SINAD, IMD, Dynamic Range, Frequency Repsonse

[wgb113]

These are several measurements @amirm does for his site but what are the cumulative effects and what is the end result.

Take noise for instance - does the worst measuring piece in your system serve as the limitation of the entire system or does each piece contribute to an aggregate whole?

With something like dynamic range - how much emphasis should we place on these measurements if the average noise level of our room is 40dB? Would the natural noise level of the room place more or less importance on this measurement?

For frequency response measurements I find the predicted in room response curious. Since our room are all so different how useful is this measurement? What are the dimensions of this hypothetical room?

 

[Plcamp]

On noise, if the worst device in your system is significantly higher noise than other devices, then that number would dominate. If you had two devices with same level noise they could add together and increase overall noise assuming their noise spectrums are similar.

 

[abdo123]

if the average noise level of our room is 40dB

context is important, 40 dB ambient noise at 100 Hz is not the same as 1KHz.

usually the room is dead quiet above 1KHz. if someone's average noise level is 40 dB in the 1KHz to 5KHz range I would pity them lol.

 

[DVDdoug]

Take noise for instance - does the worst measuring piece in your system serve as the limitation of the entire system or does each piece contribute to an aggregate whole?

Yes, noise (and frequency response variations and distortion) accumulates. Usually one of the noise sources is a lot worse than the others so the so the lower-level noise is masked (drowned-out) and doesn't contribute significantly. But if the the noises are different, such as hum & hiss they won't mask very well.

The same goes with distortion... Usually the speakers/headphones contribute most of the distortion and the other components aren't significant. (Unless you drive the amp into distortion or something is seriously wrong.)

It's possible that two frequency response variations can cancel-out but Murphy's Law and nature (entropy?) tend to work against you. And of course, the speakers or headphones are the worst contributors.

With something like dynamic range - how much emphasis should we place on these measurements if the average noise level of our room is 40dB? Would the natural noise level of the room place more or less importance on this measurement?

Again, the room noise tends to mask any electronic noise, depending on the nature of the noise. But the direction of the noise also makes a difference so you might hear 30db SPL noise from the speaker even with 40dB SPL acoustic noise.


...I'll let someone else answer the predicted room response.

 

[FrantzM]

Hi

Very interesting thread/post. I have been asking myself similar questions...

We all vie/dream/try to achieve 110 dB of SINAD... meanwhile, many find LP satisfying. LP barely reaches 70 dB under ideal conditions 50 dB is the norm and commercially available LPs, even during the so-called golden years could not achieve 40 dB figures.
I am OK and the engineer in me is quite excited by his very high SINAD headphones system ( JDS Labs Atom + Topping D10 DAC + PC as source) but is even more entertained/pleased/satisfied by my JBL LSR 308 + Denon AVR-3400 system .. I am sure that system does not break 80 dB of SINAD, if that much...
IMD.. I sincerely, am asking what level of IMD is objectionable... 1%, 3 %? ... Most people even those on this board routinely tolerate 10% distortion in the low frequencies ... under 200 Hz ... What are the objectionable levels of THD or IMD? The thresholds?
Dynamic Range ... related to SINAD ... that bears the question can most people achieve 90 dB of Dynamic range? ... even if such recordings were available ( I don't know any...) ... Keep in mind that for most living rooms, 30 dB is quiet, very quiet... Taking 30 dB as the noise floor ... 90 dB of dynamic range would be 120 dB at the listening position ... Very few systems are capable of such within the audio range... and very few individuals would tolerate such ...

I do understand that the effect of noise and distortion are cumulative. I also understand the notion of weakest link ...Still I find a system with too many AD to DA to count very satisfying and am sure not to be alone. Just going by lowest SINAD is a sure but potentially expensive way, low THD and IMD? No longer an issue in most electronics... We vilify the 80 dB some AVRs provide but this could well be enough for most musical and movie situations ...
What should look for when building an Audio System? What are the thresholds? I thought I knew.. I don't.

 

[rdenney]

Noise accumulates, but not that fast. If a preamp's background hiss is 20 dB SPL (masked by room noise in most environments), and, say, the equalizer's base noise is also 20 dB SPL, then the resultant base noise of both will be 23 dB SPL--also not audible in most rooms but barely noticeable as a difference even in a soundproof box.

But if one device produces a hiss at 50 dB SPL, then another device that produces a 20-dB hiss won't make any noticeable difference.

Quiescent hiss (produced in the absence of intentional signal) and noise aren't the same thing, of course.

Dynamic range is important. The base noise in my listening room is no better than about 35 dBA SPL. The background noise has a strong component of equipment rumble (I measure when the AC is running as a worst case), which A-weighting discounts. The same room is more like 55 dBC SPL. Given that the background noise is dominated by low frequencies, the minimum discernable musical signal--say, a solo violin in an orchestra context, muted, playing ppp--may be at 25 dB SPL, though we may prefer that it be louder. If the music has 60 dB dynamic range, the peaks--a cymbal crash or fortissimo trumpet--might be at 85 dB, which is loud but not overpoweringly so. LP records don't have this much, of course, but CD's might if recorded without any gain-riding or compression.

If simulating an orchestra-hall experience, the peaks might need to be more like 100 dB, so the quietest bits may be no less than 40 dB. A 40-dB (C-weighted) light tap on a bass drum might not be heard over my AC system, but the same 40 dB from a muted violin will be easy to hear. 40 dB of hiss might be fully masked.

But that cymbal crash has most assuredly been compressed at some point, either intentionally or as a byproduct of something in the recording chain. Even the recording studio room will probably damp the percussive peaks a bit. A really dynamic digital recording might preserve it better than in the past, but I doubt any recording has more than about 70 dB dynamic range. (I suspect most recordings are more like 20 to 30 if perceived as dynamic, and compressed pop music might be less than 10).

Even for those super-dynamic recordings, where we attempt to place peaks near live levels (say, 110 dB SPL or more), the quietest bits will have to be pulled out of the noise in a typical listening room, depending on where it is in the spectrum.

Predicted in-room response for loudspeakers is based on assumptions about early reflections. If a typical listening space has side walls of half the distance between the speaker and the listener, the reflecting angles will be 45 degrees off axis. If the side wall is the same distance as the speaker is from the listener, the angle will be 60 degrees. If the speakers are on the narrow end of a small room and the listener is at the other end of the room, the angle to the early reflections might be as little as 25 or 30 degrees. Thus, if we think that for most rooms, the early reflections will come from surfaces with a reflecting angle of 20-45 degrees, then the frequency response of the speaker at that off-axis angle can be measured, and will predict what the coloration of the early reflections will be. If the early reflections demonstrate a hump not seen on axis, the effect might be really annoying. There is an expectation that the early reflections will have less treble energy than the on-axis direct sound, because high frequencies beam somewhat even with the best waveguide designs. This drop in high-frequency energy off-axis predicts that the combined sound that might be seen on a spectrogram would tilt downward as the frequency increases. If it is smooth, people will hear it as a room effect and filter it out naturally. At least that's the idea. When you think about those angles, it's a fairly unusual room that would impose reflection angles (particularly laterally and from the floor) that would differ much from that "reflecting window" assumption. The more those reflections are damped, the less downward tilt in the in-room response, and the brighter the sound, but to my thinking not unrealistically so.

Movies with sound effects are a whole other consideration.

Rick "there is noise and there is noise" Denney

 

[Blumlein 88]

It's complicated.

 

[wgb113]

context is important, 40 dB ambient noise at 100 Hz is not the same as 1KHz.

usually the room is dead quiet above 1KHz. if someone's average noise level is 40 dB in the 1KHz to 5KHz range I would pity them lol.

Looks like it starts to get louder than 20dB below 200Hz and peaks @ 50Hz AT 35-40dB.

 

[wgb113]

Predicted in-room response for loudspeakers is based on assumptions about early reflections. If a typical listening space has side walls of half the distance between the speaker and the listener, the reflecting angles will be 45 degrees off axis. If the side wall is the same distance as the speaker is from the listener, the angle will be 60 degrees. If the speakers are on the narrow end of a small room and the listener is at the other end of the room, the angle to the early reflections might be as little as 25 or 30 degrees. Thus, if we think that for most rooms, the early reflections will come from surfaces with a reflecting angle of 20-45 degrees, then the frequency response of the speaker at that off-axis angle can be measured, and will predict what the coloration of the early reflections will be. If the early reflections demonstrate a hump not seen on axis, the effect might be really annoying. There is an expectation that the early reflections will have less treble energy than the on-axis direct sound, because high frequencies beam somewhat even with the best waveguide designs. This drop in high-frequency energy off-axis predicts that the combined sound that might be seen on a spectrogram would tilt downward as the frequency increases. If it is smooth, people will hear it as a room effect and filter it out naturally. At least that's the idea. When you think about those angles, it's a fairly unusual room that would impose reflection angles (particularly laterally and from the floor) that would differ much from that "reflecting window" assumption. The more those reflections are damped, the less downward tilt in the in-room response, and the brighter the sound, but to my thinking not unrealistically so.

I get the theory behind it but in my experience and in simply looking at others setups there are so many variables at play that I feel this theoretical measurement is useless. We've all got rooms of differing dimensions and constructed of different materials. We place different speakers and ourselves in different positions on different types of furniture.

I'm just trying to take these measurements, which I see value in, and apply them to the real world - what matters most and what doesn't really matter, or, to what extent does a measurement matter.

You and I Rick, we could buy identical systems and play identical music but get vastly different sounds in our rooms. If you were in a basement with poured-concrete walls dressed in staggered studs with two layers of drywall, heavy carpeting and professionally installed acoustic treatments in a nice rectangular room it would sound much different than my room that's an L-shape with one side open to an adjacent room, a suspended wooden floor, lots of windows, and no acoustic treatments. Then there's the thorny issue of preference which I think gets dismissed too quickly here.

I know we all love to hate car analogies but it sometimes is like looking at the specs and performance test results for a car you're interested in but as you drive it home and around town it's a very different experience. Top speed and g-forces don't matter much for most of us on public roads. It's nice to know a 0-60 time or braking distance but in the real-world?

Just trying to make sense of it all but starting to think it's still somewhat of a crapshoot and then it's down to the person and what they want out of their system to take the time and experiment with as many things as they can to find something they're happy with.

Bill

 

[FrantzM]

It's complicated.

+1...

I know

 

[FrantzM]

<snip>

Just trying to make sense of it all but starting to think it's still somewhat of a crapshoot and then it's down to the person and what they want out of their system to take the time and experiment with as many things as they can to find something they're happy with.

Bill

I believe not knowing or having a good sense of what's needed or required is what makes is a "crapshoot" ...
There are things that once they are met or achieved tend to provide decent to excellent results.

• Good directivity helps in most rooms.
• Multiple subs tend to provide smooth low frequency response.
• Flat in axis and off-axis FR is desirable. Speakers that present these characteristics end to sound the same in very different rooms.
• Equalization is (almost?) mandatory in the low bass.
• Overall low distortion helps in the sensation of a "clean" sound.
• So do Dynamic linearity and capacity. (Scaling linearly with signal level and being able to provide ample SPL at the listening position. Many audiophiles claim they listen at 65 dB, almost with pride but live music, even from one person playing a guitar in a medium size room , routinely goes above that .. peaks of 100 dB are not uncommon in such situations.

Issues that I don't understand enough and frankly don't think they make that much of a difference are :

• Group Delay
• Phase

I'd like to be able to assign some specific numbers to all those parameters...

 

[rdenney]

I get the theory behind it but in my experience and in simply looking at others setups there are so many variables at play that I feel this theoretical measurement is useless. We've all got rooms of differing dimensions and constructed of different materials. We place different speakers and ourselves in different positions on different types of furniture.

I'm just trying to take these measurements, which I see value in, and apply them to the real world - what matters most and what doesn't really matter, or, to what extent does a measurement matter.

You and I Rick, we could buy identical systems and play identical music but get vastly different sounds in our rooms. If you were in a basement with poured-concrete walls dressed in staggered studs with two layers of drywall, heavy carpeting and professionally installed acoustic treatments in a nice rectangular room it would sound much different than my room that's an L-shape with one side open to an adjacent room, a suspended wooden floor, lots of windows, and no acoustic treatments. Then there's the thorny issue of preference which I think gets dismissed too quickly here.

I know we all love to hate car analogies but it sometimes is like looking at the specs and performance test results for a car you're interested in but as you drive it home and around town it's a very different experience. Top speed and g-forces don't matter much for most of us on public roads. It's nice to know a 0-60 time or braking distance but in the real-world?

Just trying to make sense of it all but starting to think it's still somewhat of a crapshoot and then it's down to the person and what they want out of their system to take the time and experiment with as many things as they can to find something they're happy with.

Bill

Not arguing--I was just explaining the prediction approach.

Believe me, my room is as complicated as yours.

It looks almost normal in plan view. But here's the section:

The red boxes are Revel Concerta F12's.

I hear a little boominess in the bass, but haven't nailed it down yet. I never heard boominess from the Advents. But I also never heard imaging from the Advents where they were located, and the Revel's lay it all out from the listening position beautifully.

There was a music stand that rings when I play the tuba--turning it a little resolved that. The space under the piano is filled with soft tuba gig bags, so not much room to resonate anything down there. I can make the piano's soundboard ring with the tuba, too, but I don't hear it with the sound system.

But I have all the time in the world to dial it in. No rush.

Rick "no ceiling first reflections, except for what comes out of the speaker straight up" Denney

 

[wgb113]

@FrantzM all good points! I've chosen speakers with decent directivity and multiple subs. The integrated amp I've chosen to work with has a solid DAC and phono stage along with excellent bass management and room correction facilities. They're not SOTA but I'm hoping they're good enough to add up to enjoyable sound.

@rdenney My new room is actually much different than described and I'm not even in it yet so I have no idea how it will sound! It's nearly a square but it's a dedicated room so I'm hoping the above - and a lot of experimentation - helps!

 

[Blumlein 88]

Your questions are good ones and unfortunately easy answers on a checklist aren't available. The conservative approach is to sledge hammer everything. Make all the specs beyond all chance of audibility just to be sure.

The bottleneck is the speaker. I'd feel pretty safe saying if your total system SINAD is -80 db at the speaker terminals and you have flat response you'll have nothing to worry about. Even that one number is quite the simplification. One can come up with unusual situations where that isn't quite enough, and one can show examples where less is still just fine. But that is a number that while not necessarily a 100% guarantee will rarely leave you wanting. It also is a number that is attainable without going to extreme measures these days.

Having said that I think SINAD is a bad choice. I think you'll come much closer to your goals if you separate distortion and noise and frequency response. Frequency response that is +/- .1 db from 20-20,000 hz is fine and readily attainable. Distortion with music is fine if all of it is below -60 db. Getting it down to -80 db gives margins of error, and is fine. Noise is one of the worst specs to put a single number on. It almost always is much higher at low frequencies and negligible from the mid-range on up. OTOH, if you can turn your gear up to the max you'll ever listen and don't hear anything then you are good on noise. Sometimes hiss is an issue and of course 60 hz hum. You should be able to get those to the point they audibly disappear without too much trouble.

Speakers are where the rubber meets the road once everything else is taken care of. I always suggest choosing your speakers and working backwards from there. No speakers are fully flat, they pretty much always have higher distortion than everything else combined in the chain of gear, and you have directivity/room interactions. Still a good sounding speaker will generally sound good in most all reasonably normal room situations. I guess I'll stop here before getting more complications involved.

 

[FeddyLost]

but what are the cumulative effects and what is the end result

If we are looking into measurable "synthetic" parameters of modern electronic equipment, we need to keep in mind that ANY real manufacturer is trying to stretch these numbers as high as it gets. In real life we'll have a lot of EMI interference, vibration, ground loops and all other issues that we can't really predict so most probably real parameters in customer use case will be at least not better.
So, really bad measurements currently is kind of sympthom that manufacturer is totally irresponsible either don't value ordinary measurements as high as common scientific consensus which is also barely good sign.
Personally I prefer to get measurements of gear as good as possible within budget and still evaluate sound personally if equipment is not extremely cheap.