mount/x2y/JDI_X2Y_2013-04.pdf

I hate the price, but I really like the one component solution.

Anyone have any experience to share with EMI reduction / EMC with brushed electric motors?  I'm looking at the data sheet below and it all seems quite straightforward, but I'd really appreciate any feedback before I send out PCBs for fab.

http://www.x2y.com/appnotes/dcmotors/4001.pdf

Thanks!

Let me introduce myself as one who has long consulted to X2Y on applications.  The basic thrill of X2Y comes in two attributes:  Low mounted inductance, and very tight symmetry.  In a number of applications these attributes have proven to be extremely effective versus any competing approach.  X2Y has done spectacularly well suppressing wideband EMI particularly from things like brush motors where they are extensively employed by the automotive industry.  If you make a low inductance connection between the G1/G2 pads and the motor case, and then use one of the A or B pads for the motor red lead, and the other for the motor black lead, then what you end up with is a very tight and matched pair of filters for each the red lead and the black lead to the motor case.  Where that is going to outperform building discrete low pass filters is in mode conversion particularly near the filter cut-off frequency.

[T3sl4co1l]

1nF/100nF caps won't do sod all at 900MHz.  You typically need 80-90pF caps @900MHz and about 33pF at 1800MHz.

Howso?  The impedances are essentially identical, except by lucky coincidence if you tune for series resonance (notch filter).  But that's a very special case and will only work for a narrow band, not a true wideband lowpass.

Tim

[sweir]

Thanks, Tim. And I just read that this guy agrees with you:
http://speedingedge.com/PDF-Files/X2Y%20vs%200402%20081605.pdf

I really wanted a miracle produc, though. What a letdown. 

Well, you need to be careful about that report.  It was well intended, but suffers serious deficiencies that lead to incorrect results.  The fixture design was done in such a way that it obscured the performance of all of the capacitors tested.  Basically what the fixture measured was mostly the five parallel power ground cavities in the board.  That was because the spectrum analyzer connections were both close to each other in the upper left hand corner of the PCB, and parallel connections of the power ground cavities.  If the idea is that there would only be a single IC on the board located at one of the spectrum analyzer ports in that corner of the board, and it would be powered by a single voltage distributed through five cavities bypassed by capacitors across the board, then the results could be considered fair.  But that is not how any rational board is constructed, and so the fixture for all its good intentions failed to test the performance of the various bypass capacitor types as was its good intentions. 

If  one models the circuit, what you end up with is the discrete capacitors of all types tested at the far end of the spreading inductance and vertical attachment inductance operating in parallel with those big planes.  Consequently, even IDC capacitors which Intel used to rely on very heavily did not show very well.  By contrast, a number of years ago X2Y at Altera's request redid the bypass networks on one of their boards with a 3 and 6 Gbps SERDES.  We did not alter the board stack-up.  What we did was to reduce the capacitor count by over 30%, freeing up space to put the capacitors where we wanted, and detune the resonances.  The end result was much better jitter performance in their SERDES, around 34% reduction.

Above a MHz or two, the way that an IC sees the power distribution network is through a very inductive looking interconnect out to the closest bypass capacitors.  If the capacitors are far enough away, or the planes are far apart, then the interconnects are what the IC sees and you could have perfect capacitors effectively shorting the planes together and they would be completely limited by the inductance of those interconnects.  What that means is that X2Y caps are not a panacea.  They are a very effective tool for appropriate circumstances.  For EMI filtering they are stellar.  If you want to tame EMI they can be extremely effective.  If you want to use a minimum number of capacitors to reach a particular inductance target across a power cavity (planes) they can be very effective.  If you want to detune a resonance in the power distribution system, their very low attached inductance is hard to beat, even if you use conventional MLCC packages everywhere else.

Steve

[sweir]

1nF/100nF caps won't do sod all at 900MHz.  You typically need 80-90pF caps @900MHz and about 33pF at 1800MHz.

Howso?  The impedances are essentially identical, except by lucky coincidence if you tune for series resonance (notch filter).  But that's a very special case and will only work for a narrow band, not a true wideband lowpass.

Tim

Exactly.  The capacitor performance reflecting RF away from the IC power is dictated by the mounted inductance which with a decent layout is really, really good with X2Y's.  The capacitance at those frequencies could be any value from dozens of pF on up. 

Of course X2Y would love it if only their capacitors were used everywhere a capacitor can go.  In the real world, they are incomparable in many applications, competitive in other situations, and a square peg for a round hole in others.  While you can reduced the attached inductance of 0402s by drilling more via holes, it does not scale the same way as X2Y's because of the way the vias couple to each other due to the mechanical pattern.  On the one hand that gives X2Y's much better via utilization, and on the other it can complicate routing.  So like most things in signal and power integrity whether X2Y's make sense for things like power bypassing depends on the circumstances. 

I recommend X2Y capacitors to customers where they make sense, and offer other solutions where those solutions are more appropriate.  If there is any generic advice I can offer it is always be mindful of keeping stray inductance down.  It costs real money to compensate for the effects of stray inductance, so be very aware of it, and you will be in the best position to get the performance you need no matter what parts you choose.

[LabSpokane]

Here's some waveforms for what they are worth.  The first is the raw EMI pulse coming off the motor.  That is the sum of a whole bunch of garbage, not a simple fundamental frequency.  It's comprised as of both brush noise and the pulse of the field collapse as the motor poles switch.  I can remove a lot of the field collapse noise with a simple 1N400X across the terminals.  The second image is what I have the EMI beat down to with the ceramic caps I have in stock. 

Since I have to build a PCB anyway, I'm going to order some X2Ys off Digikey and build up a board per the app notes just for giggles. 

I really appreciate the discussion that's going on. Thanks all!

[sweir]

As you can see from the spike and the ringing, the 1N400x diode is pretty slow for that more or less 100ns rise time.  If you spend a couple pennies on an ultrafast diode like a UF4002 instead of a straight 1N4002 you should see considerable improvement.

Steve

[LabSpokane]

I followed the X2Y guidelines fairly closely and had a PCB built to place the X2Y cap between the motor terminals. I covered all the openings in the exterior of the motor with copper foil.  I put a ferrite on the leads off the motor just for good measure. The results are impressive at least to me.

I'd like to thank Steve, OSH Park, and lamb vindaloo for their assistance.

[RayeR]

Hi,
thanks for this thread. We also have a trouble with one DC motor. It's about 4x6cm size, 24V, 0,1A at free run, 4-5A blocking current. We used DRV8872 from TI on our control board to drive the motor. We did some measurements in EMC lab, and failed to pass at all. We also tested DC motor alone without control board, just fed from a DC PSU. It was emitting wide spectrum noise with high peaks of hundreds of MHz, even high peak at 1GHz. I was trying to lower the EMI with some ferrite ring, motor lead twisting, metal shiled cover on back of the motor (covering leads) and simple LC filter with 2 WE7427511 in series and 22nF block but all this attempts failed to pass, it lowers EMI a bit but still left some peaks over the limit. I didn't know about X2Y parts so I hope it would help.
But I see one serious disadvantage of this kind of pars. As I see in some datasheets this components have a large capacitance of hundreds nF - this will deny use of PWM control. The DRV8872 use PWM mode for smooth spin-up and for current limiting if motor is mechanically blocked. If the driver see large capacitive load during PWM it will trigger OCP and assert a fault pin. So a different driver would be needed. But the most important is to denoise the motor (our customer don't want to change motor) and then change the driver circuit...
We will order some X2Y components, make a PCB and further measurement. I'll let you know how it helps.

[wraper]

As I see in some datasheets this components have a large capacitance of hundreds nF - this will deny use of PWM control.

You can get them in tens pF to uF range.
http://uk.farnell.com/webapp/wcs/stores/servlet/Search?catalogId=15001&langId=371&storeId=10166&categoryId=700000005423&st=x2y&beginIndex=1&showResults=true&aa=true

[T3sl4co1l]

Ferrite beads may help more inside the motor, too.

Tim

[LabSpokane]

I'd also make get a proto PCB from somewhere like OSH Park and put the X2Y right between the motor terminals.  I put a ground ring around the perimeter of the PCB and copper foil taped that to the case of the motor. 

[wraper]

[LabSpokane]

Wraper,

No one's trying to debate here. I had a problem, I tried a potential solution, and posted my before and after results. If you think my result is incorrect, fine, but post facts, not cartoons.

I had an excellent result and will keep using the X2Y solution for my brushed motors despite the stupid high cost.