Supply capacitance: how much is too much?

First post intro:
The name's Dallas and I'm a multi-disciplined engineer in Denver, Colorado. I can handle a solder iron without killing surface mounted bits, build most things out of wood, and I beat/bend/cut red hot iron into hinges and such for sport and to annoy my neighbors. Electronics isn't a total mystery but kits are the way to go for me. I actually learn more and get frustrated less.

In short:
1) What is the safe limit on supply capacitance when considering the equipment only?
2) Do coupling capacitors below 0.47uF really need to be matched for human ears even when there is significant low-end signal as well as driver support?

Discussion:
While daydreaming on the final far-future enclosure design for a near-future build, it occurred to me that there could be such a thing as too much supply capacitance. Could this be true?

(By the way, Bob, I'm going to be ordering a ST-70 kit with EL34s and timing relay from you tomorrow. Yeah, that kind of near future build.)

The reason this occurred to me is that I am staring at a rough couple hundred variable frequency drives - some are years old and others new - that are available for me to harvest parts from. I could gather six huge electrolytics in less than fifteen minutes. And, although a six-pack of high-voltage 5700uFs would be visually appealing to the geek in me, I'm not so sure I would want them all collapsing at once into my near-future rig. (Bleed-offs would help.) What would be the point at which there is no longer a benefit in an increased volume of potential? To be clear, I'm not charging the phasor banks on the Enterprise.

And now, a few application notes:
This amp will be driving a pair of (properly refurbished) McIntosh ML-1Cs. I have already finished a did-it-myself version of the MQ-102 (infringement?) to go between a GoodWill-found-and-repaired Rotel 1068 and the future ST-70. (That is, there will be a +13 or perhaps +17dB boost at the far bottom end <200Hz to compensate for the stunningly overdamped woofer.) The end-game is an ST-70-based integrated amplifier with auxillary inputs to an S/PDIF & USB DAC (thanks, AMB); properly scaled mA meter with output [TUBE BIAS] circuit selector switch for biasing; and retro-fitted and re-scaled wattmeters (from an old realistic APM-200 watt meter). It will all be stuffed inside a yet-to-be designed (easy-to-service?) hardwood case paneled with easy-to-read machine etched stainless plates. (Hey, dream big, right?)
  That written, a six-pack of beer can -sized polished aluminum capacitors would fit right in... somewhere... I'm certain of it. Unless, I could blow up my living room turning it off.

'Big'(-ish) DC filter capacitors:
This morning I de-soldered four 0.47uF PIO West-Caps from a board and measured them out. They range from 0.465uF to 0.502uF. Am I correct in assuming that these would all make fine coupling capacitors even though their specific values are not 'matched'? It seems the knee frequency would be beyond even the 16-ish Hz my chest could feel at a foot from the ML-1C's woofer at +17dB low gain and pushing an ear-crushing 20watts? (89dB/w/m speakers.)

p.s. thoughts: maybe a pre-amped +17dB at that low of an end might be an issue but I don't know quite yet why... seriously. Time to visit Roger Russel's website for a bit... 35w/channel? I don't really crank the volume, though. 17'x19' room with 8' ceiling, brick/windows/bookshelves on three sides, concrete floor with huge rug and basic furniture, speakers 10' apart centered on the wide side and 3' high stands... anyone else ever get pigeonholed as AADD? Time to bust out the SPL meter and the Adcom 555ii for testing. Oh, and I better put the APM-200 back together; one needs proper data...

Download the datasheet for your favorite rectifier to find Cmax (~60µF max on pin #8 for the typical 5AR4).

My take-away:
-Datasheets can tell the curious soul just how much supply capacitance a rectifier can handle. Since most folks don't like to replace their rectifiers needlessly, these ought be observed.
-Decoupling caps larger than 0.33uF-ish are most likely wasted on human ears and [might] compromise the stability of the output signal circuit.

j beede,
Excellent pointer I didn't even think of: read the datasheet!

Bob,
Thank you for the insight. Though the visual effect to the uneducated eye would be impressive, a six-pack of beefy caps clearly wouldn't make for a healthier circuit. Concerning the 0.47uF caps, I just ran across those and thought it could save a few bucks and provide an unobservable sonic bonus. Considering their age and such, I'm inclined to stick with what you ship with the kit for now.
However, that said, I would be interested in experimenting between values and types. I am thinking something along the lines of breadboarding a few groups of four matched capacitors (0.22 films, 0.33 PIO, 0.33 films, and 0.11-ish) with short harnesses. Then I would pull a complementary harness out from under the board and keep the whole package as short as possible. I could swap around and develop my own subjective opinion. I would assume there would be some very minor reactance differences between a harness-to-test platform, but I do enjoy finding odd ways to pass the time.


(In the end, though, my hearing test results indicate anything below 42Hz is lost on me. After that I can only feel through my skeleton. Yes, hearing tests are one way to pick out headphones: find complementary frequency response profiles!)

With tubes, a lot depends on the rectifier used as well. The newer ones aren't as strong as the classics, and the ST amps push a 5AR4 pretty hard as well. A solid state WZ68 copper cap can handle twice the load gracefully. I like big bottles, so I use a Mullard GZ37 in my ST120 - much stronger tube that can handle a lot more current.

I try to go by the +30% max rule when reworking the power rails as it can be difficult to find caps in the original values. Recent project substituted 12000uf @100v for the spec 10000uf @60v/ Voltage isn't as important ... determining factor there is usually size. I'd think that was also a big part of why certain values were selected in the factory back in the day ... they wanted the most bang they could get yet still be able to fit their rigs in a standard rack.

Filters and signal, I try to get as close to spec as possible, even sometimes running a bit of math and using parallel components if there's no other option.

Nothing that says you still can't use your ginormous caps in a build for the bling ... just don't hook em up!

True, true... But I could hook them up to an external ground-isolated cage as a theft deterant.
Yeah, feel *that* music...


Nah, I have a kid and he's the kind that would dare his friends to touch Dad's stereo-of-pain.

Those are not "capacitors". They are INCAPACITATORS. {bah dum BAH}

Energy storage in a capacitor is a function of capacitance times the square of the voltage in the formula  E = 1/2CV^2.   So filtering in tube amps appears whimpy just based upon capacitance alone.  Don't let that deceive you.  As an example a 1,000uF cap charged to 100V holds the same energy of a 40uF cap charged to 500V.  And be safe.  These little critters can kill!

And I try to avoid using too much capacitance when used as stage couplers.  Too much capacitance can throw off bias if DC offsets are present in the audio signal that then slowly settles back to equilibrium..  This can cause distortion and tube cutoff if severe.  So use just enough capacitance for good low frequencies in the passband like down to 10Hz-15Hz but avoid -3dB cutoffs at 1Hz or lower.  After putting a calculator to it the values selected for the ST-70 and ST-120 low gain board (0.1uF and 0.15uF) appear optimum - not too little, not too much.

Back in the 1980s it was popular to bandwidth limit Dynaco tube amps to improve their performance in the audible range. Think of it this way... just because you can't hear it doesn't mean the amplifier isn't amplifying it.

I'm with arle* on this one. Bigger is not always better wrt coupling cap size.

Yup, I'm sold. Anything below my hearing range is wasted power anyway.

Me being me, I am planning out a test platform. The final evaluation will involve a number (4-5-ish) of capacitor variations along a breadboard with leads running to the coupling capacitor connections on the VTA and a switching station between. The switch will be operated by a very supportive seven-year old. I'm not sure what factor of a blind test that will count for. Regardless, it will be interesting. Once I'm ready with results, I'll pick this thread up again with description, photos, results, and discussion. The question really isn't about what is great for everyone, but what makes my listening most pleasurable given: lossless media (which I'll list later) optical S/PDIF -> Rotel 1068 -> MQ-102 -> Latino ST-70 -> ML-1C loudspeakers.
That is the rough-out. This may take months before I post the full sha'bangs. But I'm going to do it for me anyway, so why not post it.
Until then...