This Danish thing is indeed one of the self oscillating versions of a pwm amp.
The ICE power modules by B&O are self oscillating and also some amplifiers by Philips.
The oscillating frequency of the above circuit is determined by the integrators timeconstant and the following schmitt-trigger's hysteresys voltage (it is in fact the very same principle that you use when building multivibrators with OP-AMP, one capacitor and three resistors).
Some more modern self oscillating amps make use of phase shift alone to determine the oscillating frequency (i.e. they use a comparator rather than a Scmitt-trigger). Some proactively include the output filter in order to achieve the necessary phase-shift to achieve reliable oscillation (specially the Philips one , which some even claim to be one of the best amplifier principles available today).
Regards
Charles
The ICE power modules by B&O are self oscillating and also some amplifiers by Philips.
The oscillating frequency of the above circuit is determined by the integrators timeconstant and the following schmitt-trigger's hysteresys voltage (it is in fact the very same principle that you use when building multivibrators with OP-AMP, one capacitor and three resistors).
Some more modern self oscillating amps make use of phase shift alone to determine the oscillating frequency (i.e. they use a comparator rather than a Scmitt-trigger). Some proactively include the output filter in order to achieve the necessary phase-shift to achieve reliable oscillation (specially the Philips one , which some even claim to be one of the best amplifier principles available today).
Regards
Charles
I personally don't like the approach, though it has two major advantages:
1.) simplicity
2.) higher NFB factor due to the fact that the unity gain point is a the switching frequency rather than half of it as it is the case for an ordinary PWM amp.
The disadvantage is that you can't synchronise multiple amps to the same frequency (or even to an integer multiple or divider of an SMPS switching frequency), which could reduce IM problems on the supply rails (which in turn can feed through to the output).
As you can see, I used a quartz controlled square-wave oscillator whose output was made available to off-board devices.
Regards
Charles
1.) simplicity
2.) higher NFB factor due to the fact that the unity gain point is a the switching frequency rather than half of it as it is the case for an ordinary PWM amp.
The disadvantage is that you can't synchronise multiple amps to the same frequency (or even to an integer multiple or divider of an SMPS switching frequency), which could reduce IM problems on the supply rails (which in turn can feed through to the output).
As you can see, I used a quartz controlled square-wave oscillator whose output was made available to off-board devices.
Regards
Charles
NFB factor is the gain ratio between open-loop gain and closed-loop gain. It determines by how much THD, IMD and output impedance is reduced by NFB.
The unity-gain point of a class-d amp (being a device that does discrete-time processing) is defined by it's switching frequency.
In simple terms: If you can move the unity-gain point up by one octave (which is automatically the case for the self-oscillating principle) you will achieve an additional 6dB of feedback for a single first-order feedback loop.
Regards
Charles
The unity-gain point of a class-d amp (being a device that does discrete-time processing) is defined by it's switching frequency.
In simple terms: If you can move the unity-gain point up by one octave (which is automatically the case for the self-oscillating principle) you will achieve an additional 6dB of feedback for a single first-order feedback loop.
Regards
Charles
"The unity-gain point of a class-d amp (being a device that does discrete-time processing) is defined by it's switching frequency.
In simple terms: If you can move the unity-gain point up by one octave (which is automatically the case for the self-oscillating principle) you will achieve an additional 6dB of feedback for a single first-order feedback loop."
I find this whole thread to be very interesting. I had a couple of questions.
Regarding the "unity gain point", what exactly is this defined in reference to? For the conventional PWM PS or amp, I assume the duty-cycle "d" is the input. I believe the gist of what you are saying is that the response of the power stage to changes in "d" decreases with increasing frequency of "d". Is not the response more limited by the output filter, or are you talking about a response of the averaged power waveform before an output filter?
Also, is the frequency of the self-oscillating system defined, or does it vary with input signal? The input is no longer "d", but instead a reference voltage to an integrator, if I understand correctly.
Thanks for an interesting discussion,
John
In simple terms: If you can move the unity-gain point up by one octave (which is automatically the case for the self-oscillating principle) you will achieve an additional 6dB of feedback for a single first-order feedback loop."
I find this whole thread to be very interesting. I had a couple of questions.
Regarding the "unity gain point", what exactly is this defined in reference to? For the conventional PWM PS or amp, I assume the duty-cycle "d" is the input. I believe the gist of what you are saying is that the response of the power stage to changes in "d" decreases with increasing frequency of "d". Is not the response more limited by the output filter, or are you talking about a response of the averaged power waveform before an output filter?
Also, is the frequency of the self-oscillating system defined, or does it vary with input signal? The input is no longer "d", but instead a reference voltage to an integrator, if I understand correctly.
Thanks for an interesting discussion,
John
With unity gain point I mean the frequency, where the gain of the feedback LOOP drops to 1. The phase-shift at this frequency determines the so called phase margin (i.e. the amount missing to 180 degrees) and hence by a large part the step response of the system.
The desired phase margin is 90 degrees usually though this is hard to achieve if you take feedback from the filter.
If you take it before the filter, this target is quite hard to MISS (something designers of linear amplifiers only dream of !!).
The frequency of most self-scillating designs is quite independant of input signal.
The switching frequency of delta-sigma amplifiershowever is heavily signal dependant.
Regards
Charles
The desired phase margin is 90 degrees usually though this is hard to achieve if you take feedback from the filter.
If you take it before the filter, this target is quite hard to MISS (something designers of linear amplifiers only dream of !!).
The frequency of most self-scillating designs is quite independant of input signal.
The switching frequency of delta-sigma amplifiershowever is heavily signal dependant.
Regards
Charles
Gotcha. I'm familiar with basic analog linear feedback methods, but your response clarifies what you meant.
It is difficult with an analog loop to take the feedback before the filter, because you need to get rid of the switching ripple somehow. I suppose a separate filter might be easier to design since it does not need to handle the power, but then you are not including the output filter in the loop. This may be fine
I have little familiarity with discrete methods, although I'm doing some studying up on them. The self-oscillating amp in the link mentioned earlier looks more like a delta-sigma converter to me. Is this correct, or am I missing the boat somewhere?
I have designed self-oscillating resonant inverters whose frequency was quite stable, but these used a different method than that referred to here.
John
It is difficult with an analog loop to take the feedback before the filter, because you need to get rid of the switching ripple somehow. I suppose a separate filter might be easier to design since it does not need to handle the power, but then you are not including the output filter in the loop. This may be fine
I have little familiarity with discrete methods, although I'm doing some studying up on them. The self-oscillating amp in the link mentioned earlier looks more like a delta-sigma converter to me. Is this correct, or am I missing the boat somewhere?
I have designed self-oscillating resonant inverters whose frequency was quite stable, but these used a different method than that referred to here.
John
The following old Sony patent describes quite well how the feedback of non-self-oscillating PWM amps works (including mine).
http://l2.espacenet.com/espacenet/viewer?PN=AU8738275&CY=ch&LG=de&DB=EPD
Regards
Charles
http://l2.espacenet.com/espacenet/viewer?PN=AU8738275&CY=ch&LG=de&DB=EPD
Regards
Charles
"To me, such a self-oscillating PWM amp looks more or less like a high-power square-wave generator having a duty-cycle-control- voltage input."
Charles,
I will have to take a closer look. I admit not having analyzed the circuit. As much as I try, my company has little interest in audio-band amplifiers, so it will be a while before I get to it.
Thanks for the additional link.
Best regards,
John
Charles,
I will have to take a closer look. I admit not having analyzed the circuit. As much as I try, my company has little interest in audio-band amplifiers, so it will be a while before I get to it.
Thanks for the additional link.
Best regards,
John
Charles;
"The frequency of most self-scillating designs is quite independant of input signal.
The switching frequency of delta-sigma amplifiershowever is heavily signal dependant."
How would you describe the difference between a self-oscillating PWM amp and a delta-sigma PWM amp.?
And the "soerensen" amp would then be a delta-sigma , right??
"To me, such a self-oscillating PWM amp looks more or less like a high-power square-wave generator having a duty-cycle-control- voltage input"
Do I detect a certain less positive attitude towards this technology
Well ,as I see it, it is a rather elegant way of dealing with the problem regarding nonlinearities in the triangular waveform, and this I belive is a major problem in digital amps.
For what it is worth, iI have auditioned the TacT at several occasions, but at shows and with unfamiliar equipment and surroundings, and it was very good sounding. Just how good I can´t say.
I have in house the ICE power modules, and I had the oppertunity to compare these with the LC-audio modules and a High-end tube/MOSFet hybrid under controlled conditions and known equipment:
BOW ZZ-8 CD player
DVD-rom with upsampling prototype DAC (sounded best)
The original developed prototype of the Gryphon loudspeakers
The two PWM amps. were sounding suprisingly good AND suprisingly different: ICE power was a little lazy/undynamic sounding, but with excellent bas.
LC-audio was very energetic, but also a bit flat and strident sounding.
None of them, however, could match, overall, the tube/MOSfet hybrid even though the ICE power was the best in the bass.
The hybrid is , I should mention, a VERY good sounding amp.!!
I belive that the technology has very great potential even when it comes to sound quality per se, as the ICE power modules uses less than stellar op-amps and passive components and has not been developed from an audiophile standpoint.
I still hope I haven´t turned the focus of this thread away from the original subject??
Koldby
"The frequency of most self-scillating designs is quite independant of input signal.
The switching frequency of delta-sigma amplifiershowever is heavily signal dependant."
How would you describe the difference between a self-oscillating PWM amp and a delta-sigma PWM amp.?
And the "soerensen" amp would then be a delta-sigma , right??
"To me, such a self-oscillating PWM amp looks more or less like a high-power square-wave generator having a duty-cycle-control- voltage input"
Do I detect a certain less positive attitude towards this technology
Well ,as I see it, it is a rather elegant way of dealing with the problem regarding nonlinearities in the triangular waveform, and this I belive is a major problem in digital amps.
For what it is worth, iI have auditioned the TacT at several occasions, but at shows and with unfamiliar equipment and surroundings, and it was very good sounding. Just how good I can´t say.
I have in house the ICE power modules, and I had the oppertunity to compare these with the LC-audio modules and a High-end tube/MOSFet hybrid under controlled conditions and known equipment:
BOW ZZ-8 CD player
DVD-rom with upsampling prototype DAC (sounded best)
The original developed prototype of the Gryphon loudspeakers
The two PWM amps. were sounding suprisingly good AND suprisingly different: ICE power was a little lazy/undynamic sounding, but with excellent bas.
LC-audio was very energetic, but also a bit flat and strident sounding.
None of them, however, could match, overall, the tube/MOSfet hybrid even though the ICE power was the best in the bass.
The hybrid is , I should mention, a VERY good sounding amp.!!
I belive that the technology has very great potential even when it comes to sound quality per se, as the ICE power modules uses less than stellar op-amps and passive components and has not been developed from an audiophile standpoint.
I still hope I haven´t turned the focus of this thread away from the original subject??
Koldby
Hi Koldby
A delta-sigma amp like the Sharp SM-SX100 for instance is outputting an RF-noise like switching signal and not a pulse-width modulated rectangular of constant frequency.
I once posted such apattern within another thread but I have to find it again.
The "sorensen" amp is an ordinary self-oscillating PWM amp IMO.
What I wanted to express is just a short and simple description of how such a self-oscillating PWM actually works like.
I personally don't think that the triangular linearity is the main culprit for the class-d problems, since this can be generated with quite good accuracy.
Regards
Charles
A delta-sigma amp like the Sharp SM-SX100 for instance is outputting an RF-noise like switching signal and not a pulse-width modulated rectangular of constant frequency.
I once posted such apattern within another thread but I have to find it again.
The "sorensen" amp is an ordinary self-oscillating PWM amp IMO.
What I wanted to express is just a short and simple description of how such a self-oscillating PWM actually works like.
I personally don't think that the triangular linearity is the main culprit for the class-d problems, since this can be generated with quite good accuracy.
Regards
Charles
The simulated DS signal can be found here:
http://www.diyaudio.com/forums/attachment.php?s=&postid=179974
Be aware to watch it at 100% size or it will look quite strange.
Regards
Charles
http://www.diyaudio.com/forums/attachment.php?s=&postid=179974
Be aware to watch it at 100% size or it will look quite strange.
Regards
Charles
Charles;
No this I cannot do, but that is irrelevant in my oppinion, as the errors from a gain stage is nonlinerarities that is very difficult for the ears to detect (second order harmonics can be quite high in an gain stage and be very accurate sounding all the same) but nonlinearities in the triangular wave form is not related to the musical signal an thereby much more detectable to the ear.
To go from 16 to 20 bits of resolution in a PCM signal is easily detectable for the ear, even through an amplifier having .1 % harmonic distortion!!
Koldby
No this I cannot do, but that is irrelevant in my oppinion, as the errors from a gain stage is nonlinerarities that is very difficult for the ears to detect (second order harmonics can be quite high in an gain stage and be very accurate sounding all the same) but nonlinearities in the triangular wave form is not related to the musical signal an thereby much more detectable to the ear.
To go from 16 to 20 bits of resolution in a PCM signal is easily detectable for the ear, even through an amplifier having .1 % harmonic distortion!!
Koldby
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.