Undistorted? I think you can not make a 100W amplifier, 5V/uS slew rate, with THD below 0,01% at 20kHz, 40V peak, 8 Ohm load. Even in simulation....
It does. In almost all amps the slew rate limit is set by some signal current (which has a limit) charging a capacitance. As the signal current increases, but below the limit, it is likely that distortion increases in whatever active device is providing that signal current. How this happens depends on the amp topology, which is why people with different amps argue about what slew rate limit is needed.mchambin said:
If 5V/us is enough for an undistorted 20kHz at 40V peak, then if you could make an amp which could do this with low distortion then it would be fine for audio. It is difficult because of the distortion generated on the approach to slew rate limiting because of the extra signal current needed. Hence we go for more slew rate - and then some people mistakenly think that their amp/ears/speakers are capable of appreciating 50V/us. You may need a design level of 50V/us in order to ensure that the 5V/us actually needed by the signal can be delivered with low distortion. In reality you may need less than this as the maximim slew rate seen in any real audio signal corresponds to full amplitude sine at some kHz, not 20kHz.
So you have a 50V/us amp to deliver 5V/us cleanly, but in reality your music never has more than 2V/us in the signal - but bigger numbers look better in marketing or bragging.
Do signals we listen for our pleasure ever exceed the slew rate of a full amplitude of a 2 kHz sine wave ? Are the volume of an audio system often set such as the maximal output of a power amplifier is reached ?
For the lab, a low harmonic distortion at 10 kHz is a largely sufficient clue showing an adequate slew-rate handling. It's a criterion used by Bob Cordell, I think.
Agree.
But if you can make 90V/uS amp with low THD, and stable, why limit it for lower slew rate?
In electronics, as in all engineering, increasing one parameter always carries the risk of reducing another one either because physics says so or because the attention of the designer has been diverted. Hence there is a need for caution in aiming higher than necessary, especially after a margin for error has already been included.bimo said:
For example, what if raising the slew rate means raising quiescent current but that means more noise? How do you balance noise against slew rate? Or raise slew rate by reducing the relevant capacitance but that leads to instability with some loads. How do you balance load tolerance against slew rate?
If you use same method of ancient technology, I agree
Just make amplifier with usual topology, blameless topology with usual compensation, the ordinary Miller compensation. You will go nowhere. If you more creative and clever, you can make it.Slew rate requirements are driven by frequency (which drives the dV/dt) and peak amplitude, and fortunately the distribution of these two entities in real music are strongly inversely proportional.
IOW, high amplitudes happen at low frequencies, and high frequencies have very limited amplitude.
IOW, high amplitudes happen at low frequencies, and high frequencies have very limited amplitude.
It doesn't really work that way. Real transients are not actually just combinations of sine waves that slowly drop off with frequency, although a spectrum analyzer seems to imply it. They are caused by banging things together normally, or momentary clipping of a signal channel.
To understand this, just note what the rise-time of a 100Hz SQUARE WAVE can be. It can be almost infinite, YET the spectrum of the 100 Hz square wave looks pretty tame.
How about my scope calibration square wave? What do you think a 1ns rise-time does to audio equipment if you are not careful? Yet the spectrum analyzer will show typical musical rolloff.
No, you NEED a defined rise-time that is usually related to the closed loop bandwidth of the amp under test, and it could be fast enough to cause internal slew rate limiting of the very same amp. This is caused by high feedback in amps. Zero feedback amps do not have this problem. Usually we define a 10us rise-time as an appropriate test signal for testing solid state audio products.
To understand this, just note what the rise-time of a 100Hz SQUARE WAVE can be. It can be almost infinite, YET the spectrum of the 100 Hz square wave looks pretty tame.
How about my scope calibration square wave? What do you think a 1ns rise-time does to audio equipment if you are not careful? Yet the spectrum analyzer will show typical musical rolloff.
No, you NEED a defined rise-time that is usually related to the closed loop bandwidth of the amp under test, and it could be fast enough to cause internal slew rate limiting of the very same amp. This is caused by high feedback in amps. Zero feedback amps do not have this problem. Usually we define a 10us rise-time as an appropriate test signal for testing solid state audio products.
Last edited:
Except we're dealing with band-limited input. Sure you can generate all kinds of horrible signals with an arbitrary waveform generator, but what's coming off of a DAC post-filter? Or from a cartridge? I'd worry more about fast recovery from a pop/glitch on vinyl than having insane power bandwidth (unless you can get both easily).
Agreed, clipping changes the picture, as it effectively introduces signals which can have high frequency and high amplitude combinations.
But that is not normal music spectrum anymore.
If you get into clipping, then low slew rate and high NFB can get you into trouble, but if you hit clipping you're half in trouble anyway.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.