Re: Distortion in amplifiers.



Ian Bell wrote:

Patrick Turner wrote:
I will re examine the THD of a single BJT, we all need a refresher
course sometimes,
but I will use my schematic, not yours,
because you have not presented it anywhere so one click will have it on
my screen.

OK click on this and you will see it.

http://i103.photobucket.com/albums/m124/ruffrecords/patrick.jpg

Now I challenge you to build and test it and publish the results.

Ian


I'm glad you are still talking to me. Hundreds of less courageous ppl
would have ignored me.

That schematic appeared fine with one click, and that's more like it.

Before I try it, I need to just talk about a couple of things.

The issue we were discussing was concerning BJTs and their Voltage
linearity,
which you say is good because you get low THD from this circuit.

Have I got you right? I hope so, because I can't sleep if I misrepresent
anyone.

Now in your circuit you would have what I have to assume to be
a low impedance signal source feeding a 10uF dc blocking cap followed by
a
high R bias divider circuit, and a series 33k resistor to the base.
The emitter is effectively grounded for ac, the collector load is
the 47k dc load, with 50k ac load in parallel, which makes a total ac
load = approx 24k.
The bias circuit R are effectively tight coupled to the sig gene, so we
may ignore there
presence for now.

The data on the 2n3904 is at

http://www.fairchildsemi.com/ds/2N/2N3904.pdf

The dc gain hfe could be anywhere between about 70 and 300 for when Ic =
1.5mA,
and temp is about room temp, but let us suppose hfe is 100.
( BJTs are not like tubes, their characteristics vary with batches and
temperature,
so its not much use having fixed data....)

Allow me to guess at what we might measure in your circuit if we had
the meters able to do it without the sometimes low voltages being
obscured by
hum and noise. Noise drew me to many wrong conclusions, until i learnt
to measure things without
adding noise, shunting the signal i wanted to measure, or causing
oscillations, or all 3.

This means that if you have say 2.4Vrms at the collector output, load
current Ic = 0.1mA.

So base current will be 1/100 of this, = 0.001mA.

To get this much current change into the base via the 33k, you need a
voltage across the 33k =
33k x 0.001mA = 0.033Vrms.

Let us assume the base input resistance was say 2k.
I have no idea exactly what it would measure, just like I have no idea
what the collector resistance is either.
These parameters are not just nicely listed in the data, of if they are
it will be obtained by reading the many
graphs for the device, but to really know, you have to measure it.

So if you have 0.033Vrms input across the 33k, and you have 2k from base
to 0V,
then there must be a voltage change at the base of 2,000 x 0.001mA, so
the base ac voltage = 0.002Vrms, not much, and the VOLTAGE gain between
base and
collector = 2.4V / 0.002V = 1,200.

The voltage at the source will be 0.033Vrms + 0.002vrms = 0.035Vrms, and
so voltage gain between source and collector output = 2.4 / 0.035 =
68.6, a dramatically lower
amount of gain than actually exists between the base and the collector.
The relationship between the collector current and base current is
essentially linear, in fact as
linear as the voltage amplification in a triode.
But lets not confuse voltage linearity with current linearity.

Suppose a -ve distortion voltage appears at the collector.

Call it -Dn. Then you will have an IDn, and at the base, IDn / 100 will
flow because
all currents that flow at the collector are all divided down by the hfe
figure.
A -Dn V causes MORE Ic, so you will get MORE Ib, so Idn at the base will
cause the base voltage to
tend to go negative, tending to turn off the bjt, and to cause the
collector voltage to rise,
as Ic and IRL becomes LESS. In other words, the tiny fraction of the
output distortion current
that flows in RL is applied to the base automatically to oppose its own
production,
and this is shunt NFB in action, and quite a bit too there is.

Now, suppose you place a link across the 33k, and increase the input cap
from 10uF
to 1,000uF, and make sure the signal generator output resistance is 10
ohms, then
you will prevent the negative current FB from acting, since the input R
has been reduced to
negligible levels. What you will see now is a heck of a lot more gain,
probably about 1,200 like I said you would, but a lot more distortion in
the collector signal and as you crank up
the input voltage so output = 15Vrms, possible with a +60V dc supply,
the distortion should be quite bad.

Looking at the situation more simply, 33k only has 0.033Vrms across it
to give 2.4vrms output.

The amount of base current change is tiny!, so we say that the input is
being fed by a
virtual 'current' source.
And it because the 33k is so much larger than the resistance load it
powers, ie, the base input resistance.

In a tube, Rin to the grid is many megohms, and not easily measured, so
a 33k series grid R can still be called
low impedance signal source, or a 'voltage' source.
In a tube circuit, a 33k series grid R has no NFB function, since no
current flows in it.

In another recent post by John Byrns, the same proposition has been made
about
building a BJT amp with no NFB, but his intitial idea had a 10k series
base input R, which is a FB element.
If you want to improve on the outcome of your circuit and want a gain of
only 10, for a line stage,
then you could connect another resistance from the base to collector as
an extra FB element in addition
to the internal resistance between collector and base that exists.

The first time I connected up a BJT without any series R from signal
gene to base the
THD was appalling, and very soon when I inserted a series R to measure
the base input resistance
the distortion went a lot lower, along with the gain. Hmm, I thought,
NFB.

To raise base input resistance to a maximum, because it varies with
collector load, one can use an emitter follower buffer
from the collector, and have a CCS dc supplyto the collector, but then
the dc stability will be lousy,
so one answer is to have a -60V supply, and extend the emitter
resistance of say 100k,
while leaving the 1,000uF emitter bypass cap where it is.
Still you will find Rb in to be too low, so a darlington pair should be
used, and for really nice work,
use only darlington pairs for gain devices, with a dp for emitter
follower, and finally you will begin to see the
silicon shine, and the base to collector gain will maybe go up to 5,00,
and far more than you would ever need,
and so will the THD.
But simple shunt resistance from the dp follower emitter to input base
of the gain dp
will control runaway gain and you will have a stage with extremely low
THD.
But its all due to NFB, not the inherent voltage linearity of the BJT,
which does not exist;
such linearity is a triode owned thing, because of the triode internal
NFB which you cannot
include in any external loop like you can with currents around a BJT.

I could tell about doing a µ-follower circuit with bjts instead of
triodes,
and such a stage with shunt NFB will work just great, and local current
NFB can be
used in the emitter circuit and THD can be somewhat tailored by choosing
the collector load to the bottom gain pair.

J-fets have high Rin like tubes, and do not have such a thing as a hfe,
or base resistance and so can be treated exactly like
small pentode tubes, with a gm, Rd, and µ.

The 2sk369 at Id = 5mA has gm = 40mA/V, Rd = 80k, and µ = 3,200. However
with mant high gm fets and mosfets
with high gain, the miller C need only be 5pF drain to gate, abd if gain
= 1,000,
then C Miller = 0.005uFm way to high for most audio apps.

BJTs also have Miller C and other bandwidth limiting issues, and input
stage gain should be tailored to maintain
high Rin, and low Cin.

Patrick Turner.
.

Relevant Pages

  • Re: Distortion in amplifiers.
    ... The issue we were discussing was concerning BJTs and their Voltage ... The dc gain hfe could be anywhere between about 70 and 300 for when Ic = ... This means that if you have say 2.4Vrms at the collector output, ... Let us assume the base input resistance was say 2k. ...
    (rec.audio.tubes)
  • Re: How to pick the right transistor?
    ... I would be looking for a transistor with an advertised maximum current 3 to 10 times higher than such a requirement, since most transistors hit their peak current gain at 1/3 to 1/10th of maximum rated Ic (collector current). ... You don't mention whether you need PNP or NPN, the highest collector voltage the device must withstand, whether you are switching 220 mA or linearly amplifying at that current, the signal frequency, how important low on state voltage drop is. ...
    (sci.electronics.basics)
  • Re: Speaker Mic - Common Base Config
    ... went in search of a speaker microphone circuit. ... away from the CB gain stage. ... will bob up and down with the wiggles in voltage across it. ... The gain of a common base stage is approximately given by /, where Rc is the collector impedance, Re is the dynamic emitter resistance, and Rs is the source impedance looking back out of the emitter. ...
    (sci.electronics.design)
  • Re: BJT not gate (was) Re: a question (extreme newbie alert)
    ... resistor is at voltage level V ... If the collector to emitter voltage approaches zero, the gain heads toward zero, so if this Icalculation shows a drop across Rgreater than 5 volts, you just assume the collector to emitter voltage is less than .6 volts (the transistor is saturated on). ...
    (sci.electronics.basics)
  • Re: JJ ECC99 for CCDA
    ... collector current is not, because beta is changing. ... and a high enough resistance to be labelled a constant current ... but in the first place because beta is changing. ... Beta is not constant if collector voltage is swinging. ...
    (rec.audio.tubes)