Jan,
Yet not a single amplifier in the world does that.
And once the amplifier distorts, then:
The critical point here is: "below the threshold of the ear to hear a difference."
And this threshold varies with SPL, frequency and order of the distortion products, both HD and IMD.
The very fact that you still cite THD shows that the whole argument was not understood. THD is not capable of giving information about distortion audibility.
In fact, it is trivial to construct a system that distorts a signal to give 5% THD at maximum level and which is judged as undistorted by listeners (Geddes for example did so) and it is equally trivial to construct a system that gives 0.01% THD at maximum level and which is judged by listeners as extremely distorted in an objectionable manner (Geddes did something similar).
So low THD does not ensure low audible distortion and high THD does not make for high audible distortion, hence THD is meaningless.
Again it is highly relevant.
Let's make this crystal clear.
As long as the distortion in the recording and playback system is substantially lower at all harmonics and resulting IMD products and with the correct phase AND of a pattern similar to the ears, it will not be audible...
Ciao T
Yet not a single amplifier in the world does that.
And once the amplifier distorts, then:
The critical point here is: "below the threshold of the ear to hear a difference."
And this threshold varies with SPL, frequency and order of the distortion products, both HD and IMD.
The very fact that you still cite THD shows that the whole argument was not understood. THD is not capable of giving information about distortion audibility.
In fact, it is trivial to construct a system that distorts a signal to give 5% THD at maximum level and which is judged as undistorted by listeners (Geddes for example did so) and it is equally trivial to construct a system that gives 0.01% THD at maximum level and which is judged by listeners as extremely distorted in an objectionable manner (Geddes did something similar).
So low THD does not ensure low audible distortion and high THD does not make for high audible distortion, hence THD is meaningless.
Again it is highly relevant.
Let's make this crystal clear.
As long as the distortion in the recording and playback system is substantially lower at all harmonics and resulting IMD products and with the correct phase AND of a pattern similar to the ears, it will not be audible...
Ciao T
Thorsten, you really go out on a limb here. This post really shows you don't know your a from your e. Flat frequency response is one of the most important design goals for loudspeakers and has been recognized as such for a long time, at least by the two companies I have been working with. Clear design goals work. Some of the better speakers easily stay in between +/- 2dB nowadays. No averaging, no tricks.
Your wholesale dismissal of baffle step compensation on the notion that this equates to a 6dB LF boost shows a gap in your knowledge. Read up on the subject, please, so that you understand the physics behind it. And look at how it is implemented by knowledgeable companies. 6dB really is the outer limit, usually it is much less. Yet, you often need it to compensate for Mother Nature at work.
What really disappoints me though, and I made this statement before, is that high end speaker manufacturers almost never show the FR of their products, nor do they publish impedence graphs. Incomprehensible for such expensive technical products that can easily measured. You are no exception.
For good measure, please find below the sort of FR, unsmoothed, not averaged, good design can accomplish with average quality components.
Everything below 400 hz needs to be taken with a grain of salt, they haven't gone out for anaechoïc measurement yet.
vac
Thorsten,
"A pattern similar to the ears" You should refrain from statements like that if you can't even give a substantial response to my post #14878.
vac
Quote
"So low THD does not ensure low audible distortion and high THD does not make for high audible distortion, hence THD is meaningless."
With all due respect, this is nonsense.
It is right and proper to aspire to a reproduction signal chain that produces ZERO distortion. Once commted to tape, or whatever storage medium you care to name, by the artist and recording engineer, the reporduction signal chain ideally should have NO influence on the signal coming off the source. And,by ZERO distortion, we mean any kind you want to name - IMD, PMD, Transient etc etc. There's a whole community working on this and hoping they can get there one day.
The fact that designers **** this concept up with argumants against feedback, for feedback, for tubes and againt tubes, or use techno babble that the ear can hear -100dB is the sad bit.
"So low THD does not ensure low audible distortion and high THD does not make for high audible distortion, hence THD is meaningless."
With all due respect, this is nonsense.
It is right and proper to aspire to a reproduction signal chain that produces ZERO distortion. Once commted to tape, or whatever storage medium you care to name, by the artist and recording engineer, the reporduction signal chain ideally should have NO influence on the signal coming off the source. And,by ZERO distortion, we mean any kind you want to name - IMD, PMD, Transient etc etc. There's a whole community working on this and hoping they can get there one day.
The fact that designers **** this concept up with argumants against feedback, for feedback, for tubes and againt tubes, or use techno babble that the ear can hear -100dB is the sad bit.
Last edited:
Hi,
I have yet to measure anything like that.
Maybe we use different measurement methods.
In fact, I find it difficult to get results that are sufficiently repeatable with better than +/-1dB (unsmoothed and without averaging) even with a fair bit of attention paid to orientation, microphone positioning etc. using the tools I have.
I said it produces a 6dB LF boost in the power response of the speaker (which it does). Normal rooms produce a resultant frequency response that is significantly boosted at low frequencies if they have sufficiently stiff walls. This is actually easy to measure as well.
So we are back to bending the response of what should be a flat speaker to match what the ear wants to hear?
First, I personally am not a high end speaker manufacturer, nor do I control any policy of what such a company chooses to release or not.
Secondly, given that even small shifts in microphone position tend to cause substantial changes in frequency response (without smoothing) the whole "frequency response" thing is yet another "how long is a piece of string".
Good for you. I never get anything like such results unless I use smoothing, so clearly I am doing something wrong. Could you elaborate on the measurement process that produced these results without?
Ciao T
I have yet to measure anything like that.
Maybe we use different measurement methods.
In fact, I find it difficult to get results that are sufficiently repeatable with better than +/-1dB (unsmoothed and without averaging) even with a fair bit of attention paid to orientation, microphone positioning etc. using the tools I have.
I said it produces a 6dB LF boost in the power response of the speaker (which it does). Normal rooms produce a resultant frequency response that is significantly boosted at low frequencies if they have sufficiently stiff walls. This is actually easy to measure as well.
So we are back to bending the response of what should be a flat speaker to match what the ear wants to hear?
First, I personally am not a high end speaker manufacturer, nor do I control any policy of what such a company chooses to release or not.
Secondly, given that even small shifts in microphone position tend to cause substantial changes in frequency response (without smoothing) the whole "frequency response" thing is yet another "how long is a piece of string".
Good for you. I never get anything like such results unless I use smoothing, so clearly I am doing something wrong. Could you elaborate on the measurement process that produced these results without?
Ciao T
How many points on your FFT? How many averages? Are you exciting with an oscillator or pink noise?
The old fashioned pen recorders had adjustable damping to smooth the plots. One of Harry Olson's secrets to getting good plots was to set the damping at maximum.
The example you show is not = +/- 2 db and 10 db variation can also be expressed as +/- 5.
Whenever I have looked at a loudspeaker even a single driver requires smoothing to get useful curves. Throwing pink noise at a driver and then looking at many averages to smooth things out gives you much the same effect as turning up the damping. That allows you to see things you can fix, not a forest of trees.
T, we cross posted!
This kind of aspiration is useless, since even a single resistor, as well as any part of the circuit, including isolated wire, produces MEASURABLE distortion.
And you are right, speaking about a need to MINIMIZE a wide spectrum of distortions.
All the boomy reaction of many people in this thread, just caused by the popular practice to play with THD, everywhere, at any corner, at any thread, by using a multitude of simulation tools. An opposite camp just trying to say, stop cheeting less experienced buyers and DIYers, audio reproduction is mainly not a THD, and the big love to THD simply precludes from better understanding of a real situation, that is contradictory and compromised, where improving one parameter one almost enevitably make others worse. The position, all kind of distortion should be nulled, is obviously impossible. Even by increasing a number of stages in the circuit, we inevitably accumulate distortions of passive parts, this is one of the sides of inherent contradictions.
Last edited:
Actually THD has an interesting characteristic, if you stack two inverting stages of the same gain and distortion some of the products will cancel!
Obviously that doesn't work for IMD!
ES
Sorry, but You are the one and only talking simplifying only about THD
.
Nonsense..
Try to think about real time difference input/output, this can catch all "deformation" of signal. No difference in signal (90-100dB under useful signal)= no audible differencies. So simple to understand. This is real situation, I tried many times.
But something as in attached picture ("audiophile", highly valued component ..) has clearly audible "footprint".
Attachments
Last edited:
Thorsten:
Time windowed pulsed sine, 250 sweep points (25 per octave). S&L measuring equipment does not do smoothening. As you can see on the tail end, it gets ragged (diffraction effects round tweeter, solved by now).
When small shifts in microphone position cause substantial changes in FR, there are either diffraction or xover issues. What I show are 3 measurements at 0-45-90. No wild swings.
On baffle step compensation, you got me wrong. Mother Nature at work is not what the ear wants to hear, but how sound waves behave emanating from a speaker mounted in a baffle. When the frontal baffle area can no longer support the longer wavelength frequencies, the baffle moves out of its piston band. Acoustic radiation makes a transition from half space to full space. Net loss of low frequencies as a result. So, in order to get a flat speaker, you need BSC because of the way Mother Nature deals with acoustic loading.
Time windowed pulsed sine, 250 sweep points (25 per octave). S&L measuring equipment does not do smoothening. As you can see on the tail end, it gets ragged (diffraction effects round tweeter, solved by now).
When small shifts in microphone position cause substantial changes in FR, there are either diffraction or xover issues. What I show are 3 measurements at 0-45-90. No wild swings.
On baffle step compensation, you got me wrong. Mother Nature at work is not what the ear wants to hear, but how sound waves behave emanating from a speaker mounted in a baffle. When the frontal baffle area can no longer support the longer wavelength frequencies, the baffle moves out of its piston band. Acoustic radiation makes a transition from half space to full space. Net loss of low frequencies as a result. So, in order to get a flat speaker, you need BSC because of the way Mother Nature deals with acoustic loading.
Agree, this example is more like +/- 2.5 dB, but has flattened since. Look at the scale, it is 5dB in stead of more usual 10dB. Nothing averaged or dampened, pulsed sign sweep, repeatable, gradual roll off in horizontal plane.
Smoothening doesn't make any sense in a development trajectory. A FR is like putting a speaker under a microscope, and the more resolution, the better. Peaks and throughs point in the direction of issues that have to be resolved, like the one I posted (which was midpoint in the project I have on hand right now). Above 7Khz it gets rough. Cause is diffraction of tweeter surround/mount. Instrument maker just finished brass inserts that should improve situation. With smoothening, this might have escaped attention.
vac
Since all amps distort none will ever be perfect, and let's not focus just on amps include the entire reproduction chain. The issue in design is what trade offs to make. Prioritizing those issues is the essence of design.
Unless you are maintaining that all distortion is the 'same', then what distortion you focus on minimizing matters. What trade offs does one make.
The goal is to enjoy reproduced music, not to have bragging rights on parameters that may or may not be important.
Solving one problem often creates another problem.
It doesn't need to be perfect. It only needs to be good enough so that listeners cannot hear a difference between the input signal and the output signal.
I maintain that such amplifiers exist today.
jan didden
- Status
- Not open for further replies.