Iron Powder versus Ferrite Inductor for Buck regulator

[pieterc]

Hi everyone 

I have selected the Diodes AP61100Z6-7 buck regulator (https://www.diodes.com/assets/Datasheets/AP61100-AP61102.pdf) and it has a 2.2 MHz switching frequency. I am trying to make a choice between two inductors:
1. Iron Powder core: Murata DFE252012P-1R5M=P2 (https://www.digikey.com/en/products/detail/murata-electronics/DFE252012P-1R5M-P2/5247259)
2. Ferrite core: Taiyo Yuden NRH3012T1R5NN (https://www.digikey.com/en/products/detail/taiyo-yuden/NRH3012T1R5NN/2329088)

The Murata inductor looks attractive from a package / footprint point of view but I not sure if it is the right choice for this buck converter application regarding efficiency and EMI. Here are more details:
https://www.murata.com/en-eu/products/productdetail?partno=DFE252012P-1R5M%23

I understand that the inductor's self resonant frequency must ideally be 10 x higher than the switching frequency so that it behaves like a low pass filter for 3rd, 5th, 7th and 9th square wave switching node harmonics but the Murata inductor's graphs are perplexing. The maximum Q is low at 0.7 MHz but the inductance stays flat up to 10 MHz with a peak at 46 MHz. The impedance is ~ 90 Ohm at 10 MHz and then it climbs rapidly as frequency increases.

So is this iron powder core inductor better suited for DC power line filtering but not suitable for buck regulator applications?

The Tayo Yuden inductor is a shielded drum core design but not fully enclosed by ferrite. If I connect the buck regulator's switching node to the winding closest to the center of the core would that provide sufficient EMI margin? It reduces capacitive coupling but not inductive coupling.

Comments are much appreciated!

Thanks in advance,
Pieter

[moffy]

I would lean towards the ferrite, self resonance is higher than the powdered metal alloy core.

[T3sl4co1l]

I'm not too concerned about either one; the lower SRF on one means somewhat higher capacitance and thus high-frequency noise at the output; layout will be a few dB more critical, or perhaps you need more distance (to avoid the local switching-loop high-frequency currents) or a small ferrite bead or something (for a few more dB attenuation at such frequencies).

Ferrite might be lower loss, but they don't exactly give Q, so who knows.

I might choose between them based merely on cost, and any size limitations if applicable.

You might also consider a larger inductance, since the COT control doesn't have particular limitations like a peak current mode control for instance does, but I haven't looked at the datasheet in detail, and mind the limitations of switching and compensation of the integrated regulator.  More inductance trades off with lower SRF to reduce current ripple and core loss, but increasing EMI some.

There's also EMI in terms of near fields, which I don't think either part has much going for it; the ferrite has a spool design with fringing fields around all sides, while the powder will have some leakage out of the body -- while it's more self-shielding than the ferrite part, you don't know what orientation the winding is within it, and so how to avoid that field.  Both are at least mildly conductive so the near E-field will be about the same.

Tim

[mtwieg]

It's very difficult to suggest one over the other since:
1. You didn't specify your intended maximum output current and ripple current
2. Neither part gives you any guidance on estimating core loss
3. It's impossible to predict how they will compare in terms of near-field emissions, as you mention.
4. Both seem to have fairly similar saturation curves (likely because the Taiyo Yuden part is only "semi-shielded" and has an effective air gap)

I typically avoid both types of inductors (semishielded drum cores and "chip" type forms) unless I'm very space constrained and know that some near-field coupling won't be an issue.

[pieterc]

@moffy, @T3sl4co1l and @mtwieg,

Thanks for the valuable comments 

It is giving me pointers to think, calculate and research in the right direction. For example that the semi-shielded inductor has a softer saturation curve, because of the effective air gap. I understand non-ideal Voltage "stuff" (e.g. caps, MOSFETs, E fields) much better than non-ideal Current "stuff" (e.g. inductors, transistors, H fields).

Best regards,
Pieter

[iMo]

Perhaps not much relevant, but the DS of the chip recommends DC resistance of the coils "less than 50mOhm" while both types above are 60. The 1uH ones are lower though..

[pieterc]

For the benefit of others reading this post too: Bourns are kind enough to provide an equivalent circuit of their inductors so I am able to SPICE it and see the trade-offs:

Bourns SRR4818A-1R0Y (shielded drum core; Ferrite):
https://www.digikey.com/en/products/detail/bourns-inc/SRR4818A-1R0Y/7555996
https://www.bourns.com/docs/product-datasheets/srr4818a.pdf
https://www.bourns.com/engineering/SRR4818A/SRR4818A_Series%20Equivalent%20Circuits.xlsx
L = 1uH; Rser = 19.2 mOhm; Rpar = 1276 Ohm; Cpar = 0.855 pF (Isat = 3.6A)
$0.82 in single quantities

Bourns SRP3212A-1R0M (shielded metal core):
https://www.digikey.com/en/products/detail/bourns-inc/SRP3212A-1R0M/16546451
https://www.bourns.com/docs/product-datasheets/srp3212a.pdf
https://www.bourns.com/engineering/SRP3212/SRP3212_Series%20Equivalent%20Circuits.xlsx
L = 1uH; Rser = 32 mOhm; Rpar = 936 Ohm; Cpar = 9.88 pF (Isat = 6.5A)
$0.26 in single quantities

There is a 10 x order difference of Cpar between the shielded drum core and the shielded metal core!

Pieter

[T3sl4co1l]

You may find this of interest:
https://seventransistorlabs.com/Calc/Coilcraft1.html

Cpar isn't necessarily a physical parameter, and may be subject to fitting error.  Generally it'll determine the SRF and falling slope, but depending on how the part has been modeled (the older Coilcraft models are discussed here) it might also affect HF losses, whatever HF asymptote if any is modeled, or series-resonant dip, etc.

In general, an inductor is a transmission line component, i.e. it's made of a wire near another conductor(s) (including itself, i.e. a coil is a transmission line with dispersive characteristic).  The parallel SRF (Z peak) is given by electrical length multiplied by permeability, since it interacts strongly with the core; the series resonance is 1/2 wave and may not interact with the core (much, or at all?).  There may be higher modes, or they may be dampened out by core and dielectric losses, giving a slower-varying impedance spectrum beyond cutoff.

Typically only parallel SRF is measured and modeled, but the capacitive slope and series resonance (and beyond, as far as harmonics from the inverter extend) are relevant to overall converter response.

Tim

[mtwieg]

Again, I'll emphasize that predicting EMC/EMI behavior of an inductor based on the crude info in the datasheet is generally a waste of time. But if you're really interested, I'd look for white papers and app notes from vendors showing real-world data. For example, this appnote from Wurth (it's not clear how they actually made these measurements though). One thing worth keeping in mind is that which pin is connected to the switching node can affect EMC performance, which is why many inductors will have a pin 1 indicator.

I've never encountered a design where the SRF or parasitic capacitance on an inductor in a switchmode power supply had any noticeable impact on performance. The only reason I might care about it is if there's a specific frequency band I want to avoid contaminating, I might avoid selecting an inductor with a SRF near that band.