rshayes wrote:The output impedance will be quite low. The dynamic resistance of the transistor output in the emiter follower connection is about 26 ohms divided by the emittter current in milliamps. Since the emitter current is about 4 milliamps, the transistor output impedance will ideally be about 6.5 ohms. There will be some bulk resistance (prbably a few ohms) inside the transistor in series with this impedance, probably raising it to the 10 ohm level. There is another series term due to the source impedance feeding the emitter follower divided by the beta of the transistor. Due to the feedback in the previous stage, this impedance will be low, and the division by beta will make it even lower. This might raise the output impedance by an ohm or two. This will be in parallel with R10, which is 1000 ohms. This will lower the output impedance by about 1 percent. As a rough estimate, the output impedance will be in the 10 to 15 ohm range.
A resistor in the 100 ohm range is series with the output might be a good idea. This will raise the output impedance, but it will also make the emitter follower more stable if a capacitive load is applied (3 or 4 feet of audio cable with a capacitance of about 30 picofarad per foot).
Q1 and Q2 form a feedback amplifier, and may need some additional parts to allow operation without oscillation. One possibility is a fes picofarads between the collector and base of Q2.
The most effective bias filtering would be obtained by making R2 and R4 equal (possibly 150K).
The input impedance can be raised and some temperature compensation obtained by adding a PNP emitter follower in front of Q1. This will cancel the effect of temperature on the base-emitter voltage of Q1.
Hi again rshayes,
Im not sure if you read mine and Roberts posts, but the output impedance
cant be that low because we usually dont want to bias too close to the
Vcc supply line.
With the output bias point set at 4.5v (one half Vcc) the transistor
impedance has to be 1k or else we couldnt get that dc output voltage,
and since the Vcc line is reqarded as 0 ohms that puts the transistor
CE in parallel with that 1k resistor, which works out simply to 500 ohms.
Of course that's just an estimate, and as the transistor pulls the output
up by 1v (2v peak to peak output) that puts the new output voltage
at 5.5v and the transistor dyno R at 636 ohms, which in parallel with
1k comes out to 389 ohms. As the transistor allows the output to
dip down to -1v from the bias point the new output is 3.5v, which
makes the transistor dyno R equal 1571 ohms, and in parallel with
1k that comes out to 611 ohms. Thus, the average output impedance
is (611+389)/2, which is 500 ohms, but since we care more about
the lowest output impedance we should go with 389 ohms, which means
that anything too much lower than that will cause bad distortion.
We also know that when we have a voltage source that has an impedance
of Rx ohms, which loaded by an equal resistor of Rx ohms, gives us
an output voltage of one half the open circuit source voltage, or V/2.
Thus, if we load a source with a resistor equal to its source impedance
we will measure one half the open circuit voltage.
If we were to load this circuit with 36 ohms, we would cause bad
distortion and although the output voltage would drop to possibly 1/2 on
one peak, it would be so badly distorted on the other peak we couldnt use
the amplifer very well.
Why does this happen?
It happens because although the output impedance of the transistor
itself can go down to 10 ohms (perhaps) that doesnt mean that is
a symmetrical impedance. Sure, it can make the output appear
to 'source' an output voltage with very low impedance, but it can
not help to 'sink' the output current very well on the other half of
the cycle. What makes up the other half of the
impedance is the lower resistor (1k in this circuit), and that is partly
responsible for the entire makeup of the output waveshape.
To approximate this, i take the total output resistance and divide by
two, because if we approximate the transistor Z by itself as very
low like 10 ohms and add this to the output resistor which is 1k
we get 1010 ohms, and the average of that is close to 1/2 of the output
resistor value, or 500 ohms. It is an approximation of course, and
we can get a little lower than this in real life.
If we were to bias closer to the Vcc supply rail we could also lower
the output impedance, but at the cost of an increase in quiescent
power dissipation.
Another way of looking at this is that although the transistor impedance
is very low when it is turned 'on', it is also very very high when it is
turned 'off', and the linear operation is somewhere in between, which
makes its overall impedance work more in accordance with the output
resistor than according to its own impedance which is usually much
lower.
Still yet another way to think about it is as in a switching circuit, where
the transistor is switched from 'on' to 'off', with no linear mode.
In such a circuit, the impedance is close to 1/2 of the resistance of
the resistor in series with the transistor (C or E) for a 50 percent
duty cycle switch cycle.
I think if you take another look you will agree.
LEDs vs Bulbs, LEDs are winning.