This review page is supported in part by the sponsors whose ad banners are displayed below
Good dog. Off its leash here's what you get: two boards connected by ribbon cable, an Alps 206 10KAX2 pot on a shaft. In capacitor city we net 4 x MCap Z-foil 4.7μF/100VDC units, 4 x 1000μF/50V versions and 16 x 470μF/35V. For AC/DC converters there are two Mornsun LH10-10A24 to transform incoming 85~264VAC at 50/60Hz into ±24V/200mA out; and one Mornsun LD05-20B12 for a 12V/420mA out. We note the fully encapsulated module in its aluminum casing and wonder what's inside.
But we clearly make out two complimentary pairs of transistors. Those are Fairchild Semi issue and enhancement mode Mosfets called QFet "with proprietary
planar stripe and DMOS technology". Here the actual parts are FQP 65N06 and 27P06. There's another 2+2 vertical grouping of them right behind the Mornsun canisters. From the Zetex catalogue of Diodes Inc. and in ultra-compact SOT223 form factors grouped around the Alps pot come four complimentary pairs of NPN/PNP 75V silicon-planar high-voltage transistors. Their specific parts numbers are FZT 1054A and 792A.
Except for assorted micro SMD and certain unidentifiable bits, that's essentially it. Time for Roland to talk secret sauce. To the onlooker his circuit would seem to go to great lengths with the filtration, regulation and stabilization of power. And that would be the big loud credo for high-end electronics where "the power supply is everything" (or at least 90%). Once you reconsider the job at hand—driving headphones—the CHA-1 seems chock full of parts since a TI headphone driver IC could do the job. For these boards the chassis is simply oversized. That makes it look a bit empty until you do the math. The obvious thing missing is the power transformer we'd expect and get with a linear supply.
Jotting down parts numbers and googling their identities is one thing, knowing what they accomplish at each circuit juncture quite another. Here's what Ronald Krammer was comfortable sharing.
The discrete high-bandwidth linear voltage regulator modules are highlighted in red; the servo DC output regulators are green; the cross regulators are yellow.
"The Fairchild Mosfets aren't used as output transistors. They work as linear voltage regulators for the encapsulated output stage.
"The special benefit of this arrangement is that the regulator's <0.2Ω impedance doesn't rise above 1kHz. It remains constant up to 1.5MHz. My experience with high-speed circuit design is that nothing's worse for the signal path than varying Ω. The actual value (0.01Ω or 0.1Ω) isn't essential. Linearity across the broadest bandwidth is. Whilst showing a still useful value at 20kHz, very good integrated regulators like the popular LM318 do take off at 1kHz and at 100kHz already exceed 1Ω. Lesser parts will hit 10Ω instead. See the right graph and compare it to ours below.
"In my opinion the impedance linearity of your voltage regulation is inversely proportionate to the quality of your supply voltage. On sonic impact load regulation is actually secondary. What's the point if a load sees some 100mV when your circuit can only do it below 1kHz and regulation fails already at 20kHz? (The upper graph of the LM318 actually shows idealized behavior that's not attainable in the real world. Its C(Adj) is 10μF, ours is 1nF.)
"Our filter and storage capacitors smooth out the currents. Here I always calculate the necessary values according to Dr. Dirks' recipe to create the very same behavior as for my regulators: constant impedance across the highest possible bandwidth where we want at least 1MHz. I observe the same phenomenon in low-signal cable. Thicker conductors trigger skin and proximity effects. Those increase resistance with frequency. Impedance at 20kHz might be three times higher than at 100Hz. That's higher conduction resistance and very easily heard.
Hence our very significant efforts to design a regulator with constant impedance to 1MHz so the workings of our actual amplifier module are stabilized. This meant very significant R&D costs because suitable turnkey solutions didn't exist. Creating your own is most time-consuming. If your accountant had any say in it—here reality bites as I'll explain later—time really ought to be money."