[SysteX]

Quote:

Originally Posted by TheVinylRipper /img/forum/go_quote.gif Adding or subtracting the logs of two numbers is equivalent to multiplying or dividing the numbers themselves.

Yes, but when using decibels, you are already dealing with the logs of "other numbers"--measurements of power. So saying:

x dB - y dB

is analogous to

x' watts / y' watts

I think much of the confusion comes from the fact that there is an implied reference level in a single decibel measurement. Technically, there is no absolute decibel scale--it only measures relative intensity. When someone talks about a 30dB noise, it is relative to a reference level, 0dB. Without that defined reference level, the decibel calculation makes no sense, as it is a calculation of the ratio of the reference level to the measured level.

 

[Arjisme]

Quote:

Originally Posted by SysteX /img/forum/go_quote.gif Eh, kind of. When dealing with logs: Log(A) - Log(B) = Log(A/B) Similarly, Log(B) - Log(A) = Log(B/A) Even though we are (ostensibly) dealing with fractions, it is indeed true that Log(A/B) = -Log(B/A)

Ok, I did the math.

Log(.0001) - Log(.00316) = Log (.0001/.00316) = Log (.0316) = -1.5
Converting to decibels, we get 20 * -1.5 = -30dB.

Similarly,
Log(.00316) - Log(.0001) = Log (.00316/.0001) = Log (31.6) = 1.5
Converting to decibels, we get 20 * 1.5 = 30dB.

Decibels are the logs of other numbers. So, I can work straight with the decibels when determining the difference: -50 - -80 = 30dB, for example.

The thing is, the OP plugged 30dB into the formula site and got 31.6 V. How do we know whether it is correct to plug in 30dB as opposed to -30dB?

 

[b0dhi]

While were on the topic of distortion, I thought these graphs were interesting (it's hard to see them on the same graph):

Senn 650

Orpheus

Ety

Ultrasone 650

There are some real surprises on that list. The Etys for example almost go to -40dB in the third harmonic, which is not a pleasant sound like the even harmonics are. -40dB would be easily audible.

The Orpheus doesn't really stand out. But the Ultrasone 650 is the cleanest of all the headphones on headroom's list. It has the same sharp line as the Etys, which use an armature driver, except with much less distortion. The Porta-pro and the PX100 do really well too.

To bring this back around to the amp discussion - I've been testing with a tone generator (NCH Tone Generator), which runs through my M3 amp and Juli@ sound card. My hearing is not exceptional, I got -36 on my only try on the test. Yet when I do a 30 second sine sweep on the tone generator, I can clearly hear many harmonics fading in and out along with the main sine wave. The GS1000 should have harmonics at or below -62dB:

Since the transducer has the most distortion of all my gear, that must be where most of the harmonics are coming from. -62dB, taken as a percentage is 0.079%. If I had read that figure on a spec sheet without hearing what it sounds like, I would've assumed it to be barely obvious at all, not glaringly obvious like it is. If anyone else wants to compare the measurements to what they can actually hear I'd be interested in hearing the results. Similar results for intermodulation distortion were surprisingly audible to those that participated, even on electrostats, and I'm wondering if harmonic distortion will be the same story?

EDIT: link to HeadRoom graphing page

 

[TheVinylRipper]

For sound measurement purposes zero db is defined as the lower limit of human hearing, the softest sound that can be heard by normal, unimpaired human ears.

The reason decibels are used to represent sound levels is that the human ear responds in a roughly logarithmic manner to increases or decreases in sound level.

 

[TheVinylRipper]

I still think there is something wrong with those graphs.

I have Diamond Cut Audio's DC6 which is essentially a forensic audio application, like you see on CSI and the like.

The graphs I get of transducer response curves are quite sharp at the top and gradually broaden out toward the base, rather like the shape of the Eiffel tower.

I'm watching my grandkids right now and am not at my computer so I can't show you directly.

However you can download the demo version of DC6 if you wish and test transducers for yourselves.

http://www.diamondcut.com/Downloads/Downloads.htm

 

[b0dhi]

Quote:

Originally Posted by TheVinylRipper /img/forum/go_quote.gif For sound measurement purposes zero db is defined as the lower limit of human hearing, the softest sound that can be heard by normal, unimpaired human ears.

This is true when you hear general phrases of how loud something is in "absolute" terms. However, for this discussion (and generally), 0db is always the amplitude (or voltage) of the input signal. For example, the graphs above, 0db would be the actual amplitude of the input wave seen on the left, not the softest sound the ear can hear.

Quote:

Originally Posted by TheVinylRipper /img/forum/go_quote.gif I still think there is something wrong with those graphs. I have Diamond Cut Audio's DC6 which is essentially a forensic audio application, like you see on CSI and the like. The graphs I get of transducer response curves are quite sharp at the top and gradually broaden out toward the base, rather like the shape of the Eiffel tower.

Firstly DC6 is not something used by crime labs anywhere. "Forensic" is a marketing term - it does nothing any good studio (mine included) can't do, except maybe removing clicks and pops which is specific to ripping vinyl.

Secondly, the broadening of the base you mentioned is bound to happen because even digital audio has a dynamic range of 140dB, so it becomes flat at -140dB. If you look at the headroom graphs, the signal never goes below about -100dB, which is probably close to the SNR of the recording gear. Also, when you mentioned the "broadening" thing about 10 pages ago, someone pointed out that the windowing method will change how "sharp" or "broad" the bases are. If it was sharp all the way down to -140dB, I'd be suspicious, but those graphs are perfectly reasonable.

 

[Arjisme]

Quote:

Originally Posted by TheVinylRipper /img/forum/go_quote.gif For sound measurement purposes zero db is defined as the lower limit of human hearing, the softest sound that can be heard by normal, unimpaired human ears.

Yes, but that doesn't apply to the recent dB discussion above. You are talking about SPL and the dB discussion has been about voltage.

 

[tourmaline]

One thing i didn't see come up in this discussion is hearing. Wouldn't it be much better to test the hearing of every participant to make sure he/she can hear the whole spectrum a system or cable is capable of?!

This would ensure the participants are able to notice the subtlest differences in cables , systems and sources.

I know (much) younger people then me with much worse hearing!
Assuming everybody hears the same is maybe the biggest flaw conducting those tests!

 

[kwkarth]

Quote:

Originally Posted by tourmaline /img/forum/go_quote.gif One thing i didn't see come up in this discussion is hearing. Wouldn't it be much better to test the hearing of every participant to make sure he/she can hear the whole spectrum a system or cable is capable of?! This would ensure the participants are able to notice the subtlest differences in cables , systems and sources. I know (much) younger people then me with much worse hearing! Assuming everybody hears the same is maybe the biggest flaw conducting those tests!

A trained, but defective ear is often better than a perfect, but untrained ear.

 

[TheVinylRipper]

I'm glad to know that the microphones with which those graphs were taken were perfect and added no distortion or frequency response irregularites at all.

Considering that I am the only one on this thread to have posted *any* sort of frequency response graphs taken by themselves then I am so far the leading authority on those graphs.

When someone else on the threads posts graphs which they themselves have taken with their own equipment and software then I shall concede that they have a leg to stand on.

Anyone who has the means to take such graphs and does not post such is suspect in my mind.

 

[mdarnton]

We get the same thing in the violin world. There are the people who hear no difference (and can be quite strident about it), then we have people who can tell the difference. I have a friend who can identify a Stradivari vs a Guarneri on recordings, very consistently, which pretty much proves it can be done, but people who rely on instrumentation rather than their ears usually don't get it, and can often be very strident about it. There are lots of "double-blind" tests "proving" that modern instruments are better than old ones, but they rely on using people who don't have the necessary discriminatory skill, so what does that prove?

Check this out: http://www.nagyvaryviolins.com/ On his lead page there are two FFTs attempting to prove that his violins sound like Strads. If you can read an FFT, you'll get the point. He was on the Nova TV show years ago, playing one of his violins against a Strad. On my TV I could tell the difference, which was extreme, but he pulled print-outs off his computer "proving", to quote exactly, if I remember right after all these years "they are the same".

In summary, instrumentation don't mean squat beyond measuring the very crudest aspects of sound. I follow this very closely in the violin world, where the job should be easier because it's more defined and limited, and though there's lots of expensive instrumentation floating around, it's very crude, relative to what the ears can do when it comes to actually measuring what's heard.

And, after all, it's the ears that matter, not what some instrumentation says. In the context of technology and science supposedly designing headphones. Are we to suppose that headphone manufacturers make design decisions based on numbers, not what they hear? I doubt it very much.

Though the process of and information from a discussion like this one can be interesting, I rarely see anyone proving anything to anyone--for one thing, you can't prove anything about hearing to someone who's functionally deaf and proud of it.

 

[TheVinylRipper]

I don't really see where the hearing of the test subjects matters in a DBT. One can either hear the differences or one cannot.

The mere claim that one can hear the differences should be adequate evidence that one's hearing is up to the job.

 

[kwkarth]

Quote:

Originally Posted by TheVinylRipper /img/forum/go_quote.gif [size=xx-small]I'm glad to know that the microphones with which those graphs were taken were perfect and added no distortion or frequency response irregularites at all. Considering that I am the only one on this thread to have posted *any* sort of frequency response graphs taken by themselves then I am so far the leading authority on those graphs. When someone else on the threads posts graphs which they themselves have taken with their own equipment and software then I shall concede that they have a leg to stand on. Anyone who has the means to take such graphs and does not post such is suspect in my mind.[/size]

Not sure I follow you. Can you amplify your thoughts please?

 

[kwkarth]

Quote:

Originally Posted by mdarnton /img/forum/go_quote.gif [size=xx-small]We get the same thing in the violin world. There are the people who hear no difference (and can be quite strident about it), then we have people who can tell the difference. I have a friend who can identify a Stradivari vs a Guarneri on recordings, very consistently, which pretty much proves it can be done, but people who rely on instrumentation rather than their ears usually don't get it, and can often be very strident about it. There are lots of "double-blind" tests "proving" that modern instrument are better than old ones, but they rely on using people who don't have the necessary discriminatory skill, so what does that prove? Check this out: http://www.nagyvaryviolins.com/ On his lead page there are two FFTs attempting to prove that his violins sound like Strads. If you can read an FFT, you'll get the point. He was on the Nova TV show years ago, playing one of his violins against a Strad. On my TV I could tell the difference, which was extreme, but he pulled print-outs off his computer "proving", to quote exactly, if I remember right after all these years "they are the same". In summary, instrumentation don't mean squat beyond measuring the very crudest aspects of sound. I follow this very closely in the violin world, where the job should be easier because it's more defined and limited, and though there's lots of expensive instrumentation floating around, it's very crude, relative to what the ears can do when it comes to actually measuring what's heard. And, after all, it's the ears that matter, not what some instrumentation says. In the context of technology and science supposedly designing headphones. Are we to suppose that headphone manufacturers make design decisions based on numbers, not what they hear? I doubt it very much. Though the process of and information from a discussion like this one can be interesting, I rarely see anyone proving anything to anyone--for one thing, you can't prove anything about hearing to someone who's functionally deaf and proud of it.[/size]

Totally agreed.

 

[kwkarth]

Quote:

Originally Posted by TheVinylRipper /img/forum/go_quote.gif [size=xx-small]I don't really see where the hearing of the test subjects matters in a DBT. One can either hear the differences or one cannot. The mere claim that one can hear the differences should be adequate evidence that one's hearing is up to the job.[/size]

I completely agree.