What matters about SINAD measurements

[amirm]

RCA guy thanks for talking common sense. If you can’t hear it, what’s the use for engineering excellence? Isn’t engineering supposed to accomplish a practical task ?

Tell me this: how long do you wash your hands and why?

 

[QAMatt]

Your point underscores the importance of independent measurements, such as by amirm and presumably others on this forum (I'm new so don't know you all yet.) Denon's claim as you report it is false, but I would hope that it is not true that "all the silicon vendors are spec'ing their max power." I traffic mostly not in consumer but professional audio, where such deceitful marketing (it's not engineering) would not be tolerated for long.

Yes, I think Denon is quoting their max power at 10% distortion based on Amir's measurement and I've attached TI's plot on the TPA3255. This is from the front page of the TPA3255 spec. They call this a "315W stereo...Ultra-HD" class D part. The 315W into 4 ohms is at 10%--they don't try to hide it at all.

Agree, too, that commercial stuff tends to spec THD at rated power. But they spec a bit beyond the knee too. A Crown CT3000 specs rated power at 0.35% THD. In the case of TI, their part would make 250W at that distortion level, but the 10% figure pushes it to 315W. In other words, they are only overselling it 25% because the curve is so steep in that region. In fact, both amps are in trouble that point.

 

[garbulky]

Tell me this: how long do you wash your hands and why?

20 seconds. Usually to clean them. Now my turn how long does it take you to eat cereal in the morning and why?

 

[RCAguy]

A few points. Whether someone can detect non-linear distortion is highly depends on perception qualities of the listener. In testing lossy compression for example, we know that trained listeners do a ton better than non-trained. So this is not a static number.

On why measure, in every case I show SINAD it is accompanied with FFT. It is the FFT that is instructive of whether the distortions can be audible. I realize people mostly focus on the SINAD since it is easier to digest but I do make an attempt to show the picture at the same time.

SINAD by itself then is best thought of engineering excellence when not focused on too accurately. I show the SINAD figures in four buckets for this reason. No question in my mind that a device in forth bucket is much worse design than those in first or even second bucket. In this manner, we use SINAD as a rough figure of merit without falling into the pitfall of using it too accurately when it doesn't have such predictive power.

The other reason to show SINAD is that the industry uses THD+N which is the same thing as SINAD. By measuring this we can tell how truthful the manufacturer is in their specifications.

Par1: Agree. Likely similar to AudSciRwv members, the trained subjects I have used for blind ANOVA analysis correlate well in their opinions. When they differ, it is often more because of acoustic variations.

Par2: Agree and I referred to your FFTs in my OP.

Par3&4: Agree. In the 1959s when I first got into audio, manufacturers were more engineers than marketers, so specs were more honest and transparent. Magazines published reviews with testing by Hirsch-Houck and other independent labs, not the dictates of their ad sales dept. Much of my systems integration work is T&M in shootouts among mfgrs, which I weed out by their misdeeds after returning their equipment (twice just last week).

 

[RCAguy]

It's about engineering excellence/competence; it's not about audibility.
If it was about audibility we wouldn't have much to discuss.
There are lots of sometimes interesting discussion on the merits of this design, or that. Most competently designed equipment reaches a performance standard that most people will find acceptable and indistinguishable from another such product.
By the way, It hasn't been my experience that professional audio is above all the bullshit.

...and yet here we are discussing audibility. And it is well that we do, because for all intents and purposes, hearing perception is the underlying application. Engineering - applied science - is often about knowing what you can get away with. From how much reinforced concrete is safe for a bridge v. how much is just wasted, to how much distoriuon to allow for a given application in audio, ranging from boom boxes, car audio, and low-end "hi-fi" with 10% distortion to audiophile and studio audio with <0.35% quoted above. The former operate to the right of the onset of clipping "knee" while the latter keep below onset - members of this forum should keep to the left. And while “prosumer” audio has strong marketing ties to consumer, truly professional, commercial, broadcast audio has resisted the BS, such as 1200w amplifiers measuring 0.04% or monitor speakers with <1% - unheard of a decade ago.

 

[amirm]

20 seconds. Usually to clean them.

How do you know that is sufficient to kill all the germs that may make you sick?

 

[RCAguy]

Yes, I think Denon is quoting their max power at 10% distortion based on Amir's measurement and I've attached TI's plot on the TPA3255. This is from the front page of the TPA3255 spec. They call this a "315W stereo...Ultra-HD" class D part. The 315W into 4 ohms is at 10%--they don't try to hide it at all.

Agree, too, that commercial stuff tends to spec THD at rated power. But they spec a bit beyond the knee too. A Crown CT3000 specs rated power at 0.35% THD. In the case of TI, their part would make 250W at that distortion level, but the 10% figure pushes it to 315W. In other words, they are only overselling it 25% because the curve is so steep in that region. In fact, both amps are in trouble that point.

I unconsciously turn a blind eye to inflated "marketing" data, so I forgot TI's rating of 315w at 10% for the otherwise marvelous TPA3255. They also give 1% ratings, but I give it 100w at 8ohms -0.004% or 200w at 4ohms <0.1% - nothing higher registers with me, nor would I recommend higher for you. These numbers are plenty for home use, even a small recording control room - but a stiff power supply is required, which is a lot more massive and costly than the amplifier!

 

[DonH56]

RCA guy thanks for talking common sense. If you can’t hear it, what’s the use for engineering excellence? Isn’t engineering supposed to accomplish a practical task ? So if one doesn’t hear a difference in real life listening, I would say the task is accomplished. I say this because I have listened to gear referred to as “poor engineering” which is completely inaudible to me. I wish people realized the very low impact a difference between 70/80 and 120 dB makes.

I do realize the impact, and that is why I set the bar at around 80 dB instead of 40 or 60 dB. I have heard what I (and some equally crazy others) considered excessive hiss if the amplifier's DNR is below 80 dB, particularly with highly-sensitive speakers. You obviously think my position bonkers and that's OK. Fortunately the companies I have worked for through the years do value engineering excellence, as long as it doesn't cost too much.

BTW, engineering is to make research practical in the sense that we implement the science to create a functional device (hopefully). That does not always mean the device is of practical use; sometimes it is to validate basic research, and practical uses come later. A lot of products started that way... A piece of wire tickling a crystal may not be considered of any practical use but there are a lot of radios around today that started from that rather impractical project.

But I do appreciate the reminder of just how low the general public holds engineers.

 

[Tks]

But I do appreciate the reminder of just how low the general public holds engineers.

I'd outsource all governments and institutions to engineers and scientists... And eventually even they would outstay their welcome at some point, AI would be better after that (with respect to impartiality and even the last hints of bias would be removed when *well-being* is the overarching goal of the AI's programming). After AI though... Yeah, that's beyond our lifetime.

I love engineers and scientists!

 

[RayDunzl]

I suspect musicians - those with direct experience with musical instruments - would more easily detect added distortions. However, they are rarely the ones to ask about quality reproduction, from what I've read.

I say added distortions, because music is layer upon layer of harmonic distortions. That's part of what gives different instruments their different sound.

Here is the electrical signal of a single note on an electric bass guitar - from the jack on the guitar to the ADC of a Behringer 202HD, to PC via USB, RTA displayed in REW.

No "effects" added.

Varying the pick technique will vary the relative levels of the harmonics - which changes the "tone" expressed.

The above was a pretty straightforward lick with a plastic pick, a few tries taken to get a nice series onscreen, without the second harmonic exceeding the fundamental (it is easy to do), so the distortion display would make sense, so the software would not be looking for the wrong sequence of frequencies for the "distortion" measures.

What's the SINAD there? About 5?

Note the "multitone" sequence, with little grass between the frequencies.

How many harmonics are present there? Thirty five or so?

The little spikes at 30 and 40 and 55 are the three lowest strings, not fully damped.

 

[garbulky]

How do you know that is sufficient to kill all the germs that may make you sick?

It's not usually for germs. I just do it till the greasy feeling goes away. The CDC reccomends 20 seconds. FWIW handwashing doesn't kill all germs. Handwashing reduces the bacterial load on your hand typically to about 8%. Plain water washing gets it down to 23%
Also handwashing is not a great protection against droplet and airborne infections.
(Now what about the cereal?)

 

[garbulky]

I do realize the impact, and that is why I set the bar at around 80 dB instead of 40 or 60 dB. I have heard what I (and some equally crazy others) considered excessive hiss if the amplifier's DNR is below 80 dB, particularly with highly-sensitive speakers. You obviously think my position bonkers and that's OK. Fortunately the companies I have worked for through the years do value engineering excellence, as long as it doesn't cost too much.

BTW, engineering is to make research practical in the sense that we implement the science to create a functional device (hopefully). That does not always mean the device is of practical use; sometimes it is to validate basic research, and practical uses come later. A lot of products started that way... A piece of wire tickling a crystal may not be considered of any practical use but there are a lot of radios around today that started from that rather impractical project.

But I do appreciate the reminder of just how low the general public holds engineers.

I hold engineers in high esteem! I don't think you are crazy at all. So sorry it came off that way. Our world would not be possible without engineers like you! I completely appreciate the tickling crystal idea! A lot of our world is built upon what appears to be mundane or boring phenomena that people have figured out how to exploit.

I too think 80 db is a great bar. But also, I have been unable to hear -60 (ish) db distortions in music listening. Now hiss on the other hand, I have heard as I think we all have. I do think thinking a product as being not good because it doesn't hit numbers a lot higher than 80 is pointless for practical listening purposes.

For instance if I had a product with audible distortion and that distortion is "mild hiss" or "barely perceptible treble roll off" - well that's not really a poorly engineered product. It may be a product with room for a small amount of improvement (if anybody cares). I'm talking about the audible distortions as evidenced by measurements here not -95 db distortions or whatever.

I am of the opinion that (if I'm not imagining it), any differences in electronics I hear must be in the audible range in some way which means they are probably not buried deep down in the signal.

 

[garbulky]

I suspect musicians - those with direct experience with musical instruments - would more easily detect added distortions. However, they are rarely the ones to ask about quality reproduction, from what I've read.

I say added distortions, because music is layer upon layer of harmonic distortions. That's part of what gives different instruments their different sound.

Here is the electrical signal of a single note on an electric bass guitar - from the jack on the guitar to the ADC of a Behringer 202HD, to PC via USB, RTA displayed in REW.

No "effects" added.

View attachment 25236

Varying the pick technique will vary the relative levels of the harmonics - which changes the "tone" expressed.

The above was a pretty straightforward lick with a plastic pick, a few tries taken to get a nice series onscreen, without the second harmonic exceeding the fundamental (it is easy to do), so the distortion display would make sense, so the software would not be looking for the wrong sequence of frequencies for the "distortion" measures.

What's the SINAD there? About 5?

Note the "multitone" sequence, with little grass between the frequencies.

How many harmonics are present there? Thirty five or so?

The little spikes at 30 and 40 and 55 are the three lowest strings, not fully damped.

Do you play bass? My acoustic electric Ibanez bass's neck snapped

 

[amirm]

It's not usually for germs. I just do it till the greasy feeling goes away. The CDC reccomends 20 seconds. FWIW handwashing doesn't kill all germs. Handwashing reduces the bacterial load on your hand typically to about 8%. Plain water washing gets it down to 23%
Also handwashing is not a great protection against droplet and airborne infections.
(Now what about the cereal?)

The 20 second is CDC's SINAD. They don't know how many germs are on my hands and how well I wash during that 20 seconds. They picked a number that they hope is on the safe side.

Circling back, the idea of how long to wash your hands should be so that you don't get sick from said germs. No one can give you that assurance as you imply in your answer. Same with SINAD. It is one metric out of many and has certain predictive strength just like the 20 seconds. I would love to be able to tell you what is transparent to you for music you listen to but I can't do that with low values of SINAD with high assurance.

 

[amirm]

Now my turn how long does it take you to eat cereal in the morning and why?

I don't eat cereal in the morning or otherwise.

 

[DonH56]

I hold engineers in high esteem! I don't think you are crazy at all. So sorry it came off that way. Our world would not be possible without engineers like you! I completely appreciate the tickling crystal idea! A lot of our world is built upon what appears to be mundane or boring phenomena that people have figured out how to exploit.

I too think 80 db is a great bar. But also, I have been unable to hear -60 (ish) db distortions in music listening. Now hiss on the other hand, I have heard as I think we all have. I do think thinking a product as being not good because it doesn't hit numbers a lot higher than 80 is pointless for practical listening purposes.

For instance if I had a product with audible distortion and that distortion is "mild hiss" or "barely perceptible treble roll off" - well that's not really a poorly engineered product. It may be a product with room for a small amount of improvement (if anybody cares). I'm talking about the audible distortions as evidenced by measurements here not -95 db distortions or whatever.

I am of the opinion that (if I'm not imagining it), any differences in electronics I hear must be in the audible range in some way which means they are probably not buried deep down in the signal.

Ah, a misunderstanding, my apologies. Too little sleep lately...

The thread is about SINAD, so noise and distortion are included. Noise we've discussed. Distortion I find a little trickier to define. Like most folk, I can tell if distortion around 1% (-40 dB) is added to music, but frankly that is an A-B comparison. Given the complexity of most music I might not tell at -40 dB for some music or complex passages, but for a solo flute (about the closest to a pure tone) or other solo instrument I know well I might detect around -60 dB (0.1%). At least so I have found in my brief recent excursions into adding distortion to sounds.

Another thing to consider is that IMD (intermodulation distortion) happens whenever you have harmonic distortion, is not harmonically related to the signal so is easier to hear (i.e. is more obnoxious), and for e.g. two tones of equal amplitude whose sum matches that of a single tone's peak amplitude the IMD will be about 9.542 dB higher in amplitude than HD (harmonic distortion) thus roughly twice as loud. I find it much easier to detect 0.1% IMD than the same level of HD even on a pure tone.

Another factor in the desire for very low noise levels is the way it is measured and referenced on the datasheet. For a power amplifier, it is typically referenced to full-scale output, thus distortion normally falls as the signal level drops, but the noise floor is relatively constant. That means if the SINAD is dominated by noise, typical at low output, then if you are listening at 50 dB below the max output you are getting much closer to the noise floor. My system can put out around 105 dB or so from a pair of speakers but I usually listen around 65~70 dB (though often louder for music). So if the amp only has 80 dB SNR (or SINAD limited by noise) relative to maximum (full-scale) output the noise floor is only about 50 dB down. Different speakers, different amps, and the result may be better or worse. 100 dB from a high-power amp driving sensitive horn speakers may not be good enough while 60 dB driving insensitive speakers might be fine.

Finally, you have to watch how noise is weighted when it is reported. A weighting mimics our hearing but also weights out LF and HF noise. Broadband results often yield 3 to 10 dB worse SNR. I tend to look more at wideband noise specs, especially when the noise gets dominated by HF hiss (alas, not much a problem for me anymore) or LF power spurs (typically 50/60 and 100/120 Hz depending upon where you live). Those power supply spurs can be obnoxious and thus are one thing I pay attention to in Amir's results.

IME/IMO/HTH/etc. - Don

 

[DonH56]

I suspect musicians - those with direct experience with musical instruments - would more easily detect added distortions. However, they are rarely the ones to ask about quality reproduction, from what I've read.

Speaking as an engineer and musician, I have long said my musician friends are much more likely to comment about the poor intonation by not dropping the third of the chord, poor timing, or lack of dynamics and so forth than anything the amp is doing. There are certainly exceptions, but my long-ago observation (original so far as I know) is that audiophiles tend to listen to the gear; musicians tend to listen to the music. Not all the time, but in general, and IME. - Don

 

[RayDunzl]

Do you play bass?

I have one, and pick it up now and then to remind myself how lame I am with it, yes...

Ibanez 5 String Bass Guitar Right Handed, Weathered Black GSR205BWK

 

[garbulky]

Ah, a misunderstanding, my apologies. Too little sleep lately...

The thread is about SINAD, so noise and distortion are included. Noise we've discussed. Distortion I find a little trickier to define. Like most folk, I can tell if distortion around 1% (-40 dB) is added to music, but frankly that is an A-B comparison. Given the complexity of most music I might not tell at -40 dB for some music or complex passages, but for a solo flute (about the closest to a pure tone) or other solo instrument I know well I might detect around -60 dB (0.1%). At least so I have found in my brief recent excursions into adding distortion to sounds.

Another thing to consider is that IMD (intermodulation distortion) happens whenever you have harmonic distortion, is not harmonically related to the signal so is easier to hear (i.e. is more obnoxious), and for e.g. two tones of equal amplitude whose sum matches that of a single tone's peak amplitude the IMD will be about 9.542 dB higher in amplitude than HD (harmonic distortion) thus roughly twice as loud. I find it much easier to detect 0.1% IMD than the same level of HD even on a pure tone.

Another factor in the desire for very low noise levels is the way it is measured and referenced on the datasheet. For a power amplifier, it is typically referenced to full-scale output, thus distortion normally falls as the signal level drops, but the noise floor is relatively constant. That means if the SINAD is dominated by noise, typical at low output, then if you are listening at 50 dB below the max output you are getting much closer to the noise floor. My system can put out around 105 dB or so from a pair of speakers but I usually listen around 65~70 dB (though often louder for music). So if the amp only has 80 dB SNR (or SINAD limited by noise) relative to maximum (full-scale) output the noise floor is only about 50 dB down. Different speakers, different amps, and the result may be better or worse. 100 dB from a high-power amp driving sensitive horn speakers may not be good enough while 60 dB driving insensitive speakers might be fine.

Finally, you have to watch how noise is weighted when it is reported. A weighting mimics our hearing but also weights out LF and HF noise. Broadband results often yield 3 to 10 dB worse SNR. I tend to look more at wideband noise specs, especially when the noise gets dominated by HF hiss (alas, not much a problem for me anymore) or LF power spurs (typically 50/60 and 100/120 Hz depending upon where you live). Those power supply spurs can be obnoxious and thus are one thing I pay attention to in Amir's results.

IME/IMO/HTH/etc. - Don

Well explained. Thanks for the input

 

[Shadrach]

RCA guy thanks for talking common sense. If you can’t hear it, what’s the use for engineering excellence? Isn’t engineering supposed to accomplish a practical task ? So if one doesn’t hear a difference in real life listening, I would say the task is accomplished. I say this because I have listened to gear referred to as “poor engineering” which is completely inaudible to me. I wish people realized the very low impact a difference between 70/80 and 120 dB makes.

...and yet here we are discussing audibility. And it is well that we do, because for all intents and purposes, hearing perception is the underlying application. Engineering - applied science - is often about knowing what you can get away with. From how much reinforced concrete is safe for a bridge v. how much is just wasted, to how much distoriuon to allow for a given application in audio, ranging from boom boxes, car audio, and low-end "hi-fi" with 10% distortion to audiophile and studio audio with <0.35% quoted above. The former operate to the right of the onset of clipping "knee" while the latter keep below onset - members of this forum should keep to the left. And while “prosumer” audio has strong marketing ties to consumer, truly professional, commercial, broadcast audio has resisted the BS, such as 1200w amplifiers measuring 0.04% or monitor speakers with <1% - unheard of a decade ago.

I may not have made myself clear.
The data that amirm presents is about measured performance. One product often compared to another. There isn't much to discuss unless one believes the data is wrong. Based on that data amirm states that one product performs to a higher specification than another.
At this point the reader has then to decide what level of performance is important to them.
One can always apply the' fit for purpose' argument to products. The problem is often each application requires differing levels.
We do have a measure of what is and what isn't audible. A brief tour of the various 'hi fi ' boards should show that this measure of what is audible isn't accepted by many.
You may believe and possibly even have some data to present, that backs up your view that a particular set of measurements is 'good enough'.
However, while it may be true that under test conditions one can demonstrate that the 'good enough' standard is indistinguishable from a higher standard, if one applied that to say automobiles then we would all be driving Volkswagen beetles equipped with speed limiters.
Thankfully, my scientific/engineering background isn't in domestic audio, or even 'professional' audio. I was an Avionics engineer working with acoustics for military applications so much of the nonsense that gets applied to 'recreational' audio didn't get established in my views.
For me domestic audio is a hobby. I build much of my own equipment, used to belong to an audio club where the hobby aspect dominated.
I rather like amirm's measurements. They allow me to make a better informed choice should I wish to purchase a product based on a data set rather than the many opinions of audiophile experts, be they domestic or 'professional'.