I thought one of terry j's comments (which were regarding measuring the difference between two pre's with regards to bass) would be a great opener for another round of civil discussions. This discussion is not just with regards to bass (or pre's) but the entire audible spectrum (regardless of how its produced).

A simple in room measurement will tell you if there is more bass or not, praps no need to measure the unit itself.

I believe it was also terry j that at one stage said that an 18" PA bass driver 'CAN' have more 'slam' or something to that effect than smaller bass drivers even given that they had the same FR ? (I'd have to go search for the exact quote and I'm sure he will correct me if Im remembering it wrong). I remember some interesting discussions were had surrounding this where 'wave front' was brought up, not sure what was meant by that with regards to bass and sub bass ???

But it begs the question, does FR tell us everything?

If so, wouldn't bass / sub bass drivers that measure the same sound the same (like in your 18" bass driver when compared to others)?

I guess a secondary question linked to that is where and how do you measure? Outdoors, Anechoic chamber, Indoors (at the listening position, 10cm from the driver being measured or anywhere in between). I mean some speakers may measure beautifully on axis but terribly off axis and some speakers have drivers that 'beam' at certain frequencies (Polar Response).

If FR can't tell us everything what other measurements should we be looking at?

What if a subwoofer measures perfectly flat with regards to FR but has 30% distortion (I know there are all types) in the lower regions and another subwoofer measure exactly the same but has lower levels of distortion? An example can be seen here where an 18" TC Sounds LMS-5400 was measured in a 75L box compared to a 200L box

I know theres a lot of question in here, but it should make for a great thread on its own right I would have thought ???

Edited April 25, 200916 yr by Drizt

I believe it was also terry j that at one stage said that an 18" PA bass driver 'CAN' have more 'slam' or something to that effect than smaller bass drivers even given that they had the same FR ?

At what SPL? You can have a 5" driver with the same FR as TJ's Franks, but they will go to pieces above about 90dB as they simply cannot displace enough air.

But it begs the question, does FR tell us everything?

No, and I don't know any speaker designer that would say it does. For electronics, it merely tells if it has been competently designed (flat response) or tailored to a particular curve.

If so, wouldn't bass / sub bass drivers that measure the same sound the same (like in your 18" bass driver when compared to others)?

No.

I guess a secondary question linked to that is where and how do you measure? Outdoors, Anechoic chamber, Indoors (at the listening position, 10cm from the driver being measured or anywhere in between). I mean some speakers may measure beautifully on axis but terribly off axis and some speakers have drivers that 'beam' at certain frequencies (Polar Response).

Directivity is largely a function of cone size, and to a lesser degree cone shape profile. Few speaker designers are concerned with off axis response from what I have seen.

Good in room measurements are easy with MLS techniques (above a few hundred Hz), and outdoor groundplane measurements can bring up the rest. Stitch the two curves together and you can get something almost as good as anechoic measurements.

If FR can't tell us everything what other measurements should we be looking at?

Power response, non linear distortion, diffraction effects, directivity, power compression, IMD,, time domain effects (ringing, resonance etc), box modes and panel resonace/vibration. Plus how these change with amplitude, and if all drivers change the same way.

What if a subwoofer measures perfectly flat with regards to FR but has 30% distortion (I know there are all types) in the lower regions and another subwoofer measure exactly the same but has lower levels of distortion? An example can be seen here where an 18" TC Sounds LMS-5400 was measured in a 75L box compared to a 200L box

We are considerably less sensitive to LF distortion, simply because we are less sensitive to LF. In the midrange, much lower distortion is readily audible.

The additional non linearity shown in the link is simply because of the additional air load of the box and is an expected affect; I agree with Illkas results that bigger is almost always better. The trouble is that most households will not happily accomodate a 200L enclosure. Also not that the distortion numbers displayed in the graphs are taken over a much greater bandwidth than that driver would typically be used for and so are much higher than they would be if band limited.

You have actually asked a huge question, ie what (measurable) aspects of a speaker's performance make an audible contribution to it's sound. People like Toole have written books on it and not fully encompassed the subject.

Thanks for the great reply A9X.

Flat Response and EQ etc comes up in a lot of threads and I thought it might be good to have one central thread for the topic / discussion that other threads can point to as the need arises.

Thanks for the great reply A9X.

Flat Response and EQ etc comes up in a lot of threads and I thought it might be good to have one central thread for the topic / discussion that other threads can point to as the need arises.

EQ is best used in the modal region where it can be useful in removing peaks in the total (room + speaker) response.

Frequency response cant tell us everything but it can tell us a lot.

For electronics flat frequency response would seem a necessary but not sufficient condition for accurate reproduction.

For loudspeakers things are more complex as there are so many "frequency responses" however getting the frequency responses right is fundamental to building a good loudspeaker. Research at NRC Canada and Harman Itnl has shown a strong correlation between frequency response in loudspeakers and listener preference.

I have talked to lotsa people about why larger bass drivers seem to sound different ( all other things being equal ) but have never heard a convincing argument about why this should be so. Maybe its just the all other things are never equal, or even close to equal.

As A9x says directivity is largely a function of cone size and to some extent profile. I would add "enclosure " design and crossover design.

I think more and more designers are becoming more concerned with directivity as we are seeing many speakers particularly from European designers incorporating tweeter waveguides in order to reduce off axis flare around the tweeter crossover point. This maybe because it is now easier to achieve the basic flat on axis response with newer drivers and CAD crossover design.

I think A9x is a little optimistic saying MLS plus outdoor ground plane is more or less as good as anechoic measurement. MLS type measurements are often restricted in frequency resolution due to short time windows and some serious problems can be hidden this way. Maybe this is why some speakers that appear to measure well dont sound that way.

I think the effects of power compression which seem rarely discussed may be underestimated. The measurements done at NRC Canada and shown with the loudspeaker reviews at www.Soundstage.com show quite pronounced differences between frequency responses taken at 70dB and 90 or 95 dB.

In terms of where ,what and how ( Max are you listening ? ) you should measure in terms of frequency response one answer might be :

all of , on axis , close to axis, early reflections, total sound power all with 2 Hz resolution.

Then in room at the listening position or spacially averaged for a number points around the listening position again with high resolution and maybe 1/20 octave smoothing for bass frequencies only.

In the end you need to know how to interpret these measurements in terms of what they indicate you are likely to hear. The connection between the measured and the heard is a huge largely unexplored area of music reproduction at home which is why Floyd Tooles book is so valuable to anyone interested in this area.

Cheers Mike

Great response mate. Im glad theres been some great technical responses so far.

music is sound pressure level (amplitude) over time. different drivers made from different materials using different glues and geometries will have different amounts of energy storage and resonance behaviours.

often a ragged FR curve will indicate that there are resonances, but this is another dimension to the sound that FR doesn't always reveal. the resonant properties of a driver are quite audible, and often when a driver sounds like it has a peak in the FR it doesn't, but it has a resonance at that freq.

stereophiles speaker reviews often show this with the reviewer commenting on an aspect of the sound and then the measurements revealing it to be a resonance issue.

music is sound pressure level (amplitude) over time. different drivers made from different materials using different glues and geometries will have different amounts of energy storage and resonance behaviours.

often a ragged FR curve will indicate that there are resonances, but this is another dimension to the sound that FR doesn't always reveal. the resonant properties of a driver are quite audible, and often when a driver sounds like it has a peak in the FR it doesn't, but it has a resonance at that freq.

stereophiles speaker reviews often show this with the reviewer commenting on an aspect of the sound and then the measurements revealing it to be a resonance issue.

Hello Kye, Could you explain what you are saying here a little more. I would have thought that the resonant behavior of a driver would always be visible in its frequency response as long as it was measured with sufficient resolution. The resonant behavior of a driver or system may be made more obvious with a waterfall plot.

There would seem to be some good evidence that above 200 Hz or so we detect resonant behavior using spectral information rather than temporal clues.

Cheers Mike

I've linked this one before..

http://stereophile.com/interviews/608kev/

Revel's Kevin Voecks interviewed by Larry Greenhill, June 2008 - Stereophile.

Voecks: Listening tests over the past 10 years have taught us one other thing. Above the midprice range of loudspeakers, there is no correlation between the sound quality and the loudspeaker's price. Although many high-priced loudspeakers do perform adequately in our listening tests, the most expensive speaker in a given double-blind listening test may be the least preferred by our listening panel.

Voecks: We can easily predict which loudspeakers will be preferred and which will not. For example, a speaker that measures poorly on axis will never be preferred in our double-blind listening tests. Also, listening tests correlated best with our measurements conducted in an anechoic chamber, like the one we have at the Harman facility in California....

....Another listening preference that is reliably reproduced [is the effect of] speaker resonances. Research shows that resonances are audible and influence listener preferences. Speakers with resonances are not preferred by our listening panel....

This next quote may answer why Larger/More Bass Drivers perceptibly sound better..

Voecks: The Salon2 moves more air and has greater output, particularly in the bass. The Salon2's three 8" woofers have a combined area equivalent to a 14" woofer, but the heat generated is spread out among three voice-coils. This means that you won't get the heat buildup that leads to dynamic compression. (As voice-coils heat up, impedance goes up and leads to a mismatch in a speaker's filter network.) The Salon2 is more resistant to dynamic compression than the Studio2 because it has more drivers to dissipate the heat.

Hello Kye, Could you explain what you are saying here a little more. I would have thought that the resonant behavior of a driver would always be visible in its frequency response as long as it was measured with sufficient resolution. The resonant behavior of a driver or system may be made more obvious with a waterfall plot.

There would seem to be some good evidence that above 200 Hz or so we detect resonant behavior using spectral information rather than temporal clues.

Cheers Mike

Not always. Many FR's are done by sweeping a sine wave as signal source, with the receiving gear having a narrow band filter that tracks the source. If done at modest levels, it's very possible for resonances to not be excited, or to be very small compared to how they would look in a CSD as the filter has already moved past where the driver is misbehaving.

This next quote may answer why Larger/More Bass Drivers perceptibly sound better..

Even better is to design the driver motor with better heat control in the first place.

http://www.aespeakers.com/Lambda001-1.php

Hello Kye, Could you explain what you are saying here a little more. I would have thought that the resonant behavior of a driver would always be visible in its frequency response as long as it was measured with sufficient resolution. The resonant behavior of a driver or system may be made more obvious with a waterfall plot.

There would seem to be some good evidence that above 200 Hz or so we detect resonant behavior using spectral information rather than temporal clues.

Cheers Mike

hi ando,

here are some examples of waterfall plots to have a look at.. the FR plot is the highest curve at the rear of the diagram.

notice in the above that while some of the resonances have corresponding dips/peaks in the FR, not all of the dips/peaks indicate resonance issues.

this one not only has resonances from dips/peaks, but it has resonances that either go up and down in frequency, or decay and then get boosted again. very strange, and no way to tell that from the FR, which around the 2k mark looks pretty smooth and nothing to worry about.

this last one is a great example - lots of dips and peaks with nasty resonances, but notice the huge dip at about 22k has no resonance attached to it, but lies between two resonances that are formed by the peak in FR higher in freq and the nastier resonance lower than it, which isn't formed by a peak or dip, but occurs on the slope going from the dip up to the peak.

Not always. Many FR's are done by sweeping a sine wave as signal source, with the receiving gear having a narrow band filter that tracks the source. If done at modest levels, it's very possible for resonances to not be excited, or to be very small compared to how they would look in a CSD as the filter has already moved past where the driver is misbehaving.

good point.

if you read enough of the reviews and the associated comments by JA you will begin to see how difficult it is to really measure a loudspeaker properly. I think he may have even written an article about it.

I have done extensive measuring of loudspeakers over the last decade (I estimate i've done several thousand FR plots - I had a joke with my housemate and co-speaker designer that pink noise was our favourite song ). These tests were mostly done as part of crossover designing for speaker prototypes over the years, although some were done to track down resonances and alleviate them.

All of these plots were done like this:

- make a change

- listen to the change and have a think about what I think needs to happen

- measure the speaker and plot the FR, either confirming or correcting my impressions

As such, I got to be quite familiar with what different FRs sound like.

One thing that was especially interesting was how you did your test. I did two kinds of tests, an anechoic test (via a trick I worked out) and a test where I played continuous pink noise for 5-20 seconds and waved the mic around to evenly sample the 1m x 1m x 1m cube around the listening position.

The first test gives the plot of the speaker, the second gives the speaker and all the room interactions and resonances etc. What we hear (ie, what I would expect to see on a FR plot after listening to the change) was about half-way between these two plots.

In the end I came to the conclusion that when designing crossovers there is far to little resolution on a FR plot to integrate drivers properly when only using test tones / noise (and I was working at 24bit 96kHz). I did the above two FR plots, waterfall plots, time-alignment measurements, and some other tricky things to find resonances and other time-based phenomena.

My method now is to simply use my ears to integrate things because I know what i'm listening for, and when it all snaps into place.

I have a friend who owns a six-figure active system and he obsesses about getting the levels right on his active crossovers. He first adjusts for integration, however there will be a range of adjustment where this occurs. Then within that range he adjusts for another critieria. Then within that range he adjusts for another critieria. These adjustments are tremendously fine and in a real situation there would be no way you could measure to be this close.

I am yet to try his secret method, but will do so once I have the means in place - his system is truly holographic.

FR is one aspect of a system, and it's the aspect that your ears adjust to over time.

Personally, I think that within a certain range (perhaps +-10dB) it is one of the least important things to aim for.

Hello Mika, It is interesting given what Voecks and other Harman employees say about the correlation between measurements and listening that Revel publish so little in the way of measurement results, not even a basic on axis frequency response curve. I have actually twice tried to ask Revel about this but never received a reply.

Given Revels close connection to JBL you would hope they would have the ability to build speakers with minimal thermal compression effects.

Minimizing thermal compression almost certainly is a reason larger speakers or multiple smaller ones sound different but the proponents of the "larger speaker sound better " theory often claim that this true "all other things being equal" , that is the difference is due to the greater cone area and nothing else and some claim that multiple smaller woofers are still inferior. Cheers Mike

Not always. Many FR's are done by sweeping a sine wave as signal source, with the receiving gear having a narrow band filter that tracks the source. If done at modest levels, it's very possible for resonances to not be excited, or to be very small compared to how they would look in a CSD as the filter has already moved past where the driver is misbehaving.

Hi Brett, Wish I really understood more about this .

As I understand it the great majority of peaks in a loudspeakers frequency response are caused by resonances.

If the frequency response is measured in such a way that the effects of resonances are reduced ( or made less visible ) and so the response flattened is this a measure of the " true " frequency response ?

A slow sweep response taken under anechoic conditions would seem to take into account the full effects of resonances but as you say for the resonance to be fully excited it needs a certain amount of energy and time of excitation. This may not happen when playing music .

I am not sure if in the swept sine wave situation the changing frequency is excited long enough for resonances to fully develop or not. Maybe it is. As I understand the purpose of the system is to remove room effects. Measurements made with time windowed impulse systems are open long enough even in fairly small rooms for some resonant behavior, at least at higher frequencies to be examined via waterfall plots. If the resonances can develop in the time allowed in impulse systems then maybe they can develop and affect frequency response in time allowed by swept sine wave type measurements.

As I write this seems less likely so we are left with a variety of frequency response measurements. Great. Back to the same old if make a measurement you better know how to interpret it and what it is actually good for. Cheers Mike

PS Got any good references about this stuff ?

Hello Kye, Thanks for taking the trouble to write such a long and detailed response.

hi ando,

here are some examples of waterfall plots to have a look at.. the FR plot is the highest curve at the rear of the diagram.

notice in the above that while some of the resonances have corresponding dips/peaks in the FR, not all of the dips/peaks indicate resonance issues.

Agreed not ALL peaks and dips are resonance issues but I think most peaks in loudspeaker frequency response curves are caused by resonances

this one not only has resonances from dips/peaks' date=' but it has resonances that either go up and down in frequency, or decay and then get boosted again. very strange, and no way to tell that from the FR, which around the 2k mark looks pretty smooth and nothing to worry about.[/quote']

This may be because the frequency response curve has insufficient resolution. If this came from Stereophile I believe the frequency response curves have a resolution of about 300Hz and this waterfall plot probably has considerably less than that as it is usual to reduce frequency resolution to increase time resolution for these plots. Evidence of this can be seen in Stereophile if you compare the frequency response plots to the highest curve of the waterfalll plot which is generally much smoother indicating less resolution in the frequency domain.

this last one is a great example - lots of dips and peaks with nasty resonances' date=' but notice the huge dip at about 22k has no resonance attached to it, but lies between two resonances that are formed by the peak in FR higher in freq and the nastier resonance lower than it, which isn't formed by a peak or dip, but occurs on the slope going from the dip up to the peak.[/quote']

Do you mean lots of resonances with lots of peaks and dips? Arent the resonances the most likely cause of the peaks and dips?

if you read enough of the reviews and the associated comments by JA you will begin to see how difficult it is to really measure a loudspeaker properly.

I dont think there is much doubt about this

FR is one aspect of a system' date=' and it's the aspect that your ears adjust to over time.

Personally, I think that within a certain range (perhaps +-10dB) it is one of the least important things to aim for.

[/quote'] This surprises me. Are you saying if a musician plays two notes at the same volume and they are recorded as such it is ok for a reproduction system to play one back up to 100 times as loud as the other ?

Cheers Mike

Hi Brett, Wish I really understood more about this .

As I understand it the great majority of peaks in a loudspeakers frequency response are caused by resonances.

If the frequency response is measured in such a way that the effects of resonances are reduced ( or made less visible ) and so the response flattened is this a measure of the " true " frequency response ?

A slow sweep response taken under anechoic conditions would seem to take into account the full effects of resonances but as you say for the resonance to be fully excited it needs a certain amount of energy and time of excitation. This may not happen when playing music .

I am not sure if in the swept sine wave situation the changing frequency is excited long enough for resonances to fully develop or not. Maybe it is. As I understand the purpose of the system is to remove room effects. Measurements made with time windowed impulse systems are open long enough even in fairly small rooms for some resonant behavior, at least at higher frequencies to be examined via waterfall plots. If the resonances can develop in the time allowed in impulse systems then maybe they can develop and affect frequency response in time allowed by swept sine wave type measurements.

As I write this seems less likely so we are left with a variety of frequency response measurements. Great. Back to the same old if make a measurement you better know how to interpret it and what it is actually good for. Cheers Mike

PS Got any good references about this stuff ?

Short answer - will expand when I'm back online after the move.

Sweeps do not remove room effects, but MLS will do above a certain frequency depending upon the window.

As you have seen in some CSD,some resonances are relatively long. On the sweep, the narrow band filter on the measured signal is to ensure that anything NOT at the frequency being input is rejected as all it is doing is measuring amplitude, at that F, at that moment. Because of this, a resonance excited can still be ringing away nicely and not show up because the filter has rejected it strongly because it's now looking at some other F, not the one at resonance.

Hello Mika, It is interesting given what Voecks and other Harman employees say about the correlation between measurements and listening that Revel publish so little in the way of measurement results, not even a basic on axis frequency response curve. I have actually twice tried to ask Revel about this but never received a reply.

Given who their customers are, I reckon publishing lots of info into general domain would not be to their advantage. Most audiophiles are anally retentive about the most trivial stuff, and do not understand what they are seeing, so data would just confuse and invite comparisons made on that data by people who can't do it. Those who are interested can read it in AES of published papers.

No, it doesn't.

My ears are my point of view.

I suspect that a good solid state amplifier and pre-amp measures "flatter" than my McIntosh.

I am almost certain that most good speakers have a better frequency response measurement than the Quad ESL57s.

However, I enjoy listening to this combination much more than theoretically better things, both ones I own, have owned, and have listened to.

I consider lack of distortion, speed and "energy" more important than frequency response, though obviously a respectable band of frequency has to be displayed with some degree of accuracy.

Hello Kye, Thanks for taking the trouble to write such a long and detailed response.

Agreed not ALL peaks and dips are resonance issues but I think most peaks in loudspeaker frequency response curves are caused by resonances.

This may be because the frequency response curve has insufficient resolution. If this came from Stereophile I believe the frequency response curves have a resolution of about 300Hz and this waterfall plot probably has considerably less than that as it is usual to reduce frequency resolution to increase time resolution for these plots. Evidence of this can be seen in Stereophile if you compare the frequency response plots to the highest curve of the waterfalll plot which is generally much smoother indicating less resolution in the frequency domain.

Do you mean lots of resonances with lots of peaks and dips? Arent the resonances the most likely cause of the peaks and dips?

I dont think there is much doubt about this This surprises me. Are you saying if a musician plays two notes at the same volume and they are recorded as such it is ok for a reproduction system to play one back up to 100 times as loud as the other ?

Cheers Mike

hi ando..

you may be right, that with enough resolution the FR will reveal where there are issues. so, if you have something that has no peaks/dips, it may indicate there are no resonance issues.

however, the above graphs indicate pretty clearly that there isn't a resonance exactly matching every dip and peak. I think, therefore, that if your hypothesis is true, then the FR would predict a resonance if there was only one, but when there is more than one, it might be more difficult to predict them. perhaps some kind of detailed analysis of the FR might do it, but it's not intuitive to the naked eye.

FR is a funny thing.

If you have ever listened to two drivers of greatly differing efficiency you will know immediately that frequency response is not the only element of good sound. my experiments (comparing a vifa 5" paper cone ~88dB and the Audax 6" paper cone ~100dB) confirmed this for me.

there are many ways to build a driver to be (relatively) flat, all of which produce very different sound. IMHO the traditional "make the cone stiff and make the surround damped" gives the worst sound, but of course it gives the best looking FR plot, and therefore sells speakers.

my comment about 10dB is a deceiving one, because there are some things you need to understand:

- a flat speaker, when put into a relatively normal room, will have a FR that can have peaks/dips of 20dB or more

- the human ear is sensitive to the area under the FR curve, not the height of it - if you have a peak 2Hz wide and 20dB tall it will be basically undetectable to the ear, whereas a 1dB peak that's an octave wide will be quite audible (this is why level matching in blind testing is very important)

- 10dB is a huge difference electrically, but only sounds like double the volume

taking into account the above, I maintain that (to my ears) FR optimised beyond a certain level is a low priority.

if you have a computer you can easily try some things yourself.

try:

- 1db boost between 50 and 200Hz

- 10db boost between 250Hz and 270Hz

- 20dB dip between 450 and 500Hz

- 80dB dip between 5000 and 5050Hz

next time you're on the phone for a while, swap ears with the handset and listen to the persons voice change EQ. FR is one thing that your ears adjust to, they don't adjust to other things like resonances, dulling of transients etc.

You can produce a huge variety of graphs to describe how speakers perform, some are more relevant and others are less..

Given who their customers are, I reckon publishing lots of info into general domain would not be to their advantage. Most audiophiles are anally retentive about the most trivial stuff, and do not understand what they are seeing, so data would just confuse and invite comparisons made on that data by people who can't do it. Those who are interested can read it in AES of published papers.

Brett, Thats a little cynical of you. I can see it makes sense from a marketing point of view but its frustrating for people who want to see their actual measurements for particular models in their range .These are not available via AES or anywhere else I know of.

To me, Harman claim that the difference between speakers can essentially be described in measurements. The measurement information they provide would indicate their lesser line speakers perform as well as their best. As the best cost twice as much or more and are said to sound better I wonder how the audible improvements are reflected in tyhe detailed measurements. Just me I guess.

Hope house moving isnt proving too irksome. Never much fun but I guess going from a rental to your own place is good motivation. One more tie to beautiful Sydney though! Cheers Mike

You can produce a huge variety of graphs to describe how speakers perform' date=' some are more relevant and others are less..[/quote']

Hi Kye, at least we are ageed about this !

Dont know if you hve seen www.stereophile.com/features/105kh/index1.html where Keith Howard discusses the problems of frequency response and waterfall plots taken with varying time windows . If not it may interest you.

Cheers Mike

Brett, Thats a little cynical of you. I can see it makes sense from a marketing point of view but its frustrating for people who want to see their actual measurements for particular models in their range .These are not available via AES or anywhere else I know of.

it may be cynical (and I think it is) but I suspect that it's true too.

I have heard that hifi circles in other countries (Singapore and lots of the rest of asia) are mostly made up of people who are in it to brag about how much their systems cost rather than what they sound like. according to this logic a system that cost half cannot be better than the system worth double.

I would suggest that along with cost, other factors that somewhat correlate with sound quality would also be used as the 'one and only' way to judge sound. FR is one of these things.

I have read accounts from sales people that when manufacturers started providing FR plots, their jobs became so much easier when trying to sell the expensive stuff. Previously when the sales person suggested the $10k speakers were better than the $3k ones the people were thinking 'the sales guy wants to hit his sales targets' but as soon as you throw in a little graph, when the sales guy says it's better you will believe him because there is some science involved as well.

if you look closely to product packaging you will be surprised to note the number of packages that have scientific looking graphs that are actually meaningless (no labelled or scaled axes, measuring the wrong thing, or obviously just made up to look the part). science is the religion of the western masses at the current time in history.

if you look closely to product packaging you will be surprised to note the number of packages that have scientific looking graphs that are actually meaningless (no labelled or scaled axes, measuring the wrong thing, or obviously just made up to look the part). science is the religion of the western masses at the current time in history.

Darn right, kye. Not that I am dismissing science, what gets my goat is that most people don't realize that science can be used to misinform as much as it enlightens. It takes a bit of sophistication to interpret a result and know what it tells you (or does not tell you). The mistake is to take that single result in isolation and magnify it beyond its actual importance.

Of course I am talking about my own field here, but I am sure it is common in many other fields of science.

Keith, Kye I dont think we are in disagreement. I think many people want to use " science " to support their preformed beliefs so when this doesnt happen they dismiss the bits they dont like. Of course in some cases "science " is used simply to sell things.

Both things seem to be rather prevalent in audio.

To me the best insurance against the misuse of science ( after all science does have a few entries on the positive side of the ledger ) is wider understanding and discussion/examination of processes and results. Bad science needs to be identified as such. I certainly agree sience as a religion would not be a good thing but nor can I see dismissing something as central to our lives as science as useful.

I think we are likely to get better audio equipment through trying to correlate what we hear with the physical attributes of components. This can lead to improving areas that most probably affect what we hear rather than improving things that dont. To do this I think we would be better served by scientific studies ( which shouldnt be above criticism just because someone claims they are "scientific " ) than by attribution based mostly on gut feeling or wishful thinking.

Cheers Mike

I certainly agree sience as a religion would not be a good thing - ??? Why not Mike, better than all the other religions.

I think we are likely to get better audio equipment through trying to correlate what we hear with the physical attributes of components. This can lead to improving areas that most probably affect what we hear rather than improving things that dont. To do this I think we would be better served by scientific studies ( which shouldnt be above criticism just because someone claims they are "scientific " ) than by attribution based mostly on gut feeling or wishful thinking.

Cheers Mike

Mike, I believe in psychoacoustic perception in audio. So I don't trust what most people (me included) say when describing a sound. It is just so easy to be wrong. eg more than a few people swear by cable burn-in when there is no evidence. Measurement is evidence. A measurement probably doesn't explain everything so one needs to know what and why one is measuring. While ears are a relatively accurate instrument, the full experience of just a few seconds of music seems to be too vast an amount to store reliably in memory. Just a few secs is too long to be reliable. ciao for nao

- the human ear is sensitive to the area under the FR curve, not the height of it - if you have a peak 2Hz wide and 20dB tall it will be basically undetectable to the ear

Hello Kye

That might be true when the speaker in question is fed with a typical music signal that covers a broad frequency range, such as a band, ensemble or orchestra ± singer, but for single instruments playing individual notes it isn't.

Consider the bottom few octaves of the piano, notes typically differ by anything between 2 to 5 Hz. When playing a real piano it is "easy" to hear the difference between adjacent notes which only differ by these 2 to 5 Hz. Telling the difference between these notes would be even easier if one were 20dB louder than the other.

The question then is, if a piano recording with simple keying in the lower octaves were played over a hifi system using

A. a speaker which had a dead flat frequency response and then

B. the same speaker with 20dB of narrow band (2Hz wide) EQ

would we (or the pianist!!) be able to tell the difference? The answer I suggest is YES.

This probably goes somewhat* to the heart of why people often comment that using recordings of the piano is a good tool in assesing a system. *Also having to reproduce all the harmonics and percussion makes the piano a good challenge as well.

This is also probably true of some of the tuned instruments played solo, like the organ or harp.

Of course if we are not listening to the lower octaves of the piano or organ and say a "typical" broad band music signal, I accept that it would be very hard to single out this small 2 Hz peak.

Best

JA

Edited May 1, 200916 yr by JA