CD-Quality vs. High Resolution Streaming

[tmallin]

I find that telling the difference between 16/44 and higher-res versions is sometimes difficult. I find it to be highly recording dependent.

Context

My current main audio system is outlined at "My Current Audio Systems" thread. As you can see, I currently use the Dutch & Dutch 8c speakers in my main system. Those who know those speakers will will know that these active speakers contain their own DACs which,, while they will accept PCM inputs up through 24/192, convert all digital inputs to 24/48. The speakers also do not accept DSD streams. Thus, you may say, my take on this topic is invalid on its face since my current speakers don't allow me to hear the full benefits of 24/88, 24/96, 24/192 PCM recordings or any sort of DSD recordings or DSD resampling.

But prior to my recent acquisition of the D&D speakers, I was feeding the analog output of my Lumin X1 streamer through a pair of Benchmark AHB2 amps to Gradient 1.4 speakers, and before that into Harbeth M40.2 speakers. Via the Gradient 1.4 speakers especially I did extensive experiments comparing the native resolution of all programs with upsampled or resampled versions. The upsampling/resampling was done with either the Lumin X1 or Roon through my Roon Nucleus+. I compared the quality of upsampling/resampling of the same type done by the X1 or via Roon. I compared various levels of PCM upsampling with both the X1 and Roon. I compared PCM upsampling with DSD resampling at various levels of resampling up through DSD 512 since the X1 handles such DSD 512 signals natively. I tried all the various options Roon gives for DSD resampling.

I also compared in native format (not upsampled or resampled) the MQA versions of recordings available through Tidal with the High Resolution versions available through Qobuz. I further compared the CD Redbook versions of the same recordings on Tidal with the MQA versions and the CD Redbook version on Qobuz with the High Resolution version on Qobuz. Keep in mind that the Lumin X1, from its analog outputs does full MQA unfolding and rendering.

Results as to Upsampling/Resampling

A rather consistent pattern emerged as to upsampling and resampling. In most cases upsampling/resampling done to any particular level via Roon bested that done by the Lumin X1. Via Roon the highest level of PCM upsampling/resampling, 24/768, and the highest level of DSD resampling, DSD 512, sounded best of all the upsampling/resampling options available via Roon and the DSD 512 sounded better than PCM 24/768. There were also certain DSD 512 options provided by Roon which, if engaged, further improved what DSD 512 sounded like.

My considerable sighted experimentation with applying these processes at the playback end through my Lumin X1 and Roon suggests to me that these processes do indeed change the sonics, usually in terms of an apparent enlargement of space and spreading out of images within that space, together with a concomitant shift of tonal center toward the high frequencies. The higher the level of PCM or DSD upsampling/resampling, the more extreme these effects become. I'm using "extreme" here as a relative term; the effects are subtle compared, just for example, to the sonic differences between two different models of excellent phono cartridges used in the same tonearm on the same turntable. Still, while you may or may not like the overall effect of such processing, I think most who have experimented with this sort of thing would acknowledge that such processing does indeed make sonic differences.

I liked the spatial changes in many cases. Sometimes things got a little too "pulled apart," but by an large a bigger space was preferable. What I did not generally like was the shift of the tonal center of recordings toward the higher frequencies. Most recordings are balanced too brightly to begin with, I find, compared to the sound of real acoustic instruments as heard from a favorable concert hall position. While some listeners may interpret this tonal shift as added clarity or transparency, I came to hear any increase in apparent clarity as merely the result of the tonal shift toward the treble.

Thus, eventually I began reverting to listening to most material at its native resolution. Native resolution sounds more relaxing and naturally, not hyped. The one exception to this is for low-resolution internet radio programs. For those, I often find that the resampling Roon automatically does of 16/44 signals to a 24/48 level adds openness wiithout affecting the tonal center too much. But even with internet radio program, I find there often are tradeoffs between the native 16/44 provided by the Lumin X1 radio station and the Roon 24/48 radio station.

Results as to Tidal MQA vs. Qobuz High Resolution

In short, the results of such comparisons are highly recording dependent and inconclusive. Roon allows easy A/B comparison of the Tidal MQA and Qobuz Hi Res versions with only a few seconds delay. I don't do any intentional blind tests.

I have no reason to believe either one sounds better. As you know, some powers that be in audio came out quite strongly for how much better MQA sounded than anything else. Others believed otherwise. That controversy seems to have receded into the background these days since folks can more easily compare the two since Qobuz Hi Res came on the scene.

I don't hear one as truly superior to the other, but I do find the two often sound quite different from each other. This was true when I was listening to the Lumin X1 doing full MQA decoding/unfolding and equally true now that I only hear the first unfold via Lumin or Roon into the non-MQA DACs of the Dutch & Dutch 8c speakers.

There does seem to be a general pattern in such comparisons. The Tidal MQA versions generally sound a bit "softer" and easier on the ears while at the same time often resolving bass and midrange details in a more natural way. On the other hand, the top two octaves of information from the Qobuz High Res version is usually more resolved, filligreed, and just sounds more extended without sounding bright. More inner detail is apparent without any overt change in tonal balance. Bass has more drive/punch. I can well understand how different listeners could hear these differences as better or worse, but most often I just hear the presentations as different.

[Continued below]

 

[tmallin]

Results as to CD vs. High Resolution

So much of the comparative quality judgment of CD vs. High Resolution versions of a particular recording depends on whether you are really comparing apples to apples. I have a very strong suspicion that in many if not most cases apples are actually being compared with oranges, especially if the higher-res versions of the recording are released over time as "new and improved" versions with different engineers supervising the final product each time.

These professionals most often can't resist second-guessing earlier engineering efforts and they are each looking to sonically distinguish their version in some way which at least some listeners will hear as a "definite improvement" as one "progresses" from CD 16/44 to 24/44, 24/48, 24/96, 24/192, DSD, etc. We're talking economics here since what is being urged is rebuying of the same artistic material for sonic reasons only. And, from a less skeptical viewpoint, it is certainly true that in releasing a new, higher-resolution version of an older recording, everyone involved in the production may take a lot of extra care to make sure that their efforts are as clean and transparent as possible.

One easy way to make later high-res versions sonically distinctive is more or less compression. This sort of change makes true level matching very difficult to accomplish. For many listeners, more compression is better, oddly enough. Or at least many engineers seem to think that way for the rock/pop market; the loudness wars continue. That's probably because more compression results in higher average loudness which, up to a point at least, is perceived as better sound since increased volume is usually heard as better sound.

Other obvious ways to produce distinctive sound involve changing the frequency balance of the recording: more or less treble or bass, or a more projected midrange. These frequency balance changes don't have to be large--half a dB over a broad range is easily heard, for example. Over time, many reissues seem to shift the tonal balance toward the treble as this seems more "revealing" and is seen as better comporting with modern musical taste.

Or, if the original recording was overly compressed, reducing that compression by electronic means can easily be heard as creating a more dynamic and cleaner, lower distortion sound. This may all be true, but it is true regardless of whether the sample rate or depth is changed.

Many multi-track pop recordings are blatantly remixed in later remasterings because some creative people think they know better than the original engineers and band how the tunes should sound. The Steven Wilson-labeled remixes of 1970s-era Yes, Jethro Tull, and other bands' titles come to mind. Many people think these remixes are better from a musical standpoint than the originals, and certainly the sound of the recordings also is significantly different. Such recordings definitely should not be used when trying to determine what effect, if any, sample rate and depth have on recorded sound.

Thus, for any but brand-new recordings simultaneously released in both CD and High Resolution versions, I think it's best to presume that there has been some manipulation of the recording in the years between the original and later high res version.

For digital recordings originally made at 16/44 or 24/48, the creation of 24/96, 24/192, and all levels of DSD products requires upsampling/resampling. As discussed above, my considerable sighted experimentation with applying these processes at the playback end through my Lumin X1 and Roon suggests to me that these processes do indeed change the sonics, usually in terms of an apparent enlargement of space and spreading out of images within that space, together with a concomitant shift of tonal center toward the high frequencies.

Note that a lot of "high res" recordings are derived from 24/48 recordings which seem to be the de facto standard for a lot of studio recordings these days. Thus, what is labeled as 24/96 or 24/192 may just be an upsampled version of a 24/48 original recording. Upsampling, at least the type done at the playback end via the likes of Roon and Lumin, can clearly result in a different sound.

Note that Qobuz often identifies recordings as "Hi Res" as long as they are somewhat better than 16/44. Qobuz labels even 24/44 recordings as "Hi Res."

Also, I think there are also often differences between the Tidal and Qobuz 44/16 versions of a given recording; very generally Tidal has more bass and drive while Qobuz seems clearer and lower in distortion and less compressed, although I would bet both services claim not to add any compression.

I don't think that it is a simple matter to convert from one sampling rate to another if the two different rates are not integer multiples of each other. Thus, given the 16/44 CD standard, you would think that most Hi Res versions would be 24/88 or 24/176, rather than 24/48, 24/96, or 24/192. While Reference Recordings stuck with simple integer conversions of the standard 44 Redbook sampling rate, other companies, by and large have not.

This may have to do with the fact that video is usually done at 24/48, and that seems to be the defacto standard of many audio professionals as well. Thus, multiplying the 48 rate by 2 or 4 yields 96 and 192. This means that upsampling an original 16/44 CD Redbook standard recording to a higher 96 or 192 rate is not as simple a matter as it may seem.

The same goes for downsampling a newer digital recording which was originally made at 24/96 or 24/192. Downsampling to 16/44 is not a simple arithmetic process. The downsampling might produce digital artifacts, in other words, if not done very carefully. Reference Recordings mostly used 88 and 176 as their original digital sampling rates and that made for easier conversion to the 16/44 CD standard.

Bottom line: Whether a given high res version of a recording actually sounds better than the CD Redbook standard version is something which must be judged on a case by case basis. One must be careful to compare apples to apples, noting if anything other than the sample rate and bit depth has been changed between the two versions. In many if not most cases of older recordings upsampled to a higher res version, I think it should be assumed that this is not an apples to apples comparison. And in the case of newer recordings where a CD Redbook standard version is derived by downsampling from the original higher resolution digital recording, the usual lack of simple integer multiples of the sampling rate may produce digital artifacts in the 16/44 version unless the engineers are very careful about the process.

If one is satisfied that CD quality streaming is good enough that nothing more in terms of resolution matters, I will not argue. We are blessed indeed that so much very high fidelity material is available so easily for so little money via streaming.

Implications for My Use of the Dutch & Dutch 8c Speakers

Given my conclusions above, the lack of the ability of my current system to reproduce native resolutions above 24/48 and inability to handle DSD recordings is of minimal consequence. Even where prior versions of my system allowed full MQA unfolding/rendering, PCM up through 24/768 and DSD up through 512, I generally preferred native sampling and bit depth to be more realistic and better sounding than upsampling/resampling.

While I might prefer my current D&D 8c-based system to handle native PCM up through 24/192 and do full MQA unfolding/rendering, the proof is in the pudding. If I were missing a lot, my reactions to the D&D speakers would not be so overwhelmingly enthusiastic.

 

[Addicted to hifi]

I like cds better than streaming.

 

[richard_crl032]

And my upgraded Denafrips Tplus dac is not MQA capable and Tidal is not an option for me