Well, I replied on page 1 post #9 with a specific quantitative measure (and link to a less-daunting "better thread" illustrating how limited-bandwidth and cross-over both distort transcient response) -- one might say very much in line with OP's oscilloscope square-wave comp -- but he doesn't seem to have noticed or followed his thread.
Last edited:
Well your statement was expressly a proposal, not showing a comparison. (..or even asking: "what do you think about this?")
Really it reads as if "this is something else you could do", not "this correlates (or counters) with your own work in this manner".
Even reading through some of your prior thread, it's difficult to make a correlation - his testing specifically bandwidth-limits both drivers and both drivers have nearly identical Impulse Responses results, so their "envelopes" (ETC) should be very similar. Maybe I'm missing something then?
Really it reads as if "this is something else you could do", not "this correlates (or counters) with your own work in this manner".
Even reading through some of your prior thread, it's difficult to make a correlation - his testing specifically bandwidth-limits both drivers and both drivers have nearly identical Impulse Responses results, so their "envelopes" (ETC) should be very similar. Maybe I'm missing something then?
Last edited:
May I ask if "fast/slow" and "dynamic/laid-back" differences were noticed (or not) among the speakers, and what if any significant driver material/characteristics (beside directivity) may account for the post-equalization SQ differences, in your estimation? THANKS.
Last edited:
OP can consider my proposed math to quantify the qualitative comp he wrote in words in his blog. He might also consider broadening the subject under test, such as different material/construction, XO effects, etc., using various test signals. Low-hanging fruits (as researchers like to call them).Well your statement was expressly a proposal, not showing a comparison. (..or even asking: "what do you think about this?")
Really it reads as if "this is something else you could do", not "this correlates (or counters) with your own work in this manner".
Even reading through some of your prior thread, it's difficult to make a correlation - his testing specifically bandwidth-limits both drivers and both drivers have nearly identical Impulse Responses results, so their "envelopes" (ETC) should be very similar. Maybe I'm missing something then?
Yes. Seems to me, as i mentioned before, that efficiency has something to do with it. Highly efficient speakers appear more dynamic.
There is an interesting statement about perceived "speed" in this article on purifi-audio.com:
https://purifi-audio.com/2019/12/10/a-fast-driver-needs-a-light-cone-or-does-it/
Actually it focusses more on the relationship between cone weight and perceived "speed", but there are some statements, which could be interesting for this discussion as well:
and
Obviously Purifi Audio have their opinion/answer about the question in the thread title 🙂
https://purifi-audio.com/2019/12/10/a-fast-driver-needs-a-light-cone-or-does-it/
Actually it focusses more on the relationship between cone weight and perceived "speed", but there are some statements, which could be interesting for this discussion as well:
and
Obviously Purifi Audio have their opinion/answer about the question in the thread title 🙂
As I remarked elsewhere that quoted statement is almost tautological (not a value judgment); and borne out by common sense/experience -- hard-stiff vs soft-pliable membrane material. Obviously, it's only part of the picture. System bandwidth does not depend on just one driver (fullrange excepted) but on multiple drivers with XO and cabinet interactions. And transcient fidelity of the "attack" (usually lack thereof) is a function also of phase and other distortions, where XO again plays a major part.
Last edited by a moderator:
Fair point.
Discussing about "speed" and "fast" and "slow" obviously is something different, if you talk about a single driver or a complete system.
This thread (the initial post) as well as the blog entry from Joseph was referring to a single driver, so the above mentioned statements from Purifi Audio fit to the initial question.
Interesting follow up question to your point would be - are "fast" drivers needed to create a "fast" system or is a proper overall system design able to compensate "slow" drivers?
Vice versa, I think, it is not in question, that poor system design will always compromise the overall result, even, if you take the 'best ("fastest") drivers in the world' 😀
Last edited:
In principle, time/money no object, perhaps near-transcient-perfect systems could be made from (all) "slow" drivers, especially given ASP/DSP/new-tech. Ditto phase-aligned/time-coherent multi-way speakers could be designed and fine-tuned using complex XO (Dunlavy et al.). By now technical know-how, affordable test equipment, and fancy free software may even have made it possible to DIY at a very high level of complexity (well beyond me).
Fullrange and Fullrange Augmented Straight-Through speakers, especially DIY but also some classic multi-way brands (such as Ref3A, Eggleston, even Monitor Audio Studio all-alloy-first-order, flagship midrange without HPF), have historically tried to finess those difficulties by eschewing or minimizing XO -- and obtain excellent transcient speed, dynamic articulation, imaging and soundstage almost without effort -- but with well-known trade-offs (and not-so-simple solutions).
I think everyone should try a $100 pair of small bookshelf fullrange labyrinth. Maybe even this ;-) https://www.diyaudio.com/community/...r-5-2-in-evansound-3-75l.393909/#post-7219146
Fullrange and Fullrange Augmented Straight-Through speakers, especially DIY but also some classic multi-way brands (such as Ref3A, Eggleston, even Monitor Audio Studio all-alloy-first-order, flagship midrange without HPF), have historically tried to finess those difficulties by eschewing or minimizing XO -- and obtain excellent transcient speed, dynamic articulation, imaging and soundstage almost without effort -- but with well-known trade-offs (and not-so-simple solutions).
I think everyone should try a $100 pair of small bookshelf fullrange labyrinth. Maybe even this ;-) https://www.diyaudio.com/community/...r-5-2-in-evansound-3-75l.393909/#post-7219146
Last edited:
I think it's all about *bandwidth, for either a system (entire speaker) or for a driver.
System idea is probably fairly well accepted, or at least less debated.
Driver idea is bound to be debated.
My my growing opinion:
For any two drivers under comparison, i think the one with the widest bandwidth will make transients more precise....ie sound "faster"
Biggest problem in simply comparing the bandwidths, is if say the overall widths of the bandwidths are equal, but the center of their spans are different.
Then, my guess is that the driver with the higher center of its span, will generally sound "faster", given the dominance of high frequency on an impulse response.
(* by bandwidth, I mean the frequency span of flat mag and phase.
And maintains flat mag and phase through the entire span, with unclipped headroom at the max average SPL that the system or driver is specified to deliver)
System idea is probably fairly well accepted, or at least less debated.
Driver idea is bound to be debated.
My my growing opinion:
For any two drivers under comparison, i think the one with the widest bandwidth will make transients more precise....ie sound "faster"
Biggest problem in simply comparing the bandwidths, is if say the overall widths of the bandwidths are equal, but the center of their spans are different.
Then, my guess is that the driver with the higher center of its span, will generally sound "faster", given the dominance of high frequency on an impulse response.
(* by bandwidth, I mean the frequency span of flat mag and phase.
And maintains flat mag and phase through the entire span, with unclipped headroom at the max average SPL that the system or driver is specified to deliver)
Taking it at face value. The mass of the tweeter diaphragm and all of the little details about how it flexes and morphs, reflects etc, based on it's shape and materials... will have an effect. That will change the response in the real world.
However. The microphones diaphragm is subject to exactly the same multitude of details and all the others to do with getting it's diaphragm response through some filters, amplifiers and an ADC.
It's like trying to measure an elastic band with another elastic band. Comparing two lengths of strings.
If you want to measure what the tweeter is doing I would suggest using an industrial vibration camera and analysis software. It will show you in real video (DSP'd to amplify motion) and in 3D modelling (in the highly advanced cases) exactly what the tweeter is doing. You can expect waves to be rippling across it's surface interacting, bouncing off the edges, constructive and destructive feed back etc. etc. You'll find the essence in tweeter design is probably based in exactly how those ripples move, interact and cancel out when required. In the past based only on ears a mic and an oscilloscope and today with advanced telemetry, DSPs and fluid dynamics/material modelling software.
On audiophile behaviour. I find when my mum makes lasagne it tastes better. I know when I made my own OPA headphone amp it sounded better to me. I'm 100% sure if you spend £1000 on a cable pre-burner it will sound better to you.... but that's where we step bluntly into the area of audio-phoolery and snake oil.
However. The microphones diaphragm is subject to exactly the same multitude of details and all the others to do with getting it's diaphragm response through some filters, amplifiers and an ADC.
It's like trying to measure an elastic band with another elastic band. Comparing two lengths of strings.
If you want to measure what the tweeter is doing I would suggest using an industrial vibration camera and analysis software. It will show you in real video (DSP'd to amplify motion) and in 3D modelling (in the highly advanced cases) exactly what the tweeter is doing. You can expect waves to be rippling across it's surface interacting, bouncing off the edges, constructive and destructive feed back etc. etc. You'll find the essence in tweeter design is probably based in exactly how those ripples move, interact and cancel out when required. In the past based only on ears a mic and an oscilloscope and today with advanced telemetry, DSPs and fluid dynamics/material modelling software.
On audiophile behaviour. I find when my mum makes lasagne it tastes better. I know when I made my own OPA headphone amp it sounded better to me. I'm 100% sure if you spend £1000 on a cable pre-burner it will sound better to you.... but that's where we step bluntly into the area of audio-phoolery and snake oil.
How would you know the *bandwidth of a particular driver or commercial loudspeaker without measuring one (or every) yourself? And what *bandwidth might we expect of a run-of-the-mill loudspeaker (or even any loudspeaker)? The span between XO points? Nearly none? Would love to know....
And pseudo sociology. These are behavioral tropes independent of audible differences.
I wonder how long before we can measure the nerve signals from the ears or synaptic responses in the audilble regions of the brain and settle some arguments and start a whole new set of them.
I expect they already can in advanced med labs with FMRIs or what not. How long before you can tap that with a biometrics device in a Univerisity lab or your home?
How long before you can inject it directly. Even in deaf people. I'm pretty sure there is already patients using this tech, I wonder what bit depth and sample frequency they have?
I expect it's anything like vision a massive amount, like 90% will be filtered out immediately as worthless and the rest of the sound stage rebuild from our mental model and interruption. Although there aren't that many "audible illusions" compared with visual illusions so maybe we are better at sound. You might expect response to vibration and being able to determine if the vibration is a threat would have evolved long, long ago in aquatic animals. Maybe long before vision.
I expect they already can in advanced med labs with FMRIs or what not. How long before you can tap that with a biometrics device in a Univerisity lab or your home?
How long before you can inject it directly. Even in deaf people. I'm pretty sure there is already patients using this tech, I wonder what bit depth and sample frequency they have?
I expect it's anything like vision a massive amount, like 90% will be filtered out immediately as worthless and the rest of the sound stage rebuild from our mental model and interruption. Although there aren't that many "audible illusions" compared with visual illusions so maybe we are better at sound. You might expect response to vibration and being able to determine if the vibration is a threat would have evolved long, long ago in aquatic animals. Maybe long before vision.
Just happened to be in my YT list. This is what I mean by measuring vibration and doing things with it using DSP. Both this video and the one that spawned it are very interesting in just what you can do if you can remove noise and pull out a needle in a haystack!
And yet the Sean Olive/Harman preference model seems to indicate there are common loudspeaker preferences, once the test is properly set up.
I will follow your lead -- notch-filter or simple-EQ flat-ish then do a careful comp. Might take a while though. Lowther PM6A-notched 98dB will probably win both speed and dynamics, but so dry.... Jordan JX92S may lose despite extended near-flat response and phase. Others in between.
Last edited:
Measuring for one's self would of course be best.
Personally, after a driver is mounted in its box/baffle/horn....I look at the raw response and make a judgement call as to usable bandwidth.
Not hard to do really...where's f-6 or so, on the low end...and where does either breakup or rolloff on the high end begin, whichever comes first.
Then I apply filters to smooth within the passband, and to match xover target curves..
Imho, comparing processed/ filtered responses defines the bandwidth, and makes a good way of predicting transient reaction, as well as performing listening tests
Wider the bandwidth the better, in terms of transient impact of one similar driver to another...... to my ears.
Before a driver gets mounted...we are all the mercy of published curves.
But I think even that info helps predict final usable bandwidth, eh?
As to how does the bandwidth linearity hold up to full SPL? .....compression, over excursion, etc
no way around that without measuring ourselves i guess ........other than reading efforts from guys like Ricci and Erin
And i guess i should add, that just because a driver has wider bandwidth, doesn't mean i want to use it that way.
I like multi-ways, 4&5 ways that often means I'm only using a driver in the meat of its bandwidth..
For me, doing so keeps SPL sounding as clean at higher levels, as at low. (back to linearity vs level being a part of my def of bandwidth.)
Hi Joseph, i really enjoy your blog posts, as you look for ways to dig deeper with measurements.
I'm not expert with square wave testing at all, but I've done enough of it to be pretty certain it's not a good test for drivers.
The sum of odd harmonics starting with the third, that comprise a square wave simply span too much of the frequency range to give decent results, for drivers which have limited bandwidth.
For example, the 5kHz square wave used in your blog, has sine components of 5, 15, 25, 35......kHz.
How many of those frequencies are the ribbons producing, especially given the magnitude response shown by Arta?
A couple of other things I've found about square waves in general.
They can only be produced well, when both magnitude and phase are flat for the square waves fundamental frequency, and also flat through the needed set of odd harmonics.
A square wave without at least the first two odd harmonics, 3rd & 5th, gets pretty worthless imo.
So ime, we can really only use square waves at about 1/5th of the highest frequency that the DUT is capable of....so for speakers, somewhere in the zipcode of 4KHz...
Which brings me to the square waves in your blog. To me they look like what I'd expect a good soundcard's output would measure, not the ribbons.
Looks like the soundcard is running an 88.2KHz sample rate, and you can see the 5KHz fundamental, and the first two harmonics.
(Hope you don't mind a couple of snips from your blog)
I think it takes flat mag and phase from 5kHz to 25KHz to make that.
So with the ribbons rapidly rolling off high end response, I think you had to be measuring the soundcards' output of REW's generator.
Hope this has helped 🙂
Your posts on distortion have sure helped me get going with those types of measurements...
The bandwidth of the audio interface applied determines the reproduction of your 5kHz square wave. Your measurements clearly show the lowpass filter function of the measuring equipment I think. Try 192 or even 384 kHz, with appropriate analog circuitry.