[Mark Aitchison]

Patrick Turner wrote:

If we had a pure CCS load to the 2N3904 collector, its open loop gain
would be extremely high, perhaps 2,500... And with such enormous gain
comes lots of distortions


Hmm, thinking about it, if you cut down the input signal so the output
voltage is reasonable and the load impedance is highish (say 680k)
compared with the 6k8 you had before, then I think you would get low
distortion (with or without an unbypassed emitter resistor, although the
latter would be better). The high impedance (a CCS in parallel with
something like 500k-1Mohm plus test equipment, or into a good
transformer, means the percentage change in collector current would be
very low, and this is what is needed to reduce distortion. Signal
across the base-emitter junction varies with the log of the percentage
change in collector current (as a first approximation, assuming we are
biasing the transistor sensibly), and the distortion component of that
signal drops very quickly as the signal decreases.

So, if you still have the circuit laying around to test a modification,
I think it is worth giving an active load for the 2N3904 a try - I think
you could be pleasantly surprised with the results.

Mark.

[Eeyore]

Mark Aitchison wrote:

Patrick Turner wrote:

If we had a pure CCS load to the 2N3904 collector, its open loop gain
would be extremely high, perhaps 2,500... And with such enormous gain
comes lots of distortions

Hmm, thinking about it, if you cut down the input signal so the output
voltage is reasonable and the load impedance is highish (say 680k)
compared with the 6k8 you had before, then I think

Apparently not !


you would get low
distortion (with or without an unbypassed emitter resistor, although the
latter would be better). The high impedance (a CCS in parallel with
something like 500k-1Mohm plus test equipment, or into a good
transformer, means the percentage change in collector current would be
very low, and this is what is needed to reduce distortion. Signal
across the base-emitter junction varies with the log of the percentage
change in collector current (as a first approximation, assuming we are
biasing the transistor sensibly), and the distortion component of that
signal drops very quickly as the signal decreases.

So, if you still have the circuit laying around to test a modification,
I think it is worth giving an active load for the 2N3904 a try - I think
you could be pleasantly surprised with the results.

Did you study at the University of Toytown ?

Graham

[Mark Aitchison]

I wrote:

very low, and this is what is needed to reduce distortion. Signal
across the base-emitter junction varies with the log of the percentage
change in collector current (as a first approximation,...

Oops! Not quite right, the way I wrote it. "Signal" is not the best word
to use there. I mean: the base-emitter signal reduces with *percentage
change* in collector current (which can be small for voltage outputs of
decent size) as a first approximation, and the distortion in that small
signal is very small. Even for larger signals, where the change in base
voltage is not well enough approximated on a straight-line graph but
from deltas on a log curve between input voltage and output current,
arranging for the fractional change in output current to be low for a
good output voltage is still good news because the base voltage is still
shrinking as the load impedance increases, just not linearly, and the
fraction of the small signal that is distorted shrinks quickly.

And where I talk about a distortion component in the base voltage this
only makes sense in the case of emitter degeneration (or other nfb); I
often think of the input voltage being distorted with respect to the
output as being another equally valid way of thinking about the output
being distorted with respect to the input - maybe my brain thinks this
way due to a misspent youth listening to Spike Milligan; I probably
shouldn't write in these terms, as it probably confuses still-sane
readers. But thinking in terms of distortion in the base signal is
valuable because it relates nicely to the base junction I-V equation
(after differentiating and using a bit of mathematics).

Mark.

[Patrick Turner]

Mark Aitchison wrote:

Patrick Turner wrote:

If we had a pure CCS load to the 2N3904 collector, its open loop gain
would be extremely high, perhaps 2,500... And with such enormous gain
comes lots of distortions

Hmm, thinking about it, if you cut down the input signal so the output
voltage is reasonable and the load impedance is highish (say 680k)
compared with the 6k8 you had before, then I think you would get low
distortion (with or without an unbypassed emitter resistor, although the
latter would be better). The high impedance (a CCS in parallel with
something like 500k-1Mohm plus test equipment, or into a good
transformer, means the percentage change in collector current would be
very low, and this is what is needed to reduce distortion. Signal
across the base-emitter junction varies with the log of the percentage
change in collector current (as a first approximation, assuming we are
biasing the transistor sensibly), and the distortion component of that
signal drops very quickly as the signal decreases.

So, if you still have the circuit laying around to test a modification,
I think it is worth giving an active load for the 2N3904 a try - I think
you could be pleasantly surprised with the results.

Mark.

Assemble a circuit and test the measured results.

That's the only way to really find out what is fact about such things.

Bu my expereince is that were anyone to have a gain of say 2,000
from a single bjt, its utterly useless because of
the noise, distortion, poor BW and high Rout, so a shirtload of
shunt voltage NFB needs to be used.
Where the collector load is a high ohm value, ie, CCS,
massive dc displacement oocurs as well, and current FB doesn't work
because
there is such low current change.
So voltage NFB has to be used.

Setting up a darlington pair gain common emitter stage
and with darlington pair in a µ-follower topology like a tube circuit
will give you easily gain = 1,000, and a shunt FB network from the
top device emitters to base input of the bottom gain devices
to reduce gain to say 20 will result in again reduction of 50 times, and
if the
open loop THD was 5% at 10Vrms without NFB then expect 0.1% with NFB.

Do the work to find out!!!!

Analyse it all!!

Understand all of what you are doing!!

Forget trying to use a transformer in the collector circuit of a bjt.

Now why did I say that?

Patrick Turner.