Nose floor vs audible effects

[mkrzych]

Hello,

I'd like to ask someone who is good with the audio measurements or know this stuff good, how the DAC noise floor measurements go with comparison to the real audible range and level of sound we usually listen to?

 

To be more precise, let say we have measured some peaks over the noise floor of the DAC for 1kHz tone up to (just example) -102dB, does it mean that those may be audible only when we listening at the 102dB level? If so, when for the certain equipment gear the noise floor is even worst - 95db, that means that won't be audble if listening at 70dB level right?

 

Additionally, how it goes in comparison to S/N values? For high resolution files we have 24-bit depth giving the ability of the dynamic range over 120dB, so for the DAC if announced as true 24-bit, the S/N ratio must be over that mentioned 120dB - am I right?

[mkrzych]

Nobody is able to chime in with explanation?

[wgscott]

In general, noise is random, signal is not, so you do not need a signal to noise ratio over 1.0 to detect the signal.

 

But, as the thread title suggests, who nose?

[mkrzych]

;-) Well, my mistake in the title, but my nose saying that at least at reasonable sound pressure levels it may be inaudible.

[sandyk]

Perhaps some input by Miska and Paul would be welcome here, as both have been able to hear signals well below the noise floor due to their Sonar experience.

[Miska]

Hi,

 

I'd like to ask someone who is good with the audio measurements or know this stuff good, how the DAC noise floor measurements go with comparison to the real audible range and level of sound we usually listen to?

 

That is actually pretty complex question, because in many cases discrete tones are considered part of the noise floor while audibility and effect of discrete tones vs white noise are quite different.

 

To be more precise, let say we have measured some peaks over the noise floor of the DAC for 1kHz tone up to (just example) -102dB, does it mean that those may be audible only when we listening at the 102dB level? If so, when for the certain equipment gear the noise floor is even worst - 95db, that means that won't be audble if listening at 70dB level right?

 

It depends on properties of the noise floor. If noise is white, you can hear quite a bit below it's level, usually at least 20 - 40 dB below the total level (if it's a discrete tone vs noise).

 

First, if you have pretty quiet normal room, it has background noise of around 30 dB SPL. For amplifiers you need to note the reference level for the noise measurement. For comparison purposes it could be stated for example relative for 1W output power. For a powerful amplifier the full power SNR may be much better.

 

Additionally, how it goes in comparison to S/N values? For high resolution files we have 24-bit depth giving the ability of the dynamic range over 120dB, so for the DAC if announced as true 24-bit, the S/N ratio must be over that mentioned 120dB - am I right?

 

24-bit PCM has dynamic range of 20*log10(2²⁴) = 144.5 dB. You are going to have hard time finding a DAC that would have such SNR. ESS lists 135 dB DNR for their ES9018 DAC chip in mono mode and TI 132 dB DNR for their PCM1792A in mono mode.

 

Note that the noise level covers entire audio band so it is much less audible than a discrete tone at same level... So for a DAC that thas let's say 100 dB SNR, you don't have much problem measuring discrete tone at -120 dB level, or hearing it (given enough gain).

[mayhem13]

I 'think' what the OP is questioning is our ability to actually discern some of the lower referenced signal to noise ratio's in a complex signal such as music. True, many SNR for amplifiers spec at 1w which if given an average efficiency speaker system is capable of say 88db nearfield at peaks.....or in other words pretty darn loud so the spec can be quite meaningless in day to day listening, even less so with high efficiency speaker systems.

 

Ambient room noise floor........a good starting point for considering the specs of gear or file formats when considering a purchase or upgrade. You'd be surprised what you find in a room you believe to be quiet. Simple inhale and exhale sounds, ground plane low freq resonances, mechanical noises from HVAC systems, lighting ballasts and transformers......things you never consider as poluting your listening experience but yet we split hairs here over signals of lesser degree of audible amplitude.

 

So in other words, next time when considering equipment specs, best to consider their real world potential first.

[sandyk]

I 'think' what the OP is questioning is our ability to actually discern some of the lower referenced signal to noise ratio's in a complex signal such as music.

 

It may not matter too much for pop music with all of the dynamic range squeezed out of it by excessive compression and limiting, but Classical Music for example, often has very low level passages, where some people are even able to hear musician's pages being turned. Perhaps you should stick with 320MP3 and RB CD if your system is unable to resolve these things ? I may no longer be able to hear such fine detail either, but I know of others who can. Some are even able to tell from good recordings, the nature of the recording venue, whether it is dry sounding, spacious or whatever, even the precise location of individual musicians..

[dallasjustice]

IME, the file bit depth is totally unrelated to the DR. Some of the most exquisite and dynamic music ever recorded is red book.

 

Given the forgoing and even the best rooms only have maybe 90db of DR, how is 144db of DR beneficial for home reproduction?

 

It may not matter too much for pop music with all of the dynamic range squeezed out of it by excessive compression and limiting, but Classical Music for example, often has very low level passages, where some people are even able to hear musician's pages being turned. Perhaps you should stick with 320MP3 and RB CD if your system is unable to resolve these things ? I may no longer be able to hear such fine detail either, but I know of others who can. Some are even able to tell from good recordings, the nature of the recording venue, whether it is dry sounding, spacious or whatever, even the precise location of individual musicians..

[sandyk]

IME, the file bit depth is totally unrelated to the DR. Some of the most exquisite and dynamic music ever recorded is red book.

 

Given the forgoing and even the best rooms only have maybe 90db of DR, how is 144db of DR beneficial for home reproduction?

 

Get real. In the real world , not in manufacturer's Application Notes, very few DACs do much better than a bit over 120dB and very few DACs are capable of a full 24 bit resolution. Even the Data Sheet for the DSD1792A states − 127 dB (2 V rms, Stereo)

[mkrzych]

It may not matter too much for pop music with all of the dynamic range squeezed out of it by excessive compression and limiting, but Classical Music for example, often has very low level passages, where some people are even able to hear musician's pages being turned. Perhaps you should stick with 320MP3 and RB CD if your system is unable to resolve these things ? I may no longer be able to hear such fine detail either, but I know of others who can. Some are even able to tell from good recordings, the nature of the recording venue, whether it is dry sounding, spacious or whatever, even the precise location of individual musicians..

 

Well, in a good recording and even on my probably crappy system in the eyes of indoctrinated high end audiophile, I am easily able to hear the musicians pages being turned or breathing when playing on the. certain instrument. Also space ambient if resonant and well captured.

 

What I am rather interested is the fact if measurement is showing peaks around 1kHz discrete signal up to let say -90dB which is not so good assuming for a good DAC, we won't be able to hear those distortions when not listening at 90dB level. Correct?

[sandyk]

Well, in a good recording and even on my probably crappy system in the eyes of indoctrinated high end audiophile, I am easily able to hear the musicians pages being turned or breathing when playing on the. certain instrument. Also space ambient if resonant and well captured.

 

What I am rather interested is the fact if measurement is showing peaks around 1kHz discrete signal up to let say -90dB which is not so good assuming for a good DAC, we won't be able to hear those distortions when not listening at 90dB level. Correct?

 

What is the point of spending more money on far better measuring DACs if these things don't matter ?

[mkrzych]

What is the point of spending more money on far better measuring DACs if these things don't matter ?

 

I am not gonna spend money based on the measurements at all! BTW, it may be someones ego. Just want to get some technical knowledge about how to read those measurements and be idiot proof when talking with highly motivated in store seller.

 

To be more precise if you got something like below:

 

 

How we can interpret it and more importantly if it can be audible in some situation - which one if so?

[sandyk]

If you believe that measurements don't matter, why not just buy a $30 DAC from ebay ?

I doubt that anybody here will be able to advise you on translating measurements of a DAC to sound quality, unless the DAC is well below average,. and with mediocre power supplies.

[mkrzych]

If you believe that measurements don't matter, why not just buy a $30 DAC from ebay ?

 

Because it's nonsense and exaggeration this way. I just would like to understand some basic theory which is I suppose not magic.

[dallasjustice]

Get real. In the real world , not in manufacturer's Application Notes, very few DACs do much better than a bit over 120dB and very few DACs are capable of a full 24 bit resolution. Even the Data Sheet for the DSD1792A states − 127 dB (2 V rms, Stereo)

 

I am talking real rooms and real dynamic range in those rooms.

 

Do yourself a favor before you hit reply: buy a calibrated mic, a good mic pre and quality ADC. Measure your system at seated position. You could download the Audiolense trial software and use it to measure your system's DR. See if it's any better than 90db. I will be surprised if it is.

 

There are many reasons other than the specs you mentioned to buy gear.

[beanbag]

The practical noise floor of the DAC is if you can still hear your speakers hiss or make a 60 Hz buzz when you have your system turned up to full volume.

 

The answer to your second question is more complicated due to different hearing sensitivity at different frequency, wideband vs narrow band noise, whether two noise sources are at nearby frequencies. CAn you clarify the questions in the second paragraph? What part is -102db?

 

Did you ever read that xiph.org article about how hi res is useless, and redbook CD can encode signals below the noise floor?

[DavidL]

I remember reading some while ago that there is pysho-acoustic evidence that we find listening more 'comfortable' if there is some very low level random noise in the background. Apparently we become 'unsettled' if the noise floor is too low. I can't recall the source of this finding unfortunately.

 

[Don't want to derail this thread but this factor may be part of the reason many audiophiles prefer vinyl reproduction to digital - I know Michael Fremer did some blind tests not so long ago and found that listeners preferred the more 'noisy' of two high-quality turntables.]

[audiventory]

Signal/noise ratio is not peak level of signal minus peak level of noise floor.

 

Signal/noise ratio is relation "energy of signal"/"energy of noise".

 

Energy is summ of spectral comonents into defined band. For different values of band - different value of SNR.

[GregWormald]

Generally (IMO) the greater the difference between the desired signal and **all** unwanted sounds the better. If our ears were all that did the listening, then anything below the level of the signal would essentially be inaudible.

 

However, our brains do the actual listening and since hearing is a primary survival sense, absolute numbers can mean very little.

 

For instance, it is quite usual for us to be able to hear our name across a crowded room at a noisy party--even when we can't make out any other parts of the conversation at the same level.

 

In the end, buy what sounds good to you and enjoy the music.

 

Greg

[PewterTA]

This might not exactly fit into this conversation, but I think it sort of shows how we can be fooled by things. Over the weekend, WKHanna and I tested two similar amps (Rotel RB-1090 vs RB-1080). The differences are basically 200w vs 384w a channel between the two. Both amps were similarly modified by myself, using silver soldier, better wiring and a couple other things I did that I can't remember off the top of my head. These amps have been this way for the last year or so. Well this weekend, I decided to take my 1080 and modify some more by isolating the transformer. I did this by suspending the transformer using Portuguese cork and sandwiching the transformer in between. I then encapsulated the toroidal transformer with ring of Mu Metal keeping the magnetic fields away from the electronics for each channel and caps. The difference was pretty stunning in my own personal testing.

 

So that leads to this weekend, I took the 1080 over to WKHanna's house and we had a "shoot off." We went back and forth multiple times (as many as I would say 20 times) listening to various tracks on each. We could not do a double blind test, so we knew which one was playing...

 

Here's where my thoughts come into play on this thread. During our testing the first 10 times, we never volume matched the amps. After about 10 times we both agreed the 1080 was louder than the 1090. Hands down at least 3dB louder minimum. So in our haste we figured out that we had a loudness issue that was creating all of our differences between the two amps. Trumpets and percussion hits on the snare or bass drum... they were just louder. No ifs, ands, or buts about it. So when we went to volume match the amps we found out, to our utter amazement, both amps read EXACTLY the same in their loudness with multiple dB meters.

 

We literally sat stunned for about 30 minutes trying to wrap our brains around how you could hear a CLEAR volume increase...yet it does not match on any sound meter. We started into our perception of the music, which is why we continued going back and forth for another 10 times or so. On every piece of music we listened to there was a clear difference in how loud the music sounded, but not once did we get even one dB in difference between the two....

 

So I think it sort of goes to show that our brains can perceive a difference in things even when none may really exist. I wish we had had more people there to experience it, but unfortunately we didn't. And I'm not about to sit around and wait on my 1090 to modify as I like the differences I hear with the isolation.

 

So even though we can measure things such as the noise floor and say one thing is more quiet than another... it can possibly be perceived by our minds, but may no necessarily by our instruments to measure (or vice versa).

[dallasjustice]

Here's a shot of a dirty noise floor, Yo.

 

 

In general, noise is random, signal is not, so you do not need a signal to noise ratio over 1.0 to detect the signal.

 

But, as the thread title suggests, who nose?

[mayhem13]

Portugese cork!..........things just keep getting better here with no signs of slowing down I see. Keep up the great tweeks and mods guys. It's doing wonders for the credibility and survival of this wonderful hobby!

[Miska]

What I am rather interested is the fact if measurement is showing peaks around 1kHz discrete signal up to let say -90dB which is not so good assuming for a good DAC, we won't be able to hear those distortions when not listening at 90dB level. Correct?

 

Roughly yes. Symphony orchestra can reach about 120 dB SPL.

 

90 dB SPL is not yet very loud.

 

This is one of the scales also with recommended maximum exposure times (note, continuous):

Dangerous Decibels » How Loud is Too Loud?

[sandyk]

I am talking real rooms and real dynamic range in those rooms.

 

Do yourself a favor before you hit reply: buy a calibrated mic, a good mic pre and quality ADC. Measure your system at seated position. You could download the Audiolense trial software and use it to measure your system's DR. See if it's any better than 90db. I will be surprised if it is.

 

There are many reasons other than the specs you mentioned to buy gear.

 

And while you are at it, dumb it all down with additional software equalisation ?