• AVR Freaks

Hot!EMI problems using MCP2515

Page: 12 > Showing page 1 of 2
Author
atestani
Starting Member
  • Total Posts : 12
  • Reward points : 0
  • Joined: 2003/11/07 12:49:52
  • Status: offline
2019/10/13 17:58:54 (permalink)
0

EMI problems using MCP2515

I am using two MCP2515 in a product that is failing EMC testing.  The devices are using a 16MHz crystal and there are two strong spurs at 154MHz and 160MHz. The test lab is seeing these problems as well as my own testing with both h-probes and e-probes with a spectrum analyzer.  Using an RF probe on the crystal OSC1 and OSC2 show these harmonics are very low there so it seems the problem is internally generated within the MCP2515.
 
I am at a loss as to what to do to fix this problem so any help will be greatly appreciated!
 
Thanks
 
Al
 
#1

31 Replies Related Threads

    nigelwright7557
    Super Member
    • Total Posts : 308
    • Reward points : 0
    • Joined: 2006/11/06 08:15:51
    • Location: 0
    • Status: offline
    Re: EMI problems using MCP2515 2019/10/13 21:21:25 (permalink)
    0
    You could screen using grounded copper pours on the pcb.
    After that the enclosure needs screening.
     
     
    #2
    GMAprogrammer
    New Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2019/11/10 10:01:39
    • Location: 0
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/12 04:17:25 (permalink)
    +1 (1)
    Hey Al,

    got the same problem here, during precompliance the PCB failed due to a spike arround 64Mhz, which is a harmonic of the 16Mhz Clock the MCP2515 was running on. (In my case supplied by the ATmega32u4 clock out pin)

    What package type do you use in your design? I am using an SOIC-18 Package. I condensed the radiation down to the package itself. By shielding the package Top with copper foil i was able to reduce the radiation down by 10dBuV which would be enough to stay below the limit. Unfortunately for the final design this is not an option and my thinking is, that there have to be many devices out ther using the MCP2515 without shielding. So why does this particular one radiate?

    Also a test with sacrificed MCP2515 from a CAN Bus Shield revealed that the harmonic was no longer present. There where some other spikes poping up but the 64MHz harmonic was barely noticeable.
    All that brings me to beliefe that the SOIC-18 is a Noisy package design and that there are some variations in production that contribute to the problem of different harmonics being radiated out.

    An approach of fixing the issue would be to change the package to a smaller type to reduce the current loop inside the IC.

    It's all trial and error, i know but i tested many things by now and the only thing that reduced the noise level by a good amount was the shielding the package itself.

    -Alex
    #3
    atestani
    Starting Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2003/11/07 12:49:52
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/13 09:02:11 (permalink)
    +1 (1)
    Nice work, Alex.
     
    I am using the TSSOP pkg so it isn't limited to the SOIC package. I wonder if the QFN package has the same problem?  
     
    I "fixed" the problem by clocking the chip at 4MHz vs 16MHz  and initializing it that way.  The result of doing that was the harmonics were reduced but, most important, were shifted to a section of the limit line that is higher.  I don't know if the test lab we are using will now pass this but it was a solid solution in my test setup.
     
    Microchip needs to fix this problem!  Do they monitor this forum but, if not, how to we report this?
    #4
    ric
    Super Member
    • Total Posts : 24333
    • Reward points : 0
    • Joined: 2003/11/07 12:41:26
    • Location: Australia, Melbourne
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/13 12:53:39 (permalink)
    +1 (1)
    atestani
    ...
    Microchip needs to fix this problem!  Do they monitor this forum but, if not, how to we report this?

    No, they do not officially monitor this forum, although some employees occasionally browse in their own time.
    The official way to report a problem is via a "Support Ticket".
    http://microchip.com/support
     

    I also post at: PicForum
    Links to useful PIC information: http://picforum.ric323.co...opic.php?f=59&t=15
    NEW USERS: Posting images, links and code - workaround for restrictions.
    To get a useful answer, always state which PIC you are using!
    #5
    GMAprogrammer
    New Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2019/11/10 10:01:39
    • Location: 0
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/13 14:19:02 (permalink)
    +1 (1)
    BTW:

    I made a.test with only the MCP2515, a 16Mhz Quarz and nothing more. No SPI communication, etc. Guess what is visible? Right a harmonic at 64MHz with exactly the same Amplitude. For.me this is the final.prove.

    I also tested with a 8Mhz quarz and the noise performance is way better. So.i guess i will go.with the same solution as you.

    -Alex
    #6
    Nikolay_Po
    Super Member
    • Total Posts : 1913
    • Reward points : 0
    • Joined: 2012/04/01 13:49:27
    • Location: Russia, Novorossiysk
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/13 15:17:12 (permalink)
    +1 (1)
    Alex, can you test an EMI with 1..10 Ohm SMD resistor in series with Vdd pin, a resistor between the pin and any decoupling capacitor?
    Also I see that in contrast to many PIC microcontrollers, Vdd and Vss pins of MCP2515 package are old-styled, are placed at a package diagonal. The PCB design may be crucial for this case. Can you show a fragment of your PCB design with MCP2515 Vss and Vdd traces/polygons?
    #7
    atestani
    Starting Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2003/11/07 12:49:52
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/13 15:31:38 (permalink)
    +1 (1)
    I have tried all sorts of things as you suggest and nothing helps. I can't speak for Alex, but I'll bet he has done the same.
     
    This is a purely radiated issue and caused by a strong field the chip itself is generating.  Strong enough at 64MHz when operating at a 16MHz clock rate that it exceed limits and fails EMC testing.  
    #8
    Nikolay_Po
    Super Member
    • Total Posts : 1913
    • Reward points : 0
    • Joined: 2012/04/01 13:49:27
    • Location: Russia, Novorossiysk
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/13 16:21:56 (permalink)
    +1 (1)
    I can't believe that the small piece of silicon itself can effectively radiate at 64MHz. Just because of size. The interference is radiated by the leads and, most probably by Vdd/Vss pins but may be by crystal pins the same. The EMI is an evidence of spiky switching in circuit. If you will cut HF supply current the spikes will be mitigated because of lack of power.
    I'd try a ferrite bead in series with Vdd pin, may be Murata BLM15BB470SH1 or similar. It has low (Ohms) impedance at 8MHz (internal logic frequency) and tens of Ohms at 64MHz. 
    #9
    atestani
    Starting Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2003/11/07 12:49:52
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/13 19:33:21 (permalink)
    +1 (1)
    I would certainly agree with you if I hadn't worked on the things you are suggesting for several days and they all failed.  I tried ferrite beads, discrete hi and low Q LC and CLC filters on the power leads, tried to reduce the slew rate of the output (which is differentially driving a 120ohm load), and many other things.   The *only* thing that worked was changing the clock from 16MHz to 4MHz.  
     
    Another piece of data is that I used an H-field probe and spectrum analyzer to measure the field before and after of everything I tried and it was the strongest signal when placed of the center of the package and dropped off steeply as the probe was moved towards the leads.  However, none of the things I tried changed the field strength.
     
    Alex, you are also having these issues, what is your experience in trying to fix it?
    #10
    GMAprogrammer
    New Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2019/11/10 10:01:39
    • Location: 0
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/14 11:09:28 (permalink)
    0
    Nice to see that we are getting a constructive discussion going. Today i took the time to do some further testing but first i want to make something clear: Allthough i strongly believe that the MCP2515 itself is the problem it might be the case that i am wrong and it is a design fault on my side. I hope this Post will help to understand whats going on and give some more fundamental prove of what i am experiencing with the MCP2515. OK now that's out of the way let's start with my test results:

    1. Test Setup:
    As i am in my office right now and don't have the chance to occupy the emc test chamber all day i am testing with a TEM Cell and spectrum analyzer. The actual Amplitude measured is not really comparable with a 3m far field measurement but this setup gives me the easiest option of testing several interations quick. Generally everything over 30dBuV is something to be concerned about and every Spur passing 40dBuV will most likely fail in a farfield measurement.
    The Setup consist of a TBTC3 Tem Cell and a SSA3021X with EMI Option. All Measurements were taken with the following settings:
    Start Frequency: 10MHz
    Stop Frequency: 130MHz
    Filter Option: EMI
    Detector Type: Quasi Peak
    Attenduator: 0dB

    For Reference heres a Sweep without anything to measure. You can clearly see there is some backround noise in the sub 30MHz and arround 95Mhz. Sub 30Mhz is mainly from surrounding office devices and 95MHz noise is from Radio stations that couple into the USB Cable that supplies the DUT.

     
     
    Forum does not let me post everything at once so i need to split it up:


    -Alex
    #11
    GMAprogrammer
    New Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2019/11/10 10:01:39
    • Location: 0
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/14 11:24:13 (permalink)
    0
    1. Original PCB Layout Test:
    Ok, at first i establish a baseline by testing my most recent design which failed in the CISPR32 3m precompliance test. As Nikolay asked for a picture of part of my layout here is the part where the MCP2515 is located including the ATmega32u4 that supplies the 16MHz via it's CLKO Pin.
     
    Edit: The Forum won't let me upload the pictures in decent resolution. So here are the links:
    Top: https://ibb.co/6nZQngn
    Bottom: https://ibb.co/GxGjQSk


    As you see it's just a 2-Layer PCB. I tried to make the ground pour on the bottom as solid as possible. This will be further improved in the next iteration because it is basicly a good thing no matter what.
    This is what i see when taking a measurement in the TEM Cell:

    As you see the Spur at 64MHz is clearly visible and passing the 30dBuV by a good amount. Several other Spurs at 32Mhz and 128MHz are below 27dBuV and don't considered to be a problem. So in the following i will focus on the 64MHz frequency.

    -Alex

    Attached Image(s)

    #12
    Antipodean
    Super Member
    • Total Posts : 1768
    • Reward points : 0
    • Joined: 2008/12/09 10:19:08
    • Location: Didcot, United Kingdom
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/14 11:29:37 (permalink)
    +1 (1)
    GMAprogrammer
     
    As you see it's just a 2-Layer PCB. 

     
    I suspect that is your problem. I wouldn't be using those sort of frequencies on a two layer board and expect to be passing an EMC test. Four layer minimum, and things like 16 MHz clocks probably need to be in an inner layer between two ground planes.
     

    Do not use my alias in your message body when replying, your message will disappear ...

    Alan
    #13
    GMAprogrammer
    New Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2019/11/10 10:01:39
    • Location: 0
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/14 11:36:40 (permalink)
    0
     2. Shielded Package
    Ok, to check if the actual package is causing the radiation i shielded the top of the package with copper foil and connected it directly to ground and via a 100nF Cap to +5V. Note that i had desoldered the MCP2515 before (That's why it is quite messy arround it) to add some copper foil beneath it as a hacked top groundplane but that had no effect. So basicly now the package is surrounded by copper.


    Here is a picture: https://ibb.co/cFTvMs7
    This is what i got:
     
     
    As you can see the Spur at 64MHz is now about 10dBs lower than before.

    -Alex
    post edited by GMAprogrammer - 2019/11/14 11:37:56

    Attached Image(s)

    #14
    GMAprogrammer
    New Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2019/11/10 10:01:39
    • Location: 0
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/14 11:49:40 (permalink)
    +1 (1)
     3. Nikolays Suggestion:

    Ok so next up was testing Nikolays suggestions in adding a sub 10Ohms resistor in series with the suplly voltage of the IC and a decoupling cap which basicly performs like a lowpass filter and should eliminate high frequencies coming over the powerlines. I just had an 11 Ohms resistor but that was close enough for me. That's the result:
     
     
     
    Nothing changed. Still the same emissions as without the Resistor.

    -Alex

    Attached Image(s)

    #15
    atestani
    Starting Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2003/11/07 12:49:52
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/14 12:13:12 (permalink)
    +1 (1)
    I totally agree with the need to have a 4+ layer board BUT I do and still have this problem.
    #16
    GMAprogrammer
    New Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2019/11/10 10:01:39
    • Location: 0
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/14 12:56:54 (permalink)
    +1 (1)
    Ok, actually i agree with both of you. a 4-Layer Board might seem the convenient solution for everything these days. But it doubles the pcb costs by a factor of two. And i will definitivly consider it once i got figured out what the actual problem is. But i have a feeling i would experience similar problems with a 4 Layer design.

    Ok back to topic:
    After i read Alans post i though ok how can we test the system without the MCP2515 being there. I don't know if the following is 100% correct but the CLKO Pin of the ATmega32u4 sees the MCP's OSC1 Pin as a series of resistor, inductor and capacitor to ground. All that resulting in an input impedance of the ICs pin.

    So i desoldered the MCP2515 (again) and terminated the 16MHz clockline with a two 100kOhm resistors in parallel(If you have suggestions for another value let me know). I determined the value by checking the datasheet and finding that the typical input current of the osc1 pin i 5uA. Measuring the output voltage of the clock signal and it's is about 300mV RMS.  So 50kOhms should give a current of about 6uA. Close enough.

    I know it is an estimation but thats what i came up with for now. If i should change anything just let me know. I could add some capacitance in series. The datasheet specs 7pF Input capacitance but i don't know if that is also the case for the OSC1 pin.

    So my plan was to see what the actual trace and return loop for the Clock signal contributes to the emissions.
    And this is the result:



     
    There is a Peak visible at 64MHz but it's way below what i get when the MCP2515 is present.

    Let me know what you think.

    -Alex

    Attached Image(s)

    #17
    atestani
    Starting Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2003/11/07 12:49:52
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/14 15:44:02 (permalink)
    +2 (2)
    Alex,  You WILL see the same problem with a 4 layer board.  I have a 4 layer board and have the problem. 
     
    My layer stack up is:
    top
    inner ground
    inner 3.3V
    bottom
     
    Bypassing is with a 100nF capacitor next to the power pin and both sides of the cap drop to their respective planes with vias. 
     
    I tried all sorts of bypass and decoupling techniques. I believe I took the power and ground to the MCP2515 off the PCB and wired then discretely so I could "try things" including putting a current probe on the lead from the power supply to the bypass cap and then on the lead from the bypass cap to the chip power pin and didn't see anything abnormal. Bottom line nothing worked except the change in clock rate from 16MHz to 4MHz.  
     
    I can upload images of my schematic and layout but it really isn't any different than yours except I use planes for voltage and grounds and you are using traces. 
     
    I am convinced layout, bypassing and decoupling, layer count, and all of that part of the discussion is irrelevant.  We have both proven none of these things fix this problem.  However, I agree it is very strange that it appears to be the package/chip itself broadcasting a 160MHz signal.  I also would say we really don't know what the root cause is. Admittedly, my crystal change from 16MHz to 4MHz was a bandaid because I couldn't find any other way to eliminate the problem and I was out of time.
     
    I don't recall if I did any more than try to slew rate control the output.  I don't have my setup any longer to try this but what happens if the output load is disconnected, i.e. removing the termination resistor(s) on the CAN bus?  Assuming there are two 120 ohm termination resistors on the bus, the MCP2515 has to drive over 50mA.  If the spike changes, then lift the output leads of the MCP2515 and connect a 60 ohm resistor right across the lifted leads thereby eliminating the antenna created by the connection out to the CAN-bus.    
     
    Frankly. I don't think that experiment is going to show much because: 1) the CAN-bus is running at a very slow 250kHz and 2) if it is the fast edges of the output signal, why does the problem go away when I run at 4MHz... the output drivers are switching in exactly the same way and the MCP2515 is demanding the same current from the rail.  
     
     
    Al
    #18
    Nikolay_Po
    Super Member
    • Total Posts : 1913
    • Reward points : 0
    • Joined: 2012/04/01 13:49:27
    • Location: Russia, Novorossiysk
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/14 16:38:26 (permalink)
    +1 (1)
    Hi Al, hi Alex. Thank you for sharing your experience. I newer saw or even heard  that the package's (or chip's) own interference can fail EMC test without of problems at PCB or wiring level!
    Alex, your PCB seems to me completely OK, particularly for 16MHz-clocked chip. Sure if trying to mitigate certain EMI problem at PCB level you'd make small changes, but the version you showed is reasonable, completely normal.
    Guys, can you test wither the leads of problematic chip are ferromagnetic or not? Will the chip "gravitate" to a magnet?
    #19
    atestani
    Starting Member
    • Total Posts : 12
    • Reward points : 0
    • Joined: 2003/11/07 12:49:52
    • Status: offline
    Re: EMI problems using MCP2515 2019/11/14 19:54:18 (permalink)
    0
    Nikolay,
    Thanks for trying to help us on this problem.
     
    I totally agree..... this is certainly weird!  However, I have been doing some research on IC packaging and EMI and there is a lot of evidence that this is not an isolated case but a more or less widespread issue. IN this case, when we are clocking the chip at 16MHz, the problem seems to be at the 10th harmonic!  This is even stranger since it is so high.  Does anyone know what the PLL in the chip is running the internal clock at? 
     
    I was intrigued by your question but didn't have any loose MCP2515's.  I was able to take one off an earlier board and found it was not slightest bit magnetic.  The date code was 1811 which is the same as the boards I have in EMC test.   What exactly are you looking for?
     
    I think it is time we brought Microchip into this discussion so I am going to put in a "ticket" to see what happens.  The MCP2515 is a very popular chip so my guess Alex and I are not the only ones with this problem.  
     
    Al
    #20
    Page: 12 > Showing page 1 of 2
    Jump to:
    © 2019 APG vNext Commercial Version 4.5