Mono & Stereo : Interview with Dan Lavry of Lavry Engineering
We see that year by year there is kind of race for bits and high sampling rates. Where do you think this will stop?
Regarding bits: The ear can not hear more then about 126dB of dynamic range under extreme conditions. At around 6dB per bit, that amounts to 21 bits, which is what my AD122 MKIII provides (unweighted).
Regarding sample rate: The ear can not hear over 25-30KHz, therefore 60-70KHz would be ideal. Unfortunately there is no 65KHz standard, but 88.2KHz or even 96KHz is not too far from the optimal rate. 192KHz is way off the mark. It brings about higher distortions, bigger data files, increased processing costs, and all that for no up side! People that think that more samples are better, and that digital is only an approximation, do not understand the fundamentals of digital audio.
What rate and bits are enough for today music reproduction and recording?
Regarding processing bits:
For music production, for adding and mixing many channels, for various digital processing, we need more bits. One must make a distinction between processing bits and conversion bits. Say for example that you have 32 channels, each channel made out of 24 conversion bits. If you sum the channels you end up with 31 bits. At the end of the process, the 31 bit sum can be reduced back to say 24 bits, or to 16 bits, because the ear can not hear 31 bits (186dB dynamic range). It is best to have a lot of processing bits. How many, it depends on the number of channels and on the type of processing.
Regarding the rate:
One has to make a distinction between the audio sample rate and the rate of a localized process:
The audio sample rate is the rate that carries the music data itself. Roughly speaking, the audio bandwidth itself is slightly less then half the sample rate. A 44.1KHz CD can contains music to about 20KHz.
At the same time, there are many cases when we use much higher "localized rates,ƒ˘. Such higher rates do not increase the musical content. The higher rates still offer the same original bandwidth of the sample rate. We up sample or down sample between localized rates for various technical reasons. For example, virtually all modern DA's operate at 64-1024 times the sample rate speeds (in the many MHz range). Operating at such high rates simplifies the requirements of the anti imaging filter (an analog filter located after the DA conversion). The decision about the ideal localized rate depends on the technology and the task at hand. It is an engineering decision, not an ear based decision. As always a poor implementation may introduce Sonics, and it would be wise to refrain from the often encountered practice of far reaching false generalizations, so common in the audio community.
When CD came on horizon nobody talked about jitter. Now days everyone is having his own philosophy around it. Can you elaborate on this subject please. What is the real importance and how to approach this?
Jitter is not only an audio issue. I was dealing with jitter issues in medical conversion, way before the days of digital audio. Jitter is an issue for all conversion (video, instrumentation, telecom, medical, industrial controls,Ķ).
The concept of conversion is based on two requirements:
"Taking precise snap shots,ƒ˘
Taking the snap shots at evenly spaced intervals, and playing them back at the same evenly spaced intervals.
Think of a movie camera with an "unsteady motor,ƒ˘, or a playback film projector with a motor that rattles between too slow and too fast. Either case will distort the outcome, and the distortion depends on both the jitter (speed variations), and on the subject itself. Jitter would not do much harm to a steady object, but it does alter the view of a fast moving object. Similarly, in audio, the distortions due to uneven timing (jitter) is due to the interaction between the clock imperfection and the audio itself. Unlike tube, transformer or many other distortions, the outcome due to jitter is NOT predictable or repeatable. There is no such thing as "good sounding jitter,ƒ˘. There are many types and causes of jitters. What we hear is not only about the jitter amplitude and frequency. It is also about jitter type.
Conversion jitter may alter the sound significantly. At the same time, transferring data (that was already converted) between say AD and a computer, or between other digital sources and digital destinations, does not call for great jitter performance. It is only during the conversion process that jitter needs to be very low. Data transfer jitter is not much of an issue, when you are only moving ones and zeros.
We see that year by year there is kind of race for bits and high sampling rates. Where do you think this will stop?
Regarding bits: The ear can not hear more then about 126dB of dynamic range under extreme conditions. At around 6dB per bit, that amounts to 21 bits, which is what my AD122 MKIII provides (unweighted).
Regarding sample rate: The ear can not hear over 25-30KHz, therefore 60-70KHz would be ideal. Unfortunately there is no 65KHz standard, but 88.2KHz or even 96KHz is not too far from the optimal rate. 192KHz is way off the mark. It brings about higher distortions, bigger data files, increased processing costs, and all that for no up side! People that think that more samples are better, and that digital is only an approximation, do not understand the fundamentals of digital audio.
What rate and bits are enough for today music reproduction and recording?
Regarding processing bits:
For music production, for adding and mixing many channels, for various digital processing, we need more bits. One must make a distinction between processing bits and conversion bits. Say for example that you have 32 channels, each channel made out of 24 conversion bits. If you sum the channels you end up with 31 bits. At the end of the process, the 31 bit sum can be reduced back to say 24 bits, or to 16 bits, because the ear can not hear 31 bits (186dB dynamic range). It is best to have a lot of processing bits. How many, it depends on the number of channels and on the type of processing.
Regarding the rate:
One has to make a distinction between the audio sample rate and the rate of a localized process:
The audio sample rate is the rate that carries the music data itself. Roughly speaking, the audio bandwidth itself is slightly less then half the sample rate. A 44.1KHz CD can contains music to about 20KHz.
At the same time, there are many cases when we use much higher "localized rates,ƒ˘. Such higher rates do not increase the musical content. The higher rates still offer the same original bandwidth of the sample rate. We up sample or down sample between localized rates for various technical reasons. For example, virtually all modern DA's operate at 64-1024 times the sample rate speeds (in the many MHz range). Operating at such high rates simplifies the requirements of the anti imaging filter (an analog filter located after the DA conversion). The decision about the ideal localized rate depends on the technology and the task at hand. It is an engineering decision, not an ear based decision. As always a poor implementation may introduce Sonics, and it would be wise to refrain from the often encountered practice of far reaching false generalizations, so common in the audio community.
When CD came on horizon nobody talked about jitter. Now days everyone is having his own philosophy around it. Can you elaborate on this subject please. What is the real importance and how to approach this?
Jitter is not only an audio issue. I was dealing with jitter issues in medical conversion, way before the days of digital audio. Jitter is an issue for all conversion (video, instrumentation, telecom, medical, industrial controls,Ķ).
The concept of conversion is based on two requirements:
"Taking precise snap shots,ƒ˘
Taking the snap shots at evenly spaced intervals, and playing them back at the same evenly spaced intervals.
Think of a movie camera with an "unsteady motor,ƒ˘, or a playback film projector with a motor that rattles between too slow and too fast. Either case will distort the outcome, and the distortion depends on both the jitter (speed variations), and on the subject itself. Jitter would not do much harm to a steady object, but it does alter the view of a fast moving object. Similarly, in audio, the distortions due to uneven timing (jitter) is due to the interaction between the clock imperfection and the audio itself. Unlike tube, transformer or many other distortions, the outcome due to jitter is NOT predictable or repeatable. There is no such thing as "good sounding jitter,ƒ˘. There are many types and causes of jitters. What we hear is not only about the jitter amplitude and frequency. It is also about jitter type.
Conversion jitter may alter the sound significantly. At the same time, transferring data (that was already converted) between say AD and a computer, or between other digital sources and digital destinations, does not call for great jitter performance. It is only during the conversion process that jitter needs to be very low. Data transfer jitter is not much of an issue, when you are only moving ones and zeros.
"Voici mon secret. Il est très simple: on ne voit bien qu'avec le cœur. L'essentiel est invisible pour les yeux." Antoine de Saint-Exupéry