capacitor bypassing/cascading......

[prickears]

Not to open a tin of worms here.

Sellers of expensive caps would have me believe that sound quality of a given tweeter can be improved by making up the desired value of C by wiring smaller value capacitors in parallel, (cascading or bypassing is 'their' terminology for this).

Funny thing is, I don't see alot of XO designs with this feature, which kind of rings alarm bells.

Is there any real advantages by doing this or, is it just a way of flogging a few more expensive caps???

Mick.

[martin clark]

I strongly suspect the last. Having tried it (with an open mind) I'd always rather just use one cap of sufficient value/quality.

[speakerdoctor]

North Creek Music Systems has long been a proponent of bypassing. They sell caps for that very purpose and have one of their web pages devoted to the subject.

Bypassing

OTOH, I've worked on a number of the popular brand Snell vintage speakers, Peter Snell developed. He commonly used bundles of NPE's bypassed with a single pico-farad mylar cap in his famous & much sought after Model A's and others. I have pictures to prove it.
The less known Kindel Phantom speaker had a very complex crossover with numerous stacked caps of different types. Also have pics if anyone is interested.

[jeffreybehr]

Bypassing larger caps with smaller-value, higher-quality caps is a well-accepted procedure in hi-end audio. Lots of manufacturers use it; speakers I've owned that had bypass caps in the original crossovers include the Eminent Tech. LFT-8 and -12, Kindel PLS-A and other Kindels, and the truly-5-star-quality Vandersteen 5As.

Conrad-johnson and Audio Research, among dozens of others, bypass coupling caps in their hi-end preamps and poweramps.

It has to be done carefully, but it works very well, generally in the higher frequencies where the large caps are a little 'slow'-sounding. Some object to the process, saying they can hear the addition. Well of course; that's why one does it.

A typical bypassed combo would be a SoniCap Gen.1 'propylene with (in ascending prices) a MultiCap RTX ('styrene), Mundorf Silver/Oil, or SoniCap Platinum (Teflon-film) bypass.

[jcx]

keeping esr low so that if needed by the XO alignment it can be defined by intentional, accurately controlled external R is good design practice

if you can't get low enough esr in a single cap then paralleling n equal 1/n value caps can help

the question of "how accurate" is acceptable can be estimated from XO, driver circuit properties and ABX detectable frequency response thresholds:
ABX Amplitude vs. Frequency Matching Criteria

the numbers suggest nearly all tiny "fast" film/mica/whatever audiophile flavor of the month bypass cap of only a few % of the main cap value will give inaudible difference in DBT ABX testing

naive subjective "just listen" evaluation is worthless - Blinding, level matching are absolute minimum controls for valid subjective evaluation - short switching times gives better resolution - and don't move that box or anything big (like your listening position) around the room or that alone will change frequency response by clearly audible frequency response differences

adding enough C will change XO response and at some point it will become audible - but you could also change response by fiddling with the inductors - why isn't that a common tweak suggestion?

driver type, quality, unit-to-unit variation, box diffraction, orientation, room treatment are all going to give much larger measurable and audible differences and attention to these should repay your tweaking efforts with much higher "dividends"

not to say that proper XO design, implementation and trimming aren’t equally important - just that the "I'll add a audiophile approved bypass" to the caps is poor a paradigm

[speakerdoctor]

The D.F. of a Bennic 150 uF 150 V NPE was cut in half with the addition of a 0.22 uF 50 V Vitamin Q bypass cap (i.e. 4.3% > 1.95%).

[martin clark]

Jcx - thanks for typing-out most of what came to my mind!

Apart from value drift, ESR and similar effects, another problem is this: if you have some [less than perfect] cap and add a 'high quality' bypass or two - just what are you hearing ? (apart from change in total value = crossover summation drift):

- No bypass cap can make up for losses due to dialectric absorption in crap caps, mechanical Q losses, or other similar problems. It becomes a frequency-dependant loss instead. You now have two caps in parallel with differing losses, that's all.
- Adding low-esr/low-ESL/low-DA low-value film caps may well add a small peak due to basic shunt L-C resonance. Yes really: given the large loop areas involved in typical crossover wiring, such resonances from C and wiring and layout can fall well within the audible bandwidth, esp. for larger value (eg midrange-xover) caps in 3-way crossovers. It always seems to occur at the top of the presence band by measurement - 'more detail' ? Uh-huh... BT,DT, learned a lot.
- Just how much time and money are you prepared to invest in dicking-around on this stuff before buying just one good polyprop cap anyway?

[carlmart]

Bypassing caps is a controversial issue. In fact capacitors are a controversial issue.

There are some serious articles on this issue, like the one by Jung-Marsh. But there's also many people stating their opinion (and their taste) as if it was "the truth".

If you believe the lower the Dissipation Factor (DF), the better the sound (I do believe so), you should try to lower it bypassing the larger caps with smaller ones.

Now we have relatively affordable film caps available, particularly polypropylene, but you can also get bipolar electrolytics, designed for speaker use, that do seem to sound good.

So, try this: get some smaller film caps and alligator clip them to your larger capacitors. And listen if there's any change, for better or worst. Keep trying until there's a combination that you like. This routine generally pays off and it's what serious pros do. It takes time though.

[speakerdoctor]

[QUOTE=carlmart;2571925]Bypassing caps is a controversial issue. In fact capacitors are a controversial issue.

There are some serious articles on this issue, like the one by Jung-Marsh. .............

QUOTE]

This one's even more serious...

DIYCore.com - Home

[jcx]

the worst number I see on Bennic's site for "low loss" BiPolar Electros is 20% DF at 10 KHz

|Xc| = 1/(150u*2*pi*10KHz) =~ 100 mOhms

20% DF gives 20 mOhm ESR

1/(2*pi*RC) = 1/(2*pi*20m*0.22u) = 34 MHz

that is the frequency where 0.22uF "cuts DF in half"

even 100x worse DF/esr numbers won't have audio frequency effect


as mentioned the parasitic LC resonance from the parallel connection will occur lower than that


for even 4 Ohm drivers 20 mOhm is 0.5 % for ~= 0.05 dB response difference so the "high"/unbypassed ESR is not itself expected to have any audible audio frequency consequence