The truth about caps

[R.G.]

I got my 30UF 500V caps from Mouser Electronics. Trusted and used by almost everyone U.S. on this forum. Yet I got a "small cap in a big can" thing. I'm more inclined to go with R.G.'s explaination of the whole thing. They're not "fake" caps, just smaller ones. The bigger can is what the MFG has been selling to their distributors so Sprague just made it work...

I'm kind of torn on this one. All electro cap innards are smaller than the outer can by a notable amount. Cornell Dubilier has a diagram in some of their tech lit about this. The wound cap pellet is stuck into the outer can with either pitch or a mechanical holding arrangement. But there is no internal can nor any internal shrink wrap. That would impede heat flow out of the can, and that's a big consideration in power applications.

The ones with internal heat shrink cans are definitely fakes, IMHO. Just a smaller wound bundle inside is the way they're made properly.

[Steve Conner]

My favorite scheme is to put high current arrays of MOSFETs in the + and - rails, clamping them saturated-on for normal operation and turning them off when the fault detector circuit goes off. I thought this was an obvious application, but I've never seen anyone else do it, even after I posted it on my web site.

I'd guess that's because it does less than current limiting on the output stage, combined with rail fuses and a speaker relay, but costs more.

If either rail fuse blows, the DC offset detector will trip the speaker relay. And a speaker relay gives you free anti-thump.

If you're concerned about contacts in the signal path, you could probably make a speaker relay from back-to-back MOSFETs powered by a photovoltaic isolator, but Douglas Self would have fits at the THD.

[R.G.]

I'd guess that's because it does less than current limiting on the output stage, combined with rail fuses and a speaker relay, but costs more.

Hey! kEwl! A true work of engineering excellence! More bells and whistles and less real substance!

My reasoning is more esoteric - if perhaps more misguided. This is a fire- and speaker-saver. The presumption is that the output devices are already toast or may be assumed to be. Hence, current limit circuits on the outputs won't help. Or one rail fuse has opened by either a close encounter with its rating or metal fatigue.

When one rail fuse blows but not the other, the output voltage goes to the opposite rail and is only current modulated by the remaining output device in the best of cases. In the worst of cases, it sticks at the opposite rail, connecting the speakers DC-wise to the rail.

If either rail fuse blows, the DC offset detector will trip the speaker relay.

At that point, the speaker relay is supposed to open. I've worked with relays. They work great - as long as you don't really, really need them. If they're regularly interrupting anything like their rated current, they also arc, pit, and weld. Like any mechanical device, they need maintenance. I'm not concerned at all about the sonic aspects of another contact in the audio path - but I really do want to stop a transistor failure from becoming an opportunity to replace all my speakers. A good speaker relay needs to handle transient currents up in the 10-20-30A range, depending on what speakers you have attached and what kind of transients happen. If you happen to short the output line (Hey! I can, no problem. It sez right here that the output transistors are current limited...) you can get some interesting transients through those relay contacts. They'll work great the first time, and second, and ...third? fourth?

I guess put another way, I have this almost religious faith that if I quit driving the gate of a MOSFET, it will stop passing current. I believe that more than I believe in relays.

And a speaker relay gives you free anti-thump.

It does do that reliably. In my experience, thump comes from single power supply amps with capacitor outputs, and symmetrical Linn architecture amps are almost thump-free, but there's always some amp that will thump.

If you're concerned about contacts in the signal path, you could probably make a speaker relay from back-to-back MOSFETs powered by a photovoltaic isolator, but Douglas Self would have fits at the THD.

Yeah, I experimented with that, and decided that I could do the same job at the power supply with the same number of MOSFETs and PV isolators, and have the THD cancelled by the amp's feedback and power supply rejection. Actually, that is exactly the point where I started thinking about this. I don't mind the extra contact if thought I could actually rely on it. But I have this mental image of big relays being expensive and welding closed or open.

Practical or not, it does work well, and sidesteps the issues of non-symmetrical fuse blowing and relay issues.

[Merlinb]

Hey! kEwl! A true work of engineering excellence! More bells and whistles and less real substance!

Practical or not, it does work well, and sidesteps the issues of non-symmetrical fuse blowing and relay issues.

I dread to think what Douglas Self would say!
No doubt he would argue that anything standing between the PSU and the amp will usually retard the overall performance, or limit the transient current available, depending on where you put the FET relative to the reservoir caps.

Alternatively we might suggest that the output devices be overated enough that the rail fuses blow before the output devices can fail (or you might say the the rail fuses are slightly underrated for the theoretically available output current). This amps I used to build were designed this way- the fuses would always blow before the output transistors could.

Also, even if one fuse blows, the output offset should immediately force the remaining devices into cut-off, with no harm done to the speaker. Adding a relay would offer just a little extra help.

[R.G.]

I dread to think what Douglas Self would say!
No doubt he would argue that anything standing between the PSU and the amp will usually retard the overall performance, or limit the transient current available, depending on where you put the FET relative to the reservoir caps.

No doubt. Although they do look like small-ohms resistors when they're driven on as much as possible. Most output stages have good rejection of the power rails, although that's not necessarily true of the earlier stages.

Alternatively we might suggest that the output devices be overated enough that the rail fuses blow before the output devices can fail (or you might say the the rail fuses are slightly underrated for the theoretically available output current). This amps I used to build were designed this way- the fuses would always blow before the output transistors could.

That's one way to do it - excepting that bipolars tend to current hog, so you have to work at forcing current sharing or the problem still exists. In my experience, it's at best a statistical matter whether the fuse protects the transistor or vice versa. My presumption is that when you need to turn off the power rails, one or more output devices are already dead.

Also, even if one fuse blows, the output offset should immediately force the remaining devices into cut-off, with no harm done to the speaker.

It should - unless there's a signal going into the amp. Hard to say what happens in all cases.

[Steve Conner]

This is always a big worry in hi-fi amps, that can find themselves hooked up to speakers worth more than your car. There has to be a second line of defence against a failure in the amp itself that sticks the output hard against one rail: a rail fuse can't be relied on to pop in time to save the speaker.

From a safety perspective, you must work on the basis that the transistors always give their lives to save the fuses. The fuses are just there to keep the amp from catching fire once the semiconductors have failed short.

RG's idea of de-thumping the power amp by design, and using MOSFET rail circuit breakers instead of fuses, may well bear some consideration.


My own experiences in this context are:

I once had a 400 watt Class-D amp fail on me. It overheated, then one of the power MOSFETs failed short and stuck the output to the (-80V!) rail. Luckily it was on a dummy load, saving some lucky speaker from complete incineration. I had planned to use this module without a speaker relay because the maker claimed that it was "fully protected".

The only hi-fi amp I ever built was a current-feedback MOSFET design that had the worst turn-on and turn-off thumps ever. Not so much thumps, as vicious cracks and bangs that made me fear for my tweeters.

The design was supposed to be speaker-relay- and rail-fuse-less: the mains was supplied through a latching relay, and the protection circuitry (thermal and DC offset) just unlatched it, turning the whole amp off. There was also SOA protection that muted the audio, and if muting didn't bring the devices back within their SOA, it would fall through to the second stage of protection and again turn the amp off.

To get rid of the thumps, I eventually fitted a speaker relay with gold contacts, salvaged from the power supply of an old TV camera.

[R.G.]

Steve, I'm probably overly affected by my experiences with switching power supplies. We were trying to get these to work reliably back when the only devices which could switch power were high-voltage bipolars that simply were not fast enough and/or did not have enough SOA to do the job. I actually had transistors fail in a way that melted holes in the tops of steel-case TO-3s. The output section of the power supply would have charred and vaporized copper tracks (some of them 0.2" wide) and dead everything, but the "protective" fuse would be intact. So I'm always thinking in terms of how many nanoseconds I have to shut down the current flow before filling the lab with noxious fumes.

[Merlinb]

Back to capacitors...

Getting back on topic, Walt Jung wrote some good articles on cap distortion. You can find them here, about a third of the way down, "Picking capacitors".
Services

Some of it ties in with Self; polar capacitors tend to generate distortion when the AC voltage across them starts to become appreciable.

[Rhodesplyr]

For the longest time I swore by Spragues only,but I found the F&T's live up to my expectations,so I guess there could be others,but for me right now its just those two.I wont use any others.

Just out of curiosity, anyone tried Panasonic FC/FM series or Elna Silmic? Of course, these are often not available in high enough voltage for power amp filter caps, but they would certainly work for cathode bypasses. Those are my current default caps for lower voltage circuits, e.g., FM tube tuners.

[guitician]

Polypropylene and Foil are great, but are pricey.... anyone try CDE 940C30S1K-F .01 3000V $1.5 and can slew 2568 volts a microsecond!

[R.G.]

Polypropylene and Foil are great, but are pricey.... anyone try CDE 940C30S1K-F .01 3000V $1.5 and can slew 2568 volts a microsecond!

Wow! Whatta slew!

Of course, a 1000W amp output at full tilt with a 20kHz sine coming out slews a maximum of about 22V/uS, so a 2568V/uS slew may be a bit of overkill.

Reminds me of the sign: "Danger! 2 Million Ohms!"

[12Bass]

Regarding the OP: The best way to find out is to try it yourself....

Place a high quality polypropylene in parallel across an interstage coupling capacitor and see if you can hear a difference (maybe even solder just one end so you can "switch" it in and out, provided the voltages are safe enough). My experience has been that there is often a perceptible increase in clarity, particularly if the capacitor being bypassed is a large electrolytic. Everything in the circuit makes a difference, down to the solder and traces themselves. The question is whether or not the differences are significant.... and that can be a subjective matter.

[Phostenix]

so a 2568V/uS slew may be a bit of overkill.

I thought overkill was the driving force behind all of this, no?

[Steve Conner]

2568v/us across a 0.01uF cap implies a current of about 25 amps.

[guitician]

CDE's UNL ElectroFilm series has a max. peak current of 2200 A. I bought a couple 30uf/600v back when they were $15 ea. I though it was worth the extra $$ to have a little overkill

Wouldn't you want your audio engineer to have "Golden Ears", instead of just a lot of electronics theory?