It’s commonly held among audiophiles who understand electronics that well engineered and built DACs are audibly transparent. This belief comes from properly conducted double blind tests with well trained listeners, performing ABX testing, learning about how DACs work and how to measure them, reading detailed DAC measurements made by others and performing those measurements themselves.<\/p>\n
However, “well engineered and built” is a loaded phrase. Some DACs use a AA filters that start attenuating within the passband (below 20 kHz), or have passband ripple, or non-flat phase response. Some DACs have elevated IM distortion at moderate levels (the ESS IM hump<\/a>). Others have increasing distortion near full scale. Some modulate power or clock noise into the outputs due to insufficient filtering. All of these limitations can be audible under the right conditions, and have been observed with DACs considered to be well engineered and built.<\/p>\n Also, “audibly transparent” is a loaded phrase. Does it mean musically<\/strong><\/em> transparent, or perceptually<\/strong><\/em> transparent? Consider the\u00a0 difference between 96 kHz and 44.1 kHz rate, or a linear vs. minimum phase filter having equal amplitude responses. These differences are considered to be inaudible. I can differentiate them in an ABX test, but only using “appropriate source material”. In this case, a high quality recording of jangling keys or a square wave. Differentiating them with a musical signal is much more difficult, and I’m not sure I could do that. In some cases, I’ve detected these differences with high quality castanet recordings, but is that really music? I consider it on the borderline between music and test signal. We listen to music, not test signals, so while I believe a good audio system should strive for perceptual transparency, some people consider the lower bar of musical transparency to be sufficient.<\/p>\n The Corda Soul<\/a> is a DAC, preamp, and headphone amp with useful DSP functions. I’ve owned one for nearly 5 years and in some ways it’s one of the best measuring pieces of gear I have seen. Subjectively it sounds fantastic (by which I mean it is transparent, or doesn’t sound like anything – and many DACs and preamps do not achieve this, adding their own coloration to the sound) and I prefer the Soul to other high quality DAC \/ preamps in direct comparisons. I never expected to encounter a better sounding DAC…<\/p>\n I recently replaced my Tascam SS-R1 with the newer model, the DA3000. The SS-R1 still works like new but is limited to 44.1 k and 48 k sampling, where the DA3000 supports every sample rate from 44.1 k to 192 k and DSD 64 and 128. This means I can connect the Soul digital output to the DA3000 digital input, and the DA3000 will simply work regardless of the sample rate. The DA3000 analog output (balanced XLR) goes to the Soul’s analog input.<\/p>\n I tested the DA3000 and the Soul by playing test signals through the Soul and capturing its analog output on the DA3000. The Soul always had a hump in the distortion curve that peaks at -70 to -80 dB at 1,000 to 2,000 Hz, and appears at every sample rate. But only at high digital input levels (above -12 dB). More on that here<\/a>. Since I now had two recorders – both the DA3000 and the SS-R1 – I was able to narrow down what causes this by bypassing the Soul’s DA converters, using only its analog gain stage. I connected the Soul’s digital output to the DA3000, then sent the DA3000 analog output to the Soul’s input, then recorded the Soul’s analog output on the SS-R1. The hump disappeared entirely!<\/p>\n This means the Soul’s distortion hump was coming from its internal DA converters, or its FF de-emphasis curve which is implemented in DSP (both of which are bypassed when you use an external DAC). This is surprising, since the Soul goes to great lengths to ensure clean DA conversion. It uses well regulated switching power supplies and dual WM8741 DAC chips, each in mono mode, one for each channel, fully balanced. However, the Soul’s analog gain stage measured entirely transparent. Noise was below any threshold I could record. The SS-R1 is only 16 bit, so all I can say is that noise is below -96 dB even at low volume settings. Frequency response was perfectly flat. Distortion measured at -96 dB, the limits of 16-bit.<\/p>\n So: the Tascam DA3000 DA converters measured cleaner than the Soul (here<\/a> for measurement details). And not slightly cleaner, but a whopping difference: from -70 dB to below -96 dB, at least 26 dB and probably more. And this is in a frequency range where our hearing is most sensitive. But that said, not every difference you can measure, is audible… For just one example, this difference only exists at high digital input levels. From -12 dB and below there is no difference I can measure.<\/p>\n The Soul allows the use of an external DAC and has a switch to instantly switch between that and its internal DAC. The difference is readily audible, by which I mean I can hear it not only with test signals but on a wide variety of music. Perceptually and subjectively, compared to the DA3000, I characterize the Soul’s internal DAC as:<\/p>\n In contrast, the DA3000 DAC sounds a touch smoother, softer, warmer.<\/p>\n I call this a revelation because it was so unexpected. It really surprised me. Up to now, the Soul has been less edgy \/ grainy than other DACs I have owned, such as the Oppo HA-1<\/a>. Even though the difference is subtle, it is a joy to listen to my familiar recordings with a slightly smoother, warmer perspective.<\/p>\n Note: The 0.2 dB loudness difference is the obvious culprit. It’s small enough to be barely perceptible as loudness, yet perceived indirectly as “richer”, “more detailed”. Yet normally, all else equal, slightly louder is perceived as slightly better. So it’s the opposite of expected. And I hear the same subtle differences even after adjusting for the 0.2 dB loudness difference.<\/p>\n Upon further listening I made some other observations. The Tascam DA3000 DAC doesn’t resolve fine detail quite as well as the Soul. It slightly veils some of the subtle background sounds. However, in voicing and soundstage I still preferred the Tascam. So the difference was more of a trade-off.<\/p>\n The DA3000 isn’t better, it’s just different. As an analog preamp, the Soul is unmatched both subjectively and objectively: clean and transparent with noise and distortion so low I can’t measure it, and perfect channel balance at every volume setting. Its DSP functions are useful and well implemented. And it is well built with great support. D\/A conversion is its weak point, but even there, it has exemplary performance below about -6 dB digital input level.<\/p>\n To summarize DA300 vs. Soul, the Soul sounds more detailed yet thinner. The DA3000 sounds softer, warmer, but less detailed. These differences are subtle and nothing obvious in the measurements indicates why.<\/p>\n If my opinion isn’t clear by now, I’ll just say it. Well engineered DACs do not all sound the same.<\/strong> Some may sound the same, while others may have audible differences. All audible differences can be measured – if you know what to measure, under what conditions to measure it, what signals to use when measuring, and how to do it right. But most published specifications are only the most basic measurements that don’t nearly cove everything that can be heard. So just because basic specs like SINAD and FR are the same, doesn’t necessarily imply they sound the same.<\/p>\n","protected":false},"excerpt":{"rendered":" Introduction It’s commonly held among audiophiles who understand electronics that well engineered and built DACs are audibly transparent. This belief comes from properly conducted double blind tests with well trained listeners, performing ABX testing, learning about how DACs work and how to measure them, reading detailed DAC measurements made by others and performing those measurements … Continue reading Audio: DACs and Revelations<\/span> The Setup<\/h3>\n
The Revelation<\/h3>\n
\n
But Wait, There’s More!<\/h3>\n
The Conclusion<\/h3>\n