Progress Report:
All the measurements have been done other than the outdoor acoustic farfield measurements. These I hope will take place sometime next week. All the drivers were run in for a minimum of twenty hours each.
I am very happy with the results so far.
First the in-box impedance plot of the combined Purifi drivers (series-parallel):
Here we can observe a few things, first, these are all the four drivers measured in situ, in the box. With the port lengthened from 100mm to 150mm, using 90mm PVC pipe (87mm ID) and the tuning I was looking for is bang on. You can see that it is the target 30Hz and got there with -/+ 0.5Hz. Interestingly, and this seems hardly ever mentioned, the vented box tuning is independent of the drivers so that if other drivers were used, the frequency of the bottom of the 'saddle' would still be 30Hz. But of course, the rest of the impedance plot would look very different. With sealed boxes, the single peak is a combination of driver and box volume, but a Helmholz resonator frequency is independent of the driver.
Also, the ratio of the peaks looks right. Experience indicates this will be a well-damped alignment. Other modelling was used, not shown here. On paper, the voltage sensitivity will likely we 1.5dB lower than before, but that is yet to be settled.
As we can see, from 200Hz to 1KHz the impedance plot looks very smooth, this indicates the internal box resonances are well behaved. There is the slightest issue at around 4KHz, but that is not the box, but the driver. It also shows up in Purifi's own free-air impedance plot below:
The single LF peak shows this is nominally in free air measurement and a single circa 39Hz peak. You can see that same slight issue at 4KHz and I consider that not a bad thing, that two independent measurements show the same thing. I am thinking this is a reflection of the rear basket and/or the spider?
Next, the tweeter impedance plot:
Let us compare that with Scan-Speak's published measurement:
Note that we don't have the electrical phase graphed. But again we see a good correlation between two independent measurements. The use of the waveguide seems to have a minimal effect here, and that is a good thing that there are no aberrations from the waveguide showing up.
Now for the nearfield acoustic measurement:
Important: Discard response above 1KHz.
Looks very nice indeed!
This is not the normal frequency response of the driver. This is a nearfield measurement, the microphone is within 10mm of the cone, about halfway between the centre of the dust cap and the surround. This is designed to capture mainly the piston range of the driver and the data captured should be regarded as valid, at best, up to 1KHz.
The measurement you are looking at above is the sum of two measurements, that of the driver right on the cone, and the other is that of the port where the microphone is placed at the mouth of the port.
When nearfield and port responses are added together, this is known as summing the two responses into one. But still not the complete in-box response.
When the farfield measurement is available, this will be merged with the nearfield, to give us a final frequency response.
The final frequency response measurement of the Purifi driver in situ is actually THREE combined measurements. These are carefully put together and checked, double-checked, and then double-checked again.
The farfield measurement yet to come will be done outdoors and not at the usual one metre, but at two metres. I want to capture the relative phase of the drivers, relative to two metres at the height of the tweeter. That means our crossover will be modelled at something closer to the actual listening position. If it works at two metres, it will hang together nicely at realistic listening distances.
The above shows the microphone at one metre, but think of it as two metres. There will be three measurements for the On-axis, then three for the 15 degrees Off-axis, and then finally three for the 30 degrees Off-axis. In total, nine measurements.
Of these nine measurements, the first will be the tweeter On-axis. This needs to be exactly two metres. Then when the other two Off-axis measurements are done, the impulse of the first tweeter measurement will serve as a reference for the exact position of the microphone within mm accuracy. This ensures that all nine measurements are all done at the same distance, no matter what angle.
We can now model the crossover and see exactly what is going on across a 60 degrees arc in front of the speaker. Yes, we will be running three copies of SoundEasy software whilst modelling the crossover, across a wide-screen monitor.
There is something else that I do a little bit different: At two metres I use 4V (2W) rather than 2.83V (1W), and this allows me to reference at one metre (calibrated), whilst actually doing the measurement at two metres. I use a sample Vifa XT25 tweeter that has a known 91dB voltage sensitivity, to calibrate the setup.
I use an Earthworks M30 Measurement Microphone with a special custom data correction file, effectively flat response to 50KHz.
I am not sure if the above info will be appreciated as it seems to be the kind of thing that nerds are interested in. Most of you rightly just want the design information in the end and just go ahead and build them. And that is understandable.
But I also suspect that some of you will appreciate a little background on the design and that it isn't just something that is quickly cobbled together. My late friend Walter Bæk-Hansen used to refer to a lot of speaker designers (and their companies) as "Box Stuffers" - what he meant is that they would buy expensive Scan-Speak drivers and stuff them into boxes with very few if any original design ideas; just make the box look good and somebody will buy them. At least you can take it from me, this is not the case here. We are not using ultra-expensive Purifi drivers just for the sake of it, or just to look impressive. It has to be a seriously thought out design. How much better than the standard Elsinore Mk-6 remains to be seen. But we are getting there soon to find out and that will be exciting.
Cheers, Joe
All the measurements have been done other than the outdoor acoustic farfield measurements. These I hope will take place sometime next week. All the drivers were run in for a minimum of twenty hours each.
I am very happy with the results so far.
First the in-box impedance plot of the combined Purifi drivers (series-parallel):
Here we can observe a few things, first, these are all the four drivers measured in situ, in the box. With the port lengthened from 100mm to 150mm, using 90mm PVC pipe (87mm ID) and the tuning I was looking for is bang on. You can see that it is the target 30Hz and got there with -/+ 0.5Hz. Interestingly, and this seems hardly ever mentioned, the vented box tuning is independent of the drivers so that if other drivers were used, the frequency of the bottom of the 'saddle' would still be 30Hz. But of course, the rest of the impedance plot would look very different. With sealed boxes, the single peak is a combination of driver and box volume, but a Helmholz resonator frequency is independent of the driver.
Also, the ratio of the peaks looks right. Experience indicates this will be a well-damped alignment. Other modelling was used, not shown here. On paper, the voltage sensitivity will likely we 1.5dB lower than before, but that is yet to be settled.
As we can see, from 200Hz to 1KHz the impedance plot looks very smooth, this indicates the internal box resonances are well behaved. There is the slightest issue at around 4KHz, but that is not the box, but the driver. It also shows up in Purifi's own free-air impedance plot below:
The single LF peak shows this is nominally in free air measurement and a single circa 39Hz peak. You can see that same slight issue at 4KHz and I consider that not a bad thing, that two independent measurements show the same thing. I am thinking this is a reflection of the rear basket and/or the spider?
Next, the tweeter impedance plot:
Let us compare that with Scan-Speak's published measurement:
Note that we don't have the electrical phase graphed. But again we see a good correlation between two independent measurements. The use of the waveguide seems to have a minimal effect here, and that is a good thing that there are no aberrations from the waveguide showing up.
Now for the nearfield acoustic measurement:
Important: Discard response above 1KHz.
Looks very nice indeed!
This is not the normal frequency response of the driver. This is a nearfield measurement, the microphone is within 10mm of the cone, about halfway between the centre of the dust cap and the surround. This is designed to capture mainly the piston range of the driver and the data captured should be regarded as valid, at best, up to 1KHz.
The measurement you are looking at above is the sum of two measurements, that of the driver right on the cone, and the other is that of the port where the microphone is placed at the mouth of the port.
When nearfield and port responses are added together, this is known as summing the two responses into one. But still not the complete in-box response.
When the farfield measurement is available, this will be merged with the nearfield, to give us a final frequency response.
The final frequency response measurement of the Purifi driver in situ is actually THREE combined measurements. These are carefully put together and checked, double-checked, and then double-checked again.
The farfield measurement yet to come will be done outdoors and not at the usual one metre, but at two metres. I want to capture the relative phase of the drivers, relative to two metres at the height of the tweeter. That means our crossover will be modelled at something closer to the actual listening position. If it works at two metres, it will hang together nicely at realistic listening distances.
The above shows the microphone at one metre, but think of it as two metres. There will be three measurements for the On-axis, then three for the 15 degrees Off-axis, and then finally three for the 30 degrees Off-axis. In total, nine measurements.
Of these nine measurements, the first will be the tweeter On-axis. This needs to be exactly two metres. Then when the other two Off-axis measurements are done, the impulse of the first tweeter measurement will serve as a reference for the exact position of the microphone within mm accuracy. This ensures that all nine measurements are all done at the same distance, no matter what angle.
We can now model the crossover and see exactly what is going on across a 60 degrees arc in front of the speaker. Yes, we will be running three copies of SoundEasy software whilst modelling the crossover, across a wide-screen monitor.
There is something else that I do a little bit different: At two metres I use 4V (2W) rather than 2.83V (1W), and this allows me to reference at one metre (calibrated), whilst actually doing the measurement at two metres. I use a sample Vifa XT25 tweeter that has a known 91dB voltage sensitivity, to calibrate the setup.
I use an Earthworks M30 Measurement Microphone with a special custom data correction file, effectively flat response to 50KHz.
I am not sure if the above info will be appreciated as it seems to be the kind of thing that nerds are interested in. Most of you rightly just want the design information in the end and just go ahead and build them. And that is understandable.
But I also suspect that some of you will appreciate a little background on the design and that it isn't just something that is quickly cobbled together. My late friend Walter Bæk-Hansen used to refer to a lot of speaker designers (and their companies) as "Box Stuffers" - what he meant is that they would buy expensive Scan-Speak drivers and stuff them into boxes with very few if any original design ideas; just make the box look good and somebody will buy them. At least you can take it from me, this is not the case here. We are not using ultra-expensive Purifi drivers just for the sake of it, or just to look impressive. It has to be a seriously thought out design. How much better than the standard Elsinore Mk-6 remains to be seen. But we are getting there soon to find out and that will be exciting.
Cheers, Joe
Ah yes, but you can't please everybody. I am an old dog, and I like old school. Besides, putting a speaker of this size on a turntable? That is not going to happen. But the next stage will further make it clear that there are plusses to the way I am doing things and I hope you will find it interesting. I know exactly what data I want and I do something that is a bit of a surprise in the way the crosser works. I have a priority that others don't have. So this is not a stock ho-hum LR4 crossover. 
But I have been looking at VituixCad. When time presents itself, I will download it and play with it.
But I have been looking at VituixCad. When time presents itself, I will download it and play with it.
oh no, not just put the loudspeaker on a turntable, but also outside away from stuff high up in the air on a pole!
Maybe I need this?
I wonder if they make a fold-up version? Pocket size would be good.
The late John Dunleavy used a forklift truck and hoisted it up who knows how many feet in the air.
My ways are based on good, old and well-tried methods in this book:
And you can still buy it. Came out in 1998.
https://www.parts-express.com/Testing-Loudspeakers-Book-500-030
For those reading tests by John Atkinson in Stereophile, same methodologies.
I wonder if they make a fold-up version? Pocket size would be good.
The late John Dunleavy used a forklift truck and hoisted it up who knows how many feet in the air.
My ways are based on good, old and well-tried methods in this book:
And you can still buy it. Came out in 1998.
https://www.parts-express.com/Testing-Loudspeakers-Book-500-030
For those reading tests by John Atkinson in Stereophile, same methodologies.
..bah, most anechoic chambers are only good to about 80 Hz.
56 feet off the ground, that's the "good stuff".
(..err, of course you have to do so in some location practically devoid of extra sound.)
Reality check:
When I was measuring (fully) loudspeakers it was about 300 Hz up in-room with lot's of absorption more than a meter away around the speaker (and absorption under the speaker and above it). For low freq.s I'd take it outside on my long drive-way and do a ground plane measurement (..hoping I could capture the Impulse Response without to much noise).
56 feet off the ground, that's the "good stuff".
(..err, of course you have to do so in some location practically devoid of extra sound.)
Reality check:
When I was measuring (fully) loudspeakers it was about 300 Hz up in-room with lot's of absorption more than a meter away around the speaker (and absorption under the speaker and above it). For low freq.s I'd take it outside on my long drive-way and do a ground plane measurement (..hoping I could capture the Impulse Response without to much noise).
Right, 56ft is the wavelength of 20Hz and there would be few anechoic rooms that size. So we use methods that does not require it.
I hear you loud and clear. And that's the question, will the "ULD" version sound better?
But there is nothing wrong with the measurements being better and then sounding better too. I am hoping that it will.
Honestly measurement is an helper, it never said how good it sounds ?
Best for this desgin Joe
I hear you loud and clear. And that's the question, will the "ULD" version sound better?
But there is nothing wrong with the measurements being better and then sounding better too. I am hoping that it will.
Actually It doesn't matter if it sounds better, more like that it doesn't sound worse and people (who are willing to pay a lot more) believe they are getting the very best version.
-demand for those who can afford it is often relatively inelastic, though getting a good perceived *value is still a high priority.
Most only want to go through the build process once (even if they commission it out), their hobby is music listening - not loudspeaker building. Why take a chance with a lesser objective part if you've got the money? (..and those that do have the money for this often find that their time is more valuable than the cost increases.)
*group buys with drivers can help this out in the DIY context.
-demand for those who can afford it is often relatively inelastic, though getting a good perceived *value is still a high priority.
Most only want to go through the build process once (even if they commission it out), their hobby is music listening - not loudspeaker building. Why take a chance with a lesser objective part if you've got the money? (..and those that do have the money for this often find that their time is more valuable than the cost increases.)
*group buys with drivers can help this out in the DIY context.
Actually It doesn't matter if it sounds better, more like that it doesn't sound worse and people (who are willing to pay a lot more) believe they are getting the very best version.
I sorta understand where you are coming from. I am a half-full guy rather than a half-empty guy. Whatever happens, the current standard NRX and MFC versions are not going to sound less great than they already are.
The curious thing is that this might mean that more of those too will be built. There shouldn't be any losers.
I don't know which you intend, but vertical measurements would be most helpful (the drivers are stacked vertically) with the drivers playing together, to capture the crossover and driver placement issues.
Hi Allen
I hear what you are saying and will look at it later, but it will not really change the design. I already know that it will do way better than an MTM speaker with 1st order Butterworth crossovers that sum at -3dB and has a narrower window all round. So I know it will be reasonable but not as good in that respect when compared to LR4 crossovers. It's come down to priorities and I am not enamored by LR4 or even LR2.
I hear what you are saying and will look at it later, but it will not really change the design. I already know that it will do way better than an MTM speaker with 1st order Butterworth crossovers that sum at -3dB and has a narrower window all round. So I know it will be reasonable but not as good in that respect when compared to LR4 crossovers. It's come down to priorities and I am not enamored by LR4 or even LR2.
OK, been a few busy days.
Below are the modeled results and grabs from SoundEasy software. All the measurements are calibrated measurement, so SPL data is correct on the left-hand side.
1. Frequency Response
This is 15 degrees off-axis and aims to be the preferred listening angle.
2. Family of on-axis and off-axis responses:
This illustrates how toe-in will help in different rooms. Try with 15 degrees off-axis (red) and if the sound is a bit dull (or heavily damped room), some additional toe-in might help. If the sound is a bit bright, toe-out a bit.
3. Various driver contributions:
This family of responses was all taken 15 degrees off-axis. Green is the top two ULD drivers that form an MTM pair with the Tweeter in Blue. The Red is interesting, this is the bottom two ULD drivers and the dip at 4KHz is actually the two drivers canceling each other out to a degree. This 4KHz dip occurs at 2 metres, but since most will definitely be listening further back than that, the dip will increase to about 5KHz at 2.5 metres. This will very slightly boost the response around 1-2Khz and that has been accounted for.
4. Crossover and null.
Here Green is the total summed response of all four ULD drivers. The Tweeter is Blue and the final summed response, again 14 degrees off-axis, is Red. But the key one here is Purple, which is where the tweeter electrical phase has been reversed. Note that the actual crossover frequency is bang on 3KHz and that the symmetry either side is near as perfect as can be. Yes, I was very pleased when I was able to get that.
5. The system Impedance:
You won't see this is any commercial speaker system. The impedance never drops below 6 Ohm and above 25 Hertz, never goes above 8 Ohm.
6. Simulated current-drive:
I have a current source amplifier with an output impedance of 270 Ohm. I put the design on diyaudio.com, see the link at the end. Above, what you see above is Green under voltage-drive and Red is simulated current-drive. What is important is that not only does it lock in the low-frequency alignment down to below 25 Hertz, but also the crossover. One is a mechanical high-pass filter (the box) and the others are electrical filters. EQ the current of the amplifier, so that it sees the same current at all frequencies, and this will work any time, and indeed any speaker if it presents a load like this. The peak at around 17 Hertz is a good -15dB down, so if a note comes along at that frequency, then power handling is degraded, but not completely. If this kind of EQ was done with a sealed box (single peak), then the LF response will be flat.
The next step is to build the crossover and test it. Even if not 100% in the end, it will be in the ballpark area and then tweaked. The reference will be the response/balance of the standard EL-6 in-room and both occupying the same space (move one out and another one in) and test and compare with 1/6th Octave Pink noise RTA measurements. This is also what I did earlier with the Hamlet speakers.
It's late here, 1 AM. Here is the promised link:
https://www.diyaudio.com/community/...tt-transconductance-current-amplifier.239321/
Cheers, Joe
Below are the modeled results and grabs from SoundEasy software. All the measurements are calibrated measurement, so SPL data is correct on the left-hand side.
1. Frequency Response
This is 15 degrees off-axis and aims to be the preferred listening angle.
2. Family of on-axis and off-axis responses:
This illustrates how toe-in will help in different rooms. Try with 15 degrees off-axis (red) and if the sound is a bit dull (or heavily damped room), some additional toe-in might help. If the sound is a bit bright, toe-out a bit.
3. Various driver contributions:
This family of responses was all taken 15 degrees off-axis. Green is the top two ULD drivers that form an MTM pair with the Tweeter in Blue. The Red is interesting, this is the bottom two ULD drivers and the dip at 4KHz is actually the two drivers canceling each other out to a degree. This 4KHz dip occurs at 2 metres, but since most will definitely be listening further back than that, the dip will increase to about 5KHz at 2.5 metres. This will very slightly boost the response around 1-2Khz and that has been accounted for.
4. Crossover and null.
Here Green is the total summed response of all four ULD drivers. The Tweeter is Blue and the final summed response, again 14 degrees off-axis, is Red. But the key one here is Purple, which is where the tweeter electrical phase has been reversed. Note that the actual crossover frequency is bang on 3KHz and that the symmetry either side is near as perfect as can be. Yes, I was very pleased when I was able to get that.
5. The system Impedance:
You won't see this is any commercial speaker system. The impedance never drops below 6 Ohm and above 25 Hertz, never goes above 8 Ohm.
6. Simulated current-drive:
I have a current source amplifier with an output impedance of 270 Ohm. I put the design on diyaudio.com, see the link at the end. Above, what you see above is Green under voltage-drive and Red is simulated current-drive. What is important is that not only does it lock in the low-frequency alignment down to below 25 Hertz, but also the crossover. One is a mechanical high-pass filter (the box) and the others are electrical filters. EQ the current of the amplifier, so that it sees the same current at all frequencies, and this will work any time, and indeed any speaker if it presents a load like this. The peak at around 17 Hertz is a good -15dB down, so if a note comes along at that frequency, then power handling is degraded, but not completely. If this kind of EQ was done with a sealed box (single peak), then the LF response will be flat.
The next step is to build the crossover and test it. Even if not 100% in the end, it will be in the ballpark area and then tweaked. The reference will be the response/balance of the standard EL-6 in-room and both occupying the same space (move one out and another one in) and test and compare with 1/6th Octave Pink noise RTA measurements. This is also what I did earlier with the Hamlet speakers.
It's late here, 1 AM. Here is the promised link:
https://www.diyaudio.com/community/...tt-transconductance-current-amplifier.239321/
Cheers, Joe
Attachments
Last edited: