Saturday, 18 January 2014

Audiophile bullshit

Once again we see some incredible twaddle from audiophiles and once again people fall for it - so as a public service I am going to try and explain things as clearly as I can. I will try and be fair here, honest.

To start with, you need to understand a relatively simple concept, which is useful for anyone, not just those in to music. The difference between digital and analogue. It is really not as hard as it sounds.

In the real world we encounter things like sound waves. These work with our ears to allow us to hear things - they are pressure waves where the air has waves of compressed and less compressed air (sound can travel through other media). These waves trigger nerves in our ears to signal to our brain that we have heard something. We are very good at picking up the frequencies (tones) of sound within a range of frequencies. Some creatures can hear wider ranges, and the range we can hear changes as we get older.

These compression waves in the air have a pattern to them. If you drew the pattern of higher and lower pressure as a height on a graph on paper, you would see something like this image below. This shows one frequency (i.e. one tone) at a time, but sound can be a complicated mix of different tones. Here you see how louder means higher changes in pressure, and higher pitch means faster changes in pressure.


One of the problems with sound waves traveling through the air is that the sound changes. We know this. We know a sound further way sounds quieter. We know that other background sounds can make a sound harder to hear. We know that some rooms change the sound we hear, with an echo being the most obvious of changes. This is why theaters are carefully designed so that people can hear the sound as it was intended.

But some times we want to record sound or make sound, and the simplest approach to this is using analogue signals. These mean converting the sound from pressure waves to an electrical signal. Once upon a time the conversion was not even electrical - it was mechanical movement of a needle on a cylinder. We can use something simple like a different voltage for different sound pressures. We get voltage wave forms just like those shown above. A microphone does this conversion. We can later convert back from electrical signals to a sound using a speaker. The electrical signal is analogous to the sound pressure wave, and is called an analogue signal. There are several ways to do this conversion, some being better than others. For example, AM and FM radios work in different ways.

The problem with these analogue signals is that they too are not perfect. A long cable will reduce the signal (like making it quieter). Electrical circuits to amplify the signal can also add interference (noise). Some cables can pick up background noise (interference) from other electrical sources.

The quality of the cables, connectors and electronics make a difference to the analogue signal. So, if you have a system to make sound (e.g. music), AKA "HiFi system", and it uses analogue signals, then the quality of the HiFi system and its components matter. Better (more expensive) systems produce noticeably better sound. So people made and sold more expensive better systems, and people bought them. No problem here.

But the world has changed. Even the link from a sound system to a speaker can be digital now, and in practice most people are listening to music that is streamed or downloaded and played on a device that makes the sound directly (via headphones, etc).

Digital signals work in a different way. Instead of using an analogy of the sound pressure waves, the waves are measured, and those measurements conveyed as numbers. Later, those numbers are converted back to sound pressure waves.

The reason for this is that numbers are easier to transport and store reliably than analogue signals. You can reliably communicate a number. "42" is still "42" if it is quiet. Ultimately these signals may be carried as 1s and 0s on a wire, and it could simply by one voltage for a 1 and another for a 0. In practice it is way more complex now, but there are systems that just use these simple voltages.

The result is that you have digital cables, carrying one or more digital signal (these 1s and 0s) between equipment. If the cable works then every single 1 and 0 that is meant to go down the cable gets there. It does not matter if the cable is not very good, if the voltage for 0 is a bit off, as it still comes out as a 0.

Such cables can be used for some direct signaling of sound or video, such as an HDMI cable to your TV.


This presents a problem for those that sell high end cables and equipment to audiophiles. How can they sell expensive gold plated connectors and silver loaded conductors to people if any old cable will do?

You get some amusing cases of gold plated connectors for optical cables. Optical cables use light to carry the 1s and 0s rather than electricity as this avoids interference from electrical noise nearby. Even so, gold plated connectors, something that could help on an old analogue electrical cable, have been seen!

Well, with digital electrical cables, they were quite cunning. When you transmit a digital signal you need to do something called clock recovery which means you extract the timing from the signal. If you have a really cheap cable, with distortion on the signal, you may not get the exact timing of the start or end edges of the digital signals. If you had really cheap and nasty electronics that did clock recovery on a per bit basis, that could affect timing and affect the timing of the sound generated. Timing is quite important and could, in theory, affect the sound you then hear (although these variations are way above the frequency anyone can actually hear). So, whilst the argument was incredibly tenuous, it starts with a gram of fact which made it harder to just shoot down in flames. Saying "theoretically, maybe, but outside human hearing" being met with "well, I can hear the difference", and the whole thing becomes subjective.

But, once again, things have moved on. Now we have audio that is played from audio files like MP3s, which are downloaded or streamed.

It is worth pointing out that, in the digital era, there are things that make a difference. The quality of the original audio source (the recording studio), and the digitising equipment (microphones, ADCs, etc) all make a difference. The final quality of the playback systems, DACs and headphones or speakers make a difference. Both of these deal with the analogue and real world of sound ways, so that makes sense. Another big difference is the compromises made in making a digital signal, the measurement of those sound pressure waves. Ultimately you are throwing away some detail as part of the process. The different ways of coding an audio signal, and the data rates and sampling rates, all make a difference. This is why you can get difference quality formats, even just within MP3s there are different bit rates which sound different.

But when we finally get to the article I mentioned, with expensive high quality Ethernet cables, used for Ethernet for streaming audio, things get a tad special.

Here we are talking about a system that is not just about transferring bits, but a system for sending packets data - basically a system for sending a file of information which is the digital record of the music to which you wish to listen.

Ethernet cables do come in different grades and standards, and there are much higher standards for 10Gb/s cabling, which makes it expensive. But if you don't have a 10Gb/s network, then getting such cables does not help matters at all. For audio you do not need 10Gb/s, or 1Gb/s, or even 100Mb/s. A file transfer can keep up with real time at a few hundred kb/s. The only risk you suffer is streaming audio in real time is if the link is bad enough and slow enough (e.g. a really dodgy ADSL line) that the data transfer cannot keep up and the audio has to stop playing to buffer more data. This is a pretty clear audio failure which is not subjective, and not something that "better ears" can hear when others cannot.

Ethernet cables transfer data very fast, and there is no issue with clock recovery as that is all part of the protocol and the switches and Ethernet controller chips. None of that feeds in to the process of playing the sound in any way. Also, ironically, even a non fully working cable, i.e. one on which there are errors, will usually not cause an issue as the data is resent if incorrect as part of the various IP based protocols. So you have a case where you don't just need an "up to spec" working cable, you could even have a "really dodgy" cable, and still get perfect audio. If there are no pauses for buffering, the final data getting to the audio playback is the same, it is the same 1s and 0s, so it sounds the same!

There is no way whatsoever that a better Ethernet cable can ever affect the "quality" of a streamed audio playback, simple as that. What could have an impact is the quality of the sound card, and speakers or headphones, and the person's ears.

There is one other small point you see in that article in the pictures, which is direction arrows. This was one of the very special ideas that audiophiles had, and dates back to speaker cable and analogue signals. The idea was that somehow you could "condition" the cable to work better in one direction, and this was done by the manufacturer of the expensive cable. You should therefor always connect the cables the right way around with arrows from amp to speaker.

Well, this was always bullshit. Cables don't work like that, and neither do electrical signals. In practice the arrows were almost certainly there to avoid mistakenly connecting amp output to amp output and causing serious problems as a result. Having arrows and following them avoids that. But people started to believe they had some electrical signficance.

What makes this extra special is the use on Ethernet cables. The way Ethernet works is that signals go both ways anyway. Even if the bullshit was true, that a cable could be conditioned to work a certain way, the fact that Ethernet sends data both ways defeats that totally. They are just for show as connecting an Ethernet cable back to itself does not cause fires.

There are other factors when buying cables, like: The colour - is it one you like, or does it fit a colour scheme you are using for cables? How robust the cable is - is it likely to break or wear out somehow - can the cat chew through it? Some cables have nicer features like guards for the cable clip to stop it snapping off. These are all important, but if it meets cat5 or whatever standard you need for your network, then electrically, it does the job, end of story.

The one final thing I would mention is confirmation bias. This is a real thing - it means that when you expect a certain result, you inherently have a bias in your assessment of subjective matters towards supporting that outcome, and are dismissive of any counter evidence. What this means is that if you go out and spend £255 on a 12 metre Ethernet cable, and use it on your HiFi, you will be convinced that it sounds better.

So please, don't fall for bullshit.

32 comments:

  1. The thing about analogue cables is that if they *were* directional, you'd get the same amount of distortion no matter which way round the cable was, what with the signal being AC and all. Intentional directionality in an audio cable would be a stunningly bad plan.

    Oh yeah, and exponential backoff. When your protocol intentionally includes random delays which the next layer up has to account for anyway, exactly how bad would a cable have to be to cause issues on top of that?

    ReplyDelete
  2. Always loved to read the "audiophile" cable review, in particular SPDIF (the digital cable or fibre that connects CD player to digital amplifier/processor) and how this or that particular cable enhance the stereo image (I didn't know cables included DSP).... ROFL but not only audiophile are affected, I remember seeing Belkin gold plated serial/usb cable to guarantee the higher speed your modem can produce.
    The other audiophile good one, while not in the digital domain, is the power cable and how the last meter of cable will have such a massive impact on the sounds, while the fact the electricity provided by utilities is all but constant and the 500 miles before reaching you wall plug are on "cheap" non directional gold-platinium-fairy dust alloy seems to be completely ignored.

    ReplyDelete
  3. Only thing there is there are cables that are so crap even the digital signal gets distorted to the point the other end can't work out if it was a 1 or.a.0 seen a few very cheap hdmi cables fail to work at 1080p

    ReplyDelete
  4. The way to avoid confirmation bias is to do the testing as Double Blind where neither the person listening nor the person operating the controls knows which cable is in use. They just know that they've switched in A or B and the listener says whether they prefer A or B. Then at the end they check which A and B were (which will need to have been determined by a third person). The trouble is, setting up a double blind test without making a mistake that reveals some data is quite hard. It may also require special equipment to perform switching that isn't readily available. Then you have to consider the impact of any additional switching which allowed the test to be double blind.

    The other problem you get is smug audiophiles who insist they don't need to do double blind tests because they're above all this "confirmation bias" rubbish. Or you get people who insist they've done a proper test but don't describe the test, and I have to browbeat them online for some time to get their actual methodology which turns out to be flawed (ie. not double blind), I point this out and they get all upset and say "I don't care, I know I'm right and anyway I'm above confirmation bias". Sometimes you just have to walk away.

    The other one that gets me is HDMI leads which improve colours, redder reds etc. There are specs that state exactly what colour a given digital signal on HDMI should be rendered as, and if the HDMI lead changes that (eg. to produce redder reds) then the lead is not transferring data to spec and hence is faulty.

    ReplyDelete
  5. On a related topic I like this web site which points out that 24/192 sample rate music is not only of no benefit but actively harmful to sound quality due to ultrasonics aliasing in the playback chain and producing audible interference as a result. They construct a convincing case that for playback 16/44.1 (or 16/48 on DVD or Blu Ray) is as good as will ever be needed. Note that mastering and mixing is different, 24 bit really helps there to avoid accumulation of digital noise during mixing.

    http://people.xiph.org/~xiphmont/demo/neil-young.html

    I also like this site's set of test tones. You can check both your speakers and your hearing (not to medical standards) on this site:

    http://www.audiocheck.net/

    Both sites also cover aspects of sampling theory and how most people don't understand it.

    Note my interest is in lossless codecs eg. FLAC, WAV etc. I'm not that interested in MP3 or AAC since they throw real data away and can never reproduce the original signal exactly. With today's networks and storage sizes I don't see the point.

    ReplyDelete
    Replies
    1. That is much more of a factor in audio quality than fancy cabling, for sure. When you hear a FLAC file and an MP3 of the same recording played back to back, the quality difference is readily apparent. Most people today fail to appreciate this fact, or even care. How many songs you can get onto a drive is all they seem to care about.

      Delete
    2. Well most people have $20 speakers and headphones so you wouldn't hear a difference between mp3 or flac anyway.

      Delete
  6. I live near an expensive cable manufacturer. There web site promotes one of their cable as it has some important electrical property for good sound reproduction (let's say 'low capacitance' as I can't remember what it was). They also have an even better cable, which has higher capacitance than the cheaper cable. Or in other words, the feature that makes a cable better is worse on their "better" (and of course much more expensive) cable. My Hi-Fi shop was always horrified that I connected my expensive amp to the speakers using 79 pence/metre from my first ever system :-)

    ReplyDelete
  7. One funny notion I came across among audiophiles was using "medical grade" power outlets for their audio equipment. Of course, once you check the specs which make a socket "medical grade", you find it's a regular socket but with a higher insertion/extraction force, to make it harder to pull the power cord on important medical equipment out of the socket by mistake.

    Owen: I recall calculations about 24 bit audio pointing out that for the 17th bit of resolution to be audible, you'd have to be playing your audio signal at some ridiculous volume (somewhere between 'aircraft takeoff' and 'pneumatic drill' IIRC) - at which point, of course, either the necessary ear protection or the resulting hearing loss will have a far greater impact on your enjoyment than the extra bits of resolution.

    ReplyDelete
  8. I will add:

    1. In the analog domain more expensive cables and connectors are more often than not placebo (the two exceptions are when you're trying to send power down a cable with insufficient current carrying capacity, and when the noise isolation of the cable is insufficient. The latter is the reason why professional microphones use balanced XLR connectors - weak signals and long lines - in addition to the other reason of ruggedness)

    2. With few exceptions, an XW amplifier is an XW amplifier. The exceptions are those which have been intentionally designed to modify the audio signal (Vacuum tube amplifiers all fall into the latter category by the nature of vacuum tubes themselves). Amplifier design is well known, and designing a transparent amplifier relatively easy.

    (This is not to disparge those rare amplifiers which intentionally modify the signal - some people prefer their sound.)

    3. Modern DACs are much of a muchness, though obviously this depends somewhat on the quality of their power supply. Motherboard integrated sound chips are often major culprits here, having poor shielding from the digital hash the rest of the PC is producing.

    If you want a perceptual improvement, then provided your amplifier is capable enough and your DAC isn't absolute crap, the best bang for buck you can get is better transducers - speakers or headphones. They are far and away the weakest link in the chain.

    ReplyDelete
  9. Those people who make that expensive Ethernet cable have a hilarious website.. This is great:

    DIRECTIONALITY: All audio cables are directional. The correct direction is determined by listening to every batch of metal conductors used in every AudioQuest audio cable. Arrows are clearly marked on the connectors to ensure superior sound quality. For best results have the arrow pointing in the direction of the flow of music. For example, NAS to Router, Router to Network Player.

    Are they seriously saying that 'listen' to rolls of Ethernet wire to work out which way round it should go? This has to be utter tosh and something advertising standards should be involved in.

    These purveyors of snake-oil even sell special USB wires... Really..

    ReplyDelete
  10. This is the business to be in....

    http://www.futureshop.co.uk/audioquest-diamond-rje-ethernet-cable-15m-p-5848.html#.UtrDzXk4n-Q

    Seriously, this is just fantastic. Do people really buy these things?

    Do you think if I buy this and use it to connect my routers it'l make ISIS converge faster, or perhaps speediest.net will speed up?

    Perpaps.. If I buy these and connect to my home router with it the porn will look more real?

    Check this too:

    All insulation slows down the signal on the conductor inside. When insulation is unbiased, it slows down parts of the signal differently, a big problem for very time-sensitive multi-octave audio. AudioQuest’s DBS creates a strong, stable electrostatic field which saturates and polarizes (organizes) the molecules of the insulation. This minimizes both energy storage in the insulation and the multiple nonlinear time-delays that occur.

    But it isn't carrying audio, it's carrying Ethernet frames. So their cable will prevent signal slowdown, and that will achieve what? (even if it did...)

    It's just absolute bullshit. So I'm going to talk to the ASA and see what they say.


    ReplyDelete
  11. I once saw some speaker cables on Amazon at about 1000 dollars per foot or something silly, which had a 5KV electric screen around the cable as their snake oil. One of the reviews was delightful, about how when playing silence they had listened too hard and heard what man was not meant to hear, and now even with their eyes shut and the earplugs in they can't stop the noises, oh the hideous noises from beyond the void (etc. Cthulu etc.). The "was this review useful to you?" score for that review was incredibly high. It was a great piece of writing, whoever wrote that review should be writing professionally. Alas I can't find it now.

    ReplyDelete
  12. Arrows on audio cables are a good idea, but, as you rightly say, do not mean the cable is somehow "directional". They'd be very useful when connecting a recording device, as you could have them connected so that the arrows matched the signal direction - when you've got two channels each way, the possibility of mistakes is high. There is no other meaning for them, though.

    I've bought fair quality audio leads for the last few years, because of robustness, rather than perceived sound quality. Sometimes I used to make my own leads, especially when I needed unusual connector combinations.

    As for "double blind" testing of cables, the best way would be to have someone working behind the equipment to swap them over, so that there's no switches involved. This person would need to keep their work secret, of course - so that the testing really was as blind as possible.

    Could we get the snake oil salesmen interested in trying out an oxygen-free listening room? (haha!)

    ReplyDelete
    Replies
    1. Actually, to be fair -- there IS an aspect in which end of cable (hence directionality) can matter -- shield grounding practices.

      Depending on the devices at both ends of the cable etc. -- can be 'correct' to have shields grounded all at one end in cables with independent wire also carrying the 'ground' internally -- think 'star grounding' practices, e.g. problems dealing with equipment where signal ground is directly attached to case ground, ..... Also consider
      carefully balanced vs. unbalanced audio, etc.

      I don't know all the details, but this CAN be a reason to choose a particular end of a cable, not directly anything to do with "flowing" correctly. But then the cable could just be labelled with shield-grounding-end on it, rather than ''arrows''.

      Delete
    2. Actually, to be fair -- there IS an aspect in which end of cable (hence directionality) can matter -- shield grounding practices.

      Depending on the devices at both ends of the cable etc. -- can be 'correct' to have shields grounded all at one end in cables with independent wire also carrying the 'ground' internally -- think 'star grounding' practices, e.g. problems dealing with equipment where signal ground is directly attached to case ground, ..... Also consider
      carefully balanced vs. unbalanced audio, etc.

      I don't know all the details, but this CAN be a reason to choose a particular end of a cable, not directly anything to do with "flowing" correctly. But then the cable could just be labelled with shield-grounding-end on it, rather than ''arrows''.

      Delete
    3. Actually, to be fair, there is an aspect that the cable orientation can make a difference -- shield grounding practices.

      I don't know all the details, but consider in particular unbalanced feeds going between devices some of which don't separate the signal ground from case ground, etc...

      'star grounding' practice may suggest you should have the cables which contain separate ground conductor from shield, and connect the shield to the ground contact only at one end -- which could then be marked, rather than doing the whole arrows thing, helpful as that can be for some use-cases.

      I don't know all the details, but I understand this to be a believable/real point, which is not so much about direction of travel even if the cable-orientation can make a difference =).

      Delete
  13. When people post links in comments they don't appear as clickable, which makes it a faff to follow them. Why aren't they clickable?

    And why has this page changed to an orange background colour scheme today?

    ReplyDelete
    Replies
    1. Blogspot is what it is, sorry. The colour was just time for a change :-)

      Delete
  14. I think Part 2 of the review is best:

    http://www.the-ear.net/review-hardware/audioquest-ethernet-cables-pt2-ethernet-cable

    I'm obviously throwing out all of my CAT5 right now.

    ReplyDelete
  15. Linksys switches really distort the sound of MP3s because they're cheap. For even better sound quality, you need a higher quality switch that's 100 times the price, supports QoS, and connect your equipment to it. You should also buy a NAS which supports setting CoS bits of Ethernet frames, and then run the whole thing on /31s so there's no IP broadcast or network addresses to get in the way and use up your valuable 1Gbps of bandwidth.

    Believe me, you haven't heard the crisp quality of an MP3 until you've streamed it over a LAN using a /31.

    ReplyDelete
    Replies
    1. Surely you mean a /127 ?

      Delete
    2. Nope - IPv6 increases latency by having an unnecessary number of bits transferred simply to identify the two devices sending data. This increases jitter.

      Delete
  16. Is the guy just completely unfamiliar with the technology he is using? He said he is a network admin, please, keep him well away from anything I ever use.

    “A strong, stable electrostatic field which saturates and polarizes (organizes) the molecules of the insulation. This minimizes both energy storage in the insulation and the multiple nonlinear time-delays that occur.”

    So he has a NAS and likely using SMB or something to transfer IP datagrams. The media player will buffer the data before it starts playing. So... Multiple non-linear time delays?

    ReplyDelete
  17. Can someone explain something to me? I bought a 20 dollar hdmi cord, but later bought a cheap 5 dollar one, and switching between the two you can very clearly tell that the picture looks very discolored. It's not even slightly, the shades are very different.
    If all that matters are these bits are getting to the destination, then what is going on?

    ReplyDelete
    Replies
    1. Seriously, that makes no sense - the colour is based on the values which are sent as a binary code - distorting those would be a challenge as you'd have to only affect the low bits to still have anything vaguely like the right colour. The sRGB data is serial, so affecting only the lower bits would be somewhat clever.

      Delete
    2. There are three signal pairs in a HDMI cable; TMDS Data0, Data1 and Data2.

      Control signals are encoded using a very robust 2b/10b encoding (2 bits of data sent as 10 bits on the wire). Video data is sent using a less robust 8b/10 encoding. Audio is sent with extra error correction on it using video encoding - audio can thus be carried successfully down a link that's corrupting video.

      Different data gets sent down each of the three channels; in RGB mode, Data0 is B, Data1 is G and Data2 is R, while in the YCbCr modes, Data1 is Y, and the colour bits are on Data0 and Data2.

      If your cheap cable is just in spec on one channel, and just out of spec on the other two, you'll have audio coming through OK (the error correction fixes it). One of the three video channels will come through fine; the other two will be corrupt, and you'll see discolouration.

      Without a photo, I can't make a good guess at what's going on.

      You could experiment - what happens if you change resolution on the source device? My guess is that if you drop down to SD resolution (576p or 480p), you'll see everything is fine, and that it only goes wrong as you go to higher resolutions.

      Delete
    3. Simon - thanks for the detail, as ever. Yes, does sound plausible. Just to be clear, this is a case of a cable that is not actually good enough to carry the 0s and 1s properly - the issue here is between cables that do, and cables that do and are gold plated and expensive. There is always the chance of proper dodgy cables.

      Delete
    4. There is a minimum quality of cable below which the error rate is unacceptable but frankly if your cable isn't meeting that you're either in close proximity to something arcing or it's so shoddy you should take it back to the shop. Once you reach the level of "good enough" to allow error-free data transmission you can't improve beyond that no matter how good the cable is.

      Delete
    5. The key here is that there's no subtlety with a digital cable - either it works well, or the faults induced are obvious and disruptive even to the untrained. This is the so-called cliff edge effect.

      In contrast, it is believable that an analogue cable causes a subtle difference that's difficult to spot in a brief check - better transient response, say.

      Delete
  18. Hallelujah! Voice of sanity.

    ReplyDelete