• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Why SINAD is not important (article)

SIY

Master Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
6,822
Likes
14,763
Location
Alfred, NY
It's a generally very good article. One caveat: when looking at audibility of distortion, we need to be careful not to use something very non-physical as the metric. For example, it's easy to hear -40dB of 9th harmonic on a pure tone, but I haven't encountered a device that has 9th harmonic only. Almost every real-world device is dominated by lower order distortion, odd or even depending on the topology. If there's higher order, it's mixed in with MUCH larger quantities of lower order.
 

Mad_Economist

Senior Member
Technical Expert
Joined
Nov 29, 2017
Messages
425
Likes
1,193
It's a generally very good article. One caveat: when looking at audibility of distortion, we need to be careful not to use something very non-physical as the metric. For example, it's easy to hear -40dB of 9th harmonic on a pure tone, but I haven't encountered a device that has 9th harmonic only. Almost every real-world device is dominated by lower order distortion, odd or even depending on the topology. If there's higher order, it's mixed in with MUCH larger quantities of lower order.
Extremely good point - an earlier revision of the article went into more depth about nonlinearity, but it got edited down for "readability". FWIW, there will be a companion post coming using @pkane's distort software to mimic the nonlinearities of a few real amplifiers, one of which was the inspiration for the audible H3/H5 cases (as the lower harmonics were masked with pure tones, but the higher ones were not). My editor asked me to keep the FFTs and the numbers as clean and simple as I could.
 

SIY

Master Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
6,822
Likes
14,763
Location
Alfred, NY
Extremely good point - an earlier revision of the article went into more depth about nonlinearity, but it got edited down for "readability". FWIW, there will be a companion post coming using @pkane's distort software to mimic the nonlinearities of a few real amplifiers, one of which was the inspiration for the audible H3/H5 cases (as the lower harmonics were masked with pure tones, but the higher ones were not). My editor asked me to keep the FFTs and the numbers as clean and simple as I could.
If I had known it was yours, I would have tagged you. :cool: Nice work, and it does hit points about the vain desire for single number metrics. They don't work, and that main point is made very well.
 

Chocomel

Member
Joined
Feb 18, 2019
Messages
98
Likes
276
Makes sense. But if you want a single score to rank devices, what do you use?

There's a bunch of metrics that have significantly better correlation with audio quality, for example the GedLee metric i wrote a small bit about here in a similar post https://www.reddit.com/r/headphones/comments/ltwyz9
This is not the only such metric in existence either.


Edit: i would i agree with the sentiment others have mentioned here that maybe we shouldn't want a single metric, but if we wanted to, there's very much better ones.
 
Last edited:

blueone

Addicted to Fun and Learning
Forum Donor
Joined
May 11, 2019
Messages
566
Likes
762
Location
USA
Single score to rank devices is meaningless and brings useless results. Deep analysis is always needed.

I agree. Is there even one figure of merit, in any field, that is really descriptive to an expert?
 

Mad_Economist

Senior Member
Technical Expert
Joined
Nov 29, 2017
Messages
425
Likes
1,193
SINAD, IMD, etc are intended to reflect electronic characteristics (duh), nothing more.

It's an abuse to map those measures directly into the space of perceptual relevance.
At least from a historical standpoint, I don't agree with this - in the context of audio electronics, attempts to correlate nonlinearity and distortion are about as old as these primeval nonlinearity measurements themselves. E.g.
1629566621748.png

Czerwinski et al 2001 has a lovely history of nonlinearity (obviously cutting off at publication date) that stretches back to the very start.

If you mean this in the modern day, I suppose that might be the case, but that raises the question of why we're intending to quantify electronic characteristics with coarse-grained measures like SINAD instead of looking closely at the nonlinear transfer functions of the systems in question...
 

tomtoo

Major Contributor
Joined
Nov 20, 2019
Messages
2,065
Likes
2,104
Location
Germany
SINAD, IMD, etc are intended to reflect electronic characteristics (duh), nothing more.

It's an abuse to map those measures directly into the space of perceptual relevance.

Haha sure is SINAD perceptual relevant.
Nice joke my friend to say its not.
 

JJB70

Major Contributor
Forum Donor
Joined
Aug 17, 2018
Messages
2,436
Likes
5,207
Location
Singapore
I agree. Is there even one figure of merit, in any field, that is really descriptive to an expert?

Marketing people love headline "more/less is better" figures they can sell to purchasers. More pixels is better, lighter bicycles are better, more watts is better etc. It's not that some of these things aren't true per se but rather that they are one part of a much more complex picture and in some use cases may be far from the most relevant indicator. The danger with any of these metrics is that designs are over-optimized for a particular metric which can result in a worse product. I see this a lot with an energy efficiency metric in my field
 

voodooless

Major Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
2,293
Likes
3,418
Location
Netherlands
There is a reason Amir does not only measure SINAD.. the conclusion is rather obvious.

In any case, when we reach SINAD levels this vanishingly low, it’s getting rather unimportant if the major contributor is noise or distortion.. neither will be audible.
 

tomtoo

Major Contributor
Joined
Nov 20, 2019
Messages
2,065
Likes
2,104
Location
Germany
The mixture between s/n ratio and thd is inherently problematic. But that means not that it says nothing about perception.
It says a lot.
 

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
5,668
Likes
10,119
Location
Monument, CO
SINAD is a nice single figure of merit. IMO the irrelevance to listening is partly (largely?) because the vast majority of components these days have distortion and noise well below the audible threshold. Back when 1% THD was high fidelity, and many amps listed 10% THD specs, then it was much more useful since SINAD was at the edge of audibility (or above). These days, not so much.

But of course knowing the spectral content makes a big difference, along with all the other metrics. But I tend to agree with those above that it is more of academic (or engineering, in my case) interest now.
 
Top Bottom