Input EMI Filter Design

Ntip
TL;DR Summary
How do I go about selecting a reasonable frequency response and cutoff frequency?
I am starting the input filter design for a converter and only have the voltage and power specs. It will be the input filter to other converters on a parallel bus, so I could look at the datasheets for those and get some specs there. I do not have any information from the input converter to this bus, I just know the voltage is regulated and the power it will be supplying. I'm not sure what specs to look at to determine what I should set as my cutoff frequency or how to determine an appropriate target frequency response. Does having parallel converters on the output complicate things, or do I just look at the worst case converter (I'm not really sure what to look at either way)?

I was going to start off with a DM and CM filter design but I'm not sure how to determine if I need both DM and CM since the input and output will be regulated. Also, the the output and assuming the input both have DC/DC converters, they likely already have their own capacitance. How is this factored in?

I was going to start using Heavyside's criterion for stability so require Zo << Zin or Zo=1/10*Zin where Zo is the output impedance of the filter and Zin is the input impedance to the bus. Zo can be calculated using Zo=sqrt(L/C) and Zin=(Vin^2*efficiency)/Pout. Pout would be the output power from all converters on the bus added together. I guess I should use a lowest power output to increase Zin and further reduce the Zo requirement?

I have questions about setting Zo=1/10*Zin because does this still hold when the input filter is between two regulated DC/DC converters?

I have a lot of questions and not many unknowns. I'm sure I'll create more questions for myself once I really start digging into this design. Any help is much appreciated.

Nick

Mentor
Could you post a sketch of the system? It sounds sort of like you have one off-line SMPS generating a DC bus that other DC-DC converters are hooked to? Or is it something else?

Gold Member
The design problem you are describing is very difficult in practice. There are analytical solutions and engineered design approaches. However, in my experience, you rarely actually have the data you need in practice when designing new systems. Often the hardware just isn't available to characterize adequately or you don't have the time you would need to do the testing required. As a result, a large fraction of EMI compliance design is done by guess work, trial and error.

If you can, you would first characterize the conducted emissions from the hardware without an EMI filer and compare these results to the limits required. This would give you the required attenuation vs. frequency and propagation mode (DM, CM, radiated, etc.). Then you do a filter design according to theory, implement, and check to see if it worked OK with stability testing, EMI testing, functional testing etc.

The theory for SMPS input filter stability requirements was well developed in the late 1970's by Middlebrook at Caltech. There is a ton of stuff on the web about the theory, but no real cookbook procedure. Here are a couple of links:

alan123hk
Gold Member
The problem with an analytical approach to SMPS EMI is that many of the EMI generating/coupling mechanisms are parasitic, second order, effects that depend on details that you can't always fully understand in the initial design. Things like HF components of switching waveforms, snubber design effectiveness, coupling of EMI to heatsinks, enclosures, traces, etc.

Ntip
Could you post a sketch of the system? It sounds sort of like you have one off-line SMPS generating a DC bus that other DC-DC converters are hooked to? Or is it something else?

yes, you are correct. The bus voltage is generated by a SMPS. Here is a simple sketch of the system.

Attachments

• Bus Architecture.PNG
3.7 KB · Views: 67
Mentor
yes, you are correct. The bus voltage is generated by a SMPS. Here is a simple sketch of the system.
Okay, that helps. So the off-line SMPS already has its own standard input line filter for EMI compliance, right? Although maybe not, since you said this:
I do not have any information from the input converter to this bus, I just know the voltage is regulated and the power it will be supplying.
Do you know the switching frequency of that input off-line SMPS? What are its output ripple specs (and the output voltage)?

Are you intending to put one of your custom input filters in series with each of the DC-DC converters on the right?

One issue to keep in mind is that the SMPS on the left may have an issue with stability if the loads it is feeding have input current ripple near its switching frequency. When the load current changes at a frequency near the frequency of the switcher (or the first couple harmonics), that can cause stability problems.

I've seen this in my stability analysis and testing of a DC-DC converter that I worked on many years ago. The application circuits that it fed were up to the customer, and if the customer had a circuit that chopped its input current near the 150kHz switching frequency of our DC-DC circuit, that could cause stability issues and you could get very large output voltage ripple (out of spec).

Also, if you have several DC-DC converter circuits that are all connected in parallel to a supply rail, you need to be careful about them "talking to each other" if they have the same switching frequency. This also can lead to stability issues -- you can consider synchronizing them or ensuring that their switching frequencies are far enough apart not to cause this problem. If you are planning on putting one of your input filters on each of the DC-DC circuits on the right, that will help to mitigate this issue.

dlgoff
alan123hk
The design problem you are describing is very difficult in practice. There are analytical solutions and engineered design approaches. However, in my experience, you rarely actually have the data you need in practice when designing new systems. Often the hardware just isn't available to characterize adequately or you don't have the time you would need to do the testing required. As a result, a large fraction of EMI compliance design is done by guess work, trial and error.

If you can, you would first characterize the conducted emissions from the hardware without an EMI filer and compare these results to the limits required. This would give you the required attenuation vs. frequency and propagation mode (DM, CM, radiated, etc.). Then you do a filter design according to theory, implement, and check to see if it worked OK with stability testing, EMI testing, functional testing etc.

The theory for SMPS input filter stability requirements was well developed in the late 1970's by Middlebrook at Caltech. There is a ton of stuff on the web about the theory, but no real cookbook procedure

A very insightful analysis, the complete theoretical analysis and design of the switching converter input filter is not an easy task, especially if there are multiple switching converters connected in parallel, it will make things more complicated.
Please note that the input terminal of the constant load regulated output switching converter will exhibit negative resistance characteristics. If the negative resistance becomes greater than the output resistance of the input filter at a certain frequency, it may oscillate. The oscillation frequency depends on the output reactance of the input filter and the input reactance of the switching converter.

Last edited:
berkeman
Ntip
Thanks for the feedback everyone. Just the comments so far has given me a lot to look into. It looks like I have a steep learning curve ahead of me.

dlgoff and berkeman
alan123hk
Summary:: How do I go about selecting a reasonable frequency response and cutoff frequency?

I was going to start using Heavyside's criterion for stability so require Zo << Zin or Zo=1/10*Zin where Zo is the output impedance of the filter and Zin is the input impedance to the bus. Zo can be calculated using Zo=sqrt(L/C) and Zin=(Vin^2*efficiency)/Pout. Pout would be the output power from all converters on the bus added together. I guess I should use a lowest power output to increase Zin and further reduce the Zo requirement?

I have questions about setting Zo=1/10*Zin because does this still hold when the input filter is between two regulated DC/DC converters?

I think if you plan to add a filter between the two converters, or cascade connection of multiple filter sections, then you can also apply the Middlebrook criterion you mentioned to consider related issues.

According to the Middlebrook criterion, if the condition Zo << Zin is met, the added new filter has the least impact on the transfer function characteristics of the original system, that is, the added new filter should not affect its stability. Of course, the last step is that you must carefully verify that the system is operating as expected.