Products > Dodgy Technology
Frequncy standard use in audio equipment
dmills:
The best codec snafus happen in video however....
There was one of the early motion estimating Snell and Wilcox scalers that in development was notoriously prone to mistake peoples heads for footballs when the input was a game with a panned shot.
Also, grass that looked very, very, strange was a common one on compression codecs.
There is some evidence that close in phase noise on the modulator clocks can be audible, should not really be a surprise, sampling is basically mixing in the multiplication sense so noise sidebands should be expected.
Jitter was a problem, and it is rather telling that the early protools rigs sounded BETTER with an external clock (Internal XO should always beat a VCO), but that was 25 years ago, today this is something you can fuck up, but you would have to work to get it really wrong.
bdunham7:
--- Quote from: tkamiya on May 11, 2021, 07:31:21 pm ---I'm asking because I'm genuinely curious....
I have been seeing reference oscillators of all kinds, including Rb, OCXO, and GPSDO for audo use. Some report says they can "clearly hear the difference." I'm thinking for syncing digital recording, etc, regular crystal or maybe TCXO is sufficient and that human ear can't possibly distinguish the difference in phase and frequency stability.
Is this yet another case of snake oil? Some of those goes for $4KUS or so.
--- End quote ---
Just saw this topic, so here's my $.02
Jitter in the digital sample clock usually wouldn't show up as the 'wow and flutter' from the magnetic tape days (although you could perhaps simulate it) so you typically wouldn't hear a pitch or phase difference. Sample clock jitter simply translates to an amplitude error--in the file if the jitter is during recording, in the audio output if it is in the DAC clock. Uncorrelated (random) jitter shows up as noise and is less perceptible. Correlated (periodic) jitter shows up differently and depends on the audio signal, because that is what it is modulating.
There have been many arguments on this topic. The audiophool salespeople will claim that picoseconds of jitter are audible unless they are selling you products with 5 orders of magnitude more than that, then they claim it either isn't an issue or that it is the 'right kind' of jitter. I remember looking at this closely once and concluding that 5ns or less of uncorrelated jitter just couldn't be audible, even theoretically, and people who do experiments generally find much higher thresholds. Correlated jitter would be harder to make a definitive statement about, but I suspect that this is rarely an issue outside of devices that are broken or badly designed. Just keep in mind that having a nice clock doesn't necessarily eliminate jitter, it can be reintroduced by power supply noise, interference or other things that may vary the trigger threshold of the clock signal somewhere in the chain. This is one reason why 'master clocks' and synchronous data in an audio reproduction system are sort of idiotic--you want asynchronous data transfer to a buffer and then one clock right at the DAC.
http://archimago.blogspot.com/2018/08/demo-musings-lets-listen-to-some-jitter.html
Dither, which can be equated to uncorrelated jitter, can be introduced to mask distortions caused by truncation (reducing bit depth) and decimation, but can also be used to lower the apparent noise floor by pushing it up the spectrum. So on a typical CD, the jitter dither is not uniform, but is distributed so that most of the energy is in the upper octaves where it is less audible.
This stuff is all well known to competent audio engineers for many years now, but not all products take advantage of that knowledge, especially those on the very low and and, ironically, on the very high end. Of course on the high end all of their silly design mistakes can be compensated for with expensive exotic wood component stands and speaker cables starting at $10,000.
jonpaul:
Hello all et bonjour a SiliconWizard...
Been reading the thread with great interest.
We were working on digital audio since 1970s (Dolby, THX, Eventide)
Also on AES Standards commitee for digital audio transmission and digital microphones 1980s..2017. (AES-42, AES-13 I think...) So I wrote many AES papers, also SMPTE and NAB.
A few notes please;
master clocks are usually at studios that create the media. Nowadays a GPS 10 MHz is very accurate, low jitter and low cost eg the Leo Bodnar.
Most consumer devices are playback, not media creation, so clock and data jitter is much more important that any clock frequency.
Digital audio transmission (AES/EBU, SP/DIF, AES3id) is on a single serial stream, Manchester encoded with embedded clock. A PLL is ued for clock recovery.
Thus the DAC for reconstruction depends on a recovered clock, from the PLL.
The noise to jitter issue was well researched bak in 1980s, see especially our old friends Dr Steve HARRIS (Crystal/Cirrus) classis papers (1987?) as well as the fine work of Dr Julian DUNN (RIP) (Cambridge, AP).
Most of the jitter is due to CM HF asynchronous noise contamination of the clock and PLL.
We designed special TX and RX ref circuits to reduce the jitter issues, by greatly increasing CMRR and reducing primary to secondary capacitance in the required transformers.
The fine results were confirmed by both lab and listening tests.
Beware that any fair listening needs a double blind like the 1970s..1980s ABX relays with random hidden selection.
The Audiophools will object, but the ABX was a wonderful Occams Razor.
The ABX are long gone but easy to build with a flip-flop, oscillator and gold contact relay.
Hope this sparks some interesting response !
Jon
bdunham7:
--- Quote from: jonpaul on June 03, 2021, 06:55:49 am ---Most consumer devices are playback, not media creation, so clock and data jitter is much more important that any clock frequency.
--- End quote ---
You'd be surprised how many consumer devices have ADC inputs. All of the signal processing, including tone and volume, is often done digitally in a 32-bit processing environment, so any analog line-in, microphone or phono inputs may be converted to PCM.
--- Quote ---Digital audio transmission (AES/EBU, SP/DIF, AES3id) is on a single serial stream, Manchester encoded with embedded clock. A PLL is ued for clock recovery.
Thus the DAC for reconstruction depends on a recovered clock, from the PLL.
The noise to jitter issue was well researched bak in 1980s, see especially our old friends Dr Steve HARRIS (Crystal/Cirrus) classis papers (1987?) as well as the fine work of Dr Julian DUNN (RIP) (Cambridge, AP).
Most of the jitter is due to CM HF asynchronous noise contamination of the clock and PLL.
--- End quote ---
Yes, those synchronous formats use the embedded clock and top-quality systems nowadays use a dual-PLL to recover it--and that is certainly more than good enough provided the signal is good. Synchronous is needed for video, but playback from music servers nowadays--both local and remote--often uses asynchronous transfer to a buffer and the DAC uses a local clock. It either works or it doesn't--there's no possibility of degrading the clock with this setup. On my system, when I stream a radio station, play a FLAC or MP3 file or even play a CD (depending on how I do it) the streams are all buffered and clocked locally--no embedded clock and no PLL.
MazeFrame:
--- Quote from: tkamiya on May 11, 2021, 07:31:21 pm ---I'm asking because I'm genuinely curious....
I have been seeing reference oscillators of all kinds, including Rb, OCXO, and GPSDO for audo use.
--- End quote ---
World Clocks (as used in studios with lots of conversions and digital to digital interconnects) run at 10MHz via 75 Ohm BNC cables.
There are also "re-clockers" that help the less timing stable formats (TOSLINK) keep pace, mostly because some manufacturers cheap out on the inputs.
This may make sense for PA installations on concerts or bigger recording studio setups (= your job depends on the result sounding good). At home, no point to it.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version