 The HB project and the HB channel are supported by Hi-Fi Clubben. That sound kills good music. I've been attacked quite heavily for claiming that we need higher sampling rates to increase the time resolution up to that of our auditory system. Time to react. There were two kinds of reactions. From those that were taught the Nyquist theorem at school and believed the theorem over their ears. This small group was rather impolite, not to say rude and then there were those that were interested in my thoughts that are all based on the knowledge of far more clever people but didn't understand them. I have to conclude that I haven't been clear enough for both groups so let's try to put this straight. In 2008 I was offered a court gemdak for review. The little thing impressed me heavily and made me request its bigger brother, the QDB76. That impressed me even more and I decided to buy it immediately. It later was updated to the QDB76 HDSD and the review can be found on theHBproject.com. Remarkable with those converters was the impulse response in the mid-range. The thing all court DAX from that period on had in common was very long FIR filters and the longer the filters the higher the resolution of that filter. According to Robert Watts, who designed these filters for court a filter with a million tabs would approach inaudible artifacts. Court's recent flagship DAX, the DAV, has a 164.000 Paul filter the best Watts could achieve in the cost no object project. That, by the way, ended with a retail price of over 10 grand in euros. Anyway, that's where I learned that now jitter has been controlled in the better designs, the reconstruction filter still was and is the bottleneck in digital audio. The Nyquist theorem, named after Harry Nyquist, describes how an analog signal can be digitized, described in numerical values, without losing any information. Basically it states that a band-limited signal can be digitized when sampled at minimal twice the highest frequency in the band-limited signal. In plain English, if you band-limited signal at 20 kHz you need to use a sampling frequency of at least 40 kHz. This theorem stems from the 1930s and has never been proven to be falsely. I'm not clever enough to understand all the math but people I deeply respect that do have the skills and that have been critically following the audio technology see no reason to doubt the Nyquist theorem. As I've shown in the MQA Part 1, see the link in the top right corner we humans are not able to hear above 20 kHz and then only when we are rather young and not affluent enough to buy proper gear so case closed, in theory at least. The Nyquist theorem is only a theorem, a very good one but a theorem nonetheless. The Cambridge Dictionary defines theorem as a formal statement that can be shown to be true by logic. So the theory made sense but could not be proven in practice in the 1930s since no technology was available that could implement it. And that's still where the problem lies today. We simply are not able to band-limit a signal at 20 kHz during recording without the filtering producing artifacts and the same goes for the band-limiting reconstruction filter needed at playback. The filtering in digital equipment is improving over the years but good filtering still isn't cheap and then still not perfect. Remember we want 20 kHz bandwidth and sample it at 44.1 kHz. Theoretically there should not be any signal above 22.05 kHz which is half the sampling frequency. This means that we need to use a filter that attenuates 96 dBs over 2.05 kHz when we use 16-bit resolution. Whether you will hear these filtering artifacts depends on a lot of factors. There will be people that are unable to hear these artifacts due to the condition of their auditory system. But even a bigger problem can be the playback equipment being of poor quality or set up incorrectly. But even if you can't hear the filtering artifacts they are easily shown on a simple oscilloscope by playing a pulse. You will see the so called pre and post echoes. Post echoes do occur in nature as a result of resonances but pre echoes are extremely unnatural. In real life you can't hear a softer copy of a signal before the original signal is there but in digital audio you can. This proves that although the Nyquist theorem is theoretically sound the implementation using contemporary technology is not. One of the things we can do is to reduce the steepness of the filter for that will reduce the artifacts. The only way to do this without causing aliasing is to sample at a higher frequency. If we double the sampling frequency and still filter at 20 kHz we have an octave more before we need to be down at 96 dB. That's still a very steep filter but an improvement nevertheless. If we raise the sampling rate to 192 kHz there is even more room to filter and all other things being equal improve the time resolution of the reconstruction filter. The next step is to use better reconstruction filters. Some brands have been developing their own filtering for years now. I mentioned Chord but even before Chord DCS implemented their own filtering and a few years ago PS Audio introduced a DAC using proprietary filtering. Recently the manufacturers of DAC chips offer the DAC manufacturers the option to have the filtering done outside the DAC chips. Let's freeze the designer from the very limited computational powers the DAC chips usually have. Yet another way to improve the quality is to fight the artifacts of the digital filtering by compensating for them. This is what MQA claims it can do. If you know the transfer function of a filter you can predict the errors it makes and feed it with the inverse error to eliminate it. MQA at the same time claims to manage to store all relevant information of a 24-bit 192 kHz file into a file that is no bigger than a Redbook file and protect the integrity of the file. It's still too early for a definitive judgment on MQA but it looks promising. Another approach is to use only analog filters and thus no oversampling. You still need a reconstruction filter but now the designer uses an analog filter. Often these filters start somewhat below 20 kHz and are not as steep as the digital filters. Since you need a ladder converter when you don't oversample the resolution at low level is somewhat less and can be corrected by using a number of DAC chips in parallel. To me it's just another approach of the same problem and has just as much artefacts although different in nature. Therefore some will like these non-oversampling NOS for short DACs better while others like the oversampling DACs better. Yes, the Nyquist theorem proves that you can describe an analog waveform with a perfectly band-limited signal without any loss when sampled at twice the bandwidth. But Nyquist never proved you can perfectly band-limited signal and as discussed, in practice you can't. But you can choose a higher sampling frequency to have the filtering Nazis outside the audio band. This is what I strongly believe to be true and it coincides with my auditory observations. Feel free to comment on this video and to prove me wrong. But do use arguments and remain respectful to all involved in the discussion. People that don't will be removed and blocked not for disagreeing on the topic at hand but for disagreeing on maintaining proper behavior. You can also choose just to follow the discussion, if any. So subscribe to this channel, follow my Facebook or Google plus page or my Twitter account. You can also post questions but please don't ask me for personal buying advice. View my questions video to find out why the link is in the top right corner. You find more information below this video on YouTube. If you like this video, please give it a thumbs up and tell your friends on the web about it. I am Hans Beekhuyzen. Thank you for watching and see you in the next show or on theHBproject.com. And whatever you do, enjoy the music.