# Calculating interwinding capacitance of inductor

Discussion in 'Electronic Projects Design/Ideas/Reviews' started by Flyback, Aug 11, 2017.

1. ### FlybackWell-Known Member

Joined:
Jan 5, 2007
Messages:
1,990
Likes:
34
We wish to use Power inductors for filtering SMPS. We want them to have as little interwinding capacitance as possible because we want to maximally impede the high frequency common mode noise from getting through this stray capacitance.

As you know, interwinding capacitance is not listed in inductor datasheets, so please could you confirm that the “Self Resonant Frequency” is what we need to use to calculate this stray capacitance?

Thats is, w = 1/[(sqrt(LC)]

Therefore C = 1/(w^2 * L)

2. ### BobWActive Member

Joined:
Apr 28, 2010
Messages:
501
Likes:
52
Location:
Nope. Self capacitance tends to be a fixed value at frequencies well below self resonance. But as frequencies start to approach self resonance the self capacitance begins to change quite dramatically. So you can't use the self resonant frequency of the coil to calculate coil self capacitance at lower frequencies. The only accurate way to determine self capacitance is to measure it at the frequency where you intend to operate the circuit.

• Like x 2
3. ### MrAlWell-Known MemberMost Helpful Member

Joined:
Sep 7, 2008
Messages:
11,032
Likes:
952
Location:
NJ

Hi,

I would think that if you want to "maximally impede the high frequency" then you would probably want an after-the-fact solution because you cant get a component with zero capacitance.

For example, for a capacitor in series with another capacitor and excited by a an AC voltage source and with the output taken across one of the caps, the output is in inverse proportion to the ratio of the cap values. So if one of the caps is say 1uf and the other 100uf, the output for an AC excitation will be approximately 1/100 which as you know is -40db.

To figure out how much parallel capacitance you need, you can do a measurement at some frequency and add a cap in parallel to the input and see how much it attenuates, and then you'll know if you have to add more or get away with less capacitance.

Thus adding a cap helps to reduce the high frequency components of the unwanted signal. That's one way to handle it anyway

I assumed your application was some sort of line filtering.

• Like x 1

Joined:
Jan 12, 1997
Messages:
-
Likes:
0

5. ### ronsimpsonWell-Known MemberMost Helpful Member

Joined:
Apr 17, 2007
Messages:
7,259
Likes:
961
Location:
Loveland, CO USA

Flyback,
Here is a part I have used many times. As you can see the winding is broken up into two pieces with a wall in between. This pushes up the resonant frequency / reduces the cap. I have used layers of tape to do the same thing.

MrAl: Agree, I have used a transformer + small choke. The idea there is that at high frequencies the transformer starts to "short" but the small choke "opens" up to block.
BobW: Agree/ Disagree, Interwinding cap is very complex. (many small caps not one simple cap) Using spice; I take the L and the resonant frequency to find the C. Then add in the R of wire so I get a Q that is close. I think this is a good first order spice model. (not perfect)

I do not SPICE the entire power supply because there are so many errors. Above 10mhz the PCB's C and L is very important and not known. I make a fast prototype and see where the noise problems are. Then make a SPICE of the filter, knowing it is not perfect. The plan is to look at problems and look at the filter. (how to get 6db more at 10mhz and 10db more at 300mhz to 400mhz) Even if there are errors; getting "6db more" has been very good.

For capacitors I add L & R. For inductors I add R & C. My thought is, this gets me very close to the data sheet of the part. There are other errors that out way making better models.

• Like x 1
6. ### BobWActive Member

Joined:
Apr 28, 2010
Messages:
501
Likes:
52
Location:
I certainly agree that self capacitance is a complex phenomenon. I prefer not to use the term "interwinding capacitance" however, since a significant component of self capacitance is due to the transmission line time delay effect of the inductor. The other principal but smaller contributors are fringing capacitance at the lead ends, and inter-layer capacitance in coils with multiple layers. Many people (some engineers and physicists included) believe that capacitance between adjacent turns is a primary contributor, but this hypothesis has been known to be false since the early 1900's.

However, regardless of how complex it is, the notion of a lumped component fixed self capacitance works very well for many calculation purposes over a range of frequencies from near DC, up to about 2/3 of the first self resonant frequency.

I also agree that to minimize self capacitance, it's best to pick a coil with the highest self resonant frequency, and generally the higher the Q, the better. Still, knowing the self resonant frequency won't tell you what the self capacitance is going to be. And even after measuring the self-C, it probably won't give enough information by itself to determine its harmonic filtering capabilities. The design is going to require serious lab testing.

• Like x 1
7. ### MrAlWell-Known MemberMost Helpful Member

Joined:
Sep 7, 2008
Messages:
11,032
Likes:
952
Location:
NJ
Hello,

Yeah i guess i thought of it that way too more or less, that the series inductor has inductance that 'opens' the circuit a little with higher frequencies but the capacitance shorts it out to some degree with various frequencies. The only solution i think is with a shunt capacitance on the output, a capacitor that can handle the frequencies of interest.

A transformer works mostly on mutual inductance, but there's some capacitive transfer of energy in there too. I found that i could make a transformer that worked mostly on capacitance rather than mutual inductance and get reasonable transfer of energy at higher frequencies. The capacitance calculation is not going to be easy though because of so many windings that make the physical layout very complex. I suppose it is possible but i have not tried it yet partly because i dont think it would be practical.

• Like x 1