Jitter at DSD rates

[pkane2001]

Things move quickly in the Objectify forum, it seems. Just as I was typing the response to @jabbr, the thread was shut down by our benevolent overlord.

 

Here's what I was trying to type:

 

DSD content is simply a recording done at a much higher sampling rate. DSD512 is the same signal, but sampled at 512x the standard PCM rate and with 1 bit samples. Since the recorded signal is the same, and assuming the clock modulating noise/signal is also unchanged, the resulting distortions due to jitter at the output will also be the same. Now, when the circuitry is working at 22MHz, there will be other, much higher frequencies that might potentially contaminate the clock, but hopefully these will be completely out of the audible range (some of this would be correlated jitter).

 

[pkane2001]

37 minutes ago, jabbr said:

Hmmm this is what I’m considering:

 

At 44kHz a f +|- 1 Hz error will be 43kHz then 45 kHz and 20 +|- 0.5 Hz might theoretically be audible — I pick 1 Hz because the 1/f pattern tells us that phase error increases as offset decreases 

 

Yet: at 24 MHz +|- 1 Hz, I’m not seeing how this could reasonably be audible, let’s say it’s -100 dBc and so 30 dB down (from 22 MHz to 20 kHz is already -130 dB)

 

I think a sim would show this

 

Jitter sidebands are formed around the signal frequency, not around the sampling frequency. +/-1 Hz sidebands around 1kHz signal will stay at 999 and 1001Hz regardless of the sampling rate, since the signal frequency will stay at 1kHz.

 

Here's 10ns 100Hz sine-modulated jitter at 1MHz sampling rate:

 

And here it is at 48kHz:

[pkane2001]

4 minutes ago, jabbr said:

 

No. The question is whether phase error in the DSD clock will affect the reconstructed audio.
 

The “signal” of DSD512 is 22 MHz. The jitter is on the DSD clock ie 22 MHz +|- offset.
 

 

 

 

Oh, I see. If you are talking about the inherent clock jitter (not externally modulated one, as I was thinking you meant) then yes. That kind of jitter will be moved way out of audible band by the high clock rate.

 

[pkane2001]

7 hours ago, manueljenkin said:

Corrections!

 

1. Mention of DSD is "simply a recording done at higher sample rates". Actually no, in a PCM, the value of a sample constitutes the amplitude at that sample, while in DSD, it is dependent on the samples that preceded it.

 

2. It will still have to pass through analog filters in the dac and DSD typically needs quite a bit of low passing afaik. So all you jitter deviation might still be altered by the filter after the dac. Point in case, this means DSD is not the end all be all assessment for the analysis.

 

1. No need to correct. I didn't say DSD is the same as 1 bit PCM. I said: "DSD512 is the same signal, but sampled at 512x the standard PCM rate and with 1 bit samples." 

 

2. Of course DSD requires a low-pass filter. But that doesn't have any effect on any jitter that was already within the audible frequency range, say the 60Hz mains frequency. Proper PCM reconstruction also requires a low-pass (reconstruction) filter.

 

I wrote a DSD/DSF reader a while ago, and decided to make the low-pass filter completely configurable by the user. You can place the cutoff wherever you want, and decide on how wide the transition band should be:

 

 

[pkane2001]

28 minutes ago, Miska said:

 

I wonder what kind of filter slope do you have? At the moment I have 8 different filter response choices. But I prefer and default to the DSD/SACD spec one (scaled to DSD sampling rate).

 

 

I construct the filter on the fly, based on the settings specified by the user. The transition bandwidth and the cut-off are both user-controlled, while the slope is computed from these and the desired ripple settings. I pick the size internally to get the lowest pass-band ripple while keeping filter size between 10k-32k taps. 

 

[pkane2001]

1 hour ago, Miska said:

 

OK, that's one of the many ways to do it... Of course depends on your transition band specs and roll-off shape, etc.

 

 

Yes, of course. This was just to let me experiment with different cut-off frequency and shape filters, it wasn't to produce the best "sounding" filter, if there is such a thing 😄

 

 

[pkane2001]

46 minutes ago, Miska said:

 

Yeah, note that I wasn't talking about sound. As far as I can see, those parameters give different steepness, but always the same curve? That filter is both noise filter and resampling filter or the two are separate?

 

 

Same filter. Do you think that filter steepness well above the audible range is important?

[pkane2001]

58 minutes ago, Miska said:

 

I think it is more complex matter than just filter steepness. But why do you then have such adjustment in first place if it doesn't matter?

 

 

You brought up filter slope  The software has lots of settings for things I'd like to experiment with, including 20+ FFT window functions, IIR and FIR filter selections, variable frequency and bandwidth of filter transition, etc. It's not because I think all of these have an audible effect, but rather to let me play with various parameters to test how they affect the signal. There are actually a lot more settings that I play with internally in the software, but some I've elevated to the level of UI so others can play with them, if curious.

 

 

[pkane2001]

10 hours ago, Siltech817 said:

 

Including for example physical vibration/resonance, no?

 

Do you have your DAC sitting on top of a subwoofer? 

[pkane2001]

7 hours ago, Siltech817 said:

Of course not, however I was not only referring to airborne vibration, this can even be a function of the power supply, or the resonant characteristics of certain capacitors or other elements in the circuit including the placement of the oscillator itself, and the overall layout of the board.

Since everything is built to a price point, not all will be created equal with regard to careful attention paid and budget allowed for this element of the design. Not a one size fits all answer, and yes it's audible, and no I don't have any sophisticated measurement equipment, nor do I feel compelled to provide any ABX testing results.

Place a DAC on different kinds of footers, shelves, racks, etc... made from different materials. It does not all sound the same. Neither do all different DSD capable DACs sound the same, and this is likely one reason or contributing factor as to why, unless you think jitter is fake news or something like that. TI for example does not.

It's not crazy to think this sort of thing can be audible, though because the audio industry has never agreed to a set of criteria for what and how to measure jitter, and then proven what is audible and what is not, it's silly for anyone to get up on a soapbox and suggest they know everything there is to know about it, or that the only issue would be if one were so stupid as to place a DAC on top of a subwoofer.

snaa296.pdf 2.77 MB · 5 downloads

 

So the paper used 4g acceleration for sine vibration, 7.5g for random, and 500g for shock. I would hope you don't normally experience these levels anywhere near your DAC... But, if you do, then there may be some added jitter. This would place your listening room conditions to be somewhere between a jet airplane and boost stage of a missile:

 

 

 

[pkane2001]

5 hours ago, Siltech817 said:

Again, SO WHAT? 

You read the TI brief correctly of course, but that in no way means that you have solid verifiable evidence that vibration or resonance at lower levels, including that which is not airborne, bears no audible effect on audio systems.

You don't know that, and I can tell that either you've never ever tried what I described with using footers of differing materials, shelving, racks etc... under DACs (or any other audio component for that matter) or, if you did, your expectation bias (that is: you already know all there is to know about this) fully prevented you from hearing any differences.

Note I make no claim as to the exact magnitude of such a difference, nor make any claim that one can equate that difference with any particular cost delta in an exacting manner. To each their own on the cost delta.

Did your subwoofer as component support example actually prove anything relative to the paper I attached? I thought according to your post this was only an issue of jet engine type vibration, or rocket launch type vibration? Now only vibration from a subwoofer can create an audible difference in the realm of audio systems? Really? At what SPL must we have the sub's output before a problem can occur?

So you know this to be the case then, and all assertion to the contrary can easily be refuted with detailed measurements as provided by who, the "science" site?

I think not.

BTW - I see you fully skipped the part about board layout, resonant characteristics of capacitors chosen, or the power supply proximity. Shall we include EMI/RFI fields as well? Thermal influence too?

 

Wow, buddy. You are putting so many words in my mouth, I'm starting to choke. Did I claim any absolutes? Did I state that only vibration from a subwoofer can cause jitter? Did I assert anything at all, except to provide an interpretation of the paper you posted, which is perfectly consistent with the study conducted?

 

How about, instead of you telling me that I can't possibly know everything (hint: I don't), or telling me that I can't prove a negative (hint: I can't) you provide an objective evidence that you can hear the effects of a crystal vibration due to the speaker sound in your listening room (hint: you can, but only if you want to).

 

Oh, and why do you bring in EMI/RFI or thermal effects that are not related to the simple physical shock and vibration? I made no statements about these.

 

5 hours ago, Siltech817 said:

My point is that neither pkane2001, nor anyone else, possesses any absolute proof as to what is audible

 

Only a Sith Deals in Absolutes. And I'm no Sith 😜

[pkane2001]

28 minutes ago, Siltech817 said:

You made a condescending and jerky reply to a post I made that was not directed at you, rather it was directed at Miska, and it said exactly this:

 

 

I asked him a question, somehow that generated your rude subwoofer comment, perhaps I should have played along and asked you if that were the only such condition we should be concerned with, but you knew what you were inferring, or you wouldn't have said it. Then the follow on:

 

 

More condescending sarcasm, but hardly surprising, I've seen this act before.

 

 

Perhaps I did misunderstand the purpose of the thread, I apologize for asking a simple question about whether or not we can consider including the effects of vibration and resonance on jitter at DSD rates. Evidently the OP found reason to launch into condescending sarcasm and mockery as the reply. I'll take his answer as reason not to set my DAC on a subwoofer and call it a day.

 

 

I'll chalk this up to you having a bad day. You misinterpreted a very short question I asked you, proceeded to read into it something I didn't say or mean to say, and then went on a tirade. Please, stay out of this thread until you can communicate in a civil fashion and without the insults. @The Computer Audiophile, please give me moderator rights.