AI

Jan 10, 2025 Last reply: 1 year ago 10 Replies

Gentlemen,



I've just been goofing around with X's AI assistant, Grok. After feeding in a few straightforward questions and getting surprisingly impressive answers, I decided to try something a bit more subjective to see how it would cope. Here's what happened...


My question:



Which of the common 'classic' oscillator designs is the most impressive or ingenious from a technical standpoint?



Grok's Answer:



From a technical standpoint, several classic oscillator designs stand out for their ingenuity, but one that is often highlighted for its elegance and versatility is the Colpitts oscillator. Here's why:



Design Ingenuity:



Simplicity and Efficiency: The Colpitts oscillator uses a simple configuration with a transistor (or vacuum tube in older versions) and a combination of capacitors and an inductor to create a feedback loop. Despite its simplicity, it achieves stable oscillation with just a few components. Feedback Mechanism: It employs a capacitive voltage divider for feedback, which is particularly ingenious. The two capacitors in series (C1 and C2) form a voltage divider that provides the necessary phase shift for oscillation. This method naturally provides the correct phase without additional complex circuitry, making it both efficient and reliable. Frequency Stability: The frequency of oscillation in a Colpitts oscillator is determined primarily by the LC tank circuit's resonant frequency, which can be calculated as f=12πL(C1⋅C2)/(C1+C2) . This formula highlights how the design inherently stabilizes frequency through the interaction of inductance and capacitance. The use of capacitors in the feedback loop also helps in reducing the impact of transistor parameter variations on frequency stability. Versatility: The Colpitts oscillator can be adapted for use across a wide range of frequencies from audio to RF. By changing the values of L, C1, and C2, one can easily tune the oscillator to different frequencies, making it adaptable for various applications from radio transmitters to clock generators. Low Noise and High Purity: When designed correctly, the Colpitts oscillator can offer low phase noise, which is critical for applications like communication devices where signal purity is paramount. The design inherently minimizes noise by reducing the number of components in the feedback path. Ease of Adjustment: Adjusting the frequency is relatively straightforward with minor changes to the capacitor or inductor values, making it practical for both prototyping and production environments.


Historical Context:



Invented by Edwin H. Colpitts in 1918, this oscillator was a breakthrough in the era of vacuum tubes, which were later adapted for transistor use. Its principles have stood the test of time, still being taught and used today in various forms.


Comparison:



While other oscillators like the Hartley oscillator (which uses an inductive voltage divider) or the Clapp oscillator (a variation of Colpitts with an additional capacitor for better frequency stability) are also significant, the Colpitts design is often praised for its balance of simplicity, efficiency, and versatility.


In conclusion, the Colpitts oscillator stands out for its impressive combination of simplicity, effectiveness, and adaptability, making it one of the most ingenious classic oscillator designs from a technical perspective.


**************************************************************************

Any observations?


Yeah. The resonant frequency formula is wrong. To first approximation, it's f=1/(2π sqrt(L C1 C2 / (C1+C2))). (Inline algebraic notation is a dog...)

It looks like Grok has trouble reading formulae.

Of course, it doesn't really know what it's talking about. It just ladles out some phrases it gathered from various web sites on the subject.

Jeroen Belleman

Here's my modified Colpitts:

formatting link

**************************************************************************

The formula error is not Grok's mistake. The sqrt didn't survive the cut'n'paste I'm afraid. I didn't reckon on someone spotting that so didn't mention it in the OP.

We need a new word for this kind of thing. I propose "plagiamaton", meaning something like "agent that steals ideas".

For some of the more awkward phrases, I even found where it got them from. For example, the phrase starting with "This formula highlights..." seems to come almost verbatim from a particularly poorly written web page by Keysight.

Jeroen Belleman

I suppose we're still in the early days of AI in terms of this kind of usage, so can't expect perfection for anything too demanding (not that this was, but asking for an opinion from a computer is a bit more challenging than stuff like 'who holds the record for drinking the most beer?')

Is that some sort of filtering you've tacked on there?

That's one way of looking at it. Another is that it doesn't know what it is talking about, which makes the offering useless. When we throw in your own ignorance, which let you post version of the text that had lost the square root symbol, it becomes actively misleading.

Donald Trump and Zero Hedge are making the world a more dangerous place by serving up the kind of plausible nonsense that you find attractive, and we need to find a way of neutralising their destructive contributions. Educating twits like you to the point where you could do critical thinking would be a useful step in the right direction, but just shutting you up would help.

R1 L2 C3 is a phase shifter. Node R drives C2 which series resonates with the big virtual inductor that L1 and C1 form near their infinite resonant peak.

The classic Colpitts gets its phase shift from an RC, which wrecks the Q. Tank loading by the follower input further hurts Q. My objective was to keep Q up and keep phase noise (or jitter, in my world) down.

The BUF602 is dynamite. HiZ in, LoZ out, gain=1, bandwidth 1 GHz.

Wow! Thanks, I gotta get me some of them thangs.

Most of my triggered oscillators run around 50 MHz are used to make delay generator timebases. If I do a conventional Colpitts and use an NPN transistor with Ft=300 MHz, the beta at 50 MHz is only 6, so it kills the Q of the tank.

Fets are better, but have other issues. The BUF602 is really good. We have them in stock, from another project, so I gave it a try. Given the complexity of such oscillators, it's less work to just build one and see what happens.

I think tubes have some transit-time effect that makes the grid have an ohmic component at high frequencies too.

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required