Based on the recent discussion, I polished up my simulator and wrote an article summarizing the findings. I think that you will find this to be an easy read and not surprising in outcome.
Class AB Biasing
http://www.edgrochowski.com/articles4/classab.html
Comments are welcome!
Ed
Class AB Biasing
http://www.edgrochowski.com/articles4/classab.html
Comments are welcome!
Ed
Ed, nice writeup, thanks.
Seeing that the optimal point does depend on the signal level, maybe we should think about a kind of sliding bias, that depends on the signal?
Not sure what can of worms that would open though.
Jan
Seeing that the optimal point does depend on the signal level, maybe we should think about a kind of sliding bias, that depends on the signal?
Not sure what can of worms that would open though.
Jan
Thanks!
jacques antoine - Class A does not have an optimal bias for low distortion. It is a case of "more is better". 😉
jan.didden - The optimum bias does depend on signal level. I blasted auto-bias in another thread because I prefer inherently linear designs.
The biggest variable is temperature. My main takeaway from the analysis is that I may have picked an optimal bias for normal listening levels, but at high volumes, the bias will drift into the overbias region.
Ed
jacques antoine - Class A does not have an optimal bias for low distortion. It is a case of "more is better". 😉
jan.didden - The optimum bias does depend on signal level. I blasted auto-bias in another thread because I prefer inherently linear designs.
The biggest variable is temperature. My main takeaway from the analysis is that I may have picked an optimal bias for normal listening levels, but at high volumes, the bias will drift into the overbias region.
Ed
Last edited:
Aside:
Same parts sound different, which is perceived by the ear.
Different parts, such as complementary transistors, sound VERY different, which is perceived by the ear.
Same parts sound different, which is perceived by the ear.
Different parts, such as complementary transistors, sound VERY different, which is perceived by the ear.
cumbb - I think that is true only of poor designs. Good designs have no sound of their own.
Ed
Ed
First of all I would distinguish parts and design;-)
And then develop a method to make the sound of components comparable;-)
Or vice versa;-?
And then develop a method to make the sound of components comparable;-)
Or vice versa;-?
If same parts sound different to you then you should try different parts not complementary, maybe they would then sound the same. 😀
I have designed my own amps for a number of years.
Setting the bias was an unknown.
So bias to zero.
Input sine wave and monitor output on a scope.
Turn bias up until cross over distortion goes.
Sounds good so job done.
Min bias per pair seems to be about 10mA.
Peavey mode with min bias.
Setting the bias was an unknown.
So bias to zero.
Input sine wave and monitor output on a scope.
Turn bias up until cross over distortion goes.
Sounds good so job done.
Min bias per pair seems to be about 10mA.
Peavey mode with min bias.
cumbb - Start your own thread. Let's get back on topic.
nigelwright7557 - The minimum distortion is well below what can be seen on an oscilloscope. You need a spectrum analyzer, or if you are like me, just throw the problem at the computer. 😉
Ed
nigelwright7557 - The minimum distortion is well below what can be seen on an oscilloscope. You need a spectrum analyzer, or if you are like me, just throw the problem at the computer. 😉
Ed
At high volumes most drivers will be complaining well beyond the delta between default and optimal bias. Looking at the curve for 0.22 ohm resistors hints at 0.087 A as a happy medium for levels between 1 V pk and 8 V pk, with distortion no greater than 0.025%. The distortion for 16 V pk and 32 V pk is about 0.05%, but I'd sure like to see the speaker showing less than 0.025% added distortion at those drive levels.
Lower distortion is good, of course, however beyond a certain point it's more an indication the amplifier is doing the Right Thing rather than an audible benefit.
Nonetheless the data is incredibly useful. I used to think more bias up to a quarter or half an amp on class-AB was better; apparently not.
The first takeaway is the dependence on the degeneration. I would be interested in the results with 0 ohms Re.
Nelson Pass - With Re=0, the distortion decreases monotonically with increasing bias current. The local minimum and maximum disappear. This happens at all amplitudes.
Ed
Ed
One I show you needs no degeneration, if only you bother to play a bit with the posted sim.
indra1 - Your circuit is a form of Vbe multiplier. I expect the tracking will not work well enough in practice to eliminate emitter resistors.
Ed
Ed
Works well enough in sim but I need to finish a few projects yet before I have time to bench test that. I'll let you know then. Main concern shown by the sim is bias stability vs ambient temp.
Last edited:
Aside:
Since a rough and fast tonal tuning is feasible by means of Re, all tonal circumstances included, I would determine Re tonally.
The bias can be adjusted very precisely following.
The hearing is roughly distinguished: low bias: tonally brighter, more "resolving", with "markers" that let you locate sonic events better - high bias: more corporeal, less locatability of sonic events, more "saturated".
Since a rough and fast tonal tuning is feasible by means of Re, all tonal circumstances included, I would determine Re tonally.
The bias can be adjusted very precisely following.
The hearing is roughly distinguished: low bias: tonally brighter, more "resolving", with "markers" that let you locate sonic events better - high bias: more corporeal, less locatability of sonic events, more "saturated".