CLC filter - overkill or not

[bodzio_stawski]

Hello!

I am trying to analize basic low-pass filters. I found an example of electronical design that is interesting for me because of something that I cannot understand.


(source: https://hackaday.io/project/164934-funkey-project-all-your-games-on-your-keychain)
(full schematic https://cdn.hackaday.io/files/1649347056536256/FunKey%20Rev.%20B%20Schematics.pdf)

To be precise, it is about the above CLC filter.

Correct me if I am wrong here somewhere:

1) L2 - This element protects the capacitors C8 and C10 from sudden surges in current
2) C8 - because of larger capacity, this capacitor filters a relatively low frequency noise
3) C10 - filters a higher frequency noise (kHz)

My questions:

1) The high value of the current drawn from the USB gives the advantage of using the LC filter over the RC filter (power loss on R). But as this is a CLC filter, what is the role of the C7 capacitor with this tiny 10nF capacitance, which is additionally in front of the coil, not like C8 and C10 - behind the coil? Can't we just remove C7 and not worry about EMC?
2) The values of capacitors, as I see are somethimes "tendentious" (I mean, as we can precisely calculate f. ex. reseistor for LED to get the accurate brightness, capacitors are selected for unspecified, occasional phenomena). For example, for decoupling, 100 nF capacitor is most commonly used. But for coils (or ferrite beads with impedance only), how to select/calculate the right one?

[jwet]

>>1) L2 - This element protects the capacitors C8 and C10 from sudden surges in current

Not really, that is one feature but a minor one.  Inductors have ac resistance or impedance- that is reactance and some resistance, this series impedance works in concert with the Caps to produces a filter.  This inductor is likely a ferrite bead type of some sort- these are designed to be extra lossy.  These types of inductors not only creates reactance but get very lossy at high frequencies- this was traditionally called a "choke"- this name isn't used much anymore.  At high frequencies, they create AC resistance (real- non complex resistance) due to skin effect and core losses.  This makes for a more effective filter like an LC but add this AC R without adding DCR.  Judging by the part number, this device reaches 600 ohms at some high frequency- you'd have to look up the specs, it might be 10 or 100 Mhz.  This is mainly to reduce conducted EMI from going out on the power input cord and increasing the effectiveness of the two caps to its right.  Sometimes you'll see clip on lumps in AC power cords that are used to suppress high frequency on AC cords- this serves a similar purpose.

>>2) C8 - because of larger capacity, this capacitor filters a relatively low frequency noise

Correct.  It also acts as a local charge reservoir for current peaks on the load side.  It likely becomes ineffective above a few MHz which is why its paralleled with C10.

>>3) C10 - filters a higher frequency noise (kHz)

Correct.  It augments C8 because of its larger size, parasitic inductance in its leads and construction becomes ineffective at some medium high frequency.  C8 starts to looks like a C with an L in series- its leads.  C8  "opens up" when this reactance of the lead inductance is equal to the capacitance- this is called self resonance.  It might be a 1 Mhz or two.  This second cap takes over and should have a self resonance at a much higher in frequency.  Dave has done some good videos on this bypassing and self resonant stuff.

My questions:

>>1) The high value of the current drawn from the USB gives the advantage of using the LC filter over the RC filter (power loss on R). But as this is a CLC filter, what is the role of the C7 >>capacitor with this tiny 10nF capacitance, which is additionally in front of the coil, not like C8 and C10 - behind the coil? Can't we just remove C7 and not worry about EMC?

True about using a series inductor.  Any resistor here would get in the way (I=500 mA and 1 ohm would create a .5v drop!)  C7's function puts the input Vsupply nominally at AC ground for signals going in either direction.  It also bypasses the lead inductance from the USB source- perhaps 2 meters long.  Without C7, load changes would ring up and down the supply lead from the inductance.  It also serves an importance EMI role by shorting any residual RF from going up the cord and radiating- its a pretty nice antenna.  Required/advised.  You would quickly learn that you needed this when you went to do your EMI tests at a lab.

>>2) The values of capacitors, as I see are sometimes "tendentious" (I mean, as we can precisely calculate f. ex. reseistor for LED to get the accurate brightness, capacitors are selected >>for unspecified, occasional phenomena). For example, for decoupling, 100 nF capacitor is most commonly used. But for coils (or ferrite beads with impedance only), how to >>select/calculate the right one?

Tendentious is an interesting way to put it.  These kind of parts are generally plopped down with rules of thumb and experience.  In some cases, when you take your first prototype to the EMI lab, you learn the kinds of things that work, they also tend to be overdesigned but failing a sample based EMI test in production is not a lot of fun and a few cents in capacitors is pretty cheap insurance.  If you did an exhaustive analysis, you'd probably find weaknesses in the precise values of some of these parts though there are other things in the circuit that aren't visible like layout that you can't really analyze.  I've seen academic analysis of paralleling caps in which it is shown some combinations can actually make things worse.  Here is link from Cypress that explains the logic- they recommend a single medium value, high quality cap very close to the load.  Ferrite beads are generally chosen by looking at your EMI emissions on the bench and seeing where you have problems.  You then select a bead that has enough loss at your frequency to squash it down.  Its not overly scientific in most cases and is often found by experimentation in a near finished system before final production release.

https://www.infineon.com/dgdl/Infineon-AN1032_Using_Decoupling_Capacitors-ApplicationNotes-v05_00-EN.pdf?fileId=8ac78c8c7cdc391c017d073d44f06172&utm_source=cypress&utm_medium=referral&utm_campaign=202110_globe_en_all_integration-files

You're wise to look at actual circuits to guide your learning.  Jim Williams, the famous LTC apps genius, used to recommend looking at old service manuals (and repairing) for old high quality test equipment- his favorites were old Tek O'Scopes.  HP and Tektronix both provide extensive service info for their equipment and HP (Agilent and now Keysight) maintain an extensive and free archive of all their manuals available for download free from their website.  Analyzing and trying to figure out why they did what they did is a great educational tool.  This equipment was universally designed by masters and understanding them is really fantastic.  Consumer and much modern gear isn't designed with such care.

[Benta]

I think you've got it backwards. +VBUS on the connector is an output, not an input.

[Nominal Animal]

I am a hobbyist only, but the use of an inductor in series on the supply voltage line scares me .

(I've actually managed to break stuff by switching on a power supply while connected, by the initial overshoot.  Then I learned how inductors, even ferrite beads, can cause this.  I now use a cheap PSU, Gophert NPS-1601, that does not have the initial overshoot.  See e.g. Voltlog #255 around the 6 minute mark.)

So, I looked up the Murata SPICE model for this,

Code: [Select]

*----------------------------------------------------------------------
* SPICE Model generated by Murata Manufacturing Co., Ltd.
* Copyright(C) Murata Manufacturing Co., Ltd.
* MURATA P/N : BLM31PG601SN1
* Property : Z@100MHz = 600[ohm]
*----------------------------------------------------------------------
* Applicable Conditions:
*   Frequency Range = 1MHz - 3GHz
*   Temperature = 25 degC
*   DC Bias Current = 0 A
*   Small Signal Operation
*----------------------------------------------------------------------
.SUBCKT BLM31PG601SN1 port1 port2
R1 port1 1 178.4
L1 port1 1 2.592e-6
C1 port1 2 1.360e-12
R2 1 2 451.3
L2 1 2 9.670e-7
R3 2 3 28.53
L3 2 3 5.440e-7
R4 3 port2 6.000e-2
.ENDS BLM31PG601SN1
*----------------------------------------------------------------------
and ran a Qucs-S simulation using NGSPICE with your supply circuit with a 100 Ohm (50mA at 5V) load, generic capacitors (10nF, 100nF, 10µF), and the above Murata 600 Ohm ferrite bead SPICE model.

As I feared, when the connection is made, a quick transient at about 25kHz frequency occurs, with first peak at a bit over 9V (after about 20µs of supply being connected), second at 8V (at about 40µs later), third at 7V, fourth at 6.5V, fifth at 6V, and so on.  It settles within a millisecond, but that 9V overshoot may kill not only your microcontroller stuff, but possibly even the 6.3V rated 10µF capacitor –– more experienced members here can say for sure.

[janoc]

I am a hobbyist only, but the use of an inductor in series on the supply voltage line scares me .

It is a noise suppression choke. Pretty common thing to do when you don't want noise or RF "wandering around" the circuit over the power rails.  Those inductors aren't huge (so any voltage spikes are limited)  and any voltage transients on power up/down will get "eaten" by the following capacitors and voltage regulator.

Certainly not going to damage anything. While the voltage may be large, the energy is what matters - and it isn't there due to the low inductance of that bead. Semiconductors are rated for much higher "zaps" due to ESD than what this bead can cause.

[Nominal Animal]

Interesting, Janoc; thanks!

I did verify that if one adds an 1Ω resistor in series before the first capacitor, it removes the overshoot, but obviously drops the VUSB supply by I/1Ω, i.e. by 1 mV/mA; half a volt at 500 mA load.  It also limits the inrush current to max. 3.3A during the first 30µs.

I also discovered that it is possible to use a P-channel MOSFET in parallel to the initial inrush limiting resistor, to avoid the VUSB drop.  Qucs-S had an IRLML6402 model (I would've preferred DMG2305UX), so I used that (it's drain-source on resistance at 5V being under 65mΩ), with 10 nF between its source and gate, and 47kΩ from gate to ground, and upped the resistor to 10Ω (between drain and source).  The idea is to slow down how fast the gate drops to ground.  This limits inrush current to 1A.  VUSB rises to 2.5V in 125µs, then to 5V in 75µs.  The 1A inrush pulse only lasts for ~ 60µs (starting at about 135µs from initial power on).  From about 200µs onwards, the MOSFET is fully on (V_GS≃-5V), so we get less than 33mV = 0.033V voltage drop even at 500mA load.

My question is, if one wants to use a ferrite bead in the supply voltage, wouldn't it make sense to splurge for the three additional components (costing about half an USD in singles), to avoid large inrush currents and initial voltage overshoot?  (I just do not know enough about ESD and ESD suppression in practice as to what kind of pulses can different components be expected to handle.  I only know that power-on overshoot has broken my ICs before, and I've had issues with spiky USB 5V current draws on single-board computers, so I'm asking, hopefully nicely! )  Or, is there a better way?  Or is this completely useless in practice?

[Siwastaja]

Filters can be damped by adding loss, which can be achieved by a few different ways:

* Series R with C, either explicit component or high-ESR type such as electrolytic or tantalum
* Parallel R with L
* L with AC lossy core, many ferrite beads are this way but not all

Also be aware of small ferrite beads saturating at DC currents they are well rated to, losing inductance and filtering capability.

I second the use of Spice simulations, it's a good learning tool, too. Power delivery network design is non-trivial.

I try to choose both AC lossy inductor plus AC lossy capacitor. Electrolytic with large C is a good low-cost component, large C swamps any parallel MLCCs and the high ESR of the elcap dampens the resonances (overshoot) one would see with only the MLCCs in place, due to the inductance acting with it - and this inductance exists as a stray component even if you don't use an inductor!

[janoc]

My question is, if one wants to use a ferrite bead in the supply voltage, wouldn't it make sense to splurge for the three additional components (costing about half an USD in singles), to avoid large inrush currents and initial voltage overshoot?

First, you are jumping to conclusions from a partial schematic. There is likely a voltage regulator on the VUSB line that would eliminate any overshoots.

The inrush current is not a problem because you can't get 3A of inrush current from a USB port! 1Ohm load is completely unrealistic on a USB port - USB ports are limited to 500mA only, if you want more, you need to negotiate it with the host e.g. using USB PD or some battery charging protocol. If you exceed that, the USB port overcurrent protection will trip out.

Also the choke will limit any inrush current (because that's what an inductor does - current rises slowly and lags behind voltage).

(I just do not know enough about ESD and ESD suppression in practice as to what kind of pulses can different components be expected to handle.  I only know that power-on overshoot has broken my ICs before, and I've had issues with spiky USB 5V current draws on single-board computers, so I'm asking, hopefully nicely! )  Or, is there a better way?  Or is this completely useless in practice?

But that power on overshoot that you had destroy some components was because of a poor power supply that had a bad power on behavior. Power supply is going to have much more energy available for causing issues than a ferrite bead with tiny inductance can accumulate in its magnetic field. That is not comparable. Again, to damage something having voltage is not sufficient. You do need also enough current available so that there is sufficient energy to break down the PN junction or insulation in something.

Van de Graaf generators can give you 100kV zaps (and they can hurt pretty bad) - and yet it won't kill you. Touch a 3kV busbar in a substation and you will be fried to a crisp on the spot. The difference is that the generator is not able to supply any sort of significant current whereas the busbar has the entire power station behind it.

The bead is there not for ESD reasons but to help combat high frequency noise. Most likely for EMI compliance reasons to prevent conducted EMI from propagating along the USB cables - uber simplified, inductor passes DC (it is just a coiled wire!), blocks AC. Capacitor does the reverse - blocks DC and passes AC.

[Nominal Animal]

The inrush current is not a problem because you can't get 3A of inrush current from a USB port!

Well, you can, with single board computers that use a big honking 5V rail directly on the USB.
(Odroid HC1 that I have is the problematic one.  Sure, they fixed it in later revisions, but my version can supply all its supply can, about 6A in my case.)
Similarly, wall warts often can sustain pulses of currents with just a quick temporary sagging on the voltage.

Not all USB implementations are as reliable as those on laptops and PCs.

Again, to damage something having voltage is not sufficient.

I know; I'm a physicist, and know the theory, just have no practical experience, and am therefore a complete hobbyist.

I just checked the very first 40µs in the simulated circuit, with a 10Ω load.  If the USB supply can provide enough current, the device gets a 0.2mJ pulse during that time.  At its peak, 20µs after power being connected, the voltage over the load is 9V and current 900mA (over 8 W instantaneous power), which sounds to me like it could perhaps damage some delicate silica barriers used in semiconductors.  And like I wrote above, I do have USB hosts that do not implement USB port current limiting at all, so I do consider it relevant.  (The instantaneous power dissipated in the load is very close to a sin(x)² half-wave, and based on its wavelength, it corresponds to about 25kHz; that's pretty low-frequency oscillation for many circuits.)

What I do not know, is what kind of energy pulses different components can actually take.  Is this acceptable circuit to power ICs?  Or is it "safe" only in some circumstances, like when providing 5V to a voltage regulator?  I do not know.  My fears may be completely unwarranted.

The bead is there not for ESD reasons but to help combat high frequency noise.

True.  It is just that the datasheets of the various ICs I use describe their ability to withstand such pulses in terms of ESD, typically referring to some standard of human-body model.  The one Wikipedia describes yields a narrow peak at 2kV/1.3A (over 2kW instantanous power) for example, but it decays very fast, the entire pulse being over in a microsecond.

I am talking about a "similar" event that takes about 50 times longer, but peak power is just one 250th of the ESD event.  Surprisingly, the energy in the pulse is very similar, 0.2mJ, according to my Ngspice simulations using Qucs-S.


What does this have to do with the circuit OP showed?

If there is only a voltage regulator (that can handle the initial voltage spike) on the VUSB line, nothing.

But, if some hobbyist like me happens to read this thread, and sees the circuit, and thinks they could use it as-is to provide a smooth, filtered VUSB supply to their project, then what I am describing could perhaps bite them in the ass.  (That is my fear, but cannot tell whether it is practical fear or only a theoretical one.)  So, it is very much related to the question at hand: in what situations is OP's schematic reasonable, and when it should be augmented with something else to be "safe", and why.

In other words, OP's "overkill or not" is not the only important qualification here: I'm trying to add "when is it safe? when might it bite one in the butt? Is there a better way to do this?"

[Nominal Animal]

Just to be clear: I am not at all convinced my own "fears" about this kind of supply filter are warranted.  I do see ferrite beads (120 to 600 ohms at 1 MHz) used in supply lines in most well-designed microcontroller dev boards, just not in a CLC configuration.

If anyone has designed or created this kind of circuits without problems (or has observed problems with them), I'd love to hear about it.  And I'm sure it would be interesting to others reading this thread as well, because knowing upsides/downsides of CLC filtering on the supply line would be useful.

To repeat my findings above, the overvoltage pulse I see in simulations is somewhat similar to an ESD event, with very similar energy to the human body model at 2kV (0.2mJ, or 200 microjoules), but 1/250th of the peak power, taking 50 times as long.

[MrAl]

Hello!

I am trying to analize basic low-pass filters. I found an example of electronical design that is interesting for me because of something that I cannot understand.


(source: https://hackaday.io/project/164934-funkey-project-all-your-games-on-your-keychain)
(full schematic https://cdn.hackaday.io/files/1649347056536256/FunKey%20Rev.%20B%20Schematics.pdf)

To be precise, it is about the above CLC filter.

Correct me if I am wrong here somewhere:

1) L2 - This element protects the capacitors C8 and C10 from sudden surges in current
2) C8 - because of larger capacity, this capacitor filters a relatively low frequency noise
3) C10 - filters a higher frequency noise (kHz)

My questions:

1) The high value of the current drawn from the USB gives the advantage of using the LC filter over the RC filter (power loss on R). But as this is a CLC filter, what is the role of the C7 capacitor with this tiny 10nF capacitance, which is additionally in front of the coil, not like C8 and C10 - behind the coil? Can't we just remove C7 and not worry about EMC?
2) The values of capacitors, as I see are somethimes "tendentious" (I mean, as we can precisely calculate f. ex. reseistor for LED to get the accurate brightness, capacitors are selected for unspecified, occasional phenomena). For example, for decoupling, 100 nF capacitor is most commonly used. But for coils (or ferrite beads with impedance only), how to select/calculate the right one?

Hello,

There are different kinds of filters even though they may be LC filters alike.
One type is of signal filtering or signal conditioning or oscillators of low power.
The other main type is for power filtering.

The signal ones dont have to handle much power and they are usually small.
The power ones are usually larger and contain a significantly sized inductor.

For questions 2 and 3 you got it pretty much right.
For question 1 that's also somewhat right when it comes to power circuits, but the way you usually think about it is that the inductor reduces ripple current getting to the capacitor.  Electrolytic capacitors have a maximum ripple current spec that tells you how much it can take.  The inductor helps to mitigate that.  Of course the inductor also improves the filtering action because it changes the circuit from a first order filter to a second order filter which by it's nature can filter better.  This means the output will be 'smoother'.

All inductors have 'back emf' that's actually a way of thinking about how they work.  If they did not have that, they could not operate as an inductor.  Just like capacitors try to keep the voltage across them constant, and inductor tries to keep the current through it constant.  If the external path from one terminal to the other has high resistance, you will see a large voltage develop across that resistance, and if it is open in theory the voltage goes to infinity and in practice the only thing that limits it is the construction materials used in making the inductor.  This means you have to pay attention to how the inductor is switched.  This often requires either a catch diode or a transient suppressor or for lower power a zener.
One saving grace is that you can look at the inductor storage as not only current but volt seconds, which means that the higher that voltage goes the faster the inductor discharges.  This means the higher the resistance the faster it discharges and the spike ends.  That could still damage components especially if that behavior repeats.
What you could do is analyze the circuit for overshoot when it is disconnected.  The capacitors may eat up the spike, but if not, you may be able to increase their values.

To analyze this circuit for AC voltages you can do an AC analysis using complex numbers to make the analysis very simple, almost like doing a purely DC circuit.  To analyze the transient response, you can use differential equations or take the easy way out and use Laplace Transforms.  Laplace Transforms allow you to take the simpler AC analysis and transform it from the frequency domain to the time domain and therefore obtain the time dependent responses.  These techniques are WELL worth looking into if you have never done this before.
The heart of it all is Nodal Analysis or similar.  Very well worth learning you will never regret spending some time on this.

[bodzio_stawski]

Thank you for the answers. I attached to the first post a link to the original complete schematic, so that it is known what is beyond the usb circuit.

It also bypasses the lead inductance from the USB source- perhaps 2 meters long.  Without C7, load changes would ring up and down the supply lead from the inductance.  It also serves an importance EMI role by shorting any residual RF from going up the cord and radiating- its a pretty nice antenna.

The inrush current is not a problem because you can't get 3A of inrush current from a USB port!

Well, you can, with single board computers that use a big honking 5V rail directly on the USB.

Nominal Animal raised a very important and interesting point. When the usb connector is the "beginning" of some device (power source,  data receiving), then I think about what exists to the left of the device. It's a bit like with small portable devices (such as ecigarettes), which are often bought without chargers and connected to unreliable power sources (cheap car charger, for example), with risk of harm. The device from which this section of the diagram comes from seems to be non-commercial (I think), but I think that a person has the nature to do something as best he can, as long as he knows that it can be done better (not only when the company, for example, requires him to take care of the EMI issue). Therefore, somehow I feel that it should be assumed that the USB source may be simply bad (and, for example, give much more than 500 mA) and design protection for our device before we connect it to this unsafe USB power source. Wonderful topic <3 - it would be fun to find some other nice literature on designing usb devices (and protecting against interference and other problems).