Hi thanks a lot for your valuable advice. Good, bad are subjective ... accurate is not If the aim of a system is accurate reproduction than it must reproduce accurately also test signals
The problem is to understand what makes a speaker sound bad. I think it is distortion mainly. Distortion is a very unpleasant effect.
So the lower the distortion the better. I would trade SPLs for low distortion always Actually i have to limit SPLs at home for the neighbours.
But if the response is quite full range (for reproducing all instruments) and low in distortion i think that the musical experience can be very pleasant
Yes but when a speaker sounds right i hear that immediately. Like a lightning strike Maybe it is not completely right (i.e. limited in range and SPLs)
As an example a friend of mine put an old Sansui au6900 on the table and above it a pair of small vintage Grunding boxes
Immediately the speakers put out a very clear 3D soundstage in the room. I bought that Sansui immediately. My rule is if a chain sounds good also the elements of the chain must be good. He was using a cheap cd player. Another impressive chain had a 80 euro Samsung dvd player a Revox integrated and a pair of Quad esl63 A superb realism I bought the Samsung only ... the Quad are way too big for my room I would love to get that Revox
I agree The accuracy must come from measurements From testing in specific conditions Like for any other device
I would be willing to choose a device more on how it performs on a bench than on the basis of a listening test. And anyway a listening test refers to a complete chain But when noise and distortion are very low i think that the sound can be also very good.
An accurate speaker cannot sound bad unless the recording is bad of course
Accurate imho translates in flat FR and low distortion at the listening SPLs.
No speaker can ever reproduce a square wave.
We all know what happens when we feed a square wave to a speaker. Power dissipation goes way up. Hard clipping is like a square wave. Woofers can be overheated and bottom out.
When you throw an amplifier into the mix, clipping artifacts can degenerate into ringing, worst case scenario. There's a reason we use a square wave to test an amplifier: they bring out the worst in them. Square waves can destroy a loudspeaker.
We all know what happens when we feed a square wave to a speaker. Power dissipation goes way up. Hard clipping is like a square wave. Woofers can be overheated and bottom out.
When you throw an amplifier into the mix, clipping artifacts can degenerate into ringing, worst case scenario. There's a reason we use a square wave to test an amplifier: they bring out the worst in them. Square waves can destroy a loudspeaker.
First of all this statement is just plain wrong. It depends on the phase response of the loudspeaker and HF extension to some degree. Also depends on the frequency of the square wave, since it needs a bunch of higher order harmonics of the fundamental period to make up the sharp edge of the waveform, so if you try to reproduce say a 10kHz square wave there just isn't enough bandwidth to do it. But a couple hundred Hertz can be reproduced quite cleanly by the right loudspeaker. See for example:
https://audioxpress.com/article/a-loudspeaker-that-can-play-square-waves
Anyway, this "can't reproduce a square wave" trope is very tired at this point. The ear and brain are rather insensitive to the relative phase of various components of a signal. You can look up data about this topic by searching under group delay audibility. For example, at 1kHz it takes about 1 millisecond of group delay peaking to have any impact on the perceived sound, even for sharp and short synthesized click-like sounds and listening through headphones. A loudspekaer in a room moves this up to several msec. At 1kHz the period of that sine wave component is 1 msec, so you are talking about an entire period. So you can assemble a square wave from very "off" sine components and it will still sound the same, because that is how the brain works. There is no need to actually reproduce the square wave, just it's harmonic components at the right amplitude and at a sort-of-close playback timing. The same thing applies to music signals, but it turns out that the ear+brain is very forgiving when listening to music via loudspeakers in a room.
Apparently not always true. For one example where relative phase is quite audible: https://purifi-audio.com/2019/12/07/amfm/
Please note that I did not say you cannot hear it, ever - of course you can when you make the distortion large enough. One can always generate examples of large defects that will be audible. Also, the Purifi Audio example is of Doppler distortion, which is not the same thing as phase distortion. Phase distortion is a completely linear process that does not produce additional frequencies. I hestitate to even call it "distortion" because it does not really fit the definition. But that is the commonly used term, so 🤷.
I believe my carefully chosen wording was "relatively insensitive". When I mean by this is that even when you are looking at a waveform that seems to be grossly corrupted by phase distortion, to the ear it will likely still sound pure. If you look at Figure 2 the link in the audioexpress article above you will see what I mean by this, and even this is a contrived example with a frequency response peak that is obviously a bad design. The waveform is supposed to look like a square wave - it looks awful but it will sound very much like a square wave until the phase distortion gets even worse than this. And for music signals you have to go even further, so far that such corruption is not typically found in a loudspeaker, at least a well designed one...
No new frequencies were created, so Doppler distortion is technically a linear distortion. Phase distortion is a type of linear distortion.
Its like this: Phase is a property of frequency, not of time. A signal in the time domain has only one parameter, amplitude. Thus when we speak of phase we are looking at the signal from the frequency domain perspective. In the time domain the signal waveform will always change shape if phase of the signal is changed in the frequency domain. It cannot be otherwise. Moreover, the example audio files where not created by doppler distortion, they were created by adjustment of phase in the frequency domain then an inverse FFT was performed to convert back to the time domain. Therefore it is correct to say the two waveforms vary only in phase, where the term 'phase' simply implies use of the frequency domain view of the signals.
You might want to carefully re-read the Purifi tech note because it seems to directly contradicts your statement above. Quoting from it:
IMD creates new tones both above and below the fundamental. It is those tones that Purifi has ALREADY created in their example (the spectrum in the upper figure), which they then further modify to create the AM and FM examples (both create IMD).
Also, they said:
This is why I said that phase "distortion" is really not distortion. It is just changing WHEN the components of a waveform are produced from a process (like a loudspeaker, a filter, etc.) but does not "cause both harmonic and intermodulation distortion" (which both create new component(s) at different frequency(ies) than the original one(s). Maybe I should have used the term "time delay distortion" a la Heyser to be more clear.
The other important difference between Doppler distortion and what I am calling "phase distortion" is that in Doppler distortion the phase is constantly changing, e.g. with the cone position - it is "modulated" by it, whereas phase distortion that arises from the crossover filters and driver phase responses is constant in time. These are very different phenomena, which makes one much more audible than the other.
Last edited:
To say that those waveforms vary only in phase is incorrect as that phrase refers to time domain.
Again, you are wrong here. Phase is not "part of frequency" - maybe you meant it is part of the frequency DOMAIN? Phase is to the frequency domain as time is to the time domain, they both describe when some signal energy is appearing from the process. This is why causal signals can be equally represented in the frequency domain or the time domain, and you can go back and forth between these domains without losing information.
In the frequency domain, when expressed in polar form, Magnitude and Phase are both a function of frequency. The same complex number can be expressed in rectangular coordinates as the magnitudes of the Cosine and Sine components for each frequency.
In the time domain Amplitude is a function of time. Period.
In the above sense Phase is not a property of the time domain view of a signal.
In the time domain Amplitude is a function of time. Period.
In the above sense Phase is not a property of the time domain view of a signal.
Last edited:
OK I'll see you that and raise you this:
In the time domain a signal can be expressed in terms of its frequency and phase components like this:
x(t) = A_o + SUM{ A_n * sin(2nft + phi_n) } n=0..infinity
where "phi_n" is the phase angle of component n and A_n its magnitude. This is just the Fourier series representation of a time domain signal.
So using your logic, wouldn't that make phase a part of the time domain representation of a signal?
Your turn.
In the time domain a signal can be expressed in terms of its frequency and phase components like this:
x(t) = A_o + SUM{ A_n * sin(2nft + phi_n) } n=0..infinity
where "phi_n" is the phase angle of component n and A_n its magnitude. This is just the Fourier series representation of a time domain signal.
So using your logic, wouldn't that make phase a part of the time domain representation of a signal?
Your turn.
Nothing can ever reproduce a DC-to-Light square wave.
Nothing has infinite treble response. The corners are always blunt.
Many things lack DC response. The tops are often slanted.
Reproducing an approximation of a square wave is not always fatal to a speaker. Did it ALL the time back in the synth lab. (Real ARP 2500, not miniMog or Odyssey). I think it was the mix of drugs and squares that burned speakers.
As a freshman at university, I obtained access to a Large Moog with tape decks, ribbon controller, frequency variable AC source, Dynaco 400W amplifier, KEFs with the "racetrack" woofer / passive radiator. Out of all those bits, guess which one(s) were already blown, by the time I got to use it? If you guessed the tweeters in the KEFs, you'd be correct.
Regarding phase of constituent components of a square wave, one time I made (in LabVIEW software...) a graphic EQ with both amplitude and phase controls. You could "roll" the fundamental back and forth in time - see it moving underneath the square wave - as displayed on an oscilloscope. Adjusting that and higher harmonics relative phase did things like "tip the top" toward or away from the left edge of the screen. And all kinds of other mal-formations.
Why? At my work, they wanted to see what the current was (hard to measure), by measuring voltage (easy to measure) and supposedly knowing well the impedance, mag and phase. Always came out insufficient, nonsense. So I thought I'd just dial it in by hand, by making an arbitrary mag and phase processor. It worked well enough for the first few harmonics, but after that... Kinda like a polynomial fits a curve - for a while.
In audio, you have multi-band graphic EQ, after multi-band graphic EQ - and no one thought of putting "phase" as a second set of corresponding sliders. So easy to do too, but, the relative phase of any given band doesnt matter, right? The ear cant hear it, so why bother? Anyhoo, I wonder what it'd be like to use such a thing on music?
Hi ! i think the idea is to throw out something that looks quite like the input signal even if this is a SW
For example, some speakers do quite well
https://wonki.files.wordpress.com/2006/12/rehdeko-phase.jpg
http://www.stereophile.com/images/archivesart/QUADFIG6.jpg
i guess that some signal tests not present in nature can provide some information about the dynamic behaviour of a transducer ?
Imho distortion measurements at various Hz and SPLs and the usual FR should be enough
Distortion is something added not present in the input signal It is a sort of processing of the input signal
Of all the things, nah it ain't distortion. Not anymore these days.
On the list of priorities, it's one of the least important aspects these days.
On the list of subjective anxieties, distortion anxiety is in the top 3.
More importantly, a speaker will be eventually a lowpass filter, so it's fundamentally never going to be a square wave.
Since a perfect square wave has an infinite amount of odd harmonics.
I just don't see how this even has anything to do with the importance of measuring things.
As well as the rest of this topic tbh.
A couple of thoughts:
People want to use square waves for the same reason they want use steady state HD FFTs, because its easy to visually interpret if it looks ideal.
As a practical matter, square waves might be an approximation of LF transients where there is evidence humans are sensitive to phase. If we want to kick drum to sound and feel like a real one, it may help if the transient isn't too smeared out over time by more or less linear distortions.
People want to use square waves for the same reason they want use steady state HD FFTs, because its easy to visually interpret if it looks ideal.
As a practical matter, square waves might be an approximation of LF transients where there is evidence humans are sensitive to phase. If we want to kick drum to sound and feel like a real one, it may help if the transient isn't too smeared out over time by more or less linear distortions.
Hi you must be very right because i am seeing lab reports of commercial speakers with 3-4% distortion at 90dB/1m
Maybe people like distortion