Square Wave Testing of Audio Products (Video Tutorial)

[Plcamp]

I really don’t see what squarewave testing (as a characterization of amplifier performance) adds to the equation beyond what we see in already performed testing.

But If I am missing something, please advise!

 

[fpitas]

I really don’t see what squarewave testing (as a characterization of amplifier performance) adds to the equation beyond what we see in already performed testing.

But If I am missing something, please advise!

You mentioned stability testing. I use square waves as you describe, but in this case the square wave probably won't reach the amplifier input, being filtered first.

It is a fact (and a dirty little secret) that different amps have different tolerances to reactive loads; the speaker, the unterminated cable etc.

 

[Lambda]

(as a characterization of amplifier performance) adds to the equation beyond what we see in already performed testing.

Can you Accuracy predict/calculate/draw how the squarewave from an amplifier looks only by/from the "ASR standards" performed testing?

 

[Plcamp]

You mentioned stability testing. I use square waves as you describe, but in this case the square wave probably won't reach the amplifier input, being filtered first.

Anything that affects the frequency response will affect the feedback system response, and that includes any filters applied at the signal input or output. You might see 12 to 20 turn coils in series with the output of class ab amps like Haflers, they are limiting the hf open loop response (explicitly to prevent the whole thing from becoming a hf oscillator).

The difference between an oscillator and an amplifier can be slight!

If, for example, the feedback loop of an amplifier had an inadequate stability margin within the frequency band of its intended range, I would expect its distortion performance would reveal that.

I was simply using loop stability of an example of where squarewave testing is actually useful to a designer on a bench during development. At the same time, there is zero chance my peer group would have even considered the release permission of my ultimate design, without both a paper analysis of the frequency sweep expectation of a Bode plot analysis (gain vs phase vs frequency analytic sweep…we had to do that on paper at the time) combined with a frequency swept feedback loop measurement of gain vs phase that agreed.

We used square wave load testing to determine whether or not it was even worth the time to measure the feedback loop…it was the coarse quick try and not the acceptance test.

 

[Plcamp]

Can you Accuracy predict/calculate/draw how the squarewave from an amplifier looks only by/from the "ASR standards" performed testing?

An envelope within which the actual measured result resides could be drawn, I believe.

But equal to that…you might need to answer why you believe where in that envelope it measures…matters?

 

[fpitas]

Anything that affects the frequency response will affect the feedback system response, and that includes any filters applied at the signal input or output. You might see 12 to 20 turn coils in series with the output of class ab amps like Haflers, they are limiting the hf open loop response (explicitly to prevent the whole thing from becoming a hf oscillator).

The difference between an oscillator and an amplifier can be slight!

If, for example, the feedback loop of an amplifier had an inadequate stability margin within the frequency band of its intended range, I would expect its distortion performance would reveal that.

I was simply using loop stability of an example of where squarewave testing is actually useful to a designer on a bench during development. At the same time, there is zero chance my peer group would have even considered the release permission of my ultimate design, without both a paper analysis of the frequency sweep expectation of a Bode plot analysis (gain vs phase vs frequency analytic sweep…we had to do that on paper at the time) combined with a frequency swept feedback loop measurement of gain vs phase that agreed.

We used square wave load testing to determine whether or not it was even worth the time to measure the feedback loop…it was the coarse quick try and not the acceptance test.

Yes, most if not all amps have a "Zobel" at the output to keep the followers (Bipolar or FET) from oscillating at VHF. Still, amps do behave differently feeding reactive loads.

 

[Plcamp]

Still, amps do behave differently feeding reactive loads.

Now that is one I do believe matters. And that isn’t adequately tested.

Because speakers are always reactive. The tolerance to that is something I instinctively believe needs something new.

Edit: I think for example one big benefit of dsp vs post amp crossovers is the ability to eliminate a whole class of energy storage from the load. Two amp channels easily driving a nominal 8 ohm load with a minimum 6.8 ohms is going to be less stressful to any amp channel driving minimum < 3.2 ohms (incl passive xover) if the two drivers in parallel.

I can see how amps might be differentiated by testing what phase angle they can tolerate before exceeding a given threshold of claimed distortion/power.

 

[fpitas]

Now that is one I do believe matters. And that isn’t adequately tested.

I've seen a gizmo that tests audio power amps at all possible phase angles. Let me see if I can Google it.

And the followers oscillating part is true, too. Even the Radiotron Designer's Handbook has that warning lol

 

[ta240]

Looking at the conclusions section again after reading the article yesterday actually does not tell me that there is anything gained by adding square wave testing to the regimen already conducted by Amir in his characterization tests. In fact, it reinforces that the sine wave test are indeed representative and more revealing.

There has never (to my knowledge) been any debate that square wave observations by designers experimenting with prototypes are a useful thing to do from time to time (I could tell instantly whether or not my SMPS feedback loop was as-expected stable with one look at minimum load and maximum line in voltage)…so yes it is a useful test at some stages of any design process.

But I don’t see anything in that article that suggests square wave testing is a must do to enable valid comparisons of performance, given the tests already performed?

I don’t get what the objection is?

My question was simply that if square waves can, in a glance, show us things that would indicate that amps would sound different then why aren't they used.

If all amps sound the same then the square waves should look the same, correct?
Yet sites that show square waves show that they almost never look the same.

How can different rise times, different roll-offs and different amounts of ringing sound the same?

 

[Plcamp]

My question was simply that if square waves can, in a glance, show us things that would indicate that amps would sound different then why aren't they used.

If all amps sound the same then the square waves should look the same, correct?
Yet sites that show square waves show that they almost never look the same.

One tells you there is a difference, but nothing about the degree (nor whether or not you would hear it). The other (sweep tests) quantifies the degree to insane precision?

That’s what I believe.

Edit: To be more precise I believe any in-audible- band differences that might be revealed with a squarewave output test would need to be orders of magnitude higher severity than what would have already been revealed with a much less severe problem in the already performed frequency vs distortion and linearity tests already performed.

 

[Lambda]

An envelope within which the actual measured result resides could be drawn, I believe.

Only if we make a lot of assumptions. and and granted we can make a lot of assumptions for most amps.
But we don't know how they change the Phase and therefore the ordinal signal can't be reconstructed.

 

[fpitas]

Only if we make a lot of assumptions. and and granted we can make a lot of assumptions for most amps.
But we don't know how they change the Phase and therefore the ordinal signal can't be reconstructed.

Why do we think the phase is mysterious? Doesn't it follow the Hilbert transform of the amplitude response, like usual? Are they not minimum phase somehow?

 

[Lambda]

Why do we think the phase is mysterious? Doesn't it follow the Hilbert transform of the amplitude response, like usual? Are they not minimum phase somehow?

Most Amps (without DSP) are probably minimum phase. but we don’t know
As i sad:

granted we can make a lot of assumptions for most amps.

But if the amps Gets into some sort of slew rate limiting and feedback goes down it might change.
This is also why some amps show strange change in distortion over frequency.

The total Frequency response might be relatively flat but internal feedback (and phases) might change over frequency and is maybe be Power or slew rate depended.

 

[fpitas]

Most Amps (without DSP) are probably minimum phase. but we don’t know
As i sad:


But if the amps Gets into some sort of slew rate limiting and feedback goes down it might change.
This is also why some amps show strange change in distortion over frequency.

The total Frequency response might be relatively flat but internal feedback (and phases) might change over frequency and is maybe be Power or slew rate depended.

Well, conceivably. Although the ear isn't real sensitive to phase.

 

[HP9000]

When a discussion about the square wave is started it is common for someone say "an ideal square wave cannot exist because it would require infinite bandwidth", but this is not a reason to not discuss the square wave in general when an ideal sine wave seems also unable to exist given it would require zero harmonic distortion. Every sine wave test or generator I have seen has harmonic distortion. One can say that "it doesn't matter because in quasi form it is acceptable", and this is just the point. One can say that a square wave is physically impossible given it supposes instantaneity, solidifying the idea of it being unable to exist, but an ideal sawtooth wave also supposes instantaneity, and yet I don't hear people saying we shouldn't consider it when made quasi.

I was notified about a new post in this thread, however it seems it has disappeared?

I wouldn't be surprised if it was deleted by someone. In another thread I tried to carry a conversation started by user @TopG who seems to have been deleted, and I was banned. (Pg. 71) Part of their comment was about the M-Scaler, but most of their comment was showing a general concern for time response and asked why these sorts of measurements are excluded. Amir replied dismissively, and I replied to Amir (pg. 73) asking the same question.
https://www.audiosciencereview.com/...chord-m-scaler-review-upsampler.35498/page-71

Why does the square wave test not matter?

"Frequency and phase measurements suffice since they can be converted into the impulse response."
This doesn't mean that a linear frequency response with a given phase is going to convert to a linear - meaning compared input and output - square wave or impulse response when all that the frequency response requires to be called linear is 20hz - 20khz.

"It doesn’t need any improving. We cannot detect changes in amplitude with a speed that exceeds the speed of the wavelength of a 20kHz* sound."
A lot of the metrics for a lot of devices can be said to not need improving and yet they are included, so this can't be a reason.

"Absolute phase doesn't matter at high frequencies. A 12 kHz sine wave sounds exactly the same as 12 kHz square wave"
I talked about the ear/brain's limit to sensing amplitude differences in time. A 10us pulse can stimulate the ear. A 10us delay between pulses acts as another way to use energy to demonstrate the brain's sensitivity to 10us. That we can sense 10us would be irrelevant if it couldn't be differentiated from something lower, which would be to say that it is above the limit and just being processed the same as the limit, making something lower the limit instead, but we can differentiate:

"The upper edge of human hearing is 20 kHz so it is unlikely that we could detect a difference between a 25 us impulse and a 10 us impulse (or even a 1 us impulse), but assuming sufficient energy, all would be detectable. Making really short sounds with sufficient levels of energy is technically challenging and requires good transducers, but ultrasonic transducers exist."
From: https://psychology.stackexchange.co...re-the-temporal-limits-of-the-auditory-system
"Observers were asked to discriminate between a pair of 10‐μsec pulses and a single 20‐μsec pulse having the same total energy. The independent variable was the time, ΔT, between the two 10‐μsec pulses. The stimuli were also presented as elements in a periodic pulse train. The ΔT required for resolution of two clicks (two‐click threshold) was 10 μsec."
From: https://asa.scitation.org/doi/abs/10.1121/1.1912374

I explained in a post about the role played by the ear's impedance (read lowpass filter - the reason why we even have a tonal limit). We can imagine that the reason why pulses are working here is because they have enough energy. We can ask why an ultrasonic tone at high amplitude would still be inaudible. This is because right after the positive or negative half wave (half the time of the already-ultrasonic frequency, or twice the frequency) will an opposing polarity/pressure of equal energy will occur, effectively cancelling out the first. This is not solved by increasing the amplitude further. There is not enough time between the rise and fall of one polarity pulse and another, given they are equally powerful, and given the transfer of energy is kinetic. If you lower the amplitude of one halfwave it will start to become audible again, only as a pulse. There is reason to think that we can sense rise and fall times higher than 1/20000th of a sec if they are distanced from each other in time i.e. if they are working around lower frequencies (think upper-mid range/midrange/bass). Impedance does not set an equal tonal and temporal limit for the ear.

"There is no shortest sound detectable by the human ear. Any impulse with enough energy is audible.
The shortest detectable tone, identifiable as a tone, would be on the order of 100 ms. It might be shorter for tones of high pitch."
From: https://sound.stackexchange.com/que...e-shortest-sound-perceptible-to-the-human-ear
"A click, being an impulse signal, basically stimulates the entire cochlea (the inner ear), and hence represents the most powerful auditory stimulus. Pure-tone bursts are, as a consequence, weak stimuli as they stimulate a small proportion of the available hair cells. High sound levels may therefore be necessary for short tonal stimuli, in turn degrading the frequency resolution in the cochlea as excitation spreads away from the characteristic frequencies. Mostly, several milliseconds are used as tone bursts."
"By analyzing this synapse using very short tone bursts, it was shown that the vesicles of glutamate (the neurotransmitter activating the auditory nerve fibers that lead the signal to the brain) were released as shortly as 2 ms after tone onset. Tone bursts were 8 - 16 kHz, 10 ms plateau, 1 ms rise/fall (Khimich, 2005). Note the tone burstduration and rise fall times mentioned in answer PART I."
"The problem is that the experimental data do not show that a spike wouldn't have been generated at shorter stimuli. The data merely show that with an ongoing stimulus the first spike appears at 2 ms. It does not show whether 2 ms of temporal summation is necessary, and that is why I have left part I of my answer in place, as a theoretical lower bound."
From: https://biology.stackexchange.com/questions/27662/what-is-the-human-ears-temporal-resolution

~

Let's say we forget about the limit of our time sensitivity because the common device options as they are work so far under that limit*. This still doesn't mean there isn't room for improvement.

An amp can always be improved by being a high current/output impedance amp when powering an electromagnetic coil type speaker. Voltage amps are said to sound as having a tight bass (good) albeit altogether dry sound (bad). Current amps are said to sound as having a fast bass (good) and altogether clear sound (good), despite that they have been used with the commonly available high Q speakers designed around voltage amps. The reason is because voltage amps provide damping factor for the bass, but allow the inductance (think about how high the inductance can be at 20khz*…) to affect the rise time of all frequencies, including the bass. Current amps allow the inductance to affect the magnitude of the frequencies, but less so the rise time. While you might choose a low inductance/low impedance driver for the sake of impedance/gain matching, the benefit isn't being derived from the driver. Measurement comparisons of the acoustic output between a speaker being powered by a current vs a voltage amplifier show the voltage amp amounting to higher intermodulation: http://pmacura.cz/speaker_dist.htm
Impulse response (post #504): https://www.diyaudio.com/community/...sconductance-current-amplifier.239321/page-26
Given that the distortion is due to the amp-speaker combo and not innately either, and given that the magnitude nonlinearities, ringing, and noise/gain mismatch of current amps is due to the driver design, can you have all of the benefits and none of the cons by choosing a current amp and a driver designed for low Q and low impedance (a low impedance driver being more likely to be resistive and near linear).
You can lower the effect of the inductance on rise time by lowering the inductance of the driver. One can think that this addresses the problem alone, and that they can go ahead and use it with a voltage amp. The problem now is that the voltage amp will not have much a of a load, and will be sent into high power, where slew rate is a problem. Now you're back to square one time distortion-wise. Pairing a current amp with a low impedance driver provides gain matching and doubles down on lowering the effect of inductance.

It's common for people from this site to say that "there is no audible difference" regarding anything. Sometimes it's said "we can't hear a difference", this is speaking on other's behalf, which you can't do. You can say that those people are trying to justify their purchase in their mind or toward a brand they affiliate with, but there are many cases where someone hears something someone they know bought, at an event, or that they were expecting not sound good. It isn't enough when I'm going to wonder why someone else said they did hear a difference. The discrepancy demands a test be taken to put it to rest. We might dismiss a simple "it sounds better", but someone saying "it sounds fast" has more meaning, even though implicitly it says better because slow would be worse, but with a word like that we have something to go off of for comparing measurements or conducting a test experiment.

Take the Rehdeko speaker, there are square wave tests, and there are many reports saying it sounds fast. You can say it's the distortion that is making it sound fast. If this were true you would be able to test it. You can take the all of the other measurements for reverse engineering a speaker so that measures as closely as possible in all respects (possible given that there are thresholds in every respect) except the square wave. Blind test people with both.
Class d amps. Take a linear working transistor amp, get it to measure the same as the class d amp in all areas but the impulse through some combination of keeping metrics below audible threshold or artificially raising them to match.
Delta-sigma dacs. These can measure well in literally every respect except their impulse response. Design a resistor ladder dac with no form of oversampling and with an analogue lowpass filter well beyond the audible band only for the sake of assuring it's bandwidth is aligned with or lower than the proceeding devices and/or raise the bandwidth of the proceeding devices. Get the other figures to be either below audibility or artificially raise the distortion of the delta-sigma dac to match. Play natively recorded high sample rate music through both.
The key here is that there are control variables, otherwise it's not scientific.

Wiki class d quantization errors: "While some class-D amplifiers may indeed be controlled by digital circuits or include digital signal processing devices, the power stage deals with voltage and current as a function of non-quantized time. The smallest amount of noise, timing uncertainty, voltage ripple or any other non-ideality immediately results in an irreversible change of the output signal. The same errors in a digital system will only lead to incorrect results when they become so large that a signal representing a digit is distorted beyond recognition."
From: https://en.wikipedia.org/wiki/Class-D_amplifier
Now consider the system: Delta-sigma dac, class d high voltage output (indicated given there are high current output class d's to compare, and people said they like them better), high Q speakers. I am skeptical that this is 'the end' when they can measure well in all ways except their impulse response/square.

 

[restorer-john]

When a discussion about the square wave is started it is common for someone say "an ideal square wave cannot exist because it would require infinite bandwidth", but this is not a reason to not discuss the square wave in general when an ideal sine wave seems also unable to exist given it would require zero harmonic distortion. Every sine wave test or generator I have seen has harmonic distortion. One can say that "it doesn't matter because in quasi form it is acceptable", and this is just the point. One can say that a square wave is physically impossible given it supposes instantaneity, solidifying the idea of it being unable to exist, but an ideal sawtooth wave also supposes instantaneity, and yet I don't hear people saying we shouldn't consider it when made quasi.


I wouldn't be surprised if it was deleted by someone. In another thread I tried to carry a conversation started by user @TopG who seems to have been deleted, and I was banned. (Pg. 71) Part of their comment was about the M-Scaler, but most of their comment was showing a general concern for time response and asked why these sorts of measurements are excluded. Amir replied dismissively, and I replied to Amir (pg. 73) asking the same question.
https://www.audiosciencereview.com/...chord-m-scaler-review-upsampler.35498/page-71

Why does the square wave test not matter?

"Frequency and phase measurements suffice since they can be converted into the impulse response."
This doesn't mean that a linear frequency response with a given phase is going to convert to a linear - meaning compared input and output - square wave or impulse response when all that the frequency response requires to be called linear is 20hz - 20khz.

"It doesn’t need any improving. We cannot detect changes in amplitude with a speed that exceeds the speed of the wavelength of a 20kHz* sound."
A lot of the metrics for a lot of devices can be said to not need improving and yet they are included, so this can't be a reason.

"Absolute phase doesn't matter at high frequencies. A 12 kHz sine wave sounds exactly the same as 12 kHz square wave"
I talked about the ear/brain's limit to sensing amplitude differences in time. A 10us pulse can stimulate the ear. A 10us delay between pulses acts as another way to use energy to demonstrate the brain's sensitivity to 10us. That we can sense 10us would be irrelevant if it couldn't be differentiated from something lower, which would be to say that it is above the limit and just being processed the same as the limit, making something lower the limit instead, but we can differentiate:

"The upper edge of human hearing is 20 kHz so it is unlikely that we could detect a difference between a 25 us impulse and a 10 us impulse (or even a 1 us impulse), but assuming sufficient energy, all would be detectable. Making really short sounds with sufficient levels of energy is technically challenging and requires good transducers, but ultrasonic transducers exist."
From: https://psychology.stackexchange.co...re-the-temporal-limits-of-the-auditory-system
"Observers were asked to discriminate between a pair of 10‐μsec pulses and a single 20‐μsec pulse having the same total energy. The independent variable was the time, ΔT, between the two 10‐μsec pulses. The stimuli were also presented as elements in a periodic pulse train. The ΔT required for resolution of two clicks (two‐click threshold) was 10 μsec."
From: https://asa.scitation.org/doi/abs/10.1121/1.1912374

I explained in a post about the role played by the ear's impedance (read lowpass filter - the reason why we even have a tonal limit). We can imagine that the reason why pulses are working here is because they have enough energy. We can ask why an ultrasonic tone at high amplitude would still be inaudible. This is because right after the positive or negative half wave (half the time of the already-ultrasonic frequency, or twice the frequency) will an opposing polarity/pressure of equal energy will occur, effectively cancelling out the first. This is not solved by increasing the amplitude further. There is not enough time between the rise and fall of one polarity pulse and another, given they are equally powerful, and given the transfer of energy is kinetic. If you lower the amplitude of one halfwave it will start to become audible again, only as a pulse. There is reason to think that we can sense rise and fall times higher than 1/20000th of a sec if they are distanced from each other in time i.e. if they are working around lower frequencies (think upper-mid range/midrange/bass). Impedance does not set an equal tonal and temporal limit for the ear.

"There is no shortest sound detectable by the human ear. Any impulse with enough energy is audible.
The shortest detectable tone, identifiable as a tone, would be on the order of 100 ms. It might be shorter for tones of high pitch."
From: https://sound.stackexchange.com/que...e-shortest-sound-perceptible-to-the-human-ear
"A click, being an impulse signal, basically stimulates the entire cochlea (the inner ear), and hence represents the most powerful auditory stimulus. Pure-tone bursts are, as a consequence, weak stimuli as they stimulate a small proportion of the available hair cells. High sound levels may therefore be necessary for short tonal stimuli, in turn degrading the frequency resolution in the cochlea as excitation spreads away from the characteristic frequencies. Mostly, several milliseconds are used as tone bursts."
"By analyzing this synapse using very short tone bursts, it was shown that the vesicles of glutamate (the neurotransmitter activating the auditory nerve fibers that lead the signal to the brain) were released as shortly as 2 ms after tone onset. Tone bursts were 8 - 16 kHz, 10 ms plateau, 1 ms rise/fall (Khimich, 2005). Note the tone burstduration and rise fall times mentioned in answer PART I."
"The problem is that the experimental data do not show that a spike wouldn't have been generated at shorter stimuli. The data merely show that with an ongoing stimulus the first spike appears at 2 ms. It does not show whether 2 ms of temporal summation is necessary, and that is why I have left part I of my answer in place, as a theoretical lower bound."
From: https://biology.stackexchange.com/questions/27662/what-is-the-human-ears-temporal-resolution

~

Let's say we forget about the limit of our time sensitivity because the common device options as they are work so far under that limit*. This still doesn't mean there isn't room for improvement.

An amp can always be improved by being a high current/output impedance amp when powering an electromagnetic coil type speaker. Voltage amps are said to sound as having a tight bass (good) albeit altogether dry sound (bad). Current amps are said to sound as having a fast bass (good) and altogether clear sound (good), despite that they have been used with the commonly available high Q speakers designed around voltage amps. The reason is because voltage amps provide damping factor for the bass, but allow the inductance (think about how high the inductance can be at 20khz*…) to affect the rise time of all frequencies, including the bass. Current amps allow the inductance to affect the magnitude of the frequencies, but less so the rise time. While you might choose a low inductance/low impedance driver for the sake of impedance/gain matching, the benefit isn't being derived from the driver. Measurement comparisons of the acoustic output between a speaker being powered by current and voltage amplifier show the voltage amp amounting to higher intermodulation: http://pmacura.cz/speaker_dist.htm
Impulse response (post #504): https://www.diyaudio.com/community/...sconductance-current-amplifier.239321/page-26
Given that the distortion is due to the amp-speaker combo and not innately either, and given that the magnitude nonlinearities, ringing, and noise/gain mismatch of current amps is due to the driver design, can you have all of the benefits and none of the cons by choosing a current amp and a driver designed for low Q and low impedance (a low impedance driver being more likely to be resistive and near linear).
You can lower the effect of the inductance on rise time by lowering the inductance of the driver. One can think that this addresses the problem alone, and that they can go ahead and use it with a voltage amp. The problem now is that the voltage amp will not have much a of a load, and will be sent into high power, where slew rate is a problem. Now you're back to square one time distortion-wise. Pairing a current amp with a low impedance driver provides gain matching and doubles down on lowering the effect of inductance.

It's common for people from this site to say that "there is no audible difference" regarding anything. Sometimes it's said "we can't hear a difference", this is speaking on other's behalf, which you can't do. You can say that those people are trying to justify their purchase in their mind or toward a brand they affiliate with, but there are many cases where someone hears something someone they know bought, at an event, or that they were expecting not sound good. It isn't enough when I'm going to wonder why someone else said they did hear a difference. The discrepancy demands a test be taken to put it to rest. We might dismiss a simple "it sounds better", but someone saying "it sounds fast" has more meaning, even though implicitly it says better because slow would be worse, but with a word like that we have something to go off of for comparing measurements or conducting a test experiment.

Take the Rehdeko speaker, there are square wave tests, and there are many reports saying it sounds fast. You can say it's the distortion that is making it sound fast. If this were true you would be able to test it. You can take the all of the other measurements for reverse engineering a speaker so that measures as closely as possible in all respects (possible given that there are thresholds in every respect) except the square wave. Blind test people with both.
Class d amps. Take a linear working transistor amp, get it to measure the same as the class d amp in all areas but the impulse through some combination of keeping metrics below audible threshold or artificially raising them to match.
Delta-sigma dacs. These can measure well in literally every respect except their impulse response. Design a resistor ladder dac with no form of oversampling and with an analogue lowpass filter well beyond the audible band only for the sake of assuring it's bandwidth is aligned with or lower than the proceeding devices and/or raise the bandwidth of the proceeding devices. Get the other figures to be either below audibility or artificially raise the distortion of the delta-sigma dac to match. Play natively recorded high sample rate music through both.
The key here is that there are control variables, otherwise it's not scientific.

Wiki class d quantization errors: "While some class-D amplifiers may indeed be controlled by digital circuits or include digital signal processing devices, the power stage deals with voltage and current as a function of non-quantized time. The smallest amount of noise, timing uncertainty, voltage ripple or any other non-ideality immediately results in an irreversible change of the output signal. The same errors in a digital system will only lead to incorrect results when they become so large that a signal representing a digit is distorted beyond recognition."
From: https://en.wikipedia.org/wiki/Class-D_amplifier
Now consider the system: Delta-sigma dac, class d high voltage output (indicated given there are high current output class d's to compare, and people said they like them better), high Q speakers. I am skeptical that this is 'the end' when they can measure well in all ways except their impulse response/square.

Quoted for truth, and for prosperity.

 

[restorer-john]

Those of us who work with high fidelity gear every day know the extreme usefulness of the square wave. The waveform is just as real, if not moreso than any other arbitrary waveform you can pass through a device.

The fact that it can tell you myriad characteristics and details about a device with just a single trace is wonderful. It was the first signal I passed through a 46 year old integrated amplifier yesterday when assessing the functionality of the preamp, tone amps/controls, turnover frequencies, filter stages and power stages all in one go. Then a 100Hz square. The sine sweeps came later.

Honestly, don't bother trying to convince ASR members of its usefulness- they don't get it. They worship at the 1kHz sine altar.

 

[pma]

Those of us who work with high fidelity gear every day know the extreme usefulness of the square wave.

Many here just have no idea about it. Too many armchair experts around.

 

[Axo1989]

When a discussion about the square wave is started it is common for someone say "an ideal square wave cannot exist because it would require infinite bandwidth", but this is not a reason to not discuss the square wave in general when an ideal sine wave seems also unable to exist given it would require zero harmonic distortion. Every sine wave test or generator I have seen has harmonic distortion. One can say that "it doesn't matter because in quasi form it is acceptable", and this is just the point. One can say that a square wave is physically impossible given it supposes instantaneity, solidifying the idea of it being unable to exist, but an ideal sawtooth wave also supposes instantaneity, and yet I don't hear people saying we shouldn't consider it when made quasi.


I wouldn't be surprised if it was deleted by someone. In another thread I tried to carry a conversation started by user @TopG who seems to have been deleted, and I was banned. (Pg. 71) Part of their comment was about the M-Scaler, but most of their comment was showing a general concern for time response and asked why these sorts of measurements are excluded. Amir replied dismissively, and I replied to Amir (pg. 73) asking the same question.
https://www.audiosciencereview.com/...chord-m-scaler-review-upsampler.35498/page-71

Why does the square wave test not matter?

"Frequency and phase measurements suffice since they can be converted into the impulse response."
This doesn't mean that a linear frequency response with a given phase is going to convert to a linear - meaning compared input and output - square wave or impulse response when all that the frequency response requires to be called linear is 20hz - 20khz.

"It doesn’t need any improving. We cannot detect changes in amplitude with a speed that exceeds the speed of the wavelength of a 20kHz* sound."
A lot of the metrics for a lot of devices can be said to not need improving and yet they are included, so this can't be a reason.

"Absolute phase doesn't matter at high frequencies. A 12 kHz sine wave sounds exactly the same as 12 kHz square wave"
I talked about the ear/brain's limit to sensing amplitude differences in time. A 10us pulse can stimulate the ear. A 10us delay between pulses acts as another way to use energy to demonstrate the brain's sensitivity to 10us. That we can sense 10us would be irrelevant if it couldn't be differentiated from something lower, which would be to say that it is above the limit and just being processed the same as the limit, making something lower the limit instead, but we can differentiate:

"The upper edge of human hearing is 20 kHz so it is unlikely that we could detect a difference between a 25 us impulse and a 10 us impulse (or even a 1 us impulse), but assuming sufficient energy, all would be detectable. Making really short sounds with sufficient levels of energy is technically challenging and requires good transducers, but ultrasonic transducers exist."
From: https://psychology.stackexchange.co...re-the-temporal-limits-of-the-auditory-system
"Observers were asked to discriminate between a pair of 10‐μsec pulses and a single 20‐μsec pulse having the same total energy. The independent variable was the time, ΔT, between the two 10‐μsec pulses. The stimuli were also presented as elements in a periodic pulse train. The ΔT required for resolution of two clicks (two‐click threshold) was 10 μsec."
From: https://asa.scitation.org/doi/abs/10.1121/1.1912374

I explained in a post about the role played by the ear's impedance (read lowpass filter - the reason why we even have a tonal limit). We can imagine that the reason why pulses are working here is because they have enough energy. We can ask why an ultrasonic tone at high amplitude would still be inaudible. This is because right after the positive or negative half wave (half the time of the already-ultrasonic frequency, or twice the frequency) will an opposing polarity/pressure of equal energy will occur, effectively cancelling out the first. This is not solved by increasing the amplitude further. There is not enough time between the rise and fall of one polarity pulse and another, given they are equally powerful, and given the transfer of energy is kinetic. If you lower the amplitude of one halfwave it will start to become audible again, only as a pulse. There is reason to think that we can sense rise and fall times higher than 1/20000th of a sec if they are distanced from each other in time i.e. if they are working around lower frequencies (think upper-mid range/midrange/bass). Impedance does not set an equal tonal and temporal limit for the ear.

"There is no shortest sound detectable by the human ear. Any impulse with enough energy is audible.
The shortest detectable tone, identifiable as a tone, would be on the order of 100 ms. It might be shorter for tones of high pitch."
From: https://sound.stackexchange.com/que...e-shortest-sound-perceptible-to-the-human-ear
"A click, being an impulse signal, basically stimulates the entire cochlea (the inner ear), and hence represents the most powerful auditory stimulus. Pure-tone bursts are, as a consequence, weak stimuli as they stimulate a small proportion of the available hair cells. High sound levels may therefore be necessary for short tonal stimuli, in turn degrading the frequency resolution in the cochlea as excitation spreads away from the characteristic frequencies. Mostly, several milliseconds are used as tone bursts."
"By analyzing this synapse using very short tone bursts, it was shown that the vesicles of glutamate (the neurotransmitter activating the auditory nerve fibers that lead the signal to the brain) were released as shortly as 2 ms after tone onset. Tone bursts were 8 - 16 kHz, 10 ms plateau, 1 ms rise/fall (Khimich, 2005). Note the tone burstduration and rise fall times mentioned in answer PART I."
"The problem is that the experimental data do not show that a spike wouldn't have been generated at shorter stimuli. The data merely show that with an ongoing stimulus the first spike appears at 2 ms. It does not show whether 2 ms of temporal summation is necessary, and that is why I have left part I of my answer in place, as a theoretical lower bound."
From: https://biology.stackexchange.com/questions/27662/what-is-the-human-ears-temporal-resolution

~

Let's say we forget about the limit of our time sensitivity because the common device options as they are work so far under that limit*. This still doesn't mean there isn't room for improvement.

An amp can always be improved by being a high current/output impedance amp when powering an electromagnetic coil type speaker. Voltage amps are said to sound as having a tight bass (good) albeit altogether dry sound (bad). Current amps are said to sound as having a fast bass (good) and altogether clear sound (good), despite that they have been used with the commonly available high Q speakers designed around voltage amps. The reason is because voltage amps provide damping factor for the bass, but allow the inductance (think about how high the inductance can be at 20khz*…) to affect the rise time of all frequencies, including the bass. Current amps allow the inductance to affect the magnitude of the frequencies, but less so the rise time. While you might choose a low inductance/low impedance driver for the sake of impedance/gain matching, the benefit isn't being derived from the driver. Measurement comparisons of the acoustic output between a speaker being powered by a current vs a voltage amplifier show the voltage amp amounting to higher intermodulation: http://pmacura.cz/speaker_dist.htm
Impulse response (post #504): https://www.diyaudio.com/community/...sconductance-current-amplifier.239321/page-26
Given that the distortion is due to the amp-speaker combo and not innately either, and given that the magnitude nonlinearities, ringing, and noise/gain mismatch of current amps is due to the driver design, can you have all of the benefits and none of the cons by choosing a current amp and a driver designed for low Q and low impedance (a low impedance driver being more likely to be resistive and near linear).
You can lower the effect of the inductance on rise time by lowering the inductance of the driver. One can think that this addresses the problem alone, and that they can go ahead and use it with a voltage amp. The problem now is that the voltage amp will not have much a of a load, and will be sent into high power, where slew rate is a problem. Now you're back to square one time distortion-wise. Pairing a current amp with a low impedance driver provides gain matching and doubles down on lowering the effect of inductance.

It's common for people from this site to say that "there is no audible difference" regarding anything. Sometimes it's said "we can't hear a difference", this is speaking on other's behalf, which you can't do. You can say that those people are trying to justify their purchase in their mind or toward a brand they affiliate with, but there are many cases where someone hears something someone they know bought, at an event, or that they were expecting not sound good. It isn't enough when I'm going to wonder why someone else said they did hear a difference. The discrepancy demands a test be taken to put it to rest. We might dismiss a simple "it sounds better", but someone saying "it sounds fast" has more meaning, even though implicitly it says better because slow would be worse, but with a word like that we have something to go off of for comparing measurements or conducting a test experiment.

Take the Rehdeko speaker, there are square wave tests, and there are many reports saying it sounds fast. You can say it's the distortion that is making it sound fast. If this were true you would be able to test it. You can take the all of the other measurements for reverse engineering a speaker so that measures as closely as possible in all respects (possible given that there are thresholds in every respect) except the square wave. Blind test people with both.
Class d amps. Take a linear working transistor amp, get it to measure the same as the class d amp in all areas but the impulse through some combination of keeping metrics below audible threshold or artificially raising them to match.
Delta-sigma dacs. These can measure well in literally every respect except their impulse response. Design a resistor ladder dac with no form of oversampling and with an analogue lowpass filter well beyond the audible band only for the sake of assuring it's bandwidth is aligned with or lower than the proceeding devices and/or raise the bandwidth of the proceeding devices. Get the other figures to be either below audibility or artificially raise the distortion of the delta-sigma dac to match. Play natively recorded high sample rate music through both.
The key here is that there are control variables, otherwise it's not scientific.

Wiki class d quantization errors: "While some class-D amplifiers may indeed be controlled by digital circuits or include digital signal processing devices, the power stage deals with voltage and current as a function of non-quantized time. The smallest amount of noise, timing uncertainty, voltage ripple or any other non-ideality immediately results in an irreversible change of the output signal. The same errors in a digital system will only lead to incorrect results when they become so large that a signal representing a digit is distorted beyond recognition."
From: https://en.wikipedia.org/wiki/Class-D_amplifier
Now consider the system: Delta-sigma dac, class d high voltage output (indicated given there are high current output class d's to compare, and people said they like them better), high Q speakers. I am skeptical that this is 'the end' when they can measure well in all ways except their impulse response/square.

Interesting reading, thank you.

 

[levimax]

When a discussion about the square wave is started it is common for someone say "an ideal square wave cannot exist because it would require infinite bandwidth", but this is not a reason to not discuss the square wave in general when an ideal sine wave seems also unable to exist given it would require zero harmonic distortion. Every sine wave test or generator I have seen has harmonic distortion. One can say that "it doesn't matter because in quasi form it is acceptable", and this is just the point. One can say that a square wave is physically impossible given it supposes instantaneity, solidifying the idea of it being unable to exist, but an ideal sawtooth wave also supposes instantaneity, and yet I don't hear people saying we shouldn't consider it when made quasi.


I wouldn't be surprised if it was deleted by someone. In another thread I tried to carry a conversation started by user @TopG who seems to have been deleted, and I was banned. (Pg. 71) Part of their comment was about the M-Scaler, but most of their comment was showing a general concern for time response and asked why these sorts of measurements are excluded. Amir replied dismissively, and I replied to Amir (pg. 73) asking the same question.
https://www.audiosciencereview.com/...chord-m-scaler-review-upsampler.35498/page-71

Why does the square wave test not matter?

"Frequency and phase measurements suffice since they can be converted into the impulse response."
This doesn't mean that a linear frequency response with a given phase is going to convert to a linear - meaning compared input and output - square wave or impulse response when all that the frequency response requires to be called linear is 20hz - 20khz.

"It doesn’t need any improving. We cannot detect changes in amplitude with a speed that exceeds the speed of the wavelength of a 20kHz* sound."
A lot of the metrics for a lot of devices can be said to not need improving and yet they are included, so this can't be a reason.

"Absolute phase doesn't matter at high frequencies. A 12 kHz sine wave sounds exactly the same as 12 kHz square wave"
I talked about the ear/brain's limit to sensing amplitude differences in time. A 10us pulse can stimulate the ear. A 10us delay between pulses acts as another way to use energy to demonstrate the brain's sensitivity to 10us. That we can sense 10us would be irrelevant if it couldn't be differentiated from something lower, which would be to say that it is above the limit and just being processed the same as the limit, making something lower the limit instead, but we can differentiate:

"The upper edge of human hearing is 20 kHz so it is unlikely that we could detect a difference between a 25 us impulse and a 10 us impulse (or even a 1 us impulse), but assuming sufficient energy, all would be detectable. Making really short sounds with sufficient levels of energy is technically challenging and requires good transducers, but ultrasonic transducers exist."
From: https://psychology.stackexchange.co...re-the-temporal-limits-of-the-auditory-system
"Observers were asked to discriminate between a pair of 10‐μsec pulses and a single 20‐μsec pulse having the same total energy. The independent variable was the time, ΔT, between the two 10‐μsec pulses. The stimuli were also presented as elements in a periodic pulse train. The ΔT required for resolution of two clicks (two‐click threshold) was 10 μsec."
From: https://asa.scitation.org/doi/abs/10.1121/1.1912374

I explained in a post about the role played by the ear's impedance (read lowpass filter - the reason why we even have a tonal limit). We can imagine that the reason why pulses are working here is because they have enough energy. We can ask why an ultrasonic tone at high amplitude would still be inaudible. This is because right after the positive or negative half wave (half the time of the already-ultrasonic frequency, or twice the frequency) will an opposing polarity/pressure of equal energy will occur, effectively cancelling out the first. This is not solved by increasing the amplitude further. There is not enough time between the rise and fall of one polarity pulse and another, given they are equally powerful, and given the transfer of energy is kinetic. If you lower the amplitude of one halfwave it will start to become audible again, only as a pulse. There is reason to think that we can sense rise and fall times higher than 1/20000th of a sec if they are distanced from each other in time i.e. if they are working around lower frequencies (think upper-mid range/midrange/bass). Impedance does not set an equal tonal and temporal limit for the ear.

"There is no shortest sound detectable by the human ear. Any impulse with enough energy is audible.
The shortest detectable tone, identifiable as a tone, would be on the order of 100 ms. It might be shorter for tones of high pitch."
From: https://sound.stackexchange.com/que...e-shortest-sound-perceptible-to-the-human-ear
"A click, being an impulse signal, basically stimulates the entire cochlea (the inner ear), and hence represents the most powerful auditory stimulus. Pure-tone bursts are, as a consequence, weak stimuli as they stimulate a small proportion of the available hair cells. High sound levels may therefore be necessary for short tonal stimuli, in turn degrading the frequency resolution in the cochlea as excitation spreads away from the characteristic frequencies. Mostly, several milliseconds are used as tone bursts."
"By analyzing this synapse using very short tone bursts, it was shown that the vesicles of glutamate (the neurotransmitter activating the auditory nerve fibers that lead the signal to the brain) were released as shortly as 2 ms after tone onset. Tone bursts were 8 - 16 kHz, 10 ms plateau, 1 ms rise/fall (Khimich, 2005). Note the tone burstduration and rise fall times mentioned in answer PART I."
"The problem is that the experimental data do not show that a spike wouldn't have been generated at shorter stimuli. The data merely show that with an ongoing stimulus the first spike appears at 2 ms. It does not show whether 2 ms of temporal summation is necessary, and that is why I have left part I of my answer in place, as a theoretical lower bound."
From: https://biology.stackexchange.com/questions/27662/what-is-the-human-ears-temporal-resolution

~

Let's say we forget about the limit of our time sensitivity because the common device options as they are work so far under that limit*. This still doesn't mean there isn't room for improvement.

An amp can always be improved by being a high current/output impedance amp when powering an electromagnetic coil type speaker. Voltage amps are said to sound as having a tight bass (good) albeit altogether dry sound (bad). Current amps are said to sound as having a fast bass (good) and altogether clear sound (good), despite that they have been used with the commonly available high Q speakers designed around voltage amps. The reason is because voltage amps provide damping factor for the bass, but allow the inductance (think about how high the inductance can be at 20khz*…) to affect the rise time of all frequencies, including the bass. Current amps allow the inductance to affect the magnitude of the frequencies, but less so the rise time. While you might choose a low inductance/low impedance driver for the sake of impedance/gain matching, the benefit isn't being derived from the driver. Measurement comparisons of the acoustic output between a speaker being powered by a current vs a voltage amplifier show the voltage amp amounting to higher intermodulation: http://pmacura.cz/speaker_dist.htm
Impulse response (post #504): https://www.diyaudio.com/community/...sconductance-current-amplifier.239321/page-26
Given that the distortion is due to the amp-speaker combo and not innately either, and given that the magnitude nonlinearities, ringing, and noise/gain mismatch of current amps is due to the driver design, can you have all of the benefits and none of the cons by choosing a current amp and a driver designed for low Q and low impedance (a low impedance driver being more likely to be resistive and near linear).
You can lower the effect of the inductance on rise time by lowering the inductance of the driver. One can think that this addresses the problem alone, and that they can go ahead and use it with a voltage amp. The problem now is that the voltage amp will not have much a of a load, and will be sent into high power, where slew rate is a problem. Now you're back to square one time distortion-wise. Pairing a current amp with a low impedance driver provides gain matching and doubles down on lowering the effect of inductance.

It's common for people from this site to say that "there is no audible difference" regarding anything. Sometimes it's said "we can't hear a difference", this is speaking on other's behalf, which you can't do. You can say that those people are trying to justify their purchase in their mind or toward a brand they affiliate with, but there are many cases where someone hears something someone they know bought, at an event, or that they were expecting not sound good. It isn't enough when I'm going to wonder why someone else said they did hear a difference. The discrepancy demands a test be taken to put it to rest. We might dismiss a simple "it sounds better", but someone saying "it sounds fast" has more meaning, even though implicitly it says better because slow would be worse, but with a word like that we have something to go off of for comparing measurements or conducting a test experiment.

Take the Rehdeko speaker, there are square wave tests, and there are many reports saying it sounds fast. You can say it's the distortion that is making it sound fast. If this were true you would be able to test it. You can take the all of the other measurements for reverse engineering a speaker so that measures as closely as possible in all respects (possible given that there are thresholds in every respect) except the square wave. Blind test people with both.
Class d amps. Take a linear working transistor amp, get it to measure the same as the class d amp in all areas but the impulse through some combination of keeping metrics below audible threshold or artificially raising them to match.
Delta-sigma dacs. These can measure well in literally every respect except their impulse response. Design a resistor ladder dac with no form of oversampling and with an analogue lowpass filter well beyond the audible band only for the sake of assuring it's bandwidth is aligned with or lower than the proceeding devices and/or raise the bandwidth of the proceeding devices. Get the other figures to be either below audibility or artificially raise the distortion of the delta-sigma dac to match. Play natively recorded high sample rate music through both.
The key here is that there are control variables, otherwise it's not scientific.

Wiki class d quantization errors: "While some class-D amplifiers may indeed be controlled by digital circuits or include digital signal processing devices, the power stage deals with voltage and current as a function of non-quantized time. The smallest amount of noise, timing uncertainty, voltage ripple or any other non-ideality immediately results in an irreversible change of the output signal. The same errors in a digital system will only lead to incorrect results when they become so large that a signal representing a digit is distorted beyond recognition."
From: https://en.wikipedia.org/wiki/Class-D_amplifier
Now consider the system: Delta-sigma dac, class d high voltage output (indicated given there are high current output class d's to compare, and people said they like them better), high Q speakers. I am skeptical that this is 'the end' when they can measure well in all ways except their impulse response/square.

Anyone who doesn't appreciate the power and value of square wave testing to confirm an amp is working as expected or not and to diagnose problems quickly and efficiently needs to do some research as I think this is as close to a fact as there can be in the audio world.

Where things fall apart is how correlated a "near perfect" square wave is with perceived sound quality. In an extreme case if you look at a 10 Khz square wave reproduced by a high quality class D amp and a wide bandwidth class AB you would conclude the class D amp would sound terrible but that is not the case. Just like SINAD, which is also limited in it's ability to predict perceived sound quality but has it's place, a square wave also has it's limits to predict perceived sound quality.

It seems like you are claiming some type of "conspiracy" to cover up square wave testing to keep people in the dark. I do see what some could perceive as heavy handed moderating which usually happens when someone make a "scientific claim" such as better square waves sound better but can't back it up with some evidence of perceived sound quality in blind testing. In the name of not confusing people posts are deleted and in some cases people are banned. I can see both sides of this and think ASR does OK in general but the one thing that seems inconsistent to me is that SINAD is seldom if ever held to the same standard. The excuse is "better SINAD means better engineering" but that is not necessarily true as what is the point of spending extra money on better SINAD if it doesn't make any difference for it's intended use? When it comes to "square waves" it could certainly be argued that it is "better engineering" if class D amps could reproduce near perfect 10 Khz square waves but again why bother if it is not important for perceived sound quality.

To me when talking about amps everything comes back to blind testing and perceived sound quality in order to determine what is "good enough" and at that point move on to something else like better speakers rather than chasing around things that you can't hear.