One of the reasons Class D amplifiers are considered by most to be digital (technically they are not) is the need to encode the continuous analog input signal into another language to work. So it may be correct, in many minds, to classify this amplifier type as something other than analog because of the encode/decode process. However, I would disagree.
Class D’s language, Pulse Width Modulation (PWM), is very far removed from true digital audio (called PCM) and is more closely related to DSD or 1 bit audio. In a PCM language the analog voltage is converted to a number which can then be stored in a computer. Without a magic decoder ring that understands this numeric language, true digital audio cannot be understood or reproduced. When we think of DSD, we get closer to analog because the stream of bits can be converted back to analog without a magic decoder ring.
In other words, if you take PCM and put it into your preamp you get nothing out but nasty noise. DSD and PWM (very closely related) fed directly into a preamp produce music! In fact, when DSD is recorded onto a disc a type of PWM is used to record that DSD data and, if you take either out of the disc or hard drive they are stored on, you can play them directly onto a stereo.
So we don’t typically refer to something as digital unless it’s based on a binary (2) system of on and off. With only on and off you must use a numeric system that represents values that must be decoded. I have always thought it a mistake that DSD has the word digital in its naming – because you can make an argument either way – and the fact both DSD and PWM can be streamed as analog without conversion is the key for me not referring to it as digital.
If you remember from yesterday’s post we learned that PWM (and DSD) actually have 3 states, as opposed to 2. It is the addition of this third state (time) that clearly differentiates both PWM and DSD from digital audio.
Why is today’s post focused so heavily on our understanding of both DSD and PWM as coded analog instead of digital? Because getting a clear picture of the difference between the two will help us understand just how this works.
Lastly, I wanted to point out the similarity between DSD and PWM because these are the two most analog sounding technologies I know of.
Revisions
There are no revisions for this post.
Paul, I still awaiting a direct digital amplifier which able to directly playback raw DSD and SACD with volume control in digital domain then driving speaker in PWM with Class D Amplification,, eliminating lossy passive components, all boxed in one single elegant box, it’s purest way to playback DSD recorded materials.
,The closest things I can think of are Devialet D-Preimer (PCM only) and Sharp DX-SX1 and SM-SX1 combo for SACD.
The major difficulties seems to be the chipset to convert encrypted DSD stream into raw DSD signal, or just supporting raw DSD from DSD file, and DSP that able to do volume control in DSD without conversion to PCM
Well, we’re not really that far away to be honest with you. Hang in there, it’ll take time but will happen.
This posting brings out several points that may have been unintended. However expertly or inexpertly a system of any kind including an electronic audio recording playback system is engineered, each element can be viewed as a black box with a specific function to perform. Once a function is defined and black boxes that perform that function within defined limits are available, it doesn’t matter which one you select insofar as performance is concerned. Other factors such as price, reliability, efficiency dictate the most logical choice.
From an engineer’s point of view an audio power amplifier is a “two port” device with a specific electrical function. That function is defined by its required measured performance. The reason amplifiers that measure the same produce results that sound different is that the measurements are incomplete and the methods of measurement don’t reflect the way the equipment is used in the real world. For example, the frequency response of an audio amplifier is customarily tested at one watt output into a 4 or 8 ohm resistor. In the real world, amplifiers are requred to operate at much higher outputs into reactive loads that are not even passive but kick back with reverse EMF. Measurements under those conditions would likely show a very different story.
The only valid listening test I can think of at the moment which would still be imperfect is to compare an amplifier to a shunt where the input to the amplifier under test is the output of another audio amplifier that has demonstrated the least tendency to be affected itself by the load. The difference between the amplifier under test and the shunt is the coloration the test amplifier has contributed to the signal. I have yet to see bench tests where audio power amplifier output current and voltage waveforms for complex waves into a real world loads are compared to the corresponding input waveforms and analyzed in light of the load impedance they’re working into. It seems to me that given today’s technology, this once impossible method is now a realistic option. Why hasn’t it been developed? Because in reality there’s no real interest in it, this remains an esoteric class of products for a niche market that’s not particullarly interesting anymore to mainstream electrical engineers. They have much more fascinating and compelling worlds to conquer.
@Soundminded:
there is another important reason that the measurements don’t reflect the way the equipment is used in the real world: the signal constantly varies in strength and frequency.
With respect to the test you came up with: I like to think that the modern acoustic calibrations like Audyssey do something alike what you are looking for: a known signal is fed into the system and received by a calibrated microphone. Distortions are what the system contributed to the signal. Of course, not only the amplifier is measured, but the system. It does allow freedom to combine loudspeakers to amplifiers of different makes.
@paul: I can’t wait. Question is: will it be multichannel as a modern system should be and will in incorporate Audyssey or comperable calibration as the total system will always act as a filter, no matter how good the amplifier is.
“I like to think that the modern acoustic calibrations like Audyssey do something alike what you are looking for”
This sounds like automatic equalization. It’s a start but it’s just a high tech version of what is now a very old idea. Not a bad idea, in fact quite a good one. The term “room correction” has been invented to get around the fact that the audiophile industry has made the word “equalization” taboo but it’s the same thing. Equalization of this type used to be SOP for large recording studios back in the day when they cut phonograph records in a big way but no more. This gave some consistancy of spectral balance not only from recording to recording on the same label but from label to label. This I think is one reason why many people like phonograph records better than CDs, they no longer have the means to compensate for differences in spectral balance from one recording to the next and wouldn’t know how to use them effectively if they did.
However, while it’s a good start, it’s not nearly enough, it barely scratches the surface for testing. Testing technology in the analog domain is one of the most overlooked, neglected, and inadequate aspects of modern audio recording technology. The instruments may be more sensitive than they were but what is tested has barely changed in 50 years. About the only thing I can think of that changed was the invention of TIM or transient intermodulation distortion by Matti Otala working at Harman Kardon. In fact had wide bandwidth high power tesing been performed, TIM would be redundant, already known and understood. All it really expresses in the inability of amplifiers and preamplifiers to deliver high output at high frequencies. Probably in many cases this is due to a junky power supply but topology and selected components can be responsible too. So can improper use of that most powerful technique, negative feedback. The concept of gain-bandwidth product should be well known to analog electronics engineers. The concept of signal strength outpout-bandwidth product should have been invented also and is very similar. People have a tendency to make simple concepts seem complex and invent jargon to confuse those who have not parsed their technical papers. Maybe it makes them feel their discoveries are more important.
@Soundminded
For the last paragraph I think I agree. About the first paragraph, I am not sure if I explained correctly and I am sure I do not understand what you are saying
Although it is called room calibration, it actually isn’t. It is a system calibration because dynamic behavior of the pre-amp, the amp, the DAC, loudspeaker cables, the loudspeakers etc is in there.
A fundamental difference between an equalization and the calibration is that this is done using a parametric model. The filter actually is the inverse of the model and the end effect should be in an ideal world a system with no influence (if only the world would become ideal around these X-Mas days …)
Thus there are no frequency bands each with its own gain and phase parameter. Obviously, just with an equalizer sitting in your room and having fixed settings (if you don’t turn the knobs), it will not make different recordings sound more similar.
The part I don’t understand is why it shouldn’t be possible to apply this to a record released on last centuries technology like CD or the current HiRes releases. Send a calibration signal through the studio, measure the sound pressure with a calibrated microphone and determine the model parameters. Then determine a filter (inverse of the model) and you are good to go. Note that at this moment you do not know yet if the studio is a church or an acoustically dead room and what kind of music is going to be recorded. Next step: record music and apply the filter.
As I see every system as a filter, all audiophiles should apply an inverse filter to obtain the most original reproduction: without it one would not be an audiophile
Just waiting for the ideal world now to apply some none ideal filters …
I think you misunderstand the meaning of the term “parametric” as applied to equalizers. A conventional graphic equalizer has a series of sliders or knobs each for adjusting the amplitude (loudness) of a fixed frequency range over a fixed bandwidth. A parametric equalizer has a filter that has an adjustable center frequency and can be adjusted to be narrow or wide in range in its effect. The height versus width of equalization is called the Q, the more narrowly focused the filtering effect, the highe the Q. The automatic system does exactly what you would do manually, adjusts the system FR to measure flat or some other predetermined FR curve using the calibrated microphone and either a pink noise generator or swept sine wave generator played through the speakers. While this is better than nothing it does have its flaws. First of course it will be different at different points in a room so you either measure at your favorite spot or take an average of multiple points. Second and very important, while it is intended to adjust for room acoustics, no speakers I’m aware of make provisions for adjusting the reflected sound field FR or loudness independently of the direct radiatied sound, you’re adjusting both at the same time whether with equalizers, tone controls, or driver level controls.
The use of equalizers to calibrate studio monitor speakers during the LP era meant that while different companies used different speakers RCA, their own LC-1A, Columbia and many other Altec A-7 VOTA, London and EMI probably Tannoy Dual Concentric Monitors, all of them heard nearly the same thing. When they tweaked their recording consoles to get the best balanced sound, they were more or less adjusting to the same standard. But today it’s different. Without equalizing monitors, every recording is balanced to whatever monitor the balance engineer happens to be using. Usually it’s his favorite audiophile speaker. A lot used B&W 801s for awhile but many didn’t.
You are right about a reference signal. The equivalent in TV was called “VIR” or Vertical Interval Reference signal was used to keep colors in balance. Before VIR people joked NTSC stood for “never the same color.” There are no comparable reference signals for audio spectral balance. That’s why they vary so much.It’s probably one of the reasons many people don’t like them.
Indeed that is what parametric equalizers do, but that is not what modern parametric calibration systems do. The latter use a model: they don’t work is bands (of fixed or adjustable width).
Indeed you either calibrate a small sweet spot or a larger sweet spot with it. The larger the sweet spot, the lesser it is optimized for each position in the sweet spot.
That both the speakers and the room acoustics are calibrated at the same time is desired behavior from my point of view. We agree: because the speakers do not make the provisions, you have to account for them. As the calibration system doesn’t know if the disturbances it measures are caused by room acoustics or by speaker imperfections, they can both be accounted for (I imagine for instance by sending an inverse signal to compensate a delayed reflection). It goes without saying that the key lies in preventing reflections to happen and making sure your loudspeakers and the rest of your system behaves, because also the inverse signals will cause a new indirect signal etc.: as said, the filters and calibration are never perfect.
Thanks for explaining the studio calibration.