• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Why SINAD is not important (article)

C0B-1D3-9

Member
Joined
Jan 19, 2021
Messages
13
Likes
10
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.
Or perhaps it is time to shift the focal point and make other measurements more apparent on the index.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,385
Location
Seattle Area
For the record, I didn't write the headline - my original draft just said "SINAD is bad", which isn't a very catchy title, but is a headline I would fight in support of.
Your name is on the article. If you disagree with it, then you should have asked them to remove your name.

The catchy title was intended to declare war on everything we do here. And per below that is exactly what is happening. Watch Andrew's video. Not one word of appreciation for measurements in general and even THD+N in the specific being a million miles better nothing which is the world he lives in (outside of headphones).

Edit: Also, it's a bit vexing to be accused of bad faith both while I'm actively defending this community (and yourself) in the (predictably vile) Youtube comments on the video, and while you appear to be trying to "gotcha" me about proving a negative.
I had not read the comments. That they are running with what I feared, makes me feel worse about this, not better. Why allow yourself to be used this way? Andrew could have never pulled this by himself.

Anyway, good news is that our mission continues to make companies produce better engineered products. SINAD graphing has had a profound impact on the industry as far as reduction of noise and distortion in DACs and amplifiers with zero cost to us. The influence is now continuing to AV processors and receivers. Speaking of which, I have an AVP review to finish...
 

Mad_Economist

Addicted to Fun and Learning
Audio Company
Joined
Nov 29, 2017
Messages
543
Likes
1,618
Why allow yourself to be used this way? Andrew could have never pulled this by himself.
I am not being used - I have repeatedly expressed my stance against these metrics to you, across years. I continue to hold those stances, and stand by the article. If you would like to alter those stances, you have the absolute power to do so by pointing me in the direction of whatever body of work is influencing your differing stances. I promise I will read every word, and update the article properly as appropriate, and throw an almighty tantrum to Andrew if the video isn't suitably errata'd as well.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
My problem with articles like this is that they don't solve the problem. It's not the matter of intention.

First is that there are many profound indications of a single SINAD number(let alone full FFT of 1khz tone in the dashboard) in different components in different systems. It's not a simple number for audibility. A very straightforward problem is to use a dac as preamp feeding the power amp.

By using logic what is important instead of SINAD. Not mentioned here. Does it mean nothing matters as long as -80dB THD+N is met or does it mean one should use their ears to decide? Neither is correct by the way.

The issue is clearly most of the time is mixed of half science and half ignorance/misunderstanding. This actually can be more harmful in some way than blatant audiofoolery. Well we can't avoid it in some way because it's natural. When we know a little bit of something we like to express it. But while we are here someone had to clear it up like exactly what happened.

But on the other hand, if someone actually reads the review word by word. I think it's clearly different from just looking at the SINAD number. Interpretation needs experience and knowledge. It's the way for headphones measurements and it's the way for electronics. SINAD is just a measurement metric, It's how you use it and interpret it matters. And of course other measurements matter too.
 

Mad_Economist

Addicted to Fun and Learning
Audio Company
Joined
Nov 29, 2017
Messages
543
Likes
1,618
I generally think that the failure to assert a better alternative is a poor criticism of a criticism, in and of itself - one does not need to boost the signal to improve the SNR, after all, and it was difficult enough trimming this piece down to a size anyone would read without including prescriptions for different measures of nonlinearity.

That said, in spite of my prior allusion to Vanderkooy, I'm very much hoping to give the topic of better metrics for nonlinearity audibility attention in print in the future - with a specific mind to measures that can be applied to headphones, since there, we are not so blessed that we can aim for -130dB or lower ;) If you - or anyone - has input, I'm all ears, while I've followed the PEAQ stuff a bit, I've spent more time on the "negative" side (e.g. the aforementioned body of work demonstrating the relatively poor differentiation of distortion, and the poor performance of the THD and IMD metrics).
 

Asinus

Member
Joined
Mar 30, 2020
Messages
75
Likes
90
IMO as long as there is no counter-example where a device (not a synthetically generated corner case) with higher SINAD is demonstrably worse than one with lower SINAD this call for audibility metrics is just asking to be graded on a curve because you are failing the exam.

Perhaps it would be more important if we were demonstrating that paying $10K for 115dB SINAD was not worth it when a $500 device and 100dB SINAD would not have any audible difference, but when $100 devices routinely reach 100dB this call for better metrics seems to me like excuses for lesser designs. We should not give a free pass for more noise and distortion, even if inaudible, if cheaper alternatives measure better.

When a new metric is available and demonstrably better than the SINAD go ahead and promote it, otherwise disqualifying an existing metric seems disingenuous.
 

HiFidFan

Addicted to Fun and Learning
Forum Donor
Joined
Feb 22, 2021
Messages
723
Likes
906
Location
U.S.A
I generally think that the failure to assert a better alternative is a poor criticism of a criticism, in and of itself - one does not need to boost the signal to improve the SNR, after all, and it was difficult enough trimming this piece down to a size anyone would read without including prescriptions for different measures of nonlinearity.

Does this mean there will be a follow up article from you and Andrew that discusses an alternative metric to SINAD?
 

Chocomel

Active Member
Joined
Feb 18, 2019
Messages
106
Likes
322
We should not give a free pass for more noise and distortion, even if inaudible, if cheaper alternatives measure better.

For sure, if there's a cheaper equivalent we should recommend that. The case i see more often is giving a free pass to expensive products for inaudible better measurements. Or where perfectly fine products are chastised for inaudible distortion, even if they have other areas of merit that make them worthwhile.


When a new metric is available and demonstrably better than the SINAD go ahead and promote it, otherwise disqualifying an existing metric seems disingenuous.
There are a number of them mentioned in this thread.
 

Mad_Economist

Addicted to Fun and Learning
Audio Company
Joined
Nov 29, 2017
Messages
543
Likes
1,618
When a new metric is available and demonstrably better than the SINAD go ahead and promote it, otherwise disqualifying an existing metric seems disingenuous.
I'll ask the same thing here I've asked elsewhere: should we apply this same logic to another measure that can show audible impacts, but often does not, Zout? It's pretty trivial that a 100 ohm Zout amplifier - like a 20dB SINAD amplifier - can sound audibly different from something which approximates a better voltage source. However, we typically stop caring about Zout when it passes some arbitrary threshold of potential frequency response adjustment (for some, a damping factor of 8, for others, a higher ratio). Why is that? Does the market not deserve .001 ohm Zout amplifiers, as well? Of course, the audible impact there would be nonexistent, but it is clearly closer to a voltage source than a design with a mere .1 ohm Zout.


Does this mean there will be a follow up article from you and Andrew that discusses an alternative metric to SINAD?
We will do an article that recommends aggregating noise and distortion over my dead body, but I'm definitely hoping to do a piece on some of the more-subjective-results-correlated alternatives for distortion and...I mean, there really isn't anything wrong with SNR.
 

Chocomel

Active Member
Joined
Feb 18, 2019
Messages
106
Likes
322
And yet, the article and the video are not titled X is a more meaningful metric, but SINAD is not important.

Because other metrics are not the focus of it. But as mentioned before in the thread, considering we can just look at the graphs themselves the single SINAD figure is still not important.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
SINAD, as an indicator of sound quality is not a good indicator.
I have explained this quite a few times already.
In that aspect I agree with the article.

However, Amir also measures SINAD at 50mV and not just 2V or 4V (balanced) and speaker amps at 5W.
Amir also publishes the other aspects that all together do say something about the performance of the device.
Sometimes Amir touches shortly on these aspects when he feels it is important to.

My 'complaint' would be that to really understand all the plots and the relation of them and what all the numbers combined can tell is that the reviews are basically only really telling the story to technically minded people and even then mostly the ones that know about measurements.

I would dare to say that 99.9% of the casual readers do not have enough knowledge to really understand the reviews.
Amirs videos do help here.
Some of the ASR members fully grasp all the plots and a lot of members understand some of it to certain degrees.

For outsiders and ASR haters (there are plenty) the article is a bit misleading in the sense that it appears to be a 'complaint' to ASR.
To me it isn't as most aspects are clearly addressed and explained even with some audio (which is more enlightening than numbers for most).

Is SINAD pointless ?
The answer is no. Everything between 60 an 130 can be O.K. depending on circumstances and aspects.

Amir has a point though that a SINAD above 110 pretty much guarantees that there is not much to worry for folks that don't (fully) understand what all the plots and numbers combined can tell you.

No... it isn't the only number that's important and SINAD should also not be considered as the 'all important' number.

The article TITLE (haven't watched the video yet) seems more like a complaint towards ASR (the SINAD table appearing to be all important) where the article itself is enlightening and expresses the very same thing I have tried to explain many times here. It is educational and had hoped Amir had a sticky (or video) somewhere explaining the relation to audio with examples.

In short: Good and informative well written article with an unfortunate title that seems to me like click bate.
 

Asinus

Member
Joined
Mar 30, 2020
Messages
75
Likes
90
Because other metrics are not the focus of it. But as mentioned before in the thread, considering we can just look at the graphs themselves the single SINAD figure is still not important.
Comparing graphs is more problematic since differences in scale and range can cause more confusion.

I'll ask the same thing here I've asked elsewhere: should we apply this same logic to another measure that can show audible impacts, but often does not, Zout?
I don't have any issue if you use the same logic as long as you explain the relevance of a Zout comparison. However we are looking for transparency in our playback chain, so max amplitude/power, noise floor and lack of distortion are more important figures, coincidentally SINAD summarizes those in a single number even if it is not perfect and sometimes seems unforgiving.
 

Mad_Economist

Addicted to Fun and Learning
Audio Company
Joined
Nov 29, 2017
Messages
543
Likes
1,618
However, Amir also measures SINAD at 50mV and not just 2V or 4V (balanced) and speaker amps at 5W.
Amir also publishes the other aspects that all together do say something about the performance of the device.
Gotta say, I really do appreciate the 50mV SNR measurement - it may technically be redundant with a high-level SNR measurement, but there are cases where noise residual changes somewhat with output, and it's just a useful reference in general given that it's closer to a "normal" headphone level.

Appreciate your thoughts, Frans!

I don't have any issue if you use the same logic as long as you explain the relevance of a Zout comparison. However we are looking for transparency in our playback chain, so max amplitude/power, noise floor and lack of distortion are more important figures, coincidentally SINAD summarizes those in a single number even if it is not perfect and sometimes seems unforgiving.

Are they, though? Bearing in mind that Zout can directly impact frequency response, the single most important factor? Of course, ultra-low-Zout isn't changing your FR in any way that's audible...but, per the body of the lit that I have cited (which, I note, I remain open to adding new information to), nor is -80dB+THD impacting your distortion in an audible way.

Of course, -80dB noise very likely is troublesome, leading us to the "type I error" problem I was referencing earlier...
 
Last edited:

AndrewC

Member
Joined
Jun 24, 2016
Messages
25
Likes
59
Location
Singapore
I don’t see a problem with using SINAD as a benchmark to compare the engineering quality of different products. I can’t recall seeing any reviews here that suggest SINAD is a measure of audible sound quality.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,278
Location
Gold Coast, Queensland, Australia
and...I mean, there really isn't anything wrong with SNR.

Well, yes and no. SNR is deceptive in itself if people just look at the biggest number as being the best, like they do with SINAD.

More voltage/output swing and your S/N ratio goes up all other things being equal. You can have a noisier (residual) amplifier or preamp with a much higher output swing and it will achieve an SNR greater than an amplifier with little residual noise and a lower swing. Not really fair.

That's why the S/N (SNR) at a fixed low level with HPAs is so useful. S/N on power amplifiers is always referenced to full rated power, so a 500W@8R amp will get 10dB more S/N by default over a 50W@8R amplifier. There is a high possibility the 50W amplifier also has less residual noise, but the big amp wins the SNR numbers game. SNR must be referenced to an rated output value and often, manufacturers are not specific or deliberately play games to get a better number.

Referenced to 1W is a killer for big amps as they often have a decent amount of PSU ripple due to massive transformers close to the front ends etc, so A-WTD tends to get published or a 400Hz HPF to get rid of the mains harmonics.

There was a push to rate amplifiers in dBW at one point. It sort of fell by the wayside, even though it wasn't a bad idea, just that people didn't get it and wanted to see "raw power" numbers in spec sheets.
 
Last edited:

Mad_Economist

Addicted to Fun and Learning
Audio Company
Joined
Nov 29, 2017
Messages
543
Likes
1,618
Well, yes and no. SNR is deceptive in itself if people just look at the biggest number as being the best, like they do with SINAD.

More voltage/output swing and your S/N ratio goes up all other things being equal. You can have a noiser (residual) amplifier or preamp with a much higher output swing and it will achieve an SNR greater than an amplifier with little residual noise and a lower swing. Not really fair.

That's why the S/N (SNR) at a fixed low level with HPAs is so useful. S/N on power amplifiers is always referenced to full rated power, so a 500W@8R amp will get 10dB more S/N by default over a 50W@8R amplifier. There is a high possibility the 50W amplifier also has less residual noise, but the big amp wins the SNR numbers game. SNR must be referenced to an rated output value and often, manufacturers are not specific or deliberately play games to get a better number.
That's a very good clarification - my apologies for being sloppy there, I should have said "SNR referenced to a consistent output voltage". Somehow the gamesmanship of SNR with reference levels by marketing folks slipped my mind for a (blissful) moment.
 

JJB70

Major Contributor
Forum Donor
Joined
Aug 17, 2018
Messages
2,905
Likes
6,148
Location
Singapore
cf. e.g., IQ.

(edit: or standing start quarter-mile time for automobiles)

The problem with surrogate markers is it's often easy to game the system.
For some reason, I was thinkin' about VW Diesels when I typed that last sentence.

VW went a few steps further than gaming the system and illegally used defeat (cheat) devices to get engines through testing needed to revieve the necessary certification. But the point is apt.
 

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
3,422
Likes
2,407
Location
Sweden
Is there any way to compare this SINAD to what Stereophile reports, i.e. SNR for 2.83 V, 8 ohm, audio band bandwidth and A-weighted?
 
Top Bottom