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 Subwoofer Forum => Topic started by: Helge A Bentsen on May 02, 2017, 02:51:52 AM
-
Is there a reliable and practical method of determining the real-world SPL from a subwoofer "in the field"?
I'm building a prototype for a sub and wish to have some idea about how much SPL it put's out within it's intended bandwidth.
-
My usual test method is this:
- Set processing etc as normal.
- Play pink noise
- Increase volume until bad sounds start, or the amp clips/limits
- Record SPL
If you want more complicated than that, you need to start reading here: http://www.data-bass.com/systems
Chris
-
Is there a reliable and practical method of determining the real-world SPL from a subwoofer "in the field"?
I'm building a prototype for a sub and wish to have some idea about how much SPL it put's out within it's intended bandwidth.
It depends.
Are you looking for the maximum SPL that it can produce at some freq?
Or the maximum it can produce in the intended bandwidth?
Are you looking for the maximum "listenable SPL"?
What are you using to determine this SPL? What is the response time of the meter? What is the weighting used?
You can easily come up with a different answer to each of the these questions-from the same speaker cabinet.
So first you have to narrow down the meaning of the question. Different people can see the same question different ways.
-
Helge, this may be what you are looking for
http://www.prosoundtraining.com/site/synaudcon-library/the-loudspeaker-toaster-test-revisited/
Sent from my iPhone using Tapatalk Pro
-
It depends.
Are you looking for the maximum SPL that it can produce at some freq?
Or the maximum it can produce in the intended bandwidth?
Are you looking for the maximum "listenable SPL"?
What are you using to determine this SPL? What is the response time of the meter? What is the weighting used?
You can easily come up with a different answer to each of the these questions-from the same speaker cabinet.
So first you have to narrow down the meaning of the question. Different people can see the same question different ways.
Also, what SPL at what distortion level?
(Sigh)...It's never easy ::)
-
Also, what SPL at what distortion level?
(Sigh)...It's never easy ::)
What distortion at what freq? The distortion number changes with freq-kind of like the impedance changes with freq.
A "simple single number" does not accurately describe the whole operation.
-
Helge, this may be what you are looking for
http://www.prosoundtraining.com/site/synaudcon-library/the-loudspeaker-toaster-test-revisited/
Sent from my iPhone using Tapatalk Pro
That looked similar to what I've been thinking about. Thanks for the link :)
Ivan: I wish to address two things: RMS limiter threshold to keep the sub from releasing it's magic smoke and a number I can tell people who will ask "how loud is it". I'm fully avare that this is not a simple question, so let's try to narrow it down.
I have Smaart 7, a DPA 4007, a SPL calibrator and a Powersoft K20.
I wish to be able to measure the maximum SPL the sub can produce for an extended time (3minutes or more) without failure within it's intended bandwith(hp/lp and eq in place), preferably with the same kind of accurancy that any reliable manufacturer would have on a SPL number in the datasheets for a high quality subwoofer.
-
http://www.etcinc.us/assets/nl061_loudspeaker_toaster.pdf
-
That looked similar to what I've been thinking about. Thanks for the link :)
Ivan: I wish to address two things: RMS limiter threshold to keep the sub from releasing it's magic smoke and a number I can tell people who will ask "how loud is it". I'm fully avare that this is not a simple question, so let's try to narrow it down.
I have Smaart 7, a DPA 4007, a SPL calibrator and a Powersoft K20.
I wish to be able to measure the maximum SPL the sub can produce for an extended time (3minutes or more) without failure within it's intended bandwith(hp/lp and eq in place), preferably with the same kind of accurancy that any reliable manufacturer would have on a SPL number in the datasheets for a high quality subwoofer.
Hello,
Chris already posted a link to my site earlier, the link is in my sig, but I would suggest browsing around there starting with the "know how" section which briefly describes the equipment, measurements used and then check a few links to the subwoofer measurements. The "know how" section is a little out of data but it gives a close enough idea to how and why the measurements are taken. I have been doing something similar to what you are asking about for about 10 years. A quick tip is that you need to click on the system name and then the measurements tab to get to the full measurements packet for each unit. It can be a bit of information overload but as mentioned by others the correct answer will never be a simple 1 data point answer. At minimum there would have to be a number of qualifiers to that number. Anyway if you have any questions I'd be happy to help. I wish there were many more people qualifying their speakers in this manner.
-
Most often the max spl data is calculated.
Honestly i would not bother trying to measure the real max data because as a figure has very little to the real world usage.
What is the most important is the sustained spl that it can produce with AES power rating applied to the drivers.
Then there is the program power handling and program SPL. Now this is now very tricky mater. Program power handling is always calculated figure by simply doubling the power. But things are far from this simple.
Music material varies in dynamics, and while live gigs can have great dynamics and program power is not a problem to the drivers, modern recorded music is quite compressed and average dynamic is quite smaller, thus average power sent to the drivers is much higher. Meaning they will most probably blow up with program power applied.
And all this does not account for the power compression. So the max spl figure to me is a pure fiction in real world use. But everyone love putting big numbers because they look good on paper. This force everyone to do the same, or they then look weak on paper.
Its a spec war there, and it is ugly.
-
Its a spec war there, and it is ugly.
Very true.
I spent a little while yesterday arguing with a member of a sales team that a single 10" direct-radiating box cannot possibly output 136dB and still sound good, pointing out that a 10" driver on the ground would need to move a little over 3" one-way to produce that SPL at 80Hz. They were adamant that the speaker would hit it's ratings. I don't care if the box does have a well-regarded name on the front. Physics is physics.
I was encouraged to talk to the R&D department, so I might post up a thread if anything comes of that.
Chris
-
Very true.
I spent a little while yesterday arguing with a member of a sales team that a single 10" direct-radiating box cannot possibly output 136dB and still sound good, pointing out that a 10" driver on the ground would need to move a little over 3" one-way to produce that SPL at 80Hz. They were adamant that the speaker would hit it's ratings. I don't care if the box does have a well-regarded name on the front. Physics is physics.
I was encouraged to talk to the R&D department, so I might post up a thread if anything comes of that.
Chris
OF course it always begs the questions-at what distance?-At what freq?-for how long?-using what meter?
A small firecracker can produce lot of SPL-but not for very long.
I just realized that the well respected NTI XL2 meter has a "flaw" (in my opinion) regarding the peak SPL scale.
It actually measures the peak of the sound envelope-while the other scales measure the "RMS" of the envelope.
So the Peak readings are 3dB higher that what I would consider the "peak", which would be the maximum level.
I know others will disagree, and argue about the meaning of "peak".
I will agree with them to some respect-but disagree about the "intended meaning of the reading".
So simply by using a meter that reads like this-you get a "free 3dB" of extra output to put on a spec sheet.
HOW you measure will GREATLY affect the actual readings.
A typical "hardish" show can easily produce differences of 30dB (or greater) between different scales/weighting etc when measuring the same signal.
So how loud is it? All of them-it depends on how it is being measured
-
you can always crank it up until you hear the vc bottom and then back off a tad. I dont know how many times you could do that before its broke and I aint gonna try, especially not with a $600.00 driver.
-
you can always crank it up until you hear the vc bottom and then back off a tad. I dont know how many times you could do that before its broke and I aint gonna try, especially not with a $600.00 driver.
I did this by mistake and folded the cone on a Beyma 15P1200Nd. No highpass filter, lots of power, and the software started the sweep an octave down on the start value I entered. The 2" peak-to-peak travel was impressive, though, and my SPL meter read >133dB.
Chris
-
you can always crank it up until you hear the vc bottom and then back off a tad. I dont know how many times you could do that before its broke and I aint gonna try, especially not with a $600.00 driver.
That only works on the excursion limit, but in many cases the driver could burn out long before it hits the excursion limit.
It depends on the particular freq of interest.
Sometimes excursion and power are related-at other freq they are not.
-
After inital tests, it seems like my little sub will do at least 144dB peak.
Not bad for a homebrew dual 18".
-
After inital tests, it seems like my little sub will do at least 144dB peak.
Not bad for a homebrew dual 18".
Interesting, you've beaten the simulations for a single high-power 18" with 120v RMS input in a 11cu.ft (internal) tapped horn by around 10dB.
What were the conditions where you measured that SPL?
Chris
-
I took a pair of active dual 15" subs from a known manufacturer and feed them with a signal untill an increase in drive level didn't increase the output from them as verified by Smaart.
I recorded the level and was able to match the same output with my sub. I didn't test where it goes into hard limiting, just that I could match the output.
The Dual 15" has a SPL Peak rating of 138dB, and a pair of them sums +6dB on my Smaart screen.
-
I took a pair of active dual 15" subs from a known manufacturer and feed them with a signal untill an increase in drive level didn't increase the output from them as verified by Smaart.
I recorded the level and was able to match the same output with my sub. I didn't test where it goes into hard limiting, just that I could match the output.
The Dual 15" has a SPL Peak rating of 138dB, and a pair of them sums +6dB on my Smaart screen.
Hi Helge, when quoting SPL's, I really believe in using a mic calibrator....otherwise I feel like i'm building off "he said, she said"
And they are fairly cheap.
Admittedly, I have no idea how well a calibrator dialed in at 1KHz (which is all i've ever seen) behaves in sub range, but it's gotta be better than comparing one box against another box'es specs, no matter how well regarded the manufacturer is, yes?
-
Is there a reliable and practical method of determining the real-world SPL from a subwoofer "in the field"?
I'm building a prototype for a sub and wish to have some idea about how much SPL it put's out within it's intended bandwidth.
While not answering your question, assuming you have no limit on power and the sub does not fall apart, the loudest sound would be 194 dBm. Above that it is no longer sound, it is a shock wave.
-
Hi Helge, when quoting SPL's, I really believe in using a mic calibrator....otherwise I feel like i'm building off "he said, she said"
And they are fairly cheap.
Admittedly, I have no idea how well a calibrator dialed in at 1KHz (which is all i've ever seen) behaves in sub range, but it's gotta be better than comparing one box against another box'es specs, no matter how well regarded the manufacturer is, yes?
My Smaart setup is calibrated with a mic calibrator. I verify this before I start any spl measurement.
-
I took a pair of active dual 15" subs from a known manufacturer and feed them with a signal untill an increase in drive level didn't increase the output from them as verified by Smaart.
I recorded the level and was able to match the same output with my sub. I didn't test where it goes into hard limiting, just that I could match the output.
The Dual 15" has a SPL Peak rating of 138dB, and a pair of them sums +6dB on my Smaart screen.
I understand your method, but I promise that the 2x15" did not hit 138dB in the subwoofer band, no matter what the manufacturer says they can do.
I'm part-way through writing an article about this, but here's a glimpse into what manufacturers might do:
- 1x 15" speaker in a ported box, 1KW amplifier.
- At 40Hz, it's 93dB@1w input. The in-band response slopes a little, so it's 97dB @ 100Hz. There's also a cone break-up peak at 1.2kHz, where sensitivity reaches 103dB @1w.
- Even ignoring cone excursion issues and power compression, what's my maximum SPL?
- 40Hz: 123dB; 100Hz: 127dB; 1.2kHz: 133dB. Which number goes on the box?
Yep, you guessed it, this sub can technically make a sound that has an SPL of 133dB. It's at one frequency that the sub will never be used at, but that really won't stop the marketing team.
To produce the 138dB at 50Hz, the pair of 15" cones would need to move 3" one-way, or 6" peak-to-peak. This driver (https://www.parts-express.com/re-audio-xxx15-v2-4-15-cast-frame-dvc-subwoofer-driver-4-4-ohm--268-8153) won't manage that (it's about 1" short on cone travel), and it's the most bonkers 15" I can find.
Chris
-
My Smaart setup is calibrated with a mic calibrator. I verify this before I start any spl measurement.
Then I'm curious what smaart said the SPL of your 15's" were, when they were run at max.
And what type SPL...and speed?
FWIW, a real world usage measurement I've come to like, is Smaart's LEQ, using pink band-passed with the actual HP and LP filters I intend to use, in place. IMO, this gives an average that is a decent proxy for "a one number meaningful spec".
And taken one step further, I measure LEQ with whatever eq I intend to use to smooth response.
I mean, that's where the sub is going to live, with HP, LP, and eq.....what can it do there?
Using a good RMS voltmeter with averaging, vs the LEQ readings, ....then power handling and even a sensitivity specs make sense to me.
Again FWIW. I know how much experience you have, please advise if you see anything wrong with this methodology.
-
Then I'm curious what smaart said the SPL of your 15's" were, when they were run at max.
And what type SPL...and speed?
A simple SPL number on most subs doesn't really tell a whole lot.
Since most subs have a sloping response, they are louder at the higher freq in their range than at the lower freq.
So saying a number-any number MUST be attached to some freq or graph, or it is pretty meaningless.
As simple as most people want SPL to be, it is NOT an easy one number answer.
-
A simple SPL number on most subs doesn't really tell a whole lot.
Since most subs have a sloping response, they are louder at the higher freq in their range than at the lower freq.
So saying a number-any number MUST be attached to some freq or graph, or it is pretty meaningless.
As simple as most people want SPL to be, it is NOT an easy one number answer.
Hi Ivan, I kindly disagree. And I get all the games being played with single number SPL.
IMHO, a LEQ taken with HP and LP filters in place representing intended usage, with their exact parameters provided,
along with average RMS voltage that corresponds to the LEQ period,
is a pretty dang good single number.
Again, what I've been advocating in a number of threads, and what I've seen Tom suggest as well.
-
FWIW, a real world usage measurement I've come to like, is Smaart's LEQ, using pink band-passed with the actual HP and LP filters I intend to use, in place. IMO, this gives an average that is a decent proxy for "a one number meaningful spec".
And taken one step further, I measure LEQ with whatever eq I intend to use to smooth response.
I mean, that's where the sub is going to live, with HP, LP, and eq.....what can it do there?
Using a good RMS voltmeter with averaging, vs the LEQ readings, ....then power handling and even a sensitivity specs make sense to me.
Again FWIW. I know how much experience you have, please advise if you see anything wrong with this methodology.
That's actually a rather nice idea.
The reason I started this thread is that I have for years heard people bashing manufacturers with "your SPL numbers are wrong!" but I've not seen too many helpful ideas about how to find a meaningful SPL number for real world usage.
I'm not going to sell this sub to anyone, but I'm genuinely interested in knowing how loud it is within it's operating range.
Hence the reason for using a pair of known subs as a benchmark. Now I have a known comparison.
But yes, generally speaking. Someone should write a paper on how to do this properly with the least amount of gear and know-how needed. Why? Because it can't be rocket science to measure a SPL number.
I mean, I own several things in my life that can tell me how fast I'm moving, the altitude above sea level, distance walked etc. Why can't we do the same with SPL instead of shouting "SPl numbers are wrong"?
-
Hi Ivan, I kindly disagree. And I get all the games being played with single number SPL.
IMHO, a LEQ taken with HP and LP filters in place representing intended usage, with their exact parameters provided,
along with average RMS voltage that corresponds to the LEQ period,
is a pretty dang good single number.
Again, what I've been advocating in a number of threads, and what I've seen Tom suggest as well.
I would agree with you. But how often do you see spec measurements with HP and LP filters in place? That would throw the numbers off (and not in a good way.
I do agree that this would give a better idea of the loudness in the intended range. But would still "favor" the freq that are higher in level.
-
That's actually a rather nice idea.
The reason I started this thread is that I have for years heard people bashing manufacturers with "your SPL numbers are wrong!" but I've not seen too many helpful ideas about how to find a meaningful SPL number for real world usage.
I'm not going to sell this sub to anyone, but I'm genuinely interested in knowing how loud it is within it's operating range.
Hence the reason for using a pair of known subs as a benchmark. Now I have a known comparison.
But yes, generally speaking. Someone should write a paper on how to do this properly with the least amount of gear and know-how needed. Why? Because it can't be rocket science to measure a SPL number.
I mean, I own several things in my life that can tell me how fast I'm moving, the altitude above sea level, distance walked etc. Why can't we do the same with SPL instead of shouting "SPl numbers are wrong"?
The big problem with "peak SPL" is the very nature of music being dynamic.
The peaks are of very short duration.
Most SPL meters do not have a response time fast enough to measure the peaks.
Even the fast response is WAY to slow to measure the peaks
The "max" on many SPL meters is STILL not the peak. But simple rather the maximum fast response.
The reason people with an average meter cannot measure the "max SPL" is because the average level (which is what the meters read) is easily 10-20dB below the peaks.
The loudspeakers ARE reproducing those peaks (assuming no power compression), but the simple meters can't measure them
Once again-something as seemingly simple as SPL is not so simple and has lots of variables attached to it.
In your examples, the object being measured is not changing-at least not fast (like music).
Any steady state signal/object is MUCH easier to measure.
-
Yes Ivan, I know this. After all, you have been saying this for years to everyone.
Shouldn't you then, instead of telling us how difficult it is, how wrong everybody is when measuring SPL and how everybody is telling the wrong number write the paper about how to do this correctly and repeatably?
Lay down the criteria, the numbers, the signal, the procedure and tell us how to do this correctly. You obviously have the knowledge and it would be something we all cold learn from.
-
Yes Ivan, I know this. After all, you have been saying this for years to everyone.
Shouldn't you then, instead of telling us how difficult it is, how wrong everybody is when measuring SPL and how everybody is telling the wrong number write the paper about how to do this correctly and repeatably?
Lay down the criteria, the numbers, the signal, the procedure and tell us how to do this correctly. You obviously have the knowledge and it would be something we all cold learn from.
First of all, if you actually want to measure the peak, you MUST have a meter that can do so.
Most audio (pros even) do not want to spend that much money on a meter.
Without the meter, it does not matter what procedure you have.
I use an NTI XL2 meter. There are others that are less expensive.
Even some good apps. But the problem with the apps is that they are limited by the max SPL the mic or program can handle.
And that is well well below the 144dB of the NTI.
-
I would agree with you. But how often do you see spec measurements with HP and LP filters in place? That would throw the numbers off (and not in a good way.
I do agree that this would give a better idea of the loudness in the intended range. But would still "favor" the freq that are higher in level.
Hi Ivan, What difference does it make how often do we see measurements with HP and LP filters in place?
The question is whether or not a useful, valid, single SPL number exists.
And IMHO the answer is YES.
My guess is someday, an innovative company with sufficient balls and integrity will start using it, or something similar.
Maybe lead/push other stand-up companies to follow suit...
-
First of all, if you actually want to measure the peak, you MUST have a meter that can do so.
Most audio (pros even) do not want to spend that much money on a meter.
Without the meter, it does not matter what procedure you have.
I use an NTI XL2 meter. There are others that are less expensive.
Even some good apps. But the problem with the apps is that they are limited by the max SPL the mic or program can handle.
And that is well well below the 144dB of the NTI.
Hi again Ivan, I think the great majority on these forums realize how tenuous peak readings are.
Personally, they have no meaning for me, nor really do A and C weightings.
And I'm with Helge,...
It's almost like always bringing up peaking measurements as an example of what's wrong with measurements in general,
or consistently talking about prevalent misleading industry practices.........
instead of offering solutions.......
becomes part of the problem itself...no offense meant :)
-
FWIW, a real world usage measurement I've come to like, is Smaart's LEQ, using pink band-passed with the actual HP and LP filters I intend to use, in place. IMO, this gives an average that is a decent proxy for "a one number meaningful spec".
And taken one step further, I measure LEQ with whatever eq I intend to use to smooth response.
I mean, that's where the sub is going to live, with HP, LP, and eq.....what can it do there?
Using a good RMS voltmeter with averaging, vs the LEQ readings, ....then power handling and even a sensitivity specs make sense to me.
Again FWIW. I know how much experience you have, please advise if you see anything wrong with this methodology.
One issue with using averaging is that technically dB should be converted to Pa (Pascals) averaged and then converted back to dB. This method always favors speakers with a less uniform response in a direct comparison of a single SPL number. This type of method does not reveal if the subwoofer is producing the output primarily at 70-80Hz and is 20dB down at 30Hz versus another unit that has a much more uniform amount of headroom 30-80Hz for example.
-
One issue with using averaging is that technically dB should be converted to Pa (Pascals) averaged and then converted back to dB. This method always favors speakers with a less uniform response in a direct comparison of a single SPL number. This type of method does not reveal if the subwoofer is producing the output primarily at 70-80Hz and is 20dB down at 30Hz versus another unit that has a much more uniform amount of headroom 30-80Hz for example.
Thanks Josh, any idea how Smaart does the LEQ calc ?
-
I have my own opinions that I've developed over years of testing, on how to fairly spec a speaker or subwoofer. For the sake of the discussion, I thought this might be an entertaining experiment...I've uploaded some measurements from a subwoofer I designed. What I'd like to do is have anyone interested look at them and give their opinion on what the "spec sheet" should say. Just assume that these are internal measurements from design and development and that you are in charge of putting the spec sheet together for marketing purposes. I already have my own that I would personally use, if this was a commercially available device that was for sale, which it is not. I will list a brief description of each measurement and the conditions. The equipment is high quality and calibrated. I've been doing this for about 10 years now. If there are any questions just ask.
The system is a large subwoofer intended to be used between 25-100Hz.
First off we have an impedance measurement. What nominal impedance should this cab be rated at?
(http://i828.photobucket.com/albums/zz205/Kallus42/Skhorn%20Active%20All%20Vents%20Open%20Impedance_1.png)
Note for all further measurements shown here.
1.)These measurements are taken in a large, flat, outdoor area. (Half space)
2.)The cab is placed horizontally on the ground.
3.) The microphone (EW's M30) is placed on the ground at a specified distance from the cab.
4.) There is no processing at all in these measurements. No high or low pass filters and no EQ.
5.) The measurements are taken with a very slowly ascending sine wave sweep for higher noise rejection and better resolution.
6.) There is no smoothing applied to the measurements.
Let's look at a voltage sensitivity measurements.
There are 4 measurements on the chart corresponding to microphone distances of 1m, 2m, 4m and 10m from the cab. The voltage used was 14.1 volts measured at the speakon terminal to the cab for each measurement. The measurements are then normalized relative to each other back down to a level corresponding to 1.41 volts at 1m. This being a large sub with multiple radiating surfaces there is some variation in the measurements depending on the mic distance. 10m is more accurate IMO.
How would you rate the voltage sensitivity of this cab? Is 1.41 volts correct or should a higher voltage of 2 or 2.83 volts be used resulting in a 3-6dB higher sensitivity rating?
(http://i828.photobucket.com/albums/zz205/Kallus42/Skhorn%20Active%20All%20Vents%20Open%20Voltage%20Sensitivity_1.png)
Let's look at the raw frequency response.
What would you rate the frequency response of this cabinet as?
(http://i828.photobucket.com/albums/zz205/Kallus42/Skhorn%20Active%20All%20Vents%20Open%20Raw%20Frequency%20Response_1.png)
Now let's look at determining the maximum output of this active subwoofer. I'll present 2 measurement types that are on opposite ends of the spectrum.
Here's the first one. This is measured at a distance of 4 meters ground plane and normalized to 2 meters ground plane (Approximating the SPL of 1m full-space). The high pass filter is engaged for this measurement. It is a 23Hz 12dB/octave BW filter.
The signal used is an ascending sine wave sweep covering 2Hz to 120Hz. It has a duration of 23 seconds.
The amplifier is at the point of limiting and / or clipping throughout the entire measurement. This is basically "wide open" worst case scenario type stuff. Basically this tests accelerated long term heat build up in the voice coils, issues with driver over excursion or bottoming and finds "problem" frequencies. This is considered to be the output limits with long term, low crest factor signals.
Based on this what should be the " Long term" maximum output rating? Should this be rated as is as 1m full-space, or pick up the free 6dB by rating it at 1m half-space?
(http://i828.photobucket.com/albums/zz205/Kallus42/Skhorn%20Active%20All%20Vents%20Open%20Maximum%20long%20term%20Output%20Sweep_1.png)
The short term output is quantified by a different type of measurement. I use 1/3rd octave, Hann windowed, 6.5 cycle long burst signals. This signal type was popularized by Don Keele years ago and has since been incorporated into speaker, subwoofer and amplifier measurements. It is available in Clio, REW and Igor Pro softwares and possibly others. Typically distortion thresholds would also be used to determine the maximum useful output of the speaker. I prefer to test to the distortion limit and also beyond for the sake of completeness and to determine the speakers absolute limits if passive and whether it is protected well enough not to blow if active. Measurements are taken at each 1/3rd octave interval. The 2 graphs below are the same data presented different ways. The measurements are referenced and normalized to a distance of 2 meters ground plane, though the measurements may actually be made at 4 to 8 m for larger cabs such as this.(Again the SPL reported here corresponds to 1m full-space SPL.) Also the SPL is not truly a "peak" measurement it is an "rms" type of reporting. It is directly comparable to the sine wave measurements.
Whew...Ok, with that out of the way what would you rate the short term or burst SPL of this subwoofer as? Do you add the free 6dB by making it a 1m half space number? Do you add 3dB to make it a peak reading? Also note that in the top graph this subwoofer is the center blue line only! The other 2 are historic min and max.
(http://i828.photobucket.com/albums/zz205/Kallus42/Skhorn%20Active%20All%20Vents%20OpenCEA-2010%20THD%20Limited%20Burst_1.png)
(http://i828.photobucket.com/albums/zz205/Kallus42/SKHORN%2021IPAL%203V%206K%20CEA-2010%20BURST.png)
-
Well that turned out to be a much bigger post than I thought.
Thanks Josh, any idea how Smaart does the LEQ calc ?
I have no idea. I've actually never used Smaart. Maybe it has more information in the user notes of help file?
-
First off we have an impedance measurement. What nominal impedance should this cab be rated at?
In the interest of playing it safe, I'd actually call that a 2ohm cab. A more "nominal" value would be 3ohm, but when it comes to very low impedances (and choosing amplifiers to match), I think it's best to be conservative.
- I think so long as the sensitivity measurement is clear (ie, 2.83v @2ohm or whatever), then anything sensible will be fine. If you're looking at buying something like this, you should be able to convert as needed.
- I'd call it 30Hz-150Hz.
- The most common usage for subwoofers is placed on the ground, so I'd probably standardise for that, with the note that 6dB would be lost when flown high up.
- Rating the maximum SPL is tricky since the frequency response is sloped. I'd probably rate maximum continuous and burst at the top and bottom of the passband (noted above).
Chris
-
Well that turned out to be a much bigger post than I thought.
I have no idea. I've actually never used Smaart. Maybe it has more information in the user notes of help file?
ANd the bottom line is that you can get different "numbers" from the same loudspeaker-depending on what you are looking for/at.
When the only spec is "max XXdB SPL", that REALLY doesn't tell very much.
And without any supporting data, it kind of "says", we don't want people to know where the numbers come from. At least to me.
Manufacturers used to give lots of data-but it is becoming less and less these days.
Yet, measurement systems are very affordable-as compared to years past.
-
In the interest of playing it safe, I'd actually call that a 2ohm cab. A more "nominal" value would be 3ohm, but when it comes to very low impedances (and choosing amplifiers to match), I think it's best to be conservative.
- I think so long as the sensitivity measurement is clear (ie, 2.83v @2ohm or whatever), then anything sensible will be fine. If you're looking at buying something like this, you should be able to convert as needed.
- I'd call it 30Hz-150Hz.
- The most common usage for subwoofers is placed on the ground, so I'd probably standardise for that, with the note that 6dB would be lost when flown high up.
- Rating the maximum SPL is tricky since the frequency response is sloped. I'd probably rate maximum continuous and burst at the top and bottom of the passband (noted above).
Chris
Both you and Ivan make points I agree with and both of your posts, in combination with my own post of the measurements of a cab, illustrate the problem. There is no way to express the maximum SPL of speakers, in a way that would be meaningfully comparable, with only a single data point. The same could be said of voltage sensitivity.
Let's say that the pro industry started to move towards standardized ratings of the maximum SPL of a subwoofer. They still want a single data point for the consumers. A 50Hz test frequency is settled on as that's right in the wheelhouse of any subwoofer. You have to settle on a test signal too. Let's say Mr. Keele's 6.5 cycle shaped burst. You also have to agree on the measurement conditions say outdoor ground-plane and the reporting method. Say 1 or 2m half space, or whatever. A measured result is required.
Fast forward a few years and the market starts to see the new max spl @50Hz spec factoring into subwoofer sales. What happens logically? The manufacturers start to design their subs to make the most sound possible at 50Hz as it could result in increased sales. Of course they may be shortchanging the deep bass extension and headroom or other aspects of the device to do that. It's still only a single data point.
I don't think it would be that difficult to make a fairly simple industry spec for subwoofers. I'd like to see this.
1.) Graph of measured impedance for passive cabs covering 1/2 octave below and 1/2 octave above the speakers rated bandwidth of use.
2.) Graph of measured voltage sensitivity of the cab with voltage applied and other required measurement data (ground plane or flown? Mic distance, smoothing used, etc.)
3.) Graph of frequency response for active cabs. (Voltage sensitivity takes care of passive cabs.) Again the measurement specifics need to be noted.
3.) Adopt something similar to CEA-2010 to rate the maximum output. Use the 6.5 cycle shaped bursts at 1/3rd octave intervals. Use the distortion thresholds for a pass/fail result. The consumer home audio industry is far ahead in this regard in my opinion. At least they have some sort of sensible protocol for it that has gathered traction. Personally I would throw out the averaging all together and a few other things from their standard. I'd like to see the subwoofers be tested over their full bandwidth in 1/3rd octave intervals. If your sub is rated down to 20Hz (Never mind that it is well below tuning and 20dB down by then) it needs to have the maximum SPL reported at that frequency. That would prevent overzealous claims of extension that isn't useful in the real world. With active cabs this will quickly show that the flat response shape dialed in with heavy EQ at lower volume levels, does not hold up in the least as the output demands are increased. With passive cabs it would be best to use a voltage source which is limited to the maximum recommendation by the manufacturer. A format similar to mine I posted earlier makes a lot of sense.
In my opinion the 3 measurement graphs listed above would do about 90% of it. A bit of text on a spec sheet isn't really the way to do it.
Ivan,
Danley already has 2 of the 3...
-
Josh, I am sure you know this, but the problem with providing information that nobody else does, is that some people will try to use this information to "judge" the performance-without actually knowing or understanding that data.
Let's look at distortion. Many people would like to see this.
But in what way? A percentage of the output? A specific SPL at a specific distance? Is it for a specific freq? The maximum at some freq? An average across the range? If an average-what is the range?
So let's say that a manufacturer comes up with a "single number" (like amplifiers do).
Let's say it is 2%.
If the level is pretty high, that is very low for a loudspeaker.
But yet some people would see it as extremely high-as compared to amplifiers or other electronics.
So they would claim that the particular model has really high distortion.
But what about a speaker that they like? Do they know what the distortion is-and was it measured the same way?
Unless you have a USEABLE reference to compare, you have no idea if the numbers being presented are good or not.
YES-it get pretty complicated pretty quick.
-
What about a sensitivity graph, and a Max.SPL graph limited to 10% THD?
You could overlay input voltage (or power into the nominal impedance) so that people can size amplifiers accordingly.
Chris
-
What about a sensitivity graph, and a Max.SPL graph limited to 10% THD?
You could overlay input voltage (or power into the nominal impedance) so that people can size amplifiers accordingly.
Chris
Here is where the "fudging" would come into play.
Is that 10% at any one freq? What if there is one freq that stands out, but everything else is 3%?
The problem is also how to measure THD (TOTAL harmonic distortion).
It is one thing to measure the individual harmonics (2nd, 3rd, 4th etc), But measuring them all together is a different story.
Some systems can go a graph of the individual harmonics (like a freq response graph). But the total THD must be done on a Freq by freq basis, or a 1/3rd octave (skipping all the freq in between) graph like an RTA.
And when you go to measure this, does it include the recommended processing (in order to get the pretty -often misleading-numbers on the front of the spec sheet)?
Or is it the raw response?
If it is the raw response-how low do you go? To the rated -3dB point? or the "operating range" which could be -10 or -20dB?
As you can see, it might "appear" to be a nice simple number- but in reality there are all kinds of ways to "get around" it and come up with better marketable results.
And unless everybody is doing it the same way-you have nothing to compare.
We can't even get everybody to agree on what is a -3dB point!!!!!!!!
Just look at various spec sheets (those that provide an unprocessed freq response graph) and see if the numbers on the spec sheet (sensitivity and -3dB) agree.
You will find many that don't. And LOTS that you don't even get a graph to compare-so where does it come from?
In MY opinion (and some others) the sensitivity MUST be tied to the -3dB point-or else it is meaningless.
After all what does -3dB mean? It means 3dB down FROM SOMETHING? What is that "something"? DO NOT ASSUME it is sensitivity.
So given any speaker out there, the "spec sheet writer" can make a choice. If they want a higher sensitivity, they MUST settle for a higher -3dB point. If they a lower -3dB point, the MUST settle for a lower sensitivity.
Or you can disconnect them and have meaningless numbers-but people WILL use them to compare to others-without having a clue they are wrong.
I better stop now----------------
-
Josh, thanks for thoughtful discussion..I'll throw in my 2 cents...
Nominal impedance is aptly named, given that a definition of nominal is "bearing significance in name only".
Just doing an eyeball integration of the Skhorn's impedance graph, I'd call it 4 ohms, with the realization it spends alot of time in 2 ohm territory.
But I don't think we need to eyeball a number. If we used the same pink noise pass-band limited SPL LEQ vs RMS voltage technique, that I like for a sensitivity spec, and also take an average current measurement over the same period, we would get a measured average impedance.
Alternatively, a simple spreadsheet integration of the impedance curve (for same pass-band) should give the same answer as measured average readings.
Sure, that means no more 2, 4, 8, or 16 ohm rounding..but hey.
The impedance curve is a must to have: if I or the marketplace has to have a nominal value, I'd choose a measured or calculated average over eyeballing the curve.
As far as what voltage to use for sensitivity, I don't think it matters, it just has to be specified.
Personally, I like 2.0v, just because it's an easy number to base mental math off of.
Same thing with full, or half space, doesn't matter...just needs to be clearly stated.
Adequate disclosure is the first step to creating specs with value. Sure people will play games without industry uniformity in disclosures. But as disclosure evolves, uniformity eventually takes place.....IMO (in my observation)
For quoting magnitude response:
Going back to starting with a sensitivity spec derived from SPL-LEQ vs RMS average voltage...
I would make that value the reference level for the stated pass-band. Draw that value as a straight line through the measured response, and quote both the+/- dB spec , as well as the -3, -6, etc corners, against that line..
It's a simple way of how I've come to think of response, and at least for me, it makes intuitive sense and holds up in reality.
As far as max SPL:
Peak has no meaning IMO, for all the reasons Ivan provides.
Long term max :In my mind, that's a function of compression.
For compression, I'd keep raising the average voltage on the SPL-LEQ test, until I hit a number I can't live with. I don't run gear hard enough to know what that number is, but my gut says if I get 1 dB average compression for a 3dB voltage increase, I've hit max.
Short term max: In my mind, that's a function of distortion.
I really like what you advocate in measuring this.
So far, I just use REW's sine sweeps.
IME, distortion is the first and most important, output limitation that comes in to play for the sound I try to provide. And it kicks in so far below the "max output" specs, it almost makes even looking at max output seem silly.
Not sure all that was worth 2 cents :)
-
But I don't think we need to eyeball a number. If we used the same pink noise pass-band limited SPL LEQ vs RMS voltage technique, that I like for a sensitivity spec, and also take an average current measurement over the same period, we would get a measured average impedance.
For compression, I'd keep raising the average voltage on the SPL-LEQ test, until I hit a number I can't live with. I don't run gear hard enough to know what that number is, but my gut says if I get 1 dB average compression for a 3dB voltage increase, I've hit max.
The only "problem" I see with measuring the current is the bandwidth used.
As you know, the impedance varies quite a bit within the subwoofer freq range-but what is that range?
Is is some set of numbers that ALL subs are limited to?
Or is is more specifically for a particular sub?
Consider that some subs easily go more than an octave lower than others. So the impedance peaks and dips will be different at different freq.
And many "subs" only operate over about a 1 octave range. And others are over a 2 octave range. So lots of variance there.
Regarding the power compression.
I first proposed my idea of real time power compression testing to Pat Brown (Synaudcon) back in 2001.
It is now his "toaster test" which is also part of the new AES speaker power handling test.
Basically you look at a dual transfer function and look for the point at which some point in the response drops 3dB. This is a safe test that does not do any destruction to the driver.
I would disagree with the use of AES or IEC signals for subwoofers-at least in the modern world.
Modern music has much more low freq content than music of years past.
The current signals roll off the low and high freq.
But here is another BIG issue with that idea. For this test, do you use the "recommended processing" or is the raw response? What about high and low pass filters?
MANY subs use a large boost (sometimes 10dB or more) at the bottom freq in order to get the "-3dB" point.
When this boost is applied, those freq will go into limiting much sooner.
So manufacturers would not apply those boosts when doing the test-to get higher SPL numbers. But yet-the speaker is not working as intended or providing the response that is proposed on the spec sheet.
So that is not a realistic operating condition. So the test (in that case) is purely for some numbers, and not doing what it should be doing-which is providing useful numbers for the user/operator to help judge the actual performance that can be expected from the device.
The snake pit just keeps getting deeper-------------------
-
The only "problem" I see with measuring the current is the bandwidth used.
As you know, the impedance varies quite a bit within the subwoofer freq range-but what is that range?
Is is some set of numbers that ALL subs are limited to?
Or is is more specifically for a particular sub?
Sure, subs differ alot.
So IMO the manufacturer should be free to choose whatever bandwidth they want to, whatever they think best represents the design and intended usage.
They could even provide multiple sets of measurements based on different bandwidths, to show different usages. Take the TH-118. I'd like to see bandwidth specs for say 35-100Hz for use with generic mains, and say 35-140Hz (or whatever recommended freq) for use with SM-80.
It doesn't matter, as long as the HP & LP filters in place are clearly stated.
What does then matter, is for the stated bandwidth, the sensitivity that is determined through SPL-LEQ vs RMS average voltage becomes the reference for any further specifications.
Operating frequency range can be stated +xdB, -ydB......against the reference.
Relative Corners get stated against the reference SPL
Maximum output must fall within the stated bandwidth.
If the manufacturer wants to claim an unrealistic bandwidth, the sensitivity spec will suffer.
If they choose a higher than normally used LP filter, you know they are most likely trying to raise sensitivity. There is also a high probability doing so will make the low corner look comparatively worse. (Why do we ever want to see sub specs above say 150Hz anyway ?)
Your compression test and comments make a lot of sense, thx.
If a manufacturer uses low end boost, that should be clearly stated along with bandpass filters.
And using boost will make the sensitivity spec look worse, because the voltage needed to boost the corner, will show up in the SPL-LEQ vs RMS avg.
All in all, I believe the simple SPL-LEQ vs RMS avg to determine sensitivity for a stated bandwidth, and then using that sensitivity as reference for further specs, would do a hell of alot towards cleaning out the snake pit .... ;)
-
Sure, subs differ alot.
What does then matter, is for the stated bandwidth, the sensitivity that is determined through SPL-LEQ vs RMS average voltage becomes the reference for any further specifications.
Here is the problem with using a SPL meter for the sensitivity.
As the bandwidth gets wider, the SPL goes up. Even though for a single freq it does not get any louder.
An SPL meter is looking at the TOTAL energy, not the energy at any particular freq.
The "classic" way of SPL was using a chart recorder and a swept sine wave.
The SPL at each freq was plotted on the recorded and that was the spec.
I am not aware of anybody using a simple SPL meter for sensitivity.
Here is a thread I started that specifically shows how the increased bandwidth increases the simple SPL numbers- but no individual freq is any louder.
http://forums.prosoundweb.com/index.php/topic,163748.0.html
Danley uses TEF, which is a just like the classic way-except the display is on a computer screen, not a chart recorder.
A constant voltage is applied through the sweep.
-
And when you go to measure this, does it include the recommended processing (in order to get the pretty -often misleading-numbers on the front of the spec sheet)?
Or is it the raw response?
Doesn't matter.
If internal EQ is being applied, the only thing that changes is the input level to get a given output level of the amplifier.
The distortion will be produced by the driver, and will always be limited at the same SPL per frequency no matter how much EQ is applied.
IMO, they can use processing on their graphs and produce some lovely-looking frequency response curves. The Maximum Level for 10% THD vs frequency graph would be where we seperate the men from the boys.
You could run that across any bandwidth you like. All the way down to 10Hz would be fine. Again, it doesn't matter - what the graph will tell you is where you can push the cabinet loud, which will be at and above the bottom tuning frequency.
Chris
-
Doesn't matter.
If internal EQ is being applied, the only thing that changes is the input level to get a given output level of the amplifier.
The distortion will be produced by the driver, and will always be limited at the same SPL per frequency no matter how much EQ is applied.
IMO, they can use processing on their graphs and produce some lovely-looking frequency response curves. The Maximum Level for 10% THD vs frequency graph would be where we seperate the men from the boys.
You could run that across any bandwidth you like. All the way down to 10Hz would be fine. Again, it doesn't matter - what the graph will tell you is where you can push the cabinet loud, which will be at and above the bottom tuning frequency.
Chris
But the curves might not be the same with and without eq vs distortion.
In loudspeakers the distortion is not uniform across the freq range
So are you trying to measure for a number? or is the number trying to tell actual performance?
Big difference
-
Here is the problem with using a SPL meter for the sensitivity.
As the bandwidth gets wider, the SPL goes up. Even though for a single freq it does not get any louder.
I am not aware of anybody using a simple SPL meter for sensitivity.
Another it doesn't matter.
Yes, SPL goes up as bandwidth is increased (with stimulus across entire bandwidth).
But why does that matter? How else are you going to spec sensitivity other than over intended bandwidth?
And conversely, what's the point in including data outside intended bandwidth? Doing so is one of the games we all frown on.
IMO, whether or not anyone is using simple SPL-LEQ to determine sensitivity really isn't important, what is important is... is it a better mousetrap?
Maybe some folks are beginning to use it.........
As luck would have it, I went to look at a JBL mini array box brought up this morning on the long-throw thread in the lounge. ...https://www.jblpro.com/ProductAttachments/JBL_VT4886_v3.pdf
Check out the footnotes re measurement methodology.
Unless I'm missing something (easily possible !), JBL is using it.
-
Another it doesn't matter.
Yes, SPL goes up as bandwidth is increased (with stimulus across entire bandwidth).
But why does that matter? How else are you going to spec sensitivity other than over intended bandwidth?
And conversely, what's the point in including data outside intended bandwidth? Doing so is one of the games we all frown on.
I am not aware of anybody using a SPL meter for the sensitivity measurement.
In MY opinion, what you need to see is a level at a particular freq for a given input.
A SPL meter gives you no information regarding freq response-just the total level in the area of the mic.
What happens outside of the intended bandwidth is important. Not so much from a spec standpoint, but from an eq standpoint. If there is excess level, it needs to be dealt with.
It is important to SEE that data-NOT rely on it for some higher SPL specs.
-
IMO, whether or not anyone is using simple SPL-LEQ to determine sensitivity really isn't important, what is important is... is it a better mousetrap?
Maybe some folks are beginning to use it.........
As luck would have it, I went to look at a JBL mini array box brought up this morning on the long-throw thread in the lounge. ...https://www.jblpro.com/ProductAttachments/JBL_VT4886_v3.pdf
Check out the footnotes re measurement methodology.
Unless I'm missing something (easily possible !), JBL is using it.
I have never seen an LEQ measurement that gave any freq information.
Lets say the sensitivity is 100dB as per your SPL-LEQ reading.
Let's say you have a freq response that has some peaks and dips-or some major peaks
How do you know where the LEQ number (a single number) "sits" on the overall response shape?
-
I am not aware of anybody using a SPL meter for the sensitivity measurement.
Hi Ivan, did you look at the VT4886 spec sheet link ?
Check out footnotes 2 & 3.
If they are using noise, doesn't that mean they are most likely using simple SPL ?
-
I have never seen an LEQ measurement that gave any freq information.
Lets say the sensitivity is 100dB as per your SPL-LEQ reading.
Let's say you have a freq response that has some peaks and dips-or some major peaks
How do you know where the LEQ number (a single number) "sits" on the overall response shape?
Perfect questions.....
With smaart you have to do some work to relate 'whatever' reference db on the transfer function to the SPL-LEQ meter,
with REW it's easier because the transfer function is given as measured SPL magnitude response.
Once that's done, the LEQ fits through the transfer function as a weighted middle average straight line...it has to by definition.
Then you base response + or - to that line, as well as corners. Piece of cake :)
Really, LEQ is the same thing you're doing when you eyeball a sensitivity level...only it's measured !!!
The simple requirement is that you take an LEQ simultaneously to a pink noise driven transfer function.....
You could take your TEF graphs, and just draw LEQ through it, if you could equate the sine sweep to noise, I guess.
-
Perfect questions.....
With smaart you have to do some work to relate 'whatever' reference db on the transfer function to the SPL-LEQ meter,
with REW it's easier because the transfer function is given as measured SPL magnitude response.
Once that's done, the LEQ fits through the transfer function as a weighted middle average straight line...it has to by definition.
Then you base response + or - to that line, as well as corners. Piece of cake :)
Really, LEQ is the same thing you're doing when you eyeball a sensitivity level...only it's measured !!!
The simple requirement is that you take an LEQ simultaneously to a pink noise driven transfer function.....
You could take your TEF graphs, and just draw LEQ through it, if you could equate the sine sweep to noise, I guess.
The problem with noise is trying to come up with an actual-accurate reading.
I have seen meters that read the same (or close enough) on a sine wave, but on noise they read differently.
It has to do with the integration time of the meter and how it displays its "average" between display readings.
So which one is correct?
-
The problem with noise is trying to come up with an actual-accurate reading.
I have seen meters that read the same (or close enough) on a sine wave, but on noise they read differently.
It has to do with the integration time of the meter and how it displays its "average" between display readings.
So which one is correct?
I dunno here Ivan, you have boatloads more experience here than I do.
All I do know, is that I get consistent SPL-LEQ readings using either smaart, rew, or terrasonde.
And consistent voltage readings using fluke meters and scope.
This is true for fast and slow .....A, C, and Flat measurements, as well.
And LEQ is typically taken over minutes....how much integration error can there be over such a relatively long time span ?
I just don't see the problem ...
D
-
It would seem to me (a "layman" with 45+ years in this industry) that the ONLY way for specs to really be useful would be to run tests on every speaker using the same methods and equipment. Otherwise you will never have accurate numbers for comparison.
On a side note, and I believe Ivan has mentioned this several times, specs do NOT tell you what a speaker actually sounds like in the real world.
The spec I'd really love to see is WGI/WGO: what goes in vs what goes out. How accurate is the speaker in terms of reproducing EXACTLY what goes in, and how accurate within it's rated coverage angle?
I realize that digital processing can
correct" a lot of things, but I still believe you are better off starting with a speaker that doesn't have built-in limitations that have to be corrected.
-
But the curves might not be the same with and without eq vs distortion.
In loudspeakers the distortion is not uniform across the freq range
So are you trying to measure for a number? or is the number trying to tell actual performance?
Big difference
The "frequency response" curve will vary according to whatever EQ the manufacturer implements.
The Max.SPL (10% THD Limited) curve will not.
Working through an example usually helps. I know you're a smart guy, but it feels like we're not quite getting through to each other.
Lets say I have a 15" driver in a small ported box with a built-in amplifier. There's some EQ being applied to get it flat at the low end. The manufacturer has dialled in 6dB of boost towards the bottom end of its bandwidth (40Hz).
At 40Hz, the drive unit and cabinet can produce 123dB before distortion gets too high. This number won't ever ever change with any EQ that's applied. It's a limit of the drive unit in that cabinet. Since there's 6dB of boost being applied at that frequency, at other frequencies we'll probably be able to push the line level input a little further, with the caveat that you note: different frequencies will have different drive levels needed to hit the distortion limit.
But since that's what the test is all about anyway, it really doesn't matter. Play a tone, push the level until you hit 10% THD, measure the SPL. Repeat at other frequencies.
What will be produced is a graph that will show how loud the cabinet can go at any frequency before it starts sounding stressed.
This is starting to sound a lot like Josh's method, except in graph (rather than tabulated) form.
Chris
-
The "frequency response" curve will vary according to whatever EQ the manufacturer implements.
The Max.SPL (10% THD Limited) curve will not.
Working through an example usually helps. I know you're a smart guy, but it feels like we're not quite getting through to each other.
Lets say I have a 15" driver in a small ported box with a built-in amplifier. There's some EQ being applied to get it flat at the low end. The manufacturer has dialled in 6dB of boost towards the bottom end of its bandwidth (40Hz).
At 40Hz, the drive unit and cabinet can produce 123dB before distortion gets too high. This number won't ever ever change with any EQ that's applied. It's a limit of the drive unit in that cabinet. Since there's 6dB of boost being applied at that frequency, at other frequencies we'll probably be able to push the line level input a little further, with the caveat that you note: different frequencies will have different drive levels needed to hit the distortion limit.
But since that's what the test is all about anyway, it really doesn't matter. Play a tone, push the level until you hit 10% THD, measure the SPL. Repeat at other frequencies.
What will be produced is a graph that will show how loud the cabinet can go at any frequency before it starts sounding stressed.
This is starting to sound a lot like Josh's method, except in graph (rather than tabulated) form.
Chris
It all depends on how the data is "presented".
You and I are talking about 2 different ways of presenting the data.
When I say the distortion will vary depending on the freq and boosting, I was talking about using the same drive level into the processor.
The boosted freq will have have more drive signal so possibly higher distortion/
But hwat you are talking about is running each freq up to 10% and then plotting a graph.
Of course this graph would not possibly look at all like the freq response graph.
And you would have to do each freq individually.
I honestly don't think you will ever get manufacturers to do that.
Heck most don't even post a simple response graph.
It can be good to wish, but it is something you would have to do on your own I believe.
If customers don't have anything to compare to, how do they know if it is good or not?
Here is the really scary thing. I have meet a good number of "professional audio guys who cannot even read a simple response graph.
Give them a response graph and tell them "Let's assume the sensitivity is 100dB. Can you tell me what the -3dB point is?"
And they can't.
So how can we expect people to understand more complicated issues.
I guess that is why so many manufacturers only give some simple numbers, because that is all some people "think" they understand.
Most don't want to look at graphs and such, they just want a simple number to tell them how loud it is, and they assume that number gives the whole story.
Oh well----------
-
When I say the distortion will vary depending on the freq and boosting, I was talking about using the same drive level into the processor.
The boosted freq will have have more drive signal so possibly higher distortion/
But hwat you are talking about is running each freq up to 10% and then plotting a graph.
Of course this graph would not possibly look at all like the freq response graph.
And you would have to do each freq individually.
I honestly don't think you will ever get manufacturers to do that.
Heck most don't even post a simple response graph.
Gotchya. It all makes sense now.
It'd be an easy enough test to automate. If they're bothering to publish frequency response charts, it wouldn't be much more difficult to do the distortion test. I note that RCF publish a decent amount of data for their cabinets.
I'd like to think that if we had some kind of standardised test that every manufacturer could use, people would bother to learn to read the graphs. I do suspect, though, that we're in a small minority of people that care enough about these things to come up with a way of testing them. Most DJs will go to a store, do a quick audition, and buy whatever hits them in the chest the most.
Chris
-
Ivan brings up a valid point... "What is the bandwidth"? It's not really possible to cover every sub or speaker with a single generic bandwidth. Some subs go below 20Hz and others offer nothing much below 40Hz. The same could be said above 100Hz. It's not going to be fair to rate sensitivity by using the same bandwidth for every cab.
Another thing I agree with Ivan on is NOT using a noise signal for sensitivity, impedance or SPL specs, for the reasons he has brought up. I prefer a swept sine measurement signal.
Like I said in my last post I really think only 3 measurement graphs would be needed to bring subwoofer manufacturer specs back to reality. In fact only 2 for powered subs. Speakers would require a bit more most notably polar data.
1. Impedance. Covering 10-200Hz. That covers the useful frequency range of nearly any sub cab. In my opinion the nominal impedance spec on the sheet should be the minimum impedance in the rated bandwidth of the cab rounded to the nearest whole ohm. If the cab has a measured minimum impedance of 3.6 ohms that would be a 4ohm nominal cab. If the impedance minimum is 3.3 ohms call it a 3 ohm cab. Regardless of the paper spec the measured impedance graph is there to see for yourself and make a judgment call.
2. Frequency response / Voltage sensitivity. Ground plane frequency response measured with a swept sine signal. I'd prefer a 10m measurement distance. Nominally 100w, 10X voltage for sensitivity. The frequency of maximum sensitivity should be part of the spec.
3. Short term, distortion limited, 1/3rd octave, burst output. This hits a number of areas. This is measured ground plane. No calculated marketing BS with the SPL rating. It's distortion limited, thought the levels allowed can be quite high, so the cab may be starting to sound distressed but it isn't a totally garbage FART sound making that maximum SPL reading. The individual harmonics have their own thresholds. 2nd is allowed to be very high, 3rd a bit less, 6th much less, etc...The SPL at each 1/3rd octave band is known. No guessing whether the SPL rating is made up near 200Hz. This testing will also reveal powered subs using heavy EQ boost on the low end and unable to maintain their LF extension at higher output levels. This is very common and is immediately obvious once the +/3dB response at lower volumes turns into +/-12dB as the output at 80Hz runs away completely from the 31.5Hz output. This is a bit much data to put on a simple marketing spec sheet but it can easily be condensed down into a single SPL number or even a few at different frequencies for that. Above all the test frequency should be noted. 130dB@80Hz for example.
-
1. Impedance. Covering 10-200Hz. That covers the useful frequency range of nearly any sub cab. In my opinion the nominal impedance spec on the sheet should be the minimum impedance in the rated bandwidth of the cab rounded to the nearest whole ohm. If the cab has a measured minimum impedance of 3.6 ohms that would be a 4ohm nominal cab. If the impedance minimum is 3.3 ohms call it a 3 ohm cab. Regardless of the paper spec the measured impedance graph is there to see for yourself and make a judgment call.
NO NO NO :) :) :)
Trust me on this.
If you state anything other than 2-48-16 ohms, it REALLY REALLY confuses a lot of people.
The SH50 was originally speced as a 6 ohm cabinet.
This is what I felt best represented the overall load on the amplifier.
We had a impedance curve on the spec sheet.
You would not believe all the calls I got about that.
"What size amp do I need-I can't find an amp anywhere that is speced at 6 ohms" and so forth.
So I just changed the spec sheet to read 4 ohms. All of the calls about impedance stopped.
It is truly sad how many "professionals" in our industry don't understand even the basics.
I think this is why a lot of manufacturers have stopped putting useful data on the spec sheet. "Let's just sell them what they think they want and all will be happy" seems to be the current trend, no matter if it is actually correct.
When you don't put a spec, nobody can hold you to it.
I have seems specs such as "wide" "narrow" but no degrees given. So how wide is "wide"?
How about the amp rating on a powered speaker that is "adequate". Yeah-that helps. For what?
-
Alright, call it a 2ohm box and give the amp an easier life. I don't think I'd ever put a pair of 3ohm loads across an amp unless it was an absolute emergency.
Chris
-
Alright, call it a 2ohm box and give the amp an easier life. I don't think I'd ever put a pair of 3ohm loads across an amp unless it was an absolute emergency.
Chris
But the issue is not so much the number of the lowest point on the impedance curve.
But rather the overall impedance.
THAT is why a graph of the curve is important, NOT just a simple number.
What if it only drops down to the 3.3 ohm for a small range of freq-but yet the other freq in its operating range are much higher?
If those couple of notes are not hit, then it does not present a load (at least enough to be considered) on the amp.
Once again-this is where the "simple single number" gets people into trouble.
And it is not just a curve-but also the RESOLUTION of the curve. You could put a bunch of smoothing on my example and raise the "apparent" 3.3 ohm to a higher impedance.
But if you look at unsmoothed data (depending on the number of actual data points-that is another way to "cheat", use unsmoothed but few data points and it will average out) you can see more where the peaks and dips are in the impedance curve.
Then you can get a better understanding of the actual load on the amp across the operating range.
-
Ivan brings up a valid point... "What is the bandwidth"? It's not really possible to cover every sub or speaker with a single generic bandwidth. Some subs go below 20Hz and others offer nothing much below 40Hz. The same could be said above 100Hz. It's not going to be fair to rate sensitivity by using the same bandwidth for every cab.
Another thing I agree with Ivan on is NOT using a noise signal for sensitivity, impedance or SPL specs, for the reasons he has brought up. I prefer a swept sine measurement signal.
I completely agree with all your testing methodology, and have really learned a lot from your excellent site..which BTW, thanks for it!
But I still think the SPL-LEQ methodology has a lot of merit....for a sensitivity spec, and for the +/- response specs put into words.
One question first pls:
Do you get better magnitude and phase measurements with swept sine, than with noise (ala smaart or systune etc)
I've compared mag and phase many times with smaart's transfer pink noise vs REW's swept sine, and consider the measurements equal.
Especially so when any smoothing is added.
I understand that swept sine is needed for harmonic distortion, and maybe offers more gating alternatives than pink noise.
But for mag and phase only, does swept sine offer any real advantage?
Going back to LEQ using noise...I'll give another try at why I like it...
OK, say you've built a new sub and it's time to write up the specs.
You run a swept sine measurement to get raw mag and phase response (with some chosen reference voltage)
You then eye ball what the rough bandwidth looks like, and what SPL level kinda runs through the middle of the bandwidth to get a sensitivity SPL, say 102dB for example.
If you're honest, you use that eyeballed sensitivity spec to define the -3dB, -6dB, etc, response corners, which then becomes the implied if not outright claimed bandwidth. So you've ended up with a bandwidth spec after all...
Now, how well does your sensitivity spec match through that bandwidth?
Well, perfectly, everywhere the response curve hits right on 102dB.
But how about the + and - variations, how well does the sensitivity spec represent the average sensitivity through the bandwidth?
Put HP and LP filters in place that preserve the corners you listed, run a noise based LEQ-SPL at the same chosen reference voltage,
and you'll see how good you did at eyeballing the sensitivity spec, that is how close to 102dB LEQ measures.
-
But I still think the SPL-LEQ methodology has a lot of merit....for a sensitivity spec, and for the +/- response specs put into words.
One question first pls:
Do you get better magnitude and phase measurements with swept sine, than with noise (ala smaart or systune etc)
I've compared mag and phase many times with smaart's transfer pink noise vs REW's swept sine, and consider the measurements equal.
Especially so when any smoothing is added.
I understand that swept sine is needed for harmonic distortion, and maybe offers more gating alternatives than pink noise.
But for mag and phase only, does swept sine offer any real advantage?
Going back to LEQ using noise...I'll give another try at why I like it...
OK, say you've built a new sub and it's time to write up the specs.
You run a swept sine measurement to get raw mag and phase response (with some chosen reference voltage)
You then eye ball what the rough bandwidth looks like, and what SPL level kinda runs through the middle of the bandwidth to get a sensitivity SPL, say 102dB for example.
If you're honest, you use that eyeballed sensitivity spec to define the -3dB, -6dB, etc, response corners, which then becomes the implied if not outright claimed bandwidth. So you've ended up with a bandwidth spec after all...
Now, how well does your sensitivity spec match through that bandwidth?
Well, perfectly everywhere the response curve hits right on 102dB.
But how about the + and - variations, how well does the sensitivity spec represent the average sensitivity through the bandwidth?
Put HP and LP filters in place that preserve the corners you listed, run a noise based LEQ-SPL at the same chosen reference voltage,
and you'll see how good a job you did at eyeballing the sensitivity spec.
Thanks...
Sine measurements can provide more detail and higher resolution. I always use very long / slow settings for better background noise rejection. Noise based measurements are mostly the same but gloss over some edges here and there. You are correct if you will be throwing on 1/6th octave smoothing it's mostly irrelevant. An easy way to see this difference is to do an impedance measurement on a driver or cabinet. A very slow sine wave capture will result in slightly higher peaks and a bit more detail vs PN. I've found that the resulting calculated driver parameters are closer to factory specs using sines. A few of the manufacturer's I've talked to have said it is what they also use to capture the impedance curve so that makes sense that the match is better that way. See attachment.
I don't see anything inherently wrong with Leq averaging but I see potential issues in a few areas.
What weighting do you use? A, C, Z?
What type of noise? PN, WN, periodic?
What bandwidth is the noise?
2 volts of sine wave and 2 volts average noise. Crest factor of the noise signal?
The calculated SPL is an average. I mentioned in an earlier post that this requires converting to Pa and then back to dB SPL (done inside the program or meter typically) . This type of averaging is always going to anchor itself primarily to the loudest peak in response. The lions share of recorded SPL will be from the peak area. This favors speakers with a less linear response when compared to another that has ruler flat response over a greater bandwidth. This alone is why I consider a measured voltage response graph as necessary whether the single spec is Leq derived or by other means.
You don't happen to have a comparison of a sine sweep based measurement at a known voltage and the Leq SPL of the same cab at the same average voltage do you? I'd like to see a comparison. I might mess around with it the next time I have the equipment out.
-
Thanks...
Sine measurements can provide more detail and higher resolution. I always use very long / slow settings for better background noise rejection.
With the TEF, you MUST use long slow sweeps in order to get accurate low freq information.
As a general rule you need to have the resolution to be 1/10th of the lowest expected (typically -3dB point) freq.
If you use a fast sweep the display will "look" like the cabinet rolls of much higher than it actually does.
Once again-just because it shows up on a screen DOES NOT mean that it is accurate-just accurate for the data points in the measurement parameters.
This goes for EVERY measurement program. There are all sorts of places in various programs that can easily give you a erroneous reading
-
I don't see anything inherently wrong with Leq averaging but I see potential issues in a few areas.
What weighting do you use? A, C, Z?
What type of noise? PN, WN, periodic?
What bandwidth is the noise?
2 volts of sine wave and 2 volts average noise. Crest factor of the noise signal?
The calculated SPL is an average. I mentioned in an earlier post that this requires converting to Pa and then back to dB SPL (done inside the program or meter typically) . This type of averaging is always going to anchor itself primarily to the loudest peak in response. The lions share of recorded SPL will be from the peak area. This favors speakers with a less linear response when compared to another that has ruler flat response over a greater bandwidth. This alone is why I consider a measured voltage response graph as necessary whether the single spec is Leq derived or by other means.
You don't happen to have a comparison of a sine sweep based measurement at a known voltage and the Leq SPL of the same cab at the same average voltage do you? I'd like to see a comparison. I might mess around with it the next time I have the equipment out.
Thanks Josh,
I use flat full-spectrum PN (from smaart, 12 dB crest factor).
That signal gets band limited by speaker processor. Typically, I use linear phase 48dB BW HP and LP filters at chosen corners.
Hopefully, crest factors aren't an issue. I set the bandpassed PN RMS voltage to the voltage used in sine sweeps, using a fluke 189 voltmeter. I guess alot depends on how well the fluke makes RMS time-averaged measurements of pink noise.
I average the simultaneous RMS voltage and LEQ-SPL, for about two minutes or so....after not seeing any average variance past about a minute....
I'll try to dig up some old sub comparisons of REW sweeps vs smaart LEQ at the same RMS voltage. I got convinced of SPL equivalence some time ago, and haven't made any recent comparisons. If I can't find any, I'll run a new comparison since I've ballyhooed this so much.
-
Thanks Josh,
I use flat full-spectrum PN (from smaart, 12 dB crest factor).
That signal gets band limited by speaker processor. Typically, I use linear phase 48dB BW HP and LP filters at chosen corners.
Hopefully, crest factors aren't an issue. I set the bandpassed PN RMS voltage to the voltage used in sine sweeps, using a fluke 189 voltmeter. I guess alot depends on how well the fluke makes RMS time-averaged measurements of pink noise.
I average the simultaneous RMS voltage and LEQ-SPL, for about two minutes or so....after not seeing any average variance past about a minute....
I'll try to dig up some old sub comparisons of REW sweeps vs smaart LEQ at the same RMS voltage. I got convinced of SPL equivalence some time ago, and haven't made any recent comparisons. If I can't find any, I'll run a new comparison since I've ballyhooed this so much.
Good deal. I'd appreciate seeing the comparison. I'll go ahead and put this on my list of to do items also. I need to look into REW's distortion limited burst testing function as well and compare it with what I get in IgorPro with Don Keele's app, so I'll probably look into both at the same time. I don't have Smaart so I'll probably just use REW for both PN and sine signals and see what I get. Might take me a few weeks though.
-
I need to look into REW's distortion limited burst testing function as well
Didn't know it'll do that!
Is that a built-in routine, or would it need setting for the CEA burst, open RTA and manually alter drive voltage until distortion passes the limit?
The Help file doesn't mention much about it.
I've just finished some 15" subs here that I'd love to put through their paces.
Chris
-
Didn't know it'll do that!
Is that a built-in routine, or would it need setting for the CEA burst, open RTA and manually alter drive voltage until distortion passes the limit?
The Help file doesn't mention much about it.
I've just finished some 15" subs here that I'd love to put through their paces.
Chris
Chris, it's recently added functionality.
You really have to hand it to John M. He is forever improving an already super program.
https://www.avnirvana.com/resources/v5-19-beta-5-windows-installer-with-jre-allowing-multiple-instances.16/
@Josh, hope to make some new measurements today...maybe even get to try REW's stepped sine
-
Didn't know it'll do that!
Is that a built-in routine, or would it need setting for the CEA burst, open RTA and manually alter drive voltage until distortion passes the limit?
The Help file doesn't mention much about it.
I've just finished some 15" subs here that I'd love to put through their paces.
Chris
I'm not entirely sure. I very briefly messed with it when it was first added to the generator function. it still had some problems at that time but that's been well over a year and a number of updates ago. I don't think it is a full routine. I don't think it will auto check the distortion levels and give a pass or fail flag for example. I'll have to fire it up and play around to see what it is doing.
The other current systems that have this built in are IgorPro / Wavemetrics with the routine setup years back by Don K. ($600 or so for a license of IgorPro. Don's routine plug in is freeware or was many years ago.). CLIO also has this functionality. I've not tried it myself. REW would be by far the cheapest way to do the burst tests ( Good for testing subs, speakers, amps!).
Chris, it's recently added functionality.
You really have to hand it to John M. He is forever improving an already super program.
https://www.avnirvana.com/resources/v5-19-beta-5-windows-installer-with-jre-allowing-multiple-instances.16/
@Josh, hope to make some new measurements today...maybe even get to try REW's stepped sine
Awesome.
Yes this is one of the reasons I use REW. Jon is hands on and easy to communicate with. If there is a problem or suggestion with the program he will work on it and next thing you know it's fixed or implemented. The high dollar, "pro", measurement systems do not behave with this type of direct feedback, implementation, or speed in my experience. Most of them have a user interface that looks and functions like something 20 years old comparatively. Some have less functionality. Simplicity and ease of use is high on my list of requirements.
-
Ok, here's an attempt to post a simultaneous sine sweep and a SPL-LEQ, using REW.
Sine sweep is a ported sub's raw response with linear phase HP and LP filters in place. 48dB BW at 25 and 100Hz.
SPL signal is periodic pink noise, band limited 30-100Hz. I chose 30 and 100Hz because they are the f3 points.
The sub has a tuned f3 at 30Hz, the HPF is hopefully just there for protection.
Both signals were at -12dBFS.
REW is really cool in that it holds the same correct sine equivalent RMS voltage for whatever specified bandwith pink you set.
Josh, the sub btw, was inspired by your BMS box...only it's a little larger @175L with a bit more port area.
Anyway, measured LEQ was 91.7dB, at 14.3vrms. Measurements were at 8 meters.
That gives a 1 meter LEQ derived sensitivity of 95.6 dB @ 2.83v....which sounds right for this 8 ohm driver.
So on the graph below, LEQ of 91.8 (off a tenth) is plotted on the sine sweep....
Hope this all made sense...