Jitter and the ASRC
Hello Guys
Much is said about the ASRC and it ability to remove jitter , but the guys who popularised this idea where Benchmark with the DAC1 and they called it Ultralock. From a jitter improvement perpective the way it works is that it reclocks the incoming data to an onboard clock which drives the ASRC. This clock ( DAC1) is much better than the incoming clock data which has much more jitter, which is the reason for the sound quality of the DAC1.
The jitter performance of a DAC using an ASRC is a function of the quality of the onboard clock and not the ASRC. The better this clock the better the jitter performance. You cannot have a lousy clock on a ASRC and expect it to give good jitter performance.
Put another way the clock in a DAC is an analog signal that is used to set up the correct timing for digital data to be played back , if is bad you can hear it when compared to good clock with low jitter .
With regards to ASRC their affect on sound is due to in part their numerical precision but parts are available today which give you full 24bit precision and dynamic range after processing the data
Regards
Arthur
Hello Guys
Much is said about the ASRC and it ability to remove jitter , but the guys who popularised this idea where Benchmark with the DAC1 and they called it Ultralock. From a jitter improvement perpective the way it works is that it reclocks the incoming data to an onboard clock which drives the ASRC. This clock ( DAC1) is much better than the incoming clock data which has much more jitter, which is the reason for the sound quality of the DAC1.
The jitter performance of a DAC using an ASRC is a function of the quality of the onboard clock and not the ASRC. The better this clock the better the jitter performance. You cannot have a lousy clock on a ASRC and expect it to give good jitter performance.
Put another way the clock in a DAC is an analog signal that is used to set up the correct timing for digital data to be played back , if is bad you can hear it when compared to good clock with low jitter .
With regards to ASRC their affect on sound is due to in part their numerical precision but parts are available today which give you full 24bit precision and dynamic range after processing the data
Regards
Arthur
audio_tony said:
No. Not by strict definition. Resampling is just a way of reorganising the data, like when you change the pixel resolution of a picture in photoshop.
Data-rate converters resample on-the-fly. They are clocked devices. The jitter at the output may be better than the jitter at the input, because the quality of the output clock may be better than that at the input. Where resampling means upsampling then the clock period is reduced, which in one sense means that there is less 'room' to accomodate jitter. For this reason 'resampling' has become a portmanteau word carrying the implication of jitter-improvement.
theAnonymous1 said:
I can't tell you, but: -
In all probability, yes, with some margin, dependent on a reasonable implementation and how used. This is a respectable, recent chipset whose specification probably meets or exceeds that of the equipment on which many of your favourite recordings were made. A reasonable implementation is likely, because you have to have your head screwed on to a certain extent to glue these bits together at all.
AES Preprint 4826 discusses jitter on page 23 (I am informed). It says that the threshold of audibility with a 20 kHz tone is 10 nanoseconds rms, and at 4 kHz the threshold of audiblity is 100 nanoseconds rms.
The jitter in your system should be of the order of 100 picoseconds rms worst case.
'Normal' people have to listen closely to tell the difference between a good MP3 and an original CD and only reliably differentiate with the benefit of side-by-side comparison. Your ears and brain cooperate so successfully to extract the information from the distortion and noise that your long-term memory is incapable of providing a sufficient reference. No-one would suggest that if you keep increasing jitter that you would not eventually hear it, but minute or even grosser temporal displacements in a 24-bit sample stream at more than 4* the Nyquist rate for human HF audibility is not the place to look, when you can throw away more than half the data on a 16-bit 44k1 CD and still barely hear the difference listening on the best in-ear phones. Is it an MP3 they're playing on the muzak or uncompressed linear PCM?
It's not really a question of how GOOD is this stuff now, it's more a question of how CHEAP we can make it, AND getting good reliable data on what is, and what is not audible. Unfortunately the unavailability of such widely-promulgated data permits the unscrupulous to freely exploit the uncertainties and hence the gullible section of the public, although it's availability might not ultimately eliminate such practices, given the absurd claims made (and swallowed) regarding cables. Some parties have a vested interest in muddying the water. That this continues IS however a reflection of how little a 'normal' consumer cares.
Ultimately it is the capacity of a performance to shine through much-less-than-perfect reproduction that renders these concerns trivial and even counter-productive where the enjoyment of recorded or relayed music is concerned, particularly given that the general standard has become so high. If you have satellite TV or an average computer and internet access, and people all over the world do, you have access to superb sound.
w
Unfortunately the good an valuable reference articles of the AES and IEEE are not available for the general public.
Found this nice article that sheds some light on it: http://www.iet.ntnu.no/courses/fe8114/slides/upsanddownsofasrc.pdf
Found this nice article that sheds some light on it: http://www.iet.ntnu.no/courses/fe8114/slides/upsanddownsofasrc.pdf
I think of jitter as a limiting/optimization problem. I don't want jitter to be the weakest link in my playback chain, so I try to make sure that it is not. In the digital domain, there is the concept of "effective number of bits" or ENOB. ENOB is the resolution of your PCM playback stream, as limited by external factors (jitter, slew rate, etc.) When listening to a 24-bit 192kHz stream you need < 200fs jitter (integrated from 10 to 192) to be certain that jitter effects are below the least significant bit.
Once you achieve that, then you can just stop worrying about jitter. The harmonic distortion in your power amplifier is probably thousands of times worse.
Once you achieve that, then you can just stop worrying about jitter. The harmonic distortion in your power amplifier is probably thousands of times worse.
wakibaki said:
Don't be so sure about that.
Thanks for the explanation though. I'm fairly confident now that I don't need to be so concerned about this "jitter bug".
Achieving sample jitter below 100fs is not a problem. You can get this with off-the-shelf LV-PECL XOs and the AD9516 series of clock distribution chips.
jwb said:
Hello jwb-
You are correct in the data you have mentioned here, but I find that devices such as the AD9516 clock generator and the like are made to run in the 2GHz and greater clock range. From what the datasheets I have read on the Analog Devices website say regarding their low-jitter clock solutions, they all seem to be dedicated to devices that serve sample clocks to 1GS/second A-to-D converters, SONET and ATM wide-area-network transcevers, and wireless transceivers for data links like WiFi. Jitter specs in the femtosecond range are absolutely needed for these kind of devices. It also seems that most ultra-low jitter XO's also seem to be focused in the RF or GHz range as well.
As of late, I personally have not found any easy to work with components like these you have quoted. The AD9516 comes in only the 64-lead LFCSP packaging, which there may be a privileged few here possessing the skills and tools to easily implement into their personal projects. I am not among those folks! 🙂
Might you have some good suggestions for sources manufacturing fs jitter level XO's in the oscillator frequency range common to audio uses? I seem to have no luck coming across any in my searching up to this point!
Cheers-
I found myself quite liberated when I finally realized that there was no reason to use 24.576MHz or 11.2896MHz clocks in audio systems where an ASRC is employed. Now I just use 25MHz clocks everywhere, and all the major XO manufacturers produce extremely low jitter 25MHz clocks as a matter of course.
It does mean that my DACs run at 97.65625kHz sample rate instead of 96kHz, but who cares?
It does mean that my DACs run at 97.65625kHz sample rate instead of 96kHz, but who cares?
Hi jwb!
This sounds like a very good idea indeed, looking from the usual standpoint taken by most electrical engineers. Having only one clock should equal a lot less of a chance that circuit board signal routing and other sticky points of circuit design will cause one clock to "modulate" on to the other one due to harmonics and such. Great!
Oh no, that must mean that there's someone with great hearing and absolutely perfect pitch living in your neighborhood that's crying their little eyes out every time you play your hi-fi system! They have to hear those "out-of-tune" music pieces again, and the musicians are even playing them at too fast of a tempo! LOL!! 🙂
Seriously though, I doubt most people would even notice a steady 2% overspeed in the sample clock, right? I have used a well-written "time compression/expansion" algorithm in a popular digital audio editing system to speed up a 5 min. song about 10 seconds or so before, and not even the composer of that song noticed the change until I made him aware of it. The algorithm used of course does a software "re-sampling" of the audio that preserves its pitch while changing the duration of the editing selection. Pretty nifty if you asked me....
Cheers!
This sounds like a very good idea indeed, looking from the usual standpoint taken by most electrical engineers. Having only one clock should equal a lot less of a chance that circuit board signal routing and other sticky points of circuit design will cause one clock to "modulate" on to the other one due to harmonics and such. Great!
Oh no, that must mean that there's someone with great hearing and absolutely perfect pitch living in your neighborhood that's crying their little eyes out every time you play your hi-fi system! They have to hear those "out-of-tune" music pieces again, and the musicians are even playing them at too fast of a tempo! LOL!! 🙂
Seriously though, I doubt most people would even notice a steady 2% overspeed in the sample clock, right? I have used a well-written "time compression/expansion" algorithm in a popular digital audio editing system to speed up a 5 min. song about 10 seconds or so before, and not even the composer of that song noticed the change until I made him aware of it. The algorithm used of course does a software "re-sampling" of the audio that preserves its pitch while changing the duration of the editing selection. Pretty nifty if you asked me....
Cheers!
Has anyone seen Jan Didden's recent posts on ASRC jitter tests?
http://www.diyaudio.com/forums/showthread.php?postid=1718326#post1718326
http://www.diyaudio.com/forums/showthread.php?postid=1718326#post1718326
An externally hosted image should be here but it was not working when we last tested it.
his test would belong into the 3rd category i guess
jwb said:
I would be interested to know what you are using for low jitter 25Mhz oscillators. I have been researching some including using SAW stabilized higher frequency oscillators and dividing down. The issue is that the jitter is generally only specified from 10-20KHz and up. Even with sigma-delta convertors, I am not convinced that sub 10Khz phase-noise will not make itself visible in the audio band. If there is anyone seriously technical out there that knows the answer, that would be appreciated.
You're absolutely right that close-in jitter will modulate your audio outputs. Some manufacturers will specify phase noise at 10Hz, 100Hz, 1K, 10K, 100K offsets. You can get a total jitter number from these by performing a trapezoidal approximate integration.
If you're at a loss for manufacturers, you could start with Valpey Fisher.
If you're at a loss for manufacturers, you could start with Valpey Fisher.
- Status
- Not open for further replies.