effect of noise on perception of music

Left   Right
Min Sample Value:   -127.82   -179.83
Max Sample Value:   156.81   169.48
Peak Amplitude:            -46.39 dB   -45.23 dB
Possibly Clipped:   0   0
DC Offset:   -.001    .001
Minimum RMS Power:          -65.98 dB   -63.96 dB
Maximum RMS Power:         -54.38 dB        -53.1 dB
Average RMS Power:  -59.99 dB        -57.16 dB
Total RMS Power:       -59.7 dB            -56.96 dB

I would guess the answer to this depends more on psychoacoustics that pure arithmetic but clearly I don't know. It might depend most upon hardware and experience.

The current LP project has a fair amount of very uneven background noise. It varies in intensity, and somewhat in spectral content, from instant to instant. This noise isn't particularly unusual or extreme, although fortunately it isn't what I find on the majority of LPs I do. The measurement shown above was made over a 4.5 second selection between tracks, no music included. For comparison, 4.5 minutes of the track immediately preceding this noise measures:
   Left   Right
Min Sample Value:   -27199.93   -27403.25
Max Sample Value:   28073.8   31785.77
Peak Amplitude:   -1.34 dB   -.26 dB
Possibly Clipped:   0   0
DC Offset:   -.001    -.001
Minimum RMS Power:   -43.28 dB   -39.61 dB
Maximum RMS Power:   -11.99 dB   -11.04 dB
Average RMS Power:   -20.12 dB   -19.59 dB
Total RMS Power:   -19.36 dB   -18.81 dB

Tracks on this album have longish instrumental fade-outs. I just listened to it (written to CD-RW) in the living room. Even though track endings become fairly faint, I can't really notice the noise until the very end. The noise is easily heard between tracks.

Spectral View shows that the noise is the same during the fadeouts as after they've completely faded. Chances are it is the same throughout the track. Logically, if the noise exist everywhere, it effects all the music. It is part of the mix; the music is different with it in there as would your tap water be different with a sprinkling of deadly bacteria added. The question is, should the noise make an audible difference?

If one had the proper monitors in the proper set-up, and well trained hearing, should one expect to be able to hear any difference between this recording and a CD made from the master tapes -- other than on very low level passages where the noise itself would be audible. I don't believe one would generally hear the noise as a separate sound, but should one be able to do a successful blind ABX test to distinguish between the versions?

Well maybe the test conditions are not well stated. I don't want to get into irrelevant discussions about distortions or other frailties of vinyl vs CD per se. Let us please just assume the only difference is this noise or lack thereof.

If one had the proper monitors in the proper set-up, and well trained hearing, should one expect to be able to hear any difference between this recording and a CD made from the master tapes --

yes

Just to be sure I understand, you are saying that the levels I recorded, the differences between the noise level and the audio level, do represent audible differences under good conditions.

For this purpose, doing a good NR treatment on the LP recording would be the equivalent of getting the master tapes. How far would the noise have to be reduced to make the difference inaudible under the conditions specified?

Tracks on this album have longish instrumental fade-outs. I just listened to it (written to CD-RW) in the living room. Even though track endings become fairly faint, I can't really notice the noise until the very end. The noise is easily heard between tracks.

With this kind of thing I'd be very tempted to 'chase the fade' out and zero the background between tracks, with a quick fade up at the start of the next track. As long as your own fade works in sympathy with the original fade out it'll work well - the background noise will be being reduced long before it becomes an issue.

I'm with Andrew here - unless there is some compelling reason, I always chase a fade.

sorry I wasn't very clear. I have no problem about how to handle what I need to do. I am just curious about what I don't understand. I am only looking for some insight into specifically what I asked in the last three paragraphs.

For this purpose, doing a good NR treatment on the LP recording would be the equivalent of getting the master tapes. How far would the noise have to be reduced to make the difference inaudible under the conditions specified?

No, it wouldn't really be the equivalent at all, would it? You'd be eliminating a combination of noise footprints that was already 'legitimately' present in the original recording - i.e., the mic preamp and analog processing noise as well as any present in the mastering system. If you wanted to hear what the microphones picked up, then you would need to correctly identify the individual noise components separately, and only remove the relevant ones... which of course we can't do.

Ultimately, the effect of noise is to mask signals at a level significantly lower than the noise level itself - but you have to be careful with this, because dither (which is, after all, a low-level noise signal) is absolutely necessary, especially with resampled signals that are truncated to a lower bit depth, to eliminate quantising noise - which is pretty objectionable. Removing all noise from any part of a recording is usually a pretty bad idea, one way or another. You want to know how much noise to leave? Well, it depends entirely on the musical content, quite frankly, and what you can get away with. On a 16-bit integer medium like CD, fading out noise by 'chasing the fade' certainly provides an acceptable transition from one noise profile to another, but only if this fades to dither, not absolute silence. Noise is, by its very nature, present all the time whenever there is any sort of heat involved, and it's certainly present within your hearing system. Getting rid of all of it would be unnatural, and it always seems this way when something has been over-processed from this POV.

Yes, I understand all of those complications. I though this would be easier to communicate. I just want some way to talk about only the noise that the LP adds and its effect upon the music. I don't want to do anything about the noise, I just want to understand something about it.

Noise over and above the noise of the original recording is added by virtue of the recording coming from the LP. In this particular case the noise is higher (and less constant) than on many LPs. Still that noise is itself only evident where the audio signal level is very low; then it is quite audible. The inquiry is only about this vinyl disk added noise, not about the other intrinsic noise of the recordings. Even if we can't easily separate them into different signals, we can easily separate them conceptually.

While this noise varies from instant, it still has some average value when measured over a length of time. The music has some average value when measured over a length of time. The numbers I posted are of course the total noise, not just the noise I'm talking about. If we had the tape that was fed to the cutting tool we could, in theory at least, get a file containing only the added noise.

When the numbers for the noise are not much different than the numbers for the music, it is pretty reasonable to presume the noise effects what the music sounds like, even when one can't separately distinguish the noise; the noise is mixed with, and become part of, the music signal. If the noise were instead down at -110dB, it would probably be the case that it modifies the sound of the music very little.

For the amount of noise this LP has, does the medium level and high level audio sound any different than it would sound if the LP noise did not exist? If so, how much lower would the average RMS measurement for the noise have to be before that would not be true?

Yes, I understand all of those complications.
For the amount of noise this LP has, does the medium level and high level audio sound any different than it would sound if the LP noise did not exist? If so, how much lower would the average RMS measurement for the noise have to be before that would not be true?

you were correct in your original assertion that you'd have to appeal to psychoacoustics to address some of this

you you apparently have another larger conceptual issue:  noise

because not everything that is 'coming out of the LP' that is not detectable on the master tapes is not necessarily or inherently 'noise'.  

so resolving when you're talking 'noise' and when you're talking 'different' has to be the first step . . . and quite simply is probably not possible without access to not only original tapes (in original state . . . no matter how you store them analog tape is subject to 'change' over time) but to original delivery system.   And there are inevitable artifacts introduced by the mastering and 'cutting' process as well.   For the kind of analysis you're postulating you don't necesarily need the master tapes, you need the cutting master (which is perhaps in better shape then the tapes any way)

All these things are probably true but few are relevant. It is perhaps unfortunate that I mentioned vinyl or master tapes but I didn't expect that to cause such a difficulty. In spite of my request

Well maybe the test conditions are not well stated. I don't want to get into irrelevant discussions about distortions or other frailties of vinyl vs CD per se. Let us please just assume the only difference is this noise or lack thereof.

.
no one seems willing to do so. I was curious about the question in terms of the album I was then working on, but the question is independent of any particular recording or  media type. It seems like a simple question. It may well not have a simple answer, but I don't know that either.

I expected some responses along the line of
"in part it depends on the nature of the noise"
"in part it depends on the relative (to the music) loudness of the noise"
and some discussion explaining how and why, hopefully with enough insight so I would know if the album would have sounded any different if that noise didn't happen to be there. Maybe I ‘m wrong. Maybe any reasonable answer really depends mostly on whether or not sea turtles are laying their eggs in Haiti at the time one listens to the music, but it is almost certainly independent of the recording media.

I was curious if the (mostly not "the usual") noise on that LP made the recording sound different than it would sound without the noise. This is not obvious to me since most of the time I can't tell whether or not any noise is present -- but I know it is because the noise is there at every place the music level gets low enough for me to tell the difference. It seems highly improbably that the noise is present where the music level is low but actually not there when the music level is higher.

While I framed it in consideration of that LP, the question can be considered, and surely answered, to the extent that any answers are know to acoustical science (or whatever relevant discipline), in regard to any mix of any music and any noise. I suppose it is time for me to let it go. Thanks for the effort.

I expected some responses along the line of
"in part it depends on the nature of the noise"
"in part it depends on the relative (to the music) loudness of the noise"

How about " it depends on your response to the noise".

I know people who can listen to a 78 without any apparent consciousness of the noise, and others who are disturbed by the sound of traffic on the bypass a mile away; they will certainly give you different answers.  So my answer is that you must make your own judgement, as only that answer will be right for you and that LP.

I suspect the answer to your original question is "maybe".

Paul

I expected some responses along the line of
"in part it depends on the nature of the noise"
"in part it depends on the relative (to the music) loudness of the noise"
and some discussion explaining how and why, hopefully with enough insight so I would know if the album would have sounded any different if that noise didn't happen to be there. Maybe I ‘m wrong. Maybe any reasonable answer really depends mostly on whether or not sea turtles are laying their eggs in Haiti at the time one listens to the music, but it is almost certainly independent of the recording media.

Leaving aside that this is not in the slightest how you framed the question originally - hence everybody's confusion about your intention - the answer you are seeking you probably won't like, because it will require you to do a lot of reading. IOW, it's too long for me to explain in this thread. What you need to do is study the perceptual mechanism behind ATRAC and MP3, because this relies entirely on a masking process which allows what are effectively noisier parts of a signal not to be percieved. To the extent that in order to achieve the coding gain, they both deliberately make signals noisier (although it doesn't absolutely have to be like this). There is a reasonably concise explanation here - and you will see why I don't want to regurgitate it. But it is not complete, by any means - there is a good deal more to the detail than explained there.

[fusion]

I can hardly argue that I was perfectly  clear, the results won't support that argument, but it was not for lack of trying. I still don't think I connected, however. I said, last post that I expected something "along the line of ..." but that is not what I was asking, not the answer I wanted. I expected it was necessary to get all the ‘depends upon' stuff stated. Then it might be possible to put some specific values to those dependancies and work toward a specific answer.

The answer to my question is either ‘yes' or ‘no.' When the very first response was "yes," I only continued because I wondered if the quote referenced as preface to that answer might mean that the answer's focus was on the monitors, room, and such , rather than the specific thing of interest: the particular input to those monitors.

The article Steve pointed me at provides a label for my question: fusion

in fusion the masker takes on a different quality

In this case the noise is the target and the music is the masker.

We know the music is different than it would be without the noise, different in the sense of adding two wave equations to produce a new resultant wave equation, but
can the "different quality" of the masker (music) be perceived?

The answer is either yes, the music sounds different (as in a difference a person can actually hear) because the noise is there,
or no, the music does not sound different, in spite of the noise being there; the masking is complete.

This restatement is not in any way a different question than I have been asking over and over. And maybe it is still just as opaque, but maybe the inclusion of the correct technical term will help my communication.

The answer to my question is either ‘yes' or ‘no.'

One should always beware of simple answers to complex questions and vice versa. This is a complex question - as any question that effectively involves psychoacoustics is.

Technically the answer is yes, but perceptually, I'd say no - with some reservations. And anyway, to observe fusion effects, the sources have to be much closer in level to each other than the BG noise should be to even quite a poor transcription. Also this doesn't take any account of correlation effects.

But please don't try and claim that it's yes/no - that's quite unsupportable.

We have the specified situation: ideal monitoring conditions and an able and experienced observer.
We have sample A: the particular music-plus-noise in question, represented by the second set of numbers.
We have the construct sample B: the same recording without the noise.
We have the noise itself,  identified by the first set of numbers.

It may be that the difference is so obvious that the observer can easily tell the samples apart, virtually 100% of the time. I leave that fudge factor in the statement because sometimes a fella can't even be sure of his own mother, you know. He may momentarily think she is the postman, fire chief, or Chinese ambassador. But, in this case, the answer is a simple "yes."

It may be that any difference in the sound of the two samples, if any humanly identifiable difference exists, is more subtle. We have to follow proper blind A/B testing protocols to arrive at the answer. We probably use a confidence level of 95% or 99%, depending on how we want to run the trials. Here, unfortunately, the answer is not quite so straightforward, but we will still be able to say "yes, with an x% probably" or "no" with some similar appropriate descriptor.

Even good observers vary, so we really need to select an adequate sample group of observers, large enough to provide statistical validity. Then we are a little more busy when expressing the answer.

But all this testing is in the early era of developing our science. If it is a real science, because its events  happen in the physical universe and have  concrete parameters, we eventually move from experimental hypotheses to well understood engineering principles. Where are we today along that path? There seem to be a number of fairly useful lossy encoding schema, to pick one aspect for consideration.

If we could not rely on some reasonable principles, the common advice given to many of this forum's practitioners of arcane arts -- "learn to trust your ears" -- would be most ridiculous. What could it matter what your ears tell you if everyone lives in their own highly idiosyncratic universe?

When we have a science, and we know the characteristics of our samples, we can make highly accurate predictions of what the test results would be if we were to run the tests. It's true, the answer in most cases would of necessity be some sort of probability distribution. We are, however, limiting this situation to proven reasonable observers (and those specified ideal observing conditions) rather than including all the kooks ("audiophiles") and totally inept in the pool. We should not have to qualify every word of our answer too severely.

It may well be that the characteristics I provided are not the information needed to make the prediction. It is possible that I don't have the instruments necessary to make the measurement necessary to gather the data required to make the prediction. I can only provide what I know at the moment; I had to start the discussion somewhere.

My uncertainty there is why I expected some reply along the same lines that many inexperienced posters receive: "we need a little more information before we can give you a reasonable answer"or "it depends upon ...," but I could hope I had actually provided useful information. Or that I had at least made a reasonable start on getting to an understanding.

Of course I could, and did, go with trusting my ears. I know what results I got and I made my decisions on that album before I ever started this discussion. Unfortunately I don't have access to anything approaching ideal monitoring conditions, and it is unproven whether or not I am a reasonable observer. Those limitations doesn't allow me the information to satisfy my curiosity, which is what this thread is all about. Sometimes one has to stir people up as a prelude to getting any cooperation at all.