Let's talk jitter.

[LHC]

34 minutes ago, Simonon said:

Maybe he also writes equipment reviews.

It ain't so simple. That guy is Ted Smith, a digital engineer and DAC designer for PS Audio. 

[Guest scumbag]

38 minutes ago, Simonon said:

55 minutes ago, Ittaku said:

Claims, yes, but he didn't actually describe what he heard or what his methodology was but people can claim all sorts of things...  Anyway if his unsubtantiated comment can be trusted, then that puts us into the femtosecond range.

I wonder if he is also into power cords and bybees. I take these claims with a grain of salt. Maybe he also writes equipment reviews.

Cheap shot. Ted is a zero bs guy. He is humble and very knowledgeable. A rare combination. He wrote his own fpga based dsd dac algorithm that was integrated into the directstream dac. He also has good ears and trust's them. God forbid.  

[Guest Simonon]

It ain't so simple. That guy is Ted Smith, a digital engineer and DAC designer for PS Audio. 

Interesting as PS audio marketing techniques and Paul Mcgowans blog were discussed in depth by some in my audio myths and misconceptions thread. Nothing against PS audio products but some of the claims made by them I find questionable and purely for marketing to the non technical audiophile. Therefore I would take this claim with a grain of salt. Imo claims like this tarnish the name of a good product but this is my personal opinion and I accept others may differ.

[Guest scumbag]

Is ted smith an alias for paul mcGowan? 

[LHC]

5 minutes ago, Simonon said:

23 minutes ago, LHC said:

It ain't so simple. That guy is Ted Smith, a digital engineer and DAC designer for PS Audio. 

Interesting as PS audio marketing techniques and Paul Mcgowans blog were discussed in depth by some in my audio myths and misconceptions thread. Nothing against PS audio products but some of the claims made by them I find questionable and purely for marketing to the non technical audiophile. Therefore I would take this claim with a grain of salt. Imo claims like this tarnish the name of a good product but this is my personal opinion and I accept others may differ.

That is why I said it ain't so simple. 

[Guest Simonon]

That is why I said it ain't so simple. 

I beg to differ with the following points. The claim and jitter figure is based on a sighted listening test by a designer with a vested interest in marketing of products. No doubt he is a clever guy but the fact remains this claim may be based on maeketing of PS audio product and the figure should not be taken as reference for this reason. Isnt this example part of the problem with this industry and is always a touchy subject. Again my opinion

[LHC]

42 minutes ago, Simonon said:

1 hour ago, LHC said:

That is why I said it ain't so simple. 

I beg to differ with the following points. The claim and jitter figure is based on a sighted listening test by a designer with a vested interest in marketing of products. No doubt he is a clever guy but the fact remains this claim may be based on maeketing of PS audio product and the figure should not be taken as reference for this reason. Isnt this example part of the problem with this industry and is always a touchy subject. Again my opinion

We don't even know if his listening test is sighted or blind. Yes, there is the possibility of vested interest and marketing; but he also said not everyone can hear jitter at such low level (so why would he say that if he aims to sell as many products as possible?). It is true there is a problem with the industry when it comes to these things; but consumer cynicism doesn't help matter either. We need independent umpires who we can trust and have excellent trained hearing I reckon. Neutral umpires who don't have vested interests, but also not out there to prove a point or go on a crusade. 

[Guest rmpfyf]

3 hours ago, Simonon said:

The first part doesn't sufficiently characterise jitter, the latter needs unpacking - what does 'not even worth a mention' mean?

My bad with the terminology in regards to phase noise. One of the measurements I performed regulary was phase noise and jitter measurements of the local oscillator on digital broadcast transmitters using some very nice high end test gear. Phase noise actually became problematic when a precision 7v power supply became noisey in the local oscillator synthesiser module of a particular high power digital television transmitter causing a phenomenom called centre carrier breakthrough which ultimately affected the MER or modulation error ratio at the output of the transmitter. We had a separate program on the spectrum analyser to measure the phase noise of the local oscillator.

I started using this equipment to measure DA converter clock jitter and phase noise of a bunch of dacs and cd players I have out of interest and fun.

I found phase noise of the DA converter clock to be so low it was negligable being so far down in the spectrum analyzers noise floor it was hard to gain a meaningful measurement and in my opinion to not be a factor that would audibly affect sound. Likewise even on a basic clock in an entry level cd player measured jitter was in the order of pico seconds.

I agree that I may be missing something here or their were flaws in my methodology as my background is in digital broadcast and Rf which is why I find this thread interesting. I always like to be proven wrong but not without a robust argument

I would like to experiment more with digital music servers, DA converter chips and clocks but find the cost hard to justify for an experiment when the performance of modern equipment is so good.

Great thread.

 

Certainly is a great thread

 

I find different oscillators/circuits to sound different. I don't know what it is about them that sounds different.

 

F'rinstance anecdotally I'm settled that lower phase noise close to the target is a good thing; I have no idea if hitting silly low, moderately low or whatever else phase noise numbers far away from the target actually matters.

 

With my current rig (FPGA for timing, not ideal) I can also induce jitter with OS changes - far upstream of the output clock signal - says much for system fallacies in some USB audio applications.

 

I'd get yourself one of IanCanada's boards and have a play  not so expensive. This said, as I'm often reminded, it's not just the oscillator that counts...

[Guest rmpfyf]

3 hours ago, Simonon said:

So here is a question that somebody may be able to answer
Why is their a revival of interest in the perceived analogue sound of TDA 1540 equipped early generation 14 bit CD players and later 16 bit TDA1541 dac chips.
I note one popular methodology is to use a number of dac chips in parrallel to average the slight variation in timing of these chips again for a perceived more rounded analogue sound or so the marketing says.
Would the ability to vary clock jitter have a similar affect to suit personal tastes? Would a jitter free clock sound harsh and sibilant when compared to the same dac chip with a relatively high amount of clock jitter. I am trying to gain an understanding of how jitter will audibly affect sound when compared to the use of dither in digital mastering for example.

 

They're popular because they're awesome  

 

I'd imagine a jitter-free clock to sound like exactly what's on the recording regards timing... though @zenelectro has noted that this may not be ideal to taste. When considering that a jitter-free clock in playback would only leave what timing errors were afforded by jitter in recording, it could stand to reason that some jitter in playback to further randomise what's to playback to be not a bad thing.

[Guest rmpfyf]

2 hours ago, scumbag said:

Bear with me here. Say you have a dac with very high jitter. So much that it if easily audible. It is a distortion as at any point in time it represents a deviation from the original signal. Which is measured in dB. 

https://www.stereophile.com/reference/1093jitter/index.html ". Specifically, jitter with a frequency of 1kHz affecting a DAC reproducing a 7kHz sinewave will create spurious output tones at 6kHz and 8kHz. If the jitter has a frequency of 2kHz, the jitter-created artifacts will appear at 5kHz and 9kHz. These sidebands around the signal being decoded aren't harmonically related to the signal, making them particularly unpleasant. When the DAC is reproducing music (which has a constantly changing spectral content) and is controlled by a jittered clock (that may be jittered at several frequencies), the potential for generating a highly complex spectrum of jitter-induced spuriae is obvious. If the jitter is "white" (having a random spectral distribution rather than discrete frequency components), analog white noise will be added to the DAC's output signal"

 

 

I get what's being said but it's a misunderstanding in the application of signal processing methods. 

 

Jitter does not introduce new frequencies. There are no new oscillations, just their shape has changed owing to timing errors. If a perfect sine wave was the intended shape and Fourier transforms are used, there is no way to recreate the output wave inclusive of the timing error(s) without some low-amplitude contributions from sum components of other frequencies mathematically - this doesn't mean that (in your example) that there are 'output tones' at those frequencies. You're still hearing a (broadly) 7kHz oscillation in your example irrespective of the jitter - to recreate it's shape in the time domain using Fourier methods (ie using sines and cosines) you need to add components from other frequencies at lower amplitude. You'll note the amplitude of the target drops slightly, and that net energy is broadly preserved.

 

Whether that sine wave becomes a bent or squashed sine, a triangle, a sawtooth or even square wave... you're still hearing a 7kHz oscillation. 

 

If you implement a transform using (respectively) a bent or squashed sine, a triangle, a sawtooth or even square wave... you'll have one spectral point at 7kHz of identical amplitude (for infinite transforms) in every case.

 

'White' timing errors do not add white noise.

 

Amplitude distortion at any point in time is measured in amplitude units, temporal distortion should be measured in time units. A distribution of temporal distortions can be characterised by relative amplitude of relative frequency error, which is where we get dBc/Hz type measurements from. It's not 'dB of some other spurious frequencies that rocked up over temporal error'.

[MLXXX]

If you use Audacity to generate a 700Hz sinewave, at a sample rate of 48kHz, and modulate it using the vibrato plug-in effect I mentioned earlier in this thread, with the parameters set to 100Hz variation rate , and depth of 20%, you will hear a roughness.  (I myself hear distinct additional tones.)  The point I would like to emphasize is that 20% is a very substantial modulation depth. The effective sample rate (nominally 48kHz) is being varied over a huge range. I would assume that  broadly speaking it would range from roughly 39kHz up to roughly 57kHz (repeating at 100Hz).  That would be a huge quantum of artificial jitter!  I've found that if you use a vibrato depth of 10%, the effect is very much less noticeable. [I'm not able to upload the files right now, but could do so later, if anyone expresses interest.]

 

@rmpfyf, this would be a very artificial exercise, but if you did vary a  high frequency clock at a low rate like 1kHz, then a 7kHz tone sampled by that clock would not be captured as a tone of a fixed frequency but as tone varying in frequency between limits set by the extent of the sweep of the wobbling clock rate. For example if the modulation is such as to sweep the sample rate between -14% and +14% of its nominal value then the 7kHz tone will be effectively captured  as if varying between about 8kHz and 6kHz. When played back without wobble, at the nominal clock rate, the analogue tone will continually sweep between 8kHz and 6kHz  [approx].

[Guest rmpfyf]

3 hours ago, MLXXX said:

If you use Audacity to generate a 700Hz sinewave, at a sample rate of 48kHz, and modulate it using the vibrato plug-in effect I mentioned earlier in this thread, with the parameters set to 100Hz variation rate , and depth of 20%, you will hear a roughness.  (I myself hear distinct additional tones.)  The point I would like to emphasize is that 20% is a very substantial modulation depth. The effective sample rate (nominally 48kHz) is being varied over a huge range. I would assume that  broadly speaking it would range from roughly 39kHz up to roughly 57kHz (repeating at 100Hz).  That would be a huge quantum of artificial jitter!  I've found that if you use a vibrato depth of 10%, the effect is very much less noticeable. [I'm not able to upload the files right now, but could do so later, if anyone expresses interest.]

 

@rmpfyf, this would be a very artificial exercise, but if you did vary a  high frequency clock at a low rate like 1kHz, then a 7kHz tone sampled by that clock would not be captured as a tone of a fixed frequency but as tone varying in frequency between limits set by the extent of the sweep of the wobbling clock rate. For example if the modulation is such as to sweep the sample rate between -14% and +14% of its nominal value then the 7kHz tone will be effectively captured  as if varying between about 8kHz and 6kHz. When played back without wobble, at the nominal clock rate, the analogue tone will continually sweep between 8kHz and 6kHz  [approx].

 

Sure, I get that, though it's quite different to jitter and the only reason you're seeking additional spectra is that the spectral density shown is averaged over a time period where the modulation is statistically valid. You're also not seeing peaks at 6 and 8kHz representing frequency components to change a waveform shape, you're seeing broadband spectra between that range with a peak in the middle because the frequency is actually changing. 

 

Quite different to the shape of the oscillation / nature of the pressure variation changing with broadly identical periodicity.

[LHC]

6 hours ago, rmpfyf said:

When considering that a jitter-free clock in playback would only leave what timing errors were afforded by jitter in recording, it could stand to reason that some jitter in playback to further randomise what's to playback to be not a bad thing.

Oh no  , you are opening that pandora's box !

[Guest rmpfyf]

6 minutes ago, LHC said:

Oh no  , you are opening that pandora's box !

@zenelectro got there first  

[Grant Slack]

12 hours ago, LHC said:

We don't even know if his listening test is sighted or blind. Yes, there is the possibility of vested interest and marketing; but he also said not everyone can hear jitter at such low level (so why would he say that if he aims to sell as many products as possible?).

Well, it does rather let him off the hook, though, doesn't it? If he says that everyone can hear it, then what does he say when someone says "I can't hear it"? That invites trouble.

 

It is best to say that not everyone can hear it.

 

Buyer: "I can't hear it."

 

Mr Smith: "Well, I can clearly hear it. And you might, too, under different conditions or music."

 

Buyer: "Here, take my wallet."

 

Perfect!

 

Regards

Grant

[davewantsmoore]

On 12/11/2018 at 9:49 PM, Ittaku said:

The same way it was discovered that 2nd order harmonic distortion is far less audible and quite pleasant

  Myth

[Ittaku]

7 minutes ago, davewantsmoore said:

  Myth

It's a pretty established phenomenon quoted in enough literature. Prove it?

Edited November 14, 2018 by Ittaku

[davewantsmoore]

1 hour ago, Ittaku said:

Prove it?

I can't prove something isn't true, but...  I can ask, what decent studies have show that even order distortion applied to music is pleasant?

 

In a very simple sense, even-order is less audible ... when it's on it's own (which is a completely artificial scenario) .... but when it's intermingled with a complex signal, the distortion components are no longer neatly related to the fundamental... and so they are no less audible, and no less unpleasant.

 

One way to look at most of this .... is that it's a myth that these common assessments relate to audio recording or playback systems .... because in these systems the distortion (THD) is inter modulated with complex signals (like music is).... as opposed to being applied to a single tone, or a simple waveform (like a single clean musical instrument).    It is the inter-modulation with a complex signal which makes typical distortion audible (as it pushes it outside the typical boundaries where it is masked, based on level and frequency)

 

 

Rod Ellitot:

Quote

 

Just in case anyone is wondering just how I came across this intriguing phenomenon, it was while I was testing the hypothesis that even-order distortion sounds 'nicer' than odd-order distortion.

<SNIP>

.... the whole exercise belies the claims made by those who say that second harmonic distortion is pleasant and adds to the music. It doesn't do anything of the sort.

 

 

It is commonly understood that while you could add distortion for instrument tone as your heart desires.... that adding distortion to an entire mix (like we are talking about there, in a playback system) will not be pleasant, and the idea of "good distortion" is junk.

 

2 hours ago, Ittaku said:

It's a pretty established phenomenon quoted in enough literature

It is the context which it is quoted ... and whether that applies here, which is the issue.

[davewantsmoore]

On 13/11/2018 at 2:18 PM, Ittaku said:

This is pretty much what all the most expensive amplifiers do

I do not think that good amplifiers tune the harmonic distortion "profile" the way they do, because harmonic distortion sounds better that way.

 

The real reason is to do with the cause of the distortion .... and what the distortion profile says about the cause.     "monotonically decreasing" distortion .... or other "benign" spectrum of distortion components in an amplifier mean that the thing causing the distortion isn't so bad..... where as certain distortion components "sticking out" mean that the type of non-linearity in the circuit is more severe, and more "tuned" (which could be set off, by specific input signals)..... and hence be more audible under specific conditions.

[Ittaku]

At least in the case of passlabs (or was it ps) audio, I've seen a video with him saying they literally aim to make the distortion mostly out of phase 2nd harmonic. So yes, they do tune their distortion profiles. No, I can't find the link right now.

Edited November 14, 2018 by Ittaku

[Guest Simonon]

This again sounds like marketing to the audiophile that has been told that 2nd harmonic distortion sounds nicer......this is essentially a myth. The reality is that distortion levels on modern amplifiers is so low in the order of .001% for example at deafening listening levels. Frequency response is ruler flat across the audio spectrum and beyond. 2nd or third order distortion is no longer relevant in my opinion.
It is great marketing to imply an amplifier is tuned for 2nd order distortion to audiophiles as it a common myth to say valve amplifiers sound better than solid state due to the presence of second harmonic distortion. The reality is that distortion is close to zero on a modern amp at normal listening levels.

[Ittaku]

9 minutes ago, Simonon said:

This again sounds like marketing to the audiophile that has been told that 2nd harmonic distortion sounds nicer......this is essentially a myth. The reality is that distortion levels on modern amplifiers is so low in the order of .001% for example at deafening listening levels. Frequency response is ruler flat across the audio spectrum and beyond. 2nd or third order distortion is no longer relevant in my opinion.
It is great marketing to imply an amplifier is tuned for 2nd order distortion to audiophiles as it a common myth to say valve amplifiers sound better than solid state due to the presence of second harmonic distortion. The reality is that distortion is close to zero on a modern amp at normal listening levels.

Maybe. But if they're all ruler flat and have zero distortion, why do they sound different? I guess there are people that argue that amplifiers all sound the same too, barring frequency response changes from impedance loading. I beg to differ... but anyway we've gone way off topic here. Back to DACs having effectively zero jitter and them all sounding the same instead.

[Guest Simonon]

It would be great to list and discuss the types of DA converter chips in use on DACs and common circuit designs which vary considerably. You could start with 14BIT TDA1540 and work up to DA converter chips in common use today PCM 5122 for example. I suspect clock jitter has little influence here but the DA chipset and topography does make audible differences between DACS. One also needs to consider the output stage with the types of opamps used including the use of tube buffers in some, lampizator for example. The clock is only a small part of the puzzle.

[Ittaku]

They're all interesting things, but alas not the topic of this thread. Jitter's interesting in its own right, even if the conclusion is it doesn't matter any more. We probably need other threads for the rest of the discussions.

[davewantsmoore]

39 minutes ago, Ittaku said:

So yes, they do tune their distortion profiles.

Yes, it's clear that people do that (so there's no need to evidence) ..... but "why" is more pertinent.

 

"Certain distortion sounds a certain way"  (so we should try to achieve the more pleasant sounding distortion) is a somewhat oversimplification of what is happening.

 

The profile of the distortion frequency components tell us things about the non-linearity in the amplifier which caused it.    The benefit of targeting a certain distortion profile isn't about that distortion being "good sounding", but more about not letting certain types of non-linearites occur in the circuit.

 

32 minutes ago, Ittaku said:

Maybe. But if they're all ruler flat and have zero distortion, why do they sound different?

Because they don't have zero distortion.... or some other condition (aside from actual distortion of the audio) is causing the audibility.

 

32 minutes ago, Ittaku said:

Back to DACs having effectively zero jitter and them all sounding the same instead.

NLD from jitter is typically at a very low level ..... but is at frequencies potentially well away from the signal.    In a lot of cases it would not need to rise far to be unmasked.   If a DAC were "fine most of the time, but with a very jittery instant here and there" the typical ways jitter is quantified wouldn't show that.

 

As mentioned earlier.... Correlation into a certain channel seems to be a big way distortion can be unmasked ....  which is specific to digital audio due to the very high channel separation possible.