The Signal Chain: How do noise and distortion propagate through my system?

[BR52]

Thank you, Don a great hard work to bring light in the interaction of some building blocks. Often we see endless discussions about the behavior of specific systems and how they are then generalized.
Again thank you

 

[Steve H]

Thank you for a great article. I hope you can spend some time with speaker noise and distortion in the future.

 

[BR52]

Hi Don, really sad to see a so low resonance of your thread. I guess the problem is you did not give a simple answer. Most people think in simple solutions and not in more complex possibilities. Hopefully I'm wrong.

 

[DonH56]

Hi Don, really sad to see a so low resonance of your thread. I guess the problem is you did not give a simple answer. Most people think in simple solutions and not in more complex possibilities. Hopefully I'm wrong.

Yes, it's a complex subject, and I am sure I did not do enough to simplify the concepts and start with simpler examples. I had some time off and just getting the Python program running and working through the first few examples consumed way more time than I expected. Someday I may try to go back and tighten things up but for now it is what it is... For now Life and Work is ramping up (new parts coming in to test at work, older son getting married in a few weeks, so we're pretty swamped).

That said, a lot of the threads I and others (e.g. Amir) wrote get buried in the overall wash of threads on ASR, especially now that we've grown so much. A lot of basic questions could be answered, or at least understood, by wading through the various tutorials but reading and assimilating them takes time. They keep getting views, though at a low level after initial publication, so hopefully are still relevant. I rarely use a dictionary but have one on hand for the occasion that I need it and would miss it if I did not.

No worries - Don

 

[BR52]

This was no critics about your article, I enjoyed it very much. It was more related to a lot of long surface scratching threads I can see.
I'm sure there is a more important live outside HIFI, enjoy it as much as possible.
Best Bernd

 

[Sokel]

Well documented articles like this is the main reason I read ASR.
Thanks for the time and effort!

 

[sq225917]

And the result is - buy good low noise and distortion sources with plenty of output level and match the gain of your devices to limit further reamplification after the source.

 

[nerdstrike]

That was a nice read!

 

[ex audiophile]

Don, thank you for this. I didn't see in your "signature" what components you are using in your personal system. If you could suggest a few sets of components for a 2 channel system that would work well i'm sure a lot of us would be very grateful. You would NOT need to include speakers but rather just use a generic 90db speaker to make things as simple as possible.
As an example, would this be an appropriate starting point for an excellent two channel streaming setup:

digital streaming source:Benchmark DAC3B:Benchmark LA4:AHB2 (stereo or mono depending on power needs)?

For the purpose of this discussion we'd have to keep room correction, bass management and speakers out of the mix.

 

[DonH56]

Don, thank you for this. I didn't see in your "signature" what components you are using in your personal system. If you could suggest a few sets of components for a 2 channel system that would work well i'm sure a lot of us would be very grateful. You would NOT need to include speakers but rather just use a generic 90db speaker to make things as simple as possible.
As an example, would this be an appropriate starting point for an excellent two channel streaming setup:

digital streaming source:Benchmark DAC3B:Benchmark LA4:AHB2 (stereo or mono depending on power needs)?

For the purpose of this discussion we'd have to keep room correction, bass management and speakers out of the mix.

My equipment list is in second link in my signature, the one starting with "My system:". I don't have a high-performance DAC or streamer. My schedule rarely permits time for serious listening; maybe in another 147 years when I retire.

I am not the one to ask for advice, not really up on current components and don't want to wade into that fray again, and I would prefer to keep this thread focused on the technical content and not specific equipment. Better to start a separate thread for that, please.

Of course, if someone wants to give me a Benchmark or three, I'd be glad to try them out.

 

[Jomungur]

Thanks for all this work, this is very helpful for a layperson like me.

Dumb question, but hopefully a simple answer for you. You hear a lot of talk about 'jitter' in the audio world; as in something that creates usually small but audible timing distortions in the reproduced analog signal. In the way you are presenting this summary, can I think of jitter as basically noise, or related to noise, so that's picked in up in SNR? Or perhaps distortion and so it's captured in THD? Or does that not make sense and it's a wholly separate concept?

I ask because I wonder about the virtues of these devices that claim to eliminate or reduce the effects of incoming jitter, which perhaps is another way of saying they can eliminate or deal with incoming noise. I guess that should be measurable if an audio device outputs a signal with less noise than the incoming signal, and yet I've heard people claim that jitter is not always picked up in measurements like THD or SNR.

A related question, even if we had perfect THD, SNR and SINAD measurements for all stages of the system including speakers (I know this is impossible in real life), how confident can we be that we are capturing all the distortions that affect the audibility of the sound coming out of the speakers (disregarding the post-speaker chain like speaker placement, listener hearing issues, room treatments, etc.)? Just trying to better understand the limitations of these measurements.

 

[DonH56]

Thanks for all this work, this is very helpful for a layperson like me.

Dumb question, but hopefully a simple answer for you. You hear a lot of talk about 'jitter' in the audio world; as in something that creates usually small but audible timing distortions in the reproduced analog signal. In the way you are presenting this summary, can I think of jitter as basically noise, or related to noise, so that's picked in up in SNR? Or perhaps distortion and so it's captured in THD? Or does that not make sense and it's a wholly separate concept?

I ask because I wonder about the virtues of these devices that claim to eliminate or reduce the effects of incoming jitter, which perhaps is another way of saying they can eliminate or deal with incoming noise. I guess that should be measurable if an audio device outputs a signal with less noise than the incoming signal, and yet I've heard people claim that jitter is not always picked up in measurements like THD or SNR.

A related question, even if we had perfect THD, SNR and SINAD measurements for all stages of the system including speakers (I know this is impossible in real life), how confident can we be that we are capturing all the distortions that affect the audibility of the sound coming out of the speakers (disregarding the post-speaker chain like speaker placement, listener hearing issues, room treatments, etc.)? Just trying to better understand the limitations of these measurements.

Jitter adds noise to the output of a DAC (or to the samples of an ADC). There are many types of jitter but for audio typically we talk about random jitter, though clock noise can introduce correlated (non-random, fixed) jitter, so high jitter just adds to the noise floor. Check my signature for a link to my technical articles on ASR for several threads that dive deeper into jitter and other sampling ("digital audio") related topics. Jitter is a huge topic all by itself. The introductory threads on DACs are a good start.

Measurements do pick up jitter as it will degrade SNR and perhaps THD (depends on the type of jitter). Amir performs jitter tests of DACs, as do most technical reviewers (e.g. JA in Stereophile always does the J-Test to check jitter). In my world, clock jitter is one of the limitations of very high-speed data converters (ADCs and DACs operating well above 1 GS/s may require fs clock jitter levels).

Speakers were brought up earlier and are outside the scope of this thread. They will normally dominate the distortion of an audio system, and their sensitivity (how much SPL out for how many watts in) can affect the noise floor -- highly-sensitive speakers are more likely to produce audible "hiss" from the electronics. Ditto room effects, sometimes referred to as "linear distortion" as they can affect frequency response and amplitude but do not normally add nonlinearities to the signal.

HTH - Don

 

[Jomungur]

Thanks very much for the response. I'll have to educate myself on those references.

This article is helpful for someone like me, because a meaningful technical explanation does far more for me than someone just blindly asserting that I'm not hearing what I think I'm hearing. The problem is to understand what you are trying to say, you really have to dive deep into the technicalities. I mean, Fourier analysis alone takes most people weeks of study to appreciate, even with a year or two of college calculus under your belt.

The problem you face is the people who sell this stuff often have simpler explanations that seem to make intuitive sense. For example, with the jitter point, they'll say something like "Of course we're not saying bits aren't 0s and 1s ultimately, but you know timing matters. It's not just about whether all the bits get there but whether they all get there at the right time! Even a slight delay will affect things like spacing of the soundstage." And that sounds plausible to many people. E.g., I don't think I appreciated what exactly SNR and THD measurements mean until I read your article, and that, yes, they should pick up these kind of timing delays. (Although to be fair I'm still not sure why only the sum of the 1st 10 harmonics is used in THD and whether this has any implications for the measurement, but I also haven't yet read the posts you cross referenced when you mentioned that).

 

[DonH56]

Thanks very much for the response. I'll have to educate myself on those references.

This article is helpful for someone like me, because a meaningful technical explanation does far more for me than someone just blindly asserting that I'm not hearing what I think I'm hearing. The problem is to understand what you are trying to say, you really have to dive deep into the technicalities. I mean, Fourier analysis alone takes most people weeks of study to appreciate, even with a year or two of college calculus under your belt.

The problem you face is the people who sell this stuff often have simpler explanations that seem to make intuitive sense. For example, with the jitter point, they'll say something like "Of course we're not saying bits aren't 0s and 1s ultimately, but you know timing matters. It's not just about whether all the bits get there but whether they all get there at the right time! Even a slight delay will affect things like spacing of the soundstage." And that sounds plausible to many people. E.g., I don't think I appreciated what exactly SNR and THD measurements mean until I read your article, and that, yes, they should pick up these kind of timing delays. (Although to be fair I'm still not sure why only the sum of the 1st 10 harmonics is used in THD and whether this has any implications for the measurement, but I also haven't yet read the posts you cross referenced when you mentioned that).

Understanding Fourier takes some study, but most folk understand a frequency plot and mange to get the general idea.

I expect most people will not bother to put forth the effort to read even these simplified articles and continue to "poo-poo" anything but their ears. Fortunately sites like this exist to help those like yourself who really do want to go further and not just blindly follow the marketing adverts. And many companies produce ad copy that hits some fundamental truth but is misapplied to audio, or ignores the reality of how the circuits are actually implemented to spread FUD. Audiophiles do not understand how say Ethernet data transmission works, or how modern DACs resample the clock and obviate whatever fancy clock scheme people add in front of it, nor should they (I am certainly no expert on Ethernet links, but know some of the basics), and marketeers prey on that ignorance. It is not the fault of the consumer; I wish there were better ways to report false or misleading advertising not requiring thousands or hundreds of thousands in lawyer fees.

Ten harmonics for THD is an industry standard. By the tenth harmonic, it is essentially noise unless something is wrong. BTW, SINAD is the measurement I use, because it includes all noise and distortion. But, while it provides an "easy" single number, you must dig deeper to obtain critical information for your system. Is it dominated by noise, harmonic distortion, jitter, power supply harmonics, ??? A single number does not tell all. What it is useful for (to me) is weeding out things -- if the SINAD is 40 dB (1% error), I probably wouldn't bother to look at it. If it is 100 dB compared to another product at 110 dB, then I want more information to compare the products.

At the end of the day I am not really fighting the people selling stuff, nor consumers who have no interest in any sort of objective measurement. I hope to help knowledgeable "lay" people, interested audiophiles, and fellow techies who's expertise lies in other areas by showing some of the details behind the measurements and technical verbiage.

 

[SamR]

• DAC output = 1 V max, 120 dB SNR, -120 dB THD (0.0001%);

I don’t know of any DAC that outputs only 1V. I know it’s an example, but why start with something unrealistic? Most DACs output 2-4V, especially reasonably priced units that members here will be interested in.
SND of -120dB also is unrealisitic in the real world.

Interesting otherwise, but it would be great to use real world numbers.

 

[Marc v E]

@DonH56 thank you for this detailed information! It suddenly dawned on me that the complexity of the chain of devices is probably the main reason for reviews on ASR to set the bar so high, at -115db Sinad. When a device complies to this standard it is for certain it won't withhold any other device in performance.

 

[pma]

SINAD numbers of individual components are a bit autotelic. The whole chain and level diagram of all components must be taken into account.

 

[Stinius]

SINAD numbers of individual components are a bit autotelic. The whole chain and level diagram of all components must be taken into account.

This is exactly right.

 

[Beershaun]

Just found this thread. Thank you Don! It has always been an important question in my mind to understand how the complete component chain stacks in order to understand the final value inclusive of the gain structure and clipping limits. This is fantastic. If more people saw this thread I think there would be less threads about "is it audible or not..."