Wading into it

March 20, 2014
 by Paul McGowan

One of the more confusing aspects of digital audio is the two formats we have to deal with: PCM and DSD. PCM stands for Pulse Code Modulation and DSD stands for Direct Stream Digital. The first acronym, PCM, actually describes the format while the second acronym, DSD, is a marketing term invented by Sony and Phillips (DSD is more accurately titled PDM).

So, the two formats are PCM and PDM, both acronyms that describe to us engineering types what they do and how they do it.

I know, I know. Several of you have written me begging that I explain “in english” and not use acronyms that confuse. How about this as a compromise? Would it help if I refer to PCM as simply Multi-Bit Audio, and DSD as Single-Bit Audio? Let’s try that and see if that helps. I’ll include the acronyms in parenthesis for a while to help clear the way.

Multi-Bit Audio (PCM) is the standard digital audio we all grew up with, starting with CDs introduced in 1982. In later years Multi-Bit audio gained more bits and ran faster, creating what has become known as High-Resolution audio. But it’s still Multi-Bit Audio. CDs are Multi-Bit Audio with 16 bits and running at a speed of 44,000 times a second. High Resolution audio has an added 8 bits for a total of 24 and a speed of anything more than 44,000 (although generally 88,000 would be considered minimum). The basis for this format has been around for decades.

Single-Bit Audio (DSD or PDM) first came to our attention 17 years after the introduction of the Compact Disc, in 1999, in the form of the CD’s intended successor, the SACD (Super Audio Compact Disc). SACD is another marketing term that describes a DVD disc running Single-Bit Audio. Unlike Multi-Bit Audio, Single-Bit only has one bit but runs at a considerably faster speed, 2,800,000 times a second (standard DSD) or double that at 5,600,000 times a second (double DSD). The basis for this format has been around for decades as well.

Multi-Bit Audio (PCM) is a type of code that is meaningless when you look at the code with your eye. To decode and play it back you need a computer or a dedicated IC that can unravel the code before it gets turned into music. Reasonably complicated.

Single-Bit Audio (DSD) actually looks like the music it is playing. To decode and playback Single-Bit Audio all you need is a simple analog filter and out comes music. Simple.

I found this graphic for you, showing what it takes to record music to each format with an A/D Converter. It’s very simplified, but you get the idea. Note how much more “stuff” is required to convert analog to PCM.

Tomorrow we wade in a little deeper.


Subscribe to Paul's Posts

13 comments on “Wading into it”

  1. Great post , but as simple as dsd is the various devices sound very different don’t they ?

    And the extra effort you have put into yours has made the difference. But overall PCM is still the mainstream.

    But I’d agree dsd doe sound more musical than PCM . Even if you take a dsd and convert it . The sound changes
    It’s all there but different .


  2. “DSD is more accurately titled PDM.”

    Wow! If you have posted that statement before (or if anyone else has), I’ve missed it. That suddenly blows away a lot of fog in a single simple sentence.

    1. I thought this had already been discussed. PDM isn’t really a digital signal at all as was explained in the early threads about the Hypex module and class D amplifiers. Frequency modulation, amplitude modulation, phase modulation, pulse density modulation are all analog systems of encryption for broadcast or storage. The real D/A converter in the new DSD unit is the circuit which converts PCM protocol into DSD. This is where writing code comes in, making that conversion in real time as accurate as possible.

      Converting PDM back into an audio signal is even easier than an AM radio detector which requires a diode and a capacitor used as an integrator instead of an inductor. The reason PDM decoding works for this application and why AM demodulation works is that the load it works into is fixed, known, and is modeled as a simple resistor. In the case of the Hypex and other class D power amplifiers driving speaker systems, the load impedance is highly variable and often a very complex function where capacitance and inductance usually vary strongly with frequency. This presents problems for all amplifiers to one degree or another especially with high output impedances such as vacuum tube amplifiers. Class D amplifiers appear to be especially vulnerable.

      Too bad. The DSD output of this decoder could have gone straight into the Hypex module bypassing the input circuit including the PDM encoder in the module itself.

      1. Good post…a direct digital amplifier.

        Even better, it’s been done already w/o Hypex – the conversion takes place at the output.

        There are at least three existing commercial designs of what I reference.
        Not DS to the best of my knowledge.

        More, to make it clearer:

        From Paul: “Single-Bit only has one bit but runs at a considerably faster speed, 2,800,000 times a second (standard DSD)”

        One needs to calculate and compare DS to 192K/24 bits, which represents exactly what then, and why would one be superior to the other hi rez?

        What are the HF filters doing for both, and sonically, isn’t the filtering what really matters?

    1. Well, in a way, that’s sort of right – but not quite. AM radio has a steady “sample rate” but turns the level of the sample rate (the frequency it’s broadcasting at) up and down in volume. FM radio changes the frequency back and forth to the music.

      DSD can be thought of as creating energy with its bits – for each on bit, a “bit” of energy is sent to the loudspeaker. The more bits the more energy, the loudspeaker starts to move.

      See tomorrow’s post for a better explanation.

  3. Pulse density modulation is an analog format where the time between pulses varies directly with the signal strength. Therefore the change in density in time is an analog that is proportional to the change in amplitude of the input or output voltage. In AM radio, the actual information is in the envelope. not the carrier, that is at the peak of each half cycle. These peaks have to be “integrated” in the detector by a capacitor of the correct value. It’s an antique system but the basic principle is similar. The advance of the superheterodyne receiver was that no matter what the carrier frequency, the beat oscillator would bring the signal to the same IF frequency so that among other things the value of that capacitor to integrate the envelope pulses to prevent distortion is always the same due to a fixed IF (intermediate frequency.) This is not true for SSB (single sideband broadcast) where analog signals if the beat frequency oscillator is wrong are badly distorted making the sound garbled.

    True encrypted digital signals are different. Each “word” has a definite beginning and end with a fixed time window for each word and each element in the word. If you get those starting and ending points wrong the same stream of ones and zeros will have an entirely different meaning. With analog you can begin anywhere and it doesn’t matter.

    We think of cathode ray picture tubes (old TV and video displays) as working on an analog signal. In a sense it is because the strength of the video signal that illuminates the phosphors by varying their intensity is proportional to how light the image is at any point in time. However, synchronization is critical there too and “blacker than black” signals block the analog video signal output to allow time for and to trigger the beam retrace or reset. If it didn’t, you’d see the retrace on the screen. In the case of the horizontal defection it’s to allow the beam to get back to the left of the screen, in the case of vertical deflection it’s to allow beam to retrace from lower right back to upper left for the next half frame. In this sense the timing and start and end of each line and each half frame have to be defined and synchronized just as a true digital word does or the output is entirely distorted.

    I would also like to point out that the accuracy of time density of pulses in PDM is affected by the accuracy of the clock too. Whatever the type of distortion caused by the inaccuracy of the clock in PCM, the inaccuracy of the clock in PDM will cause amplitude inaccuracy. Fortunately I don’t think even John Atkinson would argue he can hear a difference in loudness of one part in two ten trillionths.

    1. I am not sure I agree with your opening statement or, at best, it’s confusing. PDM has a fixed frequency and the timing between bits doesn’t change. Running at 2.8mHz, for standard DSD, the varying density happens because bits are either on or off, but they occur at the exact same moment regardless. The greatest density occurs when all bits are on, the least when all bits are off, but the intervals between them are fixed.

      1. AM radio has the same problem. The information only occurs at fixed intervals at the peak of each cycle of the carrier frequency. In between peaks, NADA. The capacitor in the detector is the integrator, it connects the dots. To prevent the information being inadequate, the frequency of the carrier and IF frequencies must be much higher than the highest frequency in the information signal itself. So in AM radio it’s been standardized at 455 kHz. This makes the time intervals between peak voltages that constitute the envelope extremely short. The RC time constant in the detector is tuned specifically for that frequency.

  4. ALAN TAFFEL in The Absolute Sound ran some tests with REDbook, SACD, various PCM formats and DS. We need more comparison like it of hi rez formats, for perhaps eventually the industry will settle on one or two and dispense with the rest.

    176/24 and DSD, one against the other. Summing up his impressions with Ricky Lee Jones’ Traffic In Paradise, Taffel states 176/24 dispenses with some glare he hears on CD, but in every other respect fails to match the CD; the FLAC 176/24 loses the “sense of life” that imbues the CD; the DS version is “more timid and restrained” both tonally and dynamically than the FLAC. In other words, two rungs down from Redbook using the DcS Scarlatti – not exactly junk for a $35,000 player. He does suggest that other DACs may favor DS.

    Then he asks, so does anything sound better than the CD layer on the AP disc? “Why yes” he states, the SACD layer! Well, the SACD layer is DS, but not the same DS as the DS download. He then recommends that you spring for the hybrid SACD. I wonder how one can rip it without gouging one’ eyes out? I have tons of SACDs I’d love to do it to.

    He goes on to TIME OUT by Dave Brubeck that is a reference LP for many. Superhirez offers the DS, HDtracks the 176/24 FLAC. The DS is smoother sounding with a bit tighter bass….but less visceral than the176/24 FLAC due to restrained dynamics and subdued high end extension. Once again the SACD layer was superior, more open and transparent, yet no match for the 176/24 from HDtracks that he says “you should purchase immediately.”

    Taffel covers other comps, they make my heads spin, all three of them.

    None of this has anything to do with DACs and such, but for the observation that it is the recording that matters far more than the format; and the DAC that does well with CD wins.

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