I played around with the tweeter level setting a bit, and increased the shelf frequency stepwise from 5k to 7k and bumped up the shelf gain a couple of dB.
Still a dip at 15k, but FR op to 10k is now fairly flat and the 20k peak is not severe.
I have included a zoomed FR plot of the latest itteration and a second plot showing phase response.. nothing wild going on here is it?
Assuming that these home-made measurements are reasonably representative of what is actually going on, I feel that the result is one that will stand up to, and perhaps even surpass, relatively expensive "storebought" equipment.
Rather than trying to hammer out the HF response further with more shelving and filtering of the 20k peak, I think I'll leave it at this as a good compromize.
I feel I'm starting to push the limits on what is really relevant, especially considering the venue (my smallish living room) and the equipment I'm actually working with. I have to remember that my mic is a homegrown Panasonic capsule job, not a Bruel & Kjaer!
The sound is however very pleasant, there are no signs of any audible "snags".
On the record I have playing right now, the overall timbre might seem a bit bright, but I need to do some more listening to various music before passing a final judgement, some recordings ARE actullay bright . Based on that, I may or may not adjust the tweeter level a bit to get a result that actually works in a listening environment, at the end of the day that is actually more important than a ruler flat measured FR.
Once that is done, and I'm happy with the final result, I'll do a series of off-axis measurements and post for final judgement and evaluation!
Still a dip at 15k, but FR op to 10k is now fairly flat and the 20k peak is not severe.
I have included a zoomed FR plot of the latest itteration and a second plot showing phase response.. nothing wild going on here is it?
Assuming that these home-made measurements are reasonably representative of what is actually going on, I feel that the result is one that will stand up to, and perhaps even surpass, relatively expensive "storebought" equipment.
Rather than trying to hammer out the HF response further with more shelving and filtering of the 20k peak, I think I'll leave it at this as a good compromize.
I feel I'm starting to push the limits on what is really relevant, especially considering the venue (my smallish living room) and the equipment I'm actually working with. I have to remember that my mic is a homegrown Panasonic capsule job, not a Bruel & Kjaer!
The sound is however very pleasant, there are no signs of any audible "snags".
On the record I have playing right now, the overall timbre might seem a bit bright, but I need to do some more listening to various music before passing a final judgement, some recordings ARE actullay bright . Based on that, I may or may not adjust the tweeter level a bit to get a result that actually works in a listening environment, at the end of the day that is actually more important than a ruler flat measured FR.
Once that is done, and I'm happy with the final result, I'll do a series of off-axis measurements and post for final judgement and evaluation!
Attachments
Last edited:
I think I asked about this earlier in the thread but don't recall if you answered it - has your microphone ever been calibrated or compared to a reference or another mic? Do you have any certainty that the response of the capsule is flat or do you use any kind of correction?
Since you are now correcting the response by only a few dB here and there, it's important that you have a good reference to "flat". Just curious about this, otherwise it looks like you have made huge strides in getting the tweeter response under control. Kudos to you.
-Charlie
Good evening Charlie,
Regarding the mic, I buildt it, or rather did the capsule modification and amplifier according to the plans on the Linkwitz-lab page.
http://www.linkwitzlab.com/sys_test.htm#Mic
I then managed to obtain a calibration file for this mic, can't remember if it was on the HolmImpulse web-page or where it was, and this has been installed in my HolmImpulse software.
This is of course a generic calibration file, so my mic is in no way an absolute reference. Having said that, for the price and considering I'm only a modest DIY'er, it is probably plenty good enough relative to other limitations, both equipment and skill-wise!
I believe the mic and HolmImpulse on a laptop, not to mention that gee-wizz LspCad software, has made lightyears of difference to my DIY capabilities compared to the devices I was left with in the good old bad days. (ears and simple passive filter calculator programs).
Some of my past, and admittedly less ambitions projects, often ended up sounding nice, but I'm not to sure what the measurements would have looked like!
I did find some comparisons of the mic I buildt where it was lined up with some other mic's, some of them expensive pro stuff, and it stood up to comparison surprisingly well. I think I'd have to do quite a lot of serious projects to justify the cost of going to the next performance step.
I would assume, with a reasonable degree of certainty, that with the generic calibration file, the mic should be good for +/- 0,5-1 dB from 20Hz to about 20kHz.
Yes, the tweeter response is not perfect, but I think it is acceptable. Looking at the factory measurements of a lot of other tweeters (SEAS, Scanspeak, that sort of stuff), it is apparently not uncommon to have some response anomalities up in the top, so my end result is perhaps not too bad for the rreal world after all..
And thanks again to you Charlie, you played an important role in sorting this out. That unintentional difraction groove would have eluded me completely had it not been for your observation!
Regarding the mic, I buildt it, or rather did the capsule modification and amplifier according to the plans on the Linkwitz-lab page.
http://www.linkwitzlab.com/sys_test.htm#Mic
I then managed to obtain a calibration file for this mic, can't remember if it was on the HolmImpulse web-page or where it was, and this has been installed in my HolmImpulse software.
This is of course a generic calibration file, so my mic is in no way an absolute reference. Having said that, for the price and considering I'm only a modest DIY'er, it is probably plenty good enough relative to other limitations, both equipment and skill-wise!
I believe the mic and HolmImpulse on a laptop, not to mention that gee-wizz LspCad software, has made lightyears of difference to my DIY capabilities compared to the devices I was left with in the good old bad days. (ears and simple passive filter calculator programs).
Some of my past, and admittedly less ambitions projects, often ended up sounding nice, but I'm not to sure what the measurements would have looked like!
I did find some comparisons of the mic I buildt where it was lined up with some other mic's, some of them expensive pro stuff, and it stood up to comparison surprisingly well. I think I'd have to do quite a lot of serious projects to justify the cost of going to the next performance step.
I would assume, with a reasonable degree of certainty, that with the generic calibration file, the mic should be good for +/- 0,5-1 dB from 20Hz to about 20kHz.
Yes, the tweeter response is not perfect, but I think it is acceptable. Looking at the factory measurements of a lot of other tweeters (SEAS, Scanspeak, that sort of stuff), it is apparently not uncommon to have some response anomalities up in the top, so my end result is perhaps not too bad for the rreal world after all..
And thanks again to you Charlie, you played an important role in sorting this out. That unintentional difraction groove would have eluded me completely had it not been for your observation!
Last edited:
It's nice to see that you got everything working It's also nice to see that you could follow my descriptions of how to do it, as LspCAD6 isn't all that user friendly, neither is the semi complex process of getting your data and configuring the software, so well done!
It's especially nice when you see the simulated response match up to the real world measured one. The first time mine did I almost couldn't believe it
The SEAS tweeter is a very nice sounding unit imo. One thing to bare in mind is that it has a very wide dispersion when compared to other softdomes. The extra high frequency energy going into the room might explain why it sounds brighter? I personally don't find the tweeter bright at all, just incredibly smooth and well extended, but then tastes do vary.
It's situations like this though that make you realise how useful decent software is.
It's also nice to see that you could follow my descriptions of how to do it, as LspCAD6 isn't all that user friendly, neither is the semi complex process of getting your data and configuring the software, so well done!It's especially nice when you see the simulated response match up to the real world measured one. The first time mine did I almost couldn't believe it
The SEAS tweeter is a very nice sounding unit imo. One thing to bare in mind is that it has a very wide dispersion when compared to other softdomes. The extra high frequency energy going into the room might explain why it sounds brighter? I personally don't find the tweeter bright at all, just incredibly smooth and well extended, but then tastes do vary.
It's situations like this though that make you realise how useful decent software is.
yes, it was both satisfying and amazing to see what LspCad could do and to see heo close the real life result matched!
I didn't find it that hard to use really, except for the hassle of not being able to save, but hey, its a free demo! With your explanations on how to go about, it was quite a breeze actually!
So thanks a lot for introducing me to this software and helpingh me get started!
What I'm most pleased with, is the actual x-over. Getting a good 20dB reverse tweeter dip without using delay circuits impressed me, lightyears from what I got just by slapping in a standard filter slope!
I'm not quite there with the tonal balance yet...
Having listened, slept listened some more, It becomes obvious to me that the speaker could do with a bit more energy in the 1-3k area. After adding the shelving filter, I pulled the overall tweeter level down a bit, and that resulted in the "sag" between 1-3k, which is visible on the plot.
I notice this in some lack of prescence of lead vocals and solo Instruments.
We are not talking about much here, perhaps 1 dB but even 1 dB over a wide frequency range can make an audible difference in tonal balance.
On top of this, the shelving filter has given some lift in the 4-7k region of about 1-1,5 dB, and I believe that this, combioned with the sag between 1-3k makes the sound a bit thin and bright.
So I think I will bring the overall tweeter level up a bit to increase the level at 1-3k, move the shelf up a bit more, and throw in that 23k LP as the 20k peak will inevitably rise as a consequence of pushing the shelving frequency higher.
I agree with you about the DXT tweeter, it seems very smooth to me.
Unlike some other tweeter's I have heard, it dosen't seem to draw attention to it self in any way.
I didn't find it that hard to use really, except for the hassle of not being able to save, but hey, its a free demo! With your explanations on how to go about, it was quite a breeze actually!
So thanks a lot for introducing me to this software and helpingh me get started!
What I'm most pleased with, is the actual x-over. Getting a good 20dB reverse tweeter dip without using delay circuits impressed me, lightyears from what I got just by slapping in a standard filter slope!
I'm not quite there with the tonal balance yet...
Having listened, slept listened some more, It becomes obvious to me that the speaker could do with a bit more energy in the 1-3k area. After adding the shelving filter, I pulled the overall tweeter level down a bit, and that resulted in the "sag" between 1-3k, which is visible on the plot.
I notice this in some lack of prescence of lead vocals and solo Instruments.
We are not talking about much here, perhaps 1 dB but even 1 dB over a wide frequency range can make an audible difference in tonal balance.
On top of this, the shelving filter has given some lift in the 4-7k region of about 1-1,5 dB, and I believe that this, combioned with the sag between 1-3k makes the sound a bit thin and bright.
So I think I will bring the overall tweeter level up a bit to increase the level at 1-3k, move the shelf up a bit more, and throw in that 23k LP as the 20k peak will inevitably rise as a consequence of pushing the shelving frequency higher.
I agree with you about the DXT tweeter, it seems very smooth to me.
Unlike some other tweeter's I have heard, it dosen't seem to draw attention to it self in any way.
Last edited:
That took a bit longer than anticipated..
Well, I have now added a 23kHz 24dB/Oct LP filter in the tweeter channel, and it did a very nice job of lobbing of that 20k Peak!
I looked at the phase response too, and there was virtually no change.
This should give me some more latidude for playing with the shelving filter as I don't have to worry about that 20k peak getting out of hand any more.
Irrespective of that, I'll leave the 23k LP in as there seems to be no trade-offs to it!
Well, I have now added a 23kHz 24dB/Oct LP filter in the tweeter channel, and it did a very nice job of lobbing of that 20k Peak!
I looked at the phase response too, and there was virtually no change.
This should give me some more latidude for playing with the shelving filter as I don't have to worry about that 20k peak getting out of hand any more.
Irrespective of that, I'll leave the 23k LP in as there seems to be no trade-offs to it!
Attachments
The low pass would also act as a barrier against ultrasonic hash and less then adequate low pass filters in D/A converters. Also should something misbehave and oscillate it'd protect the rest of the system from possible damage.
It's nice to see you're having fun with the software and letting your imagination take you places now that you have a way to put into practise any ideas that you might have.
I can see you catching the loudspeaker bug and wanting to design more and more things!
It's nice to see you're having fun with the software and letting your imagination take you places now that you have a way to put into practise any ideas that you might have.
I can see you catching the loudspeaker bug and wanting to design more and more things!
5th,
Good point about that 23k LP keeping various HF "garbage" from being passed on to the amp, that certainly cant hurt!
And Bob,
To your credit, I can confirm that getting rid of that 15-20k peak was an improvement! Very hard to describe the effect as it was more perceivable than directly audible, but it was like removing some sort of a thin veil from the upper frequencies.. very subjective but definitively influencing the listening expereince, so despite my reasoning around sensitivity to HF information etc, I obviously underestimated the well-founded concerns you had on that subject!
I have now tweeked the shelving filter to the point where I have a +/-1,5 dB FR from below 1k (limited by measurements in room, but should be good to where room modes starts kicking in) and up to 12k. Still a dip of 2-3 dB at 15k, but I'll live with that.
Now I'll order up components for the second card and resistors to replace the trim-pots I've been using for tweeking.
Once the fixed resistors are in place and the tweeter level checked and adjusted, I'll do some final measurements including off-axis.
And most important, it sounds great! Filling in that 1dB between 1 and 3k, had the desired effect of getting the presence right, and lobbing off a dB 4-7k, got the balance just right!
The sound is very clear and detailed without being over-emphasised in any particular area..
Seing that it's possible to reach this level of performance as an amateur is really amazing, and these new-found possibilities sure will make the next project even more fun!
The bad thing about building speakers and amps is that you only need so many.. but there's allways friends and family...
Good point about that 23k LP keeping various HF "garbage" from being passed on to the amp, that certainly cant hurt!
And Bob,
To your credit, I can confirm that getting rid of that 15-20k peak was an improvement! Very hard to describe the effect as it was more perceivable than directly audible, but it was like removing some sort of a thin veil from the upper frequencies.. very subjective but definitively influencing the listening expereince, so despite my reasoning around sensitivity to HF information etc, I obviously underestimated the well-founded concerns you had on that subject!
I have now tweeked the shelving filter to the point where I have a +/-1,5 dB FR from below 1k (limited by measurements in room, but should be good to where room modes starts kicking in) and up to 12k. Still a dip of 2-3 dB at 15k, but I'll live with that.
Now I'll order up components for the second card and resistors to replace the trim-pots I've been using for tweeking.
Once the fixed resistors are in place and the tweeter level checked and adjusted, I'll do some final measurements including off-axis.
And most important, it sounds great! Filling in that 1dB between 1 and 3k, had the desired effect of getting the presence right, and lobbing off a dB 4-7k, got the balance just right!
The sound is very clear and detailed without being over-emphasised in any particular area..
Seing that it's possible to reach this level of performance as an amateur is really amazing, and these new-found possibilities sure will make the next project even more fun!
The bad thing about building speakers and amps is that you only need so many.. but there's allways friends and family...
Indeed. I have so many projects that I'd like to try out, but the question in the air goes to the tune of "But where would you put it?" Almost every room in the house that could have a hifi system has a hifi system, all the loudspeakers being my designs, two of them active, two of them passive. There might be an opening in another room though so my fingers are crossed!
So,
Got both cards for L and R finished today and did new off-axis measurements.
Again, measured in 15 degree increments, one file with 0-30 degrees and one with 30-60 degrees.
I think this looks quite OK?
Now it's time to install the cards in to the amplifier chassii and get everything wired up properly, hopefully without running in to any hum-problems!
Got both cards for L and R finished today and did new off-axis measurements.
Again, measured in 15 degree increments, one file with 0-30 degrees and one with 30-60 degrees.
I think this looks quite OK?
Now it's time to install the cards in to the amplifier chassii and get everything wired up properly, hopefully without running in to any hum-problems!
Attachments
Well, this isn't quite over yet...
Some time has now passed since I finished the x-over cards and got the amps finished as well.
Since then, I've been listening quite a bit to the system, and I can not get away from the fact that I'm not entirely happy..
I feel that the upper midrange is a bit to pronounced and at the same time not as transparent as I would like it to be..
The frequency response measured fine enough, so I'm starting to think that despite a nice on axis response, something might not be OK with driver interaction and in room response....
I decided to re-visit some of the many measurements I've done and see if I could find some clues there as to what might be going on..
I dug out the on/off axis responses done with the last x-over configuration with 1700 Hz x-over and filter slopes optimized for phase tracking (using LspCad), and the corresponding measurements done with the first x-over version that had an x-over at 3200 Hz but no optimized phase tracking.
The curves show 0, 30 and 60 degrees.
On axis, the frequency response is fairly smooth for both configurations.
Looking at the of-axis response, I see that with the 1700 HZ x-over the off axis response is not as smooth. Most noticeable, there is a hump between approx. 1700 and 3500 Hz.
So why is this not the case with the higher x-over?
I'm thinking that the reason for this must be that with the lower x-over, there is simply an ubnfavourable match of off-axis response between the SEAS woofer and the DXT tweeter.
As this thread, and my experiementation, developed, a lot of good arguments for running a low x-over was put forth, but I'm starting to think that the priorities might need to be re-visited.
Granted, a higher X-over frequency will allow for less atenuation of woofer break-up and more difficulties with phase-tracking and comb-filtering/lobing.
Having said that, moving any phase-challenges and comb-filtering artefacts up and out of the time-sensitive midrange area should reduce the significance of these effects.
Allso, I looked at the measured distortion of the two configurations. this was quite simmilar except for a slightly more pronounced distortion peak at 1k Hz and a somewhat lower distortion at the highest frequencies.
I allso had a relatively poor phase-reversal null with the high X-over, but then again this was never optimized with LspCad as for the 1700 Hz x-over.
Based on this, my plan is to do the following:
Measure off-axis response properly (unfiltered) for both tweeter and mid and pick the x-over point based on where the off-axis response matches.
I will then use LsPCad to optimize the filter slopes to get the best possible phase tracking.
Two steps forwards, and one back!
Some time has now passed since I finished the x-over cards and got the amps finished as well.
Since then, I've been listening quite a bit to the system, and I can not get away from the fact that I'm not entirely happy..
I feel that the upper midrange is a bit to pronounced and at the same time not as transparent as I would like it to be..
The frequency response measured fine enough, so I'm starting to think that despite a nice on axis response, something might not be OK with driver interaction and in room response....
I decided to re-visit some of the many measurements I've done and see if I could find some clues there as to what might be going on..
I dug out the on/off axis responses done with the last x-over configuration with 1700 Hz x-over and filter slopes optimized for phase tracking (using LspCad), and the corresponding measurements done with the first x-over version that had an x-over at 3200 Hz but no optimized phase tracking.
The curves show 0, 30 and 60 degrees.
On axis, the frequency response is fairly smooth for both configurations.
Looking at the of-axis response, I see that with the 1700 HZ x-over the off axis response is not as smooth. Most noticeable, there is a hump between approx. 1700 and 3500 Hz.
So why is this not the case with the higher x-over?
I'm thinking that the reason for this must be that with the lower x-over, there is simply an ubnfavourable match of off-axis response between the SEAS woofer and the DXT tweeter.
As this thread, and my experiementation, developed, a lot of good arguments for running a low x-over was put forth, but I'm starting to think that the priorities might need to be re-visited.
Granted, a higher X-over frequency will allow for less atenuation of woofer break-up and more difficulties with phase-tracking and comb-filtering/lobing.
Having said that, moving any phase-challenges and comb-filtering artefacts up and out of the time-sensitive midrange area should reduce the significance of these effects.
Allso, I looked at the measured distortion of the two configurations. this was quite simmilar except for a slightly more pronounced distortion peak at 1k Hz and a somewhat lower distortion at the highest frequencies.
I allso had a relatively poor phase-reversal null with the high X-over, but then again this was never optimized with LspCad as for the 1700 Hz x-over.
Based on this, my plan is to do the following:
Measure off-axis response properly (unfiltered) for both tweeter and mid and pick the x-over point based on where the off-axis response matches.
I will then use LsPCad to optimize the filter slopes to get the best possible phase tracking.
Two steps forwards, and one back!
Attachments
What surprises me the most is how flat the off axis curves are for the 3200hz xover, typically you'd expect to see a dip appear as you go off axis corresponding to the off axis droop of the mid/bass due to beaming.
I would be surprised if that small bump played such a large part in the way the loudspeaker comes across. In the 60 degree curve you don't really have a bump, what you have is a small dip at 1700hz. Fill in that dip and the bump pretty much disappears.
We are talking an error of less then +-1dB from flat though, realistically it doesn't get much better then that.
I think you could be perhaps looking at the wrong thing, your lateral off-axis plots for both x-overs are more then acceptable. Try looking at how the driver integration varies through the vertical axis, you might see something that jumps out at you. It might not be a question of crossover point or phase integration, it could be that a peak appears due to diffraction and it's this that's causing the trouble. Bare in mind however that vertical off axis plots will always look a mess, as the drivers will go in and out of phase with one another, this will create dips in the frequency response as the drivers null with one another. What you're looking for are any peaks that stand out above what would be the nominal 'flat' on axis response.
Another thing to consider is trying more baffle step compensation, a lack in this area always gives the impression of falsely forwards mids. As you've got a sub (I think you had a sub?) you can often dial in too little baffle step because it sounds decent as all the bass is there. Maybe you could try balancing the loudspeakers with the sub turned off?
Then there's the room.
Are you used to listening over headphones? Most headphones usually have a downward tilting response too, typically we prefer it this way (I know I do). Some prefer a small broad dip in the 1-3khz area too. Don't be afraid to play around with EQ, sub level and baffle step. It might not end up being ruler flat, but if you prefer it that way then stick with it. I for one much prefer it with the 'subs' running a little hot.
I would be surprised if that small bump played such a large part in the way the loudspeaker comes across. In the 60 degree curve you don't really have a bump, what you have is a small dip at 1700hz. Fill in that dip and the bump pretty much disappears.
We are talking an error of less then +-1dB from flat though, realistically it doesn't get much better then that.
I think you could be perhaps looking at the wrong thing, your lateral off-axis plots for both x-overs are more then acceptable. Try looking at how the driver integration varies through the vertical axis, you might see something that jumps out at you. It might not be a question of crossover point or phase integration, it could be that a peak appears due to diffraction and it's this that's causing the trouble. Bare in mind however that vertical off axis plots will always look a mess, as the drivers will go in and out of phase with one another, this will create dips in the frequency response as the drivers null with one another. What you're looking for are any peaks that stand out above what would be the nominal 'flat' on axis response.
Another thing to consider is trying more baffle step compensation, a lack in this area always gives the impression of falsely forwards mids. As you've got a sub (I think you had a sub?) you can often dial in too little baffle step because it sounds decent as all the bass is there. Maybe you could try balancing the loudspeakers with the sub turned off?
Then there's the room.
Are you used to listening over headphones? Most headphones usually have a downward tilting response too, typically we prefer it this way (I know I do). Some prefer a small broad dip in the 1-3khz area too. Don't be afraid to play around with EQ, sub level and baffle step. It might not end up being ruler flat, but if you prefer it that way then stick with it. I for one much prefer it with the 'subs' running a little hot.
Ah, 5th, your viewpoints and observations are allways rewarding and interresting to read!
Yes, considering "commonly accepted practice", I agree that the good result of the high x-over configuration is somewhat surprising.
I think the SEAS DXT tweeter may play a part in this. Not to long after it was released, it was discussed in various threads, and on one account, somebody compared it with other more conventional wave-guide applications, e.g. those of John Krutke Zaph|Audio .
One of the comments was that the DXT lens/waveguide was too small to have any effect at frequencies below 3-4k Hz.
From this, I'm thinking that with the 1700 Hz filter, the tweeter is starting to widen its dispersion at the lower frequencies, and that this explains the "hump" between approx 1700-3500 Hz.
I further suspect that with the higher x-over, the above mentioned frequency range is covered by the midwoofer, which is starting to display a narrower dispersion at these frequencies, a dispersion that might actually match the dispersion of the DXT tweeter where tie waveguide of the DXT lens is effective.
This might explain why this seems to work where conventional wisdom and conventional tweeters would normally dictate otherwise?
Granted, the "hump" is not severe in magnitude, but it is rather wide, and whereas a small ripple of 1-2 dB might not be very audible, such a difference in level may be audible across an octave or two.
Following my reasoning, a higher x-over point should allso reduce the general disperison of the loudspeaker and thus reduce room-reflection problems.
A while ago, I had some large voight horns with coral Flat8 8" fullranges and Coral H-40 horn tweeters crossed over at about 7k. This was before I buildt my microphone and discovered holm-impulse.
What I can say, is that the frequency response was far from linear.
What I can say was that stereo imaging and clarity was surprisingly good!
All in all, those speakers sounded far better than they had any right to do, but I suspect this was due to the high midrange and low treble beaming of the 8" fullrange and the very narrow dispersion pattern (probably arround 40 deg.) of the horn.
How could this be good?
Well, less early reflections and more direct sound, that's the only explanation I can think of!
allso, having one driver covering the important time-critical midrange area could have played an important part.
As I write this, I really regret having let go of those drivers and boxes before I discovered DIY measurement, it would have been tremendously interresting to measure what those speakers were actually up to, not to mention tweaking them based on actual measurements!
Regarding baffle step, I went all the way and added a full 6dB, and spliced nearfield measurements confirmed that that was (urprisingly) spot on both in terms of chosen frequencies and level.
Allthough this doesn't show up in the gated measurements I keep posting, my system measures flat from 25 Hz up to about 1Khz to the extent I have been able to piece toghether nearfield and large-room farfield measurements.
Room response, that's a tough one.. in principle, I tend to agree with John Krutke here with reference to his last blog-post.. Zaph|Audio
i.e. I agree that it makes sense to go for flat speaker response to get the correct direct radiated sound in to my ears.
The problem occurs when the room interacts in such a way as to blur the difference between direct and reflected soundfield, i.e. early reflections, and that takes me back to my "hang-up" on dispersion control.
Dr Earl Geddes seems to have very firm opinions on this, many of which I find it hard to disagree with.
Perhaps I should really go back to 8" midwoofers and horn-loaded tweeters crossed over high?
No, seriously, I'd very much like to make the best use of what I have right now.
The interresting thing is that I now have two speakers and two amp-modules with x-over cards up and running. Thgis means I can actually modify one set and leave the other as is for proper A/B testing.
Yes, considering "commonly accepted practice", I agree that the good result of the high x-over configuration is somewhat surprising.
I think the SEAS DXT tweeter may play a part in this. Not to long after it was released, it was discussed in various threads, and on one account, somebody compared it with other more conventional wave-guide applications, e.g. those of John Krutke Zaph|Audio .
One of the comments was that the DXT lens/waveguide was too small to have any effect at frequencies below 3-4k Hz.
From this, I'm thinking that with the 1700 Hz filter, the tweeter is starting to widen its dispersion at the lower frequencies, and that this explains the "hump" between approx 1700-3500 Hz.
I further suspect that with the higher x-over, the above mentioned frequency range is covered by the midwoofer, which is starting to display a narrower dispersion at these frequencies, a dispersion that might actually match the dispersion of the DXT tweeter where tie waveguide of the DXT lens is effective.
This might explain why this seems to work where conventional wisdom and conventional tweeters would normally dictate otherwise?
Granted, the "hump" is not severe in magnitude, but it is rather wide, and whereas a small ripple of 1-2 dB might not be very audible, such a difference in level may be audible across an octave or two.
Following my reasoning, a higher x-over point should allso reduce the general disperison of the loudspeaker and thus reduce room-reflection problems.
A while ago, I had some large voight horns with coral Flat8 8" fullranges and Coral H-40 horn tweeters crossed over at about 7k. This was before I buildt my microphone and discovered holm-impulse.
What I can say, is that the frequency response was far from linear.
What I can say was that stereo imaging and clarity was surprisingly good!
All in all, those speakers sounded far better than they had any right to do, but I suspect this was due to the high midrange and low treble beaming of the 8" fullrange and the very narrow dispersion pattern (probably arround 40 deg.) of the horn.
How could this be good?
Well, less early reflections and more direct sound, that's the only explanation I can think of!
allso, having one driver covering the important time-critical midrange area could have played an important part.
As I write this, I really regret having let go of those drivers and boxes before I discovered DIY measurement, it would have been tremendously interresting to measure what those speakers were actually up to, not to mention tweaking them based on actual measurements!
Regarding baffle step, I went all the way and added a full 6dB, and spliced nearfield measurements confirmed that that was (urprisingly) spot on both in terms of chosen frequencies and level.
Allthough this doesn't show up in the gated measurements I keep posting, my system measures flat from 25 Hz up to about 1Khz to the extent I have been able to piece toghether nearfield and large-room farfield measurements.
Room response, that's a tough one.. in principle, I tend to agree with John Krutke here with reference to his last blog-post.. Zaph|Audio
i.e. I agree that it makes sense to go for flat speaker response to get the correct direct radiated sound in to my ears.
The problem occurs when the room interacts in such a way as to blur the difference between direct and reflected soundfield, i.e. early reflections, and that takes me back to my "hang-up" on dispersion control.
Dr Earl Geddes seems to have very firm opinions on this, many of which I find it hard to disagree with.
Perhaps I should really go back to 8" midwoofers and horn-loaded tweeters crossed over high?
No, seriously, I'd very much like to make the best use of what I have right now.
The interresting thing is that I now have two speakers and two amp-modules with x-over cards up and running. Thgis means I can actually modify one set and leave the other as is for proper A/B testing.
Ah, and for vertical response, here's the plots for 15, 30, and 45 deg up and down.
As you predicted, some deep nulls from the drivers going in and out of phase, but apart from that, I can't really see anything hair-rising stuff going on?
Now, I have to point out that the chriticism I have at the moment is not severe, the system does sound good, but I know there is more to be had.
My reference is the first track on the ELO out of the blue album, "turn to stone".
This track has an arpeggiating synth track running which is very hard to resolve in the overall very complex mixdown this track constitutes.
At the very beginning of my project, I had a focal "Access 6A" 6" midrange ( from ZALYTRON) in the same speaker boxes crossed over really high (allmost 4k) passively.
Other flaws and imperfections aside (this was really an improvised set-up, including the x-over), the said synth was beautifully resolved and presented.
This is what I would love to acheive with the current drivers. Hopefully this can be done by getting x-over and EQ right..
As you predicted, some deep nulls from the drivers going in and out of phase, but apart from that, I can't really see anything hair-rising stuff going on?
Now, I have to point out that the chriticism I have at the moment is not severe, the system does sound good, but I know there is more to be had.
My reference is the first track on the ELO out of the blue album, "turn to stone".
This track has an arpeggiating synth track running which is very hard to resolve in the overall very complex mixdown this track constitutes.
At the very beginning of my project, I had a focal "Access 6A" 6" midrange ( from ZALYTRON) in the same speaker boxes crossed over really high (allmost 4k) passively.
Other flaws and imperfections aside (this was really an improvised set-up, including the x-over), the said synth was beautifully resolved and presented.
This is what I would love to acheive with the current drivers. Hopefully this can be done by getting x-over and EQ right..
Attachments
Vertical off axis plots look fine. The entire speaker seems well put together, although as you've mentioned some fine tuning will most likely be necessary.
I tend to sit in both camps on either side of the directivity fence and appreciate both view points. I have to say that I also have used an 8" up higher then it should be used a 3.5k xover in this case, obviously there was a dip in the power response, but the loudspeaker sounded fantastic with a wonderfully spacious sound and great imaging (on axis was flat and phase aligned). I find it extremely easy to support Geddes point of view of minimising room reflections. I mainly listen to headphones and from my point of view, a loudspeaker will never reach the fidelity of a good pair of cans and those have zero room reflections to contend with.
The point of the waveguide on the DXT, at least as I have understood it, is to stop the off axis response from drooping as frequency increases. I also understand that the waveguide is supposed to direct the off axis energy into a cone like behaviour, where all energy within the cone, irrespective of frequency, is kept roughly the same, but then as you go further off axis the response starts to droop. Naturally you'd prefer all the energy to be in cone and there to be nothing outside it, but again, as shown by Geddes this isn't how the world works.
As Geddes has stated, all loudspeakers tend to omni radiation at low frequencies and the degree to which directivity can be controlled seems to basically be a function of how large your wave-guide is. This can clearly be seen if you compare the plots for the 15, 12 and 8" versions of Geddes speakers, the amount of constant directivity available decreasing as diameter goes down.
Naturally you want there to be a smooth transition from the bass driver to the tweeter, rather then an abrupt change from omni to CD, this is where matching the directivity of the bass driver to the tweeter comes in.
The DXT will start to show constant directivity by around 3.5k, but below this it's tending to omni radiation. Using Geddes plots as a reference it would seem that the basic smooth transition from CD to omni takes about an octave. So by 1700hz the transition should have pretty much occurred entirely within the DXT itself (I am assuming that a wave-guide tweeter is capable of doing this if it's usable frequency response extends down far enough). If this is true, it would stand to reason that if any mid/bass you use is an omni radiator up until 1700hz, then you're going to have a good directivity match anyway as the omni to CD transition occurs entirely within the tweeters passband.
If all of that is correct then it would perhaps stand to reason that maybe the best thing for you to do is to actually lower the xover point between the DXT and the W17? That way you're assuring that more of the CD to omni transition is occurring within the DXT itself?
I would have thought that to match the directivity of the bass driver to the wave-guide would require the correct diameter of bass driver for the wave-guide in question. This seems to require a bass driver of similar diameter to the wave-guide, which would seem to make logical sense - a loudspeaker starts to beam as a function of it's diameter and a wave-guide can effectively channel energy as a function of it's diameter too, so it would stand to reason that similar diameters would match well in the middle. This would of course require a crossover point in exactly the right place too.
How far down you can usefully take a wave-guide loaded tweeter is again a function of the wave-guide and in normal circumstances, a very large wave-guide might lower the possible xover frequency from 2000hz to 1000hz. In this scenario, providing the tweeter was EQ'd flat, would you need extension down to say 400hz for the CD to onmi region to take place entirely within the tweeter? In this case you can't cross the tweeter at 400hz as the tweeter will explode, so naturally you need a woofer of an appropriate diameter so that a 1000hz crossover would match the directivity of the two drivers together.
With the DXT however, the wave-guide is very small, the extra low end extension it provides is minimal. As the wave-guide tends to omni at around 3500hz, it should be entirely over by 1500hz and usefully the tweeter can be used down this low too, so here you can mate it to any mid/bass providing the mid/bass is omni up until 1500hz too?
I have to admit I've kinda winged all of that as I haven't looked into wave-guides in too much detail before so I have no idea if what I've said about the DXT would actually happen in real life.
Once I've got my DSP sorted I'll probably take a whole series of off axis measurements of the DXT and try and construct one of those plots that Geddes makes. It should be a breeze to EQ the DXT flat with the DSP so normalising the response shouldn't be too hard either.
Edit - Now that I've thought about it, the widening of the dispersion from the DTX @ ~3.5khz could very well be why the 3200hz looks remarkably flat off axis, the increased output from the DXT helping to fill in the droop from the W17.
I tend to sit in both camps on either side of the directivity fence and appreciate both view points. I have to say that I also have used an 8" up higher then it should be used a 3.5k xover in this case, obviously there was a dip in the power response, but the loudspeaker sounded fantastic with a wonderfully spacious sound and great imaging (on axis was flat and phase aligned). I find it extremely easy to support Geddes point of view of minimising room reflections. I mainly listen to headphones and from my point of view, a loudspeaker will never reach the fidelity of a good pair of cans and those have zero room reflections to contend with.
The point of the waveguide on the DXT, at least as I have understood it, is to stop the off axis response from drooping as frequency increases. I also understand that the waveguide is supposed to direct the off axis energy into a cone like behaviour, where all energy within the cone, irrespective of frequency, is kept roughly the same, but then as you go further off axis the response starts to droop. Naturally you'd prefer all the energy to be in cone and there to be nothing outside it, but again, as shown by Geddes this isn't how the world works.
As Geddes has stated, all loudspeakers tend to omni radiation at low frequencies and the degree to which directivity can be controlled seems to basically be a function of how large your wave-guide is. This can clearly be seen if you compare the plots for the 15, 12 and 8" versions of Geddes speakers, the amount of constant directivity available decreasing as diameter goes down.
Naturally you want there to be a smooth transition from the bass driver to the tweeter, rather then an abrupt change from omni to CD, this is where matching the directivity of the bass driver to the tweeter comes in.
The DXT will start to show constant directivity by around 3.5k, but below this it's tending to omni radiation. Using Geddes plots as a reference it would seem that the basic smooth transition from CD to omni takes about an octave. So by 1700hz the transition should have pretty much occurred entirely within the DXT itself (I am assuming that a wave-guide tweeter is capable of doing this if it's usable frequency response extends down far enough). If this is true, it would stand to reason that if any mid/bass you use is an omni radiator up until 1700hz, then you're going to have a good directivity match anyway as the omni to CD transition occurs entirely within the tweeters passband.
If all of that is correct then it would perhaps stand to reason that maybe the best thing for you to do is to actually lower the xover point between the DXT and the W17? That way you're assuring that more of the CD to omni transition is occurring within the DXT itself?
I would have thought that to match the directivity of the bass driver to the wave-guide would require the correct diameter of bass driver for the wave-guide in question. This seems to require a bass driver of similar diameter to the wave-guide, which would seem to make logical sense - a loudspeaker starts to beam as a function of it's diameter and a wave-guide can effectively channel energy as a function of it's diameter too, so it would stand to reason that similar diameters would match well in the middle. This would of course require a crossover point in exactly the right place too.
How far down you can usefully take a wave-guide loaded tweeter is again a function of the wave-guide and in normal circumstances, a very large wave-guide might lower the possible xover frequency from 2000hz to 1000hz. In this scenario, providing the tweeter was EQ'd flat, would you need extension down to say 400hz for the CD to onmi region to take place entirely within the tweeter? In this case you can't cross the tweeter at 400hz as the tweeter will explode, so naturally you need a woofer of an appropriate diameter so that a 1000hz crossover would match the directivity of the two drivers together.
With the DXT however, the wave-guide is very small, the extra low end extension it provides is minimal. As the wave-guide tends to omni at around 3500hz, it should be entirely over by 1500hz and usefully the tweeter can be used down this low too, so here you can mate it to any mid/bass providing the mid/bass is omni up until 1500hz too?
I have to admit I've kinda winged all of that as I haven't looked into wave-guides in too much detail before so I have no idea if what I've said about the DXT would actually happen in real life.
Once I've got my DSP sorted I'll probably take a whole series of off axis measurements of the DXT and try and construct one of those plots that Geddes makes. It should be a breeze to EQ the DXT flat with the DSP so normalising the response shouldn't be too hard either.
Edit - Now that I've thought about it, the widening of the dispersion from the DTX @ ~3.5khz could very well be why the 3200hz looks remarkably flat off axis, the increased output from the DXT helping to fill in the droop from the W17.
Last edited:
5th,
Your reasoning around moving the X-over further down to match directivity makes sense.
What makes me sceptical about tis approach is that it places the x-over point even further bang in the middle of the sensitive midrange area, and that the result would be a wider disperison in the midrange area than with a higher x-over point. Allso, Tweeter distortion and power handling is bound to start suffering at some point, especially as my system is intended to be able to play LOUD when required.
Looking at the off-axis plots for the high-and low x-over point, it seems to me that the higher x-over point has a potential for better "constant directivity" with somewhat less dispersion in the upper midrange area.
Admittedly a far cry from what Dr Geddes acheives with his larger designs, but for making the most of what I've got, it looks worthwhile to investigate this further.
Too bad I have to pack the suitcase for the week as I'd love to do some more measurements and experimentation and A/B listening tests right now!
You are lucky to have a DSP unit, that must make iterative experimentation such a breeze! Admittedly, LspCad was a huge step forward for me, but there is still the limitation of having to replace components for testing out iterations in real life.
Your experience with a 8" woofer crossed high sounds interresting! Besides the dip, did you make any proper off-axis measurements that could be correlated to your listening experience?
Matching waveguide diameter with midwoofer diameter sounds like a good idea.
Having said that, The experiments of John Krutke apparently showed that with the larger 8 andf 10" wave-guides, there were some issues in the highest octaves, so there is perhaps some practical limitations to how big one can go..
Your reasoning around moving the X-over further down to match directivity makes sense.
What makes me sceptical about tis approach is that it places the x-over point even further bang in the middle of the sensitive midrange area, and that the result would be a wider disperison in the midrange area than with a higher x-over point. Allso, Tweeter distortion and power handling is bound to start suffering at some point, especially as my system is intended to be able to play LOUD when required.
Looking at the off-axis plots for the high-and low x-over point, it seems to me that the higher x-over point has a potential for better "constant directivity" with somewhat less dispersion in the upper midrange area.
Admittedly a far cry from what Dr Geddes acheives with his larger designs, but for making the most of what I've got, it looks worthwhile to investigate this further.
Too bad I have to pack the suitcase for the week as I'd love to do some more measurements and experimentation and A/B listening tests right now!
You are lucky to have a DSP unit, that must make iterative experimentation such a breeze! Admittedly, LspCad was a huge step forward for me, but there is still the limitation of having to replace components for testing out iterations in real life.
Your experience with a 8" woofer crossed high sounds interresting! Besides the dip, did you make any proper off-axis measurements that could be correlated to your listening experience?
Matching waveguide diameter with midwoofer diameter sounds like a good idea.
Having said that, The experiments of John Krutke apparently showed that with the larger 8 andf 10" wave-guides, there were some issues in the highest octaves, so there is perhaps some practical limitations to how big one can go..
As far as I remember Geddes has said that the last octave often gets compromised with large wave-guides, but in reality this is only of minor importance. Maybe the last octave will be somewhat ragged, but as long as it's present in some shape or form then it isn't too much of an issue. I do know that for me, rolling off the high frequencies above 10khz is a nono as it sounds far too dull. Maybe you could cross a small wave-guide loaded tweeter to a large one but I think that might be going over board!
My DSP isn't finished yet! and yes it will be fantastic not to have to alter resistor/cap values once it's done.
I think people worry about this far too much. There are a lot of poorly designed loudspeakers out there and if you try to cross over where the ear is most sensitive and make a mess of it, it's sure to sound bad. If however the drivers let you cross over far higher up, say at 5khz, even if you make a mess of it, it's unlikely to sound as bad as the mess at 2khz. Make a good job of it though and you're not going to have too many problems, in fact going high up in frequency makes some problems worse, so it's a toss up between what pros and cons you want to live with.
I think there's some confusion going on here, the mid/bass appears to have decent off axis curves till a bit below 2khz, any higher then this and the dispersion from the midrange decreases as it really starts to beam. Geddes show's this a number of times in this PDF, but look at figure 12. The typical wide/narrow/wide dispersion characteristic is due to crossing the tweeter to the mid/bass too high up. The dispersion narrows due to the mid/bass starting to beam, but then you cross over to an omni source, the tweeter, so the dispersion widens again (before narrowing due to the tweeter starting to beam). This seems like something you would want to avoid and is avoided in Geddes designs.
As the DXT has wide dispersion at 3khz itself, if crossed here you could end up with wide/narrow/wide>CD (how narrow depends on how much the mid/bass beams). I'd have thought wide>CD would be preferable.
At the end of the day though, you can't beat hearing the system though, so it'd probably be worth give both a go.
You'd have to test this to see if tweeter distortion would ever become a problem. Mark K used them as low as 1.5khz with a 2nd order filter and did have measurements of system distortion at high drive levels, although I can't find the measurements now, but it wasn't an issue.
Both have exactly the same potential for constant directivity as any CD you're getting is completely due to wave-guide tweeter above 3-4khz. The transition to CD and what happens before this is up to the crossover and mid/bass.
My DSP isn't finished yet! and yes it will be fantastic not to have to alter resistor/cap values once it's done.
I think people worry about this far too much. There are a lot of poorly designed loudspeakers out there and if you try to cross over where the ear is most sensitive and make a mess of it, it's sure to sound bad. If however the drivers let you cross over far higher up, say at 5khz, even if you make a mess of it, it's unlikely to sound as bad as the mess at 2khz. Make a good job of it though and you're not going to have too many problems, in fact going high up in frequency makes some problems worse, so it's a toss up between what pros and cons you want to live with.
I think there's some confusion going on here, the mid/bass appears to have decent off axis curves till a bit below 2khz, any higher then this and the dispersion from the midrange decreases as it really starts to beam. Geddes show's this a number of times in this PDF, but look at figure 12. The typical wide/narrow/wide dispersion characteristic is due to crossing the tweeter to the mid/bass too high up. The dispersion narrows due to the mid/bass starting to beam, but then you cross over to an omni source, the tweeter, so the dispersion widens again (before narrowing due to the tweeter starting to beam). This seems like something you would want to avoid and is avoided in Geddes designs.
As the DXT has wide dispersion at 3khz itself, if crossed here you could end up with wide/narrow/wide>CD (how narrow depends on how much the mid/bass beams). I'd have thought wide>CD would be preferable.
At the end of the day though, you can't beat hearing the system though, so it'd probably be worth give both a go.
You'd have to test this to see if tweeter distortion would ever become a problem. Mark K used them as low as 1.5khz with a 2nd order filter and did have measurements of system distortion at high drive levels, although I can't find the measurements now, but it wasn't an issue.
Both have exactly the same potential for constant directivity as any CD you're getting is completely due to wave-guide tweeter above 3-4khz. The transition to CD and what happens before this is up to the crossover and mid/bass.
It is a good point that for the highest octave, the most important thing is that something "is there", this is not the most sensitive area I agree!
I allso agree that adding a second horn/ wave-guided tweeter to cover the top octave is going a bit over the top.
Regarding the midrange area, this is a complicated one, and I do agree that part of the focus this hhas received has come from poor-x overs. As you say making a hash of it in this area is doomed to give bad results! But even assuming a good x-over implementation, I'm a bit sceptical..
What worries me, is the fact that the ear is very time sensitive in this area, so if you have part of the midrange emeanating from the tweeter and the midrange, the ear can detect the difference in position. and due to the ears sensitivity in this area, this can muddle up spatial perseption and immaging. This is just drawing assumptions from common psycho-acoustic research. If one moves up and above 2-3 kHz, the ear transits from time, or phase, sensitivity to level sensitivity. Having two simmilar signals emeanating from two sources placed close to each others will be less distinguishable as their position will be difficult to discriminate from level alone.
I agree that the midbass off-axis looks good as such except for the dip.
But the dip aside, with the higher x-over, the 30 degree response is consistently lower than on axis, and that is something I find interresting.
As I read the plots, the response for the lower x-over seems closer to figure nr 12 than the hight x-over response, whilst the high-x over more approximates that of figure 18., i.e. It seems I got more of a wide-narrow- wide issue now than with the high x-over. But had I been using a conventional tweeter, I think it is likely that the result would have been a pronounced wide-narrow-wide response as a "non-waveguided" tweeter would probably have wider dispersion at 3 khz than the DXT tweeter.
It will certainly be interresting to modify and compare!
It was mentioned earlier on in the thread that this tweeter should be good down to 1500 Hz, but will that still apply at 105 dB?
That's another test I can do, simply measure the tweter at low and not so lo level and se what happens. What is certain is that IF the tweeter starts to overload and distort in THIS frequency range, it will not be a particularly nice thing..
That 8" crossed high you mentioned.. what sort of woofer was it?
I allso agree that adding a second horn/ wave-guided tweeter to cover the top octave is going a bit over the top.
Regarding the midrange area, this is a complicated one, and I do agree that part of the focus this hhas received has come from poor-x overs. As you say making a hash of it in this area is doomed to give bad results! But even assuming a good x-over implementation, I'm a bit sceptical..
What worries me, is the fact that the ear is very time sensitive in this area, so if you have part of the midrange emeanating from the tweeter and the midrange, the ear can detect the difference in position. and due to the ears sensitivity in this area, this can muddle up spatial perseption and immaging. This is just drawing assumptions from common psycho-acoustic research. If one moves up and above 2-3 kHz, the ear transits from time, or phase, sensitivity to level sensitivity. Having two simmilar signals emeanating from two sources placed close to each others will be less distinguishable as their position will be difficult to discriminate from level alone.
I agree that the midbass off-axis looks good as such except for the dip.
But the dip aside, with the higher x-over, the 30 degree response is consistently lower than on axis, and that is something I find interresting.
As I read the plots, the response for the lower x-over seems closer to figure nr 12 than the hight x-over response, whilst the high-x over more approximates that of figure 18., i.e. It seems I got more of a wide-narrow- wide issue now than with the high x-over. But had I been using a conventional tweeter, I think it is likely that the result would have been a pronounced wide-narrow-wide response as a "non-waveguided" tweeter would probably have wider dispersion at 3 khz than the DXT tweeter.
It will certainly be interresting to modify and compare!
It was mentioned earlier on in the thread that this tweeter should be good down to 1500 Hz, but will that still apply at 105 dB?
That's another test I can do, simply measure the tweter at low and not so lo level and se what happens. What is certain is that IF the tweeter starts to overload and distort in THIS frequency range, it will not be a particularly nice thing..
That 8" crossed high you mentioned.. what sort of woofer was it?
A few years ago I experimented somewhat with different xover frequencies between mid and tweeter. I used 4th order LR acoustic slopes that were phase aligned with delay stages and tried between 2-4khz. Neither tweeter or midrange had any issues over that range with regards to distortion and C-C spacing was kept as low as possible. The only real issue was the midrange beaming a bit by 4khz. Either way having done that experiment I am less concerned with crossing over around 2khz then I was before. There were differences in the sound but nothing changed that much, maybe I listen for the wrong things, I like a strong central image with a large spacious sound stage, I don't care so much about being able to locate the specific instruments as long as I am enveloped in sound. All xovers provided those in spades and I ended up settling on 3khz.
The 8" Amidst looking through Geddes' website yesterday I totally forgot about that, it was a vifa coated paper cone, a C20WJ-19-08. Very well controlled upper breakup, I've still got the drivers on the shelf, close at hand, but if I was going to recreate the loudspeaker I'd probably go with something like this, those drivers are completely dead off axis @60 degrees by 3.5khz, they sounded great though, this was some 8 years ago so my standards have probably gone up
You know it's a real shame that the extra distortion measurements that Mark put on his blog have disappeared. I know he hammered the loudspeaker to make a point out of it, I am tempted to say it probably can though. I mean a 4th order at 1500hz has a -6dB requirement over the standard flat frequency response. The DXT gives out 93dB @ 1500hz for 2.83volts. 105dB minus 6dB arrives at a 1500hz requirement of 99dB on the tweeter, this is only 6dB above it's 2.83volt sensitivity testing or about 5.5 watts into a 6 ohm load. That's only a very rough way of looking at it though so you'd have to measure to be sure. Were you talking 105dB at one meter or 105dB at the listening position for a stereo pair?
Figure 12 shows a standard 2 way xovered at 2000hz, with a mid/bass showing off axis droop due to beaming. The tweeter is also non wave guide otherwise the black lines representing 6dB differences in output level would be more or less parallel.
Figure 18 represents Geddes flagship wave-guide design that exhibits CD down to around 1khz, where the directivity then slowly decreases as the woofer takes over, I am guessing here that the xover is at around 1khz too if I am reading things correctly.
With the DXT, you should have controlled directivity from ~3.5khz onwards.
If you are drawing comparisons to those two figures then I'd say figure 18 would represent the 1700hz xover. Here you should have CD from 3.5khz+ and below that a smooth transition to omni. This is almost identical to figure 18, except that the CD to omni transition occurs at 1khz instead of 3.5khz.
The 3200hz xover should show signs of beaming from the mid/bass and thus a narrowing of dispersion up until the tweeter takes over. This would show a similarity to figure 12, where the mid/bass to tweeter xover does exactly the same thing. Except here the similarities end as the DXT should show CD from 3.5khz+.
Yes that is interesting. I wonder if SEAS separated the off axis response by a dB or so to make it easier to look at. Either that or something caused a difference in the level the mic picked up, I mean the off axis curves haven't even converged by 100hz which is quite clearly not caused by the usual beaming. I guess if the driver was mounted 0.5 meters away from the mic, as it says and placed in a small 12 litre box. Now if you turned the loudspeaker, instead of moving the mic, to create the offaxis curves, then the driver would get slightly further away from the mic as the cabinet turned.
If you look at scanspeak graphs for a similar sized driver you see pretty much what you'd expect, an identical on and off axis response up until the point where the driver starts to beam and reasonably good 60 degree off axis curves up until around 1800hz too.
The 8"
Amidst looking through Geddes' website yesterday I totally forgot about that, it was a vifa coated paper cone, a C20WJ-19-08. Very well controlled upper breakup, I've still got the drivers on the shelf, close at hand, but if I was going to recreate the loudspeaker I'd probably go with something like this, those drivers are completely dead off axis @60 degrees by 3.5khz, they sounded great though, this was some 8 years ago so my standards have probably gone upYou know it's a real shame that the extra distortion measurements that Mark put on his blog have disappeared. I know he hammered the loudspeaker to make a point out of it, I am tempted to say it probably can though. I mean a 4th order at 1500hz has a -6dB requirement over the standard flat frequency response. The DXT gives out 93dB @ 1500hz for 2.83volts. 105dB minus 6dB arrives at a 1500hz requirement of 99dB on the tweeter, this is only 6dB above it's 2.83volt sensitivity testing or about 5.5 watts into a 6 ohm load. That's only a very rough way of looking at it though so you'd have to measure to be sure. Were you talking 105dB at one meter or 105dB at the listening position for a stereo pair?
Figure 12 shows a standard 2 way xovered at 2000hz, with a mid/bass showing off axis droop due to beaming. The tweeter is also non wave guide otherwise the black lines representing 6dB differences in output level would be more or less parallel.
Figure 18 represents Geddes flagship wave-guide design that exhibits CD down to around 1khz, where the directivity then slowly decreases as the woofer takes over, I am guessing here that the xover is at around 1khz too if I am reading things correctly.
With the DXT, you should have controlled directivity from ~3.5khz onwards.
If you are drawing comparisons to those two figures then I'd say figure 18 would represent the 1700hz xover. Here you should have CD from 3.5khz+ and below that a smooth transition to omni. This is almost identical to figure 18, except that the CD to omni transition occurs at 1khz instead of 3.5khz.
The 3200hz xover should show signs of beaming from the mid/bass and thus a narrowing of dispersion up until the tweeter takes over. This would show a similarity to figure 12, where the mid/bass to tweeter xover does exactly the same thing. Except here the similarities end as the DXT should show CD from 3.5khz+.
Yes that is interesting. I wonder if SEAS separated the off axis response by a dB or so to make it easier to look at. Either that or something caused a difference in the level the mic picked up, I mean the off axis curves haven't even converged by 100hz which is quite clearly not caused by the usual beaming. I guess if the driver was mounted 0.5 meters away from the mic, as it says and placed in a small 12 litre box. Now if you turned the loudspeaker, instead of moving the mic, to create the offaxis curves, then the driver would get slightly further away from the mic as the cabinet turned.
If you look at scanspeak graphs for a similar sized driver you see pretty much what you'd expect, an identical on and off axis response up until the point where the driver starts to beam and reasonably good 60 degree off axis curves up until around 1800hz too.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.