A mere couple of volts for the output on a 30R load with a circuit
that has about 12.7dB CLG cant be extrapolated for a power amplifier.
What happens if output voltage is 20dB higher with a CLG of say 30dB
and with a 8R load that will then pump almost two order of magintude
more current than your phone amp load, all conditions that are the ones
of a 25-50W output power.?.
Expect the numbers to be much less appaling than -100dB THD ,
actualy for a phone amp much better numbers are possible with
LTP based designs.
And this is where THD isn't telling much. The thing what dictates accuracy of transients is actual speed/slewrate. I prefer to call this reproduction resolution.
Imagine this. We're going a bit digital here. Imagine you have a 2-bit DAC. With this, you can produce something that looks like a sine but isn't, it's an awful ugly blocky thing you'll see, like stairs going up and and down.
The more bits you add, the more accurate the DAC can reproduce a sine. It's the bit resolution that defines the accuracy of the the sine wave. Take note that I am not talking about bandwidth or speed here, for all I care each DAC step takes a second. I'm trying to point out the concept of resolution.
Now, when we look at a square in the time domain and let's take a look at what a feedback amp does: The amp constantly tries to mimic the input signal through feedback/comparisson of two signals. When stability is not an issue and there is no overcompensation, you will see that the amp output rings at the end of each input step (either rising or falling edge). This ringing "dies out", called the settle-time. The frequency of the ringing is what I call the regulatory frequency of the feedback loop. Even when the ringing has died out, you can imagine the amp to be oscillating at the regulatory frequency with an amplitude of zero, always.
Now we are getting at the reproduction resolution I spoke of. The higher the regulatory frequency of the loop, the more accurate will be the transient reproduction of the signal. The defining factor here, indeed is the slew capability of the amp in relation to the signal to be reproduced.
Just like with the DAC, more bits will give you a more accurate wave form, so will slewrate give you a more accurate reproduction resolution. Remember the amp is always oscillating at its regulatory frequency, though with an amplitude of zero. The larger the step stimulus at the input, the more this regulatory frequency amplitude is swept up as can be demonstrated by the ringing of the output in response to a step stimulus, untill it dies out again to an amplitude of 0.
An amp with a slewrate barely sufficient to support producing a 20K sine at full power severely lacks the regulatory strength to "fix" the output at its intended value. When exagerated you can see a small oscillation imposed onto the 20K sine, it's the regulatory behaviour trying to keep up reproducing the input signal. In this case, the amplitude of the regulatory frequency will be non-zero.
The more slewing capability, the lower the regulatory frequency's amplitude will be (the less blocky the DAC output will be with increased resolution).
When you look in practice at what slewrate is sufficient, you just take a look at a FFT of music content. I think someone linked the spectrum content of many types of music content added together. One will see that the upper frequency of this spectrum (15K - 20K) has an extremely low amplitude compared to that of for instance the 1K signals.
Making an amp capable of staying within its slewing limits with a full power 20K sine generally makes sure there's enough slewing capability at the actual levels of real 15K - 20K content.
But that's sine waves. Now let's go on to real transients. Take a song, where all the instruments and voices kick in at the very same moment. These are comprized of milions of sines of varying amplitudes that are added together.
Now remember what a square wave is: It's an addition of a base sine and all its derived upper harmonics. THUS: Individual components (sines) of music transients can well result in signal edges that over the full output swing are near vertical, if only for that single moment in time of the transient. This requires much more slewing than needed for a single 20K full power sine wave. Cumulative edges could demand multiples of slewing needed to allow the amp to keep the amplitude of the regulatory frequency close to, or at zero.
That's why you want an amp that can produce a near perfect 20K full power square wave as this will be the ultimate theoretical max of musical transient sines added together.
All of this has jack to do with THD, this is all about transient response. That's why careful and proper supply decoupling is so extremely important (electrolytics/films/ceramics at each OPS device etc) so that the current demand of such transients can be met. Perhaps even more important than the actual slewing capability of the amp. What's slewrate worth if the supplies can't meet the near instant current delivery. (You'd get these 'knees' at the beginning/end of the transient response).
So, what slewrate is enough? In my book, you'll have enough if your amp can reproduce a 20K full power square without ringing and without noticable knees at the edges.
I have no formulae for you to calculate the required slewrate, but in this case, a scope can be a sufficient tool to determine whether your amp can produce the near perfect 20K square wave.
Whatever the harmonics content the result is a signal that has
at any given moment only a single instantaneous value.
That's entirely correct. The actual signal is an addition of all the elementary sines/wave forms that comprize an instrument or a voice. See the picture attached. You'll see the wave forms of individual instruments/tracks. The last window with the blue trace labeled MST is the sum of all the individual tracks: A single, instantaneous value at any given time.
Attachments
Magicbox, so how do you account for CD performance which has a sharp brick wall filter characteristic at circa 22 KHz? There are very few harmonics above that and yet a good recording on a decent CD player will sound very nice.
If you are going to produce near vertical transients, with the high order harmonics that implies, you are also going to be exciting all sorts of parasitics in your practical amplifier. Ergo layout becomes critical and so forth.
Good audio design is an exercise in compromise (engineering in general is - remember mother natures demands her laws are obeyed first!)
I think you have to decide the rise time and bandwidth taking into consideration the other constraints in your system. For my designs, they generally show overshoot and then I apply suitable input bandwidth limiting and the final response is a clean square wave.
If you are going to produce near vertical transients, with the high order harmonics that implies, you are also going to be exciting all sorts of parasitics in your practical amplifier. Ergo layout becomes critical and so forth.
Good audio design is an exercise in compromise (engineering in general is - remember mother natures demands her laws are obeyed first!)
I think you have to decide the rise time and bandwidth taking into consideration the other constraints in your system. For my designs, they generally show overshoot and then I apply suitable input bandwidth limiting and the final response is a clean square wave.
Last edited:
You're right about content above 20K. But you can "shift down" the same principle to for instance 50Hz, low bass. Now add everything between 50Hz and 20KHz using the same starting point in time and see what slope you end up with. One for the mathematicians amongst us to show
Last edited:
I doubt your amp is stable in the real world. And, we don't need huge loop gains and sub ppm distortion to get good sound.
There have been a lot of studies on slew rate.
I have seen some where they claim 20 V/us is enough, and another where they state 1V/us per volt of output voltage swing. So, if your amp can swing 75 V peak, the required slew rate is 75 V/us.
In VFA designs, it's quite easy to get double that as I demonstrated in the e-Amp.
The sx-Amp hits over 200 V/us and the small signal rise time is under 50 ns.
I have seen some where they claim 20 V/us is enough, and another where they state 1V/us per volt of output voltage swing. So, if your amp can swing 75 V peak, the required slew rate is 75 V/us.
In VFA designs, it's quite easy to get double that as I demonstrated in the e-Amp.
The sx-Amp hits over 200 V/us and the small signal rise time is under 50 ns.
Magic Box
You say :
- All of this has jack to do with THD, this is all about transient response. That's why careful and proper supply decoupling is so extremely important (electrolytics/films/ceramics at each OPS device etc) so that the current demand of such transients can be met. Perhaps even more important than the actual slewing capability of the amp. What's slewrate worth if the supplies can't meet the near instant current delivery. (You'd get these 'knees' at the beginning/end of the transient response).-
------------------------------------------------------------------------
Yes that`s simple truth !
BTW ,
good designed PSU is allways corner stone for good sounding Amp ( SS-CFA included) ! and is interesting to see that in case of ordinary linear PSU almost nobody insist on electrostatic schield between main transf. primary coil and secondary coil which Must be connected on AMP central ground star point .
Here is one except :
You say :
- All of this has jack to do with THD, this is all about transient response. That's why careful and proper supply decoupling is so extremely important (electrolytics/films/ceramics at each OPS device etc) so that the current demand of such transients can be met. Perhaps even more important than the actual slewing capability of the amp. What's slewrate worth if the supplies can't meet the near instant current delivery. (You'd get these 'knees' at the beginning/end of the transient response).-
------------------------------------------------------------------------
Yes that`s simple truth !
BTW ,
good designed PSU is allways corner stone for good sounding Amp ( SS-CFA included) ! and is interesting to see that in case of ordinary linear PSU almost nobody insist on electrostatic schield between main transf. primary coil and secondary coil which Must be connected on AMP central ground star point .
Here is one except :
Attachments
Last edited:
Well, don't forget we are driving speakers, too, that acts like microphones, and produce delayed motional signals that the feedback has bring back to the input stage to fight against.
What we don't measure, too, is how behave a very low signal, just after a large transient peak.
I used an OPA to compare the in and out of amplifiers in real situation, real music, loudspeaker as a charge. Same kind i use for protection. The difference (with fast OPA) is quite null whith sin waves. Listening to real music, it is an other story.
What we don't measure, too, is how behave a very low signal, just after a large transient peak.
I used an OPA to compare the in and out of amplifiers in real situation, real music, loudspeaker as a charge. Same kind i use for protection. The difference (with fast OPA) is quite null whith sin waves. Listening to real music, it is an other story.
Yeah exactly. Humans can run in circles easily, but try to do a 180 degrees and start running circles the other direction. All the kinetic energy stored in your body going one way you have to counter with your muscles trying to turn around and continue..
Edit: Anyways, I'll back off from the thread now.. I never meant to be an egomaniac or wanted to be on a soap box. I just liked sharing my thoughts and knowledge with you, 's all..
Last edited:
