NPN Common Emitter Bias

[[email protected]]

[email protected] wrote...
C2 = 1nF ??

Sure. Xc(1nF) < 0.2 ohms at 920 MHz. Actually, 100pF might be more
apropos to minimize stray inductance in the larger cap.

Yes, I know the OP didn't specify frequency and probably means/needs
audio-band operation or some such ... I was responding to Andrew's
post. When the OPs omit details, we get to make them up, don't we ?
;-)

Grins,
James Arthur

 

[Tim Wescott]

Tim Wescott wrote...



I don't have that book, but many wideband linear RF amplifiers I've
worked on used a current-sensing resistor in the collector RF path,
with a little servo circuit to establish the average value of the
base current. This is easy to do with just a few transistors. The
sense resistor need not drop more than 200 to 400mV.

If you're building a class AB linear the average collector current can
(and should!) vary quite a bit -- for these you use a diode that's in
close thermal contact with the transistors to establish the bias voltage
to the base. The circuit that I have seen came out of an old Motorola
app note (can't remember which one) and uses a 723 for the actual
regulation chore.

I _have not_ built one of these for a production system, so I can't
comment on how much care and feeding it would demand.

 

[Winfield Hill]

James Arthur wrote...

Winfield Hill wrote:

The circuit has three problems, all easily fixed. [...]

Here's a version that makes sure the inductor is bypassed at
RF frequencies, yet, per Win, avoids AM modulation of the bias
from power supply noise:

. .---------------------+----------+------- +Vcc
. | C1 | |
. R1 100n _|_ Rs
. 300mV --- 200mV
. | R3 | R4 |
. v\| .--/\/\--+--/\/\----+--------.
. |---, | 50mV 50mV | |
. /| | |/v | --- C2
. +-----+----| C| --- 1n
. | Q2 |\ C| L1 |
. R2. | C| ===
. 5mA .-. | GND
. etc | | R5 |
. | | | 1k +------------ out
. GND '-' |
. | Q1 |/
. RF IN------+-----------------|
. |\v
. |
. ===
. GND

This circuit has two problems, easily fixed.

First, the 100mV drop across R3 and R4 is determined by the base
current of Q1, the exact value of which is unknown, and which can
change by a factor of say 3 over temperature, etc. So if the drop
varies from say 50 to 150mV, then the current-sense voltage will
vary from 250 to 150mV, which means we haven't done a very good
job of setting Q1's collector current. To solve this we need to
stabilize Q2's current. We can do this with a Q1 base-to-ground
resistor, R6 = Vbe/I, sized to draw say 3x more current than Q1's
base. We can hope this new resistor, R6, will have a higher value
than Q1's RF input impedance.

The second issue is loop stability. The loop gain is roughly
gm1*Rs, which is 40*0.2=8, times R6/(R3+R4) = 750mV/100mV=7.5,
for a DC gain of about 60. Capacitor C1 must provide a dominant
pole, reducing the loop gain to below unity before the occurrence
of a second pole. The second pole could be due to Rs C2, or due
to the input coupling capacitor, C3, with R6. Either way, C1
will probably need to be larger, likely an electrolytic.

Here's the new circuit:

.. .---------------------+----------+------- +Vcc
.. | C1 | |
.. R1 elec _|_+ Rs
.. 300mV --- 200mV
.. | R3 | R4 |
.. v\| .--/\/\--+--/\/\----+--------.
.. |---, | 50mV 50mV | |
.. /| | |/v | --- C2
.. +-----+----| C| --- 1n
.. | Q2 |\ C| L1 |
.. R2. | C| ===
.. 5mA .-. | GND
.. etc | | R5 |
.. | | | not too +------------ out
.. GND '-' big |
.. | Q1 |/
.. RF IN ---||---+----+------------|
.. C3 | |\v
.. R6 |
.. | ===
.. GND GND

I question the need for R5, which at any rate must not be too
big, or drop more than a few volts. I question it because Q2's
collector is a high-Z current-source output, most likely with a
low capacitance, much smaller than Q1's base. But, if it's not
too large, it won't hurt anything. It could be an RFC.

 

[[email protected]]

James Arthur wrote...

Winfield Hill wrote:

The circuit has three problems, all easily fixed. [...]

Here's a version that makes sure the inductor is bypassed at
RF frequencies, yet, per Win, avoids AM modulation of the bias
from power supply noise:

. .---------------------+----------+------- +Vcc
. | C1 | |
. R1 100n _|_ Rs
. 300mV --- 200mV
. | R3 | R4 |
. v\| .--/\/\--+--/\/\----+--------.
. |---, | 50mV 50mV | |
. /| | |/v | --- C2
. +-----+----| C| --- 1n
. | Q2 |\ C| L1 |
. R2. | C| ===
. 5mA .-. | GND
. etc | | R5 |
. | | | 1k +------------ out
. GND '-' |
. | Q1 |/
. RF IN------+-----------------|
. |\v
. |
. ===
. GND

This circuit has two problems, easily fixed.

First, the 100mV drop across R3 and R4 is determined by the base
current of Q1, the exact value of which is unknown, and which can
change by a factor of say 3 over temperature, etc. So if the drop
varies from say 50 to 150mV, then the current-sense voltage will
vary from 250 to 150mV, which means we haven't done a very good
job of setting Q1's collector current. To solve this we need to
stabilize Q2's current. We can do this with a Q1 base-to-ground
resistor, R6 = Vbe/I, sized to draw say 3x more current than Q1's
base. We can hope this new resistor, R6, will have a higher value
than Q1's RF input impedance.

The second issue is loop stability. The loop gain is roughly
gm1*Rs, which is 40*0.2=8, times R6/(R3+R4) = 750mV/100mV=7.5,
for a DC gain of about 60. Capacitor C1 must provide a dominant
pole, reducing the loop gain to below unity before the occurrence
of a second pole. The second pole could be due to Rs C2, or due
to the input coupling capacitor, C3, with R6. Either way, C1
will probably need to be larger, likely an electrolytic.

Here's the new circuit:

. .---------------------+----------+------- +Vcc
. | C1 | |
. R1 elec _|_+ Rs
. 300mV --- 200mV
. | R3 | R4 |
. v\| .--/\/\--+--/\/\----+--------.
. |---, | 50mV 50mV | |
. /| | |/v | --- C2
. +-----+----| C| --- 1n
. | Q2 |\ C| L1 |
. R2. | C| ===
. 5mA .-. | GND
. etc | | R5 |
. | | | not too +------------ out
. GND '-' big |
. | Q1 |/
. RF IN ---||---+----+------------|
. C3 | |\v
. R6 |
. | ===
. GND GND

I question the need for R5, which at any rate must not be too
big, or drop more than a few volts. I question it because Q2's
collector is a high-Z current-source output, most likely with a
low capacitance, much smaller than Q1's base. But, if it's not
too large, it won't hurt anything. It could be an RFC.

Nice tweaks Win.

My actual circuit, as I recall now some 16 years later (there were a
few versions), was as Andrew's, with a 1-2V drop across the sense
resistor, minimizing the temperature drift problem. A Miller
integrating capacitor across the current-sense transistor ensured
stability.

The goal was to overcome the output transistor's native 6:1 beta
range (at 25C); confining Q1's idle current to a 2:1 range was
considered excellent control...the customary, competing circuit
was--get this--a fixed resistor to Vcc! Even more surprising to us
non-rf-types, was that 6:1 idle current variation wasn't really all
that bad--the actual variation was somewhat less in practice, it worked
fine, and only changed the r.f. output by say +/- 30%, i.e. just a dB
or so. I wanted to save battery power, which was at a premium, and
avoid modulating the transistor reactances and thus interfering with my
i/o network optimizations.

Of course the details of desired power output and supply voltage
(dictating choice of Q1), input and output loading, and so forth are
vital to any particular implementation.

Here's the rationale for R5:
In my case Q2 was a 2n3906, which shows about 2pF Ccb _typical_ with
a few volts Vce in the Motorola Small Signal Transistor book. That 2pF
was about double the _max_ input capacitance of Q1 alone, and added
about 90 ohms' reactance in parallel with Q1's base--enough to detune
its carefully tuned input.

Isolating Q1(b) from Q2's collector with an inductor was tempting,
except that inductors of the d.c.-choke scale were really just
capacitors in sheep's clothing at 900MHz, having unspeakable parasitics
which loaded the junction and added both resonances, and
magnetic-coupling from L1 (bad!!) to the mix.

Although simply strapping Q2(c) to Q1(b) and redesigning the input
network might've been okay, it was easier, cheaper (several million of
these devices were ultimately built) and sleep-at-night-happier to
insert an R5 dropping a volt or two--no more--and thus be sure of being
isolated, yet reactance and coupling-free. SMD resistors are
surprisingly good at UHF.

Cheers,
James Arthur

 

[Winfield Hill]

[email protected] wrote...

Here's the rationale for R5:
In my case Q2 was a 2n3906, which shows about 2pF Ccb _typical_
with a few volts Vce in the Motorola Small Signal Transistor book.
That 2pF was about double the _max_ input capacitance of Q1 alone...

Whoa, interesting, what part was Q1?

 

[[email protected]]

[email protected] wrote...

Whoa, interesting, what part was Q1?

MRF571. Sweet transistor. You're right though, I mis-read the data
sheet. Actual Cin is spec'd at 1.4pF _typical_, which would seem to
make the effect of paralleling Q2(c) slightly smaller than I indicated.

That 1.4pF figure, however, was reported at 5mA and f=1MHz. Under
bias and near 1GHz the 1.4pF figure no longer holds: the input shifts
to being just a tad inductive, best as my foggy recollection interprets
these S-parameters.

As I recall it, the non-adjustable input network used caps on the
order of 2.2pF, 8.2pF, 6.8pF, etc., so a possible 2pF extra was
significant, unwelcome, and adding R5 seemed a cheap way to ignore it
and avoid manufacturing variations.

Cheers,
James Arthur

 

[dgc]

I suspect if the OP had a negative voltage available he would want the
output to swing all the way down to it ;-)

Ian

Ian I am the OP of this post and, after reading all the replies, am certain
I am corresponding with people whose knowledge beta is well beyond mine.
Appoligies for leaving out critical data in my post. The circuit is indeed
RF (7 MHz). I am trying to squeeze 2 watts rms out of a two stage
arrangement the first of which is a crystall oscillator and the final an
NTE235 NPN in common emitter arrangement. I am trying to maintain linearity
in the output waveform and my old BKPrecision oscilloscope indicates I am
failing miserably at this. The drive from the oscillator is running about 1
volt peak (i know this is high) which may be part of the problem. The drive
is brought in from a 1 turn winding off the oscillator tuned circuit torroid
inductor. I had 500 mA of standing DC on the 235 at one point and did indeed
achieve the best output waveform at this level of bias. The heat sinked 235
was still running pretty hot at this level so I backed off. I now have a
larger heat sink (3 X 6) 1/32 aluminum which will be better I'm sure.
Frankly I may have harmonics in the output waveform as well. I'm no expert
in deciphering oscope waves unless they are pretty clean. Output load is 50
ohms, which if the equation V0^2 / 2Po is correct would indicate 50 ohms is
too high for 2 watts (another problem). Would this be compounding the
linearity issue? I got a kick out of your post. You bet, I'm the kinda guy
that wants all the peak to peak voltage available from the supply.

 

[[email protected]]

[email protected] wrote...

Whoa, interesting, what part was Q1?

Q1 was an MRF571. Sweet part. You're right though, I mis-read the
data sheet. C.in is actually about 1.4pF _typical_, which would
initially seem to reduce the effect of adding 2pF by connecting Q2(c).
That 1.4pF C.in, however, is reported at 1MHz and 5mA, and is not that
experienced in actual operation. Under bias and 920MHz it looks to my
addled rf-pate like the MRF571's base reactance is less capacitive,
shifting to just about neutral or a little inductive, depending.

Without recalling the exact particulars, the input matching network
was a fixed network with values on the order of 2.2pF, 8.2pF, 5.6pF,
etc., and had to be on-tune, so an extra 2pF was in any case unwelcome.

And that, my friends, is the tale of the resistor R5, and how she
came to be.

Cheers,
James Arthur (posted this earlier in the day, but it failed to show)

 

[Andrew Holme]

Ian I am the OP of this post and, after reading all the replies, am
certain I am corresponding with people whose knowledge beta is well
beyond mine. Appoligies for leaving out critical data in my post. The
circuit is indeed RF (7 MHz). I am trying to squeeze 2 watts rms out
of a two stage arrangement the first of which is a crystall
oscillator and the final an NTE235 NPN in common emitter arrangement.
I am trying to maintain linearity in the output waveform and my old
BKPrecision oscilloscope indicates I am failing miserably at this.
The drive from the oscillator is running about 1 volt peak (i know
this is high) which may be part of the problem. The drive is brought
in from a 1 turn winding off the oscillator tuned circuit torroid
inductor. I had 500 mA of standing DC on the 235 at one point and did
indeed achieve the best output waveform at this level of bias. The
heat sinked 235 was still running pretty hot at this level so I
backed off. I now have a larger heat sink (3 X 6) 1/32 aluminum which
will be better I'm sure. Frankly I may have harmonics in the output
waveform as well. I'm no expert in deciphering oscope waves unless
they are pretty clean. Output load is 50 ohms, which if the equation
V0^2 / 2Po is correct would indicate 50 ohms is too high for 2 watts
(another problem). Would this be compounding the linearity issue? I
got a kick out of your post. You bet, I'm the kinda guy that wants
all the peak to peak voltage available from the supply.

It is normal to use a low-pass filter network at the output of a transmitter
to remove the inevitable harmonics; although you can certainly minimise
harmonics by making the circuit as linear as possible; however, simple
2-stage oscillator-PA transmitters of the type you describe are typically
operated with the PA stage in class C i.e. no standing bias. You don't need
a linear PA unless you're amplitude modulating the drive.

As for the RL = V^2/Po issue, where RL is the load resistance "seen" by the
collector, you need an impedance transforming (matching) network at the
output which makes the 50 load "look" like the desired RL value to suck-out
the required amount of power. The transistor then needs sufficient drive to
make the requisite amount of AC collector current available.

It is normal to make the peak voltage swing at the collector close to the
power supply Vcc to maximise efficiency, but it sounds like you might be
willing to live with less than ideal efficiency. You could get 2W out with
less than a 12V swing.

One final point: it helps if you have a way to smoothly adjust the drive
level from the oscillator - perhaps by running it off a seperate variable
power supply. If the output level changes by sudden jumps, and does not
vary smoothly as you adjust the drive level, you have spurious oscillation
problems.

 

[Ian Bell]

Ian I am the OP of this post and, after reading all the replies, am
certain I am corresponding with people whose knowledge beta is well beyond
mine. Appoligies for leaving out critical data in my post. The circuit is
indeed RF (7 MHz). I am trying to squeeze 2 watts rms out of a two stage
arrangement the first of which is a crystall oscillator and the final an
NTE235 NPN in common emitter arrangement. I am trying to maintain
linearity in the output waveform and my old BKPrecision oscilloscope
indicates I am failing miserably at this. The drive from the oscillator is
running about 1 volt peak (i know this is high) which may be part of the
problem. The drive is brought in from a 1 turn winding off the oscillator
tuned circuit torroid inductor. I had 500 mA of standing DC on the 235 at
one point and did indeed achieve the best output waveform at this level of
bias. The heat sinked 235 was still running pretty hot at this level so I
backed off. I now have a larger heat sink (3 X 6) 1/32 aluminum which will
be better I'm sure. Frankly I may have harmonics in the output waveform as
well. I'm no expert in deciphering oscope waves unless they are pretty
clean. Output load is 50 ohms, which if the equation V0^2 / 2Po is correct
would indicate 50 ohms is too high for 2 watts (another problem). Would
this be compounding the
linearity issue? I got a kick out of your post. You bet, I'm the kinda
guy that wants all the peak to peak voltage available from the supply.

I am not really an RF chap, although I did do some in my youth with tubes,
so don't take what comes next as coming from an RF expert. If I wanted 2
watts out of the final stage then I would certainly be considering
operating it in class C, in which case DC biassing does not come into it.
The classic class C circuit has an resonant circuit in the collector to
remove a lot of the harmonics although some will still remain. These are
usually removed by a subsequent LC filter circuit. As for 2 watts into 50
ohms, this implies an RMS output voltage of 10V so you would need a 30V
supply - do you have this? If not the classic way to overcome this is with
an RF transformer arrangement.

HTH

Ian

 

[Andrew Holme]

Andrew Holme wrote:
[snip]

As for the RL = V^2/Po issue, where RL is the load resistance "seen"
by the collector, you need an impedance transforming (matching)
network at the output which makes the 50 load "look" like the desired
RL value to suck-out the required amount of power. The transistor
then needs sufficient drive to make the requisite amount of AC
collector current available.

I missed a factor of 2 in the above equation. It should be: RL = Vcc^2 /
(2*Po)

Although Po = Vcc^2 / (2*RL) is the maximum power available for a given
supply voltage and collector load, I also find it helpful to think about Po
= 1/2 * Icpk^2 * RL since the transistor is a current-output device.