Roger
The actual transmission through the air is indeed another valid question (work has been done in the transmission of data through ultasonic x-ducers).
If there is enough interest for the subject of this thread and no objection from the forum, i would suggest two things.
1. This thread to host general discussion and any recommendation on the possibly related topics. The list of post#19 is by no means rigid, it may be edited with new entries.
2. For details on every item of the (editable) list, a new thread with a relative name (e.g. "Wanda B1" for electromechanical ) might be opened. This would help information to be filed systematically, in an easy to find way and would prevent this thread from becoming chaotic.
Regards
George
The actual transmission through the air is indeed another valid question (work has been done in the transmission of data through ultasonic x-ducers).
If there is enough interest for the subject of this thread and no objection from the forum, i would suggest two things.
1. This thread to host general discussion and any recommendation on the possibly related topics. The list of post#19 is by no means rigid, it may be edited with new entries.
2. For details on every item of the (editable) list, a new thread with a relative name (e.g. "Wanda B1" for electromechanical ) might be opened. This would help information to be filed systematically, in an easy to find way and would prevent this thread from becoming chaotic.
Regards
George
I think it's time for someone with a measurement microphone, "good" speakers and an anechoic chamber 
to measure the acoustic differences between the output from a high quality class D & class AB (would be great to sample to 96Khz too.)
Nothing would be more satisfactory to see measured waveforms & to actually compare them with the original music score (i say music, not a swept sine or pink noise, cos it's what we listen to most often, or anyway what i listen to most often.....)
Obviously this is limited by the measurement mic, but it's a starting point....
I suspect class D is probably just reproducing the signal with the highest clarity, maybe as a result of less dynamic compression & better transient response than an equivalently rated class AB.
cheers
Rob
to measure the acoustic differences between the output from a high quality class D & class AB (would be great to sample to 96Khz too.)Nothing would be more satisfactory to see measured waveforms & to actually compare them with the original music score (i say music, not a swept sine or pink noise, cos it's what we listen to most often, or anyway what i listen to most often.....)
Obviously this is limited by the measurement mic, but it's a starting point....
I suspect class D is probably just reproducing the signal with the highest clarity, maybe as a result of less dynamic compression & better transient response than an equivalently rated class AB.
cheers
Rob
There are some older methods for measuring cone breakup effects that could make it easy to construct a simple experiment to see if there is any change in the breakup pattern.
One is to add some white powder (chalk dust or something similar) to the surface of a cone that is facing upwards. The cone vibrates and chalk dust collects at the points of minimum vibration, since this is often a dynamic effect you need to use a strobe light to photograph the cone surface while offsetting the strobe relative to the audio burst signal. This produces a series of photos mapping the effect of a tone burst. My guess would be that the frequencies of interest are in the 800Hz to 2Khz band.
Other Effects
I think the most likely audible effect of a 300Khz carrier would be to decrease the driver reverberation time (i.e. the driver releases the burst energy quicker), which is the same as saying to increase the energy in the initial wave front. If you simply measure the loudspeaker amplitude response for a swept sine wave you get a nice graph but something that is irreverent because it ignores the emphasis that the ear places on the direct sound vs sound in the reverberant filed. If this is the effect, than it should be heard as an opening up of the sound-stage, which seems to be in line with the Transparence claims. This should be visable on a waterfall plot, so theres some work for someone with a mlssa system.
Personally I've tried testing this theory a few times by adding known HRTF effects and seeing if an unknowing listener actually hears the difference better with a ClassD amp than a similar rated A/B amp, unfortunately the results were very inconclusive.
In my mind the most likely acoustic effect of the carrier would be to create a near field variation in the local air-mass loading on the cone. Any increase in the effective air-mass would result in a more efficient pulse energy transfer, which again results in a reduction in the reverberant energy of the driver, for a given excitation waveform. This could be happening because the cone is effectively acting as a much smaller cone at higher frequencies.
Anyway, that's enough thoughts to keep a PhD student busy for a while. I hope someone with lots of time is reading....
Regards
Rob
One is to add some white powder (chalk dust or something similar) to the surface of a cone that is facing upwards. The cone vibrates and chalk dust collects at the points of minimum vibration, since this is often a dynamic effect you need to use a strobe light to photograph the cone surface while offsetting the strobe relative to the audio burst signal. This produces a series of photos mapping the effect of a tone burst. My guess would be that the frequencies of interest are in the 800Hz to 2Khz band.
Other Effects
I think the most likely audible effect of a 300Khz carrier would be to decrease the driver reverberation time (i.e. the driver releases the burst energy quicker), which is the same as saying to increase the energy in the initial wave front. If you simply measure the loudspeaker amplitude response for a swept sine wave you get a nice graph but something that is irreverent because it ignores the emphasis that the ear places on the direct sound vs sound in the reverberant filed. If this is the effect, than it should be heard as an opening up of the sound-stage, which seems to be in line with the Transparence claims. This should be visable on a waterfall plot, so theres some work for someone with a mlssa system.
Personally I've tried testing this theory a few times by adding known HRTF effects and seeing if an unknowing listener actually hears the difference better with a ClassD amp than a similar rated A/B amp, unfortunately the results were very inconclusive.
In my mind the most likely acoustic effect of the carrier would be to create a near field variation in the local air-mass loading on the cone. Any increase in the effective air-mass would result in a more efficient pulse energy transfer, which again results in a reduction in the reverberant energy of the driver, for a given excitation waveform. This could be happening because the cone is effectively acting as a much smaller cone at higher frequencies.
Anyway, that's enough thoughts to keep a PhD student busy for a while. I hope someone with lots of time is reading....
Regards
Rob
I'm not convinced it can have any affect on the cone as we're talking about milliwatts here, possibly less.
I still think dither is the most reasonable explanation, I notice the same kind of thing just by oversampling a lower sampled recording, it seems to smooth it out and give it a bit of air.
It also seems reasonable it could have some affect of biasing the wire but that would only relate to the high frequencies and time smear the signal, so I also have my reservations about that one.
All the ideas are out there now though, let's let some genius write the patent ..
Regards,
Chris
I still think dither is the most reasonable explanation, I notice the same kind of thing just by oversampling a lower sampled recording, it seems to smooth it out and give it a bit of air.
It also seems reasonable it could have some affect of biasing the wire but that would only relate to the high frequencies and time smear the signal, so I also have my reservations about that one.
All the ideas are out there now though, let's let some genius write the patent ..
Regards,
Chris
Some possible general reasons:
This may be related to feedback loop characteristics: Increase the bandwidth makes the amplifier more likely to oscillate. When removing the oscillation, you also lost some bandwidth (and maybe increasing some TIM distortion).
Self-osc class D amplifiers could go to the edge of instability, while class AB can't.
It may be helpful to listen if triangular carrier based designs have this so-called "class D sound".
Another reason may be the HF oscillation removed the crossover distortion or switching distortion of class AB output stage (not the cables).
Third, some people have their copper speaker wire rusted into grey color at the speaker terminals, this may left volts of breakdown voltage for the signal to overcome.
Note:
When measuring resistance with a digital multimeter, the display number often jumps, even when the surface is almost shining new. Getting a proper connection requires to scrub the target hardly with the probe under these case.
Another case is I had met with dozenes of these DC servo motor: they have internal resistance <10 ohm, but always measures >100 ohm with a digital multimeter.
This is how typical contact points appears to low level signals.
This may be related to feedback loop characteristics: Increase the bandwidth makes the amplifier more likely to oscillate. When removing the oscillation, you also lost some bandwidth (and maybe increasing some TIM distortion).
Self-osc class D amplifiers could go to the edge of instability, while class AB can't.
It may be helpful to listen if triangular carrier based designs have this so-called "class D sound".
Another reason may be the HF oscillation removed the crossover distortion or switching distortion of class AB output stage (not the cables).
Third, some people have their copper speaker wire rusted into grey color at the speaker terminals, this may left volts of breakdown voltage for the signal to overcome.
Note:
When measuring resistance with a digital multimeter, the display number often jumps, even when the surface is almost shining new. Getting a proper connection requires to scrub the target hardly with the probe under these case.
Another case is I had met with dozenes of these DC servo motor: they have internal resistance <10 ohm, but always measures >100 ohm with a digital multimeter.
This is how typical contact points appears to low level signals.
gpapag said:
Interesting ideas here. So let me jump in with both feet.
I can't imagine that the Class D HF carrier is doing much good for the speaker cone. We are talking about 300 KHz and up. Not likely that the cone can respond at all at those frequencies. The Tripath switches at 650 KHz and up. What paper driver would respond to that? And I don't use full range speakers, so only my tweeters would get any benefit from the HF carrier.
The idea of an RF bias in the wires and voice coil does sound cool. It sure works well on tape. Makes a big difference in the sound if you tweak the bias to the tape. I used to do that on my Revox tape machines- easy to hear the difference.
Air is very non-linear in its response. Having spent over 10 years trying to build an "all digital loudspeaker", I can attest to this. Air doesn't do what you think it will at high frequencies. If the RF carrier of a Class D amp is doing something to the air, I'd love to know what.
My guess is that the mundane explanation is the closest to the truth. Class D does not force the transistors to work in a linear manner. Tubes can be very linear, transistors are not. Class D gets around this for the most part.
But I'm still holding out hope for the RF bias idea. Who is going to test it?
I can't imagine that the Class D HF carrier is doing much good for the speaker cone. We are talking about 300 KHz and up. Not likely that the cone can respond at all at those frequencies. The Tripath switches at 650 KHz and up. What paper driver would respond to that? And I don't use full range speakers, so only my tweeters would get any benefit from the HF carrier.
The idea of an RF bias in the wires and voice coil does sound cool. It sure works well on tape. Makes a big difference in the sound if you tweak the bias to the tape. I used to do that on my Revox tape machines- easy to hear the difference.
Air is very non-linear in its response. Having spent over 10 years trying to build an "all digital loudspeaker", I can attest to this. Air doesn't do what you think it will at high frequencies. If the RF carrier of a Class D amp is doing something to the air, I'd love to know what.
My guess is that the mundane explanation is the closest to the truth. Class D does not force the transistors to work in a linear manner. Tubes can be very linear, transistors are not. Class D gets around this for the most part.
But I'm still holding out hope for the RF bias idea. Who is going to test it?
Residual carrier
Yes and this low level residual is probably just enough to be a bias for the low level signals we are talking about. Bias on cables is not new. There are cables supplied with a battery attached. I know, this is not the same at all as HF. The idea of HF bias and what effect it could have on voice coil conductance as well as less than perfect cable connections is worth looking into.
Roger
Yes and this low level residual is probably just enough to be a bias for the low level signals we are talking about. Bias on cables is not new. There are cables supplied with a battery attached. I know, this is not the same at all as HF. The idea of HF bias and what effect it could have on voice coil conductance as well as less than perfect cable connections is worth looking into.
Roger
Let me be very clear I'm talking about small signal response of the system, by this I'm referring to signals in the range of -40dB to -80dB. Assuming a gain of 20dB this equates to signals with an output power range from 1mW to 100mW. Now at large signals a ripple of 500mV on a 25V signal is probably doing next to nothing, however with a classD you get increased ripple at small signals i.e. something around 1000mW of power at 300Khz to 600Khz is added to an audio band signal of 1mW to 100mW. (I'm getting these figures from some simulation results on SODA amps )
http://www.diyaudio.com/forums/showthread.php?s=&postid=778406#post778406
Typically when we perceive a good clean sound stage it is because the small signal effects are not being masked, To understand this you only need to read the theories about how we hear sound in 3D with only two ears. For instance, effects like the delay of direct sound vs sound bouncing of our shoulders plays a big role in determining vertical placement of the sound. When these effects are masked we don't "hear" a difference, all that happens is that we can no longer accurately place where the sound is coming from. For these effects the frequencies of interest are typically in the range 600Hz to 2Khz or maybe up to 5Khz.
If your speakers are 2way, than this refers mostly to the full range base speaker. Even in a 3 way this range is typically handled by a 4in to 6in mid range speaker. It is only in 4 way (and greater) systems that most of the energy will be directed to a speaker that operates in a pistonic mode over the full bandwidth of the transmitted signal. This is why I think cone breakup might be playing a role.
If we have 50mW (1/20 due to crossover) of power at 300kHZ on a 6in mid range driver than it is not going to move, the cone and voice call assembly are just to massive. However this energy has to be dissipated, (unless we want to violate the second law of thermodynamics). Some energy will be go to voice coil heating, some will be dissipated in the magnetic flux modulation and some will dissipate as a vibration, probably a surface wave on the cone.
My first guess was that the audio effect is related to hysteretic movement in a linear motor, so the modulation of the flux, along with mechanical stiction would be the two primary effects.
If this is not the cause than my second guess is that it relates to ease of breakup of the cone when it is vibrating. This probably results in a quicker release of the energy associated with the burst (signal), which is also a very good thing for improved transparence.
My third guess would be that there is some surface acoustic effect probably related to the non-linearity of air (think of it this way if the cone has some surface vibration than 1/2 the air at the surface is probably being compressed meaning the air-mass loading is increased across 1/2 the cone area. Since air mass loading directly effects the mechanical / acoustic coupling interface we could get cleaner pulse transmission with reduced energy in the reverberant field.
But again I could be completely wrong the whole effect could be to do with better damping ratios, I’m not looking for an argument, I’m just trying to explain why an inferior system seems to give better results.
Regards
Rob
http://www.diyaudio.com/forums/showthread.php?s=&postid=778406#post778406
Typically when we perceive a good clean sound stage it is because the small signal effects are not being masked, To understand this you only need to read the theories about how we hear sound in 3D with only two ears. For instance, effects like the delay of direct sound vs sound bouncing of our shoulders plays a big role in determining vertical placement of the sound. When these effects are masked we don't "hear" a difference, all that happens is that we can no longer accurately place where the sound is coming from. For these effects the frequencies of interest are typically in the range 600Hz to 2Khz or maybe up to 5Khz.
If your speakers are 2way, than this refers mostly to the full range base speaker. Even in a 3 way this range is typically handled by a 4in to 6in mid range speaker. It is only in 4 way (and greater) systems that most of the energy will be directed to a speaker that operates in a pistonic mode over the full bandwidth of the transmitted signal. This is why I think cone breakup might be playing a role.
If we have 50mW (1/20 due to crossover) of power at 300kHZ on a 6in mid range driver than it is not going to move, the cone and voice call assembly are just to massive. However this energy has to be dissipated, (unless we want to violate the second law of thermodynamics). Some energy will be go to voice coil heating, some will be dissipated in the magnetic flux modulation and some will dissipate as a vibration, probably a surface wave on the cone.
My first guess was that the audio effect is related to hysteretic movement in a linear motor, so the modulation of the flux, along with mechanical stiction would be the two primary effects.
If this is not the cause than my second guess is that it relates to ease of breakup of the cone when it is vibrating. This probably results in a quicker release of the energy associated with the burst (signal), which is also a very good thing for improved transparence.
My third guess would be that there is some surface acoustic effect probably related to the non-linearity of air (think of it this way if the cone has some surface vibration than 1/2 the air at the surface is probably being compressed meaning the air-mass loading is increased across 1/2 the cone area. Since air mass loading directly effects the mechanical / acoustic coupling interface we could get cleaner pulse transmission with reduced energy in the reverberant field.
But again I could be completely wrong the whole effect could be to do with better damping ratios, I’m not looking for an argument, I’m just trying to explain why an inferior system seems to give better results.
Regards
Rob
Lots of (in my eyes) silly arguments here ......
I tested the UcD output of my DFR9000 on my T+A TMR100 speakers. 3 meters of 1,5 mm2 in between.
The Philips puts 800 mV pp on ca. 300 kHz (300 mV RMS) on the speaker terminals.
Adding a simple parallel resonance circuit reduced this "noise" to about 40 mV pp.
No difference in soundstage or whatever . Slightly different bass, probably due to the coils that I used in the resonance circuit (0,5 Ohm).
Yes, I like thin cables, but the 0,5 Ohm forces me I should probably get some more copper in the speaker wire.
The measurement was done with a Tektronix 465B and yes I removed the resonant circuits again.
I tested the UcD output of my DFR9000 on my T+A TMR100 speakers. 3 meters of 1,5 mm2 in between.
The Philips puts 800 mV pp on ca. 300 kHz (300 mV RMS) on the speaker terminals.
Adding a simple parallel resonance circuit reduced this "noise" to about 40 mV pp.
No difference in soundstage or whatever . Slightly different bass, probably due to the coils that I used in the resonance circuit (0,5 Ohm).
Yes, I like thin cables, but the 0,5 Ohm forces me I should probably get some more copper in the speaker wire.
The measurement was done with a Tektronix 465B and yes I removed the resonant circuits again.
classd4sure said:
Interesting comparison.
But the way the "good behaviour" of the class D amp is discussed reminds me more of the guys that tried to put a machine in the air with flapping panels.
Ups, some Dutch did that recently .....
I'm not off for some new ideas, but this discussion reminds me to much of the chaos theory explaning tropical storms are derived from a butterfly taking off somewhere.
marconist said:
Interesting comparison.
But the way the "good behaviour" of the class D amp is discussed reminds me more of the guys that tried to put a machine in the air with flapping panels.
Ups, some Dutch did that recently .....
I'm not off for some new ideas, but this discussion reminds me to much of the chaos theory explaning tropical storms are derived from a butterfly taking off somewhere.
Good that you are sceptic, but do you agree that Class D has something special (soundwise)? At least I do and as an engineer, I am very interested to know why Class D sounds different than other amps I have heard so far. There must be some explanation for it, we just want to find out. I agree that the proposed theories may sound weird and I do not know whether they make sense yes or no, but at least we get a number of ideas why and maybe at some stage we find the answer.
Hope you can come up with some hints and more experiments.
Best regards
Gertjan
That's not a commonality between all class d amps, as far as I know it is more a characteristic better sounding amps share, of any class.
I think this from ClassDunce:
"Let me be very clear I'm talking about small signal response of the system, by this I'm referring to signals in the range of -40dB to -80dB. Assuming a gain of 20dB this equates to signals with an output power range from 1mW to 100mW. Now at large signals a ripple of 500mV on a 25V signal is probably doing next to nothing, however with a classD you get increased ripple at small signals i.e. something around 1000mW of power at 300Khz to 600Khz is added to an audio band signal of 1mW to 100mW."
Is a reasonable explanation. Dither was another way of saying it.
This could possibly make the wires look more ideal, right up to the voice coil, having the effect of reducing "stiction" as it were, or making it robust against non idealities which would normally smear and damp the microdetails in the signal which is everything to high end sound.
Aside from that I still don't think it can act on the cone at all, aside from the stiction aspect, possibly having the effect of reducing mass for smaller signals.
Had I the means the above is where I'd concentrate my experiments.
I think this from ClassDunce:
"Let me be very clear I'm talking about small signal response of the system, by this I'm referring to signals in the range of -40dB to -80dB. Assuming a gain of 20dB this equates to signals with an output power range from 1mW to 100mW. Now at large signals a ripple of 500mV on a 25V signal is probably doing next to nothing, however with a classD you get increased ripple at small signals i.e. something around 1000mW of power at 300Khz to 600Khz is added to an audio band signal of 1mW to 100mW."
Is a reasonable explanation. Dither was another way of saying it.
This could possibly make the wires look more ideal, right up to the voice coil, having the effect of reducing "stiction" as it were, or making it robust against non idealities which would normally smear and damp the microdetails in the signal which is everything to high end sound.
Aside from that I still don't think it can act on the cone at all, aside from the stiction aspect, possibly having the effect of reducing mass for smaller signals.
Had I the means the above is where I'd concentrate my experiments.
marconist
You can put the blame (credibility deficit) on me.
You may very well question the case. Subject is still open and will remain that way. The following from your post #33
regards
George
You can put the blame (credibility deficit) on me.
You may very well question the case. Subject is still open and will remain that way. The following from your post #33
is a valid contribution to the thread.It is an opinion expressed after a subjective listening test and some measurements. If there wil be x tenths of such opinions your opinion will be strengthened. But the questionmark will be still here.
regards
George
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.