Editor's Note: This series came about after I approached Denafrips distributor Alvin Chee about obtaining information detailing how Denafrips products work, how the digital processing is designed, and more about the company's thought process behind its highly regarded products. What follows is part 1 of 4, in a series of essays written by Zhao, the Chief Engineer at Denafrips.
Audiophile Style is pleased to publish this series but neither endorses or opposes Zhao's views expressed below. Like most subjects we encounter in life, audio has many different approaches and even more opinions about each approach. I enjoy reading and publishing these differing views and am a firm believer that there are endless paths to audio bliss. The challenge is finding one's own path while enjoying the journey.
On the insights of the design and development of digital audio equipment
Does well measuring audio equipment, by today’s standards, really equate to good sound quality?
Digital technology is a symbol of modern civilization. The developments of the past decade are especially astonishing. We have heard and experienced the big data artificial intelligence, augmented reality, digital twins, and virtual world, etc. These are the great achievements in digitization of the analog world we know, so much so that they are part of our daily lives now.
In reality, the digital world doesn’t exist. The digital signal is not analog, it is a representation. It is a convenient way for us to digitize, record, and store data; present it close to the original. In nature, what we see, hear, eat, almost everything, none of this information is digital. None of it can be expressed completely losslessly in digital form.
In the context of audio, to express and present the stored digital data in an analog signal, all we can do is use the techniques we know to convert the digital signal closest to the original analog signal, as much as possible. However, even if we believe this is arguably the closest approach, it must be built on the basis of Digital Signal Processing (DSP). Without DSP, the digital signal cannot be processed and converted into the analog signal at all.
Even with the advanced, sophisticated processing techniques, we can only achieve an approximate, closest to the original Digital To Analog conversion. It is impossible to achieve absolute losslessness. There are always various distortions, and unwanted signals introduced that do not exist in the original signal – as the conversion (AD/DA) take place.
Therefore, the High-Fidelity definition that we often talk about, can’t actually be achieved. One may achieve fidelity in certain aspects, but there are always trade offs in other aspects. Hence, there is no real, absolute high fidelity, lossless in digital audio signal processing. There are always compromises. With regard to high fidelity, if the designer does not know how and what to focus on in the design, it poses great challenges in designing great audio products, especially so for the DACs.
Industry Measurement Standards?
Often, instead of a real music signal, we use a known, lab generated sine wave, square wave, triangle wave, multi-tone signal to test audio equipment performance. This is a compromised technique as we cannot possibly use the real music signal to measure whether the design of the equipment is performing as per the desired design because music signals are constantly changing. This has caused a lot of trouble for many audio engineers, even for some highly skilled, experienced engineers who are mistakenly lead to believe that the sine/ square / triangle wave test performance index data is the gold-standard and assume that test results can completely represent the entire audio signal. This is a big fallacy!
As of todays’ technology, there is no signal that we can use to represent the audio signal we desire to hear. There is no device that can replace human ears to judge whether the sound is truly restored to the original recording, and there is no data indicator that can represent the quality of the audio signal restored from the DAC, amplifier, loudspeakers, and finally, to our ears.
There is a vast variety of audio products on the market, but there are only a handful of truly amazing, great sounding audio components. One of the reasons is that there's no gold-standard for designing a good audio equipment per se. The standards we refer to (as above) are only elementary, entry-level, industry standards.
How many industry standard makers have actually developed excellent audio equipment? The standards established today, are the basic standards, good for electronic equipment testing as a reference. These standards can only be used to determine whether the equipment is performing within specification or out-of-spec, but there is no way to judge whether the audio equipment is of excellent sound quality. The lack of a real standard causes much trouble and many challenges in the audio industry.
Talented audio designers of great audio products have their own set of design techniques up their sleeves. These techniques, often have nothing to do with the industry standard method mentioned above. The designers will never reveal to others what kind of techniques they've adopted. There are only a handful of great audio products and talented audio designers, and there are very few people who really understand and appreciate the secrets behind them. Even if the designer chooses to reveal the techniques and technologies behind certain designs, due to lack of standards, controversial debates and disagreements may arise, questioning the designer's choices.
We have seen some great audio designers, who do not use sophisticated test equipment. Simply, they make use of the basic oscilloscopes, signal generators, multimeters, as well as judging the sound with their ears, while designing the products. The products they developed, sounded much better than those developed with advanced equipment.
Why is that? As we discussed, there is no audio signal analyzer that can replace human ears to judge the sound quality of a device. And, there are no golden ears either, hence, all the measurement parameters in the industry standard may not represent the truth in terms of sound quality. Of course, it doesn’t mean that a designer can absolutely design excellent audio equipment just because he/she can hear and differentiate good/bad sound quality. Reason being, if the designer does not have a way to change poor sound performance in the direction of good performance, then it is impossible to make the equipment produce good sound.
Can the dynamic range we test, represent the dynamic range of the music signal we hear? Can the signal-to-noise ratio we test, represent the signal-to-noise ratio of the music signal we hear? Can the transient response we test with the step signal, represent the transient response of the music signal? Far from it.
The music we hear is not just the specific measurements or parameters of the audio equipment itself. What we hear is the result of audio equipment that has processed the entire music signal. As such, we can’t judge the sound quality just by looking at the parameters of a piece of audio hardware. We should look at the measurement of the processed music signal. However, sadly, there is no standard available today to measure the processed music signal. Therefore, designers have to make do with compromises, using the limited measurement techniques available as a preliminary indicator of audio hardware performance.
We use a sine wave to test equipment dynamic range. 130dB, for example, is a great one. But the real music signal may only be limited to 70-80dB. The signal-to-noise ratio we test is basically the signal-to-noise ratio of the audio equipment itself. What method can we use to measure the digital noise of the music signal unintentionally introduced, by the AD / DA conversion? With the lack of a converted digital noise measurement standard, how can the signal-to-noise ratio measurement be a good indicator?
This is one of the biggest reasons why many of the new DACs on the market, designed with advanced technology, sound "digital." The digital sound is related to the magnitude of the residual noise after digital conversion, so much so that the micro details of the relatively weak music signals will be covered up, inaudible, and adversely impact the sound quality.
Many engineers like to use square waves or triangle waves to measure the transient response of music. This isn't entirely correct. The transient response of this test, measures the transient response of the hardware device and cannot represent the transient response of the music signal.
You can do a simple test. Play a sine wave signal, a square wave signal, and a triangle wave signal with the same frequency and same amplitude. The transient response that most people hear will be a relationship of sine wave, greater than the triangle wave, greater than the square wave.
As such, how could it be possible to determine whether the transient response of the music signal is good or bad based on the square/triangle wave? Arguably, the use of a square wave signal to measure the audio equipment transient response is the most incorrect. The music signals in nature are continuous, varied in frequency and amplitude in a continuous, gradual way. Even at 100KHz, the (inaudible) music signal is gradual and continuous. It is understandable to use an impulse signal to test equipment, but can we trust the measured data as an indication of a good or bad sound? It is the biggest fallacy! Audio equipment that perfectly reproduces square wave signals must have a lot of digital signal noise, and music will most certainly sound digital.
It is puzzling and confusing at the same time. When we test the audio equipment with the industry standard, we have wonderful results, everything is great, to some extent amazing. We have the world's lowest distortion THD%, we have the highest dynamic range, and we have the best signal-to-noise ratio ever measured. Even if the equipment plays a 100KHz square wave, the equipment produces excellent transient response. But why does it sound so bad? The sound is fuzzy, heavy digital glare, poor dynamic control, and the micro dynamics aren't expressed.
Because we use A signal instead of B signal to test the equipment. If the A signal is approximately equal to the B signal, then the test is meaningful. But if the test signals are very different, then the data tested with A cannot represent B, where the B signal is the music signal we desired.
Sine waves, square waves, triangle waves, white noise, pink noise, none of these signals are remotely close to a music signal. Therefore, the measurements can only represent the hardware response to the specific, known signal.
It's like the audio equipment is a container, the music signal is water. The container itself is clean, but the container is covered with various kinds of paint. So, the water put in the container will be contaminated. The container, paints, and water are three kinds of substances, three different links, not just two substances. Most audio engineers always mistakenly believe that container = container + paints. Because most of the paints in music are invisible and cannot be tested. To us, these paints, which contaminate the water, are digital noise.
For vinyl lovers, do the pops/clicks noise produced by the turntable effect the listening pleasure? Does the inherent pops/clicks noise of vinyl records really worsen the signal-to-noise ratio of the music signal we hear? Why, in the music produced by vinyl, is the background noise often so dark and dead quiet? Are we not disturbed by pops/clicks noise? Is it because our human ears can easily distinguish pops/clicks and the background noise of recorded music?
If you were to factor in the pops/clicks noises, vinyl dynamic range may not exceed 20dB. If the obvious pops/clicks don’t effect / correlate to the music background noise, how does the standard signal-to-noise ratio measurement represent the music background noise level?
DAC – Digital To Analog Converter
In the era of digital music, a DAC is the soul of the audio equipment. The quality of the DAC determines the quality of the entire audio system. On the market, there are low-cost, economical DACs, often measuring much better than high-cost DACs. Does the low-cost DAC sound better than the higher cost DAC?
Thanks to technological advances, in today’s world, the electronic and hardware performance far exceed the needs of music signals reproduced. Therefore, in general, most of the audio equipment meets the needs of high-end audio equipment in terms of measurements. However, the digital noise in the music signal exists, it contaminates the entire music frequency band, completely superimposed with the music signal. There is no way for the human ear to distinguish and to separate the digital noise and the music. The human ear can only regard it as a part of the music signal – digital glare.
Without proper measurement, the digital noise is often ignored. We should not underestimate it. One of the key goals of DAC design is to reduce the digital signal noise after DA conversion, as much as possible.
In fact, the negative impacts of amplitude distortion, phase distortion, transient response, frequency response, etc., can be regarded as digital noise, and, these distortions are present in the entire audible frequency band. A common misinterpretation of the digital noise is, when we test the equipment with a sine wave sweep, the noise level of the entire audible frequency band was very low. One may mistakenly assume that the test result has little or no digital noise.
But DA conversion of the music signals is not the same thing, there is no digital noise in the original analog music signal. When we use AD conversion during recording, digital noise is added inevitably, artificially. When we use DSP algorithms and DA conversion to convert the digital signal to analog, these processes generate digital noise too.
Therefore, when we look at the digital signal converted by the DAC, it should be in such a relationship:
- Music signal after DAC conversion = Original analog recording + Digital noise generated by each process (AD/DA, etc)
Hence, the design focus of digital audio equipment must be to reduce the digital noise generated by each process. This has also led to different level of DSP algorithms, DAC conversion architectures, etc., which directly affect the digital noise, thereby affecting the music reproduction itself.
We can have sophisticated, advanced analyzers to test the amplitude distortion, phase curve, frequency response curve, transient response, etc. separately. But these tests are independent, isolated to each other, they do not tell us the sum of the total digital noise, or the music reproduction of the entire audible frequency band as a whole. Which is the biggest reason why all these tests, measurements, cannot be used to judge the quality of music reproduction.
The reason why I say that the standard measurements cannot determine the sound quality in a definitive way is as discussed above. It is not to say that the test standards, nor measurements are meaningless. These are means, methods and tools for us to determine the basic hardware design performance. But we must have a clear understanding of the total digital noise caused by various conversions stages in the equipment, these shall be the focus of concerns.
When we design or purchase audio equipment, we are always distracted by the specification, measurements, and the test methods. If the better specification equals better sound quality, we may have hit a home run – to design the best measured audio equipment and call it the day. If these measurements can determine the sound quality, then all audio equipment will be easier to design (for audio engineers) and easier to choose (for consumers) – just buy the one with the best measurements.
Unfortunately, it’s not. Almost all the audio equipment with great sound quality does not necessarily have the best measurements. Almost all equipment with the worst sound does not necessarily have the worst measurements either. To some extent, some bad sounding equipment may measure very well.
If we change a capacitor, a resistor, an interconnect, or a power cord, the sound we hear has changed. But do these changes affect the measurement of the equipment? Truth is, the measurements remain vastly unchanged, but the sound quality changes tremendously. Do the measurements really have a lot to do with the sound? If so, why is the sound greatly changed, but not the measurements?
So, does a well measured audio equipment by today’s standards, really equate to good sound quality?
Zhao, Chief Engineer
FormMore about Denafrips, visit https://www.denafrips.com