 Humans can hear up to 20.000 Hz and then only when they are young and unable to afford a good stereo. So why high resolution audio files? Are they snake oil or real audio quality? Not only humans are limited to 20 kHz, most studio microphones are too. There is only a very limited number of microphones that can clearly go beyond 20 kHz and they are normally not used in recording studios. Those big Neumann tube mics even roll off from 13 kHz while all great voices of the past and present are recorded with them. I know there are theories about intermodulation products above 20 kHz but given the level of high frequencies in music, tens of dBs down and the fact that intermodulation products then would be even tens of dBs further down, I think they would not be relevant if they existed. So am I saying that high-res music files are fake? There are many people hearing quality improvements when they use high-res files. Are they right? But let's first see how high-res files are produced. The only way to produce high-res music files is to record them in high-res, for instance in 192 kHz sampling 24-bit depth. Not much of a problem for modern studios, almost all produce in Pro Tools, a digital audio workstation that can record, edit and mix many tracks at that resolution. Still often albums are produced in 44.16, the CD format. The reason for this is simple. Producing in high-res takes considerable more storage space. Recording in 88.224 bits takes three times the storage space compared to 44.116 bit. 192 kHz 24-bit even takes 6.5 times the storage space. Since studios are commercial enterprises as well, they need to charge their larger storage media to the client, in most cases record companies. And it's down to the client to order and pay for the higher resolution. Many of them find MP3 already good enough to make money on. Luckily there are those that do appreciate that they can charge extra for high-res, leaving the consumer the choice. And then there are artists that are in a position just to demand high-res, although they are low in numbers. Another high-res option is to record in DSD, which is usually done using merging technology's pyramid digital audio workstation. But since digital processing in DSD is not a feasible option, it is mainly used for classical music and other music that can easily be published without processing. Editing, pasting pieces together can be done on a pyramid. Then the music around the editing point is converted to DXD, which is 352.8 kHz 24-bit PCM, spliced together and converted back to DSD64. In some cases DSD recordings need to undergo processing like EQ or Dynamics and then the entire file is converted to DXD, processed and converted back to DSD. So apart from the specific and costly recording gear you need for DSD, storage too is more costly than CD quality. By now you might wonder why I tell you all this, for I just told you that the extended frequency range high-res gives you is neither needed nor audible. Well, that is because in audio things are never as simple as they seem. And this all has to do with the filtering needed in digital recording. It is a well-known fact that prior to recording audio digitally, it needs to be filtered at half the sampling rate. So for CD quality a filter has to take out all signals above 22.05 kHz. Normally 20 kHz is taken as a filter frequency that leaves a space between 20 kHz and 20.05 to filter 96 dBs deep, for that is what 16-bit depth requires. I have mentioned in earlier videos that this implies a filter steepness of 96 dBs per single note. Normally filter steepness is expressed in dB per octave, which is 7 whole notes. So we are talking about 672 dB per octave filtering. That is an extremely steep filter and will have severe impact on the sound quality. But there is a solution. Most analog to digital PCM converters use the Sigma Delta method at a very high sampling rate, resulting in a low-bit signal that is then decimated down to PCM at the required sampling rate and bit depth. This way the initial filtering can be very mild given the extremely high sample rate. In the decimating process, where a low-bit signal of 1 to 4 bits is converted to either 16 or 24 bits, the filtering to half the PCM sampling frequency can be done in the digital domain and given sufficient DSP power, this can be done rather good. DSD converters use the same Sigma Delta conversion at one bit depth and simply register the raw DSD data, so there is no decimation. During playback PCM files have to pass the reconstruction filter. Depending on the working principle of the DAC in use, that filter can be analog, partly digital and partly analog or non-existent. Non-oversampling DACs, often abbreviated to NOS DACs, use fully analog filtering after the conversion but might even skip filtering at high sampling rate. Depending on the manufacturer, the filter settings can vary between strict and very loose. Strict filters do 96 dB from 20 kHz. Loose filters start at a somewhat lower frequency and roll off more gradually, which gives a better phase behavior while, when chosen wisely, will not produce audible aliasing products. With the emphasis on audible. Upsampling DACs work differently. Here the digital signal is first upsampled to a higher sampling frequency. In this process the digital signal has to be filtered at half the original sampling frequency too, but this time it can be done digitally. And after the conversion to analog, only a mild analog filter is needed given the now higher sampling frequency. Most modern DACs work this way where the actual conversion can be done with a ladder converter but is usually done with a Sigma Delta converter. Also, here the digital filtering can be done with a linear phase brick wall filter. On paper these filters look almost perfect. There is no phase shift and thus they are called linear phase filters. But they do produce pre-ringing, meaning they will generate sound before the real sound starts. This can be brought down from about 10 cycles pre and post-ringing to about one cycle by having a slower filter slope, comparable to what is done in the analog filters in NOSDAX. This is often called an appetizing filter. But again at the risk of aliasing products that, in a very worst case, can make voices sound like robot voices from early science fiction movies. Fortunately, in audio equipment it affects only the high frequencies while the levels are low enough to be inaudible. Still, the quality of the filter algorithm, for a large part defined by the computational power of the DSP, has enormous impact on the sound quality. That DSP, digital signal processor, can be an integral part of the DAC chip, in which case it is very limited or it can be a separate DSP chip and then can vary from more powerful than the integrated DSP to extremely powerful like in FPGA DACs. And then there is the MQA approach. That is an approach that looks a bit like computational photography. Modern cameras also have one or more DSP's aboard that know the errors in the picture sensor and compensate for them. With system and reflex cameras, where lenses can be exchanged, aberrations can be corrected too, since the lens holds information about its properties. With MQA there is a group of people that focus on the compression properties, but those are irrelevant. It was just a technique they had on the shelf. The important part is the correction of time smearing errors that are specific for a given DAC and the filtering that can repair time smearing in the recording. For the first you need a DAC that is MQA enabled, for the second you just need a player that can unfold the MQA signal. In my set of 1A I have room decode the MQA signals and then play them back over my core DAVE DAC. I more and more prefer MQA files over normal PCM files. Or my set of 1B, whether the MQA enabled MiTech Brooklyn is the DAC, MQA files are almost always the preferred choice. I more and more believe that MQA is a very good way to make affordable DACs sound better than possible otherwise. But if you want to judge MQA and you don't own software that can do this first unfold and you don't own a true high end DAC, you need an MQA enabled DAC. Judging MQA files on an affordable non-MQA DAC gives false outcomes. Ok, now I come to the point I was trying to make. On my set of 1A, using the Grim Audio Mew1 player, the Core DAVE DAC, the Acoustics AX520 amp and the PCM FACT 12 signature loudspeakers, the differences between high res files and CD quality is extremely small, probably more likely due to differences in mastering than due to the sampling rate. That's different with my set of 1B, whether DAVE DAC is replaced by the MiTech Brooklyn DAC with Ferrum Hipsis power supply. Here the time smearing is lower, the pace and rhythm is better, voices are cleaner, the stereo image is deeper and more in focus with high res files. In my set of 2A, the Denofrips Aries II DAC feeds the Marantz Ki Pearl light amp that drives the Acoustic Energy Radiance 1 loudspeakers and Abrell T5 sub for the bottom end. Here the difference is a step greater while with set of 2B, where the Bluesound Node 2 is the player, the difference is even slightly more. My very affordable set of 3 doesn't make the differences greater but the differences still are clear. I have not used MQA files for this test, just normal linear PCM files in FLAC and DSD files in DSF format. They all came from the Green Audio Mew1 that sends the audio files as server to the local players, with the exception of my set of 1 where the Mew1 is the digital player too. I more or less gave away the answer, didn't I? Apart from Jitter, that has also to do with hardware quality, there is one process in the digital player chain that has enormous influence, the reconstruction filter. Chip DACs use integrated chips that do the digital filtering and conversion. These chips are simple and have little processing power and thus poor sounding reconstruction filters. Better DACs than CD players also use DAC chips that use a separate microprocessor to do digital signal processing. And the top DACs and players use field programmable gate arrays, processors that are programmed by the manufacturer with their own code. If you send a 192 kHz PCM file to a simple DAC, it doesn't need to do the upscaling itself with its limited DSP function and thus limited algorithms. On high end FPGA DACs, there is a lot of DSP power and very good algorithms, so the differences are small to non-existent. So to me it is clear that high res files sound better when they work around the limitations of an affordable DAC. If the reason high res files sound better is because they make the work for the DAC lighter, why shouldn't we upscale CD quality files? Well, that's exactly what upscalers also called upsamplers do. Even Cort manages to improve the sound quality of the Cort DAVE if you use its upscaler. See my review. So there's even is a difference in quality with upscalers, for the external one by Cort sounds better than the one inside the DAVE DAC. Still there is a difference with true high res files. The ones that are recorded at higher sampling rates. For since they were recorded at a high sampling rate, the anti-aliasing filter in the analog to digital converter didn't need to be so steep and thus has less time smearing. The quality of upscalers differs greatly. The filter algorithms and dithering have enormous influence on the sound quality. So it's not as simple as downloading free upscalers software and converted all your CD ribs to 192 kHz or even higher. It is also questionable if extremely high sampling rates on DACs have any serious meaning. The 768 kHz sampling coming from a Cort Scaler sounded very fine but the 192 kHz upscaler from the Grim Audio player sounded even better on the same Cort DAVE. There are people that upscale their PCM recordings to DSD. That takes a lot of computational power and might need more power than a music server can deliver. That can lead to losses in audio quality. The same goes for a lesser degree for upscaling to very high PCM sample rates. It's not said that DSD always sounds better than PCM. Some DACs work better in DSD, others in PCM. There are even DACs that can switch between algorithms for PCM and DSD. In one mode PCM sounds better while in the other mode DSD sounds better. Don't forget that any DSP process is lossy. With good DSPs, the loss is very limited while the gain elsewhere is bigger. But if you use your measurement equipment instead of your ears and measure only the scalar, you will find it loses quality. Like bigger tires on cars, increased roll resistance and thus fuel consumption but gives you a better road holding in exchange. For most people high-res files will sound better since they work around the limitations of the reconstruction filter in the DAC. The higher the quality of the DAC or digital player, the smaller the difference between high-res and CD quality. Pop and Rock has limited availability in high-res. Classical and jazz music is more available in high-res, especially in DSD. Please don't look at the numbers but use your ears. Those that look at the numbers only will necessarily dumb down the arguments for audio is rather complex. Or better, audio isn't that complex our hearing is. And treating our hearing as a linear device, as technicians often do, will lead to anything but audio nirvana. And that brings us to the end of this video. As usually there will be a new video next Friday at 5pm central European time. If you don't want to miss that, subscribe to my channel or follow me on the social media so you will be informed when new videos are out. Help me reach even more people by giving this video a thumb up or a link to this video on the social media. It is much appreciated. Many thanks to those viewers that support the channel financially. It keeps me independent and lets me improve the channel further. If that makes you feel like supporting my work too, the links are on the comments below this video on YouTube. 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.