That's a pretty good restatement of what I wrote, above. There are basically two kinds of companies in this business: large (focused on production and prices, and having very good engineering resources available), and small enterprises that really don't have much of anything going for them except some audiophile reviewer endorsement of their products--i.e., "audiophile aura", a.k.a., "boutique marketing". It's a well-known phenomenon around these parts, and most folks here actually stick around to avoid that sort of marketing ploy (and 10x pricing).
Please don't miss the point that these industries generally pass from left to right on the PowerPoint chart I posted, above. They start out as in a growth industry (like Bell Labs did in the early 20th century) at the beginning of their individual macro technology lifecycles, and then ultimately end up in declining industries over time as the technology comes squarely into the public domain.
To address one of the OP's comments about "drivers costing 10x as much...". This is how these type of extremely small, largely fly-by-night companies play the game, since they can't play the long game that the big manufacturers do. Once someone on an audiophile forum or magazine writes a piece intimidating anything about how overpriced their products are, the company basically disappears and the engineering and marketing personnel show up at new companies. (MQA I think just ran into that predictable lifecycle...and "flew the coop".)
Chris
Last edited:
The number following "sweep" gives us the number of the harmonic shown. E.g. "sweep 3" are the third harmonic with series resistor 0R, 6R, 12R, 24R. The colors for the resistors are defined on the pictures. Look for 0,6,12,24.
The pictures are in #104
Best regards
Bernd
It's actually limited by the amplifier noise. If I were not lazy and used my "normal" Naim amp with 120 dB dynamic range, it would be > 90 dB.
Krglee: Since 2016, Matlab has provided a home version. $100 per main, $50 for each additional toolbox.
I am very suspicious of any AES publications. I've run into so much pure undiluted BS represented there as "audioscience" that I avoid it at all costs.
If audio were a scientific field then:
- someone has a theory
- a valid theory may not contradict a single hard-data fact of life
- as soon as such a case is represented, that theory must be discarded
Here, there is a belief that sine sweep (and other tests alike) fully characterize loudspeakers. I presented a reproducible measurement that contradicts the belief. Thus, the belief must be abandoned, no further discussion is needed. Full stop.
This is nonsense. The AES is a professional organisation and produces a journal of papers that are refereed by highly qualified experts. What theory are you referring to? I am also unaware of any paper that has claimed to fully characterize the sound quality of loudspeakers or drivers. Maybe you can direct us to one of those too?
I am not sure the 2D swept results are that complicated, and I am not sure the spectrograms add anything useful. It is nevertheless a useful piece of work, thank you. From the 2D sweep displays, I draw two conclusions:
(1) There is a decrease in distortion around the system resonance with voltage drive due to its inherent damping. A test where the coil velocity was set equal at fs in each example (as opposed to matching broadband volume) would (I suspect) narrow these differences around fs.
(2) There is a distinct increase in mid and high frequency distortion as amplifier impedance increases. This rise in distortion is indicative of eddy current non-linearity reflected in the coil impedance. Importantly also re its subjective relevance, the distortion does not appear well correlated to displacement.
I have no details on the driver used, but it would be useful to know the nature of any copper used in the motor.
May I suggest dropping the Faradey ring, splitting the concentrator irons into non-through-conductive segments to exclude eddy currents, and adding a motion sensor instead with a few nested feedback loops, and the amp? Put all inside one package with DSP that drives, controls, and oversees the operation, The efficiency ratio - without a shorting ring - shall increase drastically, and the overall energy consumption shall decrease a lot. Save the planet!
Then there will be no need to measure performance externally - the driver has the measurement sensor inside and can provide any metric of your choice anytime.
At the risk of being accused of going off-topic (which I admit this is), you can suggest it if you like.May I suggest dropping the Faradey ring, splitting the concentrator irons into non-through-conductive segments to exclude eddy currents, and adding a motion sensor instead with a few nested feedback loops, and the amp? Put all inside one package with DSP that drives, controls, and oversees the operation, The efficiency ratio - without a shorting ring - shall increase drastically, and the overall energy consumption shall decrease a lot. Save the planet!
Then there will be no need to measure performance externally - the driver has the measurement sensor inside and can provide any metric of your choice anytime.
From my experience, diametric slots cut across the pole pieces works well, but a non-conductive material would likely fair slightly better.
I would also advocate dropping a discrete sensor and implementing Birt's self-balancing bridge instead.
You will probably be aware too that some driver manufacturers foster eddy currents deliberately by adding copper tips to pole pieces and the like, partly because it does shield the iron/steel parts somewhat, but mainly I suspect because it changes the inductive impedance into a semi-inductive one and thereby increases driver efficiency at higher frequencies.
This topic should address the value of measurements. The OP tells us he likes some drivers more than others. He tries to relate it to measurements. Others present rather weird measurement setups. And the topic goes on for pages on distortion, repeating stuff from former threads. Not answering the question.
Reason is: measurements do matter. And taste cannot be measured. Not yet anyway, not until Musk gets the interface to our brains right. What more is there to say?
Hi Dave,
The point is, to brake the motion of a motor, or dynamic speaker, you short the driving coil. That's all. Whether you want to do that or not doesn't change the fact that this is what happens. An amplifier with a low output impedance does a nifty job of accelerating or braking the driver depending on what the needs are at that instant.
Anyway, as to measuring the performance of a speaker driver or system, we use the best methods we have and try to correlate that to human perception. They did study that quite a bit at the NRC. I haven't looked, but there should be papers published by Toole and those would likely appear in the Journal of the AES. Those papers are in fact peer reviewed by engineers and scientists, not by audio reviewers.
In comparing the testing results with what is heard (with a large sample size), we can improve our measuring methods and what we are trying to measure. In other words, what is important to measure.
Yes, exactly. But he principles are very similar. We studied series wound motors for subway systems.
The point is, to brake the motion of a motor, or dynamic speaker, you short the driving coil. That's all. Whether you want to do that or not doesn't change the fact that this is what happens. An amplifier with a low output impedance does a nifty job of accelerating or braking the driver depending on what the needs are at that instant.
Anyway, as to measuring the performance of a speaker driver or system, we use the best methods we have and try to correlate that to human perception. They did study that quite a bit at the NRC. I haven't looked, but there should be papers published by Toole and those would likely appear in the Journal of the AES. Those papers are in fact peer reviewed by engineers and scientists, not by audio reviewers.
In comparing the testing results with what is heard (with a large sample size), we can improve our measuring methods and what we are trying to measure. In other words, what is important to measure.
I find your testing to be interesting. But your statement here is a strawman argument. I know of no serious loudspeaker designer who feels that a sine sweep, by itself with no other measurements, can fully characterize a speaker.
Sine sweeps, or other gated measurements using periodic noise, are incredibly useful, and are one of the more powerful tools we have.
I have been trying to answer the questions posed at the end of the thread opener...
"[...] there are big variances in motor design, coil size and so on, which translates to sound, but this should be clearly visible on mesurements?
Or, we don't know what to measure, and how to combine different graphs to get proper conclusion?
If there is no measurable variance, why to pay much more for better drivers? Everyone hears, but no one can measure?"
My original post (#24) directed people to several avenues of research that have been long-published but have then been largely ignored. As far as I can tell, there is no reason for the ignorance other than there requires a certain investment in time to understand the theory - not least because most of the theory stems from axiomatic particle physics: It is certainly nothing I would recommend doing for fun!
As it stands at this point in time, we use measurements and analysis techniques that often smear and obscure data that might very well be relevant if we did a better job of looking for it. Further still, we seem tied to single frequency analysis, when every other concern I have encountered with audio, perceptual or like analysis makes routine use of the bispectrum.
So my contribution is firstly that we simply have a better job to do, and that will bring forth much more appropriate measures. My second contribution is that even then we will likely be some way from establishing some idealised measure of sound quality due to our non-linear learning capabilities. Perception and sensation are not the same thing - sometimes even having no relationship at all!
But my pessimism does not mean there is no value to be gained from measurements - even by making best use of the tools we do have readily available now. There is more value to be had in driver measurements, and therein much better correlations with subjective assessments.
Hi soundbloke,
Completely agree.
As soon as you attempt to consider what each individual brain does, or differing ear performance I think you are sunk. You can't do it.
There is but one thing we can do. Reproduce the same pressure variations as the original sound at the outside of the ear. If we can do that it will be indistinguishable from the original sound by definition. Everything that follows is well out of our control, and also to the subject listening. I think many people fall down with the argument that we need to incorporate any form of psychology in the process.
Absolutely, better define what makes a difference to what we "hear" and hone our testing methods to describe and quantify those things.
Completely agree.
As soon as you attempt to consider what each individual brain does, or differing ear performance I think you are sunk. You can't do it.
There is but one thing we can do. Reproduce the same pressure variations as the original sound at the outside of the ear. If we can do that it will be indistinguishable from the original sound by definition. Everything that follows is well out of our control, and also to the subject listening. I think many people fall down with the argument that we need to incorporate any form of psychology in the process.
Absolutely, better define what makes a difference to what we "hear" and hone our testing methods to describe and quantify those things.
I wouldn't go as far to say that how we test obscures anything. It is all information. Better testing? Sure. Better interpretation? Yes. I think we are at a point that testing of electronic systems passed some time ago. We will get there, we just aren't yet.
But the measurements we see used routinely actually do obscure relevant data. The spectrogram used in this thread, for example, and any other time-frequency waterfall plots, are smeared Wigner Distributions. And embedded/encoded/dispersed in the noise floor of such measurements are all the non-linear components too.
Hi anatech, current or voltage drive is misnomer in that sense, as you have said, for some particular acoustic output either would need to have same current, or voltage. I mean, to get same acoustic frequency response from a transducer, it doesn't matter which "drive" one uses, the current through voice coil would need to be the same for both. It's current that makes force in the motor, the acoustic output. While this driving a transducer is not important to distortion as such, it's only one side of a coin.
There is huge difference between the two, which is output impedance of the amplifier being in series with voice coil. Some distortion mechanisms in driver motor induce voltage to the coil, simplified/lumped as back EMF voltage, and if series impedance outside driver is low the induced voltage makes current flow in the circuit, which means force in the motor, which is acoustic output! Back EMF is voltage source within the transducer, and total circuit impedance determines how much it affects current, basically the amplifier output impedance is load for back EMF voltage source, which is the other side of the coin and very much affects distortion measured in acoustic domain.
Basically all electro-dynamic transducers available are made to have maximal effect on current, they are manufactured to EQ current by dominating circuit impedance assuming constant voltage amplifier, and thus a transducer impedance determines it's own acoustic output. This is fine goal, good frequemcy response, but the downside to this is also any non-linear aspects of driver impedance have maximal effect on acoustic output!
So, inspecting loudspeaker distortion from acoustic domain, only difference between voltage and current drive is how much a transducer can emit it's own distortion. Hence, this would be core to analyze acoustic distortion measurements to distinfuish distortion sources for example. Why I say amplifiers are practically irrelevant, they often have very good distortion performace in comparison to transducers, which have non-linearities both in electrical and mechanical domain, and they are even interlinked. Optimizing circuit impedance in series with voice coil is one way to reduce distortion in acoustic domain, a voltage amplifier straight to transducer allows motor distortion maximally emit into acoustic domain, while any additional impedance would decrease it, of course affecting to frequency response as well, but in active system frequency response can be manipulated independently.
To summarize, it's better forget perspective of "amplifier driving a transducer", that perspective to the system is irrelevant when toying around with speaker distortion, as you've said it's of topic, it is the wrong side of coin to look at. The other side of the same coin, load impedance for back EMF is very relevant though, and the practical difference between current drive and voltage drive is the amplifier output impedance. Output impedance of the amplifier works as load for backEMF and determines how much motor distortion mechanisms affect current and force in the motor, and emit acoustically. Amplifier has nothing to do with it, basically the driver outputs it's own distortion, which can be reduced by increasing series imepedance.
My attempt is not to preach you, I hope this would help to resolve confusion with the discussion as there seems to be some. Hopefully there is no need to fight which is better, all we care about is know what affects what, so any given system at hand could be optimized for better, what ever that is.
Last edited: