Pardon this intrusion from rec.audio.pro, but some said I might find an answer here. Here's what I understand: Fact: DSD has a sampling rate of 1 bit * 64 * 44100 per second. Assumption: Each sample raises or lowers the volume one "bit", one unit. Conclusion: A 22.05 kHz tone gets 128 samples and has a peak to peak height of -32 to +31 to -32 again (6 bits) and a 44.1 kHz tone only -16 to +15. Here's my question: How can DSD maintain a flat frequency response to almost 50 kHz at full volume? How does it keep the HF from fading to nothing?
DSD frequency response and sampling rate
Started by ●January 25, 2006
Reply by ●January 25, 20062006-01-25
"Carey Carlan" <gulfjoe@hotmail.com> wrote in message news:Xns97566DC2ADD40ccarlan@140.99.99.130...> Pardon this intrusion from rec.audio.pro, but some said I might find an > answer here. > > Here's what I understand: > > Fact: DSD has a sampling rate of 1 bit * 64 * 44100 per second. > Assumption: Each sample raises or lowers the volume one "bit", one unit. > Conclusion: A 22.05 kHz tone gets 128 samples and has a peak to peak > height of -32 to +31 to -32 again (6 bits) and a 44.1 kHz tone only -16 to > +15. > > Here's my question: > > How can DSD maintain a flat frequency response to almost 50 kHz at full > volume? How does it keep the HF from fading to nothing?Consider this: There are 64 1-bit changes 64 times for each sample at 44100, right? So, the output can change 64 bits one time at that sample rate (44100) or, it can change 32 bits twice at that sample rate, etc. Since 44100 is the typical sample rate for audio and maybe at 8 bits or 16 bits, then that's a lot and implies a sample rate that's much higher. ... as much as equivalent to (64/8)*44100. I don't know how the DSD is implemented at all but that's some technical background to lead you in a direction. fred
Reply by ●January 25, 20062006-01-25
Carey Carlan wrote:> Pardon this intrusion from rec.audio.pro, but some said I might find an > answer here. > > Here's what I understand: > > Fact: DSD has a sampling rate of 1 bit * 64 * 44100 per second. > Assumption: Each sample raises or lowers the volume one "bit", one unit. > Conclusion: A 22.05 kHz tone gets 128 samples and has a peak to peak > height of -32 to +31 to -32 again (6 bits) and a 44.1 kHz tone only -16 to > +15. > > Here's my question: > > How can DSD maintain a flat frequency response to almost 50 kHz at full > volume? How does it keep the HF from fading to nothing?What do you believe the frequency spec to be? Refer to http://www.dsdproaudio.com/html/dsd_sacd_explained.html. The highest frequency of interest is in the order of 20 KHz, same as CD. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●January 25, 20062006-01-25
Hi Carey, I'll respond to your statements below. Carey Carlan <gulfjoe@hotmail.com> writes:> Pardon this intrusion from rec.audio.pro, but some said I might find an > answer here. > > Here's what I understand: > > Fact: DSD has a sampling rate of 1 bit * 64 * 44100 per second.I don't mean to be overly pedantic, but bits have no place in sample rate. The sample rate is 64 * 44100 = 2,822,400 samples/second. The sample *size* is 1 bit. The aggregate bit rate is Fs * N, or 2,822,400 bits/second.> Assumption: Each sample raises or lowers the volume one "bit", one unit. > Conclusion: A 22.05 kHz tone gets 128 samples and has a peak to peak > height of -32 to +31 to -32 again (6 bits) and a 44.1 kHz tone only -16 to > +15.This analysis is not correct.> Here's my question: > > How can DSD maintain a flat frequency response to almost 50 kHz at full > volume? How does it keep the HF from fading to nothing?The fact is, DSD is flat to 1/2 * (64*44100) = 1.411200 MHz. The problem is, the (quantization) noise gets really loud up at that frequency. But the signal (as opposed to noise) frequency response itself is flat. Contrast this with a linear PCM system in which the signal bandwidth is also 1/2 of the sample rate, but the quantization noise power is constant throughout that bandwidth. That's the main difference between DSD and linear PCM - that the noise is "shaped" so that there is a different amount of noise in different parts of the spectrum. The reason for this has to do with the way DSD operates, and you'd need to understand what a delta sigma modulator is in order to fully explain. If you're interested, you can find some info on Wikipedia on it. We'd be happy to answer more detailed questions here if you desire. -- % Randy Yates % "Watching all the days go by... %% Fuquay-Varina, NC % Who are you and who am I?" %%% 919-577-9882 % 'Mission (A World Record)', %%%% <yates@ieee.org> % *A New World Record*, ELO http://home.earthlink.net/~yatescr
Reply by ●January 25, 20062006-01-25
Jerry Avins <jya@ieee.org> wrote in news:v_KdnQy5BrmgOErenZ2dnUVZ_t6dnZ2d@rcn.net:>> Assumption: Each sample raises or lowers the volume one "bit", one unit. >> >> How can DSD maintain a flat frequency response to almost 50 kHz at >> full volume? How does it keep the HF from fading to nothing? > > What do you believe the frequency spec to be?Here's an article about a player -6dBFS at 50 kHZ and -9 dBFS at 75 kHz. http://stereophile.com/hirezplayers/340/index7.html (from 2001)> Refer to > http://www.dsdproaudio.com/html/dsd_sacd_explained.html. The highest > frequency of interest is in the order of 20 KHz, same as CD.I've seen this description before and I don't comprehend it. See continued discussion in response to Randy Yate's post.
Reply by ●January 25, 20062006-01-25
Randy Yates <yates@ieee.org> wrote in news:k6cofda8.fsf@ieee.org:> Hi Carey, > > I'll respond to your statements below. > > Carey Carlan <gulfjoe@hotmail.com> writes: > >> Pardon this intrusion from rec.audio.pro, but some said I might find >> an answer here. >> >> Here's what I understand: >> >> Fact: DSD has a sampling rate of 1 bit * 64 * 44100 per second. > > I don't mean to be overly pedantic, but bits have no place in sample > rate. The sample rate is 64 * 44100 = 2,822,400 samples/second. The > sample *size* is 1 bit. The aggregate bit rate is Fs * N, or 2,822,400 > bits/second.Agreed, I should have said "aggregate bit rate".>> Assumption: Each sample raises or lowers the volume one "bit", one >> unit. Conclusion: A 22.05 kHz tone gets 128 samples and has a peak >> to peak height of -32 to +31 to -32 again (6 bits) and a 44.1 kHz >> tone only -16 to +15. > > This analysis is not correct.Thank you. I know it's invalid, I just don't know where.>> Here's my question: >> >> How can DSD maintain a flat frequency response to almost 50 kHz at >> full volume? How does it keep the HF from fading to nothing? > > The fact is, DSD is flat to 1/2 * (64*44100) = 1.411200 MHz. The > problem is, the (quantization) noise gets really loud up at that > frequency. But the signal (as opposed to noise) frequency response > itself is flat. > > Contrast this with a linear PCM system in which the signal bandwidth > is also 1/2 of the sample rate, but the quantization noise power is > constant throughout that bandwidth. That's the main difference between > DSD and linear PCM - that the noise is "shaped" so that there is a > different amount of noise in different parts of the spectrum. > > The reason for this has to do with the way DSD operates, and you'd > need to understand what a delta sigma modulator is in order to fully > explain. If you're interested, you can find some info on Wikipedia on > it. We'd be happy to answer more detailed questions here if you > desire.I read the Wikipedia article before coming here. Either I misunderstand or it's not complete enough for this non-intuitive thinker. The specific question is "On playback, how much change occurs in the signal for each of a series of '1' values.?" If I can get an understandable answer to that, I think I can make it the rest of the way or else fuel another question.
Reply by ●January 25, 20062006-01-25
Randy Yates <yates@ieee.org> wrote in news:k6cofda8.fsf@ieee.org:> Hi Carey, > > I'll respond to your statements below. > > Carey Carlan <gulfjoe@hotmail.com> writes: > >> Pardon this intrusion from rec.audio.pro, but some said I might find >> an answer here. >> >> Here's what I understand: >> >> Fact: DSD has a sampling rate of 1 bit * 64 * 44100 per second. > > I don't mean to be overly pedantic, but bits have no place in sample > rate. The sample rate is 64 * 44100 = 2,822,400 samples/second. The > sample *size* is 1 bit. The aggregate bit rate is Fs * N, or 2,822,400 > bits/second.Agreed, I should have said "aggregate bit rate".>> Assumption: Each sample raises or lowers the volume one "bit", one >> unit. Conclusion: A 22.05 kHz tone gets 128 samples and has a peak >> to peak height of -32 to +31 to -32 again (6 bits) and a 44.1 kHz >> tone only -16 to +15. > > This analysis is not correct.Thank you. I know it's invalid, I just don't know where.>> Here's my question: >> >> How can DSD maintain a flat frequency response to almost 50 kHz at >> full volume? How does it keep the HF from fading to nothing? > > The fact is, DSD is flat to 1/2 * (64*44100) = 1.411200 MHz. The > problem is, the (quantization) noise gets really loud up at that > frequency. But the signal (as opposed to noise) frequency response > itself is flat. > > Contrast this with a linear PCM system in which the signal bandwidth > is also 1/2 of the sample rate, but the quantization noise power is > constant throughout that bandwidth. That's the main difference between > DSD and linear PCM - that the noise is "shaped" so that there is a > different amount of noise in different parts of the spectrum. > > The reason for this has to do with the way DSD operates, and you'd > need to understand what a delta sigma modulator is in order to fully > explain. If you're interested, you can find some info on Wikipedia on > it. We'd be happy to answer more detailed questions here if you > desire.I read the Wikipedia article before coming here. Either I misunderstand or it's not complete enough for this non-intuitive thinker. The specific question is "On playback, how much change occurs in the signal for each of a series of '1' values.?" If I can get an understandable answer to that, I think I can make it the rest of the way or else fuel another question.
Reply by ●January 25, 20062006-01-25
Carey Carlan wrote:> Jerry Avins <jya@ieee.org> wrote in > news:v_KdnQy5BrmgOErenZ2dnUVZ_t6dnZ2d@rcn.net: > > >>>Assumption: Each sample raises or lowers the volume one "bit", one unit. >>> >>>How can DSD maintain a flat frequency response to almost 50 kHz at >>>full volume? How does it keep the HF from fading to nothing? >> >>What do you believe the frequency spec to be? > > > Here's an article about a player -6dBFS at 50 kHZ and -9 dBFS at 75 kHz. > http://stereophile.com/hirezplayers/340/index7.html (from 2001) > > >>Refer to >>http://www.dsdproaudio.com/html/dsd_sacd_explained.html. The highest >>frequency of interest is in the order of 20 KHz, same as CD. > > > I've seen this description before and I don't comprehend it. > See continued discussion in response to Randy Yate's post.I don't see your response yet, but Randy must be writing about the characteristics of delta-sigma converters, which can be explained fairly well with hand waving and analogies, but which need fairly detailed math for accurate understanding. You probably bring concepts from delta modulation to your understanding. They have some bearing, but there's much more to the picture. I don't know your level of technical knowledge, So instead of recommending material on them, I'll let you google for them on your own. If you have specific questions, Randy and others will be better at answering them than I. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●January 25, 20062006-01-25
Here are a couple of links I found I'm sorting through as well... analog devices decoder http://www.analog.com/UploadedFiles/Data_Sheets/422968845AD1955_0.pdf libraery of congress http://www.digitalpreservation.gov/stage/formats/fdd/fdd000230.shtml tech white paper http://www.superaudio-cd.com/technology_explained/detailed_information/whitepaper.pdf Mark
Reply by ●January 25, 20062006-01-25
Carey Carlan <gulfjoe@hotmail.com> writes:> [...] > Randy Yates <yates@ieee.org> wrote in news:k6cofda8.fsf@ieee.org: >> The reason for this has to do with the way DSD operates, and you'd >> need to understand what a delta sigma modulator is in order to fully >> explain. If you're interested, you can find some info on Wikipedia on >> it. We'd be happy to answer more detailed questions here if you >> desire. > > I read the Wikipedia article before coming here. Either I > misunderstand or it's not complete enough for this non-intuitive > thinker.Did you mean the Wikipedia article on delta sigma modulation or the Wikipedia article on DSD? I meant the former, http://en.wikipedia.org/wiki/Sigma_delta and it is quite detailed, so I'm thinking you meant the latter, which is very brief.> The specific question is "On playback, how much change occurs in the > signal for each of a series of '1' values.?"On playback, the 1-bit DSD signal is simply applied to a lowpass filter, so the precise answer to your question depends on the impulse response of that lowpass filter and the initial conditions of the system. The output is the convolution of the lowpass filter inpulse response and the sequence of 1's and 0's from the DSD bitstream.> If I can get an understandable answer to that, I think I can make it the > rest of the way or else fuel another question.Carey, why are you asking? What is the higher-level information you want to know? -- % Randy Yates % "With time with what you've learned, %% Fuquay-Varina, NC % they'll kiss the ground you walk %%% 919-577-9882 % upon." %%%% <yates@ieee.org> % '21st Century Man', *Time*, ELO http://home.earthlink.net/~yatescr
