[Ferbose]

Quote:

Originally Posted by JohnFerrier I can follow your numbers. You will more easily hear the louder signal. Noise is not easier to hear when competing with another signal. If you wish to believe so, then that is fine. What you write makes sense otherwise.

You still missed it.

Noise is easier to hear when signal is weaker. S/N ratio is poorer when the signal is weak. Because amplifier self-noise is constant in voltage, not affected by the incoming signal. It maybe -107 dB relative to the loudest full signal on CD, but only -47 dB to a softer passage in the CD.
But a noise inaudible by itself can be made audible by a (weak) signal. This is the beauty of our sensitve hearing. No, you don't really sense this small (inaudible when alone) noise as a separate entity. It is just that the distortion to the signal is audible (if you remove the noise the music sounds better), and hence we say the noise is audible.

Vice versa, a signal inaudible by itself can be made audible by adding noise. But this is unimportant in hi-fi, and only relevant if you try to receive/intercept minute signals. This is well proven in all kinds of sensory perception studies, even in fish. When you hear something interesting on TV, the signal is already audible. Hence, you don't need your friend to make noise. If they make noise the S/N will decrease, and you will be pissed. I can give you a real world example of noise making inaudible signal audible. Ever used an analog radio with tuning dials? There are some very distant stations whose signals are too weak to be audible. You tune into that frequency and you hear lots of noise but there seems to be a faint voice speaking underneath that noise, like eavesdropping on Martians. If you remove all that noise in the radio you may find that the signal is too weak to be heard (keeping the volume knob in the same position).

 

[JohnFerrier]

Quote:

Originally Posted by Ferbose You still missed it.

No, you didn't make a point that I could miss.

Quote:

Originally Posted by Ferbose But a noise inaudible by itself can be made audible by a (weak) signal.

You imagine so.

Quote:

Originally Posted by Ferbose When you hear something interesting on TV, the signal is already audible. Hence, you don't need your friend to make noise.

Well turn the volume down until the television isn't audible. Now tell me how it becomes audible by adding another signal.


.

 

[Ferbose]

I have come up with the ultimate example to bury this issue once and for all.

Say you can hear the minimum of 30 dB SPL sound in your system, and let us suppose this corresponds to a -65 dBFS signal on your CD. And your amp's self-noise is constantly at 29 dB SPL when connected to a speaker and thus inaudible. Your CD has a soft passage at -63 dBFS, equal to 32 dB SPL in your system, and while this is playing your S/N is 3 dB, which sucks because noise level is 70% of your musical signal. When there is no signal, you can't hear the noise. When a soft music is playing, you say crap, I can't hear the ambience but only harshness. Now lower the noise floor of the amp from 29 dB to 26 dB SPL using a power conditioner, from inaudible to even more inaudible. When you play your -64 dBFS = 32 dB SPL soft msuic passage again, the noise is now only 50% of your music, and S/N is 6 dB. You ears will thank you.

 

[JohnFerrier]

If you turn the down the volume on a television until it isn't audible, how does the television become audible by adding another signal? It doesn't.


.

 

[Ferbose]

Quote:

Originally Posted by JohnFerrier Well turn the volume down until the television isn't audible. Now tell me how it becomes audible by adding another signal. .

Yes, some scientific facts are counterintuitive.
For example, that space and time are all relative, or something can be a particle and a wave at the same time.
But I can assure you that an inaudible signal can be made audible by adding random noise is a fact. You just need the correct, carefully controlled experimental condition to demonstrate it, and hence we have scientific labs set up to study human sensory perception.
As I said before, using noise to make sub-threshold signals detectable is irrelevant in audio.

But audio quality is hard to quantify by scientific methods so I am not a proponent of letting engineers and scientists telling us what sounds good. In fact, OP-amp designers at TI (Burr Brown) lets qualified listeners tell them what sounds good so they can design toward that direction.

 

[JohnFerrier]

Quote:

Originally Posted by Ferbose But audio quality is hard to quantify...

I agree with this much.

 

[Born2bwire]

I pretty much agree with Ferbose here. Another way to think about what he is talking about is that the added signal acts as a carrier signal. The noise superimposes itself onto the signal and by doing so is raised up into our threshold of hearing. For example, if we turn the TV down until it is just below the threshold of hearing, and then add noise, the added amplitude of the noise plus the TV can be above the threshold.

 

[JohnFerrier]

Quote:

Originally Posted by U of GIBBERISH! I pretty much agree with Ferbose here. Another way to think about what he is talking about is that the added signal acts as a carrier signal. The noise superimposes itself onto the signal and by doing so is raised up into our threshold of hearing. For example, if we turn the TV down until it is just below the threshold of hearing, and then add noise, the added amplitude of the noise plus the TV can be above the threshold.

Got it.

 

[markl]

Someone said noise is *less* audible at loud volumes, but I disagree. It would seem to me (as a layman) that "noise floor" is variable with volume. If you listen to an analog recording, tape hiss (the "noise floor" of the recording) can be invisible at very low volumes, crank it up, and suddenly you have very loud hiss. You can hear this effect with your headphones when you crank your headamp to very loud volumes with no signal passing. The noise level increases with volume.

Question:
Another observation is that we seem to assume that this "noise" that was measured exists independent of the audio signal and is maybe only audible as extremely low-level background hiss (if it can be "heard" at all). But is he possibly measuring *distortion* which can affect the amp's ability to present sounds in the audible range? Perhaps the "noise" he measured can act on and contaminate the actual sound reproduced by the component? In that sense, it is also audible as distortion, and manifests as the perception of graininess, frayed edges, harshness, hash, etc. within the music itself? Possible?

 

[JohnFerrier]

Quote:

Originally Posted by markl The noise level increases with volume.

If the noise is from the source (CDP, tape player, etc.), then, yes, noise increases with the volume. I can't speak for every configuration, but any decent audio system today should not have audible noise with the volume turned up.

 

[Beauregard]

Just read through this thread and wanted to give a big thank-you to Ferbose! And the other folks who participated. For someone without a lot of technical competence, it's great to actually learn something worth knowing from an audio forum.

This is exactly the sort of discussion I'd love to see more frequently - an informative bridging of the usually stereotyped objectivist-subjectivist divide.

Much appreciated,
Beau

 

[Tachikoma]

Quote:

Originally Posted by Ferbose Yes, some scientific facts are counterintuitive. For example, that space and time are all relative, or something can be a particle and a wave at the same time. But I can assure you that an inaudible signal can be made audible by adding random noise is a fact. You just need the correct, carefully controlled experimental condition to demonstrate it, and hence we have scientific labs set up to study human sensory perception. As I said before, using noise to make sub-threshold signals detectable is irrelevant in audio.

This is the theory behind the "noise sharpening" plugin in foobar, isn't it? Correct me if I'm wrong

 

[Ferbose]

Quote:

Originally Posted by Tachikoma This is the theory behind the "noise sharpening" plugin in foobar, isn't it? Correct me if I'm wrong

Here is a good link about dither:
http://www.digido.com/portal/pmodule...der_page_id=27

If you want more information, Bob Katz's Mastering Audio is a fantastic book.

 

[Ferbose]

Quote:

Originally Posted by Ferbose using noise to make sub-threshold signals detectable is irrelevant in audio.

When I said this, I meant in the context of audio playback.

In audio playback, the goal is to minimize noise. There is no intent to inject noise so that sub-threshold audio signal can be heard. Although it is certainly feasible. Let me quote from Bob Katz's (the mastering engineer who made Chesky an audiophile label) book: "human beings are able to hear signals in the presence of noise of greater energy than the signal, i.e. with negative signal-to-noise ratios" (note: negative in terms of dB, as happens when 0<S/N<1). He went on to to explain that we can hear 15-20 dB into the noise. Yes we can hear signal even when S/N=-15 dB.

However, during digital recording, noise is added to make sub-threshold signal detectable more often than you think.

In fact, every self-respecting 16-bit A/D converter today would intentionally inject an analog noise (~5 dB) into the signal before the actual A/D occurs. Without doing this, CDs would sound pretty bad to audiophile ears. This process is called dithering. Although the codability of CD is 16 bit, or 96 dB dynamic range, its perceived dynamic range is close to 115 dB after proper dithering. Impossible? No, the only reason this works is because humans can hear signal below the noise. While you perform dithering in 16 bit A/D converter, the measured S/N becomes 91 dB, worse than the orginal 96 dB, but the perceived dynamic range greatly increases (115 dB). Without adding the noise prior to A/D, analog signals below the threshold of registering the last bit would be lost. The key is that the injected noise, when mixed with the incoming, sub-threshold analog signal, becomes a non-random noise that carries information which can be decoded by the ear-brain. Dithering is not adding noise. Dithering is encoding signals into the noise, and it only works because humans can hear signals below the noise. When there is no extra information to encode, adding noise just makes things worse. Dither is necessary in the analog domain during A/D (reducing quantization noise), and in the digital domain during the shortening of wordlength (reducing truncation noise) or computations (DSP). According to Bob Katz, dithering noise is almost inaudible only at -139 dBFS (got to admire our human ears).

When you have 20 or 24-bit A/D converter, injection of noise, or dithering, is basically unnecessary. The thermal noise of the equipment itself dithers at the 18th or 19th bit (self-dithering). Again, those bits beyond 19 is not useless: they still carry information because we can hear into the noise.

 

[Ferbose]

In the early days of digital audio, many audio professionals believed that quantization noise in 16-bit recording is irrelvant. Their logic was something like this:

In a hi-fi system playing CD, the loudest signal (0 dBFS) should not produce an actual SPL of higher than 105 dB. Let us keep the volume knob constant so that 0 dBFS corresponds to 105 dB SPL.
Now, quantization noise is -91 dBFS, which would correspond to a 14 dB SPL signal. In a decent recording studio, ambient noise, or silence, could be 30 dB SPL. Therefore, quantization noise at 14 dB SPL is below hearing threshold by itself. If you play a bunch of 0000,0000,0000,0001 signals, you will hear nothing. If you make a 16 bit recording without dither, analog information below -96 dBFS is lost. They also thought it is OK, because the lost analog signal would correspond to <9 dB SPL, definitely inaudible when played alone. They therefore thought 16-bit digital audio was bullet proof and hence PERFECT SOUND FOREVER.

However, with soft music passages, the quantization noise superimposed on it becomes audible because of poor S/N ratio. In louder passages no doubt the signal will begin to mask the quantization noise. Undithered quantization noise at -91 dB turns out to be detrimental to low level ambience (decaying reverberation) information, and causes perceived harshness. To ovecome this, dithering is implemented and original analog signals below -96 dB to the peak, now cleverly encoded in noise, actually become audible, giving CDs a dynamic range of 115 dB. In other words, without the help of noise, CD would never sound so good It is neither a trick or magic, it is all because our brain can decode signal buried in the noise. It you dither too much, you destroy ambience, too.

So what does this have to do with power conditioning? Well, lowering the noise floor of your electronics is really important, even if you don't hear a noise when music is not playing. When music plays, those seemingly inaudible noises may cause audible degradations to the sound. Hence, you want to supress noise whenever possible.

The again, there are devices with magical noises that makes things sound better in some contexts, such as tubes. I read that there are tube analog equalizers that sound so good mastering engineers pass signal through it without equalizing band just to make the sound better. Whe mastering becomes largely digital, some people measure the noise of the magical analog compressor and use DSP to simulate it in digital compressors. Believe it or not, the biggest reason that digital audio can sound bad is its low noise. Lots of tiny but annoying artifacts become audible when the overall noise level is so low. In the analog age, these tiny artifacts are buried under the high noise of LP playback, but euphony is still achieved.