Support Head-Fi.org by
starting all of your
Amazon.com shopping by
clicking here.
____________________________________________________________________
Today's Featured Head-Fi Blog: Jude's Blog
____________________________________________________________________
Please help
support Head-Fi by becoming a Contributing Member
CLICK
HERE -- Contributing Members, thank you
for your generous support! --
Jitter is an in depth conversation that could barry the length of this tread by a factor of 2.
I totally agree with Thomas on the pit jitter issue. Many transport companies have gone to ATAPI interface instead of SPDIF off the raw transports. This way they can read the data off the CD the same way a computer can. But I have yet to see anyone's transport re-read on an error. Still many companies still rely on the poor SPDIF coming off the raw transport as defacto standard for the data they use. Some are goo enough to reclock this but in many cases this is the cause for poor jitter on the transport end.
But in declaring low jitter, the power supply has more to do with jitter than almost anything else. Jitter is really dominated by really low frequency noise in the area below 10Hz. Most semiconductor companies that specialize in regulator technology don't even spec below 10Hz because the noise stats sky rocket. But these are exactly the frequencies that effect clocks and audio is full of clocks.
If we look at the basic preface of digital audio there is in each piece what is called "intrinsic jitter". This is the jitter that a part has no matter how the out side world effects it. This intrinsic jitter can then be amplified by the power supply. You can think of this increase in jitter like a carrier wave. The power supplies noise will basically modulate this jitter to a higher level.
Let's look at a standard oscillator used for say running an ARSC. The intrinsic jitter for these range in the sub 1ps area to as much as 150ps. Heck if you are making a microwave and need an oscillator 150ps is no big deal. But in audio we need to look to the sub 1ps units. If you talk to the designers of these clocks and ask about power supplies to drive them most of them will talk about elaborate discrete supplies. Guido Tent has some good information up on TentLabs.com. Guido work at Phillips for years in their R&D for noise reduction.
Standard linear regulators have improved in the noise area but they are still very high in the uV area. Most of them below 100uV but this in it self can cause the intrinsic jitter to amplify by 20-30dB.
The oscillator companies that spec sub 1ps jitter are using supplies lower than 10nV (below 1Hz) of noise to justify their measurements.
~~~~~~~ ARSC's
It is true that ARSC's can remove allot of the jitter coming in by their simply reclocking at an async rate. But most of the testing on this shows that ARSC's act like low pass filters to the incoming jitter.
So we have to look at the jitter equation for ARSC's kind of like this:
Jitter Out = (Jitter in/LP filter) + intrinsic jitter * power supply noise + intrinisc jitter oscillator * power supply noise
This kind of basic but shows that there is intrinsic jitter in the ARSC as well. Becuase there is tons of clocks all over the place here in both the filtering area as well as the input and output clocking areas. This jitter added to both the oscillator jitter and the attenuated jitter input is going to be applied to the next device in the chain. That being mostly the dac.
Dac's also have intrinsic jitter... in the same vein.
As Crowbar mentioned the new ESS part is getting allot of press. Dustin did a killer job on this. I have his earlier ARSC which was very good. The design uses allot of isolating techniques to assure the ARSC section does not invade the dac section and add jitter. Dustin also was awarded a patent for thier jitter reduction which is basically as Crowbar mentioned an ARSC with very low cutoff point.
Well these are just some things to think about. Sorry if it's too technical, too much coffee on my part.
Thanks
Gordon
__________________
J. Gordon Rankin
Wavelength Audio, ltd.
Why use a fixed clock for the ARCS? Most people who have tested ARSC's feel they sound better when used on a fixed multiple like 2, 4... instead of fractional.
So say for 44.1 upsample to 88.2 or 176.4 and so forth.
In several of the recording studios I work with here (Sound Images, Ashley Shephard's Audio Grotto) feel that when recording that the use of multiples are the key to better results when down sampling. Most of the recordings we have done recently are done at 24/88.2 or 24/176.4 so the output to 16/44.1 sounds better.
On the dCS gear many have talked about there ability to change the upsampling on the fly and many feel that multiple over fractionals always sound better.
Your thoughts?
Thanks
Gordon
Gordon,
The fact that the ASRC is asynchronous means that the original sample-rate is completely irrelevent to the SRC process. The data simply registers into a buffer waits there until it is ready to be processed.
We use Audio Precision (AP) hardware and software, which is the leading digital and analog audio testing platform. However, our AP interface does not have USB input, so it can only test Kmixer's bit-transparency by proxy via a USB audio device.
Elias,
I use the Prism dScope III and it has native output and input. It shows the KMIXER fluctuating error of at least one bit.
But this subject should be dead for two reason's:
1) Everyone who has every bypassed the KMIXER has heard better results. So there must be something changed in the KMIXER.
2) Since you can bypass it, then what's the big deal.
Well 3 reason's... If Microsoft says so why even try and prove them wrong. They must know a little more about it than you do since they wrote it.
Thanks
Gordon
__________________
J. Gordon Rankin
Wavelength Audio, ltd.
Hi everyone, first post from a long time lurker and audiophile.
I'm looking to buy a DAC1 PRE but unfortunately the distributor in Germany and/or Benchmark must have forgotten to do the conversion from USD to EUR. The PRE goes for EUR 1600 around here, which means a hefty $1000 premium on what it sells for in the US, surely something must be wrong here. For $1000 I could easily fly to the US and pick one up myself!
Please advise as to where I can get one at reasonable cost. Thanks!
03lab,
We don't control pricing for foriegn dealers, and that has good and bad effects. The bad effect is that some dealers charge too much. The good effect is that you can shop around to different dealers to find a better price.
Let them know that, if they want you as a customer, they've got to provide competitive pricing.
Thanks,
Elias
ps. You can begin your search for dealers/distributors here:
Quick question:
I'm using a DAC1 USB. MacMini USB --> DAC1 Balanced --> BAT VK-300xSE --> Totem Hawk speakers
I'm using the balanced outs from the DAC1, into the BAT. I have it set for "calibrated" outs and I have not touched the -20db jumper inside the DAC1 or the calibration pots yet.
I seem to have pretty good range on the volume, with normal listening position around 90 (scale goes to 140 on the BAT).
Am I going to benefit SQ by dropping the to 0db attenuation on the DAC1? The BAT has an amazingly low noise floor, so I have no noise at all at these levels.
Thanks
SpyderX,
From your description, it seems your current configuration is setup optimally. The only advantage you would gain by dropping the attenuation of the DAC1 would be a lower noise floor, but that does not seem to be a problem in your case.
I have a question about the ASRC function and the "jitter-immune" claims of the DAC1 products please.
As you said, the ASRC function of the DAC1 utilizes its own internal clock (instead of the recovered S/PDIF clock) when internally feeding samples to the DA chip. But before this happens, surely the recovered clock is being utilized by the ASRC algorithm, yes? [Otherwise the ordering of the samples as sent, would be unknown, when trying to reconstruct and zip them into the new sampling rate.]
G-U-E-S-T,
The ASRC averages the incoming clock to manage the data buffer and determine the SRC ratio. This topology allows this particular chip to achieve jitter attenuation with a corner frequency of 0.1 Hz with a very steep roll-off - effectively attenuating all jitter to levels well below the threshold of hearing (greater then -130 dB).
You have to be careful with this. The original poster spoke about jitter in the recording chain. That is very different from "pit" jitter on a CD.
The jitter that is permanently added to the signal is created by the analog to digital converter and comes from the fact that the samples are not actually takes at equi distant intervalls but at times in the jitter intervall. This jitter can not be removed.
RIGHT!
More specifically, the jitter that is present during A-to-D does not get carried on as jitter...it becomes audible distortion. At that point, it is a part of the audio...it is no longer in the clock.
Of course the ratio estimator, being an implicit ADC, suffers from quantisation. The phase between the two clocks is quantised to a time span equal to 1 period of the highest frequency clock in the chip....This error is added to the input jitter before being attenuated by the lowpass filter. Whether this effect is detectable at all depends on the spectral distribution of the quantisation error which in turn depends on the ratio of the input and output clocks.
...if you're using an ASRC as a DAC front end, use an odd ball output frequency to minimise the odds of this happening.
There's a nice AP graph there illustrating it. Additionally, ASRCs using polyphase filters (most of them) have more constraints on the sample rate ratio and performance (though the 130 dB mentioned should be sufficient in theory, plenty of people claim to hear a difference with the 175 dB of Dustin's ASRC).
__________________
Chat with us live at #diyaudio on irc.rizon.net !
"Good people do not need laws to tell them to act responsibly, while bad people will find a way around the laws." -Plato
But in declaring low jitter, the power supply has more to do with jitter than almost anything else. Jitter is really dominated by really low frequency noise in the area below 10Hz. Most semiconductor companies that specialize in regulator technology don't even spec below 10Hz because the noise stats sky rocket. But these are exactly the frequencies that effect clocks and audio is full of clocks.
This isn't exactly true. Poorly designed power supplies matched with sensitive components will result in heavy intrinsic jitter, but that is only one face of the beast.
There are a lot of converters on the market that don't attenuate any jitter below 5 kHz. That means transmission jitter (usually around 2kHz and 400 Hz), interference jitter (@ freq.'s below the cutoff), and intrinsic jitter (@ freq.'s below the cutoff) will be of equal concern with these devices.
In fact, one could argue that distortion from higher frequency jitter is more detrimental then that of low frequency jitter. This is because low freq jitter has side bands that will be very near the fundamental, and will be masked, to a various degrees, by the fundamental. High freq jitter, on the other hand, will result in side bands that are far enough away that they sound like seperate, independant, yet obnoxious spurious tones.
That being said, low frequency jitter should be dealt with properly as well, especially in professional quality converters.
Originally Posted by Wavelength
If we look at the basic preface of digital audio there is in each piece what is called "intrinsic jitter". This is the jitter that a part has no matter how the out side world effects it. This intrinsic jitter can then be amplified by the power supply. You can think of this increase in jitter like a carrier wave. The power supplies noise will basically modulate this jitter to a higher level.
I'm not sure if we share the same definition of "intrinsic jitter". In my mind, intrinsic jitter encompasses the jitter from power supply fluxuations and noise, as well as jitter from thermal noise, etc. Not to get into semantics, but its important to make sure we're talking about the same thing.
Originally Posted by Wavelength
~~~~~~~ ARSC's
It is true that ARSC's can remove allot of the jitter coming in by their simply reclocking at an async rate. But most of the testing on this shows that ARSC's act like low pass filters to the incoming jitter.
So we have to look at the jitter equation for ARSC's kind of like this:
Jitter Out = (Jitter in/LP filter) + intrinsic jitter * power supply noise + intrinisc jitter oscillator * power supply noise
This kind of basic but shows that there is intrinsic jitter in the ARSC as well. Becuase there is tons of clocks all over the place here in both the filtering area as well as the input and output clocking areas. This jitter added to both the oscillator jitter and the attenuated jitter input is going to be applied to the next device in the chain. That being mostly the dac.
Dac's also have intrinsic jitter... in the same vein.
I agree with all of this, except the equation. Poorly designed power supplies can induce and contribute to the intrinsic jitter within clocks, PLL's, transmitters, ASRC's, DAC's, etc. However, they will not amplifiy intrinsic jitter. Modulate, yes...amplify, no.
But the rest of it is very true. Simply recoverying and re-clocking data is not going to eliminate all jitter artifacts from the conversion. This is why circuit board layout and proper component selection is equally important for a high-resolution conversion system. Cookbook products will give you cookbook results.
1) Everyone who has every bypassed the KMIXER has heard better results. So there must be something changed in the KMIXER.
Be careful with this. Not only is this not true, we don't want to toss the baby with the bathwater. Subjective, biased listening test should never render scientific testing obsolete (and vice versa).
Originally Posted by Wavelength
2) Since you can bypass it, then what's the big deal.
I don't think anyone is making a big deal of it. When someone asks, I will answer.
But, I agree. This isn't a big deal for two reasons.
1. Like you said, if you prefer to bypass kmixer, then you should bypass it.
2. Bit-perfect or not, kmixer is doing little-to-no damage to the audio. The programmer at Microsoft said basically what you said...the floating point math will cause an error at the 24-th bit. If this error results in any distortion at all, it will be below the threshold of hearing.
Originally Posted by Wavelength
Well 3 reason's... If Microsoft says so why even try and prove them wrong. They must know a little more about it than you do since they wrote it.
We're not out to prove Microsoft, or anyone else, wrong. We investigate to learn the truth, no matter which side the truth resides.
The issue, as I see it, is that either AP needs to readjust their bit-transparency test, or this engineer at Microsoft is overlooking something. I wouldn't bet money on which one is at fault. But, this is a matter of resolving a technical disparity. After all, what kind of engineer can witness such a disparity and just let it be.
jitter
For $1000 I could easily fly to the US and pick one up myself!