This is why DSD is an obsolete system. PCM can give superb results, it is down to the Producer to take care and not screw it up so it sounds better on a car radio.
Imagine a 32bit parallel DSD decoder that independently decodes 32 Streams of the different Vocals and Instruments in a recording, and they are then mixed together in the analog Domain, that would be Something special.
treat it like analog and you can mix it just fine. you can change levels, send it through processors, sum channels, literally everything you can do to an analog signal. the benefit of dsd, aside from its simplicity, you get all of the information in a fraction of the space hi-res pcm uses, with fewer components and using less power, and it can be encoded into any format in the end.
DSD works this way: a 1 means the signal is going up just a little bit, a 0 means it is going down that same amount. If I want to build a large signal, I need to stream many 1's over a period of time. This is the issue: a signal level gets built over time rather than being one sample like in PCM. To reduce or increase the level of a DSD signal I need to change the amount of time I have the 1's on, essentially, but keeping the same overall density of 1's and 0's so as not to deform the signal. This is why it is hard to do.
It’s hard. It’s not impossible. Precisely as you mention - the computer program just would need to map all the 1 and 0s, map the wave, then manipulate it on amplitude There’s just no one working on it, so no viable financial solutions out there
@@rrricmat Understood. Ultimately it means grabbing a bunch of those 1's and 0's into one multibit sample, then resizing that, and finally recreate an equivalent DSD stream. This is exactly what DSD -> PCM -> DSD does.
@@rrricmat Another way is to take each single bit, assume it's a 5 bit, apply volume to that (so in essence you get a fraction of a bit) and then remap that by spreading it back into 1's an 0's that give you the new volume level. I believe this is what the Sonoma workstation does.
So if are taking a high resolution system like DSD and converting it to a lower resolution PCM system to manipulate and mix the recording then taking that lower resolution bit stream and converting it back to DSD are you really getting a DSD recording. Why not just record in PCM do all your mixing then convert to DSD and save a conversion.
You are correct to think that it would be better to record in high-resolution PCM and skip the double conversion process which adds errors and distortions. But just an FYI the PCM isn't a lower resolution. In fact, it's a higher resolution and there is a loss going back to the DSD 64 for SACD.
With DSD, analog voltage amplitude is encoded digitally by varying the density of the 1 bit pulses. No sound, no pulses. Lots of sound, lots of pulses.
I think a good way to explain it is comparing it to dithering using only fully black and fully white dots in a grid to make grayscale images. Doing so requires a much higher DPI to get similar quality to a lower resolution 8-bit grayscale display, but can be done.
So if I understand you correctly, basically you are saying that the initial DSD recording process is pretty much useless (which is what Sony figured out years ago) and that all the PDM (DSD) has to be converted to PCM for any kind of useful editing. So SACD discs by Octave Records are really just DSD (PDM) conversions from high definition PCM.
That's how the industry works to sell brain dead audiophiles useless stuff for a lot of money - the best players sell the same recording 20 times and the moron which hears expensive cables in sound will be happy because the imagination of something special - see hires where what you hear is a different mix, the same for SACD
can you connect a amp and dac combo into a audio mixer I really want to use iFi Zen CAN and a iFi Zen DAC V2 with a Razer Audio Mixer can you please help me out
OK, I get it that one cannot do a math on one bit data. But if you take multiple bits over some period of time? Let's say, a millisecond. You get a sample which you can work with, can't you? In "red book" PCM, you have 44.100 samples per second, each 16 bits. Which corresponds to about 2.800.000 samples in PDM, each one bit. So there are 63 PDM samples for each PCM sample. Can't you make calculation on 63 samples, which already form some part of the sinewave? Just a question...
Why not make copies? Take the 1 bit stream and multiply it by say 100. And from the first one. It’s volume or voltage is the max. Then that 100th copy is voltage zero. And you have your steps down in-between. So as you move the volume slider. It’s goes between those copies. More or less fixed points and I’m sure a lot of data. But would that work?
It seems if you multiplied the bitstream by 65536 then subtracted 32768 then applied a digital low-pass filter of approprate parameters, you'd basically have 16-bit PCM but at the DSD bitrate, which could be downsampled to a lower bitrate without aliasing since the low-pass filter was applied first.
So does that mean Octave records live off the floor (using only the mixing and levels on the live recording console) without any post editing? If that is true, how would an artist utilize samples, sonic effects, layering (such as doubling a vocal), and so forth? Or don’t they?
Octave Records is currently moving to a Pyramix multitrack system which converts the PDM (DSD) stream into a high definition PCM file for editing purposes such as you are looking for. The system then converts the PCM back to DSD.
@@JonAnderhub record the playback while you are doing the additional tracks. Now that I think about idk if that's how they did or not, but it's one way to do it
David, mic’ing a monitor is lossy, meaning the sound degrades. And converting PDM to PCM, must also be lossy. Too bad we can’t maintain pristine signal and have editing.
I have to respectfully disagree. While it's true that it would be impossible to make any meaningful transformation on a PDM bitstream looking at only one bit at a time, that's completely irrelevant. No one said computers were limited to processing one bit. In the same way that sample rate conversion in PCM requires looking at many consecutive samples to generate one output sample, manipulation of PDM bitstreams requires looking back tens or even hundreds of microseconds to know how to interpret the most recent bit. But computers are very good at this. Because this is fundamentally a math problem involving integers (after you've thrown away whatever information is lost when the signal is initially encoded) there is an exact, finite, solution. Again, this solution is easy to compute. Once that's identified, it is necessary one more time to decimate into the available resolution of the format. Why is this hard? It isn't. Why is it not available to consumers? If that's true, it probably has to do with market forces, format wars, back room deals between vendors, you know, the usual suspects. But there's no technical reason.
I find this explanation confusing. With PCM, the Bit Depth is used to encode the signal amplitude with a value between 0 (no sound) and about 65,000 (clipping). But with PDM, it's the DENSITY of pulses that encodes the amplitude, not the Bit Depth. No sound, no pulses. Lots of sound, lots of pulses. Medium sound, a medium density of pulses. Given that, why can't you just add more pulses to get higher amplitude? (Or take some pulses away to lower the amplitude?) What am I getting wrong?
What I'm trying to figure out is how the negative voltage gets encoded. Sound isn't just a 0V to some max V like image data, but varies from negative max V to positive max V, and 16-bit PCM encodes this as -32768 to +32768, with 0 corresonding to 0V. However, with DSD, it seems that no pulses (string of 0-bits) would have to represent negative max voltage and a string of 1-bits would have to represent positive max voltage, with 0V represented by an equal mix of 0-bits and 1-bits in a string? It's the only way I can think of for DSD to represent the negatives and positives as well, and would love an explanation if I'm wrong.
@@sub-vibes So it seems by what you're saying that it's the first derivative of the signal that controls the bitstream and of course, DSD has to use a much higher rate to be able to capture the frequencies. I was trying to think of it in comparison to dithering only pure black and pure white dots to print grayscale images, which requires a higher dot per inch than a process with 256 levels of gray would to get the same level of quality. Images don't have negative values though and sound does and that's what I was trying to work out. Then when signals are added you have to worry about normalization and not letting the signal clip.
Oh yes we could certainly use FPGA on DSD, and Sony, Korg, CEDAR, Prism etc released workstations which can do advanced processing on DSD. Problem is the licensing on the players and the edit platforms is all a legal swamp, and none of the DSD DAWs were much good. Under developed, and nobody can play the files, SACD never caught on. We can do math on bitstream, sure. Think of it as equivalent to PWM encoding and it makes fine sense, and if you look at a bitstream it's the hold-value of the bit sequence that encodes the pulse width - so a pulse will read in essence as forty thousand zeroes in a row followed by forty thousand ones in a row, and that is roughly equivalent to a 384kHz 24 bit word sample saying "null". See? Resampling and dithering are a fascinating space. We could resample a signal of 00110011 to 0101010101 and have no detectable change in the analogue output. Any mathematical operation being performed on a bitstream would require time framing and resampling, *but* in contrast to a PCM windowed system, you can have arbitrary process frame alignment with the time code. That is a massive advantage for DSP, actually.
Modern audio architectures in digital music production, distribution and playback heavily involve math and therefore DSD is not viable. Besides, any digitally mastering of music is producing PCM (DXD or other) at the output and if you down-convert it to DSD, you just made a lossy version. DSD used in the backend (mic ADC) of the recording process and for making a digital copy of an analog master is where DSD has most value. Reality is that sigma delta conversion can sound great (the argument for DSD) but PCM can also utilize that.
@@edmaster3147I don't think there is such a thing as any music that has not gone through silicone chips produced in the last 40+ years. This includes vinyl discs and tape produced entirely in analog process, where lots of op-amps are involved. I wouldn't worry about audio having gone through silicone chips as a measure of audio fidelity. Audio gets damaged by low quality or poorly engineered analog circuits, digital circuits or DSP code.
@@ThinkingBetter I think you are right. It's like what I often say, a painting will never be something else then another reality, even capable of causing an stunning experience, yet never be the real thing. It's the same with music, playback will never be reality yet can be 'the next-best thing'. And truth be told, I wouldn't like many musicians in my house in reality either.
@@edmaster3147 The notion that live music is superior in audio fidelity to studio produced music is almost 100% false. I’ve been to lots of major concerts from Pink Floyd, Dire Straits, Elton John etc. decades ago to John Mayor, Cold Play, Imagine Dragon etc. more recently and NEVER does live sound better than the same songs played on my own system. But a live experience is much more than about audio fidelity and remembering a live concert adds a lot to how you feel when listening to the same tracks at home. Music listening is all about emotions.
@@ThinkingBetter True. But it depends on the setting. Some smaller venues are really nice to be present and hear. At the big concerts its just way to loud and the earplugs are needed. And what to say of classical music in concert halls...
Because Sony and RIAA and no DAT for you mate. Home taping is killing music. All the digital encoding stuff is highly restricted apart from personal computers. We are living in a medieval society. When the Sony patents expire we will get MHz bitstream encoding in hardware and computers, but at the moment Sony made bad deals in the '80s and '90s, and the US government idiots flushed the future of high fidelity audio down the toilet.
Love PS audio. When you look at all the limitations of DSD, i.e. not MOST user friendly, it just seems the ONLY good thing about it is higher music quality. Wish is wasn't so. Most of us just want to play high quality music without going through all these difficulties. I just don't see DSD really making it for the majority of us audiophiles. There reaches a point, where someone will only going to so far to reach "perfection". For me, DSD is just too limiting and difficult. I'll stick with CD, that's just me though. I really wish Octave Records would sell "regular" CD's at an affordable price as well as their other formats. What many companies don't seem to realize is most audiophiles don't want to go buy new equipment just to use a new audio format. Equipment is way too expensive these days to give it all up and start again. Believe me, ALL the people I know are not going to invest in DSD compatible equipmet just to play a very limited amount of DSD files.
Mixing DSD isn't a problem technically. The mathematics are easy enough to work out. But there's no software and no equipment that supports working with DSD, because no-one ever uses it, because IT SUCKS! It has no advantages over PCM, at all. All it gives you is a higher quantisation noise.
You can't mix in PCM too. Both DSD and PCM need to convert to binary in order for CPU to read and manipulate the data and convert back to PCM or DSD. No one doing DSD mix in binary because DSD and PCM in conversion gonna be the same and zero losses when you set the correct sampling rate relative to one another.
This is why DSD is an obsolete system. PCM can give superb results, it is down to the Producer to take care and not screw it up so it sounds better on a car radio.
Imagine a 32bit parallel DSD decoder that independently decodes 32 Streams of the different Vocals and Instruments in a recording, and they are then mixed together in the analog Domain, that would be Something special.
Enter the analog domain would be special stupid
treat it like analog and you can mix it just fine. you can change levels, send it through processors, sum channels, literally everything you can do to an analog signal. the benefit of dsd, aside from its simplicity, you get all of the information in a fraction of the space hi-res pcm uses, with fewer components and using less power, and it can be encoded into any format in the end.
DSD works this way: a 1 means the signal is going up just a little bit, a 0 means it is going down that same amount. If I want to build a large signal, I need to stream many 1's over a period of time. This is the issue: a signal level gets built over time rather than being one sample like in PCM. To reduce or increase the level of a DSD signal I need to change the amount of time I have the 1's on, essentially, but keeping the same overall density of 1's and 0's so as not to deform the signal. This is why it is hard to do.
It’s hard. It’s not impossible.
Precisely as you mention - the computer program just would need to map all the 1 and 0s, map the wave, then manipulate it on amplitude
There’s just no one working on it, so no viable financial solutions out there
@@rrricmat Understood. Ultimately it means grabbing a bunch of those 1's and 0's into one multibit sample, then resizing that, and finally recreate an equivalent DSD stream. This is exactly what DSD -> PCM -> DSD does.
@@rrricmat Another way is to take each single bit, assume it's a 5 bit, apply volume to that (so in essence you get a fraction of a bit) and then remap that by spreading it back into 1's an 0's that give you the new volume level. I believe this is what the Sonoma workstation does.
@@miguelbarrio we wouldn’t need to move to PCM. At 1MHz sample rate, it already looks physically like a analog wave. So just need to manipulate that
@@miguelbarrio doesn’t work that way
The 1 but system provides delta information, not absolute information
Hence the need to map the entire wave first
So if are taking a high resolution system like DSD and converting it to a lower resolution PCM system to manipulate and mix the recording then taking that lower resolution bit stream and converting it back to DSD are you really getting a DSD recording. Why not just record in PCM do all your mixing then convert to DSD and save a conversion.
You are correct to think that it would be better to record in high-resolution PCM and skip the double conversion process which adds errors and distortions.
But just an FYI the PCM isn't a lower resolution.
In fact, it's a higher resolution and there is a loss going back to the DSD 64 for SACD.
Because audiophiles have no brain
Next: Why and how oes DSD/PDM use 1 bit? What is Pulse Density Modulation?
It's already there
With DSD, analog voltage amplitude is encoded digitally by varying the density of the 1 bit pulses. No sound, no pulses. Lots of sound, lots of pulses.
I think a good way to explain it is comparing it to dithering using only fully black and fully white dots in a grid to make grayscale images. Doing so requires a much higher DPI to get similar quality to a lower resolution 8-bit grayscale display, but can be done.
So if I understand you correctly, basically you are saying that the initial DSD recording process is pretty much useless (which is what Sony figured out years ago) and that all the PDM (DSD) has to be converted to PCM for any kind of useful editing.
So SACD discs by Octave Records are really just DSD (PDM) conversions from high definition PCM.
That's how the industry works to sell brain dead audiophiles useless stuff for a lot of money - the best players sell the same recording 20 times and the moron which hears expensive cables in sound will be happy because the imagination of something special - see hires where what you hear is a different mix, the same for SACD
can you connect a amp and dac combo into a audio mixer I really want to use iFi Zen CAN and a iFi Zen DAC V2 with a Razer Audio Mixer can you please help me out
Let's hear the modular synth mixed in. 😀
Great explanation Paul👍
Perfect explanation.
OK, I get it that one cannot do a math on one bit data. But if you take multiple bits over some period of time? Let's say, a millisecond. You get a sample which you can work with, can't you? In "red book" PCM, you have 44.100 samples per second, each 16 bits. Which corresponds to about 2.800.000 samples in PDM, each one bit. So there are 63 PDM samples for each PCM sample. Can't you make calculation on 63 samples, which already form some part of the sinewave? Just a question...
If I hear "DSD is a one bit system" one more time I am going to scream. What does that mean?
A bit is either 0 or 1. And the signal has only 1 bit. So it's like a light-switch. On or off. High or Low. etc.
Why not make copies? Take the 1 bit stream and multiply it by say 100. And from the first one. It’s volume or voltage is the max. Then that 100th copy is voltage zero. And you have your steps down in-between. So as you move the volume slider. It’s goes between those copies. More or less fixed points and I’m sure a lot of data. But would that work?
If you take a 1-bit stream and multiply the bits by 24 or 32 you wind up with a high-quality PCM file.
No need to multiply the bits by 100.
It seems if you multiplied the bitstream by 65536 then subtracted 32768 then applied a digital low-pass filter of approprate parameters, you'd basically have 16-bit PCM but at the DSD bitrate, which could be downsampled to a lower bitrate without aliasing since the low-pass filter was applied first.
So does that mean Octave records live off the floor (using only the mixing and levels on the live recording console) without any post editing? If that is true, how would an artist utilize samples, sonic effects, layering (such as doubling a vocal), and so forth? Or don’t they?
Octave Records is currently moving to a Pyramix multitrack system which converts the PDM (DSD) stream into a high definition PCM file for editing purposes such as you are looking for.
The system then converts the PCM back to DSD.
The same way they did it before daws and digital
@@Pete.across.the.street Multitrack tape and a razor blade?
@@JonAnderhub record the playback while you are doing the additional tracks. Now that I think about idk if that's how they did or not, but it's one way to do it
David, mic’ing a monitor is lossy, meaning the sound degrades. And converting PDM to PCM, must also be lossy. Too bad we can’t maintain pristine signal and have editing.
I have to respectfully disagree. While it's true that it would be impossible to make any meaningful transformation on a PDM bitstream looking at only one bit at a time, that's completely irrelevant. No one said computers were limited to processing one bit. In the same way that sample rate conversion in PCM requires looking at many consecutive samples to generate one output sample, manipulation of PDM bitstreams requires looking back tens or even hundreds of microseconds to know how to interpret the most recent bit. But computers are very good at this. Because this is fundamentally a math problem involving integers (after you've thrown away whatever information is lost when the signal is initially encoded) there is an exact, finite, solution. Again, this solution is easy to compute. Once that's identified, it is necessary one more time to decimate into the available resolution of the format.
Why is this hard? It isn't. Why is it not available to consumers? If that's true, it probably has to do with market forces, format wars, back room deals between vendors, you know, the usual suspects. But there's no technical reason.
I find this explanation confusing.
With PCM, the Bit Depth is used to encode the signal amplitude with a value between 0 (no sound) and about 65,000 (clipping).
But with PDM, it's the DENSITY of pulses that encodes the amplitude, not the Bit Depth. No sound, no pulses. Lots of sound, lots of pulses. Medium sound, a medium density of pulses.
Given that, why can't you just add more pulses to get higher amplitude? (Or take some pulses away to lower the amplitude?)
What am I getting wrong?
PDM (as implemented) is strictly binary (0 and 1), a _stream_ of which encodes the level.
What I'm trying to figure out is how the negative voltage gets encoded. Sound isn't just a 0V to some max V like image data, but varies from negative max V to positive max V, and 16-bit PCM encodes this as -32768 to +32768, with 0 corresonding to 0V. However, with DSD, it seems that no pulses (string of 0-bits) would have to represent negative max voltage and a string of 1-bits would have to represent positive max voltage, with 0V represented by an equal mix of 0-bits and 1-bits in a string? It's the only way I can think of for DSD to represent the negatives and positives as well, and would love an explanation if I'm wrong.
@@sub-vibes So it seems by what you're saying that it's the first derivative of the signal that controls the bitstream and of course, DSD has to use a much higher rate to be able to capture the frequencies. I was trying to think of it in comparison to dithering only pure black and pure white dots to print grayscale images, which requires a higher dot per inch than a process with 256 levels of gray would to get the same level of quality. Images don't have negative values though and sound does and that's what I was trying to work out. Then when signals are added you have to worry about normalization and not letting the signal clip.
@@3Cr15w311 Think of DSD as pure black dots ( 1 bits) on white paper (0 bits) and you will come closer to the PDM analogy.
PCM 65k step signal or DSD -1,0,+1 are not binary data that computer can understand. Why people alway get this wrong.
Quantum Computing with Qubits will FIX this problem with DSD or some other new (yet developed) music format. :)
Oh yes we could certainly use FPGA on DSD, and Sony, Korg, CEDAR, Prism etc released workstations which can do advanced processing on DSD. Problem is the licensing on the players and the edit platforms is all a legal swamp, and none of the DSD DAWs were much good. Under developed, and nobody can play the files, SACD never caught on. We can do math on bitstream, sure. Think of it as equivalent to PWM encoding and it makes fine sense, and if you look at a bitstream it's the hold-value of the bit sequence that encodes the pulse width - so a pulse will read in essence as forty thousand zeroes in a row followed by forty thousand ones in a row, and that is roughly equivalent to a 384kHz 24 bit word sample saying "null". See?
Resampling and dithering are a fascinating space. We could resample a signal of 00110011 to 0101010101 and have no detectable change in the analogue output.
Any mathematical operation being performed on a bitstream would require time framing and resampling, *but* in contrast to a PCM windowed system, you can have arbitrary process frame alignment with the time code. That is a massive advantage for DSP, actually.
Modern audio architectures in digital music production, distribution and playback heavily involve math and therefore DSD is not viable. Besides, any digitally mastering of music is producing PCM (DXD or other) at the output and if you down-convert it to DSD, you just made a lossy version. DSD used in the backend (mic ADC) of the recording process and for making a digital copy of an analog master is where DSD has most value. Reality is that sigma delta conversion can sound great (the argument for DSD) but PCM can also utilize that.
Perhaps we should accept that silicone chips will never sound perfect, nor does vinyl or tape.
@@edmaster3147I don't think there is such a thing as any music that has not gone through silicone chips produced in the last 40+ years. This includes vinyl discs and tape produced entirely in analog process, where lots of op-amps are involved. I wouldn't worry about audio having gone through silicone chips as a measure of audio fidelity. Audio gets damaged by low quality or poorly engineered analog circuits, digital circuits or DSP code.
@@ThinkingBetter I think you are right. It's like what I often say, a painting will never be something else then another reality, even capable of causing an stunning experience, yet never be the real thing. It's the same with music, playback will never be reality yet can be 'the next-best thing'. And truth be told, I wouldn't like many musicians in my house in reality either.
@@edmaster3147 The notion that live music is superior in audio fidelity to studio produced music is almost 100% false. I’ve been to lots of major concerts from Pink Floyd, Dire Straits, Elton John etc. decades ago to John Mayor, Cold Play, Imagine Dragon etc. more recently and NEVER does live sound better than the same songs played on my own system. But a live experience is much more than about audio fidelity and remembering a live concert adds a lot to how you feel when listening to the same tracks at home. Music listening is all about emotions.
@@ThinkingBetter True. But it depends on the setting. Some smaller venues are really nice to be present and hear. At the big concerts its just way to loud and the earplugs are needed. And what to say of classical music in concert halls...
Because Sony and RIAA and no DAT for you mate. Home taping is killing music. All the digital encoding stuff is highly restricted apart from personal computers. We are living in a medieval society. When the Sony patents expire we will get MHz bitstream encoding in hardware and computers, but at the moment Sony made bad deals in the '80s and '90s, and the US government idiots flushed the future of high fidelity audio down the toilet.
A/D SAMPLING !
Love PS audio. When you look at all the limitations of DSD, i.e. not MOST user friendly, it just seems the ONLY good thing about it is higher music quality. Wish is wasn't so. Most of us just want to play high quality music without going through all these difficulties. I just don't see DSD really making it for the majority of us audiophiles. There reaches a point, where someone will only going to so far to reach "perfection". For me, DSD is just too limiting and difficult. I'll stick with CD, that's just me though. I really wish Octave Records would sell "regular" CD's at an affordable price as well as their other formats. What many companies don't seem to realize is most audiophiles don't want to go buy new equipment just to use a new audio format. Equipment is way too expensive these days to give it all up and start again. Believe me, ALL the people I know are not going to invest in DSD compatible equipmet just to play a very limited amount of DSD files.
DSD doesn't produce higher quality audio than high resolution PCM. It's a bunch of nonsense.
Mixing DSD isn't a problem technically. The mathematics are easy enough to work out. But there's no software and no equipment that supports working with DSD, because no-one ever uses it, because IT SUCKS! It has no advantages over PCM, at all. All it gives you is a higher quantisation noise.
But you can sell the shit to people buying cables for thousands of dollars
You can't mix in PCM too. Both DSD and PCM need to convert to binary in order for CPU to read and manipulate the data and convert back to PCM or DSD. No one doing DSD mix in binary because DSD and PCM in conversion gonna be the same and zero losses when you set the correct sampling rate relative to one another.
Thumbs up!
"Thumbs up" You are FIRST congratulations on this amazing performance you get 🥇🏆🍾🥂👏🇺🇸 You were so fast to click!
Huh? I can’t do math either so whatever Paul said sounds good to me , But why does it have to cost so dang much! Lmao
Of course it is possible to convert, but those conversions ruin what we are trying to protect.