Common mode choke vs X2Y capacitor

[Mad ID]

Hello,
this one is for experts

My next design must pass the EMC testing and without common mode filtering/shunting on power/signals which enter the device via a cable it will certainly fail radiated emissions from my experience so far.

Lets take power cable and CAN bus cable for example.
The common wisdom is to put a common mode choke and it will solve the problem. Lately I found that X2Y capacitors from Johanson can be used for common mode filtering.
http://www.johansondielectrics.com/x2y-products.html#.VCcelPmvXwk

It there any catch? They are cheaper and smaller no matter. If I need 100mA or 10A it doesn't matter, they do the job? I used them for a client but that device was not subject to EMC testing and I don't have the equipment in the lab to compare common mode choke vs X2Y performance.

Can I use them on CAN bus also instead of common mode chokes found in most application notes?

Thanks guys.
 

[Andreas]

Can I use them on CAN bus also instead of common mode chokes found in most application notes?

Depends on the bitrate that you need.
usually CAN does not allow more than 5nF in total at high bit rates.
And this is already used up by wiring capacitance 50-100 pF / m and receiver input capacitance 3-5 pF per input.

With best regards

Andreas

[T3sl4co1l]

CAN is more or less differential, what common mode noise are you expecting?

What standards does it need to meet, what is it, what kind of crap is it going to throw off?

Usually, if it's power equipment, emissions are the biggest concern; signal equipment, susceptibility.  Often, both.

Tim

[penfold]

Passing the connection through a ferrite to form a common mode choke is the classic way of doing things.  I wouldn't advise loading a communication line with a capacitance unless you were really in trouble.

[T3sl4co1l]

Loading a com line with capacitance is perfectly acceptable, if it's consistent with the signaling standard.  A low impedance, terminated transmission line medium, like RS422/485, with moderate bandwidths (say 115.2k), can tolerate a lot of capacitance, and probably should -- because you don't want to leave excess bandwidth in your circuit.  CAN I'm not too sure about, but it's probably similar.

Some examples where it doesn't work as well:
- Weak logic outputs (TTL/CMOS, like a parallel port), where the edges are significantly slowed (which causes problems for edge triggered signals), and the voltage levels get shunted by any loading.
- Switched states, typical of RS-485 multimaster, USB and others: the unbalance and varying impedance don't filter well.  In some cases, you can deal with it (like RS-485 usually), but others just aren't any good outside of a well shielded cable.
- High bandwidth signals (USB2.0, LVDS, etc.), where the signal itself just can't tolerate much [lumped] capacitance, which means you don't have anything for the CMC to work against.

BTW, as long as the pin driver is capable of driving a characteristic impedance, the capacitance of the cable is a non sequitur.

Tim

[SeanB]

Data lines you want to avoid adding capacitance to it, it degrades signal edges which are the inportant part of the data. Common mode choke works best there.

Power lines the common mode choke and the capacitors work better, as the choke increases the line impedance so the capacitors form a better low pass filter to reduce noise. You probably will want to use a premade filter though unless you are constrained for cost, as the filter has approval so there is a good guarantee that the specs are going to be usable. Otherwise you need the capacitors at both input and output of the choke, to provide the best attenuation. The ground lead from the filter also needs to be short and direct to the mains earth input, best is to use an IEC inlet with combined filter as this takes the least space, and provides the best attenuation and no internal paths.

[Niklas]

Most of our automotive customers have their own specifications for the CAN bus interface and sometimes also a black list of banned components.

Capacitive load
- Typically less than 100 pF to GND. 18-22 pF to GND is common
- TVS diodes with low capacitance. "Normal" TVS diodes with 30 V breakdown voltage can have more than 1 nF of parasitic capacitance

Termination
- Open, 120 ohm, 2x60 ohm with center cap to GND, 2x1.3 kohm with center cap to GND. The 2x1.3 kohm option can be helpful for stub nodes.

Common mode choke
Some customers demand it and some customers specifies "if needed for compliance"

TVS diodes
Bidirectional. Unidirectional diodes can cause a lot of problems if the GND connection is lost. Breakdown voltage must be larger than the maximum supply voltage in the same wire harness. >=18-20V for passenger cars/trucks with 12 V battery, >= 33-36 V for trucks/buses with 24 V battery

CAN transceiver
Guaranteed silent when powered off. Some of the emissions is related to the unsymmetrical nature of NPN and PNP transistors. If both are activated perfectly synchronized there would be no problems, but some transceivers have a small lag between the drivers of CANH and CANL. Also make sure that you have sufficient bulk capacitance close to the transceiver. Driving a 1 V signal on a 60 ohms (120//120) bus with edge synchronization can draw some current.

[Andreas]

Capacitive load
- Typically less than 100 pF to GND. 18-22 pF to GND is common

But this will only work in a defined environment with definded number of CAN participants and defined wiring lenght.

Loading a com line with capacitance is perfectly acceptable, if it's consistent with the signaling standard.  A low impedance, terminated transmission line medium, like RS422/485, with moderate bandwidths (say 115.2k), can tolerate a lot of capacitance, and probably should -- because you don't want to leave excess bandwidth in your circuit.  CAN I'm not too sure about, but it's probably similar.

You cannot compare CAN (recessive state where only the terminator resistor discharges the capacities) with a RS422/485 line with low impedant push pull drivers.
The later tolerate about 50nF in total for 115K2.

With best regards

Andreas

[Precipice]

But this will only work in a defined environment with definded number of CAN participants and defined wiring lenght.

Which is exactly the world automotive manufacturers live in. It's their bus, in their vehicle, with a list of devices they specify.
They don't really care that it's CAN, but they do care, passionately, that it all works properly together, and doesn't cause product recalls.

(What's particularly annoying (with at least two OEMs) is that the per-device testing is unrepresentative of the configuration in the car, so you end up having to  build quite elaborate load boxes for testing. But that's a different rant, as is the fact that a DC coupled data bus in a car is a poor plan anyway).

[Mad ID]

Depends on the bitrate that you need.
usually CAN does not allow more than 5nF in total at high bit rates.
And this is already used up by wiring capacitance 50-100 pF / m and receiver input capacitance 3-5 pF per input.

My bitrate is 20kbps and maximim cable length is 25 meters so I don't expect any problems by adding some capacitance. Also, the PHY I'm using is integrated in the LPC11C24 microcontroller and does not have slew rate adjustment.

CAN is more or less differential, what common mode noise are you expecting?
What standards does it need to meet, what is it, what kind of crap is it going to throw off?
Usually, if it's power equipment, emissions are the biggest concern; signal equipment, susceptibility.  Often, both.

Differential transceivers are not 100% symmetric, i.e. when the signal transitions the sum of voltages will not be pure DC but will have some high frequency content due to non-ideal transceivers. I read this in a couple of application notes and please correct me if this is wrong? So yea, I except common mode noise from my non-slew-rate limited PHY which can/will fail the test.
I must pass the EU EMC radiated emissions, from 30MHz to 1GHz (EN 55022 I believe). 

[Niklas]

Capacitive load
- Typically less than 100 pF to GND. 18-22 pF to GND is common

But this will only work in a defined environment with definded number of CAN participants and defined wiring lenght.

Usually the capacitance to GND is specified per node. As an OEM you have no idea of where your product will be placed on the bus, that is up to the customer to decide depending on their specific needs. Point-to-point on separate bus, stub node, end node, long bus, short bus... You have no clue and the demands can change of the life time of the product.
In our case, we have designed the PCB to have 4 connections for CANH, CANH Termination, CANL and CANL Termination and also with bridging options with 0 ohm resistors.
Want termination but with only a 4 pin connector for supply and CAN? Bridge on the PCB
Want selectable termination? Use more than 4 pins in the wire harness and bridge in the truck's harness

Just to make life easier for the OEM, some customers have the same connectors, but with different pinouts depending on the market (US, EU, Asia)

[Precipice]

My bitrate is 20kbps and maximim cable length is 25 meters so I don't expect any problems by adding some capacitance.

Is the CAN bus all yours, or do you have to interoperate? I'd be very, very careful about casually adding capacitance.

I think I'd suggest, in the spirit of overengineering with a view to deleting stuff if you pass too easily:

For the CAN bus
Fit the common mode choke, possibly with an option to no-fit it and bypass with zero ohms.
(Bear in mind, common mode chokes help to suppress general coupled crap, as well as your not-quite-differential bus)
Fit the full range of capacitors - that linked X2Y device looks interesting, but I'm not seeing automotive approvals.
Full range would be across the bus, and from each line to ground, and I've also had good results from each line to the centre tap of the termination resistors, if those are on your board.


For the power - are you grounded to the chassis? If so, a common mode choke is right out. You're stuck with Ls and Cs, and whatever power protection you need for the job anyway. Bear in mind that if you have any other ground wires in your harness, that could be grounded elsewhere, current paths get messy and unpredictable.

How busy is your board? Is it just that micro and a bit of power stuff, or is it more complex? Are your connections only power & CAN, or are there a load of other signals you've not mentioned? Do they all come out on one harness, or do they head off all over the place?