ProSoundWeb Community

Sound Reinforcement - Forums for Live Sound Professionals - Your Displayed Name Must Be Your Real Full Name To Post In The Live Sound Forums => LAB Lounge => Topic started by: AllenDeneau on January 21, 2013, 03:36:24 PM

Title: Understanding speaker specs
Post by: AllenDeneau on January 21, 2013, 03:36:24 PM
As I was reading another thread it has become even more apparent to me that I know some, but certainly not as much as I would like, about most things, LOL, but especially, about speaker specs.

I have an old Peavey SP system consisting of SP2G's and SP118 subs. This question really isn't necessarily speaker specific but, more of an understanding relating to my speaker. Below will be a pic taken from my manual showing the frequency response graph. I understand what it's saying BUT, as dumb as this may sound, what frequencies do each vertical line represent? By counting it shows a total of 28 vertical lines, that must mean it's a 27 band eq?

According to the Yamaha Handbook it says that the lowest 3 frequencies are excluded from the 31 band in an 28 band eq. So am I to presume that 20, 25 and 31.5 are not reporesented in this graph?

Since I usually have a parametric or a 31 band eq, what I know is: a 31 band eq is a 1/3 octave eq, meaning every octave has 3 bands of eq for control. Every 10th frequency is a decade which divides the 31 bands into 3 decades. It appears to me the 10 to the powers, on the chart, are the decade frequencies of 200Hz, 2,000Hz and 20,000Hz BUT again, which of the 31 frequencies are missing from the 31 band eq that would tune that speaker?

So my biggest questions in relation to that graph and those specs are:
1. Which frequencies do the vertical lines represent being that a 1/3 octave eq has 31 bands and the graph has 27?
2. Why do the numbers start with 2 in each decade? (I feel like I should or do know that already)
3. Why does the spacing between the graphs vertical lines decrease when the frequencies increase?
4. Shouldn't that freq chart allow me to equalize that speaker box pretty close to flat by using that chart as a starting point? That'll allow me to account for the driver issues withing the box than to eq for the room a bit quicker?

I feel like a total bonehead asking this but it's driving me nuts and I can't seem to come up with a definite answer, at least I don't think I have.

Sorry and thanks for the help...

OK, one last question somewhat related to this subject...

Every octave is a doubling of frequency, so why do 31 band eq's shallow up their frequencies in the 2nd decade of the graph? 20, 40, 80, 160, 31.5, 630, 1.25, etc...

Even after they shallow up a bit, say 31.5 to 630, they're not exactly doubling until they get back to 2.5 to 5K. Ok, I guess I know they'll still control the same frequencies but why did the center frequencies move from exact double to just a bit off?

Weird or sad that all these years I never noticed or cared or thought about this until now...

Thanks
Title: Re: Understanding speaker specs
Post by: Joseph D. Macry on January 21, 2013, 04:24:48 PM
10-superscript-2 is 10 to the 2nd power = 100 Hz. Line to the left of that are 20 through 90 Hz.
10-3 (sorry, I can't type superscript) is 1000 HZ. Left of that are 200 through 900. 10-4 is 10,000 and the highest line is 20KHz.

So the vertical lines here are 20, 30, 40 ... 100, 200, 300, 400 ... 1K, 2K, 3K, 4K ... 10K, 20K.
True, these are not the 1/3 octave bands as found on most 31-band graphic EQs.

Graphic EQ bands generally center on ISO standard frequencies which are *approximately* 1/3 octave apart. When comparing to this plot, you may have to interpolate a bit. For instance, look between 600 and 700 to find response at 630Hz.

The vertical lines get closer as freq goes up because it is a logarithmic (base 10) plot.

Do not simply use this plot to set your graphic EQ. Use it to compare with other speakers, and to get an idea of the behavior of the speaker IN ITSELF. Your actual response will vary with the room, other gear in the signal chain, environmental factors etc etc. EQ according to RTA reading (or your ears, if very good and experienced) of pink noise played through the actual system in place.
Title: Re: Understanding speaker specs
Post by: Tim McCulloch on January 21, 2013, 04:32:08 PM
Macry beat me to it, and in fewer words... :)
Title: Re: Understanding speaker specs
Post by: AllenDeneau on January 21, 2013, 04:39:01 PM
10-superscript-2 is 10 to the 2nd power = 100 Hz. Line to the left of that are 20 through 90 Hz.
10-3 (sorry, I can't type superscript) is 1000 HZ. Left of that are 200 through 900. 10-4 is 10,000 and the highest line is 20KHz.

So the vertical lines here are 20, 30, 40 ... 100, 200, 300, 400 ... 1K, 2K, 3K, 4K ... 10K, 20K.
True, these are not the 1/3 octave bands as found on most 31-band graphic EQs.

Graphic EQ bands generally center on ISO standard frequencies which are *approximately* 1/3 octave apart. When comparing to this plot, you may have to interpolate a bit. For instance, look between 600 and 700 to find response at 630Hz.

The vertical lines get closer as freq goes up because it is a logarithmic (base 10) plot.

Do not simply use this plot to set your graphic EQ. Use it to compare with other speakers, and to get an idea of the behavior of the speaker IN ITSELF. Your actual response will vary with the room, other gear in the signal chain, environmental factors etc etc. EQ according to RTA reading (or your ears, if very good and experienced) of pink noise played through the actual system in place.

Thanks Joseph, that was helpful. Wonder why they don't measure and publish these specs as related to a 1/3 octave eq?

I undertood the center on ISO I guess I was just curious as to why they change as the frequencies go higher?

I knew I should've known that it was logarithmic :(

I agree that it shouldn't be the guide to set your eq but with a new speaker you're not used to I'd think it'll give you some sort of baseling reference to get you close to flat then using the ear to tune to the room etc... No? I've never used it but during my time staring at it I began to think, well, that's a pretty good start for flattening out major dips or peaks.. Maybe not huh?

Thanks.
Title: Re: Understanding speaker specs
Post by: Paul G. OBrien on January 21, 2013, 05:07:22 PM
Wonder why they don't measure and publish these specs as related to a 1/3 octave eq?
Because a 1/3 octave EQ is a pretty bad tool to try and correct a speakers response with, ideally you want the fully adjustable notch and shelving filters found in a DSP processor. But even with that and an RTA display to look at you're not going to get a ruler flat response no matter how many filters you apply, and that's because speaker response is a product of the amplitude and phase response of the drivers and crossover components and how all that interacts with the room it's in, so you simply cannot correct that with an amplitude only device like an EQ... and that's ignoring the phase altering side effects of those EQ filters that you have no control over.
Title: Re: Understanding speaker specs
Post by: AllenDeneau on January 21, 2013, 05:39:50 PM
Because a 1/3 octave EQ is a pretty bad tool to try and correct a speakers response with, ideally you want the fully adjustable notch and shelving filters found in a DSP processor. But even with that and an RTA display to look at you're not going to get a ruler flat response no matter how many filters you apply, and that's because speaker response is a product of the amplitude and phase response of the drivers and crossover components and how all that interacts with the room it's in, so you simply cannot correct that with an amplitude only device like an EQ... and that's ignoring the phase altering side effects of those EQ filters that you have no control over.

Great info, thanks Paul.

PS. I sent you a pm, about the sub question post I had, did you get it?
Title: Re: Understanding speaker specs
Post by: Scott Bolt on January 21, 2013, 10:02:41 PM
As I was reading another thread it has become even more apparent to me that I know some, but certainly not as much as I would like, about most things, LOL, but especially, about speaker specs.

I have an old Peavey SP system consisting of SP2G's and SP118 subs. This question really isn't necessarily speaker specific but, more of an understanding relating to my speaker. Below will be a pic taken from my manual showing the frequency response graph. I understand what it's saying BUT, as dumb as this may sound, what frequencies do each vertical line represent? By counting it shows a total of 28 vertical lines, that must mean it's a 27 band eq?

According to the Yamaha Handbook it says that the lowest 3 frequencies are excluded from the 31 band in an 28 band eq. So am I to presume that 20, 25 and 31.5 are not reporesented in this graph?

Since I usually have a parametric or a 31 band eq, what I know is: a 31 band eq is a 1/3 octave eq, meaning every octave has 3 bands of eq for control. Every 10th frequency is a decade which divides the 31 bands into 3 decades. It appears to me the 10 to the powers, on the chart, are the decade frequencies of 200Hz, 2,000Hz and 20,000Hz BUT again, which of the 31 frequencies are missing from the 31 band eq that would tune that speaker?

So my biggest questions in relation to that graph and those specs are:
1. Which frequencies do the vertical lines represent being that a 1/3 octave eq has 31 bands and the graph has 27?
2. Why do the numbers start with 2 in each decade? (I feel like I should or do know that already)
3. Why does the spacing between the graphs vertical lines decrease when the frequencies increase?
4. Shouldn't that freq chart allow me to equalize that speaker box pretty close to flat by using that chart as a starting point? That'll allow me to account for the driver issues withing the box than to eq for the room a bit quicker?

I feel like a total bonehead asking this but it's driving me nuts and I can't seem to come up with a definite answer, at least I don't think I have.

Sorry and thanks for the help...

OK, one last question somewhat related to this subject...

Every octave is a doubling of frequency, so why do 31 band eq's shallow up their frequencies in the 2nd decade of the graph? 20, 40, 80, 160, 31.5, 630, 1.25, etc...

Even after they shallow up a bit, say 31.5 to 630, they're not exactly doubling until they get back to 2.5 to 5K. Ok, I guess I know they'll still control the same frequencies but why did the center frequencies move from exact double to just a bit off?

Weird or sad that all these years I never noticed or cared or thought about this until now...

Thanks

As an aside ..... YUCK!  That is a pretty bad frequency response ;)
Title: Re: Understanding speaker specs
Post by: Paul G. OBrien on January 21, 2013, 11:22:39 PM
PS. I sent you a pm, about the sub question post I had, did you get it?
Nothing in my inbox....
Title: Re: Understanding speaker specs
Post by: David Morison on January 22, 2013, 07:34:01 AM
I agree that it shouldn't be the guide to set your eq but with a new speaker you're not used to I'd think it'll give you some sort of baseling reference to get you close to flat then using the ear to tune to the room etc... No? I've never used it but during my time staring at it I began to think, well, that's a pretty good start for flattening out major dips or peaks.. Maybe not huh?

Thanks.

To expand a little on what Paul said (and making wild assed assumptions to make this example work...), lets say you looked at that graph and decided to boost 1k6 on your graphic EQ to try and flatten the response there.

The problem is you don't know what's causing that dip there. If the crossover point of the box happens to be around there, then one possible cause for the dip would be a phase (or polarity) difference between the Mid and High drivers, causing cancellation of the acoustic output of the adjacent devices. If that is the case, then that phase/polarity difference will continue to cause cancellations even if you do stick a stronger signal in to the box at those frequencies, so you'll be making your amp and drivers work harder to no avail.
HTH,
David.

Edit - Absent Apostrophe Added
Title: Re: Understanding speaker specs
Post by: Frederik Rosenkjær on January 22, 2013, 08:56:21 AM
As an aside ..... YUCK!  That is a pretty bad frequency response ;)

Perhaps, if you're used to looking at smoothed graphs - if it is the unsmoothed plot of speaker I wouldn't consider it particularly bad?
Title: Re: Understanding speaker specs
Post by: Jonathan Johnson on January 22, 2013, 11:06:36 AM
10-3 (sorry, I can't type superscript)

FYI, the standard method for typing "powers" where you can't superscript, is to use a caret.

10^1=10
10^2=100
10^3=1,000
...
10^10=10,000,000,000

At least, that's what I learned in my math & science classes.

But if you're posting to this forum, you can encase the superscript text in [ sup ] tags (the "sup" button in the edit window does this); typing the following:

Code: [Select]
10[sup]4[/sup]=1000
gets you 104=1000
Title: Re: Understanding speaker specs
Post by: Tim McCulloch on January 22, 2013, 11:07:20 AM
Perhaps, if you're used to looking at smoothed graphs - if it is the unsmoothed plot of speaker I wouldn't consider it particularly bad?

Exactly.  Seeing unsmoothed response that is *mostly* ±5dB across the operating bandwidth is not too bad, especially for a brand/model that most gear snobs would ignore.

Allen, this isn't worth worrying about.  I did this over 25 years ago and EQ'd myself into less signal and no net audible improvement.  Besides, as has been pointed out, much of your perception of the speaker output will change in situ, so you'll be better off waiting until the speakers are in position, the band is playing, etc.
Title: Re: Understanding speaker specs
Post by: AllenDeneau on January 22, 2013, 12:30:31 PM
Exactly.  Seeing unsmoothed response that is *mostly* ±5dB across the operating bandwidth is not too bad, especially for a brand/model that most gear snobs would ignore.

Allen, this isn't worth worrying about.  I did this over 25 years ago and EQ'd myself into less signal and no net audible improvement.  Besides, as has been pointed out, much of your perception of the speaker output will change in situ, so you'll be better off waiting until the speakers are in position, the band is playing, etc.

Thanks Tim, good advice and makes sense, it just got me thinking, "why I don't use this graph's info more" is all really. Sometimes, ok a lot of times, I'll overthink some info I have and this is a result LOL!

So how does a person know if the reponse graph has been smoothed or if a nice and smooth response is a natural characteristic of that speaker? I just compared my OLD SP2G to a brand new VRX 15" monitor and for sure the VRX is smoother BUT is that after correction and smoothing or is that unsmoothed. Not to mention that technology is nearly 15 years newer than my SP2G...

To expand a little on what Paul said (and making wild assed assumptions to make this example work...), lets say you looked at that graph and decided to boost 1k6 on your graphic EQ to try and flatten the response there.

The problem is you don't know what's causing that dip there. If the crossover point of the box happens to be around there, then one possible cause for the dip would be a phase (or polarity) difference between the Mid and High drivers, causing cancellation of the acoustic output of the adjacent devices. If that is the case, then that phase/polarity difference will continue to cause cancellations even if you do stick a stronger signal in to the box at those frequencies, so youll be making your amp and drivers work harder to no avail.
HTH,
David.

Thanks David, I didn't even think about that. As you may recall from my subwoofer driver thread, speaker design and understanding it isn't something I'm, sadly, proficient with so I'd love to learn more so I can better predict the systems capabilities or process the info I'm hearing while setting up the PA. How would a person determine what's causing said valleys and peaks due to phase/polarity?

Because a 1/3 octave EQ is a pretty bad tool to try and correct a speakers response with, ideally you want the fully adjustable notch and shelving filters found in a DSP processor. But even with that and an RTA display to look at you're not going to get a ruler flat response no matter how many filters you apply, and that's because speaker response is a product of the amplitude and phase response of the drivers and crossover components and how all that interacts with the room it's in, so you simply cannot correct that with an amplitude only device like an EQ... and that's ignoring the phase altering side effects of those EQ filters that you have no control over.

Great points Paul, since I have a full Peavey SP system, minus the amps, that's why I bought the VSX26 DSP to have better control over settings and parameters. That's a big reason as to why I'm asking such questions like these. Since reloacting I've had little chance to run my system so I haven't even set the rig up with the VSX yet but, I plan on using the 3 way preset to start as it's gonna be a pretty good starting point, I'm told. Then I can continue to learn more of the finer points of system tuning vs grabbing an eq fader and going to town. At least I hope that's how it'll all play out.

I'll send you the message again, thanks.

FYI, the standard method for typing "powers" where you can't superscript, is to use a carat.

10^1=10
10^2=100
10^3=1,000
...
10^10=10,000,000,000

At least, that's what I learned in my math & science classes.

But if you're posting to this forum, you can encase the superscript text in [ sup ] tags (the "sup" button in the edit window does this); typing the following:

Code: [Select]
10[sup]4[/sup]=1000
gets you 104=1000


Thanks Jonathan, now I just need to find an excuse to use that info more, watch out for new posts, LOL! I knew there was a way, it's just been way too long since I've had a math class, man I'm old.
Title: Re: Understanding speaker specs
Post by: Brad Weber on January 22, 2013, 12:34:11 PM
Exactly.  Seeing unsmoothed response that is *mostly* ±5dB across the operating bandwidth is not too bad, especially for a brand/model that most gear snobs would ignore.
It's surprising how many speakers have published +/-XdB frequency response numbers that seem wishful thinking if you look at the associated response graph.  It's also not unheard of for speaker manufacturers to publish 'corrected' numbers or response graphs that include applying their recommended processing, something that is not always clearly identified.  What it comes down to is not just looking at the numbers but also understanding what those numbers represent.
Title: Re: Understanding speaker specs
Post by: Jonathan Johnson on January 23, 2013, 03:16:55 PM
typing the following:

Code: [Select]
10[sup]4[/sup]=1000
gets you 104=1000

and an "F" on your math paper. There should either be four zeroes there, or the power should be 3. My bad.
Title: Re: Understanding speaker specs
Post by: Scott Bolt on January 23, 2013, 08:10:33 PM
Perhaps, if you're used to looking at smoothed graphs - if it is the unsmoothed plot of speaker I wouldn't consider it particularly bad?

From here page 20 for a PRX612M:  http://www.jblpro.com/BackOffice/ProductAttachments/DOC_1672.pdf

Pretty flat.
Title: Re: Understanding speaker specs
Post by: Ivan Beaver on January 23, 2013, 09:01:38 PM
Here is an example of different amounts of smoothing applied to the same speaker response.

This is just a single driver I measured the other day.
Title: Re: Understanding speaker specs
Post by: Scott Bolt on January 26, 2013, 06:40:55 PM
Thanks Ivan.

So how is the smoothing performed?  Is it a rolling average or something similar?  It has been my experience that doing any smoothing shifts the curve you are smoothing.

In the specs we see on different speakers, is there anywhere we can tell how much smoothing was used?
Title: Re: Understanding speaker specs
Post by: Ivan Beaver on January 26, 2013, 09:48:47 PM
Thanks Ivan.

So how is the smoothing performed?  Is it a rolling average or something similar?  It has been my experience that doing any smoothing shifts the curve you are smoothing.

In the specs we see on different speakers, is there anywhere we can tell how much smoothing was used?
In the case I presented, it is simply a percentage.  The percent is in octave terms-so 100% is 1 octave (the graph with the most smoothing).  33% would be 1/3rd octave.

Another way of performing smoothing is to simply use less points in the measurement.

And then there is the "marketing smoothing" in which they just simply redraw the graph to be what they want it to be-no matter what the actual measurement is.  YES it has been done.

Or you can use a wider trace line or change the scale and so forth.

OF course in the old days of chart recorders-you just simply change the speed of the chart.

I guess the best way to find our for sure would be to look closely at the graph and see if it says-or in some small print or call the manufacturer and see if you could get through to somebody who could give you that information.
Title: Re: Understanding speaker specs
Post by: Scott Bolt on January 27, 2013, 11:55:24 AM
In the case I presented, it is simply a percentage.  The percent is in octave terms-so 100% is 1 octave (the graph with the most smoothing).  33% would be 1/3rd octave.

Another way of performing smoothing is to simply use less points in the measurement.

And then there is the "marketing smoothing" in which they just simply redraw the graph to be what they want it to be-no matter what the actual measurement is.  YES it has been done.

Or you can use a wider trace line or change the scale and so forth.

OF course in the old days of chart recorders-you just simply change the speed of the chart.

I guess the best way to find our for sure would be to look closely at the graph and see if it says-or in some small print or call the manufacturer and see if you could get through to somebody who could give you that information.

Ahh....

It would be my conclusion then that the curve should be published without any "smoothing".  It is simply becoming "not accurate" and certainly not representative in any way.
Title: Re: Understanding speaker specs
Post by: Ivan Beaver on January 27, 2013, 12:06:26 PM
Ahh....

It would be my conclusion then that the curve should be published without any "smoothing".  It is simply becoming "not accurate" and certainly not representative in any way.
Good luck finding any of those out there.

And remember that our ears do not hear "unsmoothed".  We have a built in "smoothing mechanism" that helps.

The example I posted was of a single driver-a multiway speaker is going to look a lot worse.

And even if you said "unsmoothed"-how many data points would you want?  that is a different way of smoothing.

You will never get a "standard" that everybody uses.  That is why it is important to be able to look at various graphs and try to recognize what is the same-what is different and what is important.

There does become a point at which unsmoothed data is almost useless-as there is to much detail.

Now if you are looking for a SPECIFIC thing in the data-then unsmoothed in a particular freq range can be a GOOD thing-but not overall across the full freq spectrum
Title: Re: Understanding speaker specs
Post by: Jonathan Johnson on January 27, 2013, 10:42:29 PM
The example I posted was of a single driver-a multiway speaker is going to look a lot worse.

I'll also point out that no single test of a single speaker is representative of the whole production of that model. A graph representing the response curve as included in product literature should be an average representation of several samples. The more samples you have, the smoother the graph will become. Even so, if every unit of a particular model exhibits a deep, narrow dip in response at 12 kHz for example, it is important to display that on the graph; overzealous smoothing may cause that dip to appear as an insignificant ripple.

The best we could hope for is a graph that presents an average response curve with dips and spikes typical of the model, and also includes curves representing deviation. Of course, now we have to understand what those high and low deviation curves mean, and that may be much more difficult to represent and interpret. Marketing people don't want customers to have doubt, and deviation specs create doubt (at least in the minds of the marketers).
Title: Re: Understanding speaker specs
Post by: AllenDeneau on January 28, 2013, 04:54:47 PM
great explanation guys. Is it safe to assume that what's being said is every manufacturer has their own test methods and publish what they wish to publish and it's up to us to discern it's validity?

Brand A may do multiple tests and then average their results and present it in a non-smoothed fashion but, brand B may only do 2 tests and then use the most beneficial smoothing results and, brand C takes the data straight from the manufacturer of their components and publishes them without ever testing them in their box design and so on with each different speaker brand?

You will never get a "standard" that everybody uses.  That is why it is important to be able to look at various graphs and try to recognize what is the same-what is different and what is important.

So that's my next new quest, how to determine the above... The answer was presented to my original question, now how do I interpret the results? I believe the best case scenario is to be able to audition each "box" myself and measure them out but, what if you don't have that luxury? Can you really hold the plots as useful and beneficial info as a whole?

Thanks for the info guys, it's been a real learning experience and I truly love to learn. Now I'm trying to understand why I avoided this learning for so long  :-[
Title: Re: Understanding speaker specs
Post by: Ivan Beaver on January 28, 2013, 08:20:29 PM
Here is an example of using no smoothing-but changing the 'data points".  I am showing 2 graphs.  The first one is just a few samples and one with more samples.

The second graph still shows both of those-but also has one with a lot of samples.  Notice how "cluttered" it gets.

I have never seen any manufacturer show data that looks like that-even though it is the most accurate.

Sometimes to much information is not a good thing.
Title: Re: Understanding speaker specs
Post by: Scott Bolt on January 29, 2013, 10:44:30 PM
Here is an example of using no smoothing-but changing the 'data points".  I am showing 2 graphs.  The first one is just a few samples and one with more samples.

The second graph still shows both of those-but also has one with a lot of samples.  Notice how "cluttered" it gets.

I have never seen any manufacturer show data that looks like that-even though it is the most accurate.

Sometimes to much information is not a good thing.

Sure, but both of these sets of graphs show the data I would look to see.  They show the peaks and valleys in the response clearly.  Taking this graph and showing anything that looks "flat" would simply be bogus.
Title: Re: Understanding speaker specs
Post by: Ivan Beaver on January 30, 2013, 07:30:33 AM
Sure, but both of these sets of graphs show the data I would look to see.  They show the peaks and valleys in the response clearly.  Taking this graph and showing anything that looks "flat" would simply be bogus.
so except for your own measurements-how many manufacturers are showing data that looks anything like unsmoothed data?

It all depends on what the particular person is looking for.  For a designer-there is a lot to be gained from unsmoothed data.

For the casual buyer-the unsmoothed data would turn them away-since they want something that is "smoother" sounding.  Even though they have no idea what they are really looking at.
Title: Re: Understanding speaker specs
Post by: Scott Bolt on January 30, 2013, 09:29:03 PM
so except for your own measurements-how many manufacturers are showing data that looks anything like unsmoothed data?

It all depends on what the particular person is looking for.  For a designer-there is a lot to be gained from unsmoothed data.

For the casual buyer-the unsmoothed data would turn them away-since they want something that is "smoother" sounding.  Even though they have no idea what they are really looking at.

... [chuckles]...

You are making my point for me.  So there is no mathematical explanation for the frequency response graphs most manufacturers show.  It is just another marketing tool much like how many "watts" a powered speaker has  :o

You are correct that few would understand the graph and how it may or may not translate into a good sounding speaker.... but that is still no reason to "fake" the plot to get a more marketing friendly response curve.

I would just assume that any data being trumpeted as "engineering data" actually be real engineering data.  While I am an engineer, I suspect that any sound professional would also prefer that the data be based on real readings instead of marketing's dream of what they wished the data looked like ;)
Title: Re: Understanding speaker specs
Post by: Tim McCulloch on January 31, 2013, 12:03:53 AM
... [chuckles]...

You are making my point for me.  So there is no mathematical explanation for the frequency response graphs most manufacturers show.  It is just another marketing tool much like how many "watts" a powered speaker has  :o

You are correct that few would understand the graph and how it may or may not translate into a good sounding speaker.... but that is still no reason to "fake" the plot to get a more marketing friendly response curve.

I would just assume that any data being trumpeted as "engineering data" actually be real engineering data.  While I am an engineer, I suspect that any sound professional would also prefer that the data be based on real readings instead of marketing's dream of what they wished the data looked like ;)

I think the issue is "how much information is useful"?  Looking at the higher resolution data shows you a lot more, but to the "average" buyer (and a number of presumably more sophisticated ones) the extra data points will be misinterpreted or given far too much importance.

It's very easy to make bad sounding systems by applying too much correction to too many data points, particularly if that data is from a single test sample and not the unit in situ.  Don't ask me how I know.  ;)

Title: Re: Understanding speaker specs
Post by: Ivan Beaver on January 31, 2013, 07:32:58 AM

I would just assume that any data being trumpeted as "engineering data" actually be real engineering data.  While I am an engineer, I suspect that any sound professional would also prefer that the data be based on real readings instead of marketing's dream of what they wished the data looked like ;)
But a real engineer will not want to look at the full freq response in an unsmoothed format.

When doing real engineering-you will zoom in on particular areas unsmoothed to get a better idea what is going on.  Not the whole thing.

Look at it this way-look at the door in your room-it doesn't matter.  Now step up to it and put your eye as close as you can to it-does it look different?  Now let's take a magnifying glass and look at it.  How about an electron microscope?

If you want the REAL data-then the electronmicroscope is the way to go...

But what if the question is -"Have I sanded the door smooth enough to to paint?".

If you look closely you will still see ALL KINDS of imperfections-and you will never get them out.  But for the intended purposes-eyeballing it several feet away is just fine.

HOWEVER-if you are a paint engineer and are looking at the surface to see how well different formulations of paint will stick to different surfaces-then looking a lot closer is important-but not to the average person.

There is a thing as to much data.

As the old saying goes "What are we here to do?".  What do we really need to see?
Title: Re: Understanding speaker specs
Post by: Brad Weber on January 31, 2013, 09:28:24 AM
I would just assume that any data being trumpeted as "engineering data" actually be real engineering data.  While I am an engineer, I suspect that any sound professional would also prefer that the data be based on real readings instead of marketing's dream of what they wished the data looked like ;)
A real Engineer is also going to be looking at polars or balloon data, going to want better information about maximum power (and possibly ask for voltage and current information rather than Watts), will want to see sensitivity versus frequency, will want to see phase data and possibly other technical factors that probably don't matter to or may even discourage the average purchaser.  That has always been the dilemma, providing technical data that is complete and accurate enough for some but not overwhelming or intimidating for others.
 
I think what Tim was referencing was something it took me a while to understand (and that I still am learning) and that is differentiating what is important from what is not.  I have certainly been guilty of spending hours worrying about some little anomaly in the response that probably had no practical impact - and all too often at the expense of issues that did matter.
Title: Re: Understanding speaker specs
Post by: Ivan Beaver on January 31, 2013, 10:34:41 AM
I have certainly been guilty of spending hours worrying about some little anomaly in the response that probably had no practical impact - and all too often at the expense of issues that did matter.

Agreed

That whole "Can't see the forest for the trees" thing.

We all get hung on the "little things that don't matter in the grand scheme of things" from time to time.
Title: Re: Understanding speaker specs
Post by: AllenDeneau on January 31, 2013, 12:53:15 PM
WOW, thanks all I've certainly learned a few things as well as Ivan pointed out, I have more questions as well..

I understand that, for my current level of service and equipment, that the graphs are of very little use to me as I'm not educated on what they really mean to really analyze. However, the graph along with the Polars, as Brad referred to, can help me take a quick snapshot of it's general characteristics and coverages correct?

By having some understanding of where certain things lie in the frequency spectrum the polars will help me determine if I can get good intelligibility from a single speaker in x sized width of room. Is that correct?

The graph, whether it's been smoothed or not, really helps me determine where my LP and HP filters should be applied in DSP and a general look at the speakers "real" Max SPL, given the graph is somewhat truthful. Is that correct?

Like Brad said, I too have been victim of concerning myself with issues that really aren't an issue in the scheme of things however, I seem to do that in most parts of my life, not just looking at speaker response graphs  :-\ For example, I've over thought and over thought my ideas on what to concentrate on as I'm restarting my business after moving over 500 miles from home, I just need to grab and go...

OK, so the $64,000 question is: how to determine if the plots have been smoothed a little, a lot, not at all or if the marketing department had a whack at them? Is it just a call to the relative manufacturer to get such info?

Now, I'd love to actually discuss the graph I posted regarding my OLD SP2G's. I know they aren't the best sounding speaker out there and weren't even in their day but I don't think they're completely horrible either. I've been able to get them fairly smooth with eq but it would be nice to have something a little flatter right from the get go. With that said, until I get re-established and can upgrade, these have to do and I'm ok with that for the most part.

How do I know if the freq response I posted has been smoothed and if it hasn't, is it a really bad response?

Some of you are so much further ahead of me in the game that I'm only hoping to ever acheive even a portion of your understanding. With that being said, I'm asking, if you were speaker shopping, back in the mid '90's when these were new lol, and you looked at the freq response and KNEW it was an unsmoothed graph, what would your thoughts be?
Well, everyone BUT Scott as I already know what he thinks  ;)

What I'm hoping to acheive by asking the last part is, I'd like to see through "your eyes" what the info says, helping me to develope my own "filters" of what info is really saying.. Make sense?

Here it again to save you from going back through the thread.
Title: Re: Understanding speaker specs
Post by: Ivan Beaver on January 31, 2013, 07:13:51 PM
WOW, thanks all I've certainly learned a few things as well as Ivan pointed out, I have more questions as well..

I understand that, for my current level of service and equipment, that the graphs are of very little use to me as I'm not educated on what they really mean to really analyze. However, the graph along with the Polars, as Brad referred to, can help me take a quick snapshot of it's general characteristics and coverages correct?

By having some understanding of where certain things lie in the frequency spectrum the polars will help me determine if I can get good intelligibility from a single speaker in x sized width of room. Is that correct?

The graph, whether it's been smoothed or not, really helps me determine where my LP and HP filters should be applied in DSP and a general look at the speakers "real" Max SPL, given the graph is somewhat truthful. Is that correct?

Like Brad said, I too have been victim of concerning myself with issues that really aren't an issue in the scheme of things however, I seem to do that in most parts of my life, not just looking at speaker response graphs  :-\ For example, I've over thought and over thought my ideas on what to concentrate on as I'm restarting my business after moving over 500 miles from home, I just need to grab and go...

OK, so the $64,000 question is: how to determine if the plots have been smoothed a little, a lot, not at all or if the marketing department had a whack at them? Is it just a call to the relative manufacturer to get such info?

Now, I'd love to actually discuss the graph I posted regarding my OLD SP2G's. I know they aren't the best sounding speaker out there and weren't even in their day but I don't think they're completely horrible either. I've been able to get them fairly smooth with eq but it would be nice to have something a little flatter right from the get go. With that said, until I get re-established and can upgrade, these have to do and I'm ok with that for the most part.

How do I know if the freq response I posted has been smoothed and if it hasn't, is it a really bad response?

Some of you are so much further ahead of me in the game that I'm only hoping to ever acheive even a portion of your understanding. With that being said, I'm asking, if you were speaker shopping, back in the mid '90's when these were new lol, and you looked at the freq response and KNEW it was an unsmoothed graph, what would your thoughts be?
Well, everyone BUT Scott as I already know what he thinks  ;)

What I'm hoping to acheive by asking the last part is, I'd like to see through "your eyes" what the info says, helping me to develope my own "filters" of what info is really saying.. Make sense?

Here it again to save you from going back through the thread.
Some answers.  Polar are not easy to read/understand by most people.  If you think a freq response graph is hard-then polars are out of the question.

Polars will give you no information regarding intelligibility.    Just the coverage angle of the loudspeaker at different freq.  Intelligibility is a combination of noise-reflections-phase response-interactions between different loudspeakers-freq response-listening position and so forth.

If the graph looks like the one you show-then it has been smoothed.  If you don't see some radical "spikes" like I presented-then it has been smoothed.  Good luck getting any more details from the manufacturer than they present on the website or on the spec sheet.  But it might be possible.

There is generally no way to tell if the graph presented is "correct" .  I know of cases where the manufacturer simply draws the graph the way they want it to look.

You need to look closely at whether or not any processing has been applied to get the graph.  If it has-then this can RADICALLY affect the max SPL-ESPECIALLY on the low end.  It is pretty common to add a low freq boost to make the response appear to do lower.  It will-until you turn it up.  Whatever was added by boosting-has to be subtracted from the max SPL-At the boosted freq.

The freq response is often only a small indicator of how a loudspeaker sounds.  You can have 2 speakers that measure almost exactly (within 0.25dB) and sound COMPLETELY different.  There is a lot more than just the raw on axis freq response.

Don't expect to learn everything quickly.  Many of us have spent decades in the business on many different sides-and are just beginning to get a small handle on what is going on.  And we get blindsided all the time.

But to put another answer to the "Is it smoothed or not", I can say that in 99.999999999% of the cases it IS smoothed.  I have never seen any manufacturer post any unsmoothed data.  It is simply not in their best interest to do so.  And if they did-then what would you get from it?  And compare it to?

Anybody who "needs" unsmoothed data is going to have their own way to measure it-and surely would not trust any data provided by the manufacturer.

We apply various amounts of smoothing, not in an attempt to "hide anything", but rather to provide some useful information to the viewer.

Regarding trying to use a single graph to try and determine filters-remember that the graph is usually taken in a reflection free environment.  Your usage is probably not that.  You will have boundary loading (which can increase low freq response)-all sorts of reflections, interactions with other loudspeakers and so forth.

Measurements in the existing situation are the ONLY way to do it correctly.  And then you have to do multiple measurements and average.

Don't look for simple answers.  That will result in simple WRONG answers.
Title: Re: Understanding speaker specs
Post by: Scott Bolt on January 31, 2013, 07:28:40 PM
A real Engineer is also going to be looking at polars or balloon data, going to want better information about maximum power (and possibly ask for voltage and current information rather than Watts), will want to see sensitivity versus frequency, will want to see phase data and possibly other technical factors that probably don't matter to or may even discourage the average purchaser.  That has always been the dilemma, providing technical data that is complete and accurate enough for some but not overwhelming or intimidating for others.
 
I think what Tim was referencing was something it took me a while to understand (and that I still am learning) and that is differentiating what is important from what is not.  I have certainly been guilty of spending hours worrying about some little anomaly in the response that probably had no practical impact - and all too often at the expense of issues that did matter.

Sure.  A Nyquist plot would likely be a useful tool to an engineer ..... not so much to others.

I would say that you can't get much of a story from the frequency plot in general.  You aren't going to hear how it sounds by looking at the plot.

To the OP.... from looking at the graph .... I would say it is a little weak on the bottom, has a bit of a hole at ~1.8K which may take away from vocal clarity .... or keep the speaker from sounding harsh ;)  The peak at 500 may make the speaker sound a little "honky".

Since you are using these with subs, normally, I would have suggested a cross-over frequency around 100Hz .... but this speaker appears to have a pretty decent dip right at 100Hz.  There is generally a little notch at the cross-over point anyway, this might make it worse.

I would like to say that there is much more to sound than a frequency plot at 1M on axis.

If you have a digital cross-over like the DRPA, you can setup several presets and A/B them to see what you like, and what you don't.  If not .... well, it is a little harder to change your graphic eq so quickly ;)

I had a friend who works live sound for a living come over one night and ring out my system for me for a steak dinner and a few beers.  That was surely worth every penny ;)