Zesilovace NCore Tridy D
Zesilovace NCore Tridy D
Zesilovace NCore Tridy D
Kattegat 8
9723 JP Groningen, The Netherlands
+31 50 526 4993
sales@hypex.nl
www.hypex.nl
Ncore® is the trade name for the first significant step in power amplifier performance in a decade.
Building on the stong heritage of UcD, Ncore takes the things UcD does well and does them ten times
better. Literally. Following nearly two years of sporadic unofficial demonstrations Hypex is ready to
announce this ground-breaking technology to the world.
Of course, this being audio, people will not only want to know how well it performs, but how it works. Here’s the
story.
Loop bandwidth. Self-oscillating loops are said Remarkably, an exact oscillation criterium for
2
to have a wider usable bandwidth/switching square wave oscillators can be derived on the back
3
frequency ratio. of an envelope . By adding in extra terms for the
harmonics of the square wave it predicts not only
Robustness. Loop dynamics remain largely
the exact idling frequency, but also the switching
unaffected by significant changes in operating
frequency at arbitrary duty cycles.
conditions such as load changes and compo-
nent tolerances.
arg lim− ∑
∞
( )(
1 − e −2 i⋅π⋅n⋅h ⋅ 1 − e 2 i⋅π⋅n⋅h)
⋅ H (2i ⋅ π ⋅ f ⋅ n) ⋅ e i⋅ f ⋅t = 0
t →0
n =1 2 ⋅ n
The first and last are very real, the middle one is
only true for standard triangle wave based imple-
1
mentations. Ripple-compensated modulators do A graph of the solutions with H being the loop func-
not suffer from this limitation but they are signifi- tion of a UcD style amplifier shows that the switch-
cantly more complicated. ing frequency follows a roughly elliptical trajectory:
arg ( H (2i ⋅ π ⋅ f )) = 0
Next, the nearly sinusoidal waveform at the compa-
rator is treated as an external carrier and small-
signal DC gain derived on the basis of the slope. It
took a while to realise that the classical analysis
was wrong in every way: firstly, only pure sine wave
oscillators run at the 360° phase point. A class D
amplifier is a square wave oscillator. Secondly the
resulting “carrier” waveform at the comparator
input responds immediately to the signal so it can’t This is a very significant result. Switching fre-
be treated as though it were independent. quency may be all over the place but linearity can
be optimized to near perfection. It certainly can be
2
“Globally Modulated Self-Oscillating Amplifier with Improved
Linearity”, presented at the 37th AES conference, 2009.
1 3
WO2009131440, “Method and control circuit for controlling One of those large brown ones that tend to contain mail-order
pulse width modulation” catalogues.
2/5
Technology White Paper
made better than the open-loop linearity of the in the past that it’s understandable how this myth
power stage. That wasn’t obvious from either past arose in the first place.
theory or practice. Being able to optimize linearity
UcD amplifiers are designed to have essentially
removes the single biggest obstacle posed by self-
frequency-independent distortion for psychoacous-
oscillating control.
tic reasons. They typically have a loop gain of 32dB
from DC to 20kHz and open loop THD in the order
5 …and back of 1% (or much lower at low power levels) resulting
5.1 Making practice match theory in in-band distortion products hovering around the
The original UcD circuit has a remarkably simple 0.03% mark for all audio frequencies. Given UcD’s
comparator circuit consisting of six transistors and success on the audiophile front a successor would
two diodes. This is fine until one wants to design have to follow the same spirit. Simply shifting one
optimized modulators and expects the circuit to of UcD’s two real poles to DC would have made the
behave accordingly. At that point a much more numbers look nicer, but would not have yielded a
“ideal” comparator is in order. The final design better amp. Instead, another pole was added to
contains ten transistors, which is nothing compa- obtain a total of 5 poles: one real and two complex
ted to the benefit. The new comparator circuit is a pairs (one pair being the output filter).
lot faster and handles much smaller signals. The The loop gain plot bears closer resemblance to that
match between predicted and actual distortion of a sigma-delta modulator than to that of an am-
performance is now almost exact, resulting in low plifier:
distortion right until the onset of clipping.
3/5
Technology White Paper
solutions appear as downward zero crossings. The- stated immediately after clip) insures that no noi-
re are two, one at 500kHz and one at 26kHz. ses other than regular clipping distortion remain
inside the audio band.
0.1
and holding all integrators in reset, or on letting the
0.05
integrators clip intentionally near maximum modu-
% 0.02
lation. The former has the advantage of being very
0.01
clean and responding only when needed. The latter 0.005
4/5
Technology White Paper
1 +1
T T
0.5 -0
-1
0.2
-2
0.1
-3
0.05 d
B -4
% 0.02 r
-5
0.01 A
-6
0.005
-7
0.002 -8
0.001 -9
0.0006
-10
20 50 100 200 500 1k 2k 5k 10k 20k 20 50 100 200 500 1k 2k 5k 10k 20k 50k 100k
Hz Hz
THD into 4 ohms at 500W (blue), 100W (green), 10W (red). Frequency response at 8, 4 and 2 ohms.
7 Sound
Not so long ago a reviewer, having to test “yet ano-
ther Hypex UcD based amplifier”, yammered that it
really was no fun writing about these things as
“they just sound clean and neutral and do what
they’re asked to do”.
8 Conclusion
A new class D amplifier was presented that delivers
audio performance and sound quality well excee-
ding that of the best linear designs without trading
in any of the classical benefits of switching amplifi-
cation.
5/5
Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.
Alternative Proxies: