High Resolution – The Future of Digital Audio lecture by Igor Levin, Morten Lindberg, Lothar Kerestedjian and Daniela Manger

Earlier this year, during the High End Show in Munich we were given the chance to organize a lecture about high resolution audio and the benefits it brings to listeners. In order to get several different perspectives we got together four different experts and supporters of high resolution audio, from four different fields of the audio industry.
 
The lecturers were Daniela Manger from Manger Audio, Morten Lindberg from 2L , Lothar Kerestedjian from HighResAudio and Igor Levin, CEO and Founder of Antelope Audio.
 
The title of the lecture, High Resolution – The Future of Digital Audio attracted a significant audience in front of the Technology Stage at the High End Show. Here you can watch the whole lecture divided into several short videos.
 


 
Transcription
 
Good morning and welcome to High Resolution, this early morning first talk about high res audio starting with Igor Levin, CEO of Antelope Audio who will introduce you to high res audio systematics. So, then again, welcome and ‘give a hand’ for Igor Levin.
 
Guten morgan, Good morning. Thank you for coming and… it’s actually amazing what kind of things people want to hear early in the morning. So let’s talk a little bit about the technical stuff because the guys that come after me are a lot more practical than me. I’m not a very practical kind of a guy. So let’s do the sinusoids as everybody talks about sinusoids. This is something that shows the basic aspects about audio sampling.
 
Now, there’s fundamentally three things to high-quality audio. One thing is the bit resolution, and this is what these three things illustrate. Where this is presumably a converter say, with 24-bits whereas the second picture, figuratively speaking maybe 16-bits and then 12-bits. So the bit width determines how accurately we sample the waveform vertically. The second aspect is the sample rate, which is the horizontal aspect of sampling. This shows you the same sinusoid sampled at 44.1 kHz and this is sampled at 384 kHz. So we could see that we’re taking a lot more points at the 384 kHz, which obviously gives us a much better resolution in time. That’s something that’s very important for trenchant response. That is the issue that will be addressed by one of the later speakers in the forum.
 
The third aspect to quality audio sampling reproduction is the issue of jitter. The green waveform is the waveform we wanted to have the blue waveform is the waveform we got when the points move off the curve because of the timing inaccuracies. So having accurate clock is equally important to being able to get back the waveform you started with. So these are the three most important things and that’s where the battle organizes. The battle is based on trying to get nice bit width resolution, trying to get nice, accurate and complete sampling and trying to reduce or control the jitter.
 
So the next thing we’re going to talk about is the structure of a typical D/A converter chip. Now this is becoming a bit technical. Hopefully not too shocking to most participants. This is a typical D/A converter chip. Now, first thing that happens in the D/A converter chip is the interpolation so that when the audio comes in at 44.1 kHz, you actually have to increase the sample rate because the D/A component, Number 3 in the diagram, does not work at the sample rate that you think the audio is playing on. So, it typically works that sixteen times the sample rate or thirty two times the sample rate, so inside a D/A converter chip there is an interpolator that upsamples every DAC has an up-sampler, whether you realize it or not, that’s fundamental to how the D/A converter chips work the Sigma-Delta converters. So one of the issues with the sound quality is how good is this interpolator? …Because obviously if the interpolator is not good then what you’re getting is an artifact of interpolation.
 
Because that was quite a big issue in the early designs of the chips, people were thinking, is there something we can do to avoid the issue of interpolation to avoid the distortions that come from the upsampling. So the idea of the “DSD” was born. This part here represents the A/D converter, which starts with the Sigma-Delta modulator and that has a digital filter that downsamples and then… we are connecting this to a D/A converter. So we’re kind of starting with a high sample rate, down sampling it where it comes out at 44.1 kHz for CD. Now, we’re upsampling it again to bring it up to high frequencies, typically several megahertz and then we’re applying it to the DAC.
 
So the insight on the DSD was: Okay, why do we start with a high frequency then throw it away going down and then going back up again? -So let’s get rid of this digital filter. Let’s get rid of this interpolator and this is what the DSD is we’re taking the output at high rate from the modulator at typically several megahertz and we’re applying it directly to the DAC. So, in principle the idea seems brilliant. It’s a brilliant idea to get rid of the two components that are quite responsible for distortion. In practice however the picture didn’t work out as good as people had hoped…
 

 
Transcription
 
So what is the problem with the DSD? Well, here I put down some of the issues, but perhaps one of the most important issues is the issue number three, that all the mastering software, everything that we have, that everybody uses because the CD you listen to is not… it’s not directly from the microphone to your speaker, but it’s been processed. There was a recording engineer; there was a mastering engineer. All mastering today that exists, exists only in PCM format. There is not much you can do with DSD directly. You can perhaps change the volume, but that’s about it. So as a result, even when the record says it’s DSD, it has effectively been converted to PCM for mastering purposes and so at the end, you can convert it back to DSD if you like and that’s what’s done, but unfortunately that kills the original idea of getting rid of those interpolation filters. So we’re back to using the interpolation filters and then at the end we are artificially converting it to a DSD. That is the issue number 3 that is perhaps the weakest point of the DSD idea.
 
The issue number two is equally important. When the DSD was originally created it was a 1-bit format. Since then the world moved on. The chip designers quickly realized that 1-bit DAC does not give you the dynamic range that you want and the one bit designs were abandoned approximately ten fifteen years ago towards 3-bit designs and then later to 5-bit modulators. That’s something that’s not taken into account. So as a result, when you’re converting it to 1-bit, you’re actually throwing away the bits that the DAC is capable of because the DAC itself. You cannot buy a commercial DAC chip today that’s a 1-bit DAC chip. All the chips that receive the DSD formats, they are actually 3-bit or 5-bit modulators which will then artificially which are not using all the bits that the chip is capable of. So as a result, this 1-bit issue is sort of… again there is a loss involved, that it’s mismatched with the current technology that’s commonly available.
 
Then there are issues of standardization and most importantly as well from a technical point of view. The DSD system uses a lot of dithering in order to work properly and this heavy amount of dithering it’s actually high-frequency noise, which when it gets into analog circuits like preamplifiers and power amplifiers after it tends to react with the with the input circuits in such a way as to produce another set of distortions. So a huge amount of dithering noise is always an issue with the analog circuits that follow. So as a result because of some of these reasons DSD format as well as standardisation issues and equipment never really obtained penetration and in my view the idea was excellent, but the implementation did not quite work out how it was intended to be. So now let’s see – what can we do with today’s technology? Times move on, we’re now to 3-bit modulators, 5-bit modulators and we are up to the chips that can convert at a much higher sample rate.
 
So the argument I’m going to make is that 384 kHz audio actually has a chance to become what DSD wanted to be but did not actually become. Now what is so cool? Well, the cool thing about high sample rates is that they… whereas they do use a little bit of filters in the chip these are not the long and deep filters with very very long-time trenchant responses. These are affectively we’re taking a signal as if almost directly from the modulator with a small amount of filtering, very very gentle. We are able to convert it to 384 kHz, PCM and we are able to go away from the problems with the upsampling and the downsampling. We now, because it’s a PCM format, we now become compatible with all the mastering plugins and it is now a format that’s easily stored and distributable in files, so you can store them in WAV files, you can store them in FLAC files and that’s something that makes it very compatible with the modern internet age. And, because of the very very dense sampling in time – it reproduces the time response, the trenchant a lot better, which is thought to be responsible for creating a sound which comes much closer to the analog sound that the digital has always wanted to reproduce.
 
So now, very very shortly the advantages of higher sample rates: It allows for frequency response much above 20 kHz. Now this is a little bit of a controversial issue. Some people think it matters, some people think that it doesn’t,
 
- your dog will hear it for sure
- you may not.
 
It reduces the processing distortions in the DACS. It improves the mastering and it comes closer to the analog sound. So now, what can you do with the stuff? Well, one of the things you want to do with this stuff, is if you are an LP lover, you have some records, it would be nice to digitize them at much higher sample rates because it’s going to sound a lot closer to your original LP. Now, what kind of equipment can you use With it? There’s a few things available on the market including our own unit that we’re going to be showing… We are showing it today and that’s called Rubicon. It’s a 384 kHz mastering and processing center for audiophile. You have 384 kHz A/D converter, D/A, you have a USB interface, you have an atomic clock that takes care of the jitter issues and you have a nice headphone preamp. So come out, check out this unit and I’m going to pass the microphone to the next presenter who’s going to talk a little bit about – how are the high sample rate recordings actually mastered.
 
Watch all the videos here.