Chapter 6: Broadband amplifiers
Broadband is more than one octave -3dB pass band
Second order harmonics are in the pass band
Passive broad banding
CS amplifier and Miller effect
Miller effect causes a dominant pole at the input when source impedance and vgain are high, But when CL*RL > 1/omegaM, this will be the dominant pole
Remedies:
ð Add an inductor in series with the collector/drain resistor
ð Inductor impedance rises with frequency new BW limit is LC resonant f
Shunt peaked amplifier
Load impedance: choose the inductance
Check 3dB BW difference with and without L=> omega2/omega1
Select a criterion: Max BW, ZL = R at omega1 amplitude flatness,
flat group delay
Series peaked amplifier
Separating the load capacitance with series inductance
No LF zero
Combination (plus extra L)
Shunt and double series peaking
Shunt peaked L is biggest, has most delay
So a change in CGD can load first, than later CL so faster rise time
A trade-off between BW and delay
Three coils can be replaced by one T-coil
Pole-zero cancellation
ONLY for low order networks
Example of the probe
We use a tuneable capacitor
So the TF is a pole zero doublet
Other approach
ð Introduce a zero in a CS amplifier at the source
LF: gain is reduced; HF gain rises due to lower impedance of C
Chose R.C = RL.CL dominant (if the case) is cancelled
Method of open-circuit time constants
How to search the dominant pole? => Open-circuit time-constants OCtau
Underestimates the true bandwidth, accurate for dominant pole systems
OCtau does not take imaginary parts in to account
Beware: some C are not BW limiting (coupling C => short them)
Inductance: short L, add the L/R tau’s afterwards (but L generate complex pole)
Feedback techniques
The basics
Negative feedback lowers closed-loop gain
Transc, current => ideal current source others voltage source
Voltage transimpedance
Transconductance current
Impact of negative feedback on the in/output impedance
Impact of negative feedback on linearity
Higher feedback gain => linearization
Impact of negative feedback on noise
Noisy amplifier will introduce noise in the network; feedback can’t improve that (only worsen it)
vEN inputs shorted, outputs set equal
Voltage feedback opamp: compensation
Bandwidth is extended by T (A*B) A: 1st order
Compensation: make opamp with dominant pole
ð BW limitation
ð linearity problems
ð poorer slew rate (cannot be improved by feedback)
Current feedback opamp: speed (is better than voltage)
+ input => high ZIN; - input => LOW ZIN (voltage buffer!)
Bandwidth and closed loop gain are independent
Little slew rate limitation
Series vs. Shunt resp. feedback on 1 transistor
Shunt-series feedback
Effective transconductance
Design procedure: Av en R
Rf => input impedance
ð Re => gain
Consider for HF a perfect BJT (rpi & r0 = 0)
Multi stage feedback
Dual transistor broadband amplifier
Problem: every stage causes phase shift
Stability issue
Worst case scenario for stability: B equals 1
Bode stability criterion: starting at phase(A) = 0°
ð phase crossover frequency (omegaPC) when phase(A) = -180°
ð gain crossover frequency (omegaGC) when log(mod(A)) = 0 dB
At omegaPC, the loop gain A(omegaPC) must be less than 0dB
1st order: always stable when B <= 1
Caveat
Not always able to determine A, nor B
Return ratio analysis
We look for the return ratio of the error-amplifier
Caveat: an open-loop interrupts any back-propagating signals (unilaterality is assumed)
ð So look for the best unilateral device CE/S or even better a cascade output
Active broad banding
Passive is expensive (large area) and, not very good/accurate
Active components are cheaper, can match each other better
Miller effect
(The effective input capacitance of an inverting amplifier rises due to negative feedback caused by capacitance between the input and output)
The input impedance forms together with the capacitance a LPF
Miller pole:
Miller effect: non ideal amp
Use small scale approximation.
Use unilateral approximation (UA)
ð Neglect: I that leaves collector through CCB,
ð But don’t neglect the I that leaves the base
Reduce Miller effect 1: Cascode amplifier
We use a CS and a CG
ZIN of M2 is low => reduction in voltage gain of the CS
Result: lower voltage gain and Miller multiplication
Use Source substitution: replace the source in series by a new source in shunt + resistor
Conclusion: BW higher, input pole is no longer dominant but more complex and reduced output level swing and more dissipation
Reduce Miller effect 2: Source coupled amplifier
Prevents the miller effect (is halved) by eliminating an voltage inversion
Reduce Miller effect 3: Neutralization
Positive feedback: (STABILITY still OK??)
= compensation of the miller effect
ð Inject an inverted C current into the input node to
cancel the Miller C current
ð Voltage at the top is in anti phase with ID
CN (= CGD) feeds this negative current
So the INET from the gate in CN+CGD is zero
Needs to be differential, care for positive feedback
FT doublers: clever circuits with increased speed
Lower input C, but gM must stay equal
Differential pair => 2 identical C in series
ð But three devices (including current source)
Not possible? Use Battjes doubler
ð Looks like Darlington but different bias
ð Same technique as differential pair,
Bootstrapping
Form of positive feedback
Create a node with very high impedance
Two examples
ð (left) High impedance at M2 drain so Cgs becomes useless
ð (right) CGD is also bootstrapped by buffer A1