bypassing signal capacitors

[neverlistentome]

Can anyone help explain how jacking a hi-quality poly cap across a cheap electrolitic exactly works? I realize the theory is that poly-caps handle the hi-frequency material better and there-fore "take the burden" off the larger value electrolitic chosen to pass more low end, but doesn't the cheap electrolitic still pass the distorted hi-end material?

Does this arrangement cause any weird accidental cross-over effects? Is there any guide line for selecting component values when mixing these two caps together?

Also I've heard of bypassing rectifier diodes in the power supply with .1 uF poly caps to cut down on switching noise...does this make any sense? If this does should the caps strap across the actual diodes of a bridge rectifier or should they go from the + and - ouputs of the rectifier (I guess just like a first filter stage?)

 

[gyraf]

explain how jacking a hi-quality poly cap across a cheap electrolitic exactly works

Poly's have MUCH lower ESR at high frequencies.

bypassing rectifier diodes in the power supply with .1 uF poly caps to cut down on switching noise...does this make any sense

It does make sense sometimes, as the diodes tends to make a switching noise spike when they go in and out of conduction.

Jakob E.

 

[Samuel Groner]

Poly's have MUCH lower ESR at high frequencies.

Did anyone ever measure the effect on ESR? I very much suspect that the usual 100 nF (or even 1 uF) bypass cap doesn't change anything within the audio band--at least with low-ESR 'lytics.

Samuel

 

[Svart]

Quote:
bypassing rectifier diodes in the power supply with .1 uF poly caps to cut down on switching noise...does this make any sense


It does make sense sometimes, as the diodes tends to make a switching noise spike when they go in and out of conduction.



Yes, this is common for consumer devices that have failed FCC and/or have some issue with the noise coupling to other parts of the circuit. We discussed this in another thread some time ago but I don't remember which one. the caps across the diodes are little more than a band-aid to the problem and generally shift the noise into another spectrum rather than "fixing" it. Of course this is all dependant on the diodes, caps and circuit themselves.

but I'm sure Bcarso will be along to school us shortly.

 

[BradAvenson]

[quote author="Svart"]

Yes, this is common for consumer devices that have failed FCC and/or have some issue with the noise coupling to other parts of the circuit. We discussed this in another thread some time ago but I don't remember which one. the caps across the diodes are little more than a band-aid to the problem and generally shift the noise into another spectrum rather than "fixing" it. Of course this is all dependant on the diodes, caps and circuit themselves.

but I'm sure Bcarso will be along to school us shortly.[/quote]

I'm not that brad, but here is a good bit on snubber design:
http://www.hagtech.com/pdf/snubber.pdf

 

[bcarso]

That's a good discussion for the most part by Hagerman.

I take issue with his direct attribution of high frequency energy as due to the nonlinearity of the junction capacitance. The real culprit is the precise way in which the stored charge following forward conduction is discharged and how much of it there is. That's related to the junction C no doubt, but he moves fast here (as a philosophy prof once remarked on a paper I got a B- on). This leads to the classification of diodes into fast vs. slow recovery and soft vs. unspecified recovery (nobody says "hard recovery" that I've heard of, other than in connection with economic debacles and maybe erectile dysfunction).

Normal rectification with a perfect diode from a sinusoid-driven transformer with leakage L will still generate some ringing. In addition to the transformer leakage inductance-based ringing, the stuff that really fouls things up is the very high frequency energy from the real diode reverse-recovery events, which radiates like crazy and gets rectified by nonlinear circuits elsewhere, which then produce audio artifacts based on the envelope of this VHF-UHF energy. The bypassing can be tuned to lower this local ringing, which Hagerman points out then radiates less effectively, although even a 0.1 uF disc may have too much inductance to be effective. As well as damping R's, sometimes ferrite beads can help a lot too.

So Hagerman's analysis is good for the lower frequency stuff, and that may well be most of the problem in some situations. When you get into the GHz range the whole system starts to require a distributed circuit model and the maths get really messy. Also, the code packages that can deal with it get really expensive, so cut-and-try may be the best policy for DIY.

The other line of attack is to make the rest of the circuit more immune to the rectification effects. Go for small loop areas, ground planes, some teeny C local bypassing of P-N junctions (of the bipolars especially), and sometimes the use of FETs in lieu of bipolars.

 

[franklinh]

Does this arrangement cause any weird accidental cross-over effects? Is there any guide line for selecting component values when mixing these two caps together?

As far as bypassing capacitors with capacitors in the signal path, the above is my question of the day also. I feel like I'm at the same spot with my understanding of it.

Trial and error with values, composition and context yields best results is my guess - but before you can do that there's theory/math we need to know, then the circuit as a whole to consider. I'll bet there is only the most vague guideline and the difference between "eh" and "YEAH" comes down to hard work and sometimes happy accidents.

Semi-pertinent page from the meta:
http://www.groupdiy.com/index.php?topic=11376

 

[kkrafs]

[quote author="bcarso"]That's a good discussion for the most part by Hagerman.

I take issue with his direct attribution of high frequency energy as due to the nonlinearity of the junction capacitance. The real culprit is the precise way in which the stored charge following forward conduction is discharged and how much of it there is. That's related to the junction C no doubt, but he moves fast here (as a philosophy prof once remarked on a paper I got a B- on). This leads to the classification of diodes into fast vs. slow recovery and soft vs. unspecified recovery (nobody says "hard recovery" that I've heard of, other than in connection with economic debacles and maybe erectile dysfunction).

Normal rectification with a perfect diode from a sinusoid-driven transformer with leakage L will still generate some ringing. In addition to the transformer leakage inductance-based ringing, the stuff that really fouls things up is the very high frequency energy from the real diode reverse-recovery events, which radiates like crazy and gets rectified by nonlinear circuits elsewhere, which then produce audio artifacts based on the envelope of this VHF-UHF energy. The bypassing can be tuned to lower this local ringing, which Hagerman points out then radiates less effectively, although even a 0.1 uF disc may have too much inductance to be effective. As well as damping R's, sometimes ferrite beads can help a lot too.

So Hagerman's analysis is good for the lower frequency stuff, and that may well be most of the problem in some situations. When you get into the GHz range the whole system starts to require a distributed circuit model and the maths get really messy. Also, the code packages that can deal with it get really expensive, so cut-and-try may be the best policy for DIY.

The other line of attack is to make the rest of the circuit more immune to the rectification effects. Go for small loop areas, ground planes, some teeny C local bypassing of P-N junctions especially, and sometimes the use of FETs in lieu of bipolars.[/quote]

Well said! And interesting info! :thumb:

kkrafs

 

[Svart]

Yes I always look forward to Bcarso coming in. I always learn something new from him, usually with much more ease than trying to find the answers in books or online.

I don't think there is a *real* relationship between the larger, higher esr cap and the smaller, lower esr cap. there are too many variables to include such as foil thickness, material, dialectric material and properties, value, etc.

I have the mind to believe that a cap with a very high ESR reading and a small film cap with a very low ESR reading could theoretically form a band-cut filter but this would be an extreme case for sure.

I think there was a paper on this and the outcome was that a *good* electrolytic cap was sufficient enough for all audio frequencies and poor performing caps suffered from designs that were not meant for audio coupling at all. I doubt anyone would use a Black Gate for powersupply decoupling either.

My own tests confirm this, a "bad" cap either being out of spec or just a pisspoor cheapy usually sounded bad. other than sounding bad, they usually work pretty poorly as decouplers, snubbers, smoothers and so forth. A quality cap usually performed well in all instances. for instance, a Nichicon HE sounds slightly different than their PW series when you have 20 of them in an audio chain but neither sound *bad* at all. 20 cheap caps sounded like I was listening through a towel.

I guess it all boils down to money. If you plan on recapping, buy good caps and you shouldn't need the bypassing caps, BUT it won't hurt to stick them in if you have the will/money/time to do so.

 

[PRR]

DIY electrolytics

 

[Svart]

:thumb:

 

[bcarso]

EGADS cyanide capacitors!

I see a Columbo episode:

Oh...sir...I'm sorry sir. Just one more question sir, a thing that is still bothering me. You said your brother was an electronics do-it-youselfer? And he was doing some experiments on making...what did you call them, sir...lemme look at my notes here...ah here it is, "capacitors"?

 

[PRR]

> EGADS cyanide capacitors!

But look at the last electrolyte. Borax. $1/ton, fireproofs cloth, kills cockroaches, and gives a 500V film. It appears that a standard 8uFd 450V electrolytic can be made with 500 square inches of food-foil and a spoon of borax from the cleaning closet.

The suggested bucket of plates follows standard lead-battery practice. They do it for batteries because the lead-crud has to fall off the plates. The cap film is much thinner and should not fall off: we can take some paper towel and wind-up the food-foil in a little can.

The tough part is insulating the lead-in. Where the wires come out of the liquid, the film stops and there is some/much electrical leakage, especially if you slosh the liquid. Sealed totally-filled caps are the answer, but a bit tricky.

I don't really think you can toss-up a better cap than Panasonic sells. If nothing else, cap shelf-life has increased ten-fold since 1926, because well-paid chemists worked years to refine the formula: less impurities, and specific buffering agents. But hey, DIY is not always about Best Practices.

The carbon rod inside a battery is a low-R resistor.....

> I see a Columbo episode:

"It will take a while to explain capacitors. While you are listening, why don't you try some of this castor-bean soup?"

 

[CJ]

Cyanide, PCB's, it's all good.

Some people swear by those darn PCB oil filled cans for single ended audio.

"Honey, is that a new perfume your wearing?
"Why yes, dear, it's PCB Pitchouli, but actually, I think you are running your B+ a bit high."

Rock over London.
Rock on Chicago.
Dupont. Better living thru chemistry. And PCB's.

 

[PRR]

Walt Jung has several essays on capacitors.

http://waltjung.org/Classic_Articles.html

The 1985 article is new to me, and thought-provoking. A simple obvious circuit shows astonishing differences between caps.

 

[Svart]

yes some circuits can sound completely different with different caps, but I have also run across circuits that sound almost the same between caps, of course meaning *quality* caps in all instances.

it would be interesting to investigate what makes a circuit affected by it's coupling caps in a tonality way, not a LPF/HPF way that we commonly see.

 

[Gus]

One of the older mags Speaker builder? had an article with math and measurements of electro caps in loudspeaker crossovers. It did show measurements of electroa in crossovers IIRC.

 

[CJ]

Cool articles. What is the percent residue refer to as far as the scope shots? I do not know how they are measuring the percentage from the pic.