 Again a popular question amongst my viewers. What is better, higher sPCM or DSD? Well, that depends. Digital audio started with pulsecode modulation, abbreviated to PCM. It is based on the theorems of Nyquist and Shannon that in essence say that you can code any band limited signal in numerical values if you sample it at least twice the band limit frequency. I've been investigating flaws in these theorems, journalistically not scientifically, but couldn't find even the slightest doubt that people immensely more knowledgeable than I am. Those theorems stand solid. But why then can a digital audio sound so poor there? The problem lies in the band limiting, in layman's terms cutting off the audio at a given frequency, for CD 20 kHz is chosen. But if you cut off the audio at 20 kHz, the filter that does that will cause problems in the audio that is left. Especially the nasties that occur in the area around 3 kHz where our hearing is the most sensitive are rather audible. Many people started theorizing about sounds above 20 kHz being heard by humans only to explain why digital audio couldn't sound good. A more productive approach was to double the sampling frequency to 88.2 kHz and filter at 40 kHz, one octave higher. This shifts the filter nasties also one octave up and away from the most critical part of the spectrum. That made digital audio sound better, but by quadrupling the sampling frequency to 176.4 kHz further improvement was achieved. But there is a variable in the equation. The quality of the reconstruction filter in the player. The better that filter is, the less nasties will occur and the less the impact of higher sampling rates will be. Already in my year of birth CC Cutler described Delta Sigma conversion that forms the basis of Direct Stream Digital, DSD for short, the alternative for PCM. See my video DSD explained part 1 and 2 if you want a more detailed explanation. For here it suffices to mention that an extremely high sampling rate frequency is chosen, 64 times that of the CD. Since that generates loads of bits, only one bit amplitude resolution is used. This means that a signal to noise ratio of only 6 dBs can be achieved. By applying a technique called noise shaping, that noise can be moved out of band. This way the signal to noise ratio in the audio band is more than sufficient but above the audible spectrum there will be a lot of noise that might not be appreciated by your wide band equipment. So often a mild roll off filter is applied that starts at somewhere between 30 and 50 kHz. Over the years DSD at even higher sampling rates are used by some. 128 and even 256 times 44.1 kHz, so the name DSD was completed with that figure. DSD64, DSD128 and DSD256. In the beginning of digital audio, digital analog conversion was done uniquely using ladder converters. Although 16 bit resolution was claimed that usually was effectively 14 bits. The system works quite simple. The most significant bit, MSB for short, opened the switch that passed 1 volt output. The second significant bit, 2 SB abbreviated, opened the switch that passes half a volt and so on. If all bits are 1, all switches are opened resulting in a total of 2 volts output if the DAC adheres to the red book specifications. Once you get to the 21st bit, we're talking one microvolt, a voltage that is extremely hard to maintain exact due to thermal noise and other nasties. I have measured quite some digital to analog converters over the years, standalone and integrated in CD players but only one or two came close to 21 bit resolution. Most good ones offer 20 bit resolution and the rest was less, despite the 24 bit sticker on the front. Modern ladder converter chips are still limited to that 20 bit but by stacking up more converters, chips piggyback, the low level errors can be averaged out and can result in a higher amplitude resolution. These converters are nowadays called NOS DACs from non-oversampling digital to analog converters. When delta sigma converters became available, that like DSD worked with only one bit resolution and noise shaping, the linearity problem seemed solved since only one bit was used. But this technique moved the amplitude linearity to the time domain so precision clocks became extremely important. In the meantime there are DAC chips that use techniques somewhere in between like 3 or 4 bit resolution but between a player and a DAC the information, the audio is normally sent as either PCM or DSD. Many computer programs that play out audio to a DAC can convert audio from one format to the other and change the sampling rate within one format. The audio data has to be packed and labeled to make sense to the player. The most basic way to pack audio data is the waveform. It stores the audio data perfectly and is a lightweight container, as the pros call it. But the support of metadata was poor for a long time and still is not supported by many players. AIF or AIFF stores the audio data about the same way but the metadata is far better supported. Both containers store the bits the way they are in the audio stream. Then there are lossless compression containers that make the files up to 45 percent smaller without any loss of information. The most used is the free lossless audio codec, flag for short. The other popular one is Apple lossless audio codec, abbreviated ALAC. Both are widely supported in contrast to the lossless Windows Media format or WMA. The WMA format also knows lossy variants, thus wants to throw away information and as a result will sound inferior on a decent stereo. The same goes for MP3 and AAC that also degrade the sound quality as trade-off for file size and are the popular ones. Although I do use AAC in the car, it is the better of the two. I would not use either on my stereos, not even in my sub 100 euro setup 3. Frequently I come across people that claim that Flag sounds inferior to Wave. This means that the playback system doesn't handle Flag very well. Flag works like zip compression on a computer. The data is packed more efficiently but that takes some computing power to unpack. If the player is not powerful enough or has poorly performing code to unpack, chances are this will impact the sound quality. A well-known case from the past was a lin player, but that's long solved. Although I have not heard problems with apple lossless, the same could occur with it too, of course. Whether Flag or DSD sounds better depends largely on your equipment. If you play Redbook files, those files that are used on CD, chances are your DAC will upsample them. The question is, will your DAC do a proper job or does the player or player's software in the computer do a better job? Often, but not always, the software does the better job since a computer usually has more processing power than the processor in a DAC. The quality of the player's software has to be taken into account of course. If you use a DAC that does have a very high quality upsampling, I have this experience with cord DACs and there will be others. Let it do the upsampling. The question whether Wave or Flag sounds better also is a hardware and software dependent and that goes for the other lossless formats too. Whether you should prefer DSD over Flag or any other PCM format again is a matter of hardware and software. I will give you an extreme example. If you use a Nostag that does up to 192 kHz, DSD files have to be decimated to PCM 192 kHz. At best this gives a little loss but you lose quality anyway. Using 192 kHz PCM files would be the best option and if you also play 44.1 kHz files having quality player software upsample them to 176.4 kHz might give an improvement in sound quality. Why? Not because the highest sampling rate holds more information. There can't be more information than there was in the original file. But as we have seen before, at higher sampling rates the reconstruction filter gets an easier job and its nasty shift to a part of the frequency spectrum that is far less critical to our ears. If you use a hardware player that plays all sampling rates in PCM and DSD, you will not have to transcode the audio files to other sampling rates from DSD to PCM or vice versa. The best way always is to buy files that perform optimally on your system. But if you now own a system that only does 44.1 and 48 kHz and want to upgrade later on, you could buy high res files and make 44.1 or 48 kHz copies using downsampling software. The original files are carefully kept elsewhere of course. The conversion will cause a slight loss but if you use proper software the damage will be small and might not be audible on your current player. Since there is no single optimal solution, you can only compare differences between possibilities yourself. Try to be as structural as possible. Write down the possible solutions and use each possibility for a week or so, provided you play music regularly. Don't switch every few minutes for then you focus on the wrong things. If you don't hear a difference when you switch after a week, you might wonder if there is a difference and if there should be a difference if it's worth bothering. If you do hear a difference, go for the best option to your ears anyway and compare that with the other options of your system. I would start with bit perfect playback so no upsampling, downsampling or other processing in the player or player software and listen for a week. Then pick another option and again listen for a week. Always use integer upsampling, so upsample 44.1 kHz to 88.2 or 176.4 kHz and upsample 48 kHz to 96 or 192 kHz. DSD always is an integer multiple of 44.1 kHz, so if you have to convert DSD to PCM always convert to 44.1, 88.2 or 176.4 kHz. Conversion to a non-integer sampling rate needs more complex calculations and thus should be avoided. Trust your ears. If you don't hear a difference, accept that despite of what others say. And if you do hear a difference, accept that too despite what others say. Always take longer periods for listening and use a wide collection of music. So an easy question leads to a less easy answer as is always the case in audio. Easy answers are suspect and although they can be comfortable to you, they seldom are right. Welcome to my world. The next week there will be another video as always at 5pm central european time. If you don't want to miss that, subscribe to this channel or follow me on the social media, so you will be warned when new videos are out. If you liked this video, give it a thumbs up. Many thanks to all that support this channel financially, it keeps me independent and thus trustworthy. If you also feel like supporting my work, the links are in 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.