Thought this might be of interest.
Fired up MLSSA yesterday and took some frequency response measures of the TB W4-675SB. Baffle was 6.5"x10", 0.5" roundovers, and woofer center was 3.5" up from baffle bottom.
Due to the rectangular FFT window used, response is good down to about 250 Hz.
To compensate for its primary break up mode, I'd go with a 5 dB notch centered at 7 kHz. Any more and you'll have no high end, any less and you'll still keep that aggressive sound. I'll measure impedance tonight and come up with something.
You really need to be right on axis with this, even 15 degrees off and it drops like a stone above 8 kHz.
It's also plagued by a 5 dB hole an octave wide, centered on 4kHz. As a result, intelligibility with dialog suffers.
Hoped to be able to use this full range but plan is to cross at about 3 kHz to an Aura NSW2. Woofer will have 8 kHz notch + one inductor. Aura will probably be low Q second order, but we'll see what the measures say.
Not shown is low end response; its 3.3L sealed for fb=99Hz and Q = 0.61. It'll be high passed with a second order active, Q=0.9. Will give a nice cross at 100Hz to a 4th order low pass on a sub. Its xmax in this config gives an honest 105dB.
Graph of
response:
http://www3.sympatico.ca/dalfarra/W4657SB.gif
Excel file of responses:
http://www3.sympatico.ca/dalfarra/W4657.xls
DDF
Fired up MLSSA yesterday and took some frequency response measures of the TB W4-675SB. Baffle was 6.5"x10", 0.5" roundovers, and woofer center was 3.5" up from baffle bottom.
Due to the rectangular FFT window used, response is good down to about 250 Hz.
To compensate for its primary break up mode, I'd go with a 5 dB notch centered at 7 kHz. Any more and you'll have no high end, any less and you'll still keep that aggressive sound. I'll measure impedance tonight and come up with something.
You really need to be right on axis with this, even 15 degrees off and it drops like a stone above 8 kHz.
It's also plagued by a 5 dB hole an octave wide, centered on 4kHz. As a result, intelligibility with dialog suffers.
Hoped to be able to use this full range but plan is to cross at about 3 kHz to an Aura NSW2. Woofer will have 8 kHz notch + one inductor. Aura will probably be low Q second order, but we'll see what the measures say.
Not shown is low end response; its 3.3L sealed for fb=99Hz and Q = 0.61. It'll be high passed with a second order active, Q=0.9. Will give a nice cross at 100Hz to a 4th order low pass on a sub. Its xmax in this config gives an honest 105dB.
Graph of
response:
http://www3.sympatico.ca/dalfarra/W4657SB.gif
Excel file of responses:
http://www3.sympatico.ca/dalfarra/W4657.xls
DDF
Took the impedance tonight at 3 drive levels (small, med and large signal, all within the published xmax). Results graphically at:
http://www3.sympatico.ca/dalfarra/W4657SBZ.gif
and the excel file's been updated with the impedance data, at the link posted earlier.
It’s pretty apparent that this is a fairly basic motor. Large inductance, no copper. As input power increases, so does inductance. This isn't too big an issue as a load for a trap cct, but it betrays that the distortion in the high end will be a bit heavier than average due to inductance modulation. Given that the inductance modulation is minimal below 3 kHz, that’s where the first stab at the crossover will be. The damping at resonance also shifts, even with moderate input power. This means the suspension is either non linear or the published xmax is hugely optimistic. Finally, an impedance anomaly tracks the break up at 6.8kHz.
One nice outcome was that fs was within 1% of Unibox's prediction. The prediction assumed "heavy stuffing" (which it was) meaning that Unibox looks to assume a realistic Ql for stuffing.
DDF
http://www3.sympatico.ca/dalfarra/W4657SBZ.gif
and the excel file's been updated with the impedance data, at the link posted earlier.
It’s pretty apparent that this is a fairly basic motor. Large inductance, no copper. As input power increases, so does inductance. This isn't too big an issue as a load for a trap cct, but it betrays that the distortion in the high end will be a bit heavier than average due to inductance modulation. Given that the inductance modulation is minimal below 3 kHz, that’s where the first stab at the crossover will be. The damping at resonance also shifts, even with moderate input power. This means the suspension is either non linear or the published xmax is hugely optimistic. Finally, an impedance anomaly tracks the break up at 6.8kHz.
One nice outcome was that fs was within 1% of Unibox's prediction. The prediction assumed "heavy stuffing" (which it was) meaning that Unibox looks to assume a realistic Ql for stuffing.
DDF
I have a pair of 657s and 871s from a swap meet that I never got around to using. Thanks DDF for the measurements! 
I'm thinking of using the w4-657 and w3-871 in a 2-way with the 871 crossed over around 700 hz. I'm leaning this way for a few reasons:
First project after reading up on speaker building for the past months.
700 hz appears to be the approx. range where bsc needs to be applied. Hopefully, I can avoid the bsc and just pad down the 871s?
I'm trying to avoid too many components in the crossover (i.e. trying to stay away from the 8khz breakup and the 4khz dip.
Need to use these drivers up before I can justify purchasing more.
If I get ambitious, I might try the drivers in a transmission line.
I'm hoping that these drivers will work well as rear surround speakers.
Thanks again for the measurements and comments.
regards
Dave
I'm thinking of using the w4-657 and w3-871 in a 2-way with the 871 crossed over around 700 hz. I'm leaning this way for a few reasons:
First project after reading up on speaker building for the past months.
700 hz appears to be the approx. range where bsc needs to be applied. Hopefully, I can avoid the bsc and just pad down the 871s?
I'm trying to avoid too many components in the crossover (i.e. trying to stay away from the 8khz breakup and the 4khz dip.
Need to use these drivers up before I can justify purchasing more.
If I get ambitious, I might try the drivers in a transmission line.
I'm hoping that these drivers will work well as rear surround speakers.
Thanks again for the measurements and comments.
regards
Dave
Very interesting
I was thinking of using these drivers in a 3-way like you have. My comments were posted here. I now guess that these drivers don't deserve too much money spent on good accompanying components... I'll try to get an inexpensive tweeter and woofer to tie up with these things.
Your box dimensions too are useful.... I can use them ready-made for my purpose. Do you have the T/S parameters handy, specially the Vas?
I was thinking of using these drivers in a 3-way like you have. My comments were posted here. I now guess that these drivers don't deserve too much money spent on good accompanying components... I'll try to get an inexpensive tweeter and woofer to tie up with these things.
Your box dimensions too are useful.... I can use them ready-made for my purpose. Do you have the T/S parameters handy, specially the Vas?
Hi,
TS Params for my two drivers here:
http://www3.sympatico.ca/dalfarra/W4657TS.xls
They tend to get gritty up top. I plan to cross them to D26NCs. Very linear tweeter, should be nice, and it has a small faceplate so it should make for a really nice and smooth off axis response.
Low frequency end depends on what you want to do with it. The real sweet spot for these drivers is in a small box against wall delivering good bass up to 105 dB. If you want to use as a mid for a big room system, don't spend too much money on the woofer, the mids aren't worth it in that application. Sensitivity is about 85 dB SPL without BDC, so go with a moderately low sensitivity woofer.
I'd look at woofers with a 2 pi sensitivity of 85 dB. Dayton/Silver flute maybe, cross em about 300 Hz in a manner to minimize floor contribution.
Dave
TS Params for my two drivers here:
http://www3.sympatico.ca/dalfarra/W4657TS.xls
They tend to get gritty up top. I plan to cross them to D26NCs. Very linear tweeter, should be nice, and it has a small faceplate so it should make for a really nice and smooth off axis response.
Low frequency end depends on what you want to do with it. The real sweet spot for these drivers is in a small box against wall delivering good bass up to 105 dB. If you want to use as a mid for a big room system, don't spend too much money on the woofer, the mids aren't worth it in that application. Sensitivity is about 85 dB SPL without BDC, so go with a moderately low sensitivity woofer.
I'd look at woofers with a 2 pi sensitivity of 85 dB. Dayton/Silver flute maybe, cross em about 300 Hz in a manner to minimize floor contribution.
Dave
A reconstructed impulse and the time file are close, but not exactly the same thing.
Still, is the frq file available? I must have missed it somehow. Your Web site seems to contain only gifs and xls files.
Also, if you don't mind getting the word out, while Tang Band allows plus or minus 12 Hz deviation from FS/FO, I have only measured four -inch transducers with FS/FO between 10 to 15 Hz higher than specified. It is rare to find one that is lower than spec. Something to keep in mind when selecting and designing enclosures for four and five-inch TB transducers.
Mark
Still, is the frq file available? I must have missed it somehow. Your Web site seems to contain only gifs and xls files.
Also, if you don't mind getting the word out, while Tang Band allows plus or minus 12 Hz deviation from FS/FO, I have only measured four -inch transducers with FS/FO between 10 to 15 Hz higher than specified. It is rare to find one that is lower than spec. Something to keep in mind when selecting and designing enclosures for four and five-inch TB transducers.
Mark
Mark, my first post links to an excel file.
The measure is convolved with the window of my choosing to make it quasi anechoic, but if you accept that, then the impulse response will be attainable straight away as an IFT of the data from the excel file.
The driver was min phase.
Bigger issue for you is that the measure is on the baffle of my design and choice.
The measure is convolved with the window of my choosing to make it quasi anechoic, but if you accept that, then the impulse response will be attainable straight away as an IFT of the data from the excel file.
The driver was min phase.
Bigger issue for you is that the measure is on the baffle of my design and choice.
And that is kinda the point. I have some experience with this transducer. I also have a modification design to improve the transducer's performance.
There are, however, differences in our measurements. This also gets to where I want to release the modification design.
diyaudio is easy. Brick and mortar magazines are less easy, but AudioXpress and Popular Science have larger readerships.
I have attached a gif of my measurements. These are far field measurements off of an infinite baffle (the term applies both to the baffle dimensions and the complete separation of the backwave from the test environment).
The response is typical of a metal diaphragm. Lots of undesireable vibration modes of high Q. There is also an edge hole/surround problem below 1 kHz. It is barely visible in the stock response because the mid treble diaphragm vibration modes swamp the response down to 1 kHz. When you control for the mid treble vibration modes (whether mechanically or electrically with a notch filter) the edge hole/surround problem becomes visible. As a result, a surround modification is required in addition to the cone modification.
The onset response shows that the mass and VC inductance are just too high for 20 kHz response. Yet with just a simple two component passive equalizer I can get the on-axis frequency response out to 20 kHz. This equalizer is providing a high frequency boost. Yet doing so will make the impulse response (greatly improved by the cone and surround modification) worse. Yet the transient misbehaviors are isolated to frequencies above 13 kHz and so are not as audible as they might be at lower frequencies. Part of the problem is enhancing yet another vibration mode with a center frequency just below 20 kHz.
I would judge the modification to be of moderate difficulty. It requires some mechanical skill and a steady hand. The cone modification can be done from the underside (with increasing difficultly). The basket has very narrow supports and thus allows better than average access to the underside of the cone.
The metal diaphragm is fairly robust. With care, the modification is removable without damaging the cone or surround. Should you mess up in application, you should be able to start over. Still, as with all mechanical modifications, it has to be a proceed at your own risk.
In sum, if you were going to try to redesign a metal diaphragmed transducer, this is a good one to choose.
But all of this modification stuff is for later. For now, it is just about the stock performance under standard and comparable conditions.
There are, however, differences in our measurements. This also gets to where I want to release the modification design.
diyaudio is easy. Brick and mortar magazines are less easy, but AudioXpress and Popular Science have larger readerships.
I have attached a gif of my measurements. These are far field measurements off of an infinite baffle (the term applies both to the baffle dimensions and the complete separation of the backwave from the test environment).
The response is typical of a metal diaphragm. Lots of undesireable vibration modes of high Q. There is also an edge hole/surround problem below 1 kHz. It is barely visible in the stock response because the mid treble diaphragm vibration modes swamp the response down to 1 kHz. When you control for the mid treble vibration modes (whether mechanically or electrically with a notch filter) the edge hole/surround problem becomes visible. As a result, a surround modification is required in addition to the cone modification.
The onset response shows that the mass and VC inductance are just too high for 20 kHz response. Yet with just a simple two component passive equalizer I can get the on-axis frequency response out to 20 kHz. This equalizer is providing a high frequency boost. Yet doing so will make the impulse response (greatly improved by the cone and surround modification) worse. Yet the transient misbehaviors are isolated to frequencies above 13 kHz and so are not as audible as they might be at lower frequencies. Part of the problem is enhancing yet another vibration mode with a center frequency just below 20 kHz.
I would judge the modification to be of moderate difficulty. It requires some mechanical skill and a steady hand. The cone modification can be done from the underside (with increasing difficultly). The basket has very narrow supports and thus allows better than average access to the underside of the cone.
The metal diaphragm is fairly robust. With care, the modification is removable without damaging the cone or surround. Should you mess up in application, you should be able to start over. Still, as with all mechanical modifications, it has to be a proceed at your own risk.
In sum, if you were going to try to redesign a metal diaphragmed transducer, this is a good one to choose.
But all of this modification stuff is for later. For now, it is just about the stock performance under standard and comparable conditions.
Attachments
very interesting
I have a pair of these laying around that I had used in a BL horn experiment some years ago. At the time I was impressed with their performance in the horn (for the price). I look forward to you posting your driver mods. I'd like to give them a try, and maybe build another small backhorn to put them in.
PJN
I have a pair of these laying around that I had used in a BL horn experiment some years ago. At the time I was impressed with their performance in the horn (for the price). I look forward to you posting your driver mods. I'd like to give them a try, and maybe build another small backhorn to put them in.
PJN
I remember now, seeing your measure when I first started playing with the driver.
Rest assured, what you see in my plots is what I got. These are properly taken quasi anechoic measures, via MLSSA/B&K windowed rectangular to maintain resolution. Trust me, they're good. Looks like the QC on these drivers may be a bit wild, yours and mine look like different drivers in some respects.
I've placed the tim files back on my ftp site, in a folder "w4657". Let me know how you make out with them.
I also included BDC models of the driver in cabinet, plus a jpg. You can work backwards to determine the quasi anechoic response in infinite baffle. Surprisingly enough, the driver must have a rising low end as there's little diffractive roll off in my samples. The balanced sounding bass to mids bare this out.
FWIW, the inducance characteristics and high freq distortion of this driver are a bit excessive IME, and boosting the treble would leave you with more crummy high end. I settled on D26NCs to drop in, some day.
Rest assured, what you see in my plots is what I got. These are properly taken quasi anechoic measures, via MLSSA/B&K windowed rectangular to maintain resolution. Trust me, they're good. Looks like the QC on these drivers may be a bit wild, yours and mine look like different drivers in some respects.
I've placed the tim files back on my ftp site, in a folder "w4657". Let me know how you make out with them.
I also included BDC models of the driver in cabinet, plus a jpg. You can work backwards to determine the quasi anechoic response in infinite baffle. Surprisingly enough, the driver must have a rising low end as there's little diffractive roll off in my samples. The balanced sounding bass to mids bare this out.
FWIW, the inducance characteristics and high freq distortion of this driver are a bit excessive IME, and boosting the treble would leave you with more crummy high end. I settled on D26NCs to drop in, some day.
The only concern with differences in measurement is what they may mean for a modification. I am not criticizing your measurements. If there is a wide range of performance differences amongst stock units, then it becomes impossible to predict what you may get whether you stay with stock or with a modified unit. I do agree with you that there is little correspondence between your postings and my postings. (And my versions in stock condition have the midrange compressed bright sound consistent with the measured rise in output.)
So, one avenue of investigation is to ask that others test and post. Can we cobble together data on more units?
Also, equalization is just an option. The question is whether the end performance of a crossover or equalized top end is better (or less crummy). Because I do worry about transient response, my choice would be to equalize rather than crossover to a tweeter. Regardless, I have documented in fine resolution the crumminess of equalization.
Also, another point to all of this . . . assuming that we can come to an understanding on the measured differences that make the modification viable, what do we have? What should we use as a reference for comparison? We might use the new Jordan. This URL was posted as part of a Bandor/Jordan thread. It shows the mlssa performance of the new 53 mm Jordan (JXR6).
http://www.hifisound.de/oxid/out/ox...EJJ-1110856.pdf
Limiting the analysis to 20 kHz, the response in the Hobby HIFI measurements is plus or minus six db. The unmodified Tang Band cannot do plus or minus six dB.
The modified Tang Band cannot do 30 kHz. It can, however, make it to 20 kHz and it can go lower. It has a larger cone and a lower Fs (even when you consider the tendency for the actual unit's Fs to measure higher than rated). My modified units measure better than plus or minus six dB. They do about plus or minus four dB, with fewer dips and peaks than the Jordans in the Hobby HIFI tests.
Just for fun, I would like to use the Hobby HIFI tests of the JXR6 as a reference point and compare and contrast the modified Tang Band to it. This would just be for fun. And experimenting with the Tang Band is a lot cheaper than smearing tooth paste on the fragile metal cone of the JXR6 (something suggested in that other thread).
Lastly, edge holes . . . a phenonema of the outer edge of the cone. This is where the cone couples (or not) to the surround. Generally characterized by a rise in output at frequencies just below a sharper dip in the response. The center frequency of the dip is related to the diameter of the cone. The larger the cone, the lower the frequency of the edge hole. For the Tang Band, the entire region impacted by the edge hole phenomena would be from about 800 to 3 kHz.
Interesting (and back to that other thread) Jordan's own waterfall plots for the JX53 and JX92S show what appear to be edge holes. They do not appear in Jordan's frequency response tests or in the Hobby HIFI tests. Also, Tang Band's frequency response testing never shows edge holes but their drivers do suffer from the problem and many of their three-inch transducers have a design feature (similar to the second modification for the W4-657s) that minimizes the edge hole problem.
So, one avenue of investigation is to ask that others test and post. Can we cobble together data on more units?
Also, equalization is just an option. The question is whether the end performance of a crossover or equalized top end is better (or less crummy). Because I do worry about transient response, my choice would be to equalize rather than crossover to a tweeter. Regardless, I have documented in fine resolution the crumminess of equalization.
Also, another point to all of this . . . assuming that we can come to an understanding on the measured differences that make the modification viable, what do we have? What should we use as a reference for comparison? We might use the new Jordan. This URL was posted as part of a Bandor/Jordan thread. It shows the mlssa performance of the new 53 mm Jordan (JXR6).
http://www.hifisound.de/oxid/out/ox...EJJ-1110856.pdf
Limiting the analysis to 20 kHz, the response in the Hobby HIFI measurements is plus or minus six db. The unmodified Tang Band cannot do plus or minus six dB.
The modified Tang Band cannot do 30 kHz. It can, however, make it to 20 kHz and it can go lower. It has a larger cone and a lower Fs (even when you consider the tendency for the actual unit's Fs to measure higher than rated). My modified units measure better than plus or minus six dB. They do about plus or minus four dB, with fewer dips and peaks than the Jordans in the Hobby HIFI tests.
Just for fun, I would like to use the Hobby HIFI tests of the JXR6 as a reference point and compare and contrast the modified Tang Band to it. This would just be for fun. And experimenting with the Tang Band is a lot cheaper than smearing tooth paste on the fragile metal cone of the JXR6 (something suggested in that other thread).
Lastly, edge holes . . . a phenonema of the outer edge of the cone. This is where the cone couples (or not) to the surround. Generally characterized by a rise in output at frequencies just below a sharper dip in the response. The center frequency of the dip is related to the diameter of the cone. The larger the cone, the lower the frequency of the edge hole. For the Tang Band, the entire region impacted by the edge hole phenomena would be from about 800 to 3 kHz.
Interesting (and back to that other thread) Jordan's own waterfall plots for the JX53 and JX92S show what appear to be edge holes. They do not appear in Jordan's frequency response tests or in the Hobby HIFI tests. Also, Tang Band's frequency response testing never shows edge holes but their drivers do suffer from the problem and many of their three-inch transducers have a design feature (similar to the second modification for the W4-657s) that minimizes the edge hole problem.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.