TH-18 Flat to 35hz! (Xoc1's design)

[Josh Ricci]

Originally Posted by Tom Danley

I really liked the parameters of the 18sound woofer in question but they had a problem at high powers the edge suspension / cone attachment would fail too often. They may have fixed (they were working on it) that but we did not go back to using that one. They do make some very nice drivers though.
Best,
Tom

Thanks for that info Tom. I have been interested in the 18Sound drivers for a while. When you say "edge suspension / cone attachment" do you mean the edge of the cone near the surround or the spiders near the triple joint with the cone and former? I suspect the former. I recently damaged one of my B&C 21's in a TH near the cone edge and surround so I am interested if you observed the same failure mode in the 18Sound drivers. Admittedly I was doing a "what will she do?" type of testing when the B&C cone was damaged.

[Josh Ricci]

On the subject of the "missing 3dB" from the simulations to Danley's measurements, here is some supportive evidence for Danley's data.

Here are measurements of 3 different TH's taken groundplane at 1m and 10m and compared with HR simulations. The drive levels used are voltages that should allow not more than 1W or 100W into any cabinets minimum impedance as measured. This results in some funky drive voltages that are not 2 or 2.83v. 2 of the cabs are of my own development and one is the DTS-10. As Tom has mentioned there are some differences between the real world cab and the simulated one and you also have calibration and equipment tolerances and small voltage, distance and setup errors that can all add up. Still as can be seen the large size of these enclosures does provide some extra apparent gain to the microphone in the 1m versus the 10m results primarily in the low bass, but regardless both distances result in quite a bit of extra output being apparent when compared with the simulation data. 2-5dB in fact over large frequency ranges. I repeated many of the measurements on separate occasions second guessing myself as to whether I had made some errors in the setup which resulted in these discrepancies from the models but they are there.

Food for thought.


Attachment 295376

Attachment 295377

Attachment 295378

If you look at the DTS10 measurements and compare with DSL's while accounting for the fact that I measured with the drivers in series for a roughly 10ohm min impedance and used a 3.3v input versus DSL using a 2.83v input into parallel wired drivers you will see that they match rather well. The difference in drive level is right about 5dB stronger in the DSL measurement. Obviously there are also some measurement, setup, smoothing and equipment differences there too, but if you scale up my 10M DTS10 measurement (The red line)another 5dB to account for the increased drive level it is hitting just about 100dB at 30Hz which is darn close to what is shown in DSL's chart.

[Tom Danley]

Hi Josh
The failures were at the outer edge of the cone. They reminded me of the failures when I was working out the Servodrive subwoofers.
Those could fold the cone into a sort of a star shape and produce creases that ran from the center to the outer edge.
To get a better feel for what was going on then, we made a Plexiglass window and found at a specific frequency, one could fold up the cones on command at that “just right” frequency (very disheartening at the time). The solution was to make a very strong cone, one a 300 LB could stand on and not collapse.

With the 18sound, there was usually a single crease where the edge suspension let go (not symmetric) and the sound that flapping edge suspension made alerted the operators there was something wrong..
That failure would not happen using a pink noise signal but with a sine wave at just the right frequency and above a certain level you could find a “hot spot” on the cone where that small section was moving farther than anywhere else (a failure point).
This was in the very early production of the driver so it isn’t surprising there was a problem, also being a horn loading that has a reasonably high air pressure, it may have been the compression ratio past safe for that cone.

I am not at the shop, but I believe 18 sound worked on the design and sent a sample but I don’t recall anything about the tests. Meanwhile, production had moved that on to a B&C driver, a company that has been very nice to work with and eager to make what we need.

That kind of issue may have been the problem you had “pushing” the 21inch, that has a lot of cone area, a large distance from the VC attachment to the cone edge. You can make scary bass with larger radiator areas but maybe a better way to get there is with multiple drivers, that way you get more area and more motor.
Also, if the upper bass range measured differently than the prediction (less) cone flex may be the cause.
On the efficiency thing, all I can say is a proper measurement trumps a computer model. By proper, you need a piston mic calibrator or a mic with known calibration (we use earthworks measurement mics), a real wide band RMS Volt meter (we have HP400’s and HP3456’s to set the drive level and something like a TEF machine which accepts the mic voltage sensitivity etc.
Then you need to be far enough away so the box isn’t distortion the “1 meter” radius assumption. Keep in mind too that in airborne acoustics 1 dB is nothing, in loudspeaker driver production, there is a normally + - 2 dB tolerance on sensitivity just with the raw drivers.

Want to hear something cool (the latest big synergy horns in use)? If you have headphones and facebook, go to the link, go to “posts by others” hit see all and scroll down to the first video posted by Mike (a hifi recording, Diana the Krall).
That was taken 700+ feet from the loudspeakers, it sounds like that everywhere and using point sources, the measured SPL variance is only + - 2 or 3 dB (depending which meter you used) over all the seats from under the system to 780 feet away..

Danley Sound Labs, Inc. | Facebook
Best,
Tom

[NEO Dan]

Originally Posted by Josh Ricci

Links not working

Hey stranger, did you get your driver reconed?
It'd be great if you could add a standardized voltage sensitivity plot like DSL with the one voltage or,,, possibly even adjusted for total nominal driver impedance(20v/4Ω/10m...) so those of us trying to make comparisons can play along


I'm ~100 yards from the Police station so I've been doing 10v/5m for 4Ω cabs, even that is loud enough that I don't take more than a few sweeps at a time All the dogs within ~1/8th mile radius go NUTS like alien squirrels from mars have just invaded.

...

This makes me wonder how Ivan gets through testing with the residential street so close to the shop. I guess that zoning would provide some protection from complaints...

[iand]

Originally Posted by Josh Ricci

On the subject of the "missing 3dB" from the simulations to Danley's measurements, here is some supportive evidence for Danley's data.

Here are measurements of 3 different TH's taken groundplane at 1m and 10m and compared with HR simulations. The drive levels used are voltages that should allow not more than 1W or 100W into any cabinets minimum impedance as measured. This results in some funky drive voltages that are not 2 or 2.83v. 2 of the cabs are of my own development and one is the DTS-10. As Tom has mentioned there are some differences between the real world cab and the simulated one and you also have calibration and equipment tolerances and small voltage, distance and setup errors that can all add up. Still as can be seen the large size of these enclosures does provide some extra apparent gain to the microphone in the 1m versus the 10m results primarily in the low bass, but regardless both distances result in quite a bit of extra output being apparent when compared with the simulation data. 2-5dB in fact over large frequency ranges. I repeated many of the measurements on separate occasions second guessing myself as to whether I had made some errors in the setup which resulted in these discrepancies from the models but they are there.

Food for thought.


Attachment 295376

Attachment 295377

Attachment 295378

If you look at the DTS10 measurements and compare with DSL's while accounting for the fact that I measured with the drivers in series for a roughly 10ohm min impedance and used a 3.3v input versus DSL using a 2.83v input into parallel wired drivers you will see that they match rather well. The difference in drive level is right about 5dB stronger in the DSL measurement. Obviously there are also some measurement, setup, smoothing and equipment differences there too, but if you scale up my 10M DTS10 measurement (The red line)another 5dB to account for the increased drive level it is hitting just about 100dB at 30Hz which is darn close to what is shown in DSL's chart.

I suspect one cause of such cone failures is the shape. Straight-sided cones (which most sub drivers used) are the strongest if the main stress is force from the voice coil in the plane of the cone (or standing on it), but not if there is a large force from air pressure like in a horn with high CR.

In this case a better solution might be "inverted dome" bowl-type "cones" like some of the high-power car audio drivers use, usually pressed aluminium or titanium. Unfortunately this shape precludes using some of the really stiff/low mass materials which need a straight line (no curvature) in one direction.

Maybe the ideal would be a double-skinned bowl shape with low-mass structural foam in between?

[David McBean]

Originally Posted by djk

"It is interesting that the Klipschorn has a claimed sensitivity of 105 dB @ 1W/1M"

In 0.5Pi, Tom is speaking of 2Pi.

113.67dB/W/1M in 0.5Pi

Hi djk,

This would suggest that the output of a TH-118 tapped horn in eighth space is at least 6 dB higher than that of a Klipschorn bass corner horn, even after taking the difference in sensitivity definitions into account. It is interesting that the performance of a tapped horn can be superior to that of a purpose-designed reactance-annulled corner horn. I am somewhat surprised that the difference is as large as it is, even after allowing for the possible driver differences as mentioned by Tom. A 6 dB improvement in SPL means that 4 times as much acoustical power is being radiated for the same input - a remarkable increase in efficiency / sensitivity.

Incidentally - Klipsch refers to SPL in dB @ 1W/1M as "sensitivity" when technically it is really a measure of power conversion efficiency. SPL in dB @ 2.83V/1M as used by Tom, is the more appropriate definition of sensitivity.

Kind regards,

David

[Josh Ricci]

Tom,

Re: Post #693

The 21" cone failed when doing exactly the type of sine wave testing that you describe for the 18Sound.

Also I acquired an assortment of ACO Pacific gear and mics to go with my M30's earlier this year.

(NeoDan no recone yet. Many many applications of rubberized CA glue later I'm going to turn the cone and try it again while paying more attention. That way I can save my other good driver. If it goes again or not hopefully then a recone kit will be ordered. With a little more knowledge.)

[djlivex]

Anyone tried the new P-Audio C18-650 ELF, 8.0mm Xmax? Thanks!

[kctess5]

I'm starting the build of 4 of these babies in about 2 weeks.

What I'm wondering is what high pass filter I should use. Looking at the hornresp excursion and frequency response graphs I am thinking a 24db/octave Butterworth filter at 28Hz. That should leave 40Hz almost unaffected and be 6db down at 28Hz. When I model the excursion with 6dB less power at 28Hz it is well under the first excursion peak in the passband.

Any thoughts?

BTW the drivers are B&C 18PS76's. I will only be using 300 watts per so excursion shouldn't be too bad

[NEO Dan]

300w on the 18's is nothing, if you don't already have the 18's you could easily do cheap 15's in there and save a few hundred dollars.