back to group delay :
I found that with some signals (sawtooth), low freq GD is quite audible : curiously, longer GD gives "more" bass. I am wondering if some of you had same findings.
You can also try it with my soft here
I had the same questions about audibility of GD in low frequencies so I did a software to change GD without changing the FR (an allpass filter).
I found that with some signals (sawtooth), low freq GD is quite audible : curiously, longer GD gives "more" bass. I am wondering if some of you had same findings.
You can also try it with my soft here
Hi jlo,
Amplitude, phase and group delay are inextricably linked.
EQing and crossover modification of amplitude response affect the other two.
The sounds upon which I found sub/LF phase change most noticeable were an outdoor recording of a distant canon firing, and the hall ambience background of recorded music. Both changed considerably with LF phase change.
There are those who claim we are incapable of hearing LF phase change, however, I just wish they would speak for themselves only !
With regard to group delay, I found this has a separate effect.
Take the kick drum. Where there is EQ boost induced group delay, the initial thud sounds as if it has less amplitude than the following membraneous vibration, and again sounds quite different, though if in phase can be reasonably tolerable.
Lynn is proposing to increase LF output by increasing driver area with reducing frequency, and if carefully done this should not develop phase and group delay induced deviation or EQ amplitude modification of either.
You mention adjusting group delay, but you must simultaneously have also been altering either the amplitude or phase response.
Is this what provides 'more' bass.
Cheers .......... Graham.
Amplitude, phase and group delay are inextricably linked.
EQing and crossover modification of amplitude response affect the other two.
The sounds upon which I found sub/LF phase change most noticeable were an outdoor recording of a distant canon firing, and the hall ambience background of recorded music. Both changed considerably with LF phase change.
There are those who claim we are incapable of hearing LF phase change, however, I just wish they would speak for themselves only !
With regard to group delay, I found this has a separate effect.
Take the kick drum. Where there is EQ boost induced group delay, the initial thud sounds as if it has less amplitude than the following membraneous vibration, and again sounds quite different, though if in phase can be reasonably tolerable.
Lynn is proposing to increase LF output by increasing driver area with reducing frequency, and if carefully done this should not develop phase and group delay induced deviation or EQ amplitude modification of either.
You mention adjusting group delay, but you must simultaneously have also been altering either the amplitude or phase response.
Is this what provides 'more' bass.
Cheers .......... Graham.
How to measure a driver for use as a dipole
Hi
JohnK, here it is and as it is only a few clicks more I include the CSD plot as well:
( As you know for sure and indicated by the yellow bar THIS measurement is only valid above 200 Hz due to the 5ms window )
What can be seen is that this driver in this baffle refuses to show the first baffle peak, instead – almost exactly - there is a 4 dB dip at 600 - 700 Hz where the 4 dB peak should be. Hope it is of any help though.
Anyway I wouldn't recommend that driver. Not because of that FR which always can be EQed out ( no resonance can be seen in the CSD plot between 300 - 1000 Hz ). This driver also has a good motor, extended FR, low moving mass and low distortion (3rd HD up) BUT one thing that bothers me the longer the more is that tupperware sound coloration of the plastic ( magnesium filled ) cone material. This may have increased since I replaced the original foam by rubber surroundings.
I am also exploring a high sensitive WESTRA paper cone chassis around 10 years older then the almost ancient Dynaudio above. This was kind of el cheapo speaker at that days but in contrast to the high priced Dynaudio after roughly 35 years they are still in perfect working condition having a flawless linen surrounding and NO broken voice coil braid either as is very common with Dynaudio chassis.
Sound is fast and transparent with a nice singing in the bass department. I really have to measure it more extensively.
A interesting detail is the acoustically transparent dust cap made of black gauze – a early version of what now is done with phase plugs.
I am curious with which speaker Lynn will come up for the mids. It is easy to find PRO speakers with paper ( maybe even hemp ) cones but it isn't easy to find PRO speakers with phase plugs avoiding the rear coughing through the pole piece ventilation hole - essential for OB's IMO.
Greetings
Michael
Hi
JohnK, here it is and as it is only a few clicks more I include the CSD plot as well:
( As you know for sure and indicated by the yellow bar THIS measurement is only valid above 200 Hz due to the 5ms window )
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
What can be seen is that this driver in this baffle refuses to show the first baffle peak, instead – almost exactly - there is a 4 dB dip at 600 - 700 Hz where the 4 dB peak should be. Hope it is of any help though.
Anyway I wouldn't recommend that driver. Not because of that FR which always can be EQed out ( no resonance can be seen in the CSD plot between 300 - 1000 Hz ). This driver also has a good motor, extended FR, low moving mass and low distortion (3rd HD up) BUT one thing that bothers me the longer the more is that tupperware sound coloration of the plastic ( magnesium filled ) cone material. This may have increased since I replaced the original foam by rubber surroundings.
I am also exploring a high sensitive WESTRA paper cone chassis around 10 years older then the almost ancient Dynaudio above. This was kind of el cheapo speaker at that days but in contrast to the high priced Dynaudio after roughly 35 years they are still in perfect working condition having a flawless linen surrounding and NO broken voice coil braid either as is very common with Dynaudio chassis.
Sound is fast and transparent with a nice singing in the bass department. I really have to measure it more extensively.
A interesting detail is the acoustically transparent dust cap made of black gauze – a early version of what now is done with phase plugs.
I am curious with which speaker Lynn will come up for the mids. It is easy to find PRO speakers with paper ( maybe even hemp ) cones but it isn't easy to find PRO speakers with phase plugs avoiding the rear coughing through the pole piece ventilation hole - essential for OB's IMO.
Greetings
Michael
Thanks guys. I have now changed my mind and ditched the WWMTMWW 4 way and LR4 at low XOs. I would follow Lynn's OB, however, my plan had been drawn up well before this thread started and I have all the drivers on hand already so I will press on. I have now opted for a MTMWW 3 way + sub, passive acoustic LR4 at 2k, active acoustic B3 at 120Hz. Sealed box for 120Hz- with a natural LR2 roll off at 50Hz (no electrical high pass), Sub with LR2 low pass at 50Hz. MTM is still necessary to get the sensitivity to 93dB with a maximum SPL within linear excursion 118dB from 120Hz up (from effective wide baffle). Technically, I would go with TM or MTM but not TMM, given the M is a 6.5". I don't know if using a sealed box for 120Hz- would degrade the sound more than the GD introduced in LR4 in a U-frame. Power response is less even, but perhaps some power increase at the low end is better since our ears are less sensitive at low frequencies. However, I have to work with the drivers I have on hands, which would give up the SPL capability of 115dB at 80Hz in a U-frame so U-frame does not seem to be an option unless I use a LR4 for driver protection. So at the end, for reduced GD while maintaining high SPL capability I have to go with a sealed woofer box from 120Hz-.
Good choice, backing down on the complexity.
Voicing MTM's can take a very long time in my experience. Even JohnK alludes to it not being a quick or trivial process, and he's using a lot more sophisticated simulators than when I voiced the Ariels in 1991.
The discrepancy between measurements, simulations, and actual audition is a big warning sign you're in over your head. To some extent, all of us experience this on a new design, but the odds go way up with complex radiation patterns, which have a way of defeating expectations. In the end, the ear is the final guide to neutrality, but that assessment can only reached with a lot of tedious listening to pink-noise and music with a wide selection of quite dissimilar amplifiers.
By "dissimilar amplifiers" I mean radically different technologies - a variety of different price-point Class AB transistor amplifiers, Class AB pentode amplifiers with feedback (vintage style), and non-feedback Class A triode amplifiers. A good speaker will be even-handed, reveal the sonic differences inherent to the technologies, but not fail completely with one topology or the other.
Voicing MTM's can take a very long time in my experience. Even JohnK alludes to it not being a quick or trivial process, and he's using a lot more sophisticated simulators than when I voiced the Ariels in 1991.
The discrepancy between measurements, simulations, and actual audition is a big warning sign you're in over your head. To some extent, all of us experience this on a new design, but the odds go way up with complex radiation patterns, which have a way of defeating expectations. In the end, the ear is the final guide to neutrality, but that assessment can only reached with a lot of tedious listening to pink-noise and music with a wide selection of quite dissimilar amplifiers.
By "dissimilar amplifiers" I mean radically different technologies - a variety of different price-point Class AB transistor amplifiers, Class AB pentode amplifiers with feedback (vintage style), and non-feedback Class A triode amplifiers. A good speaker will be even-handed, reveal the sonic differences inherent to the technologies, but not fail completely with one topology or the other.
Lynn,
Many thanks for your posts. Your posts are always very enlightening and very delightful to read.
I have tried to tune the XO of my previous WWMTMWW (designed by somebody, I have taken out the drivers for the new OB speakers) and understood how hard it was. It took me a very long time (many nights and long weekends over one year) and numerous trials of different XOs to get it right. I only got the MTM right, but not necessarily the WW -MTM- WW right. But that is a different story because the woofer boxes were not well designed.
I read an article by somebody who wanted to investigate why MTM gave a particular sound and he found that at 45 degree off axis there was a 20dB dip while for a TM the dip was much less. My guess is that although that experiment certainly showed something, it can be far more complicated. On that particular axis with the given drivers, baffle and XO the MTM would produce a 20dB dip but under different parameters the TM may work out worse. The mathematics involved would be too complicated to comprehend and it is hard to theoretically prove one way or another. Since a number of people mentioned the same thing I guess from a DIYer (like myself) perspective one should approach MTM with cautions, and I do.
There are many factors to consider. If I need to have two Ms to make up the sensitivity and SPL, I can only do it either MTM or TMM. For TMM with a 6.5" M, I consider that the polar response would be unacceptable because the distance between the T and the second M is too long comparing to the wave length of the XO frequency. Secondly, TMM would certainly place the midwoofers 200mm closer to the floor if we maintain the ear level of the tweeter (floor reflection is increased). At last, even if we can prove that the off axis behaviour of TM is generally better than MTM, we can not deny one advantage of MTM - the superiority of its vertical response over that of the TM. So, I chose MTM over TMM, and prefer to deal with the complexity of the XO.
Of course, I would choose a 8" or 10" pro driver to make a TM instead of two 6.5" to make a MTM if I could. But I am not ready to give up my existing 4 x 6.5" drivers yet because they are some of the best of the HiFi drivers and cost a fortune. If I did not have them, for sure, I would try a TM with a pro driver.
I would be very interested in how your design goes. I can see a lot of goodies in it already. The only thing I still have a question is how the dipole peaks and nulls are dealt with. Using dipole simulation software, I have never got it right with a large baffle unless, I imagine, the back waves are damped. A large baffle makes the first dipole peak and null low, which is hard to be dealt with in a XO. John K's approach is to make a narrow baffle to shift the first dipole null to the XO region to be dealt with by the XO. I would like to have a lot more driver excursion headroom so I would not want a narrow baffle. What I will try is to use a narrow front rounded panel to get the diffraction very smooth, with extended, angled side wings to make it effectively a large baffle to preserve sensitivity, SPL / headroom (with dipole=monopole at 100Hz). In addition, the side wings will allow me to place 100mm to 150mm thick stuffing material. If the back waves, which I don't necessarily want in a small room, can be reduced by 20dB, for example, the dipole peaks and nulls will no longer be a problem. Why not a monopole then? because I don't want the "box" sound. One may argue that the stuffing materials are not linear. My thought is that it is for damping the back waves, and if one observes the measurement it is still far more linear than the reflections from the front wall. It is fairly linear and effective from 250Hz to 4k, in the range we want it to be. Below 250Hz, the stuffing materials gradually become ineffective, making the radiation pattern closer to a dipole. At higher frequencies, it is closer to a monopole, giving an even power response when blending with a monopole tweeter (I have tried back firing tweeter and did not really like it).
By the way, how close stuffing material to the driver magnet before it creates the "mass loading" effect?
Regards,
Bill
Many thanks for your posts. Your posts are always very enlightening and very delightful to read.
I have tried to tune the XO of my previous WWMTMWW (designed by somebody, I have taken out the drivers for the new OB speakers) and understood how hard it was. It took me a very long time (many nights and long weekends over one year) and numerous trials of different XOs to get it right. I only got the MTM right, but not necessarily the WW -MTM- WW right. But that is a different story because the woofer boxes were not well designed.
I read an article by somebody who wanted to investigate why MTM gave a particular sound and he found that at 45 degree off axis there was a 20dB dip while for a TM the dip was much less. My guess is that although that experiment certainly showed something, it can be far more complicated. On that particular axis with the given drivers, baffle and XO the MTM would produce a 20dB dip but under different parameters the TM may work out worse. The mathematics involved would be too complicated to comprehend and it is hard to theoretically prove one way or another. Since a number of people mentioned the same thing I guess from a DIYer (like myself) perspective one should approach MTM with cautions, and I do.
There are many factors to consider. If I need to have two Ms to make up the sensitivity and SPL, I can only do it either MTM or TMM. For TMM with a 6.5" M, I consider that the polar response would be unacceptable because the distance between the T and the second M is too long comparing to the wave length of the XO frequency. Secondly, TMM would certainly place the midwoofers 200mm closer to the floor if we maintain the ear level of the tweeter (floor reflection is increased). At last, even if we can prove that the off axis behaviour of TM is generally better than MTM, we can not deny one advantage of MTM - the superiority of its vertical response over that of the TM. So, I chose MTM over TMM, and prefer to deal with the complexity of the XO.
Of course, I would choose a 8" or 10" pro driver to make a TM instead of two 6.5" to make a MTM if I could. But I am not ready to give up my existing 4 x 6.5" drivers yet because they are some of the best of the HiFi drivers and cost a fortune. If I did not have them, for sure, I would try a TM with a pro driver.
I would be very interested in how your design goes. I can see a lot of goodies in it already. The only thing I still have a question is how the dipole peaks and nulls are dealt with. Using dipole simulation software, I have never got it right with a large baffle unless, I imagine, the back waves are damped. A large baffle makes the first dipole peak and null low, which is hard to be dealt with in a XO. John K's approach is to make a narrow baffle to shift the first dipole null to the XO region to be dealt with by the XO. I would like to have a lot more driver excursion headroom so I would not want a narrow baffle. What I will try is to use a narrow front rounded panel to get the diffraction very smooth, with extended, angled side wings to make it effectively a large baffle to preserve sensitivity, SPL / headroom (with dipole=monopole at 100Hz). In addition, the side wings will allow me to place 100mm to 150mm thick stuffing material. If the back waves, which I don't necessarily want in a small room, can be reduced by 20dB, for example, the dipole peaks and nulls will no longer be a problem. Why not a monopole then? because I don't want the "box" sound. One may argue that the stuffing materials are not linear. My thought is that it is for damping the back waves, and if one observes the measurement it is still far more linear than the reflections from the front wall. It is fairly linear and effective from 250Hz to 4k, in the range we want it to be. Below 250Hz, the stuffing materials gradually become ineffective, making the radiation pattern closer to a dipole. At higher frequencies, it is closer to a monopole, giving an even power response when blending with a monopole tweeter (I have tried back firing tweeter and did not really like it).
By the way, how close stuffing material to the driver magnet before it creates the "mass loading" effect?
Regards,
Bill
with an allpass filter, you only modify phase (and group delay, group delay being only another view of the phase/frequency curve) but you do not modify frequency response (it' not anymore a minimal phase system). So you can really hear if the phase itself is audible or not.
Now those digital filters (FIR and IIR) and real-time softwares, you can easily simulate lot of psychoacoustics effects and really decide if something is audible or not without modifying other parameters (I played with phase and GD, crossover types, box diffraction, early reflections, room modes,...have a try with the softs, it's free)
jlo,
Thanks for letting us know your software which I am sure is very interesting to many of us. I love to play with them but rarely have the time (I use internet only at work). I promise that someday I will. Before that happens, do you have a summary page in which you tell us your findings in your numerous software (even though it represents your personal opinions)? it would be extremely helpful.
Regards,
Bill
Thanks for letting us know your software which I am sure is very interesting to many of us. I love to play with them but rarely have the time (I use internet only at work). I promise that someday I will. Before that happens, do you have a summary page in which you tell us your findings in your numerous software (even though it represents your personal opinions)? it would be extremely helpful.
Regards,
Bill
Hi jlo,
I think it might be illuminating to study superimposed 'scope sims of input/output for first 2 cyles of sine at LF.
Group delay in RLC circuits is related to steady sine, not initial start of wave, as with kick-drum, bass guitar picking.....
The same in/out relationships need to be observed with digital processing, for the reproduction becomes altered where fundamental tones are shifted in time wrt their own unaltered harmonics.
Cheers ....... Graham.
I think it might be illuminating to study superimposed 'scope sims of input/output for first 2 cyles of sine at LF.
Group delay in RLC circuits is related to steady sine, not initial start of wave, as with kick-drum, bass guitar picking.....
The same in/out relationships need to be observed with digital processing, for the reproduction becomes altered where fundamental tones are shifted in time wrt their own unaltered harmonics.
Cheers ....... Graham.
HiFiNutNut said:
In practice, mass loading can be mostly avoided if the (wool/cotton) felt is kept at least 2~3" away from the cone surface. (Some kind of simple frame or spacer will be needed.) These are minimums - bigger cones probably need a little more distance.
Much of the "low-efficiency" reputation of transmission-line enclosures is due to the filling getting too close to the drivers, thus mass-coupling, which results in substantially degraded midrange and HF and lower efficiency. If the MF/HF response of the driver gets congested, "closed-in" sounding, or starts to measure different, the filling is too close, and interacting with the cone in an unfavorable way.
You should seriously re-evaluate your position on the drivers. Either they're good enough, or they're not, and that decision should be rigorously isolated from sunk costs or arbitrary decisions about the system topology. You really must be ruthless here, otherwise you could literally spend years trying to force-fit XYZ drivers into a design where things are never going to be optimal. This is part of the reason why companies with in-house drivers are frequently less successful than companies that are not tied to any one driver vendor - if your hands are tied to XYZ drivers and ABC design goals, frankly, the amount of work goes way up, and the chances of success go way down.
Don't mean to be harsh, but the only speakers I ever designed that I liked were ones where I started with a clean piece of paper - a process that inherently means throwing out a lot of drivers and also throwing out topologies that just don't work. Plenty of stuff looks good on paper, or in a computer simulation, and sounds terrible in the real world. Even though I'm kinda-sorta retired, I value my time, and I have a lot better things to do than force drivers to do things they don't want to do. That is the biggest time-waster of all, with the least to show for it.
I would do the simplest, most basic thing of all - listen to XYZ drivers with no crossover, full range, on a baffle. Do you like them or not? Do they put a smile on your face, or are you trying to convince yourself to like them, because they're "supposed" to sound or measure good. Pay attention to your emotional reaction; are you tensing up, or relaxing? Are you mentally justifying the sound, making excuses for the system? These initial reactions in the first minutes of listening are highly significant and relate directly to long-term satisfaction. As the Buddhists like to say, "First Thought, Best Thought."
For me, drivers are a go/no-go decision. If they are unlikable, that's it, out they go, no second chance, no 100-hour "break-in", none of that audiophile BS. If the vendor can't make a driver that sounds musical on a flat baffle right out of the box, forget it. Just about any driver out of a dusty 1955 console TV set can pass this test, why not an overpriced techno-wonder driver?
The absence of gross and obvious break-up on the FR curve is only the first pass; after that, it has to sound good - as in, similar to music, with a sense of realism and a "breath of life" that good drivers have. These qualities come down to subtle manufacturing decisions about types of glues between VC former and cone, speed-of-sound mismatches between the VC former material and cone material, and small asymmetries in the spider assembly. These manufacturing decisions only slightly show up in the waterfall/CSD curves - they're right down in the noise - but make all the difference between a musical and unmusical (hifi) sounding driver.
As a preview of what I'm going to do, I'm going to measuring and auditioning a lot of 8~12 inch drivers, listening to them full-range on an OB, no crossover, just as-is. The leading candidates are the 18Sound 8NMB420 and 12NDA520 mentioned earlier, an EnABL'ed Alnico Lowther, and an assortment of hemp-cone drivers from Hemp Acoustics and Tone Tubby. Not a Fostex fan, no plans to audition those. The Fertin has an exotic charm, but I've never heard one, and they're way too expensive to buy and then re-sell if I don't like 'em.
Fooling around with the "variable-geometry" concept and the assorted LF shaping networks follow the choice of wideband driver, and will be complementary to the overall sonic character of the primary driver, extending the dynamic range and sense of LF "presence" and tactility. I plan to do a lot of A/B switching between the LF array + primary driver and the primary driver alone to make sure the basic character remains the same.
Hi
Lynn, you several times mentioned to do your auditioning with pink noise.
I also use this test but only to compare relative balance between speakers in surround systems.
What exactly are you listening for in mono / stereo with pink noise ?
BTW what you point out in
is absolutely true for the cheap Westra (Japanese manufacturer ? ) compared to the expensive Dynaudio I mentioned - though the Westra measures awful 3rd HD and has a strong break up resonance at around 3.5 kHz it sounds way more appealing.
Centuries ago I have listened to that Westra 3-way speaker and fortunately they were still in my brothers cellar where I dig them out recently. Its amazing that we have such a good long time memory to sonic patterns.
Though it has been a closed box design, the mid-woofers sonic pattern shines through also in OB.
If RCF is available in US, also have a look there as they may have one or two candidates for your project.
Greetings
Michael
Lynn, you several times mentioned to do your auditioning with pink noise.
I also use this test but only to compare relative balance between speakers in surround systems.
What exactly are you listening for in mono / stereo with pink noise ?
BTW what you point out in
is absolutely true for the cheap Westra (Japanese manufacturer ? ) compared to the expensive Dynaudio I mentioned - though the Westra measures awful 3rd HD and has a strong break up resonance at around 3.5 kHz it sounds way more appealing.
Centuries ago I have listened to that Westra 3-way speaker and fortunately they were still in my brothers cellar where I dig them out recently. Its amazing that we have such a good long time memory to sonic patterns.
Though it has been a closed box design, the mid-woofers sonic pattern shines through also in OB.
If RCF is available in US, also have a look there as they may have one or two candidates for your project.
Greetings
Michael
How to measure a driver for use as a dipole
Hi
Here is another OB frequency response.
This time its the Westra WS-250-1302 measured in the well known 30 x 46 cm OB at 1M distance and around 5ms window which provides valid results above roughly 200-300 Hz, as indicated by the yellow bar.
As can be seen, also this chassis refuses to show the first baffle peak for one reason or another - though I can't see the reason why. ( Here the speaker was mounted from behind the baffle which should not affect the measurement too much ).
----------------------
Is there a significant difference with respect to group delay between this approach ( splitting the load ) and a simple 6 dB XO ?
Greetings
Michael
Hi
Here is another OB frequency response.
This time its the Westra WS-250-1302 measured in the well known 30 x 46 cm OB at 1M distance and around 5ms window which provides valid results above roughly 200-300 Hz, as indicated by the yellow bar.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
As can be seen, also this chassis refuses to show the first baffle peak for one reason or another - though I can't see the reason why. ( Here the speaker was mounted from behind the baffle which should not affect the measurement too much ).
----------------------
Is there a significant difference with respect to group delay between this approach ( splitting the load ) and a simple 6 dB XO ?
Greetings
Michael
Hi Michael,
It's hard to say what is going on with your drivers/baffle. That is one of the problems with these types of simulations. It is very difficult to accounf for the behavior of the rear respopnse from the driver.
FYI, he is a comparison for a Peerless 6.5" excludive on a 33 x 107 baffle with the driver at 16.5, 47.
The measurement is a combination for the near field/1M measurements with a merge point around 200 Hz. For the simulation (the blue trace is relevant) I emulated the same thing. I ran the sim for 1 M and then for 100 M. Then I patched the 100 M result below 200 Hz to the 1M result above 200 Hz. This was done because the way the near filed measurement is manipulated to get the low frequency dipole response the result is equivalent to being far from the source.
Not perfect but the trends are pretty good. Above 2K Hz the driver has a breakup peak not reflected in the sims.
I think the problems you are having may be related to driver directionality and asymmetry of the front and rear response.
These problems are very complex because the strength of the rear wave depends on so many things. The sims are usually excellent for low frequency where the front and rear are omni-directional, but as you move towards the dipole peak things can be critically dependent on the specific drivers. Again, that is why I recommend the midrange design being based on actual measurement.
It's hard to say what is going on with your drivers/baffle. That is one of the problems with these types of simulations. It is very difficult to accounf for the behavior of the rear respopnse from the driver.
FYI, he is a comparison for a Peerless 6.5" excludive on a 33 x 107 baffle with the driver at 16.5, 47.
An externally hosted image should be here but it was not working when we last tested it.
The measurement is a combination for the near field/1M measurements with a merge point around 200 Hz. For the simulation (the blue trace is relevant) I emulated the same thing. I ran the sim for 1 M and then for 100 M. Then I patched the 100 M result below 200 Hz to the 1M result above 200 Hz. This was done because the way the near filed measurement is manipulated to get the low frequency dipole response the result is equivalent to being far from the source.
Not perfect but the trends are pretty good. Above 2K Hz the driver has a breakup peak not reflected in the sims.
I think the problems you are having may be related to driver directionality and asymmetry of the front and rear response.
These problems are very complex because the strength of the rear wave depends on so many things. The sims are usually excellent for low frequency where the front and rear are omni-directional, but as you move towards the dipole peak things can be critically dependent on the specific drivers. Again, that is why I recommend the midrange design being based on actual measurement.
How to measure a driver for use as a dipole
Hi
JohnK, the GOOD thing about test baffles is that they are cut quick.
Fortunately I have a 6.5" Peerless at hand.
Though maybe not exactly the one you used but the HDS series of Peerless speakers differ only very little having ruler flat FR and their break up at around 3-4 kHz as can be seen at several data sheets at their page.
An interesting detail can be seen in the last picture. The ventilation of the VC / dust cap is done through holes in the membrane close to the VC – hadn't listened enough to decide if this has adverse effects to the rear sound ?
----------------------
Made a simulation and a 1M measurement to compare for my 30x46 cm OB
As can be seen we come closer now but there still is a severe gap between measurement and simulation in that there are additional peaks at around 1 kHz and 1.5 kHz and the dip at 1.8 kHz is 3-4 dB deeper as expected.
All in all not as smooth as your OB. Does your OB have any edge treatment ( mine is a simple 16mm MDF ) ?
I agree 100%
Anybody else who has measurements to compare ?
Greetings
Michael
Hi
JohnK, the GOOD thing about test baffles is that they are cut quick.
Fortunately I have a 6.5" Peerless at hand.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Though maybe not exactly the one you used but the HDS series of Peerless speakers differ only very little having ruler flat FR and their break up at around 3-4 kHz as can be seen at several data sheets at their page.
An interesting detail can be seen in the last picture. The ventilation of the VC / dust cap is done through holes in the membrane close to the VC – hadn't listened enough to decide if this has adverse effects to the rear sound ?
----------------------
Made a simulation and a 1M measurement to compare for my 30x46 cm OB
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
As can be seen we come closer now but there still is a severe gap between measurement and simulation in that there are additional peaks at around 1 kHz and 1.5 kHz and the dip at 1.8 kHz is 3-4 dB deeper as expected.
All in all not as smooth as your OB. Does your OB have any edge treatment ( mine is a simple 16mm MDF ) ?
I agree 100%
Anybody else who has measurements to compare ?
Greetings
Michael
mige0 said:
An easy question. Pink noise should sound just like falling water, nothing else, no mechanical, metallic, harsh, or peaky characteristics. The BBC discovered in the early Sixties (DEL Shorter, if I recall right) that audibility of "buried resonances" (CSD artifacts to us now) is 10~15 dB greater with pink-noise than music, although with extended listening, resonances become more noticeable with time and more annoying.
Now there are certain things to keep in mind with subjective auditioning of pink noise. With nearly all but the most uncolored speakers, it is very fatiguing and tiring to listen to, and will degrade the acuity of the ear if you listen more than a couple of minutes - it's like looking into a bright light, you don't want to make subtle assessments after too lengthy of an exposure.
I mostly use subjective pink-noise assessments to make quick A/B/A assessments of crossover adjustments or compare this or that driver against each other. Outright peaks of even 1/2 dB are audible in the midrange, and buried resonances as much as 20~30 dB below the main signal are audible as well. By "buried resonances" I mean a peak in a LF driver that is attenuated by, say, 20 dB in the low-pass crossover, so it doesn't appear in the main frequency-response measurement except as a very tiny ripple lost in the noise, but is readily audible with pink-noise or just the right selection of music, where the coloration will come and go with no apparent pattern.
This is why the "black-box" FR curve for an entire speaker system - or coax driver - is almost useless for assessing the real performance of the drivers. The resonances that are concealed by the crossover only appear when you look at the responses of the individual drivers, with or without the crossover filters, where you can see the resonances without the concealment of the "other" driver. Since the ear has properties akin to a synchronous detector, it can look into the noise and readily identify the resonance, unlike the FR curve, which is no more than an averaged-over-time measurement of the collective response.
So pink-noise is good for identifying peaks or more subtle buried resonances. It does not identify holes in the response - in fact, a quick A/B/A test might favor a slightly dished response over the flat one. It take a little experience not to be fooled by the "smoother" sound of a slightly dished response. As far as I know, narrow notches are not audible at all with pink noise - listening to transient clicks for phase distortion is probably a better way of pinning down these artifacts.
Pink noise also has the odd characteristic of not being sensitive at all to amplifier distortion, so the usual tweak business of swapping amplifiers to twiddle with the subjective balance does not apply. Most any amplifier is good enough for pink-noise audition, and by playing with a 0 to 2 ohm series resistor, you can find out for yourself how sensitive the system is various source impedances (damping factor). I aim for low sensitivity to source Z, since power amps can be kind of unpredictable under dynamic conditions - they're a lot less close to perfect zero-ohm voltage sources than you might think.
I've been mentioning an A/B/A test protocol. I've found the initial A/B (old/new) transition can be initially confusing, but returning to the known (the B/A part) clears things up right away. Thus, A/B/A, under the direct control of the auditioner. I'm trying to improve acuity, not fool myself, so I familiarize myself with (A), switch quickly to the new (B), listen for awhile and pay close attention to immediate impressions (with music or noise), then switch back (A). On switching back to the known, differences are immediately exposed - I may write down my impressions at that point, while they're fresh in mind.
The double-blind crowd will screech, squawk, wave their arms at an imaginary blackboard, and say that preconceptions will always bias a "sighted" test, especially one under the control of the auditioner (I make the switch, when I want, and know what I'm comparing).
My only response is that I wish that were true. Much of the time - in fact, the majority of the time - I desperately want XYZ driver, circuit, or amplifier to sound better, because "it should". I respect the designer as a friend, the design philosophy aligns with my own prejudices, the whole thing just looks right - it all comes collapsing down in that first minute. At least half the time, I have a huge sinking feeling when I flip the switch, especially back to the (A) position.
The first impression is only rarely reversed - it usually gets worse, much worse, over time. I've had at least three products - from designers who are personal friends - that I really wanted to like, but just sounded worse and worse the longer I had them in my home. It takes all of my limited supply of tact to gently return them without comment.
Exotic drivers, famous-name-designer parts or amps, the first A/B/A tells all, and it hasn't been very kind to the high-end industry or the latest fads in triode amps or high-efficiency speakers. This is why I don't trust any reviewer (or Internet pundit) any more - the stuff they like sounds just awful when I listen to it. If I'm lucky I can make my escape before they ask what I think.
The design process, for me, is trying to get the measurements and subjective impressions to converge - in a favorable direction. Rapid-comparison pink-noise assessment was a primary design tool at the BBC for many decades, and results in a bit of "BBC Sound" when you use it for loudspeaker design - not a bad thing in my book, since we're fighting against driver and cabinet colorations just as much as in the old days of the Fifties and Sixties.
By the way, Magnetar has a very interesting combination of technologies in his HT system. As I mentioned in the post above the referenced link, using quite dissimilar technologies for different parts of the spectrum - some academically respectable, others really "out there", can work just fine.
Before we get too cocky about our wonderful computer simulations and how accurately they model "real life", we should remember it took more than ten years before Neville Theile's bass-alignment theory made it from Australia to the AES Society in America. It was actually the simplification and generalization of Theile's theories in Richard Small's doctoral thesis that resulted in Bob Ashley of the AES finally noticing and getting both sets of articles in the AES Journal in 1971~73.
It's taken more than twenty to thirty years for the prosound community to start paying attention to CSD and buried resonances - despite all of DEL Shorter's BBC papers in Wireless World magazine in the late Fifties and Sixties, and the widely-ignored papers in Audio magazine by Richard Heyser, inventor of TDS testing.
Today, more than thirty to forty years after DEL Shorter and Heyser published their first articles, time-domain and CSD measurements are still all too rare in prosound horn and waveguide specifications, and we continue to see the abominable practice, dating from the Fifties, of "smoothing" ugly-looking frequency-response curves. Back then, it was "pen-damping", although we can look at Fifties magazines and see curves that look suspiciously hand-drawn, as if they were re-drawn by an artist from actual B&K machine curves that weren't so pretty. These days, with no "pen damping" to speak of, what's the excuse? Pixel-damping, maybe?
I don't think we can pat ourselves on the back with this track record of the highest-quality research being stolidly ignored for decades by the rest of the industry. Frankly, the industry-wide track record is little short of pitiful - Not Invented Here writ across a continent.
(Don't forget the development of magnetic tape and AC bias - ignored in the English-speaking world, and developed in Germany in the chaos of WWII, despite the vastly greater R&D capability of Bell Labs, RCA, EMI, and others.)
Extrapolating from this sad history, we can confident that we, too, are ignoring some very important things, only to be revealed ten, twenty or even thirty years from now. That's why listening matters - we can be pretty certain, looking at the stop-start history of audio, the best theories, simulations, and measurements of today are still leaving a lot out.
Before we get too cocky about our wonderful computer simulations and how accurately they model "real life", we should remember it took more than ten years before Neville Theile's bass-alignment theory made it from Australia to the AES Society in America. It was actually the simplification and generalization of Theile's theories in Richard Small's doctoral thesis that resulted in Bob Ashley of the AES finally noticing and getting both sets of articles in the AES Journal in 1971~73.
It's taken more than twenty to thirty years for the prosound community to start paying attention to CSD and buried resonances - despite all of DEL Shorter's BBC papers in Wireless World magazine in the late Fifties and Sixties, and the widely-ignored papers in Audio magazine by Richard Heyser, inventor of TDS testing.
Today, more than thirty to forty years after DEL Shorter and Heyser published their first articles, time-domain and CSD measurements are still all too rare in prosound horn and waveguide specifications, and we continue to see the abominable practice, dating from the Fifties, of "smoothing" ugly-looking frequency-response curves. Back then, it was "pen-damping", although we can look at Fifties magazines and see curves that look suspiciously hand-drawn, as if they were re-drawn by an artist from actual B&K machine curves that weren't so pretty. These days, with no "pen damping" to speak of, what's the excuse? Pixel-damping, maybe?
I don't think we can pat ourselves on the back with this track record of the highest-quality research being stolidly ignored for decades by the rest of the industry. Frankly, the industry-wide track record is little short of pitiful - Not Invented Here writ across a continent.
(Don't forget the development of magnetic tape and AC bias - ignored in the English-speaking world, and developed in Germany in the chaos of WWII, despite the vastly greater R&D capability of Bell Labs, RCA, EMI, and others.)
Extrapolating from this sad history, we can confident that we, too, are ignoring some very important things, only to be revealed ten, twenty or even thirty years from now. That's why listening matters - we can be pretty certain, looking at the stop-start history of audio, the best theories, simulations, and measurements of today are still leaving a lot out.
What you are seeing is the response from the lower driver on my NaO Mini baffle. There is no edge treatment. It's just a 90 degree edge of a rectangular baffle.
I agree with Lynn that you shouldn't take the baffle sims too seriously. They work great at frequencies below where driver directionality or asymmetry come into play but above that it's a question of how well the directionality models represent the real world drivers. BUT, the usefulness of the simulation is that while if they say the result is good you really can't be sure until you build and test, however, if they say the result is bad it probably won't be better in the real world.
By the way. I posted a DP/OP page at my site. It's not meant to be comprehensive, just a basic progression of design considerations.
http://www.musicanddesign.com/Dipoles_and_open_baffles.html
It probably still has a number of typos. But note the last setence of the closing comments,
I should also add that how the driver is mounted (i.e. cavity resonances and such) also affect the result.
Simulate, evaluate, cut, test....
I agree with Lynn that you shouldn't take the baffle sims too seriously. They work great at frequencies below where driver directionality or asymmetry come into play but above that it's a question of how well the directionality models represent the real world drivers. BUT, the usefulness of the simulation is that while if they say the result is good you really can't be sure until you build and test, however, if they say the result is bad it probably won't be better in the real world.
By the way. I posted a DP/OP page at my site. It's not meant to be comprehensive, just a basic progression of design considerations.
http://www.musicanddesign.com/Dipoles_and_open_baffles.html
It probably still has a number of typos. But note the last setence of the closing comments,
I should also add that how the driver is mounted (i.e. cavity resonances and such) also affect the result.
Simulate, evaluate, cut, test....
john k... said:
Yup, no substitute for doin' the work. I intensely look forward to being physically capable of getting started - some time after the RMAF, at the rate things are going.
Many many many thanks for all the superb contributions I've seen over the last several months, it's really exceeded my expectations in evey way. Thanks also to ever-patient moderators for dousing the flames when appropriate, so much nicer than the "other" audio-forums where things can spin out of control for months at a time.
Lynn Olson said:
Thanks to you for starting the thread in the first place, and trusting us to assist in whatever way we can.
I think the difference here is that all of us mods recognise that we still have a lot to learn, and want to do so, and so any &**^%&% that gets in the way of that is gently assisted to change their attitude.