Does anyone remember the work of Stanley Lipschitz, John Vanderkooy, and Paul Young on the audibility of differences in amplifiers and preamplifiers in ABX (or double blind) testing when gain and frequency response are held equal? Whether you agree with their results or not, they did find that even “minute” differences in the frequency responses of electronics were audible.
Now, loudspeakers (and loudspeaker drivers) excel at producing not only minute differences in frequency response but major differences too. Even the very best performers in loudspeakers and drivers must be rated at plus or minus 2 or 3 db instead of the less than .25 db band pass rating of electronics.
As a designer, these differences in frequency response (and impulse response) are important to me. I use such characteristics to select drivers and design crossovers and corrective networks, always striving for the goal of coherent impulse response and flat frequency response.
To my fellow designers and builders in the diyaudio forum, I ask, “Does the measured or tested performance of drivers matter to you?” Do you want to know how a driver may change information contained in the electrical signal as it transforms that information into an acoustical signal (in musical terms we might say that the driver colored the sound)? And if you did know the performance of the driver, would you attempt to improve the pass band output of that driver through design?
At least at the lowest frequencies I know that most of you already do this. When you design an enclosure you choose a particular design school (transmission line, ported [Thiele/Small], or sealed) and follow the precepts of that design school to construct an enclosure. You are not satisfied with just what the driver does, you change its acoustic output through air mass loading, tuning, and dampening of the enclosure.
Are you as concerned with the rest of the pass band of the drivers? When you design the crossover, do you attempt to make the acoustic output (pass band or stop band) better than what the driver produces in its natural state? Are you interested in how response variations in the stop band performance of drivers can interfere with correct summing of the crossover? And are you ever concerned with the precision and detail of the manufacturer supplied statistics and performance measurements you might use as the basis of your design efforts?
For example, in a recent thread a member talked about using one of Tang Band’s new 2 by 3 full range drivers in a line array design. Now, I like Tang Band. I find the drivers they produce interesting. Still, before I design something using one of their drivers (or anyone else’s drivers) I would like to know as much as I can about the driver’s performance.
For evaluation purposes, I purchased three of the Tang Band W23-972S drivers. This version of the driver features the “ppm” cone. They included a three-page set of specifications and test results with my order. I include the “log” frequency response graph supplied by Tang Band. I am new to this so I hope this works.
This graph shows the driver meeting the specified frequency response with a plus or minus 6 db range. This is not an unusual quality of performance in a modern driver. Still, as a designer, I am interested in how the peak output from the driver is shown occurring at just below 8 kHz, how it is down about 3 db by 10 kHz, and is down an additional 4 db just shy of 15 kHz. I am also interested in the falling response in the lower frequencies and the slight roughness in output between 1 kHz and 3 kHz. Output falls nearly 10 db between 800 Hz and 150 Hz and there are two 4 db dips at 1.2 kHz and 2.5 or 2.6 kHz. These are all areas I would like to smooth in my final design.
Most of the time I like to do my own testing to supplement the manufacturers testing. The test system used for the above test is not the only kind of acoustical testing available and my test system is different. I happen to prefer impulse testing, and I like to examine frequency response graphs using a linear scale instead of a log scale. Unless otherwise noted, you can assume that the testing is done at one meter. I also find value in converting the impulse response to frequency response in two segments to produce three frequency response graphs. I want to see the summed frequency response, and I want to see the frequency response of the onset portion of the impulse and the frequency response of the decay portion of the impulse response. Often the onset and decay response will be markedly different than the summed response, and I don’t believe it ought to be. This is another goal in my design criteria; to get the onset and summed response as close as possible, and to minimize the decay contribution to the output of the loudspeaker or driver.
I include my testing of the W23-972S drivers. I also changed graph color and superimposed the frequency response graphs of the three samples to show production variations in driver output. Consistent with how resonantly active the cone is, my frequency response graph is a bit more “peaky” than Tang Band’s. While our overall rating of the frequency response is close, I am finding more unevenness in the fine detail of the response than Tang Band’s testing. I believe this will relate to the “sound” of this driver and it is something I would try to correct if I were going to use this driver in a design. It is not, even with my test results, a terrible or unusable driver. It shows certain characteristics, however, that I would try to understand. Once I understood what in its design is making it perform the way it does, it just requires a different set of crossover/loudspeaker design approaches than other drivers to maximize the accuracy of the acoustic output.
Despite how important adherence to a design goal and testing is to me, I do not want to claim that testing and adherence to my set of design goals is the only way to enjoyable sound reproduction. There have been past driver designs and past loudspeakers that have had legions of devoted fans, yet were not perfect in testing. I have included two graphs (the original testing done over 20 years ago) here to show this point. The first is of the Jordan 50 mm Module. This was a well-loved driver in its day, yet, while it lacks the small signal variability of the W23-972S, it is no more flat in its pass band output. There is a huge bell mode breakup resonance, little output above it, generally falling output below it, and a bit of small signal unevenness. Despite it test results, I designed a loudspeaker using this driver; I just designed certain corrective networks to make the output closer to my ideal. Indeed, this is how I started working with Don Spangler, I sent him a letter through Speaker Builder detailing my corrective efforts with the Jordan 50 mm Module.
The last graph is of the Quad ESL 63. It was in production at about the same time as the Jordan 50mm. Like the Jordan, there were discerning audiophiles who loved this loudspeaker. In both cases I can understand why. Even in comparison to most of today’s tweeters, they both produced a more coherent onset response and no worse decay response.
In sum, I would welcome more acoustical testing of drivers and loudspeakers in this forum. I would love to see results from all of the four major testing schemes currently available. I would also be willing to do more testing and submit the results of those tests to this forum if others are as interested as I am.
Mark
Now, loudspeakers (and loudspeaker drivers) excel at producing not only minute differences in frequency response but major differences too. Even the very best performers in loudspeakers and drivers must be rated at plus or minus 2 or 3 db instead of the less than .25 db band pass rating of electronics.
As a designer, these differences in frequency response (and impulse response) are important to me. I use such characteristics to select drivers and design crossovers and corrective networks, always striving for the goal of coherent impulse response and flat frequency response.
To my fellow designers and builders in the diyaudio forum, I ask, “Does the measured or tested performance of drivers matter to you?” Do you want to know how a driver may change information contained in the electrical signal as it transforms that information into an acoustical signal (in musical terms we might say that the driver colored the sound)? And if you did know the performance of the driver, would you attempt to improve the pass band output of that driver through design?
At least at the lowest frequencies I know that most of you already do this. When you design an enclosure you choose a particular design school (transmission line, ported [Thiele/Small], or sealed) and follow the precepts of that design school to construct an enclosure. You are not satisfied with just what the driver does, you change its acoustic output through air mass loading, tuning, and dampening of the enclosure.
Are you as concerned with the rest of the pass band of the drivers? When you design the crossover, do you attempt to make the acoustic output (pass band or stop band) better than what the driver produces in its natural state? Are you interested in how response variations in the stop band performance of drivers can interfere with correct summing of the crossover? And are you ever concerned with the precision and detail of the manufacturer supplied statistics and performance measurements you might use as the basis of your design efforts?
For example, in a recent thread a member talked about using one of Tang Band’s new 2 by 3 full range drivers in a line array design. Now, I like Tang Band. I find the drivers they produce interesting. Still, before I design something using one of their drivers (or anyone else’s drivers) I would like to know as much as I can about the driver’s performance.
For evaluation purposes, I purchased three of the Tang Band W23-972S drivers. This version of the driver features the “ppm” cone. They included a three-page set of specifications and test results with my order. I include the “log” frequency response graph supplied by Tang Band. I am new to this so I hope this works.
An externally hosted image should be here but it was not working when we last tested it.
This graph shows the driver meeting the specified frequency response with a plus or minus 6 db range. This is not an unusual quality of performance in a modern driver. Still, as a designer, I am interested in how the peak output from the driver is shown occurring at just below 8 kHz, how it is down about 3 db by 10 kHz, and is down an additional 4 db just shy of 15 kHz. I am also interested in the falling response in the lower frequencies and the slight roughness in output between 1 kHz and 3 kHz. Output falls nearly 10 db between 800 Hz and 150 Hz and there are two 4 db dips at 1.2 kHz and 2.5 or 2.6 kHz. These are all areas I would like to smooth in my final design.
Most of the time I like to do my own testing to supplement the manufacturers testing. The test system used for the above test is not the only kind of acoustical testing available and my test system is different. I happen to prefer impulse testing, and I like to examine frequency response graphs using a linear scale instead of a log scale. Unless otherwise noted, you can assume that the testing is done at one meter. I also find value in converting the impulse response to frequency response in two segments to produce three frequency response graphs. I want to see the summed frequency response, and I want to see the frequency response of the onset portion of the impulse and the frequency response of the decay portion of the impulse response. Often the onset and decay response will be markedly different than the summed response, and I don’t believe it ought to be. This is another goal in my design criteria; to get the onset and summed response as close as possible, and to minimize the decay contribution to the output of the loudspeaker or driver.
I include my testing of the W23-972S drivers. I also changed graph color and superimposed the frequency response graphs of the three samples to show production variations in driver output. Consistent with how resonantly active the cone is, my frequency response graph is a bit more “peaky” than Tang Band’s. While our overall rating of the frequency response is close, I am finding more unevenness in the fine detail of the response than Tang Band’s testing. I believe this will relate to the “sound” of this driver and it is something I would try to correct if I were going to use this driver in a design. It is not, even with my test results, a terrible or unusable driver. It shows certain characteristics, however, that I would try to understand. Once I understood what in its design is making it perform the way it does, it just requires a different set of crossover/loudspeaker design approaches than other drivers to maximize the accuracy of the acoustic output.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Despite how important adherence to a design goal and testing is to me, I do not want to claim that testing and adherence to my set of design goals is the only way to enjoyable sound reproduction. There have been past driver designs and past loudspeakers that have had legions of devoted fans, yet were not perfect in testing. I have included two graphs (the original testing done over 20 years ago) here to show this point. The first is of the Jordan 50 mm Module. This was a well-loved driver in its day, yet, while it lacks the small signal variability of the W23-972S, it is no more flat in its pass band output. There is a huge bell mode breakup resonance, little output above it, generally falling output below it, and a bit of small signal unevenness. Despite it test results, I designed a loudspeaker using this driver; I just designed certain corrective networks to make the output closer to my ideal. Indeed, this is how I started working with Don Spangler, I sent him a letter through Speaker Builder detailing my corrective efforts with the Jordan 50 mm Module.
An externally hosted image should be here but it was not working when we last tested it.
The last graph is of the Quad ESL 63. It was in production at about the same time as the Jordan 50mm. Like the Jordan, there were discerning audiophiles who loved this loudspeaker. In both cases I can understand why. Even in comparison to most of today’s tweeters, they both produced a more coherent onset response and no worse decay response.
An externally hosted image should be here but it was not working when we last tested it.
In sum, I would welcome more acoustical testing of drivers and loudspeakers in this forum. I would love to see results from all of the four major testing schemes currently available. I would also be willing to do more testing and submit the results of those tests to this forum if others are as interested as I am.
Mark