AnandTech PS article

[NeilBlanchard]

Hello Mike:

The AnandTech article seems to have been corrected: it (now) reads that 8dB represents an approximate doubling of perceived volume.

So by that standard, the quietest (Antec) is 4.5X to 8X quieter than the noisest (Vantec) (19-24dB difference). Yikes! The only problem with these noise figures is they don't say what the load was -- I'd have to assume then, that it is at idle.

[MikeC]

8 dBA is more correct than 3 but I don't understand why they don't just go with the same standard that has been used in acoustics and psychoacoustics for 50+ years. 10 dBA = a doubling of perceived loudness. What is the point of saying it's 8? Any acoustics expert will say that's not right, the research doesn't support it. It is a reference point: Without references there is chaos. Now there will be another 100,000 people scratching their heads and saying oh, 8 dBA is double? ok... Misinformation is dumb.

OK sorry for the rant.

[UncleAstro]

I guess that the folks at Anand don't know what a logarithmic scale is. Unless we were working with a base 8 numbering system an 8db difference can't be double. By definition, a 10db difference is a doubling in a base 10 logarithmic scale. How they could come up with their figures is beyond me.

Brett

[MikeC]

Someone emailed me this very interesting & sophisticated thread from the Google Group: rec . audio . pro It's a obviously an audio / recording discussion group.

The thread deals with DIFFERENT dB? (How do they differ? Practical uses?)

The particular postpointed out to me says:

In 'electrical' terms, if you're comparing power, then double = 10log(2) = 3.01dB (rounded to 3dB). But power is proportional to sqrt(P), so with voltage, 6.02dB (rounded to 6dB) is double.

In 'audio' terms, you may be looking at voltage, power, or sound pressure level (i.e. the closest real term to perceived loudness). Typically voltage is in dBu or dBV (they're different, see the webpage), power is in dBm, and sound pressure is in dBspl (or usually just dB). Sound pressure is analogous with voltage, i.e. 6dBspl is double. Sound pressure is what they are referring to when they say "a jet engine is 120dB, a jackhammer is 110db" etc.

However, a few other posts most succintly identified the crux of the loudness issue for me:

Graham is right in suggesting that there is no real measure of psychoacoustic doubling of apparent loudness. He is also right in
saying that on average, a doubling of loudness requires more like 10dB(voltage) increase in level rather than 3 or 6dB sometimes mooted. The dB scale has more to do with convenience in handling units of large numerical range than psychoacoustics.

How loud we percieve things is a very complex issue which depends on the
spectrum, absolute SPL, and dynamics of the signal, to name but a few (not to mention atmospheric pressure and recently consumed beer!).

For instance an increase in signal level of 6dB has a greater effect at low
levels than high. This is because the ears response to absolute SPL is not
linear and cramps at the highest SPLs. This is why we get get a naturally compressed sound from listening at high levels. And why we can to some extent re-create the sensation of high levels by compression etc.. Also this is why we need to take care with monitoring levels when mixing!

However the adjustment the ear makes to high SPLs has delay in it and is thus a dynamic process. So transients at high SPLs are less cramped than the underlying music, This is why the percussion often stands out when we listen loudly and why lack-lustre mixes often sound better if we thrash the monitors.

AND

In the whole development of music we still only have a vague ppp to fff to describe loudness. With other sensations it is quite easy to judge what double is e.g. you can take two equal length sticks and put them together and the human eye is very good at perceiving the angle subtended whatever the distance. Similarly the ear is good at pitch perception especially for an octave interval, but loudness is like colour in that the concept of "double" does not seem appropriate. "Double" requires an fixed linear scale for reference and loudness does not have this for pitch or absolute level before we even start taking into account exposure to previous loud sounds.

I think this colour analogy is a good one. We can measure the wavelength of light, but we still only see a rainbow. We do not talk about red being
double violet although it nearly is.

[NeilBlanchard]

Hello Mike:

You grabbed some good bits. I used to be an audiophile of sorts, and I'm remembering something called the Fletcher/Munson curve. It has to do with a quite wide change in response of "our" (human) ears depending on the frequency. Bass (low) frequencies are the strangest: we have relatively poor response when there isn't much volume, but then there is a range of volumes where we hear small changes as bigger than if the same increase were higher up in frequency. Then, somewhere around 5kHz we have the ability to hear very well even at very low volumes, but we are less able to hear differences as the volume gets greater.

Since fans generate medium to medium-high frequencies, it probable that it's the oversensitivity at low volumes with poor distiction at higher volumes in the 5K range that is coming into play here...

I'm sure I've got the details wrong at least, but the general concept is: knowing what consititutes a "doubling" of volume would require that you 1) know what frequency you were refering to, and B) what volume you were starting at -- cuz' without both of those, it's probably meaningless to try!

Another thing to remember is that frequency doubles with each octave. So, the theoretical limits of "our" hearing is 10 octaves:

20-40Hz Bottom octave -- below most music.

40-80Hz the top of this octave is that annoying boom that "those boys" produce to excess in their cars.

80-160hz this starts to touch on the midrange

160-320Hz lower midrange

320-640Hz right in the midrange

640-1280Hz the upper part of this is probably higher than the highest human singing

1280-2560Hz lower high-range

2560-5120Hz definite tweeter territory

5120-10240Hz cymbals are in this range

10240-20480Hz most people's hearing drops off somewhere in this range. The squeal in monitors and TV's is about 15-16kHz. Lotsa' "old guys" can't hear above 5kHz!

[ez2remember]

Hello,

Neil do you have any idea what frequency PC Noise would drop into?

1. Quiet, very low noise PC? (quiet undervolted fans, quiet decoupled Seagate HD)

2. Noisy PC that comes standard from most vendors?

Thanx in advance for any info.

[BaconTastesGood]

One of the things the folks at [url]www.g4noise.com[/url] have done is taken their loud Macs and sampled the audio and run it through a spectrum analyzer trying to understand which components make which noise.

This may be something that Mike could look into -- absolute sound pressure isn't nearly as useful if we can't identify which bands are causing the most noise.

[MikeC]

You guys are right that lots of sphisticated analysis can be applied to quieting PCs, and that may come. But the truth is that whether the loudest noise is at 10 dBA or at 40 dBA, at 100 Hz or at 5000Hz, as long as it is not completely obscured by ambient noise, you will learn to hear it, if not right away, then within some finite length of time.

In another life, I found that a speaker system with a flat measured frequency response usually sounded a bit annoying. After a great deal of experimentation, I discovered that a slight dip in the response, as little as 1-2 dB, between about 3.5 and 6 KHz often made most systems sound that much clearer, more natural. It had to do with our hearing acuity and the various distortions in the music sources in that range. But once you did that, it became easier to spot the next performance anomaly, where it be a lumpy bass, poor dynamics or whatever.

What's the point? Hmmm... Something along the lines that the worst sonic weakness can usually be heard, even if that weakness is really small.

The simplest methods are usually the best and often the cheapest.

• If you can minimize the CPU heat, maximize the heat transfer into the HS, then use the quietest method to eliminate the heat from the case...
  If you can use the quietest fanned PSU...
  If you can decouple mount the quietest HDDs and optical drives...
  If you can minimize airflow impedances throughout the case...

Then there is no need for any audio analysis. The thing will be quiet enough for just about anyone, and it will not be practical to get it any quieter.

Must be Friday afternoon -- someone tell me if that's gibberish up there...

[JarsOfFart]

20-40Hz Bottom octave -- below most music.

40-80Hz the top of this octave is that annoying boom that "those boys" produce to excess in their cars.

80-160hz this starts to touch on the midrange

160-320Hz lower midrange

320-640Hz right in the midrange

640-1280Hz the upper part of this is probably higher than the highest human singing

1280-2560Hz lower high-range

2560-5120Hz definite tweeter territory

5120-10240Hz cymbals are in this range

10240-20480Hz most people's hearing drops off somewhere in this range. The squeal in monitors and TV's is about 15-16kHz. Lotsa' "old guys" can't hear above 5kHz!

To add to that, concert A on the staff is 440 Hz. When frequency is doubled that represents a jump in an octave too.