Is there such a thing as a "predictive" rail switching (what some would call class G) amplifier, as opposed to a "basic" rail switching amplifier?
I've read in previous posts that the Carver M400 uses a predictive rail switching system (is this referring to their magnetic input circuit for the AC line, or is it referring to their output stage?). QSC Audio, in a white paper on their Powerlight amplifiers, states that their rail switching system:
"...a predictive switching scheme guarantees that each supply voltage will be switched on just before the signal needs it, and switched off as soon as possible after the signal falls, in order to squeeze maximum efficiency from the four-tier supply." (To me, this sounds like what any regular rail switching amplifier does).
To me, the word "predictive" sounds like a rail switching output stage where the high voltage rails are switched on and off based on some linear prediction algorithm. Then again, predictive could be a juicy marketing word used to spice up a product.
I've read in previous posts that the Carver M400 uses a predictive rail switching system (is this referring to their magnetic input circuit for the AC line, or is it referring to their output stage?). QSC Audio, in a white paper on their Powerlight amplifiers, states that their rail switching system:
"...a predictive switching scheme guarantees that each supply voltage will be switched on just before the signal needs it, and switched off as soon as possible after the signal falls, in order to squeeze maximum efficiency from the four-tier supply." (To me, this sounds like what any regular rail switching amplifier does).
To me, the word "predictive" sounds like a rail switching output stage where the high voltage rails are switched on and off based on some linear prediction algorithm. Then again, predictive could be a juicy marketing word used to spice up a product.
most "predictive" rail switching schemes simply switch in the higher rail when the output gets within a few volts of the lower rail. you will hear about amps that do this not sounding good, which in some cases can be true, since it has an effect similar to crossover distortion if not properly done. class f amps use a linear scheme (variable higher rails that follow the envelope of the audio), and there are amps that use a PWM circuit in the power supply do do this also. the class f is more "predictive" since the rails rise a little faster than the audio envelope, but there is no "magic" here, just a bit higher gain in the power supply controller than exists in the amp driver. remember, most of these amps that pioneered these extended classes were engineered at a time when almost everything was done analog. class d amps were invented around the same time, but IIRC the only commercially produced D amp at the time (Sony) had a clock frequency of 50Khz, and some very interesting (pyrotechnic) failure modes .
The switching procedure itself takes some time to start, and also some more time to be completed because it's usually done at a controlled dV/dt slope to keep electromagnetic interference well under control. On the other hand, the requested amplifier output voltage (the music signal) is continuously changing, either rising or falling. As a consequence, leaving enough voltage margin for proper switching of trebble frequencies to avoid temporary clipping causes premature switching with lower frequencies and reduces the overall efficiency of the system well below its theoretical maximum.
Predictive, frequency-sensitive (actually slew-rate-sensitive) switching, is easily achieved by adding frequency compensation to the voltage sense mechanism, so that higher frequencies appear to have higher amplitudes in order to trigger the switching event sooner.
Predictive, frequency-sensitive (actually slew-rate-sensitive) switching, is easily achieved by adding frequency compensation to the voltage sense mechanism, so that higher frequencies appear to have higher amplitudes in order to trigger the switching event sooner.
There are two "simple" ways of doing this that I can think off.
Essentially, you use a diode from the low voltage supply to the output decoupling cap, and a regulator from the high voltage supply to this same cap.
Method 1 involves having the regulator follow the signal voltage with a DC offset, down to a minimum of the basic voltage rail (to avoid reverse bias). This gives a constant Vce during high voltage swings, which may or may not be what you want.
Method 2 just drives the regulator by following the input but with higher gain, again with a minimum of the basic voltage rail. This is slightly less efficient, and gives a Vce that follows the input signal. Again, may or may not be what you want. Somewhat simpler to build.
Essentially, you use a diode from the low voltage supply to the output decoupling cap, and a regulator from the high voltage supply to this same cap.
Method 1 involves having the regulator follow the signal voltage with a DC offset, down to a minimum of the basic voltage rail (to avoid reverse bias). This gives a constant Vce during high voltage swings, which may or may not be what you want.
Method 2 just drives the regulator by following the input but with higher gain, again with a minimum of the basic voltage rail. This is slightly less efficient, and gives a Vce that follows the input signal. Again, may or may not be what you want. Somewhat simpler to build.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.