Naked Science Forum
Non Life Sciences => Technology => Topic started by: Gregory Anderson on 05/05/2010 10:30:03
-
Gregory Anderson asked the Naked Scientists:
Hello Naked Scientists!
Absolutely love the show (http://www.thenakedscientists.com/HTML/podcasts/)! I always listen to your podcasts (http://www.thenakedscientists.com/HTML/podcasts/) while I work away in the lab here in Toronto, Ontario, Canada!
Anyway, onwards to the question. I always hear people who consider themselves to be "audiophiles" say that the sound quality of vinyl records (which produce analog sound waves) is much richer and fuller than CDs (which produce digital sound waves).
I'm of the opinion that the human ear cannot detect such small sound wave patterns, and thus, it makes no difference.
Has anyone ever looked into this to determine if the human ear can tell the difference between a record and a CD?
Thanks again Dr. Chris & gang!
Greg
What do you think?
-
Greg,
IMHO yes the human ear can hear the difference.
Firstly, you must remember that the sound that reaches the ear, and the electric signal driving the speakers is always analogue! The information from a CD goes through a Digital-Analogue Converter which transforms a series of steps into a smooth curve (the mad audiophile has this DAC in a seperate box to avoid contamination from the actual motor that spins the CD).
Secondly, I was of the opinion that a lot of the ideas about musicplayers were mere pretention until I was invited to listen to good music reproduction equipment in the test room of an audiophile shop. the differences are marginal and easily swamped by outside interference; but they are there. Until you are prepared to spend silly amounts of money on equipment the differences are not appreciable. Unless you are prepared to invest in pricey interconnects, state of the art amps and pre-amps, expensive speaker wire etc; then the difference is un-noticable
A final point is that it is much easier to hear the differences in a relative test setting - ie when you listen to the same recording through different systems one after the other. I could not walk into a room and tell whether the sounds from the speakers were analogue or digital sourced; but changing from one source to another, I can.
Frankly, the best bet would be to pop along to any decent audio shop with two versions of a recording you know well and ask if you could try out a set up with both LP and CP. I'd bet you could hear the difference!
Cheers - Matthew
-
If the digital processing is done to a sufficient standard then the reproduction is better than the best analogue systems. The ear may tell the difference because the analogue will have more background noise or more distortion products perhaps.
The problem is that not all systems are designed well and this does not always correlate with price. The digital bitstream from a sigma-delta D to A coverter is well out of the audio range and it should get filtered to a high degree to remove frequencies beyond the audio band anyway (typically >20kHz). This is sometimes not done so well because, as I said, it is outside the audio band. This is OK if the amplifier used is of high linearity but not so good with a poorer amplifier which can allow these high frequencies to mix down back into the audio band. The result can be audible artifacts.
So a good digital system with a good amplifier is going to generally be beeter than an analogue line-up. Speakers are a serious limitation but the same for both. This leaves the starting material - a CD vs vinyl or tape. Most recording is done digitally now anyway and the studio equipment is at 20 to 24 bit precision. This gives a dynamic range way beyond what the human ear could discern and far more than can be handled on a vinyl record (which use compression for most recordings). Again this may result in a different listening experience but there is no doubt that the digital system is going to be higher fidelity.
So, yes, you probably can tell the difference in a typical recording but there is no doubt that the digital recording should be, and generally is, better.
-
In short, when done properly, a CD reproduction should be as good as your ears - in terms of frequency response, signal-to-noise (a.k.a. dynamic range), linearity etc.
A vinyl records inherently has a more limited frequency response, limitations on amplitude especically of high frequency sounds, and limited dynamic range (background hiss and rumble).
When preparing a studio master for a vinyl release, it is likely to be 'tweaked' to fit within the limitations of the vinyl recording medium, (for a classical recording, boosting the quiet sounds to reduce the dynamic range), and maybe limiting other aspects of the recording. This tweaking may give it a perceptually different 'feel'. In general, the LP may be likely to be 'warmer' and less 'analytical' than a CD recording from the same studio master.
That said, the degree of tweaking required and the perceptual difference between a CD and a vinyl release will probably vary a lot depending on the type of music/recording.
Of course you could apply the same processing to a CD... but why would you? Something else you could try is recording a vinyl record onto CD, then playing back the CD. I think people would be hard-pressed to tell the difference between the CD copy and the LP original.
The listening environment will also affect the perception of any difference: it's probably most obvious on a pair of 'analytical' studio headphones, and least different on modest speakers in a poorly damped room with background traffic noise!
-
My ear can. My analogue recordings are so messed up with scratches and muck that I find them almost impossible to listen too now. Prior to going digital, I didn't think there was much wrong with them [;D]
I have tried digitizing some of them so I could clean them up a bit, but I gave it up as a bad job.
On a slightly more serious note (ha ha ha ha ha ha! - ahem, sorry) we should not overlook the fact that the high frequency response of the senior lug hole goes downhill rapidly. It's quite likely someone could pinch the tweeters from my speakers and I'd be none the wiser.
-
With digital its all to do with the number of bits used to encode the sound we hear.
Lots of people say that DAB is no good for music stations and that we should move - if we must have digital radio - to DAB+, this is all to do with the CODECS that each of these broadcast standards support. The latter in particular supports something called AAC+ which means better encoding of the music within the broadcast bandwidth available.
On those rare ocassions that I listen to music on DAB I do have a radio that is engineered enough to support DAB+ should it become available on this island so it sounds ok to me - but then I am 49 so maybe I wouldn't notice anyway.
The same device also acts as my "record player" as it picks up the music from my computer and plays it at me. It also picks up the Naked Scientist Podcast as it has far better speakers than those on my computer.
So is it noticeable? Probably - to the young - less so to us older folk - and it probably depends upon the quality of engineering of your reproduction device too.
Regards
Neil
-
I did copy my LP records onto CD, using my computer and a good sound card. Recording at 44.1kHz and then converting the resulting 600M plus wave files to MP3 gave opportunity to see how I can hear. Playing back the uncompressed audio sounded the same, and If compresses to 128K I could hear the difference, but could not at 256k. I then did the data reduction to MP3 at 320k, so that my ear would be the worst part of the chain, rather than anything else creating a audible distortion. Some later recordings were converted to Ogg Vorbis at 256k, they sound as good as the 320k MP3 but have a much smaller file size, plus some MP3 players have issues if the encoding is not at 128K, and either do not play, or just produce static ( hard coding the decode parameters in the firmware to reduce memory needed for decoding in some players).
One record was recorded at 48kHz, as it was a 16.66RPM record and I wanted to resample it so that it would sound ok, as my record player does not support this very old standard. I did the same for 78's played at 45RPM. The 78's were mixed to mon, and severely filtered to reduce hiss, and probably came out a lot closer to the original than they would on any player with a steel needle.
-
Before (http://pdf23ds.net/projects/compressor/snow-unc.mp3) and after (http://pdf23ds.net/projects/compressor/snow-c1.5.mp3) examples of the dynamic range compression Techmind was referring to (http://www.thenakedscientists.com/forum/index.php?topic=31382.msg308557#msg308557) can be heard by clicking on the links.
The compressed "after" version is similar to a vinyl recording and is audibly different from the original digital "before" version, (the quiet bits are louder).
-
With digital its all to do with the number of bits used to encode the sound we hear.
Just to reiterate (from another thread) the number of bits used in digital sampling just sets a resolution limit on the dynamic range i.e. the loudness, that is recorded (although note that the bit-depth mapping to dynamic range can mean that the most quiet of sounds may only use two or three bits of the total bit depth - if it's very quiet then the most significant bits will all be zero). The upper frequency limit is dictated by the sampling rate alone and is not dependent upon the bit-depth of the sample.
I'll just add that DAB, like MP3 and Ogg Vorbis, is a lossy codec and the bit-rate doesn't have a linear relationship with either the depth of the sample or the sampling rate. Neither DAB (in any of its variants), MP3 or Ogg Vorbis can be regarded as Hi-Fi (where Fidelity is the key word).
However, the primary issue, when comparing analogue and digital music sources, is not one of resolution, either in sampling depth, or directly, in the sampling rate, but is one of phase coherency.
According to the Nyquist-Shannon sampling theorem, any frequency can be accurately sampled if the sampling rate is at least twice the frequency of the sampled signal. So far, so good. However, frequencies higher than half the sampling rate are not just degraded but may be regarded as completely random; they're not just a bit wrong, but totally unreliable.
Because you can't vary the sampling frequency as you record the original music to ensure that you're always sampling at a high enough rate, because if you're sampling it you can only know the frequency after you've sampled it, a fixed sampling rate is used and frequencies above half of the sampling rate are filtered out.
The real issue arises because the filters needed to remove the frequencies that are too high for the sampling rate have an adverse effect on the phase of the signals passing though them, with the result that the phases of different frequencies are shifted by different amounts; although all the different frequencies go into the filter 'in-phase', when they come out of the filter they are no longer so.
This, and the quietness problem, are the biggest problems that audiophiles have with digital sound sources.
-
With digital its all to do with the number of bits used to encode the sound we hear.
The upper frequency limit is dictated by the sampling rate alone and is not dependent upon the bit-depth of the sample.
I believe Lee is 100% correct!
-
I think Lee is right in most of what he said though I would disagree about phase distortion being a major problem. The ear is rather insensitive to phase distortion and, in any case, high order digital filters used in the better systems are FIR filters and do not have phase distortion. There is usually much more phase distortion in analogue systems because the filters (either deliberately employed or because the equipment is naturally limited in frequency response) have a much greater phase distortion. The largest distortion contributor is due to the limits to linearity (i.e. the deviation from a straight line response of number of bits out to amplitude in). This results in harmonics and cross products of the input frequencies being produced. Again such distortion is also present, and to a higher degree, in analogue systems.
The biggest source of lack of fidelity are the loudspeakers which are common to both digital and analogue systems. The principle is to move a diaphragm back and forth to generate the sound wave. Although a speaker can be driven by a perfect signal it is inherently distorting. What is called intermodulation distortion results from, for example, two perfect sine waves of different frequencies being applied. This produces a pattern of harmonics because the two frequencies "mix". It can be imagined easily from thinking that the lower frequency is pushing the diaphragm back and forth whilst the higher one is superimposed on top. The lower frequency is frequency modulating the higher one because of the Doppler effect from the moving cone. This cannot be prevented but the effects can be minimised by having multiple speakers covering different bands, especially by seperating the low frequency woofers and/or having a large area of speaker for the required volume to minimise the movement of the cone, although this has other problems related to the cone not moving as a rigid body. So called electrostatic speakers, backed up by subwoofers are not a bad compromise but the distortion effects cannot be eliminated. Earphones can be much higher fidelity because the diaphragm's physical movement is very small at the low volumes needed. The only high frequency tweeter that fully overcomes this is something like the Fane Ionophone which modulates a stream of moving air. It cannot deliver much power without creating a wind though :-)
If you want the best fidelity then digital processing is hugely superior to analogue systems. The limitations on quality are imposed by systems that, for good reasons, have been limited by the need to minimise the cost and/or the amount of the data needed stored or transferred. MP3 and many other standards are good but are compromises to quality for economic reasons and because we want compact media.
-
One of the things I find astonishing is people still pay astronomical amounts for valve amplifiers and the like. Surely these days, a perfect 'map' of a valve's dynamic response can be made electronically - giving the warmth (or distortion) at a fraction of the cost.
Maybe someone with a greater grasp of audio electronics than I can explain why valves (aka tubes) still can not be emulated. Or is it just an aesthetic discern?
-
One of the things I find astonishing is people still pay astronomical amounts for valve amplifiers and the like. Surely these days, a perfect 'map' of a valve's dynamic response can be made electronically - giving the warmth (or distortion) at a fraction of the cost.
Maybe someone with a greater grasp of audio electronics than I can explain why valves (aka tubes) still can not be emulated. Or is it just an aesthetic discern?
I have a highly unscientific theory that they are paying more to satify their sense of smell than anything else. Valve (tube) equipment produces a very nostalgic aroma, and nostalgia is an extremely powerful emotion.
-
I have a highly unscientific theory that they are paying more to satify their sense of smell than anything else. Valve (tube) equipment produces a very nostalgic aroma, and nostalgia is an extremely powerful emotion.
Yep. Nostalgia can be an expensive hobby [:D]
Can think of ways to fake that valve glow, but not so sure about the smell?
Industrial-grade 50's dust anyone?
[;D]
-
...in any case, high order digital filters used in the better systems are FIR filters and do not have phase distortion.
But how can you use a digital filter without first digitising the signal? As far as I understand, there's no way of avoiding an analogue filter before the sampler. However, one of the ways to reduce the phase incoherency produced by this initial filtering is to reduce the steepness of the filter slope, which increases the bandwidth (albeit attenuated at the upper end) in combination with a higher initial sampling frequency to handle the increased, but attenuated, higher bandwidth. Once that's been done you can then start to use digital filters and down-sample to a lower rate without affecting phase.
The main attraction of valve (thermionic tube) amplification over solid state amplifiers is believed to be due to the nature of the harmonic distortion they produce. Both types of amplifier produce harmonic distortion but valve amplifiers tend to produce predominantly even order harmonic distortion whereas solid state amplifiers tend to produce predominantly odd order harmonic distortion, and in listening tests it seems that people prefer the sound of even order harmonic distortion to odd order harmonic distortion.
Valve amplifiers have been widely emulated, at least in the field of musical instrument amplification, and especially so for electric guitars. However, the analogue source signal from the instrument still needs to be sampled and digitised before it can be passed through the digitally emulated valve amplifier. This is less of an issue for musical instruments with an intrinsically limited bandwidth, like electric guitars, which sound pretty weird if played through full-range loudspeakers.
-
I've heard the arguments for valves, but the object of any good amplifier is to introduce as little distortion as possible. It seems some people prefer the distortion that valves produce, but these days there are probably much less expensive ways to achieve the same effect.
Mind you, if some people obtain satisfaction from spending ridiculous amounts of money to get a distorted signal, that's up to them. They probably also spend $200 for a bottle of Scotch.
-
But how can you use a digital filter without first digitising the signal? As far as I understand, there's no way of avoiding an analogue filter before the sampler. However, one of the ways to reduce the phase incoherency produced by this initial filtering is to reduce the steepness of the filter slope, which increases the bandwidth (albeit attenuated at the upper end) in combination with a higher initial sampling frequency to handle the increased, but attenuated, higher bandwidth. Once that's been done you can then start to use digital filters and down-sample to a lower rate without affecting phase.
I think you are speaking of anti-alias filters prior to digitisation. The way it's done is that the recording sampling is done at a very much higher frequency (it is a natural result from the use of delta-sigma conversion), usually in the Megahertz region. The analogue anti-alias filter prior to this then can be made easily flat and phase linear in the band of interest. Digital FIR filters (decimation filters) are then used to filter with near perfect precision so that the signal can be downsampled to the required frequency (44.1kHz, say, for CD systems).
Your reference to exchanging a flat frequency response with a phase linear response is right for an analogue filter (unless exceedingly complex). This is the differernce between a Butterworth filter (flattest frequency response) and a Bessel filter (most phase linear response). Providing the frequency poles are made high enough above the wanted band, the effects can be made negligible and even then these effects can be compensated in the digital domain.
-
The main attraction of valve (thermionic tube) amplification over solid state amplifiers is believed to be due to the nature of the harmonic distortion they produce. Both types of amplifier produce harmonic distortion but valve amplifiers tend to produce predominantly even order harmonic distortion whereas solid state amplifiers tend to produce predominantly odd order harmonic distortion, and in listening tests it seems that people prefer the sound of even order harmonic distortion to odd order harmonic distortion.
Can you explain why there is this difference?
I have a pretty good idea how a transistor works, but valves I'm less sure of.
I can't understand why bipolars or FETs would intrinsically create any harmonics if well biased and matched.
-
Valve amplifier stages tend to be relatively simple (class A) with fairly low gain and not much enhancement of the linearity by use of negative feedback. The valve has a naturally more linear response, and so this works, but produces a transfer characteristic which is not symmetric about the zero signal point. This introduces even harmonic distortion (it adds extra frequencies at 2x, 4x, 6x ... the input frequency and also mixes two or more input frequencies to produce frequency related to their sum and difference).
Transistors are much less linear but are cheap and so more can be used and amplifier stages built with much more open loop gain. The amplifiers use negative feedback to improve the linearity considerably and the overall response can be made much better than that of a valve amplifier. A poorly designed amplifier (of either sort) will not be very good but a sensibly designed transistor amplifier will be better. The output stage of a transistor amplifier is typically class AB (though it does not have to be if you dont mind employing huge heatsinks) and is symmetric about the zero signal point. Any distortion from this will be predominantly 3rd harmonic although should be only noticeable if overdriven. The distortion components are different (a single frequency producing odd harmonics 3x, 5x, 7x etc) but it would be debatable as to which are more acceptable - I think it depends on the music. In any case, I think the biggest market (perhaps strangely) for valve amps is for guitar amplifiers where it is often intended to introduce odd harmonics with the typically used distortion techniques being applied (nowadays using DSP technology and digital processing).
A valve's harmonic behaviour is not very dissimilar to that of an FET, peppercorn.
-
I wasn't sure why valve amps produced even order harmonics and transistor amplifiers produced odd order harmonics either - thanks for the explanation graham.d.
-
Wow - thanks a lot everyone! Clearly there are some real experts out there! You've all certainly cleared up this question for me (and given me WAY more to think about now!).
-
Here's a god article:
http://stason.org/TULARC/entertainment/audio/general/11-13-Do-tube-amps-sound-better-than-transistor-amps-FETs.html
Also:
TUMOS, the MOSFET power amp module with real Tube sound
http://www.dddac.de/op06.htm