Hi all
Am I allowed to throw a speculative question here (like talking to myself)?
I wonder (Wanda) if the perceived superiority in resolution, detail, transparency of the the class D amplifiers over class A, AB amps is really attributed to the different way these amps process the electrical signal (hardware level).
What I’m about to say? I will explain myself . But first I have to explain how did I start to question the technical nature of the problem.
They were two cases (one pre and one power amplifier) that were actually modulating the music signal with high frequency signal, in other words, they were oscillating. Nature of oscillation was such that it was sucking power and the music signal was starving low in amplitude. That was the reason they were brought to the bench. Acoustically, they were anaemic, thin, but oddly enough, artificially transparent. After oscillation problem was rectified, signal balance returned to normal, but this transparency went away.
This happened more than 5 years ago.
A few months ago, I had the chance to listen to a class D amp for the first time. The sound was very detailed and transparent. Next, I could not resist to measure it with the scope. To my ignorance and to my surprise, the output signal was modulated by a high freq. signal. It was the participants of this forum that convinced me that this was normal with such amps.
These qualities of detailed and transparent sound I experienced (to a higher degree) again with a TA2020 chip amp.
Now, thinking over why and how is it possible for these amps to produce these results, the case with the oscillating analogue amplifiers came back to my mind.
Is the correlation between the (accidental for the analogue, intentional for the class D) high frequency modulation of the output signal and the perceived detailed-transparent sound a mere coincidence, or is it something more systematic behind it?
What do I imply? Might be the case that it is the nature of the modulated acoustic waves that excites the auditory system in a more efficient way than the non modulated ones? (in technical jargon: does this modulation works to the auditory system like the HF bias works to the magnetic recording media)?
Absurd? Maybe.
Please note the following:
In the field of visual perception the following is true:
We are visually inspecting a certain surface with defined subjective conditions (i.e. lighting, angle of view, distance ect.) trying to locate any small flaw. We can increase our ability to locate such a flaw by introducing a relative motion between our eyes and the inspection surface. This relative motion has to be in a plane normal to the axis of our sight, to be small in amplitude and to be repetitive. The thus increased
discrimination capability of the visual system is said to be attributed to two factors:
The first factor is the electrochemical way that the eye cones and rodes function. They do saturate when the stimulus ( image projected over them) remains unchanged. Literally speaking, this happens only when motional nerves are neutralized by anaesthesia. Otherwise, the eye is constantly under unconscious, unintended micro-motion just for to prevent reaching saturation level. So one can say that this unconscious micro-motion is the way to keep cones and rodes in a condition just over the saturation-no impulse output-level. The intentionally applied relative motion during visual inspection adds to this level, thus increasing the resolution-discrimination- capabilities of the eye.
The second factor has to do with the access priority of the peripheral vision (most sensitive in detecting motion) over the central vision (more specialized in details and colour perception) in reaching and stimulating the alert command section of mamal's brain (this is the way the evolution process has set weighing factors and priorities into the structure of the visual perception system).
Regards
George
Am I allowed to throw a speculative question here (like talking to myself)?
I wonder (Wanda) if the perceived superiority in resolution, detail, transparency of the the class D amplifiers over class A, AB amps is really attributed to the different way these amps process the electrical signal (hardware level).
What I’m about to say? I will explain myself . But first I have to explain how did I start to question the technical nature of the problem.
They were two cases (one pre and one power amplifier) that were actually modulating the music signal with high frequency signal, in other words, they were oscillating. Nature of oscillation was such that it was sucking power and the music signal was starving low in amplitude. That was the reason they were brought to the bench. Acoustically, they were anaemic, thin, but oddly enough, artificially transparent. After oscillation problem was rectified, signal balance returned to normal, but this transparency went away.
This happened more than 5 years ago.
A few months ago, I had the chance to listen to a class D amp for the first time. The sound was very detailed and transparent. Next, I could not resist to measure it with the scope. To my ignorance and to my surprise, the output signal was modulated by a high freq. signal. It was the participants of this forum that convinced me that this was normal with such amps.
These qualities of detailed and transparent sound I experienced (to a higher degree) again with a TA2020 chip amp.
Now, thinking over why and how is it possible for these amps to produce these results, the case with the oscillating analogue amplifiers came back to my mind.
Is the correlation between the (accidental for the analogue, intentional for the class D) high frequency modulation of the output signal and the perceived detailed-transparent sound a mere coincidence, or is it something more systematic behind it?
What do I imply? Might be the case that it is the nature of the modulated acoustic waves that excites the auditory system in a more efficient way than the non modulated ones? (in technical jargon: does this modulation works to the auditory system like the HF bias works to the magnetic recording media)?
Absurd? Maybe.
Please note the following:
In the field of visual perception the following is true:
We are visually inspecting a certain surface with defined subjective conditions (i.e. lighting, angle of view, distance ect.) trying to locate any small flaw. We can increase our ability to locate such a flaw by introducing a relative motion between our eyes and the inspection surface. This relative motion has to be in a plane normal to the axis of our sight, to be small in amplitude and to be repetitive. The thus increased
discrimination capability of the visual system is said to be attributed to two factors:
The first factor is the electrochemical way that the eye cones and rodes function. They do saturate when the stimulus ( image projected over them) remains unchanged. Literally speaking, this happens only when motional nerves are neutralized by anaesthesia. Otherwise, the eye is constantly under unconscious, unintended micro-motion just for to prevent reaching saturation level. So one can say that this unconscious micro-motion is the way to keep cones and rodes in a condition just over the saturation-no impulse output-level. The intentionally applied relative motion during visual inspection adds to this level, thus increasing the resolution-discrimination- capabilities of the eye.
The second factor has to do with the access priority of the peripheral vision (most sensitive in detecting motion) over the central vision (more specialized in details and colour perception) in reaching and stimulating the alert command section of mamal's brain (this is the way the evolution process has set weighing factors and priorities into the structure of the visual perception system).
Regards
George
Paragraphs buddy, learn to use the [enter] key!
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Hi,
I think you need to learn a little more about classD. You are right, the sound does achive a very high purety and sound extremly detailed but it is nothing what so ever to do with the HF bleeding through the output. The HF bleed is mearly a by-product of the amplifiying methods employed in this class of amp. The reason it sounds so detailed and clear is simply because the audio really IS that good. I can't understand why everyone has this perception that classD sounds crumby
Regards
Mad.P
I think you need to learn a little more about classD. You are right, the sound does achive a very high purety and sound extremly detailed but it is nothing what so ever to do with the HF bleeding through the output. The HF bleed is mearly a by-product of the amplifiying methods employed in this class of amp. The reason it sounds so detailed and clear is simply because the audio really IS that good. I can't understand why everyone has this perception that classD sounds crumby
Regards
Mad.P
Mad P.
Yes. I do have a limited knowledge on anything in electronics, let alone class d, but this i have already understood. What i was really asking in thread #1 is if this mere by-product has by chance the ability to bias the auditory system in a beneficial way. A way for answering this question is having a very effective and ideal PL output filter that zeroes out any residues of HF. If the sound is still transparent and detailed after this filtering, then what i suggest as a possibility is plain wrong.
Please, can you elaborate?
Regards
George
Greetings,
When i say the sound is 'That good' ( clear ) i mean just that. The sound is an accurate representation of the input signal. Very little distortions and very low unwanted noise.
I can see what your getting at with the HF makeing sound appear brighter and cleaner than it really is, but even if this phenomenone does indeed exist ( and it might, i like to think it will certaintly alter our perception of what we are hearing ) I don't think it is at play here because the speakers will be unable to convert the high frequencys to sound waves. We can see it on the scope thats for sure but is the high frequency being transmitted into the air as sound waves? I don't think any speaker is capable of that! The only other avenue of thought here is the possibility of the HF altering the way the speakers are converting the electrical waves into sound waves. Maybe the presence of HF interferes at this point. Who knows, i don't think it is but sombody somewere may know different. I personally don't know much about speakers in this respect so sombody a little more qualified may be able to enlighten us on this.
Regards
Mad.P
When i say the sound is 'That good' ( clear ) i mean just that. The sound is an accurate representation of the input signal. Very little distortions and very low unwanted noise.
I can see what your getting at with the HF makeing sound appear brighter and cleaner than it really is, but even if this phenomenone does indeed exist ( and it might, i like to think it will certaintly alter our perception of what we are hearing ) I don't think it is at play here because the speakers will be unable to convert the high frequencys to sound waves. We can see it on the scope thats for sure but is the high frequency being transmitted into the air as sound waves? I don't think any speaker is capable of that! The only other avenue of thought here is the possibility of the HF altering the way the speakers are converting the electrical waves into sound waves. Maybe the presence of HF interferes at this point. Who knows, i don't think it is but sombody somewere may know different. I personally don't know much about speakers in this respect so sombody a little more qualified may be able to enlighten us on this.
Regards
Mad.P
It's been said before, but lack of any sort of crossover distortion may have something to do with it. Class d never tries to amplify a signal in the sense that the output follows the input signal. It modulates the audio signal (only has to switch rapidly between two voltage levels as opposed to "tracing" the input signal at a higher voltage) and allows the filter to recover that audio information.
The principles of class d operation are based on Fourier's theorem which basically says that a square wave is composed of an infinite number of sine waves of varying frequencies (sounds like music, eh?). Interestingly enough our ears can be thought of as Fourier analyzer.
Off topic a bit, the Tripath chips do add dither to the modulation signal prior to comparing it with the audio signal. In a 1-bit system like an A/D converter, signal information below the least significant bit can be coded if dither is added to the signal. This effectively codes the information in a PWM signal. This can be compared to a class d amp which is similar to a 1-bit system in that the output signal has only two discrete levels but is continuous in the time domain.
I don't know if any of that helped but I threw it out there anyway
The principles of class d operation are based on Fourier's theorem which basically says that a square wave is composed of an infinite number of sine waves of varying frequencies (sounds like music, eh?). Interestingly enough our ears can be thought of as Fourier analyzer.
Off topic a bit, the Tripath chips do add dither to the modulation signal prior to comparing it with the audio signal. In a 1-bit system like an A/D converter, signal information below the least significant bit can be coded if dither is added to the signal. This effectively codes the information in a PWM signal. This can be compared to a class d amp which is similar to a 1-bit system in that the output signal has only two discrete levels but is continuous in the time domain.
I don't know if any of that helped but I threw it out there anyway
I have thought long and hard into why class D amplifiers sound better than linear amplifers. Part of the answer is that they don't, a really good linear ampilfier will sound as good. The difference is that class D is easier to get to sound like a very good linear amp.
Now for why I think that class D does make things sound better. First class D does not increase in distortion at the same rate as linear amplification in general. All amplifiers will increase in distortion with an increase in frequency response. This is directly related to frequency response extention. A divice increases in distortion toward is limits, in almost every way, now class A/B B amps can have more distortion at lower frequencies because often their crossover distortion is greater at lower powers. This was one reason why people started to listen to music louder, because the amp had less distortion at higher volumes, other than the discovery of how the Fletcher/Munson curve applied.
At any rate class D ampifiers don't increase in distortion like linear amps do, they do but not nearly on the same degree, or in the same way. These are not absolute statements I am making, more generally speaking. Likewise the frequency extention in the upper range of class D amplifiers can go quite a bit higher than most linear ampilfiers. Our ears precieve location because of the timing difference of when the sound hits each ear, along with amplitude, but the later is secondary. So a frequency response higher than 20K is beneficial in that it helps us precieve a "cleaner" sound stage and imaging of the signal. It gives more amient information to the sound and generally helps things appear clearer.
All in all the reason things sound better is directly related to distortion, the less we have in the signal the better it will sound, and class D amplifiers don't have the distortions that linear amplifers have in the same ways. With class D most of what we measure with THD is often noise, not true distortion. And in real world situations other than test bench situations class D performs better. Understanding this will also help designers design in such a way that they limit distortions, or in other words design with these things in mine.
Terry
Now for why I think that class D does make things sound better. First class D does not increase in distortion at the same rate as linear amplification in general. All amplifiers will increase in distortion with an increase in frequency response. This is directly related to frequency response extention. A divice increases in distortion toward is limits, in almost every way, now class A/B B amps can have more distortion at lower frequencies because often their crossover distortion is greater at lower powers. This was one reason why people started to listen to music louder, because the amp had less distortion at higher volumes, other than the discovery of how the Fletcher/Munson curve applied.
At any rate class D ampifiers don't increase in distortion like linear amps do, they do but not nearly on the same degree, or in the same way. These are not absolute statements I am making, more generally speaking. Likewise the frequency extention in the upper range of class D amplifiers can go quite a bit higher than most linear ampilfiers. Our ears precieve location because of the timing difference of when the sound hits each ear, along with amplitude, but the later is secondary. So a frequency response higher than 20K is beneficial in that it helps us precieve a "cleaner" sound stage and imaging of the signal. It gives more amient information to the sound and generally helps things appear clearer.
All in all the reason things sound better is directly related to distortion, the less we have in the signal the better it will sound, and class D amplifiers don't have the distortions that linear amplifers have in the same ways. With class D most of what we measure with THD is often noise, not true distortion. And in real world situations other than test bench situations class D performs better. Understanding this will also help designers design in such a way that they limit distortions, or in other words design with these things in mine.
Terry
From what I can gather, because I wasn't around in the days it was considered "nasty", is today we have higher switching frequencies availible that enable true full range Class D amplifiers. This is made possible by advances in solid stage devices, like MOSFETs that have become available in the last five years. Likewise their have been many circuit advances in the last five to ten years, as a direct result of the aforementioned.
Terry
Terry
matjans said:bwrx, isn't class d more like an inverse fourier analysis? (re)creating a sine wave with rectangular ones?
(i can hear the marketing people go wild already!)
You better register that before someone else does
"Our class d amps utilize IFA® (Inverse Fourier Analysis), the latest breakthrough in signal processing, to deliver the finest audio reproduction on the planet!"
Basic question on class D
Hi all,
I haven’t been checking in lately and see I should have. In one of the other threads some time ago I explored this phenomena and came to the conclusion the answer is in the very nature of how current flows in conductors. All conductors are composed of many crystals with inherent boundaries between them. It requires energy to cross these boundaries and this lost energy represents lost low level signal information. With a HF signal superimposed on top of the wanted analog signal as in the class D design, or any other switching design for that matter, this boundary energy requirement is met by the HF and not the audio signal. This means that the low level stuff has a clear and open path to transverse and doesn’t get lost. Think of it as hysteresis that has to be overcome. Old style chopper amplifiers used this HF switching to control DC drift but also excelled in low level linearity.
Adding a small amount of HF energy to any normal audio signals will do the same thing and there are products around to do this. One I know of is being marketed as an instant audio improvement tweak. It is nothing more than a HF oscillator to put into your equipment and spread the noise around. It does seem to work.
Roger
Hi all,
I haven’t been checking in lately and see I should have. In one of the other threads some time ago I explored this phenomena and came to the conclusion the answer is in the very nature of how current flows in conductors. All conductors are composed of many crystals with inherent boundaries between them. It requires energy to cross these boundaries and this lost energy represents lost low level signal information. With a HF signal superimposed on top of the wanted analog signal as in the class D design, or any other switching design for that matter, this boundary energy requirement is met by the HF and not the audio signal. This means that the low level stuff has a clear and open path to transverse and doesn’t get lost. Think of it as hysteresis that has to be overcome. Old style chopper amplifiers used this HF switching to control DC drift but also excelled in low level linearity.
Adding a small amount of HF energy to any normal audio signals will do the same thing and there are products around to do this. One I know of is being marketed as an instant audio improvement tweak. It is nothing more than a HF oscillator to put into your equipment and spread the noise around. It does seem to work.
Roger
Hi,
I concure with the possibility of this as of yet unproven effect and yet am not willing to nail it down to any one thing.
The micro diode/bias is one possible explanation.
Another is that it's simply easier for a high frequency ripple to propagate down the wire/line than a strickly DC signal, it could in effect help to emulate an ideal wire by creating a lower impedance via all the stay capacitances, which could be from microdiodes to any other number of things.
It could also be a simple matter of dithering, which I think is my favorite so far.
Regards,
Chris
I concure with the possibility of this as of yet unproven effect and yet am not willing to nail it down to any one thing.
The micro diode/bias is one possible explanation.
Another is that it's simply easier for a high frequency ripple to propagate down the wire/line than a strickly DC signal, it could in effect help to emulate an ideal wire by creating a lower impedance via all the stay capacitances, which could be from microdiodes to any other number of things.
It could also be a simple matter of dithering, which I think is my favorite so far.
Regards,
Chris
I think the reason for classD's often sounding better than A/B's with similar distortion is in the way that "out of band" ripple modulates the linear motor section of a normal loudspeaker and vibrates the cone.
In precision motor drive you can often achieve better results if PWM modulation is used in place of voltage or current regulation. The simple reason for this is that an unfiltered PWM signal has sufficient energy to overcome the Stiction effects of a stationary motor whereas a voltage drive signal needs to provide a certain minimum power before there is any movement in the motor, at which time the motor often moves in a larger than desired step. While I understand that stiction in linear motor is very different problem to stiction in a rotary motor I still think that this effect is part of the reason for the ClassD sound.
The second interesting thing that poorly filtered classD does is to vibrate the cone. Vibrating the cone at frequencies of 300Khz and above does not actually result in any net movement of the cone because it is typically far too massive, however what it does do is to promote the easy breakup of the cone surface. Many years ago I was using an interferometer to measure some cone breakup effects and I noticed that the threshold for breakup was much lower if a polyphonic tone was used than a single sine wave, at the time I attributed this to the higher peak to average ratio but the effect was actually more pronounced than could be explained by some material hysteretic effect being overcome by the higher peak energy. I no longer have access to a suitable interferometer but I would challenge anyone that does to do actually measure the response cone surface when a classD amp is used and compare this to the results when an A/B amp is used. The trick is to make these measurements in the small signal domain -40db to -60db.
ClassDunce
In precision motor drive you can often achieve better results if PWM modulation is used in place of voltage or current regulation. The simple reason for this is that an unfiltered PWM signal has sufficient energy to overcome the Stiction effects of a stationary motor whereas a voltage drive signal needs to provide a certain minimum power before there is any movement in the motor, at which time the motor often moves in a larger than desired step. While I understand that stiction in linear motor is very different problem to stiction in a rotary motor I still think that this effect is part of the reason for the ClassD sound.
The second interesting thing that poorly filtered classD does is to vibrate the cone. Vibrating the cone at frequencies of 300Khz and above does not actually result in any net movement of the cone because it is typically far too massive, however what it does do is to promote the easy breakup of the cone surface. Many years ago I was using an interferometer to measure some cone breakup effects and I noticed that the threshold for breakup was much lower if a polyphonic tone was used than a single sine wave, at the time I attributed this to the higher peak to average ratio but the effect was actually more pronounced than could be explained by some material hysteretic effect being overcome by the higher peak energy. I no longer have access to a suitable interferometer but I would challenge anyone that does to do actually measure the response cone surface when a classD amp is used and compare this to the results when an A/B amp is used. The trick is to make these measurements in the small signal domain -40db to -60db.
ClassDunce
I'd imagine we're also talking about overcomming stiction in the wires themselves, which is also different not having a mechanical aspect as we'd normally view it. Why not extend that theory right to the eardrum?
The true answer could easily lie in all of the above.
Clearly many have been thinking about this but no one has yet proven it, there was even a level of shyness about admitting to it publicly. Good for the OP
Regards,
Chris
PS, Matjans I think your idea is an effective one for verifying the IF of it, but then to verify the why would be a more difficult matter.
Wouldn't it be hilarious in a few years to see all the high end class A amps mixing the signal with a high frequency ripple?
The true answer could easily lie in all of the above.
Clearly many have been thinking about this but no one has yet proven it, there was even a level of shyness about admitting to it publicly. Good for the OP
Regards,
Chris
PS, Matjans I think your idea is an effective one for verifying the IF of it, but then to verify the why would be a more difficult matter.
Wouldn't it be hilarious in a few years to see all the high end class A amps mixing the signal with a high frequency ripple?
Hi all
It is well known that the more man knows about something, the less he is willing to spell unsupported clues. Me, is one of the unqualified (in electronics and in particular digital part of it) ones who is allowed to ask a controversial question.
If (and only if) the electrical modulation of a musical signal by a HF signal really produces a beneficial end effect, then a bit systematic approach might be needed. With the help of the participants in this thread so far, I can see the following sub-questions as important:
A. Non technical area:
A1. (Is it due) to any physiological effect (i.e. mechanical modulation of any sensory part of the ear)?
A2. To any psychoacoustic effect (i.e. brain interpretation mechanism involved)?
A3. To a combination of A1 and A2?
B. Technical area:
B1. To any electromechanical reason (coil and cone motion of the speaker)?
B2. To any change in the amplifier load (cable and speaker) electrical impedance?
B3. To any change in the micro-level (conductor's non-bulk properties)
B4. To any internal (within the amplifier) effect?
B5. To any combination of B1-B4
C. To any combination of A, B
D. To any other factor
Forgive my persistence, but I believe that categorizing helps in the discussion by ruling-out or supporting with evidence any of mechanisms involved. Unique ideas mostly about testing each case or guidance to any relative internet link will be appreciated.
As far as to which or how many concequences such an effect might have -if first proven-, well...
Regards
George
It is well known that the more man knows about something, the less he is willing to spell unsupported clues. Me, is one of the unqualified (in electronics and in particular digital part of it) ones who is allowed to ask a controversial question.
If (and only if) the electrical modulation of a musical signal by a HF signal really produces a beneficial end effect, then a bit systematic approach might be needed. With the help of the participants in this thread so far, I can see the following sub-questions as important:
A. Non technical area:
A1. (Is it due) to any physiological effect (i.e. mechanical modulation of any sensory part of the ear)?
A2. To any psychoacoustic effect (i.e. brain interpretation mechanism involved)?
A3. To a combination of A1 and A2?
B. Technical area:
B1. To any electromechanical reason (coil and cone motion of the speaker)?
B2. To any change in the amplifier load (cable and speaker) electrical impedance?
B3. To any change in the micro-level (conductor's non-bulk properties)
B4. To any internal (within the amplifier) effect?
B5. To any combination of B1-B4
C. To any combination of A, B
D. To any other factor
Forgive my persistence, but I believe that categorizing helps in the discussion by ruling-out or supporting with evidence any of mechanisms involved. Unique ideas mostly about testing each case or guidance to any relative internet link will be appreciated.
As far as to which or how many concequences such an effect might have -if first proven-, well...
Regards
George
Amazing!
Hi all,
I am very glad to see that this subject is actually being taken seriously. There is hope! I can see a series of experiments being set up to prove/disprove each stage of these questions. Some would be relativity simple like injecting some RF into a long wire while measuring the DC resistance. Others like setting up psychoacoustic experiments would have way too many variables and each of these would each have to be worked out. This looks like a very good thing for a bright individual to do as a doctorial thesis.
I think we agree the main thrust would be investigating various forms of hysteresis from one end of the chain to the other. I think this would be the key whether it is electrical or mechanical in form.
There is another phenomenon that could be involved, actual transmission through the air. If you move your hand through the air you feel no resistance but if you stick your hand out a moving car window you defiantly do feel it. Take the case of a bee’s beating wings, at this speed the air acts almost solid, if it didn’t they couldn’t fly. Now if you make the air act solid with a HF bias signal first how would this effect the transmission of normal analog signals that are riding along? I think a relevant question. I do know work has been done in this area and some interesting products developed because of it but what does it do the actual sound?
More to think about for sure,
Roger
Hi all,
I am very glad to see that this subject is actually being taken seriously. There is hope! I can see a series of experiments being set up to prove/disprove each stage of these questions. Some would be relativity simple like injecting some RF into a long wire while measuring the DC resistance. Others like setting up psychoacoustic experiments would have way too many variables and each of these would each have to be worked out. This looks like a very good thing for a bright individual to do as a doctorial thesis.
I think we agree the main thrust would be investigating various forms of hysteresis from one end of the chain to the other. I think this would be the key whether it is electrical or mechanical in form.
There is another phenomenon that could be involved, actual transmission through the air. If you move your hand through the air you feel no resistance but if you stick your hand out a moving car window you defiantly do feel it. Take the case of a bee’s beating wings, at this speed the air acts almost solid, if it didn’t they couldn’t fly. Now if you make the air act solid with a HF bias signal first how would this effect the transmission of normal analog signals that are riding along? I think a relevant question. I do know work has been done in this area and some interesting products developed because of it but what does it do the actual sound?
More to think about for sure,
Roger
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.