Hi Roland,
Well, the digital information begins as an RF signal. There is another clock synced to the data rate coming off the CD, or more accurately, the speed of the disc is synced to a clock that is not the one that runs the DSP and DAC. So, there is noise and amplitude errors in the RF wave form that causes data errors to start with. Then there are timing errors associated with the rotation of the CD. Once the RF signal is processed, the wave is sampled (actually not quite that simple, but this will do for now). There is an intermediate signal called the EFM wave form that is actually processed to interpret what the digital information is. Once this is done, the information is clocked into a FIFO ram. The input is asynchronous and the output is clocked out at the DSP - DAC clock source. So this memory is always varying as to how full it is depending on how different the clock source for the CD motor (hence data rate) is to the DSP clock.
What happens in the DSP depends on the quality of the chip and processing. The tons of digital errors that occur in the reading process are put in a valid digital format that may or not represent the original information. Because the data points are repeated in different frames it can often be corrected, but often it cannot be corrected. This is important to recognize. The DSP may in those cases repeat data, mute or use interpolation to generate a close guess. Depends on the DSP chip. So much for bit perfect.
It is also critical to understand that the error correction used in data discs is not the same used in audio discs.
Once you have a valid data stream being clocked out of the DSP (valid in that the format is correct), you can run into issues with the clock quality, the PCB layout that causes timing errors, and cross talk of signals that may put odd glitches in to the clock or data paths. Then into the DAC where hopefully the high speed clock is laid out properly within the chip as it has the same issues the PCB has, or the quality of the voltage reference and noise. That depends on the decoding type. Understand that every method has issues and practical problems dependent on the layout of the actual chip.
Also, each DAC has an ENOB, or effective number of bits that is less than the advertised bit depth.
All of this impacts directly the quality of analogue signal you get out of it. I really wish that things were simple and worked perfectly, but that is not reality. I also wish that an improvement in clock quality would always improve the sound quality, but it doesn't. Its our old friend, the law of diminishing returns. Once a clock reaches some level of quality, improving the clock further does not improve the audio signal. That will vary depending on the entire system.
Second and third generation CD players tend to have the cleanest RF (eye) patterns. The newer ones have horrible, noisy signals that create extremely high data errors. That absolutely does create more unrecoverable digital errors. CD transports and heads have become as cheaply made as will function these days. They rely on DSP actions to provide "error correction and concealment". Note that that says concealment.
So this is the real truth about CD audio reproduction. If the audio signal is stored in a digital data format the errors are much better recovered, but errors can still occur.
-Chris
Well, the digital information begins as an RF signal. There is another clock synced to the data rate coming off the CD, or more accurately, the speed of the disc is synced to a clock that is not the one that runs the DSP and DAC. So, there is noise and amplitude errors in the RF wave form that causes data errors to start with. Then there are timing errors associated with the rotation of the CD. Once the RF signal is processed, the wave is sampled (actually not quite that simple, but this will do for now). There is an intermediate signal called the EFM wave form that is actually processed to interpret what the digital information is. Once this is done, the information is clocked into a FIFO ram. The input is asynchronous and the output is clocked out at the DSP - DAC clock source. So this memory is always varying as to how full it is depending on how different the clock source for the CD motor (hence data rate) is to the DSP clock.
What happens in the DSP depends on the quality of the chip and processing. The tons of digital errors that occur in the reading process are put in a valid digital format that may or not represent the original information. Because the data points are repeated in different frames it can often be corrected, but often it cannot be corrected. This is important to recognize. The DSP may in those cases repeat data, mute or use interpolation to generate a close guess. Depends on the DSP chip. So much for bit perfect.
It is also critical to understand that the error correction used in data discs is not the same used in audio discs.
Once you have a valid data stream being clocked out of the DSP (valid in that the format is correct), you can run into issues with the clock quality, the PCB layout that causes timing errors, and cross talk of signals that may put odd glitches in to the clock or data paths. Then into the DAC where hopefully the high speed clock is laid out properly within the chip as it has the same issues the PCB has, or the quality of the voltage reference and noise. That depends on the decoding type. Understand that every method has issues and practical problems dependent on the layout of the actual chip.
Also, each DAC has an ENOB, or effective number of bits that is less than the advertised bit depth.
All of this impacts directly the quality of analogue signal you get out of it. I really wish that things were simple and worked perfectly, but that is not reality. I also wish that an improvement in clock quality would always improve the sound quality, but it doesn't. Its our old friend, the law of diminishing returns. Once a clock reaches some level of quality, improving the clock further does not improve the audio signal. That will vary depending on the entire system.
Second and third generation CD players tend to have the cleanest RF (eye) patterns. The newer ones have horrible, noisy signals that create extremely high data errors. That absolutely does create more unrecoverable digital errors. CD transports and heads have become as cheaply made as will function these days. They rely on DSP actions to provide "error correction and concealment". Note that that says concealment.
So this is the real truth about CD audio reproduction. If the audio signal is stored in a digital data format the errors are much better recovered, but errors can still occur.
-Chris
I use an old laptop with a built-in CD drive. That and EAC (Exact Audio Copy, which is freeware) works fine to read CDs, normally without any errors at all. If there were to be any errors that the CD drive knows about (correctable and or uncorrectable) then they are enumerated by EAC for review by the user. Also, EAC can check online databases to see if the checksum for a known CD matches what everyone else gets from a presumably error-free read. That gives a 3rd check on read accuracy.
The bottom line for me is that ripping CDs to a PC is far superior to using a CD player in real time.
The bottom line for me is that ripping CDs to a PC is far superior to using a CD player in real time.
I just don't understand this continous talk, insistence about CD players.
Very clearly the typical use case for these clocks here - are custom dacs, typically DIY dacs, which already have an extremely superior system: many of these devices accept external master clock input, which is directly distributed to the dac chip/ chips. (my case, for example). Ian FIFO works like that. JLS boards work like that.
The only part missing/ needed is a high quality external clock driving those direct-in clock inputs.
Here a never mentioned fine detail is that there exist no commercial audio clock oscillators (AFAIK) with 50 ohm (100ohm) drive capability.. all XOs are low mA, Cmos output drivers, NOT designed to drive 50 ohm terminations. (without serious signal loss, and not clear consequencies for the oscillator circuit itself)
As a minimum, these little sh¦t pieces should be equipped a proper, low output impedance, high drive capability, low phase noise output buffer (Like it is done in the clocks of Andrea!!!)
Because of this problem, the typical onboard clock lines DO NOT have proper terminations, no signal integrity criterions are guaranteed...
Chris forgot to talk about these little particulars..
Very clearly the typical use case for these clocks here - are custom dacs, typically DIY dacs, which already have an extremely superior system: many of these devices accept external master clock input, which is directly distributed to the dac chip/ chips. (my case, for example). Ian FIFO works like that. JLS boards work like that.
The only part missing/ needed is a high quality external clock driving those direct-in clock inputs.
Here a never mentioned fine detail is that there exist no commercial audio clock oscillators (AFAIK) with 50 ohm (100ohm) drive capability.. all XOs are low mA, Cmos output drivers, NOT designed to drive 50 ohm terminations. (without serious signal loss, and not clear consequencies for the oscillator circuit itself)
As a minimum, these little sh¦t pieces should be equipped a proper, low output impedance, high drive capability, low phase noise output buffer (Like it is done in the clocks of Andrea!!!)
Because of this problem, the typical onboard clock lines DO NOT have proper terminations, no signal integrity criterions are guaranteed...
Chris forgot to talk about these little particulars..
Superior in terms of convenience.The bottom line for me is that ripping CDs to a PC is far superior to using a CD player in real time.
Not sure I agree with that. The old CD-80 I had would be up and running with a CD faster than I can find one on my server to play. But it was too big so had to go.
Here a never mentioned fine detail is that there exist no commercial audio clock oscillators (AFAIK) with 50 ohm (100ohm) drive capability.. all XOs are low mA, Cmos output drivers, NOT designed to drive 50 ohm terminations. (without serious signal loss, and not clear consequencies for the oscillator circuit itself)
- Why do you think all audio XO's are "Cmos output drivers, NOT designed to drive 50 ohm terminations"?
- What do you mean by "serious signal loss" above, and what would be the possible "consequences for the oscillator circuit itself" from a high load impedance, as intended, beyond the loading effects always described in all XO data sheets.
As a minimum, these little sh¦t pieces should be equipped a proper, low output impedance, high drive capability, low phase noise output buffer. Because of this problem, the typical onboard clock lines DO NOT have proper terminations, no signal integrity criterions are guaranteed...
- What audio properties are affected/improved by these "low output impedance, high drive capability, low phase noise output buffer" requirements?
- Why did the world wait for Andrea implementation to realize how poor the commercial audio "little sh¦t pieces" XO's are?
Faster..
Not an appropriate word for my audiophool vocabulary.
I have patience, and then I can enjoy it far better..
For this I still would have, my still working Rotel RCD 990. CDM 9 pro, PCM63, PMD100. It is strikingly good, still.
No way it gets close to my actual setup.
Not an appropriate word for my audiophool vocabulary.
I have patience, and then I can enjoy it far better..
For this I still would have, my still working Rotel RCD 990. CDM 9 pro, PCM63, PMD100. It is strikingly good, still.
No way it gets close to my actual setup.
Each to their own. Having to boot up a computer to just play a CD is sometimes inconvenient for me. For an evening of listening it's less of a problem. Sometimes its nice to be in a darkened room with no damned screen glowing at me. But that's just my preference.
I concur. the exact same preferences. no glowing screens- it's called a small display, turned sideways, quick action screensaver.. 🙂
Each to their own. Having to boot up a computer to just play a CD is sometimes inconvenient for me.
There are things called "streamers" intended to play, among others, server ripped CDs. And some software called DLNA running on the streamers, intended precisely for sorting and organizing media libraries, making playing a CD a two or three finger touches. A Raspberry running Volumio and an USB or I2S DAC is all that's needed. And it takes a few watt of powere, two orders of magnitude lower than a PC, so you can leave it on forever and then something.
Still needs a front end. All the phone interfaces suck majorly for my use case in the living room. In the kitchen I have a 9" tablet that works well though. But I accept my burden as an outlier.
Just curious, why do they suck? I find them quite decent, in particular for Volumio, where multiple options to choose exist in the Apple Store or Google Play.
Otherwise, Bluesound Node 2i HW/SW is there, but it's x3...x4 more expensive compared to an rPi solution.
Otherwise, Bluesound Node 2i HW/SW is there, but it's x3...x4 more expensive compared to an rPi solution.
Last edited:
Rip everything to SSD. Audirvana feeds 32 bit out and takes exclusive mode control of the DAC so all sample rate changes automatically follow the source file. Each song also has metadata to set absolute polarity. Proper polarity is more noticeable than the difference between some common clocks with time aligned speakers. The well tempered clock pulls the data at .5X the master clock over USB through an amanero in slave mode and then a Cronus re-clocks the three I2S signals to the master clock before feeding all including the master clock to the synchronous DAC. If you aren't running the DAC in synchronous mode and aren't using time aligned speakers that preserve the phase relationships of the music I would be far more inclined to wonder about the value I.E. audibility of better clocks.
Hi Andrea,
If you are lurking, just a reminder that there's a small group of users that may need your clocks for USB, network switches, PC motherboards, etc.:
Anyone need a 25MHz Crystal/Clock to theirs Buffalo switch, NUC`s or? At Andrea Mori`s Groupbuy at Diyaudio. - Networking, Networked Audio, and Streaming - Audiophile Style
The users over there may assume your clocks may no longer be available, so it would be good to touch base with that crowd as they are in the dark.
Some users here may gravitate over there if you need an outlet to discuss. Just start a thread. If I notice, I'll post a link.
If you are lurking, just a reminder that there's a small group of users that may need your clocks for USB, network switches, PC motherboards, etc.:
Anyone need a 25MHz Crystal/Clock to theirs Buffalo switch, NUC`s or? At Andrea Mori`s Groupbuy at Diyaudio. - Networking, Networked Audio, and Streaming - Audiophile Style
The users over there may assume your clocks may no longer be available, so it would be good to touch base with that crowd as they are in the dark.
Some users here may gravitate over there if you need an outlet to discuss. Just start a thread. If I notice, I'll post a link.
Last edited:
Oh, I forgot... I believe they also let a few users moderate their own thread.
I recommend you contact austinpop if you decide to create an account and thread. He can help walk you through on how to mod your own thread.
austinpop - Audiophile Style
I recommend you contact austinpop if you decide to create an account and thread. He can help walk you through on how to mod your own thread.
austinpop - Audiophile Style
Last edited:
- Why do you think all audio XO's are "Cmos output drivers, NOT designed to drive 50 ohm terminations"?
- What do you mean by "serious signal loss" above, and what would be the possible "consequences for the oscillator circuit itself" from a high load impedance, as intended, beyond the loading effects always described in all XO data sheets.
- What audio properties are affected/improved by these "low output impedance, high drive capability, low phase noise output buffer" requirements?
- Why did the world wait for Andrea implementation to realize how poor the commercial audio "little sh¦t pieces" XO's are?
I hear your answers loud and clear, thank you very much, QED.
Just curious, why do they suck?
Well starting with the caveats. I use mpd for playing music on the main system, I have no fruity products and my smart phone is 8 years old. I also have a significant digital music library.
All the clients I've seen are focused on playlists, which I don't do (another apparant oddity about me). I want to be able to quickly wade through the collection in the same way I'd scan the record racks for something to play. One day I might just give up, join the 21st century and go Roon, but not ready for that yet.
Oh and whatever I do needs to be usable by the better half and not be a temptation to the rugrats.
Second and third generation CD players tend to have the cleanest RF (eye) patterns. The newer ones have horrible, noisy signals that create extremely high data errors. That absolutely does create more unrecoverable digital errors. CD transports and heads have become as cheaply made as will function these days. They rely on DSP actions to provide "error correction and concealment". Note that that says concealment.
So this is the real truth about CD audio reproduction. If the audio signal is stored in a digital data format the errors are much better recovered, but errors can still occur.
-Chris
Hi Chris,
Thanks for the great sum up of how players work!
Unlike many I still use a player. I have 2000 titles in my collection and 1500 of them are sacds. I need a player.
I use one of the new Marantz players, KI-Ruby. It has a Marantz designed transport they call SACD-M3. Are you familiar with this? Marantz uses it in 4 players, SA-10, SA-12, SACDN30, and the Ruby.
The new and reliable transport was the main reason I bought the player in 2020 as my older Marantz SA-8260 has a far less reliable laser unit that is also no longer available.
Steve
PS. I know this thread is about clocks, but it has diverged a bit at the moment. So hope no issue with my post.
In the meantime Andrea and partner has created a little Wordpress website with some infos about their projects
THE WELL Audio – diy audio devices
It's still very simple, presenting all their designs.
In the download section is the windows app for the fifo …. Can be used to have a look or check for compatibility with all kind of DACs off line
TWSAFB-LT – THE WELL Audio
THE WELL Audio – diy audio devices
It's still very simple, presenting all their designs.
In the download section is the windows app for the fifo …. Can be used to have a look or check for compatibility with all kind of DACs off line
TWSAFB-LT – THE WELL Audio
Last edited:
- Status
- Not open for further replies.
- Home
- Source & Line
- Digital Line Level
- The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator