CONTACT DETAILS
Brand: Bel Canto Model: DAC2.7 Category: DAC/Preamplifier RRP: $3,950 Warranty: One Year Distributor: Absolute HiEnd Address: PO Box 370 Ormond VIC 3204 T: (04) 8877 7999 E: info@absolutehiend.com W: www.absolutehiend.com
• Balance trim • Superb sound • Feature-packed • State-of-the-art build • DSD • Defaults to AES • d.c. offset • Standby mode
LAB REPORT ON
Newport Test Labs measured the performance of the Bel Canto DAC2.7 using two different digital standards: the ‘Red Book’ CD standard using 44.1kHz/16-bit digital signals and the AES-17 standard, using 48kHz/24-bit digital signals, plus it also measured the DAC2.7’s performance using its analogue inputs and outputs. In all cases, its performance was superb, but due to the amount of data that resulted, I haven’t included all the results from all the tests, merely a representative sample.
Frequency response was excellent using both digital standards and also the analogue inputs. I have shown only the AES-17 digital result, in Graph 7, with an expanded scale of just 1dB per vertical division, and you can see that even despite this, the response is literally ‘ruler-flat’ from 50Hz up to 5kHz and only 0.02dB down at 20kHz, with a 0.1dB droop between 20Hz and 50Hz. Interestingly, the 44.1kHz/16-bit response didn’t exhibit this l.f. droop, being flat down to 20Hz, but rolled off a little earlier at high frequencies to be 0.5dB down at 20kHz.
The frequency response of the analogue section is shown in Graph 11 and you can see that it has 1dB downpoints at 13Hz and 40kHz. Across the 20Hz to 20kHz audio band, it measured at ±0.2dB. Channel separation was excellent, returning results of better than 120dB in the 44.1kHz/16-bit testing (as tabulated) but interestingly returning results of around 110dB in the AES-17 tests, as shown in Graph 8. I suspect this is more to do with the different test methods used for the two different digital formats than any real difference in separation, meaning that if the AES-17 test were to be re-done using the same methodology as used for the ‘Red Book’ test it, too, would return results of better than 120dB. Either way it’s not important, since even the 110dB result is far, far better than necessary to deliver perfect aural channel separation and perfect imaging.
Newport Test Labs measured channel balance at 0.08dB at 1kHz, which is outstandingly good. Channel phase was also outstandingly good, just 0.02° at 20Hz and just 0.71° at 20kHz. Group delay was typical of an oversampling delta-sigma DAC, speaking of which it was good to see that Bel Canto has implemented the de-emphasis circuitry in the DAC and that it’s very accurate, with only a 0.171dB error at 20kHz. Many DAC manufacturers don’t bother to implement de-emphasis, which means that all CDs recorded more than 25 years ago will be replayed incorrectly on those DACs, though since so few will still exist, it’s rather a moot point.
The DAC2.7’s signal-to-noise ratio was
Frequency response was excellent using both digital standards and also the analogue input
measured at around 123dB (weighted) with either test standard, which means the Bel Canto will not only be exceedingly quiet, but will also contribute less noise to the music than whatever amplifier you’re using, so it’s self-evidently an excellent result.
Total harmonic distortion (THD) was vanishingly low. The distortion components have been graphed individually for the 44.1kHz/16-bit digital tests in Graphs 1 through 5, as a summed total for the 48kHz/24-bit digital tests in Graph 9, and again individually for analogue-in/analogue-out in Graph 12. You can see in Graph 1 that there’s a little output stage distortion at the 0dB recorded level that lifts the levels of the second and third harmonic distortion components to around –100dB (0.001% THD), with the fourth and fifth harmonics at around –110dB (0.0003% THD) and all others below –120dB (0.0001% THD). At –10dB recorded level, the second and third components drop to –110dB (0.0003% THD), there’s a fifth harmonic at around –125dB (0.00005% THD), and a sixth and eighth at –130dB (0.00003% THD) with all others below the –140dB noise floor. The slight distortion at 0dB recorded level could have been partly due to the high output voltage of the DAC2.7 at this level, of around 4.4-volts. It’s also a ‘worst case’ result since no music will ever be recorded at this high a level.
Graph 3 shows the distortion with a –60dB test signal and you can see that there’s no harmonic distortion at all, just quantisation noise (the ‘grass’ across the bottom of the graph). This quantisation noise is just an artefact of the test signal not being dithered and would not occur when playing music, which is always dithered to remove quantisation errors. The effect is clearly illustrated in Graphs 4 and 5, which show the same test
signal, but with and without dither. Graph 4 shows an undithered test signal at 1kHz at –91.24dB at far left. All the spikes across the graph are quantisation errors. Graph 5 shows a dithered test signal at 1kHz at –90.31dB and you can see that because of the dithering, all the quantisation errors have vanished. The only penalty for dithering is an increase in the level of the noise floor, but in the case of the Bel Canto DAC2.7, that noise floor is down at around –140dB even with dithering, so it really doesn’t matter.
Graph 9 shows THD vs. Frequency at two levels right across the audio band, and you can see that it’s almost perfectly linear, and down around 0.001%, where it would be totally inaudible. Analogue THD+N (Graph 12) is also low, with all components except the third harmonic (which is at –97dB, or 0.0014% THD) more than 110dB down (0.0003% THD). Intermodulation distortion was very low but—and again interestingly—it was very slightly lower for the analogue input than for either of the digital inputs. Analogue CCIF IMD is shown in Graph 13 and you can see there are only two sidebands either side of the test signal, both sitting at –110dB (0.0003%). There is an unwanted regenerated signal at 1kHz (the difference signal between the two test signals at 19kHz and 20kHz), but it’s sitting down at –117dB (0.00014%). Digital CCIF IMD is shown in Graph 6 and the two h.f. sidebands are at –105dB (%) and the regenerated signal at –104dB (0.00056%). This graph also shows some higher-order components that are caused by interaction with the sampling frequency. Although the differences between the graphs are interesting, they’re of academic interest only, since in all cases the IMD is so low it would be completely inaudible at any playback level.
Absolute phase on the Bel Canto is switchable, but defaults to ‘Normal’ which is as you’d expect, and linearity errors are vanishingly low, as listed in the tabulated results. I was a bit surprised that there is no standby mode, but given that the DAC2.7 will draw only 8.12-watts from your 240V mains supply, you could safely leave it on all the time without fearing for your utility bill, or the planet. That said, it would have been ‘greener’ of Bel Canto to fit a stand-by mode.
Overall, I found the performance of the Bel Canto DAC2.7 to be outstandingly good and, taking into account differences in measurement techniques, also completely in accord with Bel Canto’s own specifications for the DAC2.7, so it delivers all the performance that’s promised.
Readers should note that the results mentioned in the report, tabulated in performance charts and/or displayed using graphs and/or photographs should be construed as applying only to the specific sample tested.