I think you shoud add a timer so when the comparator fires the light stays on long enough for you to see it. Also add a steep low pass filter so the clock noise doesn't throw off the circuit.
I've seen quite a few good limiter circuits posted in these forums. That might be a better idea.
I would monitor the input for when the signal is high enough to cause the amp the clip. There are many input monitoring methods and multiple LEDs. These are usually easy to adjust to your listening tastes.
Just for class-d, I suppose you could compare the percentage of modulation. That would make the circuit special just for class-d.
I've seen quite a few good limiter circuits posted in these forums. That might be a better idea.
I would monitor the input for when the signal is high enough to cause the amp the clip. There are many input monitoring methods and multiple LEDs. These are usually easy to adjust to your listening tastes.
Just for class-d, I suppose you could compare the percentage of modulation. That would make the circuit special just for class-d.
I have been using this one, I have simulated it first to get on idea on what resistors and capacitors to use, and it was pretty close in real circuit
But if you want to go all out,
Power Amp Clipping Indicator
would be great... for upper one, I am not sure, if you have rail tracking or not, circuit is similar, but can't remember who did this one
But if you want to go all out,
Power Amp Clipping Indicator
would be great... for upper one, I am not sure, if you have rail tracking or not, circuit is similar, but can't remember who did this one
Short answer... Using the speaker output potential, there should be no difference in clip detection (and level of concern) between class D and analog classes. As long as you look at the (filtered) speaker output, NOT the PWM (pre-filter) potential, which is ALWAYS clipped. This is readily known since you don't purchase special class D speakers to use with your class D amp, and we care about inaudible clipping on account of the speakers.
Not long-enough winded? Read on...
Feeling the need to chime in here even though I just read this 2-month old thread... I think there are 2 major aspects to clipping. (1) One is the literal ELECTRICAL clipping which is the signal voltage's instantaneous "change of course" (on Voltage vs. Time graph paper) upon collision with the supply voltage (minus any and all Vds, Vce, etc. penalties in series). Definition could get pretty long considering tubes with their pre-clip "compression" tendency, and MOSFETs which probably "CLIP" at higher Vds than the average bipolar saturation Vbe. Ultimately within the analog circuit, clipping could be loosely defined as that signal excursion that takes the output devices out of their wheelhouse, "uncomfortably" close to the supply voltage such that device performance changes (diminishes) as a result of that closeness. This could be very close to, or many volts away from supply depending on the devices and topology. Detection on (1) basis is via voltage approach criteria.
(2) The other aspect is the SYNTHESIS of clipping, arising from the fact that ALL signals may be expressed as a summation of transcendental functions of varying phase, frequency, and amplitude. The instantaneous "changes of course" at clipping embody Fourier syntheses of what may be extremely high frequencies (derivative of waveform approaches discontinuity). This excess HF energy is what damages tweeters (because the filters direct it there). Arguably a third subjective aspect could be described in that superfluous 'information' is supplied with the outgoing signal, which does not resemble the musical input, but this is not a direct factor in clip detection. Detection on (2) basis could probably be implemented by sampling for spurious HF.
WITHIN ITSELF, a class D amp basically 'CLIPS' full time by definition (1), because output devices are constantly alternating between supplying MIN and MAX signal levels (preferably much more quickly than the ultimate musical signals so that a filter can separate the two forms of information). In a sense, class D pushes WELL BEYOND clipping during each non-zero duty cycle since in order to realize the benefits of class D, the transistors need to alternate between CURRENT with ZERO VOLTAGE and VOLTAGE with ZERO CURRENT as efficiently as possible. They heat because of slewing, during which VOLTAGE and CURRENT are both present. It suffices to say that at times, the difference between PWM output and supply voltage is at the absolute minimum available. By this reasoning it is pointless to consider clipping in class D because an indicator will function about like your "POWER ON" lamp. It would vary in brightness since modulated analog signals dictate varied duty cycles (time spent in CLIP at the pulse width modulation (PWM) frequency).
Speakers don't speak fluent PWM. Ultimately, the signals must be filtered to limit HF to audible range and reduce HF energy delivery. To consider clipping in THIS realm ultimately, high enough PWM (class D) duty cycle should be able to drive the filter such that its output resembles that of a clipping analog amplifier.
Point is, a class D amplifier must be built (i.e. inclusive of low pass filter) so in a sense the speakers can't tell they are being driven by something besides pure analog signals, so ultimately the analog clipping indicator should work just the same out of any amplifier, provided the evaluated signal is the final (20-20kHz) one going to the speakers. Fooling with the criteria points would yield optimum results. I recommend ESP's design at the beginning of this thread, but at the expense of elegance, I personally favor buffering the voltage dividers to the transistor emitters.
NOW... A final thought... Since class D amplifiers utilize pulse-oriented modulation and require low-pass filters to deliver only musical energy, it seems that assuming these filters are tuned "just right", CLIPPING (2) is obviated since the HF energy that would be dangerous clipping can't be delivered through the filter. CLIPPING (1) should be preserved for sufficiently long duty cycles... While HF energy does not transit the filter, there is nothing in it that prevents its output from being equal to the supply voltage. Each 'CLIP' event should comprise an exponential decay toward supply voltage at the filter time constant. It would be DISTORTED, but it MIGHT not damage tweeters. I'd say some experimentation with DUMMY loads and RTA is required to determine whether a clipping indicator is needed.
Not long-enough winded? Read on...
Feeling the need to chime in here even though I just read this 2-month old thread... I think there are 2 major aspects to clipping. (1) One is the literal ELECTRICAL clipping which is the signal voltage's instantaneous "change of course" (on Voltage vs. Time graph paper) upon collision with the supply voltage (minus any and all Vds, Vce, etc. penalties in series). Definition could get pretty long considering tubes with their pre-clip "compression" tendency, and MOSFETs which probably "CLIP" at higher Vds than the average bipolar saturation Vbe. Ultimately within the analog circuit, clipping could be loosely defined as that signal excursion that takes the output devices out of their wheelhouse, "uncomfortably" close to the supply voltage such that device performance changes (diminishes) as a result of that closeness. This could be very close to, or many volts away from supply depending on the devices and topology. Detection on (1) basis is via voltage approach criteria.
(2) The other aspect is the SYNTHESIS of clipping, arising from the fact that ALL signals may be expressed as a summation of transcendental functions of varying phase, frequency, and amplitude. The instantaneous "changes of course" at clipping embody Fourier syntheses of what may be extremely high frequencies (derivative of waveform approaches discontinuity). This excess HF energy is what damages tweeters (because the filters direct it there). Arguably a third subjective aspect could be described in that superfluous 'information' is supplied with the outgoing signal, which does not resemble the musical input, but this is not a direct factor in clip detection. Detection on (2) basis could probably be implemented by sampling for spurious HF.
WITHIN ITSELF, a class D amp basically 'CLIPS' full time by definition (1), because output devices are constantly alternating between supplying MIN and MAX signal levels (preferably much more quickly than the ultimate musical signals so that a filter can separate the two forms of information). In a sense, class D pushes WELL BEYOND clipping during each non-zero duty cycle since in order to realize the benefits of class D, the transistors need to alternate between CURRENT with ZERO VOLTAGE and VOLTAGE with ZERO CURRENT as efficiently as possible. They heat because of slewing, during which VOLTAGE and CURRENT are both present. It suffices to say that at times, the difference between PWM output and supply voltage is at the absolute minimum available. By this reasoning it is pointless to consider clipping in class D because an indicator will function about like your "POWER ON" lamp. It would vary in brightness since modulated analog signals dictate varied duty cycles (time spent in CLIP at the pulse width modulation (PWM) frequency).
Speakers don't speak fluent PWM. Ultimately, the signals must be filtered to limit HF to audible range and reduce HF energy delivery. To consider clipping in THIS realm ultimately, high enough PWM (class D) duty cycle should be able to drive the filter such that its output resembles that of a clipping analog amplifier.
Point is, a class D amplifier must be built (i.e. inclusive of low pass filter) so in a sense the speakers can't tell they are being driven by something besides pure analog signals, so ultimately the analog clipping indicator should work just the same out of any amplifier, provided the evaluated signal is the final (20-20kHz) one going to the speakers. Fooling with the criteria points would yield optimum results. I recommend ESP's design at the beginning of this thread, but at the expense of elegance, I personally favor buffering the voltage dividers to the transistor emitters.
NOW... A final thought... Since class D amplifiers utilize pulse-oriented modulation and require low-pass filters to deliver only musical energy, it seems that assuming these filters are tuned "just right", CLIPPING (2) is obviated since the HF energy that would be dangerous clipping can't be delivered through the filter. CLIPPING (1) should be preserved for sufficiently long duty cycles... While HF energy does not transit the filter, there is nothing in it that prevents its output from being equal to the supply voltage. Each 'CLIP' event should comprise an exponential decay toward supply voltage at the filter time constant. It would be DISTORTED, but it MIGHT not damage tweeters. I'd say some experimentation with DUMMY loads and RTA is required to determine whether a clipping indicator is needed.
Wanted to mention vs. raintalk's indication of measuring timing and comparing duty cycle to some reasonable maximum and calling it clipping if it is exceeded... It is a 'sound' plan and can be nicely implemented in the digital domain, but I think it does not consider loading effects... I do like the idea of doing class D stuff in class D amps though.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.