[jtwrace]

My plans are to box the Ncores and the smps separately.

I was going to do this and asked before I did.  I was told it will not make any difference.  So, I went with the nicest, most reasonably priced chassis I could find without making one.

[serengetiplains]

Yeah, I suspect it's not a big deal.  I'm going to be bypassing the smps output with some film capacitors, which will spray copious amounts of EMI that is best kept from the Ncore modules, so in my application shielding makes more sense.

[serengetiplains]

Ncore electrolytics appear to be all bypassed.  Bypassing seems not such a bad idea after all, I suppose.  These onboard bypasses are of course damped, and bypass only for VHF.

[serengetiplains]

And here's the VDD on the comparator.  It looks like the VDD lead connects to the second pin on that unused pin assembly (?) on the top of the board, which should make reading the VDD voltage relatively easy.  Feeding the comparator with a squeaky clean supply looks doable.  Now I'm scared because I'm tempted ...

[serengetiplains]

This appears to be the comparator VDD regulator.  It sits about 1" away from the comparator.

[*Scotty*]

serengetiplains, If you own an oscilloscope you might look at the noise on the rails you are proposing to by-pass to see how much is there. The supply to the comparator may already be squeaky clean. Because the Ncore is an RF amplifier the by-pass caps have to be effective at lowering the impedance of the power supply at radio frequencies. If you replace these caps with ones that do not function at these frequencies due to lead inductance you could create a problem that at present does not exist.
Scotty

[serengetiplains]

serengetiplains, If you own an oscilloscope you might look at the noise on the rails you are proposing to by-pass to see how much is there. The supply to the comparator may already be squeaky clean.

It can't be, Scotty, or at least that's my opinion.  I'm talking *very* clean here, something much quieter than a shared supply can offer.  The comparator looks to be powered by either the +V rail for the buffers or for the gate.  I would replace that with a dedicated shunt fed by two series regs in series, so a three-stage reg fed by its own transformer.  That, my friend, is quiet.

Because the Ncore is an RF amplifier the by-pass caps have to be effective at lowering the impedance of the power supply at radio frequencies. If you replace these caps with ones that do not function at these frequencies due to lead inductance you could create a problem that at present does not exist.
Scotty

I appreciate the inductance problem and don't plan to remove the smd shunt caps.  I'll simply add honking-big film caps with long leads spraying RFI *everywhere.*  I expect to hear a sonic improvement.

Per my question a little further above, wouldn't lead inductance effectively render the pulse-train an analogue-like waveform of a lowish frequency?  I see an inductor in series with two capacitors in parallel, the film add-on and the original electrolytic. 

I'm not being facetious here.  I've added such caps to PWM amps before (every one I've owned, in fact).  I wouldn't now listen without them.  The change is that large.

[*Scotty*]

The lead inductance and self inductance of the capacitor raises the caps impedance to the point that any RF present will not enter the cap in any significant quantity. The cap will look more or less like a closed door at these frequencies. It won't spray RF all over because RF won't go into in to it in the first place.
 You could put together a dedicated power supply complete with double regulation, but depending on the operational frequency of the comparator circuit, the inductance of the power supply's wiring and the wiring between the first and the second regulator and the point of use may defeat your efforts. You can't spread RF circuitry all over the place adding lead inductance or board trace inductance to the circuit and have it function as intended. Usually you have to take every measure possible to minimize stray inductance at every step in the circuit to insure its stability and then you sometimes have to start adding pico-farads of capacitance here and there adjacent to components to make the circuit stable. This kind of problem can be a real SOB to avoid and a bitch to fix after it crops up.
Scotty

[*Scotty*]

I might point out that there is a very large difference between the Tact amplifiers and the Ncore. The Tact amplifier has has no error correction possible due its design. The Ncore has error correction applied and as such has a 23dB lower THD figure and correspondingly lower IM distortion figures as well.
The Tact amplifier is much more sensitive to power supply non-linearities than an amplifier with error correction. You can reap relatively larger rewards from power supply improvements because of this compared to an amplifier design with properly applied negative feedback.
Apples and oranges as it were.
Scotty

[serengetiplains]

The lead inductance and self inductance of the capacitor raises the caps impedance to the point that any RF present will not enter the cap in any significant quantity. The cap will look more or less like a closed door at these frequencies. It won't spray RF all over because RF won't go into in to it in the first place.

Thanks for clarifying; I know practically nothing about radiation, when it emits, etc.  And what you say there underlies what I was questioning.  You have a pulse-train current modulating the power supply, creating voltage modulations on that supply.  Wouldn't those larger capacitors operate on that voltage modulation at lower frequencies?  If the inductance filters radio frequency voltage modulations, wouldn't lower-than-RF modulations still get through?  The larger capacitors would then work at the very frequencies at which I want them to work: in the audio realm.  I'm searching here for an explanation why using film capacitors to bypass supply electrolytics on a PWM amplifier works, to my ear, and down to the bass frequencies.

You could put together a dedicated power supply complete with double regulation, but depending on the operational frequency of the comparator circuit, the inductance of the power supply's wiring and the wiring between the first and the second regulator and the point of use may defeat your efforts. You can't spread RF circuitry all over the place adding lead inductance or board trace inductance to the circuit and have it function as intended. Usually you have to take every measure possible to minimize stray inductance at every step in the circuit to insure its stability and then you sometimes have to start adding pico-farads of capacitance here and there adjacent to components to make the circuit stable. This kind of problem can be a real SOB to avoid and a bitch to fix after it crops up.
Scotty

Fair enough.  I would put a very fast shunt right on the pin of the comparator.  If an appropriate ground is near, that would seem to me ideal.  The preregs can be placed very closely, but with the shunt being the most important, getting less than 100% from the preregs would still be good enough for me.  The preregs would themselves be 1x1.5" circuit boards; wiring between the boards would be very short, probably less than a half an inch.

[serengetiplains]

I might point out that there is a very large difference between the Tact amplifiers and the Ncore. The Tact amplifier has has no error correction possible due its design. The Ncore has error correction applied and as such has a 23dB lower THD figure and correspondingly lower IM distortion figures as well.
The Tact amplifier is much more sensitive to power supply non-linearities than an amplifier with error correction. You can reap relatively larger rewards from power supply improvements because of this compared to an amplifier design with properly applied negative feedback.
Apples and oranges as it were.
Scotty

I can hear the benefit of the Ncore's feedback---I have a couple modules playing now.  Their stock sound is better than that of a stock Tact, but the Ncore has a way to go to surpass what I hear with my modified Tact.  The Ncores don't have quite the mud of an unmodified Tact, but I can hear a measure of it in the Ncores.  The mud I'm referring to is the peculiar sound of electrolytic capacitors.  I'm fairly confident the Ncore will sound better in ways the Tact did when I bypassed its supply capacitors.  Fwiw, I've never heard feedback to eliminate the veiling caused by electrolytics---tube amps, SS amps, class D amps, class A amps, etc.  I'd settle for 123dB on that veiling.

[Rclark]

 Very nice close in photos! Good luck with the experiment. Hopefully Mgalusha can measure yours against his stock benchmark.

 Neat 

[serengetiplains]

Thank you, R.  Here's a photo of the area in question.  You can see the comparator on the right.  Speaker leads are on the left.  The double array of solder-points is the power input connector.  The power trace (one of the pair passing through the white dot) runs from the comparator up to T77, which you can barely make out near the centre at the top of the picture.  I haven't traced the power source any further.

[Rclark]

Would you be willing to take close in, hi res pics of the full surface of the amp? Could make a cool poster.

[serengetiplains]

And a few more of that corner, a bit closer.

[serengetiplains]

Would you be willing to take close in, hi res pics of the full surface of the amp? Could make a cool poster.

Sure, I could do that tomorrow.  Its nearly impossible to see the traces on these photos, so I don't think posting a high-rez picture would do a disservice to Bruno.  If anyone has a concern, please let me know.

[serengetiplains]

Here you go!

[Rclark]

Saved. 

 Which camera are you using for that? Those are just excellent close ups. And also fascinating that those tiny caps are also bypassed as well.

[serengetiplains]

I bypassed some output capacitors today.  To reduce inductance, I match two capacitors, tie them together, then run four twisted wires, two +/– to the left side, two –/+ to the right.  This keeps lead length smaller (about 1" per lead depending on the circumference of the cap, = 2" per capacitor).

I'll hold off on reporting much detail right now.  The capacitors I added are new teflons that need time to break in, the DAC is a DragonFly (ok, but not great), I only have one channel bypassed, I want to add bridged modules, the amps and power supplies are still breaking in, etc.  That aside, adding the teflons did what I expected.  Though I hear HF edge (some of which is attributable to amp break-in, teflon cap break-in, jittered output from the DragonFly), the amp hit moments of shocking realism.  When I first turned the amp on (left channel bypassed, right channel stock), I thought I possibly destroyed my tweeter in the L bypassed channel.  I was playing a song with a long lead-in, bass-lower-mid riff and could hear nothing from the tweeter.  After a jolt of anxiety wondering how I could have ruined my tweeter (a ribbon), I realized the tweeter was fine.  What triggered my wondering was I could sense differing HF pressure levels from L and R channels, with the L channel dead.  The difference, I take it, was some form of noise.

Here's another interesting observation.  I can run the amp flat out and the L channel woofer (a mere 4" job) doesn't flap and remains well-controlled.  The sound is clear and undistorted.  I wanted to turn up the volume more, but couldn't---the amp was maxed.  The right channel woofer, for its part, began hitting its excursion limit somewhere near full volume.  When it does, it emits a nasty blurrp.  Not so the L channel.

The HF edge will reduce with time, particularly with a good DAC, better attention to wiring.  A good NOS seems a good fit here.  It will be interesting to hear how bridged operation affects the sound.  Bridged should reduce power supply noise further (nasty electrolytic DA, among other things).

[serengetiplains]

Saved. 

 Which camera are you using for that? Those are just excellent close ups. And also fascinating that those tiny caps are also bypassed as well.

It's a Panasonic Lumix portable, the cheapest way I could get a Leica lens.  It puts out 10MB raw files.  I think the one I posted was a 4.7MB jpg compression, which I'm sure this website reduced further.