#389 Two-stage Common Emitter Amplifier
Designing a two-stage common-emitter BJT amplifier.
Notes
Cascading two common-emitter amplifiers is a means of achieving high voltage gain. Voltage gains from several hundred to several thousand are possible.
I’m going to try and noodle out a theoretical design of a two-stage Class A amplifier, and then test the actual performance. Do not take my calculations as gospel. My main sources for the theory were Electronic Principles by Albert Paul Malvino and a Multistage Transistor Amplifiers YouTube tutorial by The Offset Volt:
Designing a Single Stage for Gain ~ 10
Parameter | Design/Spec Value |
---|---|
A (gain) | 10 |
Vcc | 9V |
Icq | 4mA |
Vceq | 4.5V |
ß (hFE) | 100 - 400 |
hie | 1kΩ - 10kΩ |
Vbe | 0.7V |
RL | 2.2kΩ |
Also:
- r’e = hie/hFE = 10kΩ/400 = 25Ω
Calculate collector + emmiter resistance for desired gain at the Q point
- Rc + Re = (Vcc - Vceq) / Icq
- Rc + Re = (9V - 4.5V)/4mA = 1.13kΩ
- assuming A ≅ Rc/Re
- Re = 1.13kΩ - Rc
- Re = 1.13kΩ/11 =100Ω
- Rc = 1.03kΩ, say 1kΩ
With the selected components, the theoretical gain is thus 10
Calculate the combined bias gang resistance
Base current at the q point
- Ib = 4mA / 100 = 0.04mA
Assume current through the gang at 10 x Ib as a rule of thumb to ensure “stiff” biasing i.e. 0.4 mA
So combined resistance = 22.5kΩ
Calculate the resistance of R1 and R2 components of the bias gang
Lower resistor R2:
voltage = 0.7 + Ic x Re = 1.1V
therefore R2 = 2.75kΩ say 3kΩ (standard value)
and therefore R1 = 19.5kΩ say 20kΩ (standard value)
Measure Performance - Single Stage
With 0.2V peak-peak 10kHz input, and Re = 100Ω (without emitter bypass), measured results:
- 0.2V input peak-peak
- 1.98V output peak-peak
- actual gain = 9.9
Conclusion: very close to design gain of 10, undistorted Class A performance.
Adjusting Single Stage Design with Bypass
I decided to add another 100Ω/100µF emitter bypass to the design, for a few reasons:
- it’s a common feature of such designs
- it provides negative feedback to stabilise for transistor variations (AC signals will vary the emitter resistance with inverse relationship to input/output differential, thus combating variation)
- reduces the gain a bit (so I can work with larger input signals in two-stage configuration while staying within the 9V supply limits)
Recalculating with this variation. DC characteristics are now:
- Re = 200Ω
- A ≅ Rc/Re = 1kΩ/200Ω = 5
combined base bias resistance should still be = 22.5kΩ
Lower resistor R2 is now:
voltage = 0.7 + Ic x Re = 1.5V
therefore R2 = 3.75kΩ say 3.6kΩ (standard value)
and therefore R1 = 18.9kΩ say 20kΩ (standard value)
Calculating Two-stage Performance with RL load
With:
- component values duplicated the from the single-stage design for two stages
- with the extra
100Ω||100µF
emitter bypass - and RL = 2.2kΩ
predicted performance calculated as follows:
DC analysis:
- Vb = Vcc*Rb2/(Rb1 + Rb2) = 9V * 3.6kΩ/(20kΩ + 3.6kΩ) = 1.37V
- Ve = Vb - 0.7V = 0.67V
- Ie = Ve/Re = 0.67V/(200Ω) = 3.35mA
- Ie ≅ Ic
- Vrc = Ic * Rc = 3.35mA * 1kΩ = 3.35V
- Vce = Vcc - (Vrc + Ve) = 9V - (3.35V + 0.67V) = 4.98V
- Isat = Vcc/(Re + Rc) = 9V/(200Ω + 1kΩ) = 7.5mA
Second stage:
- assuming thermal voltage VT = kT/q = 25 mV
- r’ej = VT/Ie = 7.5Ω
- A2 =
Rc||RL/(Re2 + r'ej)
=1k||2.2kΩ/(100Ω + 7.5Ω)
= 6.4 - Rin(base) = ß(Re2 + r’ej) = 100(100Ω + 7.5Ω) = 10.75kΩ
- Zin =
Rin(base)||Rb1||Rb2
=10.75kΩ||20kΩ||3.6kΩ
= 2.38kΩ - A1 =
Rc1||Zin/(Re1 + r'ej)
=1kΩ||2.38kΩ/(100Ω + 7.5Ω)
= 6.55 - Total gain A = A1 * A2 = 6.4 * 6.55 = 41.92
Assuming headroom for say 7V peak-peak, input limit would be around 0.17V peak-peak before clipping.
Measure Performance - Two Stage
With 0.1V peak-peak 10kHz input, and Re = 200Ω (with 100Ω/100µF bypass), measured results:
- 0.1V input peak-peak
- 3.88V output peak-peak
- actual gain = 38.8 = 31.8 dB
- error in the theoretical gain calc = 7%
Conclusion: predicted performance was quite close to the actual performance. And .. I have a functioning two-stage Class A amplifier to boot, with performance “in the ballpark” of the design.
Performance With an Ugly-style Build
With an ugly-style build, I see even better performance, and improved gain:
- 0.1V input peak-peak
- 4.32V output peak-peak
- actual gain = 43.2 = 32.71 dB
- error in the theoretical gain calc = -3%
More gain when built “properly”, this time overshooting estimates by a small margin (less than the tolerance on components).
At 400kHz, perfromance is maintained:
Construction
Testing on a breadboard:
Testing with an ugly-style build: