HiFi Music: Listening, Sources, Tracks? | A good recording...

Free Standard Shipping in the US on orders over $99

Shop Audiophile

Audiophile COMMUNITY

Questions

Audiophile

Showing 1 of 59 conversations about:

HiFi Music: Listening, Sources, Tracks?

Duncan

EntropyCollector

Mar 24, 2019

A good recording makes a huge difference, especially the mastering phase. Most sound engineers tend to compress the sound (don't confuse this with data compression) so that it sounds louder (google for loudness wars). This usually means that parts of the audio spectrum are misrepresented and basically forces you to a specific type of listening experience that dates back to radio broadcasting (on radio broadcast you had to be loud to survive the RF noise floor). This "tradition" is present on a lot of studio releases, even now that we have better ways of storing, reproducing, and transmitting audio. You 'll have to look for a release (or a label) that focuses on sound quality, and doesn't mess up with the mix, but presents it in the most neutral/natural way possible so that -if you want- you can "tune" it later on to your own preferences, or just listen to the thing as it really was. No matter how much money you are going to pay for your HiFi system or the X subscription service, nothing can fix a bad recording ! The best experience to me is a live performance, nothing beats the real thing, especially for jazz or classical music, you also better support the artists this way, it's a win win situation. So to answer questions about what details you should look, check out the various releases of an album you are after and get the one that has the best reviews in terms of sound quality, communities such as discogs or musicbrainz can help you with that (whatcd used to have a lot of information on the sound quality of each release, too bad it's dead). Once you have a good recording, you are looking for a good media that can hold it. Studios use ADCs and processing tools, usually with 24bits of dynamic range and 96KHz of bandwidth. This way they can sample audio signals (google for Shannon-Nyquist sampling theorem) up to 96KHz / 2 = 48KHz, with a dynamic range of 144dB (~6bits per dB so 24*6 = 144, in reality it's mostly 118 - 120dB at best). Our ears have a dynamic range of almost 120dB under ideal conditions, so 24bits are just what we need since it goes in multiples of 8bits (an octet), 16bits (what the CDs use) give us ~96dBs which is mostly OK and 32bits are just too much (but it's more "natural" for e.g. a PC so it helps when processing). As for the frequencies of the audio signal, the audible spectrum is (theoretically) 20Hz - 20Khz, we can feel sounds outside that spectrum (especially the lower end of it, with proper speakers) but in most cases, most people are not able to listen to anything above 20KHz, especially as we get older, where we loose perception of higher frequencies. So why do we go for 48KHz instead ? In digital recording the audio signal is split in pieces/samples (or if you prefer quanta), it's not continuous as in an analog recording, this affects both the recording and the reproduction of the analog signal. To put it differently there are volume changes that are too small or too short to be represented. A rough example for a small change is to have a value of 1.4 and only be able to write 1, 2, 3 etc, so 1.4 will become 1 instead and some accuracy will be lost. This is known as rounding error (or quantization error) and is a thing both ADCs and DACs has to deal with (google for dithering). A rough example of a short change is to think of a repeating pulse, if for example there is a beep every 1 second and we sample every 2 seconds we can't distinguish this beep from another beep that goes on every 2 seconds or from a continuous beep. In a worse scenario if we miss the bit for less than a second we'll never hear it ! So from a beep that goes every 1 second we get three possible signals (including the silent one) or "aliases" along with the original. A similar thing happens on the DAC side when in the process of re-creating the analog signal, if we get the 2 beeps/sec signal, and the DAC reads 2 samples/sec, it will also produce a constant beep along any other analog signal that could result these 2 beeps/sec, e.g. a 1 beep/sec, 1beep/0.5 sec etc, all with higher frequencies than the original, another version of "aliasing" also known as "imaging" in this case. In order to avoid such noise, we need to have a low-pass filter on the ADC input so that signals with frequencies higher than the maximum frequency the ADC can sample do not get recorded, and on the output of the DAC so that any generated "aliases"/"images" don't reach our ears. This filter is called an anti-aliasing filter when used on ADCs, and a reconstruction filter when used on DACs. Its cutoff frequency is the ADC's/DAC's Nyquist frequency, which is half its bandwidth, so for an ADC/DAC with 96KHz bandwidth, the cutoff frequency would be 48KHz and for one with 44KHz bandwidth it would be 22KHz which is just above the audio spectrum. So now you know why the CDs have a bandwidth of 44KHz ;-) Since the aliases produced by the DAC will be above the audible spectrum, people have argued that there is no need for a reconstruction filter or any further processing needed. This makes sense especially for DACs that operate on a single bandwidth such as the DACs on CD players that only work on 44KHz, and have an adequate dynamic range of at least 16bits. This approach is known as filter-less, non-oversampling (NOS) DAC design and is only possible for DACs, on ADCs we always need an anti-aliasing filter, as for the oversampling, we'll come back to that later on. In practice a filter is not ideal, you can't have a filter that goes from 100 to 0dBs instantly, the same way a car can't stop instantly, it needs some "space" to work (aka transient response), which means that you have to start filtering before 22KHz in order to reach the desired level at 22KHz. So an ADC/DAC that operates on 44KHz of bandwidth will need to start filtering from ~18KHz in order to reach e.g. -60dBs when it reaches 22KHz, which means you will loose some of the audio on high frequencies (to some people this seems like a "warmer" sound and they are OK with it). To deal with this, we may use a higher bandwidth, so for example instead of 44KHz, we can use 48KHz and filter anything above its Nyquist frequency (24KHz), only now the DAC will have 2 more KHz to work with so it will start filtering from ~20KHz instead, and get closer to the "20 - 20" range. We could be fine with 48KHz (and many people are) but there is still another issue we have to deal with. The filter's response is very important for the sound quality, it doesn't only affect the level of signals above the cut-off frequency (in this case the Nyquist frequency), but also their phase. In other words the signal gets a delay while it gets filtered, relative to its frequency, which in turn affects among other things the stereo imaging of a stereo recording. Stereophonic audio depends on the L and R signals being in-sync, if you create a time difference between them -aka change the signal's phase- you break this and you get less separation between the 2 channels, hence the "sound stage" becomes narrower. This distortion may happen on frequencies even lower than the cutoff frequency which means we need to move the cutoff frequency (hence the Nyquist frequency / bandwidth) even more, since the "steeper" the filter the greater the distortion. Creating good filters is a challenge, especially in the analog domain, on the digital domain on the other hand we have more options available. Instead of filtering the signal outside the ADC/DAC, we can do the filtering on the data itself, at a much higher bandwidth than the one used for A/D conversion, a process called oversampling. So for example in the case of the ADC we'll get an input signal of 24KHz bandwidth, do the A/D convertion using 48KHz bandwidth, oversample it to 96KHz (*2), 192KHz(*4), 384KHz (*8), 768KHz (*16) or more, do the filtering there at a much better resolution and either output the oversampled signal for processing, e.g. a 24/96KHz signal, or down-sample it to the desired bandwidth. This way we'll not only be able to filter out the aliases, we'll also be able to further filter out the noise created due to quantization. On the DAC side, an input signal of 48KHz will arrive, it'll get again oversampled to 96KHz (*2) or more, do the filtering there and output either the oversampled signal, or a down-sampled version of it. That oversampling now can be a huge mess because the DAC needs to "invent" new samples to place in-between the existing ones, a process called interpolation. If this is done poorly, instead of making things better, it can add distortion to the audio and "dry" it up, it may also create phase distortion. Even worse the bandwidth of the oversampled output signal may not be a multiple of that of the input signal, for example 48KHz is half of 96KHz, so if you get an input signal of 48KHz and oversample it to 96KHz you need exactly twice the samples, which makes it easier since you'll just need to place one sample between each two successive samples. If however you get an input signal of 44.1KHz and the output bandwidth is 96KHz you are in trouble, you'll first need to resample the input signal to a rate that is divisible by both 96 and 44 (aka Least Common Multiple), do the processing there and then downsample it to the output bandwidth (44 -> *24 -> 1056 -> / 11 -> 96), or even use multiple resampling steps (e.g. 44 ->* 12 -> 528 -> / 11 -> 48 -> *2 -> 96, so that it re uses 48 to 96 conversion). To avoid this some DACs also support native handling of 44.1KHz so they can also output 88.2KHz (*2), 176.4KHz (*4) and so on. In general resampling is tricky and can add distortion, especially the "uneven" kind, you can get an idea here -> http://src.infinitewave.ca/. The reason I mention all this technical stuff is to give you an idea of how complex a DAC can be, a hint of why e.g. a 32/768 DAC may not be better than a 24/96 DAC, and also to point out that this whole marketing with higher sampling rates and bit rates doesn't make much sense if the rest of the DAC is not implemented properly. Other things that are important for ADCs/DACs is to have a solid clock, one that doesn't loose ticks (google for clock drifting), when that happens the ADC/DAC will skip samples and generate a very ugly noise known as jitter or "digititis". A poorly designed filter even with good resampling in place, can be too steep for example and generate "ringing" noise. In general don't let the bit rate and sampling rate numbers fool you, as with any audio equipment, you should listen to a DAC before buying it, or at least look for trusted reviews. Just don't settle for anything less than 24/96 and look for DACs that can also do multiples of 44.1KHz as well such as 88.2KHz (*2), 176,4KHz (*4), 352.8 (*8) etc, also it's a good idea to check the specs of the DAC chip, since some times the manufacturers of the device will fail to mention that capability since the marketing is mostly on multiples of 48KHz. In terms of digital audio encoding, FLAC is probably your best option, then comes ALAC that's also lossless, and then you go for lossy encodings such as OGG/Vorbis with a high quality setting and then MP3 above 192Kbit/s with multiple passes. What you want is to have the whole spectrum available so that you don't loose any spectral content and the more lossy the encoding, the more spectral content you loose. Obviously if you get a lossy file and re-encode it to lossless, that won't fix your problems and there are marketing people out there that do just that so watch for it. You may use a tool such as spek to figure out if your file is truly lossless and retains the full spectrum. Because of the lost spectral content, if you try to post-process a lossy MP3 or OGG, you'll get various artifacts because the filter or the audio compressor / limiter will see an incomplete spectrum and add noise for the missing parts, you can get away with it with various re-construction tools that try to re-create the lost spectrum by generating harmonics (look for "bass enchancer" or "crystalizer" or "exciter" etc) but again these need a lot of tweaking to get it right and are not suited for all types of recordings. Analog media such as Vinyl or Tapes have a dynamic range of less than 110dBs (110 is a perfect vinyl, it's usualy 60 - 70dBs) and in case of Vinyl you can go up to 50KHz in terms of frequencies (quadraphonic records used such frequencies to encode the rear channels). You don't need anti-aliasing filters used on ADCs/DACs but you need pre-epmhasis/de-emphasis filters such as the RIAA curve on Vinyl (that's what a phono stage does) and the various Dolby filters on Tape. In general the analog path is closer to the way our ears operate and it sounds more naturally, for example in case of overdrive (think of electric guitar distortion), a digital recording behaves worse than an analog one (google for clipping), it distorts the sound pretty badly, where the analog recording will behave more like the real thing (obviously there are high quality tools for dealing with this during mastering so hopefully you won't get such issues on the final mix). However in order to get that experience you'll have to pay a lot of money for your equipment and properly maintain/store your media since they tend to degrade with time. It's also useless to go for analog media when the recording itself is digital, unless your only alternative is the CD version which can indeed be worse since it has to be downsampled to 16/44.1. Another thing to keep in mind is that some engineers do a different mix for the digital media than for the Vinyl version so the Vinyl version may sound better because of that since it's more focused for audiophiles, but in general Vinyl or Tape is not better or worse than a digital recording, again there are lots of details to take into account. As for the rest of the system:

For the amplifier check out its linearity and total harmonic distortion. The best amplifiers are those that behave the same across all frequencies (aka linear amplifiers), they are the most expensive and eat the most power, this property is found mostly in A-class amplifiers. If you think of a speaker moving back and forth, A-class amplifiers amplify both the forth and the back movement, B-class amplify only the forth movement and let the speaker come back on its own due to its elasticity, saving power in the process, AB are a hybrid between them, C is garbage, D use PWM for more efficiency and are more common and cheap, you can get high-quality D-class amplifiers but it takes a lot of work to design one.
For the Headphones/Speakers the larger the better IMHO, in case of speakers make sure you get something that fits your room specs, getting speakers to loud or too weak for your room will do you no good. Listen before you buy, even reviews here won't help you much, get something that matches your ears. It's easier to get the Speakers wrong than headphones so I suggest you start with a good pair of headphones such as the HD6xx, it'll cost you less and you also don't depend on the room. A good reference is if you have frequency response diagrams from a trusted source.
For the cables, prefer balanced systems over unbalanced if your cable lengths are longer than a few meters to avoid interference. The lengths of the cables matter, especially for multi-channel setups, but not that much, most people don't notice anything related to the stereo imaging and different cable lengths. The materials of the cables are mostly a marketing thing, just make sure it's not crap and the connectors are properly attached.

Since you asked about tracks, this is what I use for determining the quality of audio gear:

Nardis from Bill Evans (the Riverside recordings) to check how well the bass is represented (it's hard to get jazz bass right, it's usually "muddy")
Because from Beatles (the original stereo mix) to check the stereo imaging (huge stereo separation on this song)
Various tracks from Ozric Tentacles (I usually go with Kick Muck) for the dynamic range
Various tracks from Alan Parsons Project (Alan Parsons is a legendary sound engineer, among other things), especially for verifying a neutral post-processing (these are engineered so well that any post-processing messes them up, just don't touch them !)
The Dark Side Of The Moon by Pink Floyd (look for the Japanese releases) (Alan Parsons was the sound engineer on this one btw) because I've heard it a thousand times
Rudy Van Gelder (RVG) recordings from Blue Note (another legendary sound engineer)
Symphonic music in general (I love Tchaikovsky, look for Swan Lake, check out Deutsche Grammophon and other relevant labels, also check out for any good recordings of Moriconne)
Various disco / synth-wave etc tracks for checking dynamic range (80s productions are in general awesome, for more recent stuff check out Daft Punk and more recently I found Chrome Brulee that I enjoyed)

In general pick some tracks that you are familiar with, you'll instantly notice the difference, and most importantly don't let anyone tell you what "sounds good" or not, just because it's not FLAC or it's 16/44.1 instead of 24/96 or 32/768. Hope this helps + it wasn't too much :-) P.S. For more information https://www.audiosciencereview.com is a great resource ! It also contains trusted reviews on DACs and other equipment. P.S.2 Here is an interesting site that monitors the dynamic range of various recordings, it should help you verify if a recording was overly compressed or not -> http://dr.loudness-war.info/

(Edited)

erokit

Apr 10, 2019

EntropyCollectorThank you so much for the depth. I knew some of this stuff but you put it all together in what actually was a concise and brief digest.

A community member

Apr 16, 2019

EntropyCollectorMy goodness! A Primer IF I've ever seen one,....and I am a retired mechanical engineer! How generous of you to take so much time to respond to a fellow enthusiast's query! As I am sitting here writing this on the evening of the "Notre Dame," tragedy, I have to appreciate how wonderful SOME of humanity can be. It is just a shame that the same "passion" that motivates your kind response and also motivated the building of that glorious structure, regardless of one's beliefs OR lack thereof, for that matter; cannot be so "ecumenically evident" in too many of our fellow "travelers!?"

EntropyCollector

Apr 20, 2019

Thank you very much for your kind words ! For those of you who are interested in more detailed technical information and a lot of DAC reviews etc, check out https://www.audiosciencereview.com its forum is a great resource.

Vote

jansuz

Apr 20, 2019

EntropyCollectorThank you, clearly explained and eloquently presented.

613Member

May 4, 2019

EntropyCollectorWoW...👍

Vote

peterleyenaar

Jul 3, 2019

EntropyCollectorHere is a chart with the frequency range produced by most musical instruments, it shows a range of approximately 25 hz to 14Khz , CD quality 16/44.1 is more than adequate to represent any kind of music , anything beyond that is totally redundant and mostly marketing hype (like audio cables above a $100) if one considers that most people can't hear any frequencies above 18Khz and musical instruments don't produce any frequencies above say 15khz

Royaume

Oct 15, 2020

EntropyCollectorOh no! Audiosciencereview?? A "great resource"? You must be kidding me. Read "a new methodology for audio frequency power amplifier testing" by Daniel Cheever. Essentially, THD may be useful for engineers, but it is COMPLETELY INADEQUATE as a figure of merit! ASR assumes that THD is a figure of merit, and ranks DACs as such. However, there is in fact no correlation between the measured THD of a DAC and how good it sounds. Stop spreading ignorance and superstitious scientism.

Vote

acousticmood

May 30, 2021

peterleyenaarIt’s not just the frequency response (like the old time stereo salesmen that touted the specs of whatever they wanted to sell - specs don’t tell the whole story) - the better recordings played through better equipment at yes higher resolution provides benefits that are audible. A more defined soundstage for instance. I don’t know if anyone mentioned it but if you have a cd collection- rip them to a drive. Then consider Roon to manage it all, your library, tidal’s library and several cd quality streaming radio stations in whatever genre you’d like. I bought a lifetime subscription to Roon and use it almost every day.

Vote