Hey, thanks Paul! I was going to buy one of those fancy DACs (Digital to Audio Converters) of yours but thanks to you I found out all I need is a low pass filter and I could play my SACD. Except it didn't work. It turns out that DSD (PDM) is a digital signal and it requires a digital to analog converter to produce an analog signal. But on the bright side, you said I could see the analog in a DSD (PDM) signal so I took two leads and hooked up the SACD to an oscilloscope. That didn't work either. It turns out I need a digital to analog converter to do that too! But the good news is I could see the waveforms after I hooked up the DAC to the oscilloscope. The bad news is they didn't look anything like the original analog signal at all. There was this huge amount of noise that the original analog signal didn't have. I guess that's why additional filtering is required because the original DSD (PDM) file doesn't look anything like the original analog signal. On a lighter note, after using a digital to analog converter and a low pass filter I was finally able to get the DSD (PDM) to look just like the analog signal I recorded with PCM.
Jon, take a deep breath. I appreciate your snarky remarks. They remind me of me! PDM requires nothing more than a low pass filter to play analog. It's what we do in our "fancy DAC" that you could have mythically purchased. The output of the FPGA inside our DAC is a PDM stream and we simply run it through level shifters to increase the output from the LVDS output to a reasonable level, lkow pass it, and that's analog. But, thanks for the wonderfully snarky comment. Made my evening. :)
@@Paulmcgowanpsaudio So do I understand you correctly in that you are saying your field-programmable gate array is not programmed to be a digital to audio converter?
@@JonAnderhub The function is a D/A converter, yes. But what's relevant to our discussion is that the output of the FPGA in the DAC is PDM. That PDM stream is low pass filtered to become analog.
The question of whether DSD is digital is beside the point I think... The question is whether the signal medium (in this case it's digits, in others it is differences in magnetic flux as in tape, or physical grooves in vinyl) is 'analogous' - i.e. that it varies in a direct proportionate way the same as the source. I think that the density of ones in the DSD signal varies in an analogous manner to the loudnes of the original signal and therefore can be seen as 'analog' in the same way as vinyl records or tape. - Richard'
I see DSD as a representation of compressions and rarefaction of air pressure caused by sound. Compressions are represented by a greater concentration of ones. With PCM, compressions are represented by a higher 16 bit (or whatever word length) number. it's harder to see patterns relating to the sound when the 0s and 1s of the 16 bit numbers are laid out in sequence.
Hey Paul I've always loved your videos and have learned a lot from you. I'm just writing because I believe there was a bit of misunderstanding in the example of 24 frames per second you mentioned about movies. Actually movies are post processed at 24 fps because it gives the traditional cinematic effect most people are used to in movies, but our brains can discern motion at much higher frames per second.
I love that Paul recognizes that there is a point (frames per second) where humans see it as continuous (even though he might not be right about the 24 FPS thing) but he thinks that people can perceive the difference between 96K and 192K sampling rates, and he thinks that $5400 IEC cables sound better.
@@Wizardofgosz to be fair, he stated that about 25fps is where our brains start to se images as a continuous sequence, which is what the video industry settled for back in the days when video in the form of film had relatively limited performance. Nowadays we have monitors pulling 240fps (maybe more) and it's defenetly perceived easily by our minds. Audio bitstreaming speeds and cables, under the right chain of audio gear, are also easily discernible by us humans.
@@LouieFerdinand They're not. Cables can be shown to be identical with null tests, and I have yet to see a well done double blind test where anyone can tell the sample rates apart. Otherwise it's just narcissism and belief in magic.
@@Wizardofgosz I used to think like you until I gained more experience, but call it narcissism or magic was never among any of my thoughts related to this matter. You are a very funny man. Cheers!
@@LouieFerdinand You mean you used to believe evidence and science? But now you'd much rather believe that you can hear better than the rest of us, and that your stereo is better than everyone else's and can "resolve" more details? Geez, which one sounds like narcissism and arrogance to you? So what is your stereo resolving that mine isn't? Did mine forget to play a bass note or a cymbal crash? And did it do it on purpose? Like out of spite? Once you've paid $5 for a cable (probably $1) there is no audible difference between cables.
Hi Paul, thank you very much for your very interesting videos - I enjoy them very much since quite a while now! After watching this video I decided to send my very first question: Here in Germany many sellers of audiophile equipment use CDs of the well known label Chesky Records as 'reference' to impress possible customers. These are 128 x oversampled recordings on normal CD. Please correct me if I'm wrong, but this would mean 128 times 22.500 samples, approx. 2,2 million samples per second. This has to be reduced back to 2 x oversampling, of course, because im PCM this would mean 64 CD's of storage, resulting in 64 x 700 MB, about 45 GB of digital data... ...in times of Blu-Ray technology not really a big deal, but I only can guess that this may require powerful DA-decoders. So, since DSD 64 produces about 3 million bits per second (if I understood it right), the resulting file should be about 1/8 of a PCM 128 x oversampled file... ...a clear advantage - but comparing these two directly, not the downsampled 44.100 Hz PCM with DSD 64: Are there more advantages of DSD to mention? I understand your Statement, that in the end everything we do ist digital in some way - and I totally agree! So, keeping this in mind, both should be quite comparable? But maybe my argumentation is wrong at some point? Maybe you find this question interesting enough to discuss it officially... ...but maybe you discussed this already and I simply wasn't able to find the right video? All the best for you, your beloved ones and Octave Records, of course! Gerhard from Bavaria, Germany
Please give us samples of direct-to-DSD vs direct-to-PCM recordings so we can hear the difference ourselves. You should have plenty of these at Octave Records..
I wish it were so easy. RUclips of course doesn't play DSD nor does it allow for higher sample rate PCM. Perhaps at some point we can make some files available for download. Problem again is that not many people can natively play DSD files. But, good thought.
@@Paulmcgowanpsaudio Of course not on RUclips, you have websites right, can’t you upload samples there? You made dozens of videos praising the amazing benefits of DSD vs PCM for capturing analog, but won’t publish a couple 5 second samples to substantiate your claim?
@0:50 "Actually, I don't know why people get upset when I say this?" Who gets upset? Do you deem disagreement to be people being emotionally upset? Do people write to you that they are upset?
I would say, your body can sense frames up to 1000Hz frequencies. There were plenty of tests done but you can see one frame of light at 1000Hz, video games played upwards of 240Hz+ can be "felt" in a sense. Film is at 24fps because frame persistence is more cinematic and makes it look continuous. 60fps video and beyond just looks smoother and smoother. To say that that we can't sense it is false. Eyes have a radial acuity for a "pixel" in a sense for pixel density of film/movies/photos, it's the whole "retina" resolution from Apple. I'd feel there's a same way with hearing audio sampling, but it's a matter that you want to effectively raise your sampling rate (i.e. frame rate, pixels, etc) so that goes beyond the capabilities of the human body to tell a difference. You're basically at the point where you want to over sample but then the way you really tell analog vs digital is that digital is a binary result, it's either there or not. Analog has noise/artifacts that can still be introduced to the signal but still present the same. But it is interesting that you could play DSD through a speaker direct and get audio. Maybe that would be something to demonstrate in a video playing back raw binary through a speaker (I know you'd need to attenuate a PCM file, but it would still be good to demo it) and compare with oscilloscope readings of each compared to an analog signal.
Hi Paul, Is DSD recording same as the old Laserdisc which is purely analog system only stored as 1 and 0 on the disk? Which explains why pure DSD recordings sounds analog? Thank you.
Would like to see (hear actually) a demonstation of DSD going directly (no DAC or low pass filter) to a amplified speaker. Will that be in an upcoming video?
I would like to see that myself. But if the DSD signal has about 3M transitions per second, in order for any signal to reach the speaker that amplifier would have to have an impressive frequency response.
@Douglas Blake Don't be silly. What's feeding the filter? An output from a digital device or buffer? How much power supply noise is there? How symmetrical are the edges? How steady are the logic levels? You can't get low distortion and noise, directly using the output of a digital device. It has to be "squared up" by a 1-bit DAC - which is simply a precision analog switch which selects one of two reference voltages. The same thing that's needed in the sigma-delta modulator loop that created the bit stream in the first place.
@@marianneoelund2940 Thanks for the clarification. So DSD to analog is much more involved than "just adding a simple low pass filter" that is suggested in the video.
Planck, Length and Time. Laws of physics. Proposed by and named after Max Planck, Planck length is the shortest distance measurable. You can divide things, cut them in half, ... to that point. You can not cut something that is a Planck Length in half, trim off an end... Planck Time is the shortest measurable instant. So yes we are made up of little tiny Plank Length sized bits.
@0:55 "The whole world is made up of bits" @1:27 "The idea of analog is false" You go on to conflate continuity with analog. That is a faulty premise. Analog maintains its state. It does not transition from state to state. Whereas, digital takes a voltage and applies a mathematical formula to that voltage and a value (a number) is derived from that voltage/math marriage. You no longer have the voltage. You now have a number for that instance of voltage. Yes, the whole world is made up of bits. But not all bits are the same as other bits. Some bits remain the bits that they are. Other bits get recorded as a number for the purpose of representing those bits. Digital can continue to improve over time, even to the point of sounding far better than analog and at a fraction of today's cost. But it is still not analog. It is still digital. A cartoon can be made to look as real as can be. But it is still what it is.
@Douglas Blake DSD/PDM is not FM at all. DSD is digital entirely and FM is analog entirely. The perception of what is "most analog" between PDM and PCM is a matter of how you view it between slope sampling vs level sampling. Either way, you get closer to analog by higher specs (sample rate, resolution).
@Douglas Blake the wave form of FM is entirely analog. You frequency modulate a carrier sine wave with an analog audio wave. You wouldn’t want digital steps on an FM signal because it would mean out of band harmonics (FCC violations).
@Douglas Blake Modulation can not be actual digital RF on FM but you can send digital data using different frequencies. RDS, HD Radio, DAB radio etc. do such thing on the FM band.
@Douglas Blake LOL, yes of course I can see the pulse density varies DIGITALLY according to the analog wave forms. The pulses will be 0s or 1s in the slots of the clock but everything is entirely DIGITAL. You can't make half a bit value or add a pulse in a slot between two slots. Everything is clocked precisely with bits that can be only binary. This is how DIGITAL data flows look like. You can claim it looks analog, but it's with a digital "state space" of values like PCM is. When you observe PCM data you need to look at the data value per sample (e.g. 16 bits) and you can also argue it looks analog. Still it's digital.
@Douglas Blake No, the pulse density is digitally derived from the analog wave and is digital by nature as the density is limited to the time slots where you can sample the sigma delta bits equal to how PCM is limited to the time slots where you sample the level. Either way, you can't just map the signal flexibly through the ADC to match the exact curvature of the analog signal and quantization errors are introduced. DSD is Direct Stream Digital and not Direct Stream Analog. Theoretically, someone could make the PDM with flexible timing gaps but when you translate it into a computer, that would mean your time gaps would have to be digitized also into numerical values thus even such concept would still be digital.
why not just use analogue; you can edit that! actually the world is not made of bits, everything is a wave, read Carver Mead. It is particles that are the illusion. whether or not something appears as a wave or a particle depends on coherence
The Beauty of a human mind is we create the illusion that we call "time". There is no such thing. The science behind it is called Quantum Physics. And Yes Paul! I'm on the DSD team as well! Many Thanks!
Hey, thanks Paul!
I was going to buy one of those fancy DACs (Digital to Audio Converters) of yours but thanks to you I found out all I need is a low pass filter and I could play my SACD.
Except it didn't work.
It turns out that DSD (PDM) is a digital signal and it requires a digital to analog converter to produce an analog signal.
But on the bright side, you said I could see the analog in a DSD (PDM) signal so I took two leads and hooked up the SACD to an oscilloscope.
That didn't work either.
It turns out I need a digital to analog converter to do that too!
But the good news is I could see the waveforms after I hooked up the DAC to the oscilloscope.
The bad news is they didn't look anything like the original analog signal at all.
There was this huge amount of noise that the original analog signal didn't have.
I guess that's why additional filtering is required because the original DSD (PDM) file doesn't look anything like the original analog signal.
On a lighter note, after using a digital to analog converter and a low pass filter I was finally able to get the DSD (PDM) to look just like the analog signal I recorded with PCM.
Jon, take a deep breath. I appreciate your snarky remarks. They remind me of me! PDM requires nothing more than a low pass filter to play analog. It's what we do in our "fancy DAC" that you could have mythically purchased. The output of the FPGA inside our DAC is a PDM stream and we simply run it through level shifters to increase the output from the LVDS output to a reasonable level, lkow pass it, and that's analog.
But, thanks for the wonderfully snarky comment. Made my evening. :)
@@Paulmcgowanpsaudio So do I understand you correctly in that you are saying your field-programmable gate array is not programmed to be a digital to audio converter?
@@JonAnderhub The function is a D/A converter, yes. But what's relevant to our discussion is that the output of the FPGA in the DAC is PDM. That PDM stream is low pass filtered to become analog.
The question of whether DSD is digital is beside the point I think... The question is whether the signal medium (in this case it's digits, in others it is differences in magnetic flux as in tape, or physical grooves in vinyl) is 'analogous' - i.e. that it varies in a direct proportionate way the same as the source. I think that the density of ones in the DSD signal varies in an analogous manner to the loudnes of the original signal and therefore can be seen as 'analog' in the same way as vinyl records or tape.
- Richard'
GLOSS - TER is an ancient market town in the county of GLOSS - TER - SHIRE in the UK
I see DSD as a representation of compressions and rarefaction of air pressure caused by sound. Compressions are represented by a greater concentration of ones. With PCM, compressions are represented by a higher 16 bit (or whatever word length) number. it's harder to see patterns relating to the sound when the 0s and 1s of the 16 bit numbers are laid out in sequence.
Paul - you are right. Gloucester is in the UK and FYI it's pronounced "Gloster" as in "foster" or "roster". I loved your attempt though! :-)
Hey Paul I've always loved your videos and have learned a lot from you. I'm just writing because I believe there was a bit of misunderstanding in the example of 24 frames per second you mentioned about movies. Actually movies are post processed at 24 fps because it gives the traditional cinematic effect most people are used to in movies, but our brains can discern motion at much higher frames per second.
I love that Paul recognizes that there is a point (frames per second) where humans see it as continuous (even though he might not be right about the 24 FPS thing) but he thinks that people can perceive the difference between 96K and 192K sampling rates, and he thinks that $5400 IEC cables sound better.
@@Wizardofgosz to be fair, he stated that about 25fps is where our brains start to se images as a continuous sequence, which is what the video industry settled for back in the days when video in the form of film had relatively limited performance. Nowadays we have monitors pulling 240fps (maybe more) and it's defenetly perceived easily by our minds.
Audio bitstreaming speeds and cables, under the right chain of audio gear, are also easily discernible by us humans.
@@LouieFerdinand They're not. Cables can be shown to be identical with null tests, and I have yet to see a well done double blind test where anyone can tell the sample rates apart.
Otherwise it's just narcissism and belief in magic.
@@Wizardofgosz I used to think like you until I gained more experience, but call it narcissism or magic was never among any of my thoughts related to this matter. You are a very funny man. Cheers!
@@LouieFerdinand You mean you used to believe evidence and science? But now you'd much rather believe that you can hear better than the rest of us, and that your stereo is better than everyone else's and can "resolve" more details?
Geez, which one sounds like narcissism and arrogance to you?
So what is your stereo resolving that mine isn't? Did mine forget to play a bass note or a cymbal crash? And did it do it on purpose? Like out of spite?
Once you've paid $5 for a cable (probably $1) there is no audible difference between cables.
The first example of digital communication, perhaps the telegraph? Then AM radio with a sampling rate of 1Mhz.
Hi Paul,
thank you very much for your very interesting videos - I enjoy them very much since quite a while now!
After watching this video I decided to send my very first question:
Here in Germany many sellers of audiophile equipment use CDs of the well known label Chesky Records as 'reference' to impress possible customers. These are 128 x oversampled recordings on normal CD. Please correct me if I'm wrong, but this would mean 128 times 22.500 samples, approx. 2,2 million samples per second.
This has to be reduced back to 2 x oversampling, of course, because im PCM this would mean 64 CD's of storage, resulting in 64 x 700 MB, about 45 GB of digital data... ...in times of Blu-Ray technology not really a big deal, but I only can guess that this may require powerful DA-decoders.
So, since DSD 64 produces about 3 million bits per second (if I understood it right), the resulting file should be about 1/8 of a PCM 128 x oversampled file...
...a clear advantage - but comparing these two directly, not the downsampled 44.100 Hz PCM with DSD 64: Are there more advantages of DSD to mention?
I understand your Statement, that in the end everything we do ist digital in some way - and I totally agree!
So, keeping this in mind, both should be quite comparable? But maybe my argumentation is wrong at some point?
Maybe you find this question interesting enough to discuss it officially...
...but maybe you discussed this already and I simply wasn't able to find the right video?
All the best for you, your beloved ones and Octave Records, of course!
Gerhard from Bavaria, Germany
Please give us samples of direct-to-DSD vs direct-to-PCM recordings so we can hear the difference ourselves. You should have plenty of these at Octave Records..
I wish it were so easy. RUclips of course doesn't play DSD nor does it allow for higher sample rate PCM. Perhaps at some point we can make some files available for download. Problem again is that not many people can natively play DSD files. But, good thought.
On RUclips? dream on
@@Paulmcgowanpsaudio Of course not on RUclips, you have websites right, can’t you upload samples there?
You made dozens of videos praising the amazing benefits of DSD vs PCM for capturing analog, but won’t publish a couple 5 second samples to substantiate your claim?
@@f1shb0n3d Yup we do and perhaps that's something we can get to at some point. Thanks for the suggestion.
Can I get Van Morrison on DSD?
@0:50 "Actually, I don't know why people get upset when I say this?"
Who gets upset?
Do you deem disagreement to be people being emotionally upset?
Do people write to you that they are upset?
I would say, your body can sense frames up to 1000Hz frequencies. There were plenty of tests done but you can see one frame of light at 1000Hz, video games played upwards of 240Hz+ can be "felt" in a sense. Film is at 24fps because frame persistence is more cinematic and makes it look continuous. 60fps video and beyond just looks smoother and smoother. To say that that we can't sense it is false. Eyes have a radial acuity for a "pixel" in a sense for pixel density of film/movies/photos, it's the whole "retina" resolution from Apple. I'd feel there's a same way with hearing audio sampling, but it's a matter that you want to effectively raise your sampling rate (i.e. frame rate, pixels, etc) so that goes beyond the capabilities of the human body to tell a difference. You're basically at the point where you want to over sample but then the way you really tell analog vs digital is that digital is a binary result, it's either there or not. Analog has noise/artifacts that can still be introduced to the signal but still present the same.
But it is interesting that you could play DSD through a speaker direct and get audio. Maybe that would be something to demonstrate in a video playing back raw binary through a speaker (I know you'd need to attenuate a PCM file, but it would still be good to demo it) and compare with oscilloscope readings of each compared to an analog signal.
There is no spoon
Oh, the comment section is mighty tasty.
Hi Paul,
Is DSD recording same as the old Laserdisc which is purely analog system only stored as 1 and 0 on the disk? Which explains why pure DSD recordings sounds analog?
Thank you.
If memory serves me Laserdiscs were actually analog using FM modulation
@@Paulmcgowanpsaudio Yes Paul I believe that is correct, so is DSD same like FM modulation where the distance between the 1 and 0 is pure analog?
Would like to see (hear actually) a demonstation of DSD going directly (no DAC or low pass filter) to a amplified speaker. Will that be in an upcoming video?
I would like to see that myself. But if the DSD signal has about 3M transitions per second, in order for any signal to reach the speaker that amplifier would have to have an impressive frequency response.
You still need to have a 1-bit DAC which will precisely set the two output levels, and produce clean, balanced transitions.
@Douglas Blake
Don't be silly. What's feeding the filter? An output from a digital device or buffer? How much power supply noise is there? How symmetrical are the edges? How steady are the logic levels?
You can't get low distortion and noise, directly using the output of a digital device. It has to be "squared up" by a 1-bit DAC - which is simply a precision analog switch which selects one of two reference voltages. The same thing that's needed in the sigma-delta modulator loop that created the bit stream in the first place.
@@marianneoelund2940 Thanks for the clarification. So DSD to analog is much more involved than "just adding a simple low pass filter" that is suggested in the video.
Planck, Length and Time. Laws of physics. Proposed by and named after Max Planck, Planck length is the shortest distance measurable. You can divide things, cut them in half, ... to that point. You can not cut something that is a Planck Length in half, trim off an end... Planck Time is the shortest measurable instant. So yes we are made up of little tiny Plank Length sized bits.
We are nowhere near able to sample at that level in real time though so this argument doesn't matter...yet
@@shayhan6227 It seems that sampling rate stays ahead of data transfer and storage size needs.
@0:55 "The whole world is made up of bits"
@1:27 "The idea of analog is false"
You go on to conflate continuity with analog. That is a faulty premise.
Analog maintains its state. It does not transition from state to state.
Whereas, digital takes a voltage and applies a mathematical formula to that voltage and a value (a number) is derived from that voltage/math marriage. You no longer have the voltage. You now have a number for that instance of voltage.
Yes, the whole world is made up of bits. But not all bits are the same as other bits.
Some bits remain the bits that they are. Other bits get recorded as a number for the purpose of representing those bits.
Digital can continue to improve over time, even to the point of sounding far better than analog and at a fraction of today's cost. But it is still not analog. It is still digital.
A cartoon can be made to look as real as can be. But it is still what it is.
By your definition above, as there is no math involved to playback a dsd stream (just an analog low pass filter), you should be classing it as analog.
@@alex_madeira I think that you replied to the wrong comment, as it is unrelated to my comment.
everything is waves whether or not it looks like matter depends on coherence/ incoherence of the constituent waves
DSD is digital (Direct Stream Digital).
@Douglas Blake DSD/PDM is not FM at all. DSD is digital entirely and FM is analog entirely. The perception of what is "most analog" between PDM and PCM is a matter of how you view it between slope sampling vs level sampling. Either way, you get closer to analog by higher specs (sample rate, resolution).
@Douglas Blake the wave form of FM is entirely analog. You frequency modulate a carrier sine wave with an analog audio wave. You wouldn’t want digital steps on an FM signal because it would mean out of band harmonics (FCC violations).
@Douglas Blake Modulation can not be actual digital RF on FM but you can send digital data using different frequencies. RDS, HD Radio, DAB radio etc. do such thing on the FM band.
@Douglas Blake LOL, yes of course I can see the pulse density varies DIGITALLY according to the analog wave forms. The pulses will be 0s or 1s in the slots of the clock but everything is entirely DIGITAL. You can't make half a bit value or add a pulse in a slot between two slots. Everything is clocked precisely with bits that can be only binary. This is how DIGITAL data flows look like. You can claim it looks analog, but it's with a digital "state space" of values like PCM is. When you observe PCM data you need to look at the data value per sample (e.g. 16 bits) and you can also argue it looks analog. Still it's digital.
@Douglas Blake No, the pulse density is digitally derived from the analog wave and is digital by nature as the density is limited to the time slots where you can sample the sigma delta bits equal to how PCM is limited to the time slots where you sample the level. Either way, you can't just map the signal flexibly through the ADC to match the exact curvature of the analog signal and quantization errors are introduced. DSD is Direct Stream Digital and not Direct Stream Analog. Theoretically, someone could make the PDM with flexible timing gaps but when you translate it into a computer, that would mean your time gaps would have to be digitized also into numerical values thus even such concept would still be digital.
It's pronounced Glos-ter
why not just use analogue; you can edit that!
actually the world is not made of bits, everything is a wave, read Carver Mead. It is particles that are the illusion. whether or not something appears as a wave or a particle depends on coherence
Made in Canada.
The Beauty of a human mind is we create the illusion that we call "time". There is no such thing. The science behind it is called Quantum Physics.
And Yes Paul! I'm on the DSD team as well!
Many Thanks!
You didn't explain how DSD doesn't require conversion to true analogue in order to be played via speakers. 🔊