Peachtree Audio decco65 D/A integrated amplifier Measurements

Sidebar 3: Measurements

I measured the Peachtree decco65 using Stereophile's loan sample of the top-of-the-line Audio Precision SYS2722 system (see and the January 2008 "As We See It") for both analog and digital inputs. To test the decco65 as a digital processor, I took the measurements from its Preamplifier Output jacks.

Hooking up my MacBook Pro to the Peachtree's USB input, I used the Apple USB Prober utility to identify the decco65 as "XMOS USB Audio 2.0," XMOS being the supplier of the USB data-receiver chip. Operational mode was confirmed as "isochronous asynchronous," and the sample rates handled extended up to 192kHz with 24-bit word length. The coaxial S/PDIF inputs locked to data with sample rates up to 192kHz—as did, rather surprisingly, the TosLink input. All the digital inputs preserved absolute polarity (ie, were non-inverting), and a 1kHz tone at 0dBFS resulted in a level of 4.41V at the preamplifier output with the volume control set to its maximum.

The digital inputs' frequency response with data having sample rates of 44.1, 96, and 192kHz is shown in fig.1. With each sample rate, a smooth rolloff above the audioband is interrupted by a sharp cut in output just below the Nyquist frequency (half the sample rate). Channel separation at low and middle frequencies for digital data was excellent, at 100dB in both directions, but this worsened at 20kHz to a still-adequate 72dB, due to the usual capacitive coupling between the channels, probably at the volume control.


Fig.1 Peachtree decco65, digital frequency response via S/PDIF at –12dBFS into 100k ohms with data sampled at: 44.1kHz (left channel green, right gray), 96kHz (left cyan, right magenta), 192kHz (left blue, right red) (0.25dB/vertical div.).

Increasing the bit depth from 16 to 24 dropped the noise floor by around 12dB (fig.2), implying resolution of approximately 18 bits. The peak representing the 1kHz tone at –90dBFS just reached the correct level, suggesting minimal linearity error. With 24-bit data (blue and red traces), however, some low-level enharmonic spuriae are unmasked. Fig.2 was taken with TosLink S/PDIF data; repeating the test via the decco65's USB port gave an identical result (fig.3), confirming that the USB input does correctly pass 24-bit data. With undithered 16-bit data, the waveform of a 1kHz tone at exactly –90.31dBFS was correctly reproduced (fig.4), though with more noise than the best-measuring D/A processors. With undithered 24-bit data (fig.5), the result was a well-shaped, if noisy, sinewave.


Fig.2 Peachtree decco65, FFT-derived spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with 16-bit S/PDIF data (left channel cyan, right magenta), 24-bit S/PDIF data (left blue, right red).


Fig.3 Peachtree decco65, FFT-derived spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with 24-bit USB data (left blue, right red).


Fig.4 Peachtree decco65, waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit data (left channel blue, right red).


Fig.5 Peachtree decco65, waveform of undithered 1kHz sinewave at –90.31dBFS, 24-bit data (left channel blue, right red)

I uncovered an anomaly in the D/A section's behavior: When fed a full-scale tone, the reconstructed analog signal began to clip, giving rise to a picket fence of distortion harmonics in its spectrum (fig.6) that was independent of the volume-control setting. Fortunately, with a tone at –1dBFS the spectrum was very much cleaner (fig.7), with the second harmonic at –90dB (0.003%) and the third at –96dB (0.0015%), the only harmonics of note. The same thing happened with the high-frequency intermodulation test: the full-scale mix of 19 and 20kHz tones gave rise to many distortion products (fig.8), but the same signal at –1dBFS produced a clean spectrum (fig.9). Given how rarely music has a sustained level greater than –1dBFS, this anomaly should have no subjective consequences.


Fig.6 Peachtree decco65, spectrum of 1kHz sinewave, DC–10kHz, at 0dBFS into 100k ohms, volume control at –12dB (left channel blue, right red; linear frequency scale).


Fig.7 Peachtree decco65, spectrum of 1kHz sinewave, DC–10kHz, at –1dBFS into 100k ohms, volume control at –12dB (left channel blue, right red; linear frequency scale).


Fig.8 Peachtree decco65, HF intermodulation spectrum, DC–30kHz, at 0dBFS into 100k ohms (left channel blue, right red; linear frequency scale).


Fig.9 Peachtree decco65, HF intermodulation spectrum, DC–30kHz, at –1dBFS into 100k ohms (left channel blue, right red; linear frequency scale).

The decco65 proved effective at rejecting word-clock jitter on both its S/PDIF and USB inputs. Fig.10 is the spectrum of the output at the preamplifier outputs while the Peachtree decoded a 24-bit version of the Miller-Dunn J-Test via the USB port. No significant sidebands can be seen.


Fig.10 Peachtree decco65, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz: 24-bit data via USB from MacBook Pro (left channel blue, right red). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.

Turning to the decco65's performance via its analog inputs, the maximum gain measured at the speaker outputs at 1kHz was 31.3dB, this dropping by 0.25dB when the tube was switched into circuit with the remote control. The maximum gain from the preamplifier output was 8.25dB. The output was non-inverting from both speaker and preamplifier outputs. The preamplifier output impedance was a low 50 ohms at high and middle frequencies, increasing to 1k ohm at 20Hz, presumably due to the presence of an undersized coupling capacitor. The analog input impedance was a moderately high 15.5k ohms at 20Hz and 1kHz, dropping slightly but inconsequentially to 13.5k ohms at 20kHz.

The output impedance was moderately high over most of the audioband, at 0.18 ohm, rising to >1 ohm at 20kHz, due to the presence of a low-pass filter to reduce the level of ultrasonic noise. As a result of the interaction between this high source impedance and the impedance of our standard simulated loudspeaker, there was a variation in response of ±0.4dB (fig.11, gray trace), and an increasing rolloff at the top of the audioband as the load impedance decreased. The output at 20kHz is –1dB into 8 ohms (blue and red traces) but –3dB into 4 ohms (cyan and magenta traces). This rolloff slows the rise of a 10kHz waveform (fig.12). Like all subsequent graphs other than fig.13, fig.12 was taken with an Audio Precision AUX-0025 passive low-pass filter between the dummy load and the analyzer's input, to prevent ultrasonic switching noise generated by the decco65's output stage from corrupting the measurements.


Fig.11 Peachtree decco65, volume control at maximum, analog frequency response at 2.83V into: simulated loudspeaker load (gray), 8 ohms (left channel blue, right red), 4 ohms (left cyan, right magenta), 2 ohms (green) (0.25dB/vertical div.).


Fig.12 Peachtree decco65, small-signal 10kHz squarewave into 8 ohms (with AP low-pass filter).


Fig.13 Peachtree decco65, tube in circuit, volume control at 12:00, analog frequency response at 2.83V into 8 ohms (left channel blue, right red) (0.25dB/vertical div.).

Fig.11 was taken with the volume control set to its maximum; the matching between channels can be seen to be excellent. However, at lower control settings, an imbalance of up to 1dB appeared. Fig.13, for example, was taken with the volume control set to 12 o'clock—the right channel is now 0.75dB lower than the left. The tube was switched into circuit for this graph; as well as reducing the level of both channels by 0.25dB, it rolls off the top octave by 3dB at 20kHz. Channel separation measured at the speaker outputs was modest, at 70dB in both directions in the midrange, decreasing to 42dB at 20kHz (not shown).

The decco65 had much less ultrasonic noise in its output than other class-D amplifiers I have measured recently: just 23.6mV with a center frequency of 401kHz with the inputs shorted and the volume control at its maximum compared with 790.5mV for the Bel Canto CR7 reviewed elsewhere in this issue and 851.5mV for the Anthem M1 that was reviewed in the December 2012 issue. The decco65's noise level is equivalent to an unweighted, wideband signal/noise ratio of 41.6dB ref. 2.83V into 8 ohms, which increased to 70.4dB when the Audio Precision passive low-pass filter was switched into circuit. Further restricting the measurement bandwidth to the audioband increased the S/N ratio to 74.7dB, while an A-weighting filter gave 77.5dB. These are fairly modest results, mainly due to the presence of some low-level, power-supply–related spuriae in the decco65's output (fig.14).


Fig.14 Peachtree decco65, spectrum of 1kHz sinewave, DC–1kHz, at 30W into 8 ohms (with AP low-pass filter, linear frequency scale).

Figs. 15 and 16 show how the THD+noise percentage in the Peachtree's output varied with power into 8 and 4 ohms, respectively. Specified as giving a maximum power of 65Wpc into 8 ohms (18.1dBW) or 95Wpc into 4 ohms (16.8dBW), the decco65 is shown by these graphs as clipping at 90Wpc into 8 ohms (19.5dBW) and at 120Wpc into 4 ohms (17.8dBW). (Clipping is defined to be when THD+N reaches 1%.)


Fig.15 Peachtree decco65, distortion (%) vs 1kHz continuous output power into 8 ohms (with AP low-pass filter).


Fig.16 Peachtree decco65, distortion (%) vs 1kHz continuous output power into 4 ohms (with AP low-pass filter).

Fig.17 shows how the THD+N varied with frequency into 8 ohms (blue and red traces) and 4 ohms (cyan and magenta). The decco65 turned itself off with sustained drive into 2 ohms, which is why I haven't plotted its behavior into this low impedance. (Turning the amp off and on again restored operation.) As is typical of class-D amplifiers, the THD rose at higher frequencies, especially into 4 ohms. This graph was plotted without the tube in circuit; switching it in made no difference.


Fig.17 Peachtree decco65, THD+N (%) vs frequency (with AP low-pass filter) at 15.65V into: 8 ohms (left channel blue, right red), 4 ohms (left cyan, right magenta).

At a moderate output level, 30W, into 4 ohms (fig.18) and 8 ohms (fig.19), the distortion was predominantly fifth harmonic in nature, though other harmonics are also present. At lower powers, however, the third harmonic dominated (figs.20 & 21). Finally, despite the decco65's decreasing linearity at high frequencies, it didn't perform as poorly as I was expecting on the high-frequency intermodulation test. With an equal mix of 19 and 20kHz tones a few dB below visible waveform clipping on the oscilloscope (fig.22), the difference component at 1kHz lay below –80dB (0.01%), with the higher-order components at 18 and 21kHz at –70dB (0.03%). Again, switching the tube into circuit produced no appreciable difference in these results.


Fig.18 Peachtree decco65, 1kHz waveform at 30W into 4 ohms (with AP low-pass filter), 0.045% THD+N (top); distortion and noise waveform with fundamental notched out (bottom, not to scale).


Fig.19 Peachtree decco65, spectrum of 1kHz sinewave, DC–10kHz, at 30W into 8 ohms (with AP low-pass filter, linear frequency scale).


Fig.20 Peachtree decco65, 1kHz waveform at 5W into 8 ohms (with AP low-pass filter), 0.056% THD+N (top); distortion and noise waveform with fundamental notched out (bottom, not to scale).


Fig.21 Peachtree decco65, spectrum of 1kHz sinewave, DC–10kHz, at 1W into 8 ohms (with AP low-pass filter, linear frequency scale).


Fig.22 Peachtree decco65, HF intermodulation spectrum, DC–24kHz, 19+20kHz at 30W peak into 8 ohms (with AP low-pass filter, linear frequency scale).

Considered in the light of its $999 price, the decco65's measured performance is better than you might expect. Its digital section offers relatively high resolution, and while the predominance of the fifth harmonic at high powers is not something I like to see, its amplifier offers more subjectively benign distortion products at lower powers. The only thing that bugged me was the poor volume-control tracking, which could be considered inevitable given the designer's cost constraints.—John Atkinson

Peachtree Audio
2045 120th Avenue NE
Bellevue, WA 98005
(704) 391-9337

awehns's picture

Unless you are a human FFT analyzer, figures 4 and 5 can't be seen as spectra.

awehns's picture

Figure 3 isn't a waveform!


Love all these measurements though, so thanks.

John Atkinson's picture

Sorry about the mislabeled graphs. I am traveling today but I'll fix this later today.

Update: Okay, got to where I was going today (flying from Seattle to Ashland Oregon) and have fixed the caption problem.

John Atkinson

Editor, Stereophile

mattd's picture

Is the nova125 review available online? 

John Atkinson's picture

Is the nova125 review available online?

Not yet, I m afraid.

John Atkinson

Editor, Stereophile

RoryB's picture

From the specifications of the amp section itself, it appears that the Decco65 is using the Texas Instruments TAS5613 150WPC class-D amplifier IC. This IC can deliver 65WPC into 8 ohms and 125W at 4 ohms with 1% THD, and claims both >90% efficiency and flat THD content across the entire audio bandwidth, with extended bandwidth to 80kHz. Pretty darn awesome for something the size of a postage stamp. The key to this, according to the TI data sheet for the IC, is their "PurePath HD" feedback scheme, which enables wide bandwidth and extremely low distortion (compared with other class-D ICs). A larger 300WPC amplifier IC is also available, the TAS5630. For those that are new to all this, an IC-based class-D amplifier builds everything but the input filtering caps and the output filter into the chip itself - which means the output devices are also onboard. A benefit to this is the extremely short signal paths on the amplifier IC's silicon die. By contrast, the popular ICEpower and Hypex UcD/nCore amplifiers use discrete components.

What this review says to me is that affordable class D amps (and particularly IC-based class-D amps) are finally "getting there" in terms of audio quality to where they can be used in a credible high-end audio product. I lost a channel in my class-AB amplifier recently and am using a Chinese stereo amp (Sinewave Genius200) based on the TAS5613 while the big iron is being repaired. Overall, my impressions mirror the author's - while not the last word in transparency, there is a wealth of detail being unveiled. Also, there are no sweeteners in the amp IC itself - tonally, it is on the neutral to cool side (like a class-AB amp based on bipolar FETs instead of MOSFETs). But the amp is remarkably free from the phasey 'hash' and elevated treble THD that I've heard from other class-D IC amps, mainly the low-cost Tripath variety or later ST Micro ICs (though the Tripaths were at least listenable), so imaging is much improved over what I've heard before from affordable class-D amps.

Still, the work is not done for engineers of class-D ICs, or those amp manufacturers that use them. Even the TI PurePath ICs are power-rated at THD levels  (1% THD and 10% THD) that would be considered unacceptable in a Class-AB amplifier at full power, so there is room for improvement. (The TAS5613 is rated for 0.03% THD at 1W - the foregoing statement is only meant to highlight that at full power, class-D amps still have distortion issues.) Also, this indicates that class-D amps should actually be de-rated slightly when comparing them to similarly rated class-AB units to put them on equal footing where THD is concerned. Still, I am glad to see these improved class-D ICs making their way into higher-end products as their performance rises to the necessary level.

There is still some room for designers to affect the sound of these amps, even though the 'guts' are hidden inside the IC. What amp designers can do is make careful choices in the selection of input filtering caps and also in the design of a properly-damped output filter that provides maximal bandwidth and uses components designed for high voltage and high current applications, such as metalized film capacitors, potted large-gauge inductors, and high power resistors of good quality. A well-designed output filter will be more than a simple second-order lowpass LC network - it will include a parallel leg containing a shunting capacitor in series with a damping resistor to prevent additional HF distortion from voltage spikes that overwhelm the amp IC's feedback network. The math for output filter design is very simple, so when amp manufacturers leave this stuff out, I wonder what they are thinking other than trying to shave off every penny of cost - not a mindset that will succeed in the high end.

Lastly, to stay competitive, I'd like to see ICEpower A/S and Hypex Electronics BV create their own IC designs incorporating their patented technologies, such as a second feedback loop that monitors the output of the output filter and attempts to exert greater control via the amp stage. The very short signal paths on an IC would be of clear benefit to these already class-leading technologies.