Better DC Power Filtering?

[masinyourface]

Howdy y'all!

For most of my projects up until now, I've used the standard RC low-pass filter on the incoming DC input power before it hits my circuit. Short, simple, and fairly effective.


—RC Low-Pass Filter



Lately though, I've been itching to over-engineer everything for the sake of my own vanity. Specifically, I've been considering implementing an LC low-pass filter to replace the RC filter.


—LC Low-Pass Filter

Obviously this will raise the cost of each build for little improvement, but I'm young, unmarried, and have the extra spending money so why not?!? 

Where I'm at now is choosing inductor/capacitor values. I've tried looking through the forum hoping to see if anyone else has tried doing this, but haven't found anything specifically pertaining to LC circuits. Does anyone have any advice/experience, or would you happen to know where to direct me? Any help is appreciated!!

[GibsonGM]

Wow, that's total overkill, man!    If you really insist on doing this, look up some tube amp information...old designs used these networks for power filtering quite often.  You could look up "L C power filtering" on Google or something, too.   Maybe there will be more info there about WHY they used L's in the early days, and why they don't now.    We just don't need that level of stability now (achieved other ways, actually), and better 'stuff' is available that replaced the L's...plus they contribute to noise and so on. 

Carry on if you must, but I gotta say it again....you'll never have need for this level of filtering with what we do     But maybe you just want to do it and see it in action, that's cool.

[masinyourface]

Oh, I know it's hilariously overkill, but just let me be vain this one time hahaa 

[garcho]

it has nothing to do with vanity, believe us. use your money and time learning how to do cool stuff, instead of implementing old fashioned circuitry for the sake of using a new component or being "different" (audio electronics are stone-aged electronics, unless you're TC Helicon or Eventide or something, forget about doing anything truly unique). and doing something that other people don't do is only cool if that thing is cool, in this case, it's just a boring part and a couple basic equations, so maybe not cool? it won't be "the secret sauce" to anything, that's for sure, and it won't keep your rig quieter. why not learn how to make your own bipolar or "positive ground" power supply with an IC or something like that? or find an experienced amp repair technician to show you how to make your own power supplies and how to safely wire mains and high voltage? or if you like filters, audio is alllllllll about filters, even filters that don't "do" anything (phasers), it would be a far more useful exercise and expense. are you familiar with "gyrators"? basically, using op amps and RC filters to mimic inductance. way more fun!

[R.G.]

OK, next step up on the ledge. 

Capacitors get better at shunting AC to ground as frequency increases, as their impedance goes down with frequency. Inductors get better at, well, impeding  AC because their impedance goes up linearly with frequency. Great, right? high AC impedance and low DC impedance. Gotta be better.

Um, maybe. What happens when you feed an LC circuit an AC signal that's at the frequency where their impedances are equal? Yep, you get a resonance, and unless it's damped, it will set up a resonant howl. LC resonance is one of the few ways to use passive components get out a bigger AC voltage than you put in. So L-C circuits bring with them an immediate worry about introducing damping to your power filter so it won't ring on you and make things worse in some cases.

Some situations are naturally damped. Some situations are under damped. You, the designer, has to figure out which is which, and what to do about damping things.

Or just be lucky.   

[reddesert]

Read a resource such as the wikipedia page on RLC circuits: https://en.wikipedia.org/wiki/RLC_circuit.

R.G.'s point about resonant frequencies is important. Here's also a practical issue: the amount of inductance you need to make a LPF. The low-pass filter RLC configuration is shown in Figure 6 of the wiki page.  Find that and it will also give you the formula for the low pass filter corner frequency.  The corner, in angular frequency, is omega = 1 / sqrt(LC).  In units of Hertz, it's f = 1 / (2 * pi * sqrt(LC)).

In contrast the corner frequency of an RC filter is f = 1 / (2 * pi * RC).

So if you have a RC filter with R = 100 ohms and C = 47 uF, which is pretty common, then the LPF corner frequency is 34 Hz.  That sounds good!  It's nice and low, and the parts are easily available.

Now try achieving that same corner frequency with an LC filter. Solve f = 34 Hz = 1 / (2 * pi * sqrt(L * 47 uF)).  We get L = 0.5 Henries. An 0.5 Henry inductor is a larger, more annoying and expensive part than a 100 ohm resistor. And there's still the concern about resonant behavior of the circuit.

This is one reason why designers often use gyrators to simulate inductance in filter circuits, for example: because inductors are a pain in the neck.

[GibsonGM]

I remember reading about a few of those things, esp. ringing...I knew others would come in with more practical reasons why not to use them!  They used them 'back then' because they had to.

That said, don't we ALL get a period of amazement at inductors and their properties as we're learning, and try to find real world places to use them?     I still get excited when I can use one in a SMPS!! 

[amptramp]

If you are looking at power supply filters, you can use this resource:

http://www.duncanamps.com/psud2/index.html

and compare the effects of L-C filtering and R-C filtering.  Note that choke input does result in nice low levels of ripple but only at a constant load meaning that power stages must operate in Class A push-pull.  This has been done for some audio systems but these are very inefficient amplifiers.  Your filter output impedance will be approximately SQRT (L/C).  If your load impedance is lower, you can oscillate.

If you have a switching regulator after that, the feedback makes the load a negative resistance because a rising input voltage causes a falling input current and vice versa with a constant load.  If the magnitude of the negative resistance of the regulator is the same as the positive resistance of the L-C filter or less, then you have an oscillator.

[Transmogrifox]

If you really want to step it up a notch and just do cool stuff for fun, then implement an SMPS at, say, 1 MHz with control-loop bandwidth of 100 kHz.

The SMPS control loop will take care of all your normal audio-frequency PS garbage and then you can use small-sized, small-value and low-cost components to filter the high frequency garbage from the SMPS.

The key to a quiet SMPS is a well-behaved stable control loop.  An added benefit is you could use something like a SEPIC topology to regulate anything between, say, 3V to 20V, so you could connect just about any dirty old PS to your stompbox and get a strong, steady 9.5V output.

People will point out this potentially gets you into trouble with the FCC if your SMPS is letting out garbage above Class B emissions limits.  You can somewhat mitigate this by using a common mode choke and a pair of ferrite beads on the input and feed this with short-length wiring from a supply with the FCC logo and Class B rating. Just don't go selling these things en masse and they probably won't come sniffing you out.

You probably want ferrite beads and a common mode choke on the output feeding into your stompbox circuitry as well.

As you can see, this would easily satisfy a desire for a Rube Goldberg style for a stompbox power filter.

On a more serious note, an LRC network can be an excellent filter if you have the space.  Components will get rather large if you're aiming to squelch 60 Hz hum.  This is looking like a 3300uF capacitor and a 220 mH inductor to get 40 dB rejection on 60 Hz.

For reference, the same rejection could be had with a 100 ohm resistor and the same (3300uF) capacitor.  A 220 mH inductor is likely to be very large compared to a 100 ohm 1/8 Watt resistor.  This is much like a wah inductor, which often have about 20 to 50 ohms DC resistance anyway, so you still suffer the same series voltage drop.

Inductors become much more practical at high frequencies.  This is why the suggestion for an SMPS.  An SMPS can reject the low frequency input noise, leaving only high frequency stuff to reject with an LC filter using small L & C sizes and values.

It's higher complexity, but if you were using SMT components and layout typical of a converter like this then the space would typically be less than the equivalent rejection using an all-passive solution.  You begin to lose space with an SMPS when you begin to care about RF emissions.

[PRR]

Start with what chokes you can readily get.

NO! Start with Specification of Load: voltage, current, susceptibility to crap.

(If you are powering a TL072 with any reasonable Vbias filter, the raw 9V supply can be full of crap, because of high PSRR.)

I will postulate 9V 18mA for an example.

Since 1,000uFd 16V caps are cheap, we can probably do well right off a rectifier with such a cap. Even 220uFd gives "only" 0.5V ripple. But you gotta go fancy...

In any case, you want the C facing the load to be "small" compared to load at the lowest frequency of interest, to bypass the circuit being powered. 9V/18mA= 500 Ohm load. By rule of thumb I would use >10uFd for 1K, so >20uFd for 500r.

We want the L to be larger than the load at the incoming crap frequency. Take 1 Henry. At 100Hz it is 628 Ohms.

We probably want much more than 22uFd and 1 Henry.

OK, Hammond has 154E 20H 20 ma. 1666 Ohms. That 1.7K resistance is far too large for a 9V/18mA= 500 Ohm load. We get just a couple of Volts out.

Hammond 154M 2H 100 ma. 175 Ohms is the same size (/cost) and a more likely Ohms. With just 22uFd on its output we get 13mV ripple. Going 470uFd gives less than a mV ripple.

Still you expect 175 Ohms to be a large loss against a 500 Ohm load. And since you may not know the load exactly, it may even vary in use, we prefer to aim for quite-small loss.

All of this CAN be figured on a matchbook, maybe with a sliderule. However graphs make people go "wow!" and computers graph-and-post faster than me scanning a doodle.

Duncan PSUD. Get it! Don't give any backtalk about your computer/sellfone won't run it; push the neighbor's kid off his Win-machine if you must.


Indeed 175 Ohms into 500r drops from 9V to 6.7V.

We want >1H with much lower R. One candidate is Hammond 159V 1.5H 500 ma. 27 Ohms. It will drop from 9V to 8.36V, so we can probably work with it. With 470uFd out we get 1mV ripple. It is 9X the weight (cost?) of the 154 models. It is bigger than some pedals.

Compare to a "capacitor multiplier", 1K to base of NPN, 1,000uFd to ground, load at NPN emitter. The drop will be a steady 0.7V over a fair range of current, and about 1mV ripple out. The R+NPN is far cheaper and smaller than any of these chokes.

Oddly, none of these sims "rang" significantly, even on a bang-start. It CAN happen, but may be more frightening than likely.