Hysteresis
Join Our Community Subscribe to Paul's PostsHysteresis describes how something appears or responds based on its history.
What we see in the present is that way only because of what happened in the past. Stretch a rubber band and upon its release, the band does not return to its original shape. That’s hysteresis.
Understanding hysteresis means we can put its somewhat predictable behavior to good use. Take for example your smartphone’s reaction to finger flicks.
Because hysteresis can be a dynamic lag between an input and an output that disappears if the input is varied more slowly—called rate-dependent hysteresis—a slow flick of your finger on the smartphone’s screen inches up a list while a quick flick sends the list zooming forward.
We build the same rate-dependent hysteresis in PS Audio’s volume control knobs. A slow turn inches forward the volume but a fast turn sends the level up or down quickly. This programming is not by accident.
As long as we’re still in the audio category there’s also something called hysteresis distortion. This kind of distortion occurs in audio products based on magnetic principles: loudspeaker drivers, crossover inductors, or all those magnetic components in a Class D amplifier. Once the passage of a varying audio signal magnetizes a piece of magnetic material in one direction, it retains that state. When our constantly changing audio signal then reverses direction, the magnet’s historic memory adds or subtracts from the audio signal causing distortion.
Its history affects its future outcome.
It’s not worth getting hysterical about hysteresis (I couldn’t resist those words) but maybe fun to read up on it and how it affects our everyday life.
Half an hour on Wikipedia & I have a better understanding of hysteresis; thanks Paul.
Also, I now know what happens when I get bent out of shape quickly or slowly.
So I’m assuming hysteresis plays a role in what we call “burning in” new equipment?
I had to look it up as I’d forgotten what it was. Must have been from Macroeconomics.
My conclusion was it does not apply to such a remote control as hysteresis is about systems with two known or fixed end-points and how you get from one to the other. Designing a variable sensitivity remote control is not hysteresis because the endpoints are not fixed, unless you always start and end at the same volume level, which I doubt anyone does. I suspect the process Paul is describing is a process of latency.
It also occurred to me to wonder how many people actually use the knob on the front of their stereo. I’ve not actually had one on my main system since 2009.
When I see that word I immediately think of hysteresis loops and my old Akai 4000D reel to reel tape deck which I still have. I remember as a teenager buying that machine in town and carrying it home on the bus. I used a tape head demagnetiser which, after rubbing over the tape heads, the instructions told you to move it slowly away thereby creating ever decreasing hysteresis loops. It was a ritual I performed but it always seemed very strange to me. They were just words and you couldn’t see that you’d done anything, you just had to believe, like so many things. I also still have the demagnetiser but like the tape deck it’s now unused. I think I might be a bit of a hoarder.
The listener is the ‘knob’ in front of the stereo.
Inspired by the ‘nut behind the wheel joke’.
Has “knob” entered the American vernacular?
It occurred to me that whilst I use an app a lot of the time, for vinyl I have to use my remote, which has the second biggest knob in audio history. Whether it has hysteresis or latency I don’t know. All I know is it’s grossly out of proportion, but has been a huge marketing success and is the knob that has launched a thousand purchases. The biggest must be on my son’s Muso Qb2, and that has flashing lights as well and the pause and play touch controls in the middle of it.
Inspired by Christopher Marlowe.
Paul’s made an interesting point here and I think the issue is that with a conventional volume dial there is a feedback loop between turning the dial and hearing the consequent increase in sound level. It’s our ears that take time to adjust and recalibrate. Using a dial with a digital meter or light ring like on the MuSo Qb2, we know roughly what level we are aiming for and can set it almost instantly, because the feedback loop is visual and instant.
Rich,
So Australian too…hilarious!
The distortion that occurs due to the resonance of an inductor (that does not have an air core) without a capacitor and worse with it, which is part of a passive Xover, is due to the phenomenon of hysteresis, as it has been succinctly described.
This distortion can go unnoticed as long as the hearing mechanism is not exposed to a sound that is free of it, as is the case of the sound of a woofer that is connected directly to the output transistors, or transformer (tube amp) as happens in active multi-amplification.
Try it for yourself, building a subwoofer with passive Xover (the best you can build) and the same sub, with Xover active (preferably digital) just like I have.
Once this difference is detected, there is no going back.
The phenomenon of hysteresis affects audiophiles due to a hysteresis loop in transformer steel. Because the magnetic domains not only lag but don’t retrace their magnetization in one direction positive going input-> north to south pole as they do in the opposite direction negative going direction-> south to north pole they create a hysteresis loop which matters when the transformer is in the signal circuit such as in the transformer output of most vacuum tube amplifiers.
https://eng.libretexts.org/Bookshelves/Materials_Science/Supplemental_Modules_(Materials_Science)/Magnetic_Properties/Magnetic_Hysteresis#:~:text=A%20magnetic%20hysteresis%2C%20otherwise%20known,(i.e.%20narrow%20or%20wide).
This results in a unique and peculiar type of distortion transistor amplifiers and OTL tube amplifiers can’t create. Perhaps this explains why I was able to identify by just listening, the only transistor amplifier by hearing alone that was at the vacuum tube valley trade show a dozen years ago. It may also explain why McIntosh installs autotransformers in their solid state amplifiers to deliberately created this type of distortion making them sound tubelike. You can also see how the curves flatten out at the extremes becoming progressively non-linear to the point where the curves reach saturation where an increase in the inducing electrical field creates no increase in the induced magnetic field.
Another example of a hysteresis loop is a thermostat to control a heating or air conditioning system. If it didn’t exist minute changes in temperature would have the heating or air conditioning system turning on and off every few seconds limiting its life considerably. Instead it takes a few degrees deviation from the set point before the system turns on to bring the temperature back to where it was set.
The delay caused by the phenomenon of hysteresis explains the limitations of servo systems and feedback loops. At a frequency where the delay in the feedback signal is 180 degrees out of phase with the input signal you have an oscillator reinforcing that is magnifying distortion at that frequency rather than cancelling it. Hysteresis loss is why the bumper car ride at amusement parks is so much fun. You turn the steering wheel and nothing happens at first so you oversteer. The same phenomenon in piloting large ships is not so much fun. Ship captains have to learn how to deal with it. The property that causes most hysteresis is inertia. A body at rest tends to stay at rest, a body in motion tends to stay in motion.
BTW, there are no hysteresis loops in loudspeakers. The hysteresis of the time between the application of voltage to a voicecoil and the movement of the cone is called group delay. It is due almost entirely to the inertial mass of the cone and the relatively constant force of the static magnetic field of the permanent magnet. The change in the magnetic field created by the voice coil’s changing current happens at the speed of light in a vacuum times the index of refraction of the material the coil is made from.
Paul, Ted Smith uses an output transformer to couple the DS DAC signal to the output jacks. He does this to high frequency filter the DSD data stream and thus convert it to analog. In doing so he adds the transformer hysteresis distortion to the analog signal. I wonder if this adds to the analog sound of the DS DAC?
An air core inductor is a better choice for lower distortion. This is why no loudspeaker system of any value uses iron core inductors. The ferrite core reduces cost for a given inductance but at a price. Air core inductors wound around non ferrite cores such as around cardboard tubes or plastic bobbins have no hysteresis losses but you need many more turns for the same inductance and you’ll need heavier gage wire to keep the DC resistance low for a given number of turns.
To keep hysteresis losses low the “softest” magnetic iron should be used. Supermalloy is one of them. It has a saturation point of 15,000 gauss. The curves should be close together. The short fat squat curves are magnetically hard. Their distortion and magnetic saturation are much worse. The other major losses in iron core inductors and transformers are eddy current losses. These are internal circulating current loops induced into the iron by current in the primary windings. To minimize this transformers are made with thin lamina electrically isolated from each other often with varnish or epoxy. The windings are arranged so that the circulating currents are in the direction of the thickness of the lamina minimizing them.
WHOA! I believe I finally caught you in a terminology error. You have been the most prolific and precise poster here, so I am shocked.
Loudspeaker delay is caused by several different factors. One is that the pressure output is proportional to the acceleration of the cone. There are pressure step functions, envelope step functions and Dirac impulses in music, and not enough JERK to reproduce them, partly from voice coil power limits, and from amplifier limits and voice coil inductance, which sets the HF limit in lieu of cone breakup. The response is only as fast as you suggest with constant current drive and current feedback.
Group delay is a resonance phenomenon. Near the primary spring-mass resonance, it takes time to load and unload the resonant energy store, which has to be fed through the voice coil resistance. This causes waveform and time distortions on rectangular modulation envelopes such as evidenced by bass rhythm instruments.
I use both a conrad-johnson (c-j) preamp and phono preamp in my audio system. They discovered that certain types of capacitors improved the sound of their gear. They the determined that the less hysteresis the capacitor had the better their gear sounded. Capacitors display a loop type hysteresis much like in tape decks, except it does not go away with demagnetizing. To this end c-j worked with a capacitor manufacturer to develop proprietary capacitors that use Teflon as their dielectric and have very low hysteresis. There are, however, two downsides to these capacitors, they are expensive and they take at least 100 hours of use to break-in and sound their best.
I have the CJ Premier One, the Premier Two pre-amp and the Nuvistors CJ Pre-pre, highly modified by the factory, made in the late days of Johnson and Conrad, all of them had capacitors that have no brand name .
The sound of all of them after the modification, has nothing to do with the stock one, there is an abysmal improvement.
a·bys·mal
adjective
1. extremely bad; appalling. Interesting choice of words. 🙂
abysmal = deep, profound, unfathomable, incomprehensible.
Audiomano, I am not sure what you are saying or what you are talking about. The models you are talking about are from the very early days of conrad-johnson. To be specific the models I am comparing are as follows: In 2002 I purchased a Premier 16 Series 2 preamp and a Premier 15 Series 2 phono preamp. Neither of these had Teflon capacitors and I used these amps about 17 years until in 2019 and 2020 when I purchased ( and traded-in old amps ) a GAT Series 2 Preamp and a TEA1 Series 3 phono preamp. Both of these have the Teflon capacitors. The P16 and P15 were very good sounding amps, however, the GAT and the TEA1 are extraordinary sounding amps. The purity of the sound that comes from them is remarkable.
Indeed I am speaking with pride of models at the beginning of CJ, when they received unanimous favorable criticism for the serious audio-publications.
My CJs were modified near the end of the Johnson and Conrad era, with practically incredible results, which shows the philosophy of these two icons of the audio industry, in favor of their clients.
These units are working wonderfully well, until today.
It is possible that this time I have explained it better.
As for the Johnson and Conrad era, did you know who is in charge of CJ today?
Much better!!!
Yes, It is one of their long time employees.
His name is Jeff Fischel, I have interacted with him several times, he seems to be a real knowledgeable and good guy.
This phenomenon is termed “dielectric absorbtion” and comes in linear and non-linear varieties. The lowest DA is in oil filled, skived Teflon dieletrics used for ultra-precise scientific and military applications. 99.9% of Teflon capacitors use a cast film which has slightly more distortion.
Non-linear DA is a major source of distortion in interconnects, and a minor one in speaker cables. It also accounts for the distortion of FR4 epoxy circuit boards. It can be measured with a good impedance bridge – even my 50s era GenRad 1650 was able to predict the sound of my interconnects.
I can easily hear the sound of polyester caps and barely hear polypro in passive crossovers. To hear Teflon, you need to have everything else perfect.
Why do interconnects have more non-linear DA than speaker cables?
Also FR4 PCB’s are here there and everywhere. It is hard to imagine a system that does not have them.
(A) because the non-linear DA is in essence the formation and destruction of short to medium term electret domains, which have a finite amount of charge capacity. Speaker signals have 1,000+ times more coulombs per second, and so are far less affected. Speakers signals are predominantly affected by non-linear DA in crossover caps, so we should adopt these mitigating strategies:
1) Bi-amp and tri-amp, eliminating crossover caps. This also eliminates coil distortions, either magnetic hysteresis in ferromagnetic cores; or the power compression and loss of damping factor in air core coils.
2) Use the lowest DA caps you can afford, in order: oil filled skived Teflon, cast Teflon, oil caps (PCBs), skived Polypro, cast Polypro, oil filled paper, polyester (not recommended).
(B) Teflon and Polysulfone PCBs have lower distortion. Point to point wiring (common in tube gear) eliminates this issue. You can also use an FR4 PCB to effectuate point-to-point circuitry, which also reduces EMI/RFI interference by reducing interconnection trace length. You have to arrange the components so all junction holes are adjacent.
I hand build LM3886 based amps point-to-point on the chip package leads, for a total circuit path length of 1.5″ from input to output. There are Silver Mica caps on the input and output, and the path between those RF bypass caps is more like 1/2″ including the NFB path. This design includes GHz .33uF ceramic chip caps soldered from the power supply leads to ground lead. These techniques eliminate the need for a fully shielded enclosure. The circuit has grounded Aluminum heat sink on three sides but no metal box, and has above average RFI rejection.
Paul,
Given you sell Class D amplifiers, can you show us the distortion they produce? How do you measure it? How does it compare, using the same measurement process, to the non Class D amps you sell?