Is there any merit to basing tweaks on close-mic measurements of a cone?
Pictured: tests of a new 95dB 6" mid. I measured crudely (test box with cardboard baffle), but it seems to match the manufacturer's plot pretty well, such as the 800Hz 3dB dip, so I assume my tests are good enough to base some questions on. The polars were at roughly 0, 15, 30, 45 degrees.
manufacturer's data
1) The main peaks in my polar plots seem to match peaks from the dustcap. Would damping just the dustcap be likely to give a smoother HF rolloff?
EDIT: initial results seem counter intuitive. The ~6kHz peak does seem to be from the dustcap, but damping the dustcap as the only mod (using a thin layer of silicone glue) actually makes the 6kHz peak bigger. I wonder if the peak is a cavity resonance, and the glue has made it worse by making the dustcap more airtight. This is going to need lots of trial and error.
2) Any tips on using (REW) waterfall / impulse plots to guide the tweaking process?
EDIT: a non issue / good result. These plots are already seem good, and damping makes the driver settle slightly faster.
3) Any idea what would cause that 800Hz dip / tips for reducing it?
EDIT: turns out it was the test baffle
4) This test-and-spot-tweak methodology seems like an obvious idea, but I don't recall it being done or discussed by amateurs. Why not?
It seems that dammar and other cone tweaks were discussed more often a few years ago - before the advent of cheap and easy ways to test their effectiveness.
Pictured: tests of a new 95dB 6" mid. I measured crudely (test box with cardboard baffle), but it seems to match the manufacturer's plot pretty well, such as the 800Hz 3dB dip, so I assume my tests are good enough to base some questions on. The polars were at roughly 0, 15, 30, 45 degrees.
manufacturer's data
1) The main peaks in my polar plots seem to match peaks from the dustcap. Would damping just the dustcap be likely to give a smoother HF rolloff?
EDIT: initial results seem counter intuitive. The ~6kHz peak does seem to be from the dustcap, but damping the dustcap as the only mod (using a thin layer of silicone glue) actually makes the 6kHz peak bigger. I wonder if the peak is a cavity resonance, and the glue has made it worse by making the dustcap more airtight. This is going to need lots of trial and error.
2) Any tips on using (REW) waterfall / impulse plots to guide the tweaking process?
EDIT: a non issue / good result. These plots are already seem good, and damping makes the driver settle slightly faster.
3) Any idea what would cause that 800Hz dip / tips for reducing it?
EDIT: turns out it was the test baffle
4) This test-and-spot-tweak methodology seems like an obvious idea, but I don't recall it being done or discussed by amateurs. Why not?
It seems that dammar and other cone tweaks were discussed more often a few years ago - before the advent of cheap and easy ways to test their effectiveness.
Attachments
Last edited:
Searching for bell modes is getting me some possibly useful info, thanks.
Loudspeaker and Headphone Handbook
edited by John Borwick
Attachments
First, nice, poor man's Klippel
I still have to finish my amplifiers but I would like to do similar tests eventually.
Where are you in Oz? Richard Lee is in Cooktown and is pretty experienced in this.
The 800 Hz dip shows up at the dust-cap, and the surround, and every polar.
If it's a bell mode then shouldn't there be some place where it doesn't cancel?
Maybe that's mid cone, perhaps try a sweep there.
Best wishes
David
It's probably not a bell mode if it shows up everywhere. Most LF drivers have a spider resonance in the region of 400 - 500Hz. This driver has a fairly high Fs, so perhaps the spider resonance is at a fairly high frequency.
I would suggest taking an impedance sweep. If you don't see a spider resonance before 800Hz, then this is probably it.
Paper coned pro type drivers normally have very little or no cone / surround dip, unlike paper coned hi-fi drivers with a rubber surround.
I would suggest taking an impedance sweep. If you don't see a spider resonance before 800Hz, then this is probably it.
Paper coned pro type drivers normally have very little or no cone / surround dip, unlike paper coned hi-fi drivers with a rubber surround.
A mere 3200km away, MelbourneFirst, nice, poor man's Klippel
Where are you in Oz? Richard Lee is in Cooktown and is pretty experienced in this.
Will do, cheers. If it is present everywhere, I'm not sure if that will mean:
-not a bell mode
OR
-is a bell mode, but not a valid test
A mere 3200km away, Melbourne
I will email him a "head's up" that his contribution would be useful.
Your measurements so far look like the test should be valid.
So if it doesn't show up mid-cone then, like Hennie, I would suspect that it's not a bell mode.
But you make an excellent point, many "audiophiles" don't even do tests, let alone consider their validity
What microphone do you use, a 1/4" condenser?
Best wishes
David
Last edited:
Hope so. I'll try to rig up a more consistent test system.
The close mic tests: pretty consistent around 800Hz. The dip is everywhere.
The far field tests are very position dependent. If I take many samples at almost the same spot, a few will be perfect, but most will show the 800Hz dip.
The (miniDSP) UMIK-1 is what I'm using for this.
The (Dayton Audio) iMM-6 is what I use outdoors / to get averages (e.g. I'll play pink noise to get the average output at roughly 30 degrees off axis, rather than at a specific point).
Does raise the obvious question that if it's not a bell mode then what does cause it?
These are similar 1/4" or 6 mm condensor electrets I believe, typical of what I expected but always best to check.
Why two similar mics? Just the USB convenience of the UMIK1?
Best wishes
David
I hope it has a simple cause: driver mounting.
Loudspeaker Measurements
My ripple looks very similar to what Troels demonstrates with his Vifa XT25TG flush vs/ non flush plots, at the bottom of the page.
...so I should try better mounting + the mic setup he recommends.
That and the software it enables.
I got the Dayton mic+Audio Tool first. This is basic and works very well for doing averages and quick equalisation: play pink noise, watch the bars on the phone screen, adjust eq until flat. I did this recently with my outdoor rig (laptop using Foobar as signal generator+eq, with mic+phone to measure). It allowed me to take averages over a small arc ~8m from the horn stack much faster than with REW, without an >8m cable.
The REW + UMIK1 combo is (mostly) more advanced, and much better for overlaying multiple measurements e.g. to compare several crossover configurations, or to get polar plots. It can also be refined more easily - you can mount the mic further from reflective surfaces.
Other plans:
1) Ye olde trial and error.
I've ordered a carton of drivers. With 10 units in hand, I'll be able to try out tweaks and compare the results.
2) Methodical testing.
Someone please correct me if these assumptions are wrong. If I take frequency plots and impedance sweeps under varying conditions:
-when I switch from [free air] vs. [in box], any anomalies that move will be from changing driver-air load interactions. These will ultimately depend on the final enclosure (size / stuffing / baffle placement).
-when I switch driver orientation (e.g. driver facing up vs driver facing down) , any anomalies that move will be from non-symmetry of soft parts and motor. These will be beyond the scope of hand-crafted tweaks.
-any blip that doesn't move in the above tests will be a fixed interaction between parts of the driver, e.g. bell modes, reflections off the frame, cancellation between the cone and spider. These are what I'd try to fix with cone damping and other tweaks.
1) Ye olde trial and error.
I've ordered a carton of drivers. With 10 units in hand, I'll be able to try out tweaks and compare the results.
2) Methodical testing.
Someone please correct me if these assumptions are wrong. If I take frequency plots and impedance sweeps under varying conditions:
-when I switch from [free air] vs. [in box], any anomalies that move will be from changing driver-air load interactions. These will ultimately depend on the final enclosure (size / stuffing / baffle placement).
-when I switch driver orientation (e.g. driver facing up vs driver facing down) , any anomalies that move will be from non-symmetry of soft parts and motor. These will be beyond the scope of hand-crafted tweaks.
-any blip that doesn't move in the above tests will be a fixed interaction between parts of the driver, e.g. bell modes, reflections off the frame, cancellation between the cone and spider. These are what I'd try to fix with cone damping and other tweaks.
Very nice to see a good discussion of measurement methods.
What 800 Hz dip? A few dB wouldn't be audible. If a screw (or dust cap) was loose, you'd fix it. But otherwise, hardly worth fretting about. A case of listening with your eyes rather than your ears. Funny how often in this website builders forget they have to translate the measurements into what it means for human experience, not an end in itself. And FR is really "low-hanging fruit" which isn't the whole story but it is the usual suspect lurking conveniently nearby.
What are you making? That determines what measurements you want.
B.
What 800 Hz dip? A few dB wouldn't be audible. If a screw (or dust cap) was loose, you'd fix it. But otherwise, hardly worth fretting about. A case of listening with your eyes rather than your ears. Funny how often in this website builders forget they have to translate the measurements into what it means for human experience, not an end in itself. And FR is really "low-hanging fruit" which isn't the whole story but it is the usual suspect lurking conveniently nearby.
What are you making? That determines what measurements you want.
B.
Last edited:
I understand this viewpoint. I and know how room effects (or even changing position slightly) inevitably make >3dB difference.
However:
a) I know some anomalies do get grating with time, or seem mostly benign, but then pop out on some listening material.
b) I'm keen to chase any anomalies down which have simple acoustic fixes, like damping the frame, or part of the cone.
Something like those 3-way Oris / Azura systems, which are built around ~95dB/watt wideband drivers.
I'd prefer a 4" - 6" driver with no whizzer; something a lot smoother than a Lowther, as seen here (the AH-160 plot).
My exact methodology doesn't matter, but I think probably a horn, or a short array, to get more LF bandwidth - e.g.
a) 200Hz-3,500Hz
b) 200Hz-400Hz
c) 200Hz-400Hz
d) 200Hz-400Hz
The 800Hz thing seems to be utterly swamped by baffle changes. This is the difference adding wings makes (apologies for the naff image).
No wings (green) is a ~40cm wide baffle, boosted half a dB for clarity. Black is the same thing with extensions taped on.
If I want to get really academic, and isolate driver-specific anomalies, I need to build a ~infinite baffle.
For now, it looks like I'll get most benefit / widest smooth bandwidth by:
a) Sorting the < 2kHz range mostly by finding the best (most benign) baffle size & positioning
b) damping the peaks > 4kHz with cone tweaks
No wings (green) is a ~40cm wide baffle, boosted half a dB for clarity. Black is the same thing with extensions taped on.
If I want to get really academic, and isolate driver-specific anomalies, I need to build a ~infinite baffle.
For now, it looks like I'll get most benefit / widest smooth bandwidth by:
a) Sorting the < 2kHz range mostly by finding the best (most benign) baffle size & positioning
b) damping the peaks > 4kHz with cone tweaks
Attachments
Last edited:
Still there however- I think you are correct to try to fix this.
Unlike a room response ripple, this is in the direct sound and will color the reflected sound too.
So I expect it would be audible, wouldn't be hard to check.
Best wishes
David
You think it is udible? Guess you haven't been reading Toole and psychoacoustics*. Actually, commonly accepted that a hole (as in the first charts) is not detectable on ordinary music inputs. Or I should say, it is not detectable in blind testing but very very obvious - even hideously grating and noxious - if the testing isn't blind.
Not hard to check? I wish that were true and glad you recommended an empirical approach. But double-blind testing to see if THAT driver has a hole that is detectable is hard to plan for home testing.
B.
*when people are involved, the relevant knowledge base is psychoacoustics
You guess incorrectly*, I have read Toole, in fact I consider his work so important that I paid for a copy to keep handy at home.
I did even consider to cite it in my previous post, but that seemed a bit heavy handed and pedantic.
This statement contradicts "Sound Reproduction" by Floyd E. Toole, Focal Press, 1st edition, 2008, p. 445 et seq.
What Toole does say is that "Buchlein concluded that dips are less noticeable than peaks", not as you claim, that they are "not detectable".
Toole says that "most likely the cause of a dip is a destructive acoustic interference" and continues the discussion on that premise that the dip will be very sharp, which makes it less audible.
But the dip under discussion here is not very sharp, so it's not clear how audible it will be.
Hence my idea that it would be useful to actually check.
The OP has several drivers and the dip looks pretty close to an inverted resonance, shouldn't be that hard to use a simple passive equalizer on one and do A/B, move to proper blind test if he thinks he hears a difference.
But actually I had in mind a simple test with a processed WAV file to see if a similar dip was clearly audible, then decide if it was worth a full scale blind test.
Best wishes
David
*It was a blind test!
Yea, reduced but still there. I haven't yet been exhaustive about pinning down the source; I only changed my baffle in one dimension.
Next step: make a crude soffit mount (baffle + room corner + duct tape). If that 800Hz dip is no longer evident, I can be pretty sure that it was a box / baffle anomaly.
This baffle sensitivity has surprised me. I had thought that by getting the mic very to the speaker, the enclosure would become near-irrelevant.
I think that eq to reduce a peak by 3dB = a clear improvement, but haven't tested myself on the inverse. I've always (like Ben) assumed they were a lot more benign.
There's an 800Hz dip sample here (the 3rd one) which is quite audible, but it uses a -12dB cut to make that point.
I applaud your graciousness.
The limit of detectability depends on the width of the resonance as well as whether it's a peak or dip and I haven't tested this myself either.
It seems worthwhile to read Buchlein because he tested for dips rather similar to yours, but I haven't found a copy yet, only Toole's summary.
Toole says that dips like yours are not typical and hence doesn't discuss the issue much .
My first impression is that the 5 or so dB dip and moderate Q is borderline, not obviously trivial but not clearly important either, hence the idea to test.
Even if the audible difference is small I am still quite interested as an intellectual problem, because the cause of the dip is a bit of a mystery.
And a smoother response can only help, at worst won't hurt.
Best wishes
David
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.