response curves
Thanks Mark:
The curves are nice. I wish I had the skill set to understand them. Here are a few questions that come to mind:
Will the curves change once the drivers are broken in?
The low frequency role off seems quite high when compared to the published specs (fe127) is this a part answer to my first question?
The spikes and troughs seem to be consistent amongst all three drivers with the 126/127 curves being almost identical with a slight off set. Are there treatments that will help to smooth out the curves to improve the driver or are we trying to make a silk purse out of a sow's ear?
The work you did with the TB driver was very impressive do these drivers hold similar possibility?
Thanks again for your efforts. Lot to discuss here now.
Thanks Mark:
The curves are nice. I wish I had the skill set to understand them. Here are a few questions that come to mind:
Will the curves change once the drivers are broken in?
The low frequency role off seems quite high when compared to the published specs (fe127) is this a part answer to my first question?
The spikes and troughs seem to be consistent amongst all three drivers with the 126/127 curves being almost identical with a slight off set. Are there treatments that will help to smooth out the curves to improve the driver or are we trying to make a silk purse out of a sow's ear?
The work you did with the TB driver was very impressive do these drivers hold similar possibility?
Thanks again for your efforts. Lot to discuss here now.
Lionel said:
Difficult to explain... but I try.
If I watch to the second graphic I see that the black line has only peaks and nearly no gaps between 2 and 10 k. In the contrary the red line seems to be really more tormented in this area so I guess more difficult to tame...
But perhaps I am wrong
That's a fair interpretation. My take on it is that the 127 is nearly +-3db to 5kHz before it goes beyond that. This is why I mentioned a filter centered around 7k (guess) might be all that's needed to clean it up well, IMHO. I appreciate Mark including the FE103E which is known to be a relatively smooth driver. Now we can see it.
MarkMcK said:
Mark - nice work!
I am disapointed that the published Fostex specs do not show the same magnitude at the peaks and dips nor a rising response on the 127.
Also, the efficiency is published as 89, 93 and 91 dB for the FE103, 126 and 127 respectively. I don't see that big of a difference in efficiency.
Mark, do you also have the capabilities of measuring the T/S parameters? It would be interesting to see how close they are to the published Fostex specs.
Lionel said:
No, the 103!
Cheers,
Gio.
The important question in relation to this thread is "Where do we go from here?"
Many posts ago, one member made a statement against using modified drivers. No other member opposed that position. From the results I posted and my listening to the three drivers reproduce the test signal; I do not believe we can have a reference quality loudspeaker using any of the three drivers in stock condition.
If we modify, how far are you willing to go with the modification? Do you want to limit it to prefiltering only? If so, then the major cone vibration mode cannot be controlled. It is of the wrong shape and magnitude to be complemented with a passive speaker level notch filter. Just is.
Now, about interpreting the graphs. Try to keep in mind that people often see data as they want to see it and not as it really is. Our tendencies before we see the data can influence our conclusions based upon any data set.
First, all far field, wide band anechoic or quasi-anechoic testing rolls off the low frequencies. This is a function of the size of the anechoic chamber, either the actual or the simulated chamber. It is this chamber size phenomenon that is the cause of the low frequency roll-off seen in the three driver comparison graphs. For the test done here, 200 to 300 Hz is the lowest valid frequency. You have to do one test to get the low frequency and a different test to get the everything else response of a driver. I will be posting low frequency response results at a later date.
Now, I have no vested interest in the FE103E, the FE126E, or the FE127E. When I evaluate the data in light of the criteria set up for this project, my choice is the FE126E. I have an argument for why I prefer the 126, but others may have other preferences. Anyway, here is my argument:
First, one of the conditions for this reference project was high sensitivity. The FE 126E has the highest sensitivity. You have to look at the lowest frequencies to compare sensitivity in a bandwidth region not dominated by cone vibration modes. As I have repeatedly stated, when the output of a driver may vary by as much as 20 db over the specified driver bandwidth, what does a one number sensitivity rating mean? Answer: very little.
Yet, at these lower frequencies the FE126E is more sensitive than either the FE103E or the FE127E. What may be surprising, but really isn’t when you compare motor strength, is that the FE103E is more sensitive than the FE127E.
Two, the goal should be fairly balanced octave-to-octave output. The FE126E is more balanced that the FE127E. Indeed, the vibration modes of the cones of the FE126E and the FE127E are almost identical. They may not look identical because the FE126E does not suffer from a dip in the response between 3 and 10 kHz that the FE127E does (and the cone vibration modes are shifted slightly lower in frequency for the FE126E in contrast to the FE127E). If you compensate for the lack of the dip in response, then the FE126E is almost identical to the FE127E. If you attempt to filter out the cone vibration modes, the dip in the FE127E’s response becomes important.
Any attempt at filtering is attempting to approach the ideal response level of the driver. As we pull down the peaks, if the filter response to vibration mode response is not a perfect complementary match, we will also pull down areas surrounding the peaks. With the FE127E’s response already down in this very busy vibration mode region, we will suppress the output even farther. This will impede our ability to achieve octave-to-octave balance.
So, in sum, let us assume that a plus or minus 3-db range is our goal. With the FE126E we would have more sensitivity and a greater chance of achieving a generally even octave-to-octave output within a plus or minus 3-db range.
Last, the notch just shy of 2 kHz is not a measurement aberration. Remember how I described the test set-up. Tim also sent me another driver along with the FE127E. I simply swapped out the Fostex drivers and installed this other driver. All else in the test set up remained the same. This fourth driver does not show the dip in response seen in all the Fostex drivers.
Just for comparison sake, I attach the response graph of this fourth driver.
Good designing and good building,
Mark
Many posts ago, one member made a statement against using modified drivers. No other member opposed that position. From the results I posted and my listening to the three drivers reproduce the test signal; I do not believe we can have a reference quality loudspeaker using any of the three drivers in stock condition.
If we modify, how far are you willing to go with the modification? Do you want to limit it to prefiltering only? If so, then the major cone vibration mode cannot be controlled. It is of the wrong shape and magnitude to be complemented with a passive speaker level notch filter. Just is.
Now, about interpreting the graphs. Try to keep in mind that people often see data as they want to see it and not as it really is. Our tendencies before we see the data can influence our conclusions based upon any data set.
First, all far field, wide band anechoic or quasi-anechoic testing rolls off the low frequencies. This is a function of the size of the anechoic chamber, either the actual or the simulated chamber. It is this chamber size phenomenon that is the cause of the low frequency roll-off seen in the three driver comparison graphs. For the test done here, 200 to 300 Hz is the lowest valid frequency. You have to do one test to get the low frequency and a different test to get the everything else response of a driver. I will be posting low frequency response results at a later date.
Now, I have no vested interest in the FE103E, the FE126E, or the FE127E. When I evaluate the data in light of the criteria set up for this project, my choice is the FE126E. I have an argument for why I prefer the 126, but others may have other preferences. Anyway, here is my argument:
First, one of the conditions for this reference project was high sensitivity. The FE 126E has the highest sensitivity. You have to look at the lowest frequencies to compare sensitivity in a bandwidth region not dominated by cone vibration modes. As I have repeatedly stated, when the output of a driver may vary by as much as 20 db over the specified driver bandwidth, what does a one number sensitivity rating mean? Answer: very little.
Yet, at these lower frequencies the FE126E is more sensitive than either the FE103E or the FE127E. What may be surprising, but really isn’t when you compare motor strength, is that the FE103E is more sensitive than the FE127E.
Two, the goal should be fairly balanced octave-to-octave output. The FE126E is more balanced that the FE127E. Indeed, the vibration modes of the cones of the FE126E and the FE127E are almost identical. They may not look identical because the FE126E does not suffer from a dip in the response between 3 and 10 kHz that the FE127E does (and the cone vibration modes are shifted slightly lower in frequency for the FE126E in contrast to the FE127E). If you compensate for the lack of the dip in response, then the FE126E is almost identical to the FE127E. If you attempt to filter out the cone vibration modes, the dip in the FE127E’s response becomes important.
Any attempt at filtering is attempting to approach the ideal response level of the driver. As we pull down the peaks, if the filter response to vibration mode response is not a perfect complementary match, we will also pull down areas surrounding the peaks. With the FE127E’s response already down in this very busy vibration mode region, we will suppress the output even farther. This will impede our ability to achieve octave-to-octave balance.
So, in sum, let us assume that a plus or minus 3-db range is our goal. With the FE126E we would have more sensitivity and a greater chance of achieving a generally even octave-to-octave output within a plus or minus 3-db range.
Last, the notch just shy of 2 kHz is not a measurement aberration. Remember how I described the test set-up. Tim also sent me another driver along with the FE127E. I simply swapped out the Fostex drivers and installed this other driver. All else in the test set up remained the same. This fourth driver does not show the dip in response seen in all the Fostex drivers.
Just for comparison sake, I attach the response graph of this fourth driver.
Good designing and good building,
Mark
Attachments
Looking back at the thread I think the Tangband was left out of the running due to availability more than it being a modified driver. There was a question of using a modified driver as a reference but IMHO, getting the most from low cost drivers almost necessitates some modifications. The 40-1197 is another example of this. Does anyone object to mechanical mods of the FE127E?
Some of the modifications of driver you use to propose can requiered specific "surgeon's skill". I imagine that the obtained results can vary dramatically from one person to other which is working against the original goal to built a "reference project".
I take this occasion to tank you very much for your fruitfull and generous participation to this forum.
Kind regards.
Hi
I've been following this thread with interest. Although I've no plans to build the speaker, it's proving a good education and I'm looking forward to reading about the results.
Having established that I've no right to chip in, can I chip in re the modified drivers?
The idea of a 'base-level' reference speaker, so anyone who is curious can see what all the fullrange fuss is about, is a really good idea. Don't put off anyone by doing anything too complicated to the drivers. Would it be better to use an unmodified driver to begin with and then report back on modifications later? That way you would have your reference point and could evolve the design from that point on.
Colin
I've been following this thread with interest. Although I've no plans to build the speaker, it's proving a good education and I'm looking forward to reading about the results.
Having established that I've no right to chip in, can I chip in re the modified drivers?
The idea of a 'base-level' reference speaker, so anyone who is curious can see what all the fullrange fuss is about, is a really good idea. Don't put off anyone by doing anything too complicated to the drivers. Would it be better to use an unmodified driver to begin with and then report back on modifications later? That way you would have your reference point and could evolve the design from that point on.
Colin
driver modifications
I've been following the development of this project with interest.
It is interesting to see the development process and I may well tag along with the final results.
Question: Would modifications to the driver change design characteristics of the enclosure? BSC or filter networks are easy to change but the cabinet itself is another matter.
If the set-up could be built with a stock driver then a modified driver substituted we could actually have two different/related projects: diyReference Stock and diyReference Open. It would be a chance for us to develop our "hot-rodding" skills.
I've been following the development of this project with interest.
It is interesting to see the development process and I may well tag along with the final results.
Question: Would modifications to the driver change design characteristics of the enclosure? BSC or filter networks are easy to change but the cabinet itself is another matter.
If the set-up could be built with a stock driver then a modified driver substituted we could actually have two different/related projects: diyReference Stock and diyReference Open. It would be a chance for us to develop our "hot-rodding" skills.
I'm going to risk asking Mark to come up with a mod to handle the cone vibration of the FE127E. If it's typical of his other solutions it shouldn't change the T/S parameters much, if at all. Usually it includes damping the cone at a particular point just enough to control the vibration. We may also want to consider a phase plug.
Stock vrs Modified
I think that "alandarkdale" is on to some thing.
I suspect the specific size and shape of the enclosure is not going to change a great deal, unless the modifications change the T/S parameters significantly.
I went through a similar learning experience with a Fe167e project and was quite astounded at the change in the sound of the completed project. Confidence grows with experience.
I think we already have designed a couple of enclosures that could be built to fit the stock drivers based on published specs. These enclusures will sound fine with the driver selected. This may be a good starting place fo those that are a little shy to take the scalpel to a driver.
For those of us that are more adventurous, it is fun to take a calculated risk and "play" with a driver to attempt to get more from it than the manufacturer expected or intended. The best example I can think of is the tinkering work that has been done on the Radio Shack 40-1197 driver. This driver was not expensive and the modifications although permenant are not particularly difficult. The hardest thing is mustering up the confidence to do the job. Not every one is up for that until they see it completed and hear the results. Once the modifications are completed, most people are impressed with how well they work and how easy they were.
I personally have a vested interest in the 127 driver as I already have a few of them. I suspect any modifications to the 126 will be very similar to the ones needed for the 127 so no big deal for me.
This is a DIY discussion, so lets get the best out of these driver possible. Sharpen up the scapel, get out the cone treatment glues, buy a new bottle of nail polish, roll out a few duct seal worms lets see what we can do.
I guess I am in for the modified driver.
I think that "alandarkdale" is on to some thing.
I suspect the specific size and shape of the enclosure is not going to change a great deal, unless the modifications change the T/S parameters significantly.
I went through a similar learning experience with a Fe167e project and was quite astounded at the change in the sound of the completed project. Confidence grows with experience.
I think we already have designed a couple of enclosures that could be built to fit the stock drivers based on published specs. These enclusures will sound fine with the driver selected. This may be a good starting place fo those that are a little shy to take the scalpel to a driver.
For those of us that are more adventurous, it is fun to take a calculated risk and "play" with a driver to attempt to get more from it than the manufacturer expected or intended. The best example I can think of is the tinkering work that has been done on the Radio Shack 40-1197 driver. This driver was not expensive and the modifications although permenant are not particularly difficult. The hardest thing is mustering up the confidence to do the job. Not every one is up for that until they see it completed and hear the results. Once the modifications are completed, most people are impressed with how well they work and how easy they were.
I personally have a vested interest in the 127 driver as I already have a few of them. I suspect any modifications to the 126 will be very similar to the ones needed for the 127 so no big deal for me.
This is a DIY discussion, so lets get the best out of these driver possible. Sharpen up the scapel, get out the cone treatment glues, buy a new bottle of nail polish, roll out a few duct seal worms lets see what we can do.
I guess I am in for the modified driver.
If the options are the FE126E and the FE127E, then my choice is still the FE126E. Dave has provided a cabinet design for the FE126E, so an enclosure for the low Q driver should not be a problem.
There are many problems with democracy; one of them is the tyranny of the majority. And just how many people voted for a specific driver? I believe it was seven people and the vote was not unanimous. One member stated that they would prefer a shielded driver if it performed as well as the unshielded. In this case, it does not. In all, not much of a majority. When this thread had 230 or so posts, those all came from just 14 members. Still not a large number.
Another problem is when one person appoints himself or herself to speak for all.
Anyway, except for the rather large suck out between 3 and 10 kHz, the cone vibration modes for the 126 and the 127 are the same. In my previous post I explained why that the 127 suck out could be a problem for a modification design.
I consider any obligation I had to this forum or the person donating the FE127E ended when I did the testing. That was all I offered. Of my own interest, I will be modifying the FE126E. Still, I understand that several of the people who voted bought FE127E drivers. One even posted a picture of a large collection of FE127E boxes. I might try porting the FE126E modification design to the FE127E. No promises however. Except for the fact that you have the drivers, why would you want to settle for poorer performance?
I am attaching test results of a first modification to the FE126E. A few points of explanation: Fostex’ banana paper cones are very noisy with vibration modes. They are much worse than polypropylene or the mag/AL alloy coned driver Tim just sent along.
Cone vibration modes may be independent, loosely coupled, or tightly coupled. Fortunately, several vibration modes in the FE126E cone are tightly coupled. That means that controlling one mode may take care of more than one. For the modification documented here, two regions of the cone are mechanically modified and a two-component step filter is used to control the rising frequency response. The critical regions where the mechanical modifications are made are 9 and 17 mm in from the outside edge of the cone.
With this modification, the driver almost qualifies as high fidelity. Except for the response dip just below 2 kHz, the driver does meet a plus or minus 3 db specification from 250 (the 250 is just the limit of the test and not the driver) to just under 16 kHz.
Good designing, good building, and may you experience true democracy,
Mark
There are many problems with democracy; one of them is the tyranny of the majority. And just how many people voted for a specific driver? I believe it was seven people and the vote was not unanimous. One member stated that they would prefer a shielded driver if it performed as well as the unshielded. In this case, it does not. In all, not much of a majority. When this thread had 230 or so posts, those all came from just 14 members. Still not a large number.
Another problem is when one person appoints himself or herself to speak for all.
Anyway, except for the rather large suck out between 3 and 10 kHz, the cone vibration modes for the 126 and the 127 are the same. In my previous post I explained why that the 127 suck out could be a problem for a modification design.
I consider any obligation I had to this forum or the person donating the FE127E ended when I did the testing. That was all I offered. Of my own interest, I will be modifying the FE126E. Still, I understand that several of the people who voted bought FE127E drivers. One even posted a picture of a large collection of FE127E boxes. I might try porting the FE126E modification design to the FE127E. No promises however. Except for the fact that you have the drivers, why would you want to settle for poorer performance?
I am attaching test results of a first modification to the FE126E. A few points of explanation: Fostex’ banana paper cones are very noisy with vibration modes. They are much worse than polypropylene or the mag/AL alloy coned driver Tim just sent along.
Cone vibration modes may be independent, loosely coupled, or tightly coupled. Fortunately, several vibration modes in the FE126E cone are tightly coupled. That means that controlling one mode may take care of more than one. For the modification documented here, two regions of the cone are mechanically modified and a two-component step filter is used to control the rising frequency response. The critical regions where the mechanical modifications are made are 9 and 17 mm in from the outside edge of the cone.
With this modification, the driver almost qualifies as high fidelity. Except for the response dip just below 2 kHz, the driver does meet a plus or minus 3 db specification from 250 (the 250 is just the limit of the test and not the driver) to just under 16 kHz.
Good designing, good building, and may you experience true democracy,
Mark
