Digital gaps

February 28, 2019
 by Paul McGowan

It’s common wisdom that because digital breaks up sound into bits there’s missing information between those bits—information not lost with analog. But is that accurate?

The comparison between the uninterrupted straight line picture of analog and the chunky digital copy might lead us to imagine differences that don’t necessarily exist.

If I were to side with the measurementists I could pretty definitively demonstrate there is nothing measurably missing in a proper digital recording.

If were to then take the opposite side and agree with the analog proponents that correctly point out the audible differences between digital recordings and analog recordings, we would then be at a stalemate.

If nothing is missing in digital what explains the differences in sound quality?

We can say with absolute certainty that a PCM recording of a live feed sounds different than an analog or DSD recording of the same event.

Yet, it is also true that a PCM recording of the analog playback is nearly indistinguishable from the analog playback.

Tomorrow I am going to suggest what might be going on.

Subscribe to Paul's Posts

47 comments on “Digital gaps”

  1. As true as those statements are, I never got, what the ability to exactly reproduce a by an analog concept played back recording has to do with the ability to reproduce this result by a digital recording/playback itself (in case it would be desired at all).

    It’s two very different qualifications imo, just like a digital camera can make an exact image of an analog photography but not produce the same photo itself (except trying to do this with digital effects).

    I like both for what they are. The digital ability to exactly record vinyl playback doesn’t change anything regarding the different approaches.

    My guess for the reasons for differences between them is, that both add and leave out or veil some information differently and to a different extent.

    1. But I certainly get the message that if digital can exactly reproduce an analog playback, it should also be able to exactly record and reproduce a live event. And this since a long time.

      It’s indeed interesting what’s going on, why digital improved so strongly since the days it was already capable of recording an analog playback more or less exactly and how this connects.

      No doubt analog isn’t as accurate as digital…but what’s or what was missing from digital….or isn’t there anything missing at all and some just need what analog adds as a compensation for other losses within the chain from recording to playback? Or just for their preference? The label 2xHD even puts an analog step inbetween the digital media production process…seems strange.

    2. “Yet, it is also true that a PCM recording of the analog playback is nearly indistinguishable from the analog playback.”

      Nearly indistinguishable appears to be a relatively small quantity, or is it?

      When you think about bit depth, 16 doesn’t seem like a big number compared to sampling an analog signal 44,100 times per second while recording it which appears to be a really big number, or is it?

  2. Oops. No differences in sound quality between different formats and different reconstruction filters needed for PCM? What about differences between ladder DACs and sigma-delta designs? The inherent weakness of a digital recording is the fact that it will rarely catch the peaks of steep transient signals. Thus there is at least a loss of dynamic range. No chance to reconstruct these peaks.

    1. A fast sampling rate, say 192khz, will catch all but the briefest transients. Digital enthusiasts say that this does not matter, because such transients are well above the human hearing range. I agree with them, but I would, wouldn’t I?

      1. The very limited data I’ve researched would suggest humans can detect very tiny transients sounds, bottom line is arguably PCM even at 384 sample rate is still not quite fast enough to grab them, however DSD even at 64 will capture transients less than 1 microsecond. Is this immediately detectable? all the evidence would suggest not but it’s listening to music for long periods and the possible benefit of to this that I would like to see explored with more vigour. My opinion and that of many far more experienced people I have spoken to agree there is more of an ‘ease’ to listening to recordings made at very high sampling rates.

        1. Direct quote from from AESC paper 5931Author Wieslaw Woszczyk
          “In some musical instruments, acoustic pressure builds up extremely fast during onset transients reaching tens of dBs within a few microseconds. For example, transient onset of xylophone shows waveforms with rise time of less than 10 μs with instantaneous peak output reaching 126dB SPL. Trumpet playing forte can register 120-130dB peak SPL with steep rise of the waveform within only 10μs to full signal level. Snare drum reaches 130dB and cymbals 136dB peak SPL within microseconds [9]. Rogowski ‘s values were measured during a short musical selection captured using 1/4 inch, Brüel&Kjaer 4135 microphone and a 192kHz, 12bit A/D conversion.
          Can CD-rate sampling of audio every 22.7μs register the full waveform detail of the sound of these instruments? Based on these onset requirements, to achieve a transparent recording medium, one should sample audio with less than 1μs between samples to accurately capture steep waveform changes.”

          As said my research is limited….

      2. And where are these “fast” transients coming from? Fact is they don’t exist outside of test signal generators. We have a 15psi atmosphere more or less. Show me an acoustical device that can produce a sound pressure transient that exceeds even 44.1 sampling capabilities! This is also the farce behind TIM distortion. Sure it can be easily proven with simple test signals but it never happens with real program material due to the slow risetime of natural sound.

    2. If we can’t catch these steep transients in audio then how is it possible we can digitize video and RF and other analog signals at hundreds of mhz, even into the ghz range? We have had Ghz capable digital sampling scopes since the late 1980s. And btw, no analog tape can capture fast risetimes as you claim either. The linear region of magnetic tape is very narrow. Why was Dolby such a godsend to the recording industry in the late 1960s? Because it allowed much more dynamic range by implementing reversible compression, well almost reversible. Digital recording on magnetic or any medium has none of these limitations

  3. „… just like a digital camera can make an exact image of an analog photography.”

    Interesting comparison but in my opinion not quite true. Since the material you are printing on are not the same for digital or analog photographs, that results will be different. Or how we photographers like to say: the sound of these images is different.
    Best Ulrich

  4. A feedback I got from a production responsible of the Winter & Winter Label was most unexpected as they are mainly doing digital releases and some dedicated analog produced vinyl.

    He said in spite of coloration and long time storage matters, analog for him and based on his experience is still the most lossless capturing technology. He would prefer do make digital media from analog tape. This certainly offers a lot of arguing potential for many.

  5. I too shall look forward to Paul’s continuation tomorrow. I have my own explanation which I give to people who feel that digital sampling must make the music granular. Unfortunately it involves trying to explain the underlying principles of Fourier Transforms (without naming them as such), and normally goes down like a lead balloon. I shall be interested to see what sort of a fist Paul makes of it.

  6. The thing that has really hit home to me is the massive difference between the way analogue and digital capture work. Those stupid and visually inaccurate ‘stair-step’ diagrams that still constantly get wheeled out don’t help. The continuing debate over hi res vs cd quality standard makes fascinating reading if only for the total ignorance it shows up. It seems accurate to say that a PCM 96/24 recording will contain all the frequency capture and dynamic range we could possibly want but the interesting point of debate is that at that sampling frequency maybe we still are not capturing those tiny microsecond transient effects that are really important for healthy listening? DSD even at its basic SACD level will do whilst DSD 256 blows it out the park…
    A digital capture is, it seems to me, a blueprint – an instruction manual that than has to be handed over to a clever device called a DAC that can read those instructions and literally and I mean literally build the electrical signals that are then amplified up to drive the speakers/headphones. So the starting instruction manual and the DAC configuration are going to have a major effect on the music ultimately ‘built’ by it all.
    PCM and DSD differ in the way they ‘write down’ the instructions, so again it stands to reason there should be a difference when those different instructions are followed and the music constructed by the DAC?

    1. Reading Ted‘s explanations are also most interesting, as he indicates that upsampling has many more benefits and influences than just rising the granularity of replicating the signal…and he also explained that this reverses at some point in a certain digital design and can start to have disadvantages…so reaching the maximum granularity is probably never possible, not only in terms of storage space (but probably also not most important)

      At least up to this point we all know how it improves the performance.

  7. If digital playback alters the original analog recording then it’s not reproducing it properly. A pure analog recording must still sound analog when converting to digital. If it’s sound character is digital then it’s not right. I still believe that digital cannot accurately reproduce a pure analog recording. That there are immeasurable things in analog that we do hear in analog playback that digital cannot play back properly. Yeah you can break up an analog straight line into 1’s and 0’s but can you accurately put it back together again? We know the ear can detect things that cannot be measured. There’s no better measuring device then the ear.

    There’s an immediacy or soul of the music missing when an analog recording is played back digitally that is there when played back on good analog equipment. It’s hard to describe but it’s there and it’s real. Analog playback sounds more musical more liquid and real to my ears where digital comes across as dry. More or less on certain recordings. In other words it don’t sound digital. The differences might seem subtle but to the trained ear they are not so subtle.

    Are all of us analog freaks just nuts? I’m not saying digital doesn’t have it’s advantages over analog. S/N ratio being the biggest advantage in my opinion. But that’s the non musical aspects of the sound. I would rather have something there in analog that shouldn’t be there then something missing in digital that should be there. Does that make sense to anyone? The analog versus digital debate is similar to the tube versus solid state debate or MM versus MC. It’s all about which sounds more musical. What trade offs can we live with.

    I would also like to understand more about SPARS codes and which ones actually sound better.

    I’m looking forward to Paul’s conclusion about WTF is going on here.

    1. By the way I listen to and enjoy both digital and analog playback. I don’t hate digital playback the way some tube lovers hate solid state. I’m just saying it’s not time to push aside analog playback as being inferior. And it too can get better with technology.

    2. Interesting points, I just do not see how a digital capture from an analogue recording or two recordings captured at source in both formats can sound the same because the whole process is so very different for each?

  8. IMO it’s not what’s missing between the two formats, it’s what’s being added wrt digital processing chain – namely group delay distortion. The human ear is very sensitive to temporal distortions and digital, because it is a sampled system, introduces group delay distortion.

    No one mentions this when they talk say “they measure the same” – they will not measure the same if you look at this form of distortion. That’s one reason why different digital filters sound different – they will introduce differing levels of group delay distortion (in addition to different noise shaping and pass-band responses).

    1. I cannot dispute this claim which would add an additional problem to digital but not eliminate the discussion as to whether the gaps create missing information. Analog has an inferior S/N ratio to digital but still people prefer it’s sound which is sweeter sounding even if the HF response is tilted up. If you tilt up the HF in digital it’s unbearable to listen to. Our ears bleed. Why is that considering digital’s wider S/N ratio? We tend to ignore the higher noise floor in analog as long as the music is non distorted or not missing information. In my opinion analog sounds more natural and organic. Gets closer to reality.

      1. Flatter frequency response and wider dynamic range doesn’t necessarily automatically mean more musicality or a better presentation of the soundstage. Paul talks about flat speakers sounding terrible because the sound needs to be tailored, well probably more so with a digital source. Flat speakers even with a very flat analog source don’t sound bad because all of the musicality is still intact giving us that beautiful linear analog sweet sound with that blooming layered presentation of the soundstage we love so much in analog playback.

        1. I also think analog playback on a good system sounds better played at high levels. The sound remains sweet and listenable where digital can be harsh, congested, dry and irritating at higher levels with listener fatigue setting in quicker.

          1. Not counting the source unless it’s an analog source from the instruments being played by the performers to the microphone to the speakers everything is mechanical and analog. The only time that changes is when it’s processed digitally in the non real universe. Music is analog. Digital can attempt to sound as close to analog as possible and when it does it’s considered a success and it brings with it the convenience of digital which we all love. Maybe we all just got lazy in the fast food era. Time saved is more important but comes at a cost.

        2. Speakers with flat specifications are rarely if ever flat. The measurements are a single curve, which means one microphone at one distance and one angle. If you move that microphone closer or further, left, right, up or down the speaker gets less flat.
          These less flat responses bounce around the room until they reach your ears.

          A typical spatial response distortion is baffle step. At a frequency equal to 4560/baffle width in inches the horizontal polar angle changes from 360 degrees to 180 degrees. This changes the on-axis response 6dB, so either the room response goes up 6dB or the on-axis response goes down 6dB, unless the speaker is flat against the wall like Allison Research or it is a bipole or dipole with sound radiating out the back.

          More complicated spatial fr anomalies are caused by baffle edge diffraction, driver size, cone breakup etc.

          1. If there are no phase anomalies there will be a nice presentation of soundstage. The drivers don’t have to be ruler flat for the instruments to sound linear as long as the coloration’s do not originate from the box. Box coloration’s are intolerable. The room isn’t going to be flat regardless of how flat the frequency response of the speakers are. Speakers that are linear and don’t sound good most likely have coherence problems or problems with box coloration’s that detract from the music and sound muddy and congested.

            1. “Soundstage” is most often an illusion that comes from baffle edge diffraction anomalies that project fr bobbles off-axis. I build dipole speakers with rounded edges for low diffraction, and they “image” less with a hole in the middle, and sounding better in L+R mono.

  9. Well and good, if as a music listener, I had access to the direct feed from the microphone, to compare. A nebulous sound, as described in words will never sound as sweet as the actual musical passage.
    So, we must use what we are given, The quality of the digital files which comprise my library run the gamut of superb quality to the almost unlistenable–a circumstance over which I have no control.
    What I can control is the way the music is reproduced. And that is only done by listening to the music I own, with the equipment that I could afford.
    When you reach a point of satisfaction and contentment with the system and stop concerning yourself about the next thing you need to make an incremental improvement, the true enjoyment of the listening experience can take me away, in my Headphones, to Audio Nirvana.

  10. Time for Electronics 101. I don’t know why I do this. I get no tuition money. Okay I’m going to give you an analogy a ten year old can understand if he is interested in getting a general class radio license. What ever happened to kits Paul? What ever happened to knowledge, curiosity? Kids playing with electronics or worked on cars? I think computers and the internet shoved them off a cliff. Anyway Paul will tell you that’s what American boys did when we were young, build radios, adjust carburetors, and create explosions in our bedrooms with Gilbert Chemistry sets. Outdoors it was sports sports sports. What do they do today? Sit on their asses in front of their computers on Facebook, Instagram, Twitter and get fat eating potato chips and drinking Coca Cola.

    Okay, here’s how an AM radio receiver works. An AM radio wave is broadcast from a commercial radio station on a carrier frequency anywhere from 540 khz to 1600 khz assigned to it by the FCC. The sound you hear modulates the amplitude of the signal depending on how loud the sound is at any given moment. The curve that connects the peaks is called the modulation envelope. That’s the signal they want to transmit and you want to receive and isolate. Remember it is only at the peaks of the carrier frequency that there is any information. All the time between those peaks are not continuous, they are empty gaps. When you tune your radio you are simultaneously adjusting two circuits at the same time, the carrier frequency of the station’s signal you want and an oscillator. That’s why an AM radio has two capacitor sections. The first one is part of an LC circuit that has a high impedance only at the carrier frequency you want. Then it goes to a mixer oscillator where the adjustable oscillator is simultaneously tuned so that the sum and difference frequencies appear. The difference is ALWAYS 455 khz. This difference is called the beat frequency or IF frequency. Then there is an IF amplifier stage that its tuned to 455 khz. Keep in mind the modulation envelope is still there and there are still gaps between each peak. Then the detector circuit separates the modulation envelope from the IF frequency. It consists of a diode, a resistor, and a capacitor. The diode throws away the bottom half of the carrier wave. The resistor and capacitor are tuned so that they are a precise match for the envelope modulation. Why does that work? Because the voltage cannot change across a capacitor instantaneously. It’s a filter, and integrator. An inductor in series and a capacitor in parallel are integrators. In an inductor the current cannot change instantaneously. So the capacitor in the detector has charge and therefore voltage that changes with the modulation envelope and fills in the gaps in precise timing. This is also what happens in a digital circuit where the output is confronted with the output of a DAC that has a series of pulses at different voltages just like an AM radio circuit.

    Modern integrators use op amps like the one in this example. Notice the capacitor in the feedback circuit. Note also that they are used in among other things “wave shaping circuits.”

    https://en.wikipedia.org/wiki/Op_amp_integrator

    This one is the same circuit as a detector in an AM radio minus the detector diode. You might have to copy and paste the full address into your URL address box. BTW, this one uses a PWM circuit as a source. Usually there a series inductor is used instead of a shunt capacitor.

    https://www.google.com/search?biw=1536&bih=723&tbm=isch&sa=1&ei=cv53XNjbJdHq_AaJ4boI&q=integrator+circuit+in+an+audio+DAC+output&oq=integrator+circuit+in+an+audio+DAC+output&gs_l=img.3…77088.86867..87202…3.0..0.126.3702.38j6……1….1..gws-wiz-img…….35i39j0j0i67j0i8i30j0i24.hTrJokzM1_o#imgrc=cHw4HfK6kyIo-M:

      1. Sometimes you can only hit your head against a thick brick wall to teach someone something for so long and so hard and then you realize it is much too thick for anything to penetrate. I think I had more success teaching my dogs tricks than I’ve had trying to explain anything about electronics or sound to you. So lets start with a simple lesson instead. Now SIT and I’ll give you a nice dog biscuit to chew on. SIT I SAID DAMMIT. No luck here either. You win, I give up.

    1. Excellent! Nice to have some well established electronic theory here. Also consider FM stereo. That too is a sampled signal albeit analog processing. Don’t hand any “missing gaps” there either!

    2. You took me right back to my school days soundmind, first crystal sets then on to the construction of a TRF receiver in a biscuit tin, progressing to a more complex (for its time) superhet receiver. Health & Safety didn’t really exist back then but the school feared I would either be electrocuted or burnt by hot tubes. On the subject of chemistry sets such chemicals available back then certainly would not be allowed in the modern (boreing?) chemistry sets. Thanks for the nostalgic reminder, I’m 75 now and had a wonderful career in electronics that I thoroughly enjoyed.

  11. Unfortunately, DSD is so niche in the recording albums, arguments for it are a bit academic.

    I’ve noticed some orchestral labels have switched to PCM for recording, which, I think, is going the wrong direction, but it’s reality.

    1. DSD recordings are great ..and indeed even many SACD are made of 24/96 PCM as we can read on the back side. However I enjoy the native DSD available. Certainly DSD doesn’t help if the recording done on that basis is of lower quality than other formats. Same with highres/redbook.

    2. Classical musicians and conductors can hear the difference of DSD. Record label engineers and producers typically can’t, because they spend their lives listening to studio monitor speakers instead of the music on the other side of the glass.

      Yes, there are phase coherent studio monitors. Earthworks, Dunlavy and Lipinskys have been around for a long time, but they have very limited dynamic range. I once heard a DGG Tonmeister turn up a 5.1 Lipinsky control room until the Doppler distortion and driver excursion clipping was unlistenable! Even such standard stalwarts as Genelec and PMC use 1″ domes for the treble drivers, which can’t reproduce the 18dB crest factor in the top octave of real music. (They also use non-phase coherent crossovers with temporal and transient distortion that negates the advantages of DSD). I only listen to AMT type tweeters because they have 14dB more headroom and five times the speed of the same aperture dome or cone driver.

      Symphony orchestras had to fire their record labels to record in native DSD because IT CAN’T BE EDITED. Splicing, mixing, overdubbing, EQ and added reverb are mathematically impossible. Sniveling major label engineers forgot how to get the sound right in a pair of microphones and producers read scores while recording to make sure every jot and Italian adjective is followed precisely using dozens of takes and hundreds of splices. This is flat wrong, scores are guidelines and music is living and breathing.

      Listen to the difference on LSO, Mariinsky and SFO SACDs on their own labels. I can’t stand DGG, Sony Classical, etc. they sound like the Barbie & Ken version of Classical music – smooth mounds where there should be cujones.

  12. Having at one time had an analogue system with an excellent Merrill Heirloom TT and Grado Signature tone arm. One big difference. I was using digital time delay for rear speakers. That allowed to hear what a real room would better sound like, and not simply hear sound coming from two front speakers. It was an eye opener experience. When hearing the illusion of a room it was played in, and not simply hearing into a room, what were once subtle differences in sound became apparent. It was as if I could ‘hear down’ the tone arm. That is the only way I could describe it. As if I were looking down a huge tunneled highway delivering the sound. Just damping a little on the cartridge and it was as if someone had covered the walls of the room the performance. Change TT cables? Change cartridges, etc? Analog is a sound of its own. Its not like real music. Digital gets closer. But, closer is not necessarily preferred. And, digital’s exactness leaves less room for certain errors. That is its real problem.

    You can listen to analog with inferior cables and not mind as much. And, that rubber surround holding the cartridge’s cantilever has a TREMENDOUS effect on “sweetening” the sound. Real music does not ride on a cushion. Some people prefer that soft cushy sound that they do not know has been introduced. But, if you do not mind? And, not seeking a more true realism? Then analog is what the doctor ordered. Its like people that must pour in the milk or cream so that can enjoy their coffee. Digital is for those who enjoy the taste of a great cup of coffee black. Especially with a little fennel seed ground in. In other words, analog is for some a preferred taste. Neither digital nor analog is the absolute reality, because we have so many variables involved… i.e. power cords.. AC quality… tubes.. transistors… speaker colorations… Must I go on? Simply put: When digital was introduced, a new flavor was brought forth.

  13. The last I heard, Iron oxide particles can only be magnetized in one polarity or the other. This suggests that “analog” magnetic tape really has a random sample rate.

    I suspect a lot of the degradation people hear in many digital recordings is from the cheap analog circuitry employed. The typical digital recorder’s analog stages (and power supply) aren’t even in the ballpark with the typical professional analog tape machines of thirty years ago.

    1. Magnetic domains are essentially analog with a little non-linear hysteresis. Magnetization curves are smooth until you get down to the atomic level, much like styli reading the bumps in PVC polymer chains. Even then, it is stochastic noise which ears are extremely good at ignoring from millions of years evolution around wind, rain surf and other uncorrelated background sounds.

      OTOH, quantization noise is completely foreign to our organic cognition computers. Think about the audibility of jitter at barely measurable levels. Further, quantization distortion increases with decreasing signal levels, which is opposite to analog systems. At -46dB, digital has 8 bits less resolution so Redbook sounds like an eight bit video game synthesizer chip. Live music can have 90dB of dynamic range.

  14. The first time I heard a direct DSD/Redbook comparison I was shocked, because I thought it was going to be subtle. It was a big quantum leap in realism. The situation was highly revealing – Thiel phase coherent floostanding speakers, a Bels tube amp and diffraction control of the room reflections – a ceiling absorber and diffusers at the first reflection points. What was even more intriguing was a room full of audiophiles reported little to no difference from the same experience – and no, I was not in the sweet spot.

    A little later I was treated to a scaled demonstration at the Audio Engineering Society Convention. Digital Audio Denmark made a recording feeding a mic pair at medium distance from a concert grand piano to a series of converters, so the reverb was purely physical and every resolution step was represented with everything else remaining the same. There was a large difference going from 16/44 to 24/96, a substantial difference going up to 192Ks, a clear difference between 192 and 384 and a subtle difference between 24/384 and DSD. The audition was through headphones that were not very supersonic, so the wider bandwidth of the conversion was not the answer.

    I was already convinced that Fourier mathematics was exceeded by hearing, and here was proof, albeit experiential and anecdotal. There was some scant supporting evidence from arcane research – von Bekesy had shown a just noticeable difference of 3 microseconds between left and right ears using Copper tubes and a spark gap, showing that hearing cognition had an effective time domain bandwidth at least eight times the frequency response.

    Finally I found this paper, which used conservatory trained musicians as test subjects as I predicted because von Bekesy was working pre-loudspeaker with acoustically trained listeners:

    http://phys.org/news/2013-02-human-fourier-uncertainty-principle.html

    This shows the same better time resolution in when notes start and stop than is theoretically possible in a perfectly linear system, and also that this hearing feat does not extend to the general public. Like so many cases in Biology, organisms “game” mechanical logic to achieve higher performance, but only after dedicated training.

    So, the answer is people who learn to hear music through reproduction can’t hear the improvement of HD media until after they “break in” their ears with phase coherent systems that transmit a higher information content, which is typically un-processed near coincident pair recordings (un-mixed and un-mastered) and headphones with extended bandwidth. My “break-in” to hundreds of acoustic concerts and daily doses of harpsichord sharpened my hearing beyond audio fanatics.

    1. The only certainty is that this link will be posted ever and ever again and those who contradict the thesis will never been shown here – it bores!

  15. I wonder if digital gaps create a harshness effect similar to scraping your fingernails on a chalkboard? It might explain why the straight line of analog is said to sound smoother to most people.

Leave a Reply

Stop by for a tour:
Mon-Fri, 8:30am-5pm MST

4865 Sterling Dr.
Boulder, CO 80301
1-800-PSAUDIO

Join the hi-fi family

Stop by for a tour:
4865 Sterling Dr.
Boulder, CO 80301

Join the hi-fi family

linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram