Caps across chokes trick

[001]

Hi!

I hear about old trick to eliminate ripple in C-L-C filters after rectifiers
The idea is small cap adding across choke and trimming it for ripple resonant frequency (120hz)


but I see that it is not common at factory-made gear. Why? What about practical equations? What is affordable accuracy?

[Berni]

Punch a few numbers into the calculator and see just how big of a L and C you need for this.

Then afterwards have a look at the ripple rejection graphs in the datasheets of even cheap crappy linear regulators and you will see why.

[001]

Punch a few numbers into the calculator and see just how big of a L and C you need for this.

About 0.68uF will be Ok for typical 5H choke  isnt it? It is very cheap

[Berni]

Go check how much a 5H inductor costs.

Then compare that to a cheap linear regulator IC that can be had for about 40 cents even at places like digikey in low volume. That cheap regulator IC gets you at least around 50dB of noise rejection at 120Hz. If you spend money more on a regulator IC you can get ones up about 80 to 100dB of noise rejection. This means that even the worst linear regulators can reduce the noise from 1 V to 0.003V. Or if you are after really low noise you could put two of those 80dB rejection regulators in series to get a theoretical 160dB of rejection bringing that 1V of noise down to 10nV (Of course you wouldn't actually get that because the regulator itself is probably generating more noise than that by its own internal reference)

I don't know how many dB of rejection you can get out of such a LC filter, but if its tuned like this that makes it most effective at the single tuned frequency while the mains sine wave does also have other harmonics on it, so at least some of those would get trough.

This method however was pretty common back in the days of vacuum tubes, since using an extra tube just to regulate one power rail was too costly.

[duak]

I remember reading about this in one of the 1960's ARRL handbooks for radio amateurs.  It made more sense for vacuum tube equipment where the voltage is high and the current is low.

It can work but there are two problems to be aware of: unless an air core inductor is used, its inductance varies as a function of current due to core saturation.  The second is that a resonant LC circuit can be formed and if the load draws intermittant current at the resonant frequency, high AC voltages can build up.

[T3sl4co1l]

Rarely done because two chokes are needed.

Note that the cap is a 100% bypass for high frequencies.  You notch 100/120Hz at the expense of all higher harmonics.  That are right in the middle of the peak of hearing.

Would be okay in a choke-input filter I suppose.  But again, two chokes, why bother?  Just make the one slightly bigger.

More generally, you can make an elliptic filter which has as many zeroes as poles: rather than the attenuation dropping off asymptotically, the response is shelved, the stopband attenuation is not asymptotic (it's not flat either, there are, well, zeroes in it; but the maximum value reached between zeroes is as specified).

Or more generally still, you can consider a hybrid filter type that has some zeroes -- usually strategically located -- and more poles, so it's still asymptotic but not as aggressively as an all-pole filter, and has better performance given the signal source.

Typical applications are ADC/DAC antialiasing filters, where the zeroes are placed over harmonics, giving better image rejection and somewhat higher bandwidth for the same attenuation.


Note: poles and zeroes are mathematical abstracts, but typical implementation is an LC ladder filter, where each component contributes a pole.  Zeroes are implemented by putting a cap in parallel with an inductor, or an inductor in series with a cap.

Tim

[amyk]

About 0.68uF will be Ok for typical 5H choke  isnt it?

5H

This is what a 5H inductor looks like: http://www.hammondmfg.com/jpeg2/158Q_B.jpg

Not out of place in tube equipment, as discussed above, but absolutely elephantine for normal electronics.

[T3sl4co1l]

If you look at 001's posting history you will see little "normal" about their electronic interests

Tim

[coppercone2]

because its cheaper and the analysis is ugly compared to convincing everyone in 2 seconds with a PSRR graph from a 30 cent part that looks trust worthy because the datasheet has been photocopied 600 times

[001]

If you look at 001's posting history you will see little "normal" about their electronic interests
 

Yea
eevblog is pretty place for oldtimers tales 



I googled an Italian site about it!
 http://www.sugardas.lt/~igoramps/article47.htm

[Ysjoelfir]

I googled an Italian site about it!
 http://www.sugardas.lt/~igoramps/article47.htm

cool! but sadly that isn't italian (which is actually moderately acceptable translated by google translator) but lithuanian or russian (which is absolutely horribly translated by google) and I doubt many of us here are capable of understanding what is written there. do you mind translating the relevant parts?[/b][/size]

since I am interested in "the old techniques" I would be intrigued to learn more about it.

[001]

cool! but sadly that isn't italian (which is actually moderately acceptable translated by google translator) but lithuanian or russian (which is absolutely horribly translated by google) and I doubt many of us here are capable of understanding what is written there. do you mind translating the relevant parts?

I`m using google translate too. It is strange a little to read it but author tune LC to resonance at ripple f and put it to an amp
but no any formulas exept muddy graph


Google say:

The adjustment is made by selecting the capacitor capacitance, connected in parallel with the inductor. Do this in advance, even before installing the throttle on the chassis. The setup technique is very simple. It is necessary to assemble the measuring circuit shown in Fig. 1, connect a sound generator to its input, and an oscilloscope or an AC voltmeter to the output.

If you know (at least approximately) the inductance of the inductor, then to reduce the time it takes to adjust along the curve in Fig. 2 it is necessary to determine the order of the magnitude of the capacitance required for the resonance and start the adjustment of the throttle with such a capacitance.

By connecting a capacitor to the inductor coil and setting the generator output at a frequency of 100 Hz so that the voltmeter needle is closer to the middle of the scale, the generator frequency is changed to a small extent, observing the instrument reading. In one case, the output voltage will decrease, in the other - increase
     Continuing to change the frequency of the generator in the direction of increasing the output voltage, they reach the point at which the voltage becomes maximum. This point corresponds to the actual resonant frequency of the choke-capacitor system. If it turned out to be above 100 Hz, the capacitance of the capacitor should be increased, which can be done by connecting an additional capacitor in parallel with the existing capacitor. If the actual resonant frequency is below 100 Hz, instead of the existing capacitor, you need to put another, smaller capacity.

     Thus, by gradually selecting the capacitance, the inductor is tuned exactly to a frequency of 100 Hz. Note that an inaccuracy of tuning by only 10 Hz reduces the filter efficiency by more than half.

[MagicSmoker]

...author tune LC to resonance at ripple f and put it to an amp but no any formulas exept muddy graph

The formula to find the shunt capacitor of a parallel resonant "trap" when you know the inductor and ripple frequency is just an algebraic rearrangement of the resonance equation:

C_res = 1 / [(2pi)² * f² * L]

Where C is in F, f is in Hz and L is in H.

Note that putting a cap across the choke in a choke-input filter improves the attentuation at the parallel resonant frequency but worsens attenuation for frequencies above it because it turns the LC filter into a capacitive divider, basically. This may sound irrelevant when you are dealing with mains ripple, but rectifiers produce a burst of noise that extend well into the RF region every time they undergo reverse recovery. That's why in a lot of audio(phile/phool) gear you'll see little pF-range capacitors across each bridge diode (technically a resistor should be included to make it proper RC damper...).

As others have already noted using parallel resonant traps isn't common these days, but one place I do see it used quite often is to notch out specific harmonics in 3-phase AC systems - the 5th and 7th seem to be most popular. In these cases it really doesn't matter that the choke (and capacitors) are massive because, well, so is everything else.

[001]

The formula to find the shunt capacitor of a parallel resonant "trap" when you know the inductor and ripple frequency is just an algebraic rearrangement of the resonance equation:

C_res = 1 / [(2pi)² * f² * L]

Where C is in F, f is in Hz and L is in H.

Thanx a lot! Sorry for stupid queston but what about choke DC resistance? Is it critical for Q?

That's why in a lot of audio(phile/phool) gear you'll see little pF-range capacitors across each bridge diode (technically a resistor should be included to make it proper RC damper...).

This is strange too
Someone use an ultrafast diode  but shunt it with cap 

[Berni]

Yep a high Q is needed if you want a really sharp and deep notch. But as others have brought out you may not want that in the first place because inductors with feromagnetic cores will generally loose inductance as higher DC current is passed trough them and this moves the filters center frequency.

As for diodes. Yes capacitors across diodes are a common trick. Keep in mind that fast diodes are only fast for the turning off part. A normal diode turns on almost as fast as a fast diode, but takes much longer to turn off. So you get less power loss by using fast ones, yet you still get some switching noise with normal slow ones. This is where the capacitor comes in. It smooths over the switching point of the diode so that its not as sharp, this causes the switching edge to have less harmonics in it, so you get less of the harder to deal with higher frequency noise propagating on into the circuit. It works, but don't expect to hear a instant improvement in sound by strapping diodes into the bridge of your audio amplifier. Properly designed audio equipment is designed to tolerate some mains noise anyway.

[001]

Punch a few numbers into the calculator and see just how big of a L and C you need for this.

Then afterwards have a look at the ripple rejection graphs in the datasheets of even cheap crappy linear regulators and you will see why.

It is false
Transistor regulators are not a good idea if ripple voltage is about 50V (typical for 1st filter cap at 500V 0.5A)
Because drop voltage at transistor MUST be greater than ripple
So You miss more than 50V and heat 25W with typical transistor linear regulator here
Choke works different way. It is reactive component

[Berni]

If you have 50V of ripple after rectification then the capacitor in your rectifier is too small.

Yes linear regulators are pretty lossy but your big 5H choke also likely has a quite a bit of resistance in its winding. So expect it to get pretty warm too unless its really overbuilt. And overbuilding can get out of hand quite quickly since a 5H inductor that can handle 500mA is going to need a pretty big iron core to begin with.

The 500V 500mA you are talking about is a pretty big supply, so linearly regulating that is of-course expected to need some significant heatsinking. This is why power supplies of these power levels tend to be switchmode designs these days as the transformer alone would be rather big and heavy.

Im not saying 120Hz passive notch filters are are bad idea or anything, just that these days we have other solutions that tend to be a better fit for the job. But it was a very good solution back in the tube days.

[001]

If you have 50V of ripple after rectification then the capacitor in your rectifier is too small.

Hi! it is common value. It is only 10%

I but your big 5H choke also likely has a quite a bit of resistance in its winding.

Typical 0.5A choke is about  20Ohm DC resistance and give only 10V drop  and 5W instead 50V/25W with linear regulator

 

[T3sl4co1l]

I mean, if you're worried about secondary concerns like efficiency, or cost, or size... you'd use a switching supply.  But that's just me...

Tim

[001]

I mean, if you're worried about secondary concerns like efficiency, or cost, or size... you'd use a switching supply.  But that's just me...
 

I`m boatanchor  so don`t worry about  Steel is still cheap

[MagicSmoker]

I`m boatanchor  so don`t worry about  Steel is still cheap

And ferrite + all the other components to use it is even cheaper on a $/W basis (or any other currency).

[coppercone2]

I suspect there is a split between hobbyists that have mechanical/metalworking skills and hobbyist that don't which concerns me. I suspect this split will widen as the use of thinking machines that manufacture boxes increases. People like having things work for them.

I also suspect that getting capabilities to quickly manufacture metal boxes of whatever size you want with welding and stuff is actually pretty cheap, so long you don't fall into the trap of thinking that you will also build a roll cage for a ATV one day and go over kill on things.

I kind of wonder if you can do a rough d-sub actually using just a plasma cutter with a high fidelity nozzle. I will try to make a template. For ones that can do punch through.

[001]

I suspect there is a split between hobbyists that have mechanical/metalworking skills and hobbyist that don't which concerns me. I suspect this split will widen as the use of thinking machines that manufacture boxes increases. People like having things work for them.

I also suspect that getting capabilities to quickly manufacture metal boxes of whatever size you want with welding and stuff is actually pretty cheap, so long you don't fall into the trap of thinking that you will also build a roll cage for a ATV one day and go over kill on things.

I kind of wonder if you can do a rough d-sub actually using just a plasma cutter with a high fidelity nozzle. I will try to make a template. For ones that can do punch through.

It is hard to understand for me. Is it some US wellknown story or idiom? 
This tread is about LC trick with old iron, not philosophy

[coppercone2]

some how my post ended up in this thread not the square holes thread, i had just gotten out of bed

[floobydust]

... This tread is about LC trick...

Did you mean capacitor across LC... or CLC filter?

Just because a component is old or big and ugly, electrons don't discriminate. I have 5H 300mA inductors, mostly as paper weights and one in a power supply.
These have an anti-sat air-gap and possess leakage inductance, noobs miss that on their Spice sims.

In an LC filter, the choke's leakage inductance causes big problems for the rectifiers. Doesn't matter if they are solid-state or 866A mercury vapor beasts.

I have these old scope waveforms of 600VDC PSU LC filter (choke input) and adding a snubber across the 5H inductor.
You can just put your scope probe a few inches away to see the waveform too, if there is switching hash and spikes which shows up with no snubber. If you are doing single-ended, class A amps, they can have noisy power. I wanted to try SiC rectifier diodes as they apparently have no reverse-recovery, but need around 3kV.

You can see adding a capacitor across a choke, mutes the leakage inductance spike but makes the choke ring. Traces are 100V/DIV.