Valve frequency multipliers

Why?

The Colpitts-derived xtal oscillator is an EF91 and the

Why not use a 150 MHz xtal?

One dual triode could make two injection-locked oscillators.

Because that is the challenge I have set myself.

a) I don't have a 150 Mc/s crystal, but I do have one for 15 Mc/s

b) The same crystal is used on transmit and receive, but there has to be +/- 2.5 Kc/s deviation for F.M. when transmitting and a 100 Kc/s offset on receive. I wasn't at all confident that a 150 Mc/s crystal could be pulled that far whereas these variations become 10 Kc/s and 250 c/s respectively if the crystal is followed by a x10 multiplier and I knew in advance that it should be feasible with a 15 Mc/s crystal.

~~~~~~~~~~~~~

• Long explanation:

A reactance valve works by feeding a variable amount of 90-degree phase-shifted current into the oscillatory circuit. This is uaually achieved by having a resistor-capacitor phase shift network with the resistor from anode to grid (with an appropriate blocking capacitor) and a capacitor between grid and earth. The audio is also superimposed on Grid 1 to vary the R.F. gain of the valve.

At a Mc/s or two this is satisfactory, but at 15 Mc/s the divider action of the network attenuates the signal reaching the grid. Even with no external capacitor, the internal grid-earth capacitance is large enough that virtually no signal voltage appears on the grid. If the resistor is lowered in value, it imposes such a load on the anode that the output from the valve is negligible and the resistor loads the oscillator heavily (and may stop it altogether).

I was saved from this dilemma by the "Electronic Designers Handbook" by Landee, Davis & Albrecht. It noted that the 90-degree phase shift could equally well be achieved with an R-C network or an R-L network. As there was a low impedance source of oscillator signal at the cathode of the oscillator, I used an R-L network to inject a phase-shifted signal into the cathode of the reactance valve. This had the twin advantages that the reactance resistor also acted as the cathode resistor for both valves and it freed-up Grid 1 for injection of the audio signal from a high impedance source.

~~~~~~~~~~~~~

I had considered that, but if they became unlocked for any reason (failure of the oscillator) I risk transmitting out of the allocated band or drifting all over the place. Not a good thing to do.

On Fri, 31 Jan 2025 13:51:25 +0000, snipped-for-privacy@poppyrecords.invalid.invalid (Liz Tuddenham) wrote:

Nowadays, just ask AI!

Generating 2nd and 5th harmonics from a thermionic (vacuum) tube involves specific operational conditions and circuit configurations. Here's how you can achieve this:

Conditions for 2nd Harmonic Generation:

Non-linear Operation: Grid Bias: The tube should be biased to operate in a non-linear part of its characteristic curve. This typically means biasing the grid slightly negative so that the tube's operation swings into the non-linear regions of the plate current-grid voltage characteristic. Signal Amplitude: The input signal should be of sufficient amplitude to drive the tube beyond its linear operating region, causing even harmonics (like the 2nd) to be generated due to the non-linear response of the tube. Circuit Configuration: Single-ended Amplifier: A simple single-ended triode or pentode amplifier can naturally produce even harmonics if driven into non-linearity. Cathode Bias: Using cathode bias can help in achieving the desired non-linear operation by introducing a dynamic bias that changes with signal level.

Conditions for 5th Harmonic Generation:

Non-linearity and Symmetry: Higher Order Distortion: The 5th harmonic is an odd harmonic, which means the tube's operation needs to be asymmetrical around zero to produce it effectively. However, tubes naturally produce less odd harmonics than even ones unless manipulated. Signal Drive: A very strong input signal might push the tube into regions where higher-order harmonics become significant, but this approach might also introduce unwanted distortion. Circuit Techniques: Feedback: Using feedback in the right way can amplify certain harmonics. Negative feedback can reduce even harmonics while potentially emphasizing odd ones if designed for such. Multiple Stages: Cascading multiple tube stages can increase the harmonic content by compounding the distortion at each stage. Each stage could be designed to push further into non-linearity. Harmonic Enhancement Circuits: Specific circuits like those with transformer coupling where the primary and secondary windings' interaction can be tuned for harmonic enhancement. Also, using resonant circuits or filters tuned to the 5th harmonic can help. Tube Selection: Tube Characteristics: Some tubes are naturally more harmonic rich than others. Tubes like the 12AX7 might be used for their gain and non-linearity characteristics, but for more harmonic richness, sometimes pentodes or specific triodes known for distortion characteristics are employed. Operating Point: Grid Voltage: Adjusting the grid voltage to operate closer to cutoff can increase harmonic distortion since the tube's response becomes more non-linear near cutoff. Power Supply Design: Power Supply Stiffness: A less stiff (or intentionally loose) power supply can allow tube operation to vary more with the signal, potentially increasing harmonic distortion.

In practice, generating a specific harmonic like the 5th is challenging because tubes are inherently more suited to producing even harmonics when driven hard. To get a pronounced 5th harmonic, you'd often need to tailor the circuit design, including feedback, load characteristics, and perhaps even use of additional harmonic generators or filters post-amplification.

Why not? Sounds like fun!

You'd be well into overtone territory by then. ISTR fundamentals are limited to about 20Mhz. That may or may not matter in this instance.

Crazy.

Dumb AI. We WANT distortion. Specifically about a 10% plate conduction duty cycle.

Operate it deep in cutoff, off most of the time. A high amplitude drive and grid-leak bias would be good.

Why is that?

AI platitudes, harvesting conventional wisdom.

Well, there's AI and there's AI. The above was Grok. I did try to get the 'new kid on the block' (DeepSeek) to regale us with its wisdom on the subject, but it didn't seem to have any. Chat GPT might be worth asking....

Won't it "learn" from amateur posts to chat sites? Collect opinions?

That is exactly what I am doing and it doesn't appear to be working. With 100v on the anode, 15 Mc/s at 25v pk/pk on the grid and a 22k grid leak, the peak current for one triode of an ECC91 is around 20 - 30 mA at the positive peak of the grid swing. The average anode current is around 2.5 mA, so the conduction period is about 10%.

A 75 Mc/s parallel-tuned circuit in the anode circuit is giving so little drive to the following stage that I can't see any change in the average grid voltage of that stage caused by the drive.

If you can make it generate a 25v 15MHz square wave then you should have approximately 8v of 3rd harmonic and 5v of fifth harmonic (as well as some even harmonics because of how real valves behave). Slight low pass filtering to discourage any higher harmonic content will help here.

What does the output waveform actually look like at 15MHz?

Does the valve have enough gain at 75MHz to cope with a tuned load? ECC91 sounds like it ought to unless something else gets in the way.

This is a bit tricky because my oscilloscope only goes up to 50 Mc/s and I also have to take account of the loading effect of the probe. The waveform on the grid of the 'x5 multip[lier' triode is like an assymmetric 15 Mc/s sinewave with the top of each positive half-cycle squashed and flattened a bit. The tops are just above 0v and the bottoms go down to -25v. (The ECC91 cuts off with around -5 to -8 volts on the grid, depending on anode voltage.)

There is no voltage on the cathode, which is earthed and is common to both triodes. The waveform on the anode is a bit meaningless because it would be outside the range of the 'scope but I have investigated it three other ways:

It is rated to give more than 1W output at 200 Mc/s.

Well, Grok at any rate cites its sources so you can make a qualitative judgement for yourself (assuming that's what you're asking).

I'd be more selective of the crystal frequency. Two triplers might get you there with a lot less grief.

You can do that with something like a 6CA8 - pentode side doing xtal osc and 3x with the triode another x3. See ARRL handbook 1969 (46th Ed)CH17 p239. That was for 220MHz (they stuck a P-P tripler on the output, so 3x triplers)

RL

From old school notes, 5th harmonic peaks at 50degrees conduction,

RL

That's the sort of info I wanted. I'll experiment with conduction angles but my results are so far removed from what I need that I don't think I shall get enough improvement that way..

Spice it!

LT Spice has tube models.

I regret to say I think you are right :-(

Quintuplers just don't seem to work in those circuits so I have placed an order for a 16.656 Mc/s crystal, which will triple-triple to 149.904 Mc/s. That will mean bringing the VFO down to a range of 3.904 to 5.904 Mc/s, which should be relatively easy.

Revised block diagram at:

I just hope the new crystal will work in the same circuit as the old one and I won't have all the hassle of redesigning it.

...but I haven't got Spice.

It's free, works great, and is easy to learn how to use. And it's fun to run.

I learn a lot playing with circuits in LT Spice. I discovered a new circuit mostly by accident recently, exploring hunches.

I also use it for simple things, like designing voltage dividers or calculating wavforms in RC circuits, things that would be tedious to math with a calculator.

When it sims a few gigavolts on a node, or a megawatt in a resistor, it won't hurt you!

LT Spice is free and works very well indeed. And you can't beat the price. No idea how good it's valve models are but I bet they do exist.