Common base amplifiers and oscillators

[JimB]

Common base amplifiers and oscillators.

Inspired by another thread here on ETO I thought that I would do some experimenting.
I built a common base amplifier circuit as shown in the attachments.

I based the amplifier circuit on one which had been the subject of much discussion recently here on ETO.

To make the measurements easier and well within the limits of my test equipment and very simple construction I decided that the amplifier should work at 10Mhz rather than VHF.

I designed the tuned circuit in the transistor collector circuit for a resonant frequency of about
10Mhz with inductive and capacitive reactances (Xl and Xc) equal to 140 Ohms.
This would require an inductance of 2.2uH and a capacitance of 114pF.

I used components which I had on hand, a 100pF capacitor and a coil with 17 turns on a 0.4 inch diameter former. Tests on the coil-capacitor showed a resonance at 9.76Mhz.
Sweeping the signal generator either side of resonance, the -3dB points were at 9.46 and 10.06 Mhz.
This gives a 3dB bandwidth of 0.55 Mhz, which corresponds to a Q of 17,7, not a particularly good tuned circuit, but good enough for the purpose of this experiment.

I built the circuit on a piece of plain copper clad board which is mounted in a test jig which I use for tests and experiments like this. See the attached photographs.

I used a supply of 10 volts and the current taken was 9.1 mA.

Injecting a signal at the emitter of the transistor and examining the voltage at the collector using an oscilloscope with a x10 probe to minimise the circuit loading, the gain of the amplifier peaked at 9.52Mhz.
Sweeping the signal generator either side of resonance, the -3dB points were at 8.28 and 11.01 Mhz.
This gives a 3dB bandwidth of 2.73Mhz, which corresponds to a Q of 3.54, the tuned circuit is obviously heavily loaded by the transistor.

Measuring input and output signal voltages using an oscilloscope, the gain of the circuit was calculated to be 34.

The input impedance of the amplifier appeared to be quite low as predicted by the standard textbooks on common base amplifiers.
During the gain tests, the output of the signal generator was loaded down quite heavily by the amplifier input.
Further measurements showed that the input impedance of the amplifier was around 5.8 Ohms.

In order to turn the amplifier into an oscillator, a feedback capacitor was added from the collector to the emitter.
In a common base amplifier, the output and input are in-phase so we just need to add some feedback to make it oscillate.

Using different values of feedback capacitor gave the results shown below:

Cfb    Frequency    Supply Current    Comments
185pF    6.28 Mhz        15.8mA        Waveform at the collector distorted
125pF    7.06 Mhz        11.5mA        Waveform at the collector less distorted than the 185pF case
100pF    7.76 Mhz         9.9mA        Connecting the scope probe to the collector kills the oscillation, frequency measured at the emitter
60pF             9.1mA        No oscillation

So what can be learned from this?

The value of the feedback capacitor affects both the frequency and the amplitude of the oscillation.
With a lower value of feedback capacitor, the signal at the collector is a cleaner sinewave with less distortion, but loading the oscillator will kill the oscillation.

When the oscillator is oscillating, the collector current will change.
For someone will little test equipment, this can be a good indication that the oscillator is working.
Short circuit the coil and watch the collector current change, now you have a good idea that it is oscillating you can go hunting to find the frequency using a receiver.

A transmitter built using this circuit will be very susceptible to the poorly defined load (antenna wire) connected to the collector.

JimB



Edited on 12/9/14 to restore the pictures lost in a server crash at ETO some time ago. JimB

 

[ericgibbs]

hi Jim,
Very nice demo.

For my own 'entertainment' I will run your circuits in LTSpice simulations, as a benchmark check for LTS.

Eric

Which type transistor did you use.?

 

[ericgibbs]

hi Jim.
This is the amplifier sim.

Checks out quite well, used a 2N2222 transistor

E.

 

[Winterstone]

Hi Jim,

I have simulated the circuit as well. I did a TRAN simulation (2N3904) and the circuit did oscillate very well and self-sustained slightly above 10 MHz.
Remark to Erics simulation: Your ac response shows the tank circuit response, but - as you certainly knows - it does NOT show the resultant loop gain. This looks quite different.

 

[JimB]

Which type transistor did you use.?

I used a 2N3904.

JimB

 

[Winterstone]

Quote JimB:
In order to turn the amplifier into an oscillator, a feedback capacitor was added from the collector to the emitter.
In a common base amplifier, the output and input are in-phase so we just need to add some feedback to make it oscillate.

I think, it isn`t so simple. What means "some feedback"? The problem is: When emitter and collector are in phase the feedback path must not add some additional phase shift. However, when a capacitor is used as feedback element it will cause - together with the input impedance at the emitter node - a frequency dependent phase shift.
So, what about the loop gain and the oscillation criterion?

This gave some motivation to perform some additional simulations - and I think I can now give some explanations HOW and WHY the circuit works.

* I have replaced the real transistor model by an idealized model (Qbreak in PSpice) without any frequency dependence and with a current gain of 100. All other elements as shown above.
Result: No oscillations in the time domain. This can be confirmed by loop gain simulations which show that the phase never reaches zero deg. This seems to be clear from the tank circuit which is the only one with a frequency dependence.
* Then, a emitter-base capacitance was added (50 pF).
Now, the loop gain reveals a capability to oscillate because the phase crosses the 0-deg line with a loop gain slightly larger than unity. And indeed - a TRAN analysis result in an oscillation of approx. 10 MHz.
* Summary: The circuit is able to oscillate only because there is a capacitive divider at the emitter node (consisting of the feedback C and the internal Cbe) which forms a feedback path that allows fulfillment of the oscillation condition.
*That means: The circuit under discussion is one of the classical RF topologies and can be analyzed based on the linear harmonic oscillator theory - and there is no "transistor switch-on or switch-off" effect (as mentioned earlier in another thread).
* If desired I could show the loop gain response.

 

[JimB]

Winterstone said:

I think, it isn`t so simple. What means "some feedback"? The problem is: When emitter and collector are in phase the feedback path must not add some additional phase shift. However, when a capacitor is used as feedback element it will cause - together with the input impedance at the emitter node - a frequency dependent phase shift.
So, what about the loop gain and the oscillation criterion?

Very good points, I have been thinking about that as well.

In my original post, I estimated the input impedance of the amplifier at 5.8 Ohm.
This was the magnitude, phase was not considered. Also I did not have a great deal of confidence in the accuracy of the measurement.

I have since remembered another piece of equipment which I have and had not thought of using to measure the impedance even though it is designed for such things.
It is an antenna analyser (effectively a simple Vector Network Analyser, google AIM 4170 for details), that bit of equipment shows the input impedance at 9.95Mhz as 4.55 Ohms with a phase angle of 5 degrees.
It resolves this into a series resistance of 4.539 Ohm and a series inductance of 6.6nH.
So the input impedance is predominantly resistive.

In terms of loop gain, the division ratio of the reactance of a 100pF capacitor at 9.59Mhz (159 ohms) and a resistance of 4.5 Ohms is approximately 35, which is temptingly close to my measured gain of 34, for a closed loop gain of 1.
However, this does not yet adequately explain the required phase shift.

I need to think a bit more about this.

JimB

 

[JimB]

After thinking a bit more, (and playing with the big boys toys!) oscillation is taking place at about 7.5Mhz not 9.5Mhz due to the feedback capacitor being effectively in parallel (almost) with the tuned circuit, so my calculation of the reactance of the feedback capacitor is wrong.

A bit more measurement data, using the scope and looking at the waveforms at the collector and the emitter of the transistor, the displayed phase shift is of the order of 37°, not quite what we would expect.

More thinking and playing required.

JimB

 

[ericgibbs]

* Summary: The circuit is able to oscillate only because there is a capacitive divider at the emitter node (consisting of the feedback C and the internal Cbe) which forms a feedback path that allows fulfillment of the oscillation condition.
*That means: The circuit under discussion is one of the classical RF topologies and can be analyzed based on the linear harmonic oscillator theory - and there is no "transistor switch-on or switch-off" effect (as mentioned earlier in another thread).

Hi Winterstone,
I fully agree with above statement from your thread, its basically a modified Colpitts oscillator and the capacitance divider is required.

A plot of the loop gain response would be helpful for a discussion on this topic
E.


hi Jim,
Firstly , thanks Jim for creating a technical discussion on this topic.

I did find when doing a quick LTS check on the 60pF thru 185pF, the results did not agree with your original posted frequencies, I will re-run to check my findings.
Also the Zin of 5.8R seems too low.?

Eric


EDIT:
Reverse checking the osc circuit, indicates a Cbe of 8pF is used for the 2N3904 model. checking the LTS 'standard.bjt' shows a value of 8E-12 for CJE

 

[Winterstone]

In completion of my contribution (post #6) here are my simulation results:

(1) Idealized transistor (Qbreak), see attachement for loop gain and TRAN analysis.
Tank circuit: 2.2uH||100 pF. Common-base capacitor Cb: =0.5nF.
Feedback cap: Cf=150 pF
As can be seen, the capacitive divider (Cf and Cbe=25pF) is necessary to let the phase cross 0 deg for a loop gain>0 dB.
It is interesting to note that the oscillation does NOT occur at the magnitude peak frequency.
This is in accordance with other bandpass based oscillators which oscillate at a frequency other than the bandpass resonant frequency due to amplifier phase contribution.
However, it is to be mentioned that the oscillation frequency in TRAN analyses is not identical to the loop gain cross-over frequency. This may be caused by the loop gain analysis technique which is a bit "tricky" due to loading effects at the opening of the loop. I am working on some improvements.
Of course, after the decaying switch-on transient self-sustained oscillations only start after the base capacitor Cb has reached its final state which keeps the transistor conducting.

(2) Real transistor models, simulation results not shown.
With a 2N2222 model there were oscillations at 9.8 MHz and for 2N3904 the result was 10.2 MHz.
This shows the influence of different models with different capacitive properties.

 

[Winterstone]

Correction:
Due to a reading error the given cross-over frequency (loop phase=0 deg) is 9.97 MHz (and NOT 9.4 MHz).
Thus, the difference between this frequency and the observed oscillation frequency (10.3 MHz) is not as much as it was before. I think, this difference can be attributed to the loop gain simulation which - in this case - is rather problematic due to heavy loading of the tank circuit associated with large phase excursions.

 

[lebevti]

[latex]\int[/latex]

asdf