@beetlemania
re: " Curious if you have an opinion or knowledge about how to pair "main caps" if the full capacitance is not available in a single cap. Is it best to optimize balance between caps (eg, 7+7=14) or is there some degree of leeway (eg, 10+3.9+0.1)?"
My opinion is based on Thiel experience and related observations - both approaches are valid, but will produce slightly different sonic signatures. I believe the cascading bypass format compensates for deficiencies in the large cap which produces dielectric cycle anomalies (charge lag, erratic discharge, Effective Series Resistance irregularities relative to power levels, etc.), along with possible smoothing of the signal in general. A small cap is inherently less reactive, plus a higher quality small cap is more affordable considering budget constraints. As you have previously mentioned, Thiel’s 1uF bypass cap might sometimes introduce less-than-best characteristic balance when bypassing at greater than 10% of the base cap value. We arrived at our format because we developed that great 1uF tin foil / styrene film cap for the CS5, bought 6-figures of them, and put them to work wherever they made an improvement (that we could afford.) Our main (base) cap was a single cap unless its value exceeded 100uF, where the deleterious secondary effects broke through. In those cases, we used single value multiples. You’ll see ganged 100uF caps (often electrolytic) bypassed by a good polypropylene, bypassed by the great (yellow) 1uF tin/styrene. 100uF was the maximum cap value we identified from listening, and same-value was identified as more correct than cascaded values.
Note this scheme runs contrary to common practice. Note the CS2.7 (outsource engineered after Jim’s death) uses a 400uF lytic in the midrange feed (400E/15PP/1S) and other large, cascaded lytics in that shunt circuit (330E/220E/15PP.) That scheme differs from Jim’s. Note that people who prefer the 2.7 sometimes note they like the "smoothness" or "refinement" of that model which may be attributable to this cascaded approach. Note also, that Vandersteen employs such a cascaded-value, multi type approach to cap bundles. I have wondered if that difference might be a significant contributor to the obvious character difference between Thiel and VdS. The two brands share very similar philosophies and solutions down to wire configuration, but exhibit very different personalities. Jim’s engineering approach was always to identify the solution that addressed the hard-core principles most directly, optimize the particulars and confirm by ear that the most accurate solution had been found. Little to no slack was cut to make harsh recordings sound smoother or mitigate other signal chain problems.
In summary, I would choose the 7+7 option as truer to Jim’s approach. There is, of course, leeway and your other string presents the option of using a CMR (PUR+) as the 0.1uF value without breaking the bank.
If you climb back into your 2.4s and try these options, let us know what you learn. No end to the fun.
