God of SINAD vs. reality we get from most available music files

[solderdude]

I am afraid the 'general audio gear' consumer has no idea how to use parametric EQ. I think most people even have trouble with graphic EQ when it has more than 5 bands.
All they generally do is make 'smilies' (boost bass and treble).

The RME is intended for studio usage (where they know about this) and adopted by audiophiles that like to use a device that does this function without the need of using software.

The Qudelix is a rare duck and popular with people desiring this functionality without having to use apps/programs.

Most people that are aware of the benefits of properly used parametric EQ, which by itself is not that easy to apply when not having presets or measurement gear/skills, already use software for this (even on phones).
Education/awareness is needed before someone can use it 'properly'.

 

[jhaider]

I am afraid the 'general audio gear' consumer has no idea how to use parametric EQ...
Education/awareness is needed before someone can use it 'properly'.

Bullshit. Today in the headphone space there's a wealth of information from e.g. @amirm or @oratory1990 accessible to anyone who can enter data into an electronic device. It's not hard to find, ether. Anyone who reads a thread on a decent audio forum about a given headphone will find their way to these tuning aids. That step alone is sufficient to turn (for example) a borderline-unlistenable Sennheiser HD800 into a remarkable experience.

In home audio it's a little different. Most "audiophiles" do unfortunately seem content slipping gear in and out of basically unoptimized systems.

Besides, if people can obsess about airhead nothings like "omigod my DAC isn't lighting up 192 I'm not getting all the resolutionz " they can learn about things that truly matter.

 

[antcollinet]

Bullshit. Today in the headphone space there's a wealth of information from e.g. @amirm or @oratory1990 accessible to anyone who can enter data into an electronic device. It's not hard to find, ether. Anyone who reads a thread on a decent audio forum about a given headphone will find their way to these tuning aids. That step alone is sufficient to turn (for example) a borderline-unlistenable Sennheiser HD800 into a remarkable experience.

In home audio it's a little different. Most "audiophiles" do unfortunately seem content slipping gear in and out of basically unoptimized systems.

Besides, if people can obsess about airhead nothings like "omigod my DAC isn't lighting up 192 I'm not getting all the resolutionz " they can learn about things that truly matter.

Hmm - I could count the number of people in my friends/family circle who know about the use of DSP/EQ for home audio on the fingers of one...err...finger. Even fewer would care enough to pay the money, and jump the hoops.

Until it is integrated into gear they buy anyway at the price they pay anyway, and is fully automated it will remain a niche application.

 

[solderdude]

they can learn about things that truly matter.

And that's the difficult part.... education. For that you would have to pique interest in 'average' consumers first.
Do you reckon everyone will prefer EQ, crossfeed, DSP alike things in the spatial domain and other gimmicks ?
Most people just want to switch a device on, look for the latest cat videos (and seeing advertisements everywhere) and info about the folks the follow.
There is not a huge market for 'closer to perfect sound'.
Besides the vast majority of consumers don't really care about 'accuracy'.
They want 'looks' over sound and as long as there is bass it all sounds fine to them.
It's merely a very small portion of consumers that really care enough to go though all the trouble of applying digital EQ that involves 'copying' settings or uploading it to devices and know which sites they can get the required info from and realize what that info actually is and how that data was generated and under what conditions.

Personally I would love to see more devices (that have a longer lifespan than 3 to 4 years) that can process digital EQ with ease that is of high quality (RME) and affordable and small and very easy to operate even for noobs. However, there needs to be a standard for it and an easy way to implement (via www ?) for it to ever become as simple as looking at cat videos.

 

[Blumlein 88]

And that's the difficult part.... education. For that you would have to pique interest in 'average' consumers first.
Do you reckon everyone will prefer EQ, crossfeed, DSP alike things in the spatial domain and other gimmicks ?
Most people just want to switch a device on, look for the latest cat videos (and seeing advertisements everywhere) and info about the folks the follow.
There is not a huge market for 'closer to perfect sound'.
Besides the vast majority of consumers don't really care about 'accuracy'.
They want 'looks' over sound and as long as there is bass it all sounds fine to them.
It's merely a very small portion of consumers that really care enough to go though all the trouble of applying digital EQ that involves 'copying' settings or uploading it to devices and know which sites they can get the required info from and realize what that info actually is and how that data was generated and under what conditions.

Personally I would love to see more devices (that have a longer lifespan than 3 to 4 years) that can process digital EQ with ease that is of high quality (RME) and affordable and small and very easy to operate even for noobs. However, there needs to be a standard for it and an easy way to implement (via www ?) for it to ever become as simple as looking at cat videos.

You lost me at cat videos. I like cats alright, but cat videos are the lowest common denominator. Yes, I know that was actually your point. It still rubs me the wrong way.

 

[jasonhanjk]

I am happy with the SINAD metric being the sum of noise d THD which is what matters ultimately.

My only complain is that it is measured at 1khz and 2V.

i would be happier if our beloved SINAD was the maximum SINAD measured on a range of voltage (0.5 to 2V) and frequency (5hz to 20khz).

Many thanks.

I believe that will require a lot of time to test.
A much better way is to read that THD+N vs amplitude and THD+N vs frequency graph.

 

[Vacceo]

And that's the difficult part.... education. For that you would have to pique interest in 'average' consumers first.
Do you reckon everyone will prefer EQ, crossfeed, DSP alike things in the spatial domain and other gimmicks ?
Most people just want to switch a device on, look for the latest cat videos (and seeing advertisements everywhere) and info about the folks the follow.
There is not a huge market for 'closer to perfect sound'.
Besides the vast majority of consumers don't really care about 'accuracy'.
They want 'looks' over sound and as long as there is bass it all sounds fine to them.
It's merely a very small portion of consumers that really care enough to go though all the trouble of applying digital EQ that involves 'copying' settings or uploading it to devices and know which sites they can get the required info from and realize what that info actually is and how that data was generated and under what conditions.

Personally I would love to see more devices (that have a longer lifespan than 3 to 4 years) that can process digital EQ with ease that is of high quality (RME) and affordable and small and very easy to operate even for noobs. However, there needs to be a standard for it and an easy way to implement (via www ?) for it to ever become as simple as looking at cat videos.

When I used Audyssey for the first time it was a very, very easy process. That was around 15 years ago. I have used it recently due to a change of gear and it was even easier. I guess Dirac, in it's most basic (no custom) form is not too different in terms of usability.
Going full in is not that easy, but getting the most basic sweap is quite easy.

And well, that should not be at odds with pushing a higher SINAD on the device.

 

[Cote Dazur]

I get the feeling that while subjective audiophiles often worry about what doesn't exist, objective audiophiles often worry about what doesn't matter.

This comment resume the situation to me, so many are lost in the weeds, arguing about insignificant details, when at the meantime listening in poorly optimized room to poorly placed speakers.
SINAD obsession is just the tip of the iceberg and yes, good enough, in many of the gear specification available to us for enjoying listening to music at home, will do just fine. Better specs just because it is technically available is not where audiophile should waste time and money.
We have had good enough specs in dacs and amps for decades.
An easy to use, automated, hand held, inexpensive tool that could read a room and automatically adjust with precision a full system in a listening space to achieve a spectacular 3D stereo image with a reasonably flat full range frequency response is something that would help the masses understand what good home sound reproduction can bring to the table. When we will be able to measure that and have a toll to give us a solution, then we will have something worth having and discussing.
SINAD in gears has been a fixed problem for almost 50 years.

 

[Ra1zel]

It encourages mediocracy in equipment design which results in no benefit to the consumer.

For me mediocracy in design is when the equipment fails in half a year, or when it's annoying to use, or sacrifices usability, compatibility and features to achieve high place in some arbitrary internet ranking. Maybe I'm weird.

 

[Ra1zel]

when we can get good ones with very high SINAD for no extra cost ?

Problem is sometimes you can't. You are not gonna get a 120 SINAD AVR because it doesn't exists. Surprisingly I've never seen anyone throwing his 30k$ Trinnov processor out the window because it makes mere 100dB SINAD not even enough for 16bit, so how can it be of use for MQA 18bit music enjoyers.

 

[Tangband]

For me mediocracy in design is when the equipment fails in half a year, or when it's annoying to use, or sacrifices usability, compatibility and features to achieve high place in some arbitrary internet ranking. Maybe I'm weird.

Today you can have both . My Aiyima a07 still works after two years of daily use .

 

[Tangband]

Problem is sometimes you can't. You are not gonna get a 120 SINAD AVR because it doesn't exists. Surprisingly I've never seen anyone throwing his 30k$ Trinnov processor out the window because it makes mere 100dB SINAD not even enough for 16bit, so how can it be of use for MQA 18bit music enjoyers.

When SMSL or Topping decides to make a homecinema processor with a SINAD of 110 dB for below 500 dollars theres no need for more expensive and worse products. If this would happen, every producer have to make even better measuring products, to justify a higher price.

 

[DonH56]

Disclaimer: Audio is not my day job. @amirm and others have much more insight into how the terms are used for audio (testing). I know the IEEE Standards but am less familiar with things like FTC and IHF standards for audio. I am also writing this off-the-cuff waiting for a test to finish so may have buggered up some of the explanation, sorry.

I hate to muddy the waters, but in the interest in full disclosure in the context of my experience want to add a footnote. While SINAD is usually the same as THD+N if THD+N is measured using an input signal (normal as I understand it), that is not always true.

SINAD = signal to noise and distortion ratio. The IEEE (among others) define this as essentially the signal level to everything else in the spectrum (or curve fit) with the signal applied.

THD+N = total harmonic distortion plus noise which is generally considered to be equivalent to SINAD.

I have seen SINAD not equal to THD+N in some cases, usually when THD is measured using a signal by measuring the sum of harmonic energy (typically ten harmonics) compared to the signal level, and noise (SNR) is measured with a shorted input (thus no signal) compared to a reference (typically maximum) output. In the latter case, non-harmonic spurs are ignored, and things like power supply and line noise that vary with signal level due to how and where they couple into the circuit may be neglected, leading to THD+N being better than SINAD. This is one reason SINAD was adopted for data converter (ADCs and DACs) testing since by (IEEE) definition it incorporates "everything" and directly correlates to ENOB (the effective number of bits, i.e. how many bits the converter is actually realizing). This prevents "gaming" the system.

Typical analyzers apply a signal and extract THD from the measurement, with the remainder treated as noise (N). In this case, the remainder includes everything else, and thus is actually SINAD as defined per the IEEE Standards. Noise falling into the harmonic frequencies is included in their measurements, thus when the harmonics fall to very low levels (at low signal levels) the "distortion" is dominated by noise, and THD gets worse (falls). SINAD/THD+N also falls due to higher noise with respect to the signal level, natch, and that is what most analyzers plot for THD+N (IME).

There are caveats (always) to assuming noise dominates at low signal levels, For example, an amplifier with high crossover distortion will exhibit worse THD (or THD+N) as the signal falls because crossover distortion is generally a fixed amount. Crossover distortion generally has a large component at 2*fin (2HD) plus higher-frequency spurs due to the "glitchy" wave shape of the distortion. Similarly, power-supply or line input leakage may introduce fixed levels of voltage (power, energy) independent of signal level, again leading to reduced SINAD at low signal levels. SINAD captures all of that. THD+N may capture all of that depending upon how the manufacturer (or tester) defines and measures "noise".

It is my understanding, perhaps false, that the AP and other audio analyzers measure SINAD to produce their THD+N plots -- but I could be wrong.

HTH - Don

 

[Vacceo]

When SMSL or Topping decides to make a homecinema processor with a SINAD of 110 dB for below 500 dollars theres no need for more expensive and worse products. If this would happen, every producer have to make even better measuring products, to justify a higher price.

Aren't they doing It because they may have codec licensing issues?

 

[GXAlan]

How is this for a simple summary…

From: https://www.audiosciencereview.com/...-thresholds-of-amp-and-dac-measurements.5734/

Recap of thresholds of audibility

Lenient
Dynamic range, linearity: 96 dB
THD, IMD: -66 dBFS / 0.05%
Noise: -85 dBFS / 0.005%
SINAD: 85 dB
Frequency response: ±0.5 dB
Channel balance: 1 dB

Strict
Dynamic range, linearity, SINAD: 120 dB
THD, IMD, noise, crosstalk, jitter: -120 dBFS / 0.0001%
Frequency response, channel balance: ±0.1 dB

1) Speakers have a SINAD of 70 or worse. Preference score has been shown to correlate strongly with actual preference. Thus, whenever you have the budget/space/opportunity, upgrading your speaker makes sense. We are nowhere near the limits of audibility.

2) -120 dB is the threshold of audibility. Thus, if you had 130 dB SINAD electronics across the entire chain, you would never be able to upgrade to a system that is more “transparent” to the source. At that point any change in sound that is heard would be the introduction of pleasant coloration. You also would need the frequency and channel matching to be 0.1 dB.

But if we talk about transparent to the source and use lenient thresholds…

1) FM has a dynamic range of 50 dB. Therefore, the most stringent threshold is noise at -85 db and THD at 0.05%. If your integrated amp/receiver had a dynamic range of 85 dB and THD of 0.05% at the volume you need, FM would be entirely transparent to the speaker. Thus, products with 95 dB SINAD would be plenty to ensure full transparency to FM. It’s overkill on the THD but gets you the noise level you want. If you are not sensitive to noise because your listing environment is not quite enough, the THD standards are the limiting factor and once you got to 76 dB SINAD, you probably wouldn’t be able to hear any further differences
with a FM source.

2) Vinyl has a theoretical dynamic range of 70 dB. Most commercial vinyl is below this quality. The same math as FM applies. An electronics chain with SINAD 95 would be fully transparent and if you were insensitive to noise, you’d just need 80 dB SINAD. Again, products with lower SINAD already have crossed the distortion threshold.

Since vinyl has the RIAA curve, errors less than +/- 0.5 dB would be hard to hear but if two phono stages has more than +/- 0.1 dB differences in their RIAA curve, it is possible to hear a difference. RIAA curves can vary greater than 0.5 dB.

The Cambridge Duo measured here is fully transparent to all LPs to the threshold of audibility (lenient). There may be detectable differences in the RIAA curve. In contrast, the Schiit Mani2 has mains noise that might be audible but ignoring noise, frequency response is the main issue. The ability to better match the impedance of the cartridge may give your cartridge a more accurate frequency response. In both cases, upgrading beyond these phono amps is probably not necessary?*
(Headroom to deal with pops may matter, right?)

3) CDs have a theoretical dynamic range of 96 dB. In this case the dynamic range constraint exceeds the other lenient thresholds. If you had a SINAD of 106 dB, your chain would entirely be transparent to the CD.

This then means that once you hit 106 dB SINAD, you are fully transparent. By corollary, a DAC with a SINAD of 90 dB would not be transparent…

Therefore, it is possible (though unlikely) that you would be able to hear a difference between two DACs, if one was 90 dB SINAD and the other was 106 dB SINAD and you were listening to a demo CD that took full advantage of the 16/44.1 format.

4) High res audio is less about the frequency response but more about dynamic range. DSD64 has about 120 dB dynamic range while 24-bit audio has 144 dB dynamic range.

A DSD64 or PCM recorder with SINAD > 106 would capture all of the data that might be encoded in a analog studio master to the thresholds of lenient audibility. A PCM24 recorder would fully be transparent to any analog master tape. Choosing 24/176 would allow the aliasing part to be pushed out far beyond the audible frequency range. The key to going with high resolution is ensuring that the roll off toward 20 kHz remained less than 0.5 dB for the relaxed standard and 0.1 dB for the strict standard.

5) If 130 dB SINAD is needed to be fully transparent to the the threshold of audibility by even the most skilled listeners, you would want 140 dB if applying EQ or room corrrection to stay at 130 across the entire chain.

6) Subjectivists have long said that amps have different sounds, yet ABX testing causes fatigue. Assuming that the amps are transparent in SINAD to the source being used to lenient standards of SINAD, the only explanation that isn’t placebo is:

In reactive loads like a loudspeaker, if the difference in two amps for frequency response exceeds 0.1 dB but is lower than 0.5 dB, the subjectivist could be accurate in that he or she may be able to hear the 0.4 dB over time but have trouble ABX’ing as it would be harder than the lenient thresholds.

7) 18-bits is what Amir reported Meridian stating as the very best dynamic range of real music. We get that this is a rare mastering and would not be representative of most music. That is 108 dB. Therefore hitting SINAD of 118 dB is fully transparent to all known music sampled by Meridian.

8) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4753356/#__ffn_sectitle

This paper suggests that music typically has 60dB dynamic range. Genres with a greater affinity for quiet parts, such as choir, opera, and piano, showed maximum dynamic ranges around 70dB, while “louder” rock, pop, and rap genres tended towards 60dB and below.

In this case, once you hit SINAD of 80 dB, you are probably transparent to most non-audiophile music except for noise which is masked when content is actually being played.

9) Putting this all together, SINAD is critical for identifying how to be 100% transparent to even the most strict standards (shoot for 130 dB) and it’s impressive to see the lack of correlation between performance and price.

The real world experience does support the idea that most DACS and AMPS sound similar since your SINAD threshold for most music is quite low. The audiophile experience does support the idea that audibility can be heard between two good products if one is truly transparent at strict 120 dB SINAD versus ones that are not.

It does seem that we see a lot of variability of amps in frequency response under load when seen at Stereophile. @amirm, what does it take to test amplifiers with a variety of simulated speaker loads to see the effect on frequency response?

 

[tmtomh]

I feel like there's a lot of folks talking past each other here. @jhaider - I totally get the point you're making: frequency response is the most important thing, and since the room is going to play a major role in that (at least from about 300Hz and down), DSP is more important than super-high SINAD for getting the best sound. This is similar to @Ra1zel 's point that dependability and longevity are incredibly important as well - the best performance is meaningless if the unit stops working.

But where this discussion is running off the rails a bit is that some folks are getting really adamant that these factors are somehow in competition with, or in a zero-sum game with, SINAD and similar performance measurements. As a reviewer, @amirm cannot feasibly evaluate long-term reliability. He can comment on build quality, and he can note if a particular maker has a track record of poor reliability - but beyond that he's testing gear for a short amount of time and it either works or doesn't work during that short time window in which he tests it.

Similarly, if a device offers DSP or doesn't offer DSP, that is easy to know - it's a binary question, and it makes no sense for Amir or any other reviewer to treat the absence of DSP the same way he would treat, say, poor IMD performance. While DSP impacts the performance of a given system in a given room, DSP is not a performance metric for an individual piece of gear - it's a feature. If you think it's an important or essential feature, great - then don't buy devices that lack DSP, just like you might not buy a device that lacks balanced inputs or a device without a remote control or a device that has a hard-to-read display or a difficult-to-use input-switching setup. But those are individual use-case issues. Those factors have nothing to do with whether or not devices should be tested for, and evaluated on, their SINAD performance and related metrics. This is especially true given that the use of DSP is one reason that we need a little padding/wiggle room in the form of high SINAD.

And let's not forget that we have yet to see any compelling case made for substituting SINAD with anything else - aside from proposals to split it into separate noise and THD measures (which is a distinction without a difference, and something Amir already mostly does anyway), the only alternative performance measurement proposal we have seen put into action is from @pma , the creator of this thread, who is on a crusade to try to "prove" that Class D amps are unstable and perform worse than measured, by running unfiltered high-frequency square waves into them. This is just absurd, and it needs to be called out as such.

As we always say to the subjectivists, we have to have an objective standard of reference, and noise (PSU hum, self-hiss), distortion (harmonic, IMD, jitter), linearity and frequency response are among those references. It would make no sense to downgrade any device that has a digital input but not DSP. If that were the case, then active speakers would have to get downgraded on the preference score, many DACs Amir has recommended within the lower tier of acceptability would have to change to Not Recommended, and so on. It would be mixing two different types of factors: feature/capability on the one hand, and measured performance on the other.

As a larger question of how we approach purchasing gear and setting up systems, yes, these are valid and important points: something like a MiniDSP Flex will likely produce a better real-world listening experience than a digital preamp that has better crosstalk and THD performance but lacks DSP. And something that will not break down will, as a practical matter, give you more enjoyment and much less aggravation than a device with slightly higher SINAD that might break after a few months or a couple of years.

But "let's not chase the last few dB of SINAD" is not the same thing as "SINAD is dumb - let's substitute my personal favorite feature for it."

 

[Cbdb2]

Here's Arny's post on the matter:



I believe this is the disc to which he refers.

He also wrote that prior to the BIS recording, "the high water mark for dynamic range" that he found was Ricky Lee Jones's first album, which came in at "about 73 dB dynamic range."

Its large but I'm not buying the 70dbs. Best example. Listen from 2 minutes.

 

[amirm]

For me mediocracy in design is when the equipment fails in half a year, or when it's annoying to use, or sacrifices usability, compatibility and features to achieve high place in some arbitrary internet ranking. Maybe I'm weird.

There is no evidence of connection between those. My $6000 mark levinson dac failed with a bad cap. Both of my $25000 ML amps have become flakey. And I have had a non chinese DAC fail. In testing I have blown amps from Japan, France, and Canada. None from China.

So yes, it is weird to claim causation without any data.

 

[amirm]

On DSP, I use powerful and transparent version in my software player. I have no use for it in the DAC. To wit, I don't use the one in my RME DAC.

To be sure, it is a highly valuable feature if you can't do the same in software. But if it is important to you, why don't you pick a player that can do it?

 

[Thomas_A]

There is no evidence of connection between those. My $6000 mark levinson dac failed with a bad cap. Both of my $25000 ML amps have become flakey. And I have had a non chinese DAC fail. In testing I have blown amps from Japan, France, and Canada. None from China.

So yes, it is weird to claim causation without any data.

Amir, was there any claims for brands in that post?