Bushmeister:
Well, no - it doesn't work quite like that. Those figures are for max output powers. The relevant figure is damping factor (i.e. low output impedance), which is zillions (miniscule) at low/mid frequencies - assuming a "normal" solid-state amplifier. Leading to errors less than 0.01 dB (probably!).
Mind you a SET valve amplifier wouldn't be appropriate here, as these have a relatively high output impedance (low damping factor).
Well, no - it doesn't work quite like that. Those figures are for max output powers. The relevant figure is damping factor (i.e. low output impedance), which is zillions (miniscule) at low/mid frequencies - assuming a "normal" solid-state amplifier. Leading to errors less than 0.01 dB (probably!).
Mind you a SET valve amplifier wouldn't be appropriate here, as these have a relatively high output impedance (low damping factor).
Of course you are absolutely right - hadn't thought that one through properly!
Very likely to be no discernible difference - just being overly cautious!
Some thoughts on the models and measurements
Following along, and pushed for time, so this is a bit rough (and it's taken a while to draft so may miss recent posts):
The various measurements speak for themselves - great to see every one. They are mostly in the near field for diffraction: (aperture size)^2 << (wavelength times distance).
I take the question to be something like "why is there gain over the basic +3dB from doubling the power" and believe it is not trivial.
The Linkwitz model, *with* the approximations in the last step, shows that one gets the same result by multiplying a single dipole response (amplitude) by two.
Aside: it's interesting to play with the full result numerically. It reveals where the model diverges from that of a dipole with 2 units of amplitude. As expected response shapes get complicated as the clamshell spacing and wavelength become comparable.
As I think has been noted, the Linkwitz model is for point sources - so the driver/source in front does not block radiation from the one behind.
Therefore, the big question from my point of view, is whether or not such obstruction of the rear driver is important. I do not see an easy way to answer that.
By implication, I tend to agree with those who expect air-load and coupling effects to be relatively small. Consider that mms and mmd don't differ much for the drives in question (and half of the air load is on each side). There's not much sign of changing resonant frequency away from fs (unless I've missed it). We can't see fs in William Cowan's measurements which would probably be the most revealing of any coupling, if there's enough snr at fs=32Hz.
+6dB across a wide range of frequency implies that, from the microphone's point of view, the driver in front is not blocking radiation from the driver behind. That's equivalent to the assumption of point sources and the resulting factor of two as a good approximation.
The measurements suggest that can be the case, or nearly so. I don't see a good way to see why that should be the case.
I did a quick test to try to get a feel for the effect of an obstruction placed in front of a 14" driver on a baffle (not the same situation, I know). I held a cone-sized dinner plate 7 or 8 cm in front of the cone and saw almost no change in the signal on-axis at 1m, 45cm above the floor (blue curve, no plate, others with plate). The measurement shown is the raw signal in Holm Impulse, i.e., no gating or averaging. The blip near 50Hz is due to the UK line frequency and noise dominates the low end.
The presence of a baffle makes it unclear whether this measurement has relevance to the clamshell, and I don't have a spare driver of similar size to test separately. However, I did not expect the curves to overlap as closely as they do, to better than 0.5dB from 50Hz, to 190Hz.
Arguments can be made based on diffraction (big waves, small obstacle, ...), but the measurements are well within the "Fresnel" range - I thought there might be at least hints of a shadow as close as 1m.
Nice puzzle!
Ken
I think I was too slow to post and attachment did not appear, though in preview. Trying again...
Following along, and pushed for time, so this is a bit rough (and it's taken a while to draft so may miss recent posts):
The various measurements speak for themselves - great to see every one. They are mostly in the near field for diffraction: (aperture size)^2 << (wavelength times distance).
I take the question to be something like "why is there gain over the basic +3dB from doubling the power" and believe it is not trivial.
The Linkwitz model, *with* the approximations in the last step, shows that one gets the same result by multiplying a single dipole response (amplitude) by two.
Aside: it's interesting to play with the full result numerically. It reveals where the model diverges from that of a dipole with 2 units of amplitude. As expected response shapes get complicated as the clamshell spacing and wavelength become comparable.
As I think has been noted, the Linkwitz model is for point sources - so the driver/source in front does not block radiation from the one behind.
Therefore, the big question from my point of view, is whether or not such obstruction of the rear driver is important. I do not see an easy way to answer that.
By implication, I tend to agree with those who expect air-load and coupling effects to be relatively small. Consider that mms and mmd don't differ much for the drives in question (and half of the air load is on each side). There's not much sign of changing resonant frequency away from fs (unless I've missed it). We can't see fs in William Cowan's measurements which would probably be the most revealing of any coupling, if there's enough snr at fs=32Hz.
+6dB across a wide range of frequency implies that, from the microphone's point of view, the driver in front is not blocking radiation from the driver behind. That's equivalent to the assumption of point sources and the resulting factor of two as a good approximation.
The measurements suggest that can be the case, or nearly so. I don't see a good way to see why that should be the case.
I did a quick test to try to get a feel for the effect of an obstruction placed in front of a 14" driver on a baffle (not the same situation, I know). I held a cone-sized dinner plate 7 or 8 cm in front of the cone and saw almost no change in the signal on-axis at 1m, 45cm above the floor (blue curve, no plate, others with plate). The measurement shown is the raw signal in Holm Impulse, i.e., no gating or averaging. The blip near 50Hz is due to the UK line frequency and noise dominates the low end.
The presence of a baffle makes it unclear whether this measurement has relevance to the clamshell, and I don't have a spare driver of similar size to test separately. However, I did not expect the curves to overlap as closely as they do, to better than 0.5dB from 50Hz, to 190Hz.
Arguments can be made based on diffraction (big waves, small obstacle, ...), but the measurements are well within the "Fresnel" range - I thought there might be at least hints of a shadow as close as 1m.
Nice puzzle!
Ken
I think I was too slow to post and attachment did not appear, though in preview. Trying again...
Attachments
Last edited:
Thank you - I think I have repeated this question about 5-6 times in this thread! This I believe is the crux of the issue.
+1 to everything you have written. Thanks so much for taking the time.
Last edited:
Not many amps can exactly double their output power into half the load. But that is the max output, which is typically impacted by increased current flow and causes distortion to rise prematurely (e.g. at a less than doubled power). Below that max power, one should get double the power with half the load impedance.
In short, as long as the measurements are done at low power of a few Watts, which falls well below the amp's max capabilities, the amp should not influence the SPL.
OTOH, the distortion can be slightly different at different load impedances (if one is measuring distortion) but the amp's distortion level will not be large enough to make a difference, e.g. the driver's distortion will dominate the total distortion measured.
Last edited:
Charlie - this is great news - very grateful for your insightful analysis as always and to get measurements from you would be excellent!
It appears we are all quite intrigued by this behaviour! I didn't expect this from a build thread!
And of course you are right re amps - I am just becoming paranoid regarding the scientific rigour of our experiments!
Ken,
At low frequencies up to probably 250Hz the drivers' front and rear sources are indeed very monopole-like. At these (long) wavelengths, the wavefront simply propagates around objects as if they are not there - it does not follow "line of sight" type behavior.
So I do not think, at least at low frequencies, obstruction of one source by the other would be a contributing factor to the SPL trends, etc.
@CharlieLaub,
you have measured many more speakers than I, and I note what you expect to see.
At first I thought: big waves, small obstacles. In that case the +6 dB on axis is a given (SL analysis holds at least to the level of minor coupling effects).
On the other hand, with the mic close enough to the obstruction, I expect a geometrical "shadow" on axis. I did a test with a 2-foot square glass plate, and it did cast a measurable shadow (2dB IIRC, I didn't save it, the plate was heavy and wobbling around as I held it up off the floor centered on the driver, otherwise as the dinner-plate experiment).
My current guess is that a shadow is only seen when the mic is much closer than about the size of the obstruction, provided the latter is much smaller than a wavelength.
Ken
you have measured many more speakers than I, and I note what you expect to see.
At first I thought: big waves, small obstacles. In that case the +6 dB on axis is a given (SL analysis holds at least to the level of minor coupling effects).
On the other hand, with the mic close enough to the obstruction, I expect a geometrical "shadow" on axis. I did a test with a 2-foot square glass plate, and it did cast a measurable shadow (2dB IIRC, I didn't save it, the plate was heavy and wobbling around as I held it up off the floor centered on the driver, otherwise as the dinner-plate experiment).
My current guess is that a shadow is only seen when the mic is much closer than about the size of the obstruction, provided the latter is much smaller than a wavelength.
Ken
This thread is very interesting. It’s fun to read along.
Something about the 1 vs. 2 woofer measurements is bugging me, though.
Bushmeister, for your measurements:
You measured one of your subs that has two drivers.
You have two identical amplifiers driving them.
Are they mono amplifiers?
Are they stereo amplifiers where you are testing one channel output?
Was your other clamshell sub (2 more drivers) connected during the measurements?
Is your measurement software sending a mono signal (i.e., just left or right to the amplifier)?
What I’m getting at is what load and what input are are the amplifier(s) seeing during the measurements?
I apologize for all the questions, but this may clarify comparisons of measurements around this concept.
Perhaps a comparable signal chain diagram would help.
Something about the 1 vs. 2 woofer measurements is bugging me, though.
Bushmeister, for your measurements:
You measured one of your subs that has two drivers.
You have two identical amplifiers driving them.
Are they mono amplifiers?
Are they stereo amplifiers where you are testing one channel output?
Was your other clamshell sub (2 more drivers) connected during the measurements?
Is your measurement software sending a mono signal (i.e., just left or right to the amplifier)?
What I’m getting at is what load and what input are are the amplifier(s) seeing during the measurements?
I apologize for all the questions, but this may clarify comparisons of measurements around this concept.
Perhaps a comparable signal chain diagram would help.
Amplifier - Behringer NU46000
4 channel amp.
Each channel is wired directly to each sub.
Each sub is 8 ohm.
Channel 1+2 linked (mono left)
Channel 3+4 linked (mono right).
So both subs get exactly the same signal in each speaker, and each sub has it's own amp channel.
I hope this clarifies!
I thought peak is at 650Hz and null at 800, but yes that' too low for 15cm diameter. And ok for 15".
Last edited:
Yes - physically disconnected (you can see it on the ground in the picture!)
Audio interface is miniDSP 4x10HD - nothing changed here - same signal sent to amp inputs throughout. Amp has a switch to link inputs on the back.
Only one 8 ohm load active.
I think you are right - we are going to have lots of experimental data soon hopefully, so might be a good plan, given this doesn't seem to have be looked at properly before.
Just jumped into this tread and I am curious whats it about.
So if I read some pages back it seems that the mistery is why two identical woofers in clampshell (with some distance in between) position output 6db more SPL?
Maybe I missed something, but that has always been the case... Two speakers sum coherently at low frequencies, so power is times 2 (+3dB) and SPL is times 2 (+6dB).
What essential thing am I missing in this discussion? Why is the 6dB not as expected?
This has been my starting point from the beginning. But there are many counterpoints - might be worth reading through the thread!