Dynamic Comparison of CD, DVD-A, SACD - Part 1
Recently I bought an Audiotrak Prodigy 7.1 soundcard for my HTPC. This is a soundcard based on the Via Envy24HT "Vinyl Audio" controller and a Wolfson WM8770 CODEC supporting 96/24 A/D converters with 102dB dynamic range, and 8 channels of 192/24 D/A converters with 106dB dynamic range.
I also have a very old copy of Cool Edit lying around, and I thought it might be interesting to record and compare CD, DVD-A and SACD (2 channel only) versions of Diana Krall's The Look Of Love (Verve) to see if I can gain any insights on the relative performance of each version in terms of dynamics and ultrasonic frequency content.
Why that particular album? Well, "jimby" from Universal Music assures me that The Look Of Love was entirely recorded, mixed and mastered using analog equipment, and this is one of the few titles that are duplicated across all three consumer audio delivery formats. I did not want to use a recording that was originally recorded in either DSD or PCM as that would favour one format over the other. Also, I have been told the DVD-A and SACD were transferred off the same analog master, with as much commonality in the transfer chain as possible.
I recorded the title track from the album ( The Look Of Love ) three times on separate wave files at the 96/24 sample rate, which Cool Edit then upconverted to 32-bit floating point representation. The CD version was recorded based on the analog outputs from my Sony SCD-XA777ES SACD/CD player, through the pre-amp section of my Denon AVC-A1SE+ amplifier. The DVD-A version (Group 2: 2 channel) was taken from the analog outputs from my Panasonic DVD-RP82 DVD Audio/Video player (also going through the amp), and finally the SACD 2-channel version was also taken from the Sony SCD-XA777ES analog outputs (via the amp).
I also "normalized" each wave file so that the highest peak in the signal corresponds to the maximum amplitude that can be represented using PCM. I also ripped the CD version of the track digitally using Exact Audio Copy , and then upsampled this to 96kHz 32-bit floating point using Cool Edit. Using Cool Edit, I was able to compare the waveforms of all four wave files, and do the spectral plot of the frequency distribution of each wave file, as well as at a particular point in time within the track.
The "Sound" of "Silence"
This is a frequency analysis of the "silent" bit at the beginning of the track just before the music starts on the Exact Audio Copy rip:
As you can see, there is no such thing as "absolute silence", even on a digital rip off the CD. Note, though, that there is a cliff drop at around 22kHz. Note also the rising noise below 1kHz and the "hump" around 20kHz. This is same bit, but as played back by the SCD-XA777ES via the analog outputs:
The Sony played back the "silence" reasonably well. including the rise at the bottom end and the hump around 20kHz. However, note that additional ultrasonic noise between 20-40kHz has now crept in at around -108dB. This is probably a combination of noise generated by the player, the amplifier and the A/D accuracy of the Prodigy sound card. Note that the plots never look identical because I can't guarantee that I am sampling the exact same point in time for each wave file. The three wave files are not perfectly "synchronized" to the extent that 4:29 on one wave is exactly the same point in the music as 4:29 in another file. I have tried to synchronize them by eye, but only to the nearest 0.01 second.
This is the results from the DVD-A version:
As you can see, the PCM playback is very faithful, but minus the hump around 20kHz, so arguably more accurate than the CD version.
This is the results from the SACD:
Note the absence of the hump around 20kHz, but the "rising noise floor" characteristic of DSD asserts itself for frequencies above 20kHz, taking the noise floor all the way up to around -72db (which is still very quiet, if you have ears that can hear that high). There has been various comments made that the DSD rising noise floor starts as low as 5kHz, based on published results in Stereophile magazine, but as you can see I was not able to replicate those results and DSD behaved as per spec.
