Yes, that is a good keyword for finding papers on it. You can also describe it as a fractional-order pole in the impedance response.
I meant the shortened term "seminductance". The full terms semi-inductance and semi-resistance come from long-established work on transmission lines.
I thought it was an existing term, but when I did a search for it I only found my own post here:
https://www.diyaudio.com/community/...-ferrite-neodynium-magnet.390980/post-7141051
I guess the question is, do we need the hyphen? Or can we just make a new word when we need one and go on peacefully about our day?
https://www.diyaudio.com/community/...-ferrite-neodynium-magnet.390980/post-7141051
I guess the question is, do we need the hyphen? Or can we just make a new word when we need one and go on peacefully about our day?
In that case, I think you should credit yourself with inventing the word!! If you deem it to include both the semi-inductive and semi-resistive elements as a single complex entity, then I would suggest it will have a genuine use too 🙂
Well if it makes things easier why not?
That was the idea. When you have eddy currents in an inductor you get a fractional rise in impedance with frequency. The fraction of the rise is related to the ratio of inductive reactance to resistance. This ratio stays constant with frequency because of the self-distributing nature of eddy currents, finding the path of least impedance at the surface of a conductor which is a balance between the permeabilty of air vs the resistance of the conductor (neither able to dominate because they are 90 degrees out of phase). Where eddy currents fully define the impedance you usually find a constant phase shift of around 45 degrees, meaning resistance and inductive reactance are equal at each frequency. The geometry of the conductor can change the constant.
An interesting side effect of eddy currents in steel is that with a voltage source they result in a constant level of hysteresis with frequency, since the eddy current area decreases at the same rate as impedance increases with frequency, so the flux density stays constant. So for a given alloy of steel there is a constant minimum level of hysteresis at frequencies where the skin depth is inside the steel dimensions.
That was the idea. When you have eddy currents in an inductor you get a fractional rise in impedance with frequency. The fraction of the rise is related to the ratio of inductive reactance to resistance. This ratio stays constant with frequency because of the self-distributing nature of eddy currents, finding the path of least impedance at the surface of a conductor which is a balance between the permeabilty of air vs the resistance of the conductor (neither able to dominate because they are 90 degrees out of phase). Where eddy currents fully define the impedance you usually find a constant phase shift of around 45 degrees, meaning resistance and inductive reactance are equal at each frequency. The geometry of the conductor can change the constant.
An interesting side effect of eddy currents in steel is that with a voltage source they result in a constant level of hysteresis with frequency, since the eddy current area decreases at the same rate as impedance increases with frequency, so the flux density stays constant. So for a given alloy of steel there is a constant minimum level of hysteresis at frequencies where the skin depth is inside the steel dimensions.
Last edited:
I worked on Barkhausen noise in dynamic transducers exactly 2 decades ago.
Once you've heard it without any masking sounds, you'll hear it even if it's somewhat masked.
In the audio files of the Focal Barkhausen noise dominates.
In the audio files of the AMT there is no Barkhausen noise.
Why?
My educated guess:
The Focal ps130 is a traditional voice coil in gap dynamic transducer. There is a 1" iron pole piece inside the voice coil among other iron parts near the coil.
The Sounderlink AMT-920 has only Neodymium near the conductors.
@mikets42 Thank you very much for all your work and for publishing it ! ...!! 🙂
Best regards
Bernd
Last edited:
D
Deleted member 375592
Here, I am concentrating only on non-linear distortions. I do not separate the reflections, either early or late, because I assume that these are linear - which is not true whenever there is acoustic foam right next to the cone.
I am not interested in RIR / directionality / RT60 / DRR / etc. These are linear phenomena that are ... boring.
D
Deleted member 375592
Thank you! My pleasure.
I have chosen so different drivers intentionally to display how complicated the real picture is and how easy it is to make a mistake with superficial measurements.
I do not know much about Barkhausen noise. Permanent magnets are ... kind of ... macro quantum mechanic devices, evil through and through. In scientific experiments, whenever we need a magnetic field, we generate it by current. So much more predictable, so fewer second-order effects.
By the way, HiVi research DM7500 is also low in high-freq intermodular products. I have no idea why.
Re: barkhausen noise
But presumably that would only occur in magnetic material that was not highly saturated? Something similar was described on Purifi's blog, except that they identified a memory effect, which could seem noise-like because of the somewhat unpredictable release of energy that was stored earlier.
It occurs to me that, whatever part of the magnet is under-saturated, the coil is going to use that to short-circuit its own magnetic loop. I also wouldn't put it past ferrite magnets that actual demagnetization occurs, so the iron "buffer zone" isn't always enough to prevent that flexing.
But presumably that would only occur in magnetic material that was not highly saturated? Something similar was described on Purifi's blog, except that they identified a memory effect, which could seem noise-like because of the somewhat unpredictable release of energy that was stored earlier.
It occurs to me that, whatever part of the magnet is under-saturated, the coil is going to use that to short-circuit its own magnetic loop. I also wouldn't put it past ferrite magnets that actual demagnetization occurs, so the iron "buffer zone" isn't always enough to prevent that flexing.
D
Deleted member 375592
Something in the construction of a typical driver goes through the cycles of demagnetization and remagnetization. By measuring from the outside as a black box, I can not pinpoint anything. What I can say is that producing odd-order distortions that are +20dB/decade and remain steady (re main) is practically impossible in classical physics.
In the Fourier transform, a discontinuity of N-th order of derivation produces harmonics that asymptotically decay on 1/(N+1) speed towards inf. A flat spectrum can be achieved only by a random series of delta functions. The +20dB/decade slope is a product of the feedback loop formed internally by the Faraday ring (aka shorting ring) which takes differential of distortions (and also kills sensitivity). The series of delta functions can not be produced by classical physics effects as Klippel discusses (it is relevant only for low frequencies, ~1/f^2). What remains? Barkhousen noise avalanches which are clearly audible on the residual noise recordings (being unmasked). Could there be anything else? Yes, of course, but let's first finish the worst offender and then look at what remains.
Whenever you "measure" harmonics you in reality apply Welch-styled averaging and smooth all the pickling noise. You get some statistically averaged numbers, and not very repeatable ones. What do they mean? Open for the wildest interpretations... and deceiving the public by less-than-honest vendors.
That's my 2c, for what it worth.
In the Fourier transform, a discontinuity of N-th order of derivation produces harmonics that asymptotically decay on 1/(N+1) speed towards inf. A flat spectrum can be achieved only by a random series of delta functions. The +20dB/decade slope is a product of the feedback loop formed internally by the Faraday ring (aka shorting ring) which takes differential of distortions (and also kills sensitivity). The series of delta functions can not be produced by classical physics effects as Klippel discusses (it is relevant only for low frequencies, ~1/f^2). What remains? Barkhousen noise avalanches which are clearly audible on the residual noise recordings (being unmasked). Could there be anything else? Yes, of course, but let's first finish the worst offender and then look at what remains.
Whenever you "measure" harmonics you in reality apply Welch-styled averaging and smooth all the pickling noise. You get some statistically averaged numbers, and not very repeatable ones. What do they mean? Open for the wildest interpretations... and deceiving the public by less-than-honest vendors.
That's my 2c, for what it worth.
So do I get it right that you state that your time/frequency/level plots from post 6 only contain nonlinear (motor) distortions and do not relate to diffraction and/or cone breakup induced signals?
Barkhausen Noise? Really? So why do we don't hear it holding a tweeter with ferrite magnet directly on our ear(s)?
I've read it somewhere in the past it is 60dB down and not hearable .
Tape hiss with old analogue recording tapes , was it Barkhausen noise?
I've read it somewhere in the past it is 60dB down and not hearable .
Tape hiss with old analogue recording tapes , was it Barkhausen noise?
Barhausen and hysteresis are two different but related effects. The hysteresis shows up as slowly decaying odd harmonics when doing sine sweeps (but can be very complex for audio signals). barkhausen is due to the discretisation of the magnetic domains and is more noise like. For some woofers, it is possible to hear this noise when sweeping with a sine.
Cheers
Lars @Purifi
D
Deleted member 375592
Yes, you are very close.
I examine the loudspeaker / driver as a black box. On this level, I can not distinguish between Barkhousen effect, trivial classical non-linearities produced by smooth, infinitely differentiable wide-sense polynomial (Volterra) dependencies, time / temperature-dependent invariance, resonant artifacts, etc. For me, all of them are LTI distortions off the BLUE (actually, biased BLBE) LS linearization of DUT at the "working point" for a specific recording.
Cone breakup has both linear and non-linear components. I register non-linear and/or time-variant parts only.
D
Deleted member 375592
We all have heard somewhat similar (to Barkhausen) effects when listening to low-bitrate codecs not employing long-term prediction. mu/A-law digital telephony has 37dB SNR, G.722.1 & G.722.2 are subband implementations of a similar strategy, etc. One of my "golden ears" called them "cellophane radios". As Hornli stated, once you listen to them, you start hearing them everywhere quite easily (which annoys you somewhat). Yet, we are quite tolerant of smooth non-linearities and soft limiting.
Exactly that. We have a means of correlating distortions with the signal that causes them, and therefore also a means of perceiving that which is not so well-correlated. A premise of our perceptual abilities appears to be that we are geared to tune in on the less-well correlated sensations, that likely stems from the purpose of separating objects in our perceptual field.
Hi Bluesystems,
I hear what you're saying about psychoacoustics, and if you understand what the information is pointing out, you'll make your equipment look pretty and have a good story.
We have a "loudness" button for a reason. Tone controls as well. But as systems become better and listening environments improve, you don't need those things. Being more than familiar with Floyd Toole's work, I am very aware how our understanding improved.
Personally I can't understand any argument over creating the same sound field. That is the definition of high fidelity after all. If you want to attempt to account for the human mind, I'm afraid you'll never, ever reach that target because it moves with time and person. If the sound field is the same and someone doesn't like it, that's their problem. They wouldn't have liked the original performance on a different date and maybe different venue.
So we can debate and maybe disagree over what measurements matter more. But, to be honest, without reference points you can never improve the technology. Round and round you go while various designers and manufacturers proclaim they have the only answer. Sound familiar? It should, that is the history of audio in every branch until it can be measured and quantified effectively (like the electronics today). The only people arguing measurements in the electronics are the ones who have problems, or are trying to differentiate themselves in the minds of a non-technical public. They are lying.
I hear what you're saying about psychoacoustics, and if you understand what the information is pointing out, you'll make your equipment look pretty and have a good story.
We have a "loudness" button for a reason. Tone controls as well. But as systems become better and listening environments improve, you don't need those things. Being more than familiar with Floyd Toole's work, I am very aware how our understanding improved.
I haven't the faintest clue what you are talking about here. I have correlated measurements with reports from all kinds of people over the decades. Raw data that doesn't hide anything. That is what taught me a great deal. Your conclusions about what I hold important are not supported at all. You are making huge assumptions.
No, observations and feedback from other humans do that. Psychoacoustics may explain some of it. Normal experimental procedures guide your way, along with measuring responses. I can't even begin to imagine designing anything without using the best metrics to guide you. I was doing that in the 1970's and 1980's when I was designing systems. I wish I had the tools we have today!
... and the answer is - yes. You can argue about what is measured and how effective it is. In the end you have to know where you are. Humans do not record the information in a static way, we have no internal references and so make horrible measuring instruments. We can compare pretty well if our mental processes are not contaminated, so at best we are decent at comparing sounds. Like a null meter, but that is it.
Personally I can't understand any argument over creating the same sound field. That is the definition of high fidelity after all. If you want to attempt to account for the human mind, I'm afraid you'll never, ever reach that target because it moves with time and person. If the sound field is the same and someone doesn't like it, that's their problem. They wouldn't have liked the original performance on a different date and maybe different venue.
So we can debate and maybe disagree over what measurements matter more. But, to be honest, without reference points you can never improve the technology. Round and round you go while various designers and manufacturers proclaim they have the only answer. Sound familiar? It should, that is the history of audio in every branch until it can be measured and quantified effectively (like the electronics today). The only people arguing measurements in the electronics are the ones who have problems, or are trying to differentiate themselves in the minds of a non-technical public. They are lying.
Hi,
I see a lot of theoretical discussions there : ). From my experience, most of the time measurements and listening tests are correlated, especially when it comes to loudspeakers.
Here is a recent example from three way build. A midrange driver from SB acoustics SB12MNRX2 was tested, but while listening loud the sound was not as clean as I would like.
Later I exchanged the driver to another one, Scanspeak 10F/4424G00, and the problem was not there anymore.
Then I made distortion tests with the same setup, level, etc and it confirmed that indeed scan speak driver measured considerably better in harmonics.
Another similar situation was tweeters being crossed too low. Even at 2k it did not sound right, but x over point >= 3kHz fixed that, and distortion tests confirmed listening impressions.
Regards
I see a lot of theoretical discussions there : ). From my experience, most of the time measurements and listening tests are correlated, especially when it comes to loudspeakers.
Here is a recent example from three way build. A midrange driver from SB acoustics SB12MNRX2 was tested, but while listening loud the sound was not as clean as I would like.
Later I exchanged the driver to another one, Scanspeak 10F/4424G00, and the problem was not there anymore.
Then I made distortion tests with the same setup, level, etc and it confirmed that indeed scan speak driver measured considerably better in harmonics.
Another similar situation was tweeters being crossed too low. Even at 2k it did not sound right, but x over point >= 3kHz fixed that, and distortion tests confirmed listening impressions.
Regards