Understanding Colpitts "control" elements

Question

I'm struggling to build a Colpitts oscillator that runs above about 2 MHz. Unfortunately, my understanding of this thing's theory of operation is severely minimal. My end goal is to build a crystal controlled oscillator at 28 MHz for 10 M band (so I'll need it to be stable and clean), but for now I'm running the circuit below (without a crystal).

^{simulate this circuit – Schematic created using CircuitLab}

This circuit came from an Analog Devices tutorial I found here: Analog Devices Colpitts oscillator tutorial. It works fine up to about 1 MHz, but the output swing gets smaller and smaller as the frequency goes up. At 1.5 MHz it's giving about 650 mV, but is fairly stable. By the time I get to 2 MHz, oscillation is unstable, and output when it does run is down to about 300 mV.

EDIT: Based on Tony Stewart's comments (see his answer below, and some of my follow up in an earlier edit lower down), I changed the bias on this so the DC collector current is about 8 mA (that was chosen for the 2N5770, but isn't wildly far off the 10 mA target for the 2N3904). This was very effective (though still the circuit as a whole falls short). I can now get intermittent oscillations up to 5 MHz or so (with 1 uH inductor), and it's strong up to about 3 MHz. However, it still fades and quits oscillating sometimes at 5, and clearly that's nowhere near my target of 28. So, I would like to understand the feedback of this system better if anyone can help me with that (and anything else that you think might be relevant).

Note: The reason that I'm looking at this configuration specifically is because I know how to add a crystal to it directly. (See the ARRL handbook version noted below). If you think I should use some other configuration, it would be very helpful if you can show me how to add a crystal to the circuit after I tune up the LC network first.

I found that with a 2N2222 it wouldn't run at 1 MHz, but with the 2N3904 it is better. I also tried a 2N5770 (Ft around 900 MHz) and that gives just a little more output, but still won't go past about 2 MHz.

So, part one of my question is how does one control the output level of this kind of circuit? I've tried changing several things, but nothing has so far made any obvious difference. Part two is how do I get this thing to run at a useful frequency?

A little extra background: I started with the circuit from the ARRL handbook (see this PDF--it's an older edition than my copy, but the circuit is identical ARRL handbook) It's the circuit shown in figure 9.24 on page 23. However, this didn't start at all at 16 MHz--the "slowest" crystal I have. I modified it by shorting the crystal out and changing the tank but it didn't oscillate at all until I also removed the 1k resistor they show shorting the tank. But, with the 1k removed, and the crystal replaced with a short and greatly modified L&C values for lower frequency, I can make this behave pretty much the same as the above (aside from biasing, it is essentially the same.) They're claiming that either a 2N2222 or a 2N3904 can get this thing to run up to 30 MHz, but that's not what I'm finding.

EDIT:

---

Following on from Tony Stewart's observations, I have more questions:

Tony, you said "If you do not operate in the current range specified for Ft= BW then hFE will be reduced In your case by 50%. That is specified @ 20 mA with is the near the max. of the hFE vs Ic curve."

I'm not really understanding this (lack of education on my part). I think you're referring to this part of the datasheet:

But, if this is what you meant, does that mean I should bias this thing for 10 mA? Or am I looking at the wrong thing and/or entirely misunderstanding?

(Another edit...) I see that the 2N2222 does indeed list 20 mA in the same section that I pulled the 10 mA figure from the 2N3904 datasheet. So, I think that I have interpreted your information correctly??

Next you said "You can alter this high current requirement by using more feedback or less attenuation and thus more loop gain to achieve greater than unity for oscillation." Can you tell me where the feedback is in this circuit, I don't understand it well enough (indeed, this is one of the "control" elements I was referring to in the question title). And attenuation? What components are providing that? What do I need to change?

---

Answer 1

Problems:

1. If you do not operate in the current range specified for Ft= BW then hFE will be reduced In your case by 50%. That is specified @ 20 mA with is the near the max. of the hFE vs Ic curve. Also if it saturates, hFE may reduce up to 90% or so at Vce(sat) @ Isat.

You can alter this high current requirement by using more feedback or less attenuation and thus more loop gain to achieve greater than unity for oscillation.

2. Your design is Common Base so the ratio of Zc/Ze is a major factor for the gain ratio. So reducing Re will boost the gain and raise Ic. My design is Common Emitter with negative feedback from an open-loop gain of Rc/Re.

It depends on the current gain BW product and how much saturation effects reduce hFE by up to 90%. So the configuration of current gain and impedance gain to achieve I(f)*Z(f)= V(f) voltage gain > 1 to get a stable oscillator.

Xtals in this range often use harmonics or overtones with prefilters.

Here is a simulation that oscillates with any hFE=10 at 2MHz which just about any transistor can do with GBW = 20 MHz for margin and the PN2222A has 200 MHz GBW.

Here the LC Pi filter with current limiting, the oscillator starts with a "bang" with full saturated oscillations then reduces to a nice sine wave depending on Re or other R values.

Notice the R Ratios of 100:10:1 in choices for linear gain.

With modifications to add more negative feedback and gain, the spectral purity of the sine wave can be improved greatly. This one settles in a few hundred cycles which is an indication of the Q.

The log-log spectral density is shown above using Falstad's FFT.

ONsemi datasheets indicates the optimal GBW for the PN2222A but the 2N3904 only has hFE vs IC plots.

Answer 2

The Colpitts oscillator is difficult to analyse accurately, but there are versions around which are stable enough for experimenting in simulation and real prototypes. For example:

https://www.circuitlab.com/circuit/2z4bxtfb5jtg/colpitts-6mhz/

This circuit starts oscilation in class-A and once the amplitude is high enough moves to class-C. The class-C mode gives amplitude stability and a fairly clean waveform at "collector"; better shape but lower amplitude at "emitter". The "kick-start" switch is an artifact of simulation - not needed in the real world.

I have built a breadboard / proto-board version and it works very much as the simulation.

how does one control the output level of this kind of circuit?

The tank circuit of this configuration is self-limiting with an amplitude of approx. half the supply rail.

how do I get this thing to run at a useful frequency?

The configuration in this answer has a common base amplifier, less susceptible to the Miller effect. Also probably higher Q because of series resonance (the version in your question is parallel resonant for which it is more difficult to control damping).

Answer 3

At this point, I now have this running at 28 MHz, though the output is visibly less of a clean sine wave (not terrible, but I need to investigate improvement).

Two things were key. One of them was Tony Stewart's information about the relationship between Ic and hfe (which was entirely new to me, many thanks Tony!!!). I reduced the emitter resistor to 390 ohms, which put the collector current in the range on 8 mA, which is what's wanted for the 2N5770 that I'm now using. Having done that, my peak possible frequency was right about 5 MHz--a huge improvement by itself.

The second thing was the feedback value. I suspect that Tony was trying to explain this to me, but if so I didn't--and still don't :( understand the terms of reference. However, I found this video (unusual in the youtube world in that it's competently explained rather than merely stated for rote learning, which made me very happy). Colpitts oscillator tutorial by "the offset volt" From this I discovered how the split in the capacitors works to control the feedback proportion. I had a feeling this split was involved, but didn't know how, or how to control it. I had tried changing the values by making the differential more rather than less. It turns out I needed to make the split less to get more oomph into the emitter to boost the thing's ability to sustain oscillation. So, I changed the two capacitors to equal values, and the circuit fired up far more strongly. I was also able to reduce the caps to 68 pF, and at that point, it's running at 28 MHz.

After that, I retuned it for 16 MHz (the only crystal I have at my disposal right now) and added the crystal to the circuit. That worked perfectly too, and locked the frequency dead on (well, to the two decimal places that my oscilloscope measures frequency at least).

Many thanks to everyone for your contributions, and particularly Tony. Tony, if in fact your comments were trying to tell me about the capacitor ratio, I apologize for not understanding, and not marking your answer as "accepted", certainly the part I did understand was critical and is greatly appreciated.