Another piece of the puzzle

September 21, 2022
 by Paul McGowan

I often tackle the same question multiple times. With each new answer to the same question, the different angle of attack seems to illuminate the lightbulb of understanding for at least one person.

An aha! moment. For me, that’s reason enough to keep trying.

We’re all perhaps a bit weary of me beating the DSD horse. I get it. But, I am also convinced that most of my readers don’t quite understand the difference between the two main formats: PCM and PDM. And I think it’s important to shine a light on it.

One of our HiFi Family members, Tony Plachy is a retired physicist with a gift for explaining hard-to-follow concepts.

Here’s from one of his comments:

“There is something very special about DSD256. I will try to explain what it does and how it does it. First, lets call DSD256 what it is. PDM ( Pulse Density Modulation ) which is a special case of PWM ( Pulse Width Modulation). I have been to seminars and lectures where notable mastering and recording engineers have said that the remarkable thing about PDM is that when the sampling rate is high enough ( and DSD256 is certainly high enough ) and you make a digital copy of analog music and the convert the copy back to analog what you get back sounds like the original. To say it another way is DSD256 makes exact copies.

I have a DSD recorder that i use to make copies of my best vinyl. It copies at DSD128 and the plays it back at DSD64. During the recording I can toggle between the copy and the original listening to the headphone feed from the recorder. To these old ears the mastering and recording engineers are correct.

How can this be? What does this happen with PDM and why does PCM ( Pulse Code Modulation ) seem to leave a digital footprint on the results? The answer is two things, one of witch I have all ready mentioned. First, the sampling rate must be high enough ( DSD256 or at least DSD128 ). Second, with PDM the amplitude of the analog signal is NOT ( yes, is NOT ) digitized. You can see this for yourself if you go online and find an article that shows the DSD output of digitizing a sine wave. Even if you are not an EE ( Electrical Engineer ) it should be obvious that all you need to do is use an analog low pass filter to get the sine wave back. Do not ever try this if you have the PCM output from digitizing a wave, all you will hear is horrible noise.

So does PDM do this for all music? The answer is yes, it does. To understand this you need to go online again and first look up a guy named Fourier and then look up how a square wave is made from a Fourier series. If you understand this it will be obvious that PDM can make exact copies of all music.

End of lecture, do your homework, this will be on the final exam! 😉”

Thanks, Tony!

Subscribe to Paul's Posts

36 comments on “Another piece of the puzzle”

  1. I understand and believe how good and different DSD is than PCM. I didn’t make the homework yet.

    Just for the sake of terms:

    How should „digitizing“ be understood?

    With DSD, I understand we record into a unit, that processes the analog signal in zeroes and one’s with various algorithms and then we play it back with a unit that does the same in the opposite way. Those ADC’s and DAC’s are “digital equipment”.

    I understand that the sine wave is stored and restored in a very different way than with PCM, but shouldn’t we use other terms, than “it’s not digitized”, if we store it with digital equipment in digital form on digital media?

    Another question:

    If DSD today stores and plays back the completeness of an analog signal in an indistinguishable way, what do you expect of future enhancements? If we really don’t hear differences with today’s best (or good enough) digital format and equipment anymore, why develop improvements from now on?

    Do you expect that in 2, 5 or 10 years there will be no audible improvement anymore? If you don’t believe that, how can todays statement be true, that there’s no difference?

    I’m convinced of DSD’s quality, but today’s claims somehow make me remember the 80’s.

    1. The term “digitize” means to sample an analog signal at a regular interval and then use that sampling to represent the analog signal. After reading what I wrote earlier I would change one sentence. Instead of “Second, with PDM the amplitude of the analog signal is NOT ( yes, is NOT ) digitized.” , I should have written that with PDM the amplitude is not captured as a digital quantity but as a density which is an analog quantity.

      As to predicting the future of what will be the format a decade from now, I will pass on that.

  2. Music is characterized by more or less sharp transients – never having any resemblance with a pure sine-wave. It should be most obvious that low sampling frequencies (as 44.1 kHz) will not reliably hit all sharp (!) peaks and dips at their extremes. The higher the sampling frequency the higher the probability to hit every extreme amplitude value of those peals and dips. Thus shouldn’t the question be which minimum sampling frequency is required to reliably catch the sharpest peaks and dips? I learned that 1 MHz should do the job for music signals. Thus we have to wait for PCM 24/1536 or will 24/768 do the job too? Another question then is: what is lost when DSD is mastered using DXD – a kind of PCM with a sampling frequency which isn’t resolving all peaks and dips?

  3. Well congratulations to our friend, Tony who made his explanation even more well known by getting a good headline in Paul’s post.
    To explain things of this nature is an accomplishment in itself.

    Thank you Tony & Paul.

  4. I wonder if Tony has heard PCM done right by the likes of Audio Note. May not be as exact a copy but if you listen for the soul and emotion in music, you’ll be happy knowing there are thousands and thousands of plain old Redbook CDs of great music to be had, some for a “song”.

    1. I have heard Audio Note digital gear and it does have a certain emotional quality to it. However, what I consider the best PCM digital available today is the ring DAC from dCS with its low bit quantization and its very high sampling rate.

  5. I believe in the advantages of Dsd256, even Dsd128, but what percentage of our musical heritage is available in this format, or can even be remastered from analog mater tapes to this format? I fear the format will remain like my treasured direct to disc vinyl of the 70’s, very very scarce and hideously expensive

  6. We frequently go thru all these explanations. We may as well be talking about measuring what comes out of our own systems and then equating those measurements to the sound we hear from our systems so we can move on from all these discussions and just read specs to determine what we’ll hear.

    This time the variable is the digital format. Of course then we have to agree on a physical DAC that is perfectly true to its input format, R2R ladder type or Delta Sigma type so right away there are lines that have to be crossed or blurred.

    While it’s interesting to know all the details, and people like Tony and Paul offer good explanations, what does it all mean and why should it matter? (For the end user it’s how things sound in their environment)

    According to people like Cookie, Gus, and Paul, DSD matters on the recording side.
    According to MOFI and others DSD matters for both sound quality and economics when remastering from tape.

    How really big is the sound difference on the playback side if something is recorded in 4x DSD – then distributed in 4x PCM? If the comparison is made, is it a true to the format comparison? Or is it a comparison of what format that DAC handles best?

  7. I need to give you your comment quite a bit of thought Mike. I never looked at the issue that you just brought up. There are so many issues that have been brought up just with this morning’s post that it reinforces my feeling to just listen to my music in whatever form I choose as long as it brings joy to my soul. That’s the bottom line always.

    1. Hey Stimpy2,

      The differences I hear in different playback formats are subtle at best.
      I’ve done things way differently than most, but have kept my comparisons consistent across all the DACs I have had over the years.

      So I tripped across converting redbook cd format, (ripping to AIFF and then converting that file to a 2x DSD file for playback). What I gained was a playback that I could listen to louder without any fatigue, even after continuous hours of listening. Since getting a DSDAC, anything above the the 44.1or 48KHz sample rate seems to make no sonic difference with pre conversion. In the end I’m left to conclude it has to do with the DAC topology and the way filtering, and analog section is implemented.

      The remastered or recorded in DSD files I have sound really good, so I’m left to conclude that the recording / mastering process probably has a bigger impact on sound than the actual format.

      Again it comes down to a true comparison of the format itself, and that’s almost impossible to do?

      1. As soon as you said the words mixing and mastering you struck note deep down inside of me. So the recording process in most cases is usually fine but it’s the secondary processes of mixing and mastering that hold the keys to the kingdom. Why not the actual studio or live recording techniques as being part of the means to an end as well? Perhaps the answer that I’ll get from you is that you’re always satisfied with your conversion process and you’ve concluded that it only has to do with the mixing and mastering of the original recording(?).

        1. Stimpy2

          IMO ….A lousy recording is lousy regardless of the format.

          I have some music recordings I really like but won’t listen to them on the main system because they’re not enjoyable. I save them for in the car or on the garage music system.

          I’ve reached the point where constantly listening for some detail or sound stage takes some of the enjoyment away. So rather than chase after a red herring it’s much better for me to become content with the way I do things and the sound I get. There’s always room for improvement. So when I find music I really enjoy and the recording is great, I consider it a bonus. Digital format is not a top priority in the music choice for me.

          1. Now we are in alignment with our thinking and how we listen to music with regard to the SQ versus listening to a bad recording because you enjoy that specific music. Makes perfect sense.

  8. When surfing around a bit regarding today’s topic, I found this statement, which seems a fact (?) as it’s also on Wikipedia.

    Can this be the reason why many find DSD64 sounding a bit less open than it could sound and is it a fact that with higher sampling rates than DSD64, the noise floor is so high, that the dynamic range of DSD begins to decrease at a much higher frequency than 20kHz?

    Independent of this context it may be said that DSD64 was called perfect with no audible difference to the input, too, from relevant persons…until DSD256 showed up.


    The dynamic range of DSD decreases quickly at frequencies over 20 kHz due to the use of strong noise shaping techniques which push the noise out of the audio band resulting in a rising noise floor just above 20 kHz. The dynamic range of PCM, on the other hand, is the same at all frequencies.


    1. I stopped spinning CDs when I learned that the tracks from CDs ripped to a cheap HDD recorder (playing these tracks from a huge RAM buffer) sounded much better than when played real-time from my expensive CD-transport (same DAC for both digital sources. However my vinyl records had PRaT lacking from all CDs! Despite the limited dynamic range of vinyl and despite of all these vinyl associated distortions and mechanical problems and poor channel separation thus there must have been fundamentally wrong with”digital”. I also never got significant improvements in sound quality from high resolution formats. Many recording engineers I had asked confirmed my findings having sent some demo tracks with different resolution/sampling frequencies. Obviously there are many unsolved problems with “digital audio” – galvanic isolation is the new hope for a better result. Earlier minimizing jitter was claimed to solve all problems.Then the digital interconnect was a crucial element. Then an external master-clock. Not to forget most effective shielding measures. Hey, what is going on here? What are the basic requirements for the quality of the digital components in order to get a significantly better sound quality from high resolution PCM/DoP?

  9. No thanks.

    As one of many people whose analog is converted to PCM (40/384 in my case) I don’t hear anything that I could describe as a digital footprint.

    I far prefer the benefits that PCM provides, mainly DSP, which you can’t do with DSD.

    An explanation telling people to go online and look up Fourier is no explanation at all, Tony using his “old ears” makes the whole thing questionable and there are no controls, for example can any difference be heard doing the A/D and D/A at 24/96 PCM?

    Doing an 860 mile drive home today, have been catching up in The Infinite Monkey Cage, a podcast science series produced by the BBC for the last 13 years. There is no better example of how to make science understandable and fun.

  10. Since I can’t hear anything above 13.5kHz these days, 44.1/16,
    with its limit of 22.05kHz is more than I need.
    Maybe I would’ve appreciated DSD256 forty years ago, when I
    could still hear up to 20kHz, but these days it’s a moot point.

    1. I think there are enough half deaf with super tweeters…and still for good reason…the range above hearing level still has influence to what happens below. The story that everything above 16 or 20 kHz could be cut off and is meaningless is long gone…ok it’s gone since hires was available…before that, digital certainly was perfect with the cut was always perfect, independent of its limitations 😉

      1. Fair enough; we are all entitled to our opinions.
        For now, & until proven otherwise, I’m sticking with,
        ‘If I can’t hear it, then I can’t hear it’ ✌ 😉

        1. It’s the issue of suggesting I have an inferior music collection/source and should change my audio system (that does not do DSD) and buy lots of music in another format to play something I can’t hear.

          The elephant in the room is that if something if it is audibly better, you shouldn’t have to explain it, it should be self-evident from listening.

  11. Again transferring analog to digital can never be an exact copy. Analog is continuous, real life is also continuous without missing bits between 1’s and 0’s. It might be the best way to preserve it over time due to master tape deterioration and much more convenient but it’s not better than an analog recording nor can it be an exact copy. Never can. Time to find an analog recording storage process that doesn’t deteriorate over time.

    1. It could be said that the best analogue recorder is the brain and the memory but that sure deteriorates over time. Some people lose whole days, not just 1’s and 0’s.
      More seriously, maybe it’s not about making an exact copy, though that would obviously be the best, but making something that is good enough to be convincing. Once you can fool all of the people all of the time it’s job done.

Leave a Reply

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

4865 Sterling Dr.
Boulder, CO 80301

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