Parallel filter caps & unwanted resonance?

[mcgruff]

I'm building a kind of generic power amp which will be mated with a range of separate preamps.

I need to provide some switchable options eg power supply filter caps.



You could add one of every cap value you will need to support and just activate one at a time. However, it's much more flexible to choose values which can be summed together in different ways to cover a wider range of capacitances. There may be a snag though: I've been hearing that caps in parallel can create problems with unwanted resonances - some kind of effect caused by the inductance/capacitance and ESR of parallel caps.

Anyone know anything about that?

[R.G.]

There may be a snag though: I've been hearing that caps in parallel can create problems with unwanted resonances - some kind of effect caused by the inductance/capacitance and ESR of parallel caps.

Anyone know anything about that?

Yeah, some.

My question is WHERE and FROM WHOM have you heard this nonsense?

It is absolutely true that every electro cap - well, EVERY cap, for that matter - has internal inductance, series resistance, and parallel/leakage resistance. All of them. Even ONE cap has this, so each cap could in theory resonate with itself. And in fact does - if you plot the magnitude of a cap's impedance versus frequency, then from low frequencies toward high frequencies the impedance steadily drops as theory says, then it flattens, reaches a minimum, and starts rising again. That is the "resonance" of the cap, where it's inductive and capacitive impedances are equal.

The obvious question is then why don't ALL caps have resonance issues? The answer is that they're never used under conditions that would let this resonance raise its head. As a practical matter, power supply caps are damped by the load and the internal ESR and EPR, and don't ring noticeably.

Adding another cap in parallel adds another paralleled cap/inductor/resistors, but in general those have the same limitations as one cap by itself.

It is theoretically possible to get some minor ringing, I guess, but I've been using paralleled caps in power supplies for over 40 years, including designing power supplies for a major computer company, and this has never reared its head. Frankly, it sounds like advertising gimicks to sell special caps. Either that or "internet wisdom" heavily supported by the Dunning-Kruger effect.

If you would, please post a reference. I'd love to have this debate with the source of that "information" and see their data showing it's an issue and how they found it.

Shoot, it's even possible they found one of the rare situations. People actually have been struck by meteorites, too.

[exclamationmark]

This (IMO) is generally a concern in high speed electronic designs. The impedance characteristic of most capacitors forms a 'V' shape when graphed with frequency on the X-axis and impedance on the Y-axis. This is explained by the fact that the capacitor is acting as an ideal capacitor for most frequencies, until the parasitic inductances/whatever else begins to take over and the impedance rises again. To provide adequate decoupling for some circuits, it's sometimes advised to either use a specific size of capacitor (because the decoupling characteristics are strongly related to both physical size/construction and capacitance), or to use a combination of parallel capacitors to provide a wider decoupling range (pretty much the same idea as bypassing main filter caps with smaller film caps). I've heard stories of parallel capacitors not playing nice, but that might just be a case of aggravating the underlying symptom, rather than being the actual cause.

I personally haven't payed attention to audible differences between parallel caps vs bulk capacitance, but I wouldn't think it would matter too much. I use a similar scheme for fiddling with the coupling caps in some of my pre-amps. The only thing I can think of is the lifetime of the switches probably won't be too great - the current surge the switch will see is sort of like shorting a fully charged capacitor to ground, as the fully charged capacitors attempt to equalize charge with the newly switched in capacitor(s). You could move the switches around to avoid this.

[nickb]

I think this might be one source of this idea Bypass Capacitor Boobytrap - John Dunn, Consultant, Ambertec, P.E., P.C. - IEEE Consultants' Network of Long Island

IMHO, this is really only of a concern over a few hundred MHz and would be seen rarely. The author uses a lumped model for the PCB that is justifiably questioned at the frequencies of interest too.

[R.G.]

Yeah, I'd agree with Dunn about the issues - with changing the inductance of a 50pF cap out at over 100MHz and just under a GHz. A length of PCB trace can easily enough be a distributed resonator out there. But the idea that paralleling electro caps for power supplies for tube amps will experience this stuff is ... well, let's just say that it's dubious at the very, very best. All electros are well over into almost purely inductive at 100MHz and above. Like I said - it can happen, but the situations are rare, and not something to even worry about with the stated situation.

[Enzo]

Context is everything.

When the old SSTs were flying, and to a greater extent the SR-71, they went so fast that the air friction on the skin of the plane heated them considerably. The whole fuselage would expand and contract. The design had to allow for this. So we know that air friction causes heating when something moves through it.

I imagine when your car drives down the highway, there is air friction, and I bet it could be calculated just what tiny fraction of a degree that might be if you had lab equipment sensitive enough. But when I get in my car, I will not be worrying about it.


You show a bleeder resistor, but it will only discharge the caps that are "turned on" at the time.

[Dave H]

You could put a resistor (100k?) across each switch to keep the unused caps charged up to extend the life of the switches. It would also ensure that all the caps discharge through the bleeder resistor.

[Gingertube]

The resonance thing came from various discussions in the Tube HiFi Forums. They were talking about adding a film cap ( say 100 to 470 nF) across the main B+ electrolytic. Not adding parallel electrolytics.

Years ago electrolytics were so bad that their self resonant frequency could be down in the audio band, above this frequency they look/act inductive. There was some concern about that inductance resonating with the added film bypass cap. The solution was to put a 2R2 resistor in series with the added film bypass cap to damp any resonance (a solution looking for a problem, if you like).

Electrolytics have come a long way since those days (largely due to demands of switch mode power supplies) and so the benfits of adding that additional film bypass are now debatable. I use them on the main B+ node and on the first preamp tube B+ node, and have never found it necessary to put the series damping resistor in. Worse, I use parallel electrolytics for the main B+ node, I buy 22uF/450V Panasonic ED by the bagfull and use 2 , 3, 4 or 5 in parallel to taste and then add then add a 470nF/630V film cap too.

Never had a problem.

EDIT: This was to do with ringing in response to diode switching noise spikes. If using Ultrafast Soft Recovery diodes then that would mitigate any problem too.
Cheers,
Ian

[mcgruff]

Thanks for the replies. Good to know it won't be a problem.

You could put a resistor (100k?) across each switch to keep the unused caps charged up to extend the life of the switches. It would also ensure that all the caps discharge through the bleeder resistor.

Thanks - that looks like a great idea. I knew I needed to look at discharging but hadn't got to that yet.

I'm building a kind of generic power amp which will allow me to recreate a bunch of 15-watters. There will be an option of a ss rectifier and a 5AR4 tube. The full 79u with all caps engaged would only be used with the ss rectifier to replicate a Goldtone GA15. However, according to valvewizard, this is too much for a 5AR4 (too much ripple current) - but now I'd have 100k resistors "shorting" out all the switches. Am I right in thinking the large value suggested means that isn't a problem? Is bigger better here or is 100k already plenty?

[R.G.]

I'm building a kind of generic power amp which will allow me to recreate a bunch of 15-watters. There will be an option of a ss rectifier and a 5AR4 tube. The full 79u with all caps engaged would only be used with the ss rectifier to replicate a Goldtone GA15. However, according to valvewizard, this is too much for a 5AR4 (too much ripple current) - but now I'd have 100k resistors "shorting" out all the switches. Am I right in thinking the large value suggested means that isn't a problem? Is bigger better here or is 100k already plenty?

The problem with the 5AR4 and all other tube rectifiers is that there is a fixed amount of current they can emit from their cathodes, and the higher the first filter cap, the higher the pulse currents that charge the cap. This is why tube rectifiers oftes specify a maximum filter cap.

If I may make a suggestion, it is really fairly simple to approximate a tube rectifier with solid state recitifers and some series resistance. There are people who swear they can hear the difference between a real tube rectifier and a solid state one designed to approximate it, but I would dearly love to have these people do a double blind ABX test to see if they really can.

The Weber Copper Cap rectifier modules get good reviews from people who have used them, and they are SS diodes plus other parts to fake various tube rectifier characteristics.

There is no practical upper capacitance limit for SS diodes because they have such high pulse current tolerances - often hundreds of amperes. So if you were to think and experiment to see if you could get your ears around a Copper Cap (or other) approximation to a 5AR4, you might not need to sweat the capacitance question.

Just a thought.

After all, you can't really hear a vacuum - can you?

[mcgruff]

It will be interesting to find out. This is the project I'm working on. I'm trying to create a bunch of 15W amps with one set of transformers. At least to begin with, I'm keen to do this as accurately as I can.

A tube rectifier does create a voltage drop so that will have an effect on the circuit. AFAIK the resistance also contributes to sag and so might influence the way a loud chord "blooms". The 5AR4 isn't so bad but the 5Y3 has quite a large voltage drop.