Very stable single frequency audio oscillator (discrete components)

Hi, I need a very stable audio oscillator circuit that operates at a single frequency somewhere at 30Hz.
Square wave is preffered and stability is the key point, since it will be used as a reference to lock a higher frequency oscillator.

I would use a multivibrator, but how stable it should be if using low ppm resistors and capacitors?

In that case, can it be made so that the timing resistors are made high and the capacitor very low values?
I am asking that because it is easy to find very low ppm resistors, but only capacitors up to a few nF can be made NP0.

Some ideas are here, how about the high stability version of figure 3?
https://www.nutsvolts.com/magazine/article/bipolar_transistor_cookbook_part_6

The big problem is that the Vbe of the transistors change with temperature. If you are after crystal like performance (100 ppm per deg C). Then it will be impossible for an amateur to do unless you have a temperature controlled oven to test it in and a stable oscillator to compare it with. Why not build a crystal oscillator and divide its output down to the correct frequency. Another thought put it and its PSU regulator in a low temperature oven set to say 50 degs C in use.
Frank

chuckey said:

The big problem is that the Vbe of the transistors change with temperature. If you are after crystal like performance (100 ppm per deg C). Then it will be impossible for an amateur to do unless you have a temperature controlled oven to test it in and a stable oscillator to compare it with. Why not build a crystal oscillator and divide its output down to the correct frequency. Another thought put it and its PSU regulator in a low temperature oven set to say 50 degs C in use.
Frank

Click to expand...

You cannot build stable discrete dividers in reasonable component count. This is the problem with dividing down from RF to audio.
I would say a +/-5Hz accuracy is enough, since this would be translated to +/-5Hz on RF after locking, which is not bad at all.
I see with interest the circuit in figure 3 in this page https://www.nutsvolts.com/magazine/article/bipolar_transistor_cookbook_part_6 but transform it to use darlington pairs like in figure 4. This will also allow a much higher timing resistance, which is desirable to low down the timing capacitors values.
The circuit in figure 3 states a 0.5% frequency stability, would that be adequate for my purpose?

If you worry about stability, why not use a watch crystal divided down to the intended frequency, e.g. 32 Hz = 32768 / 2^10.

Transistor multivibrators are affected by forward voltage and current gain variations. A CMOS 555 can give you better accuracy. High resistance level increases the effect of leakage currents. NP0/C0G capacitors are available up to at least several 10 nF, but probably not in a DIY shop around the corner.

Hi,

HC4060 and a crystal and come Cs, maybe Rs.

Klaus

neazoi said:

I need a very stable audio oscillator circuit that operates at a single frequency somewhere at 30Hz.
Square wave is preffered and stability is the key point, since it will be used as a reference to lock a higher frequency oscillator.

Click to expand...

You need to specify what you mean by very stable.

Theoretically speaking, a square wave is not a single frequency. By convention, the sine and cosine functions form the basic of all periodic functions.

Most likely you will need to have an oven for keeping the temperature constant to 0.1C or better.

Usually we use an accurate high frequency clock (Rb or Cs) or NMR signals (D2 lock) as reference lock- they can provide 1e-9 stability or better.

In any case, drift will be your only enemy.

c_mitra said:

You need to specify what you mean by very stable.

Theoretically speaking, a square wave is not a single frequency. By convention, the sine and cosine functions form the basic of all periodic functions.

Most likely you will need to have an oven for keeping the temperature constant to 0.1C or better.

Usually we use an accurate high frequency clock (Rb or Cs) or NMR signals (D2 lock) as reference lock- they can provide 1e-9 stability or better.

In any case, drift will be your only enemy.

Click to expand...

You are absolutely right in both of your points.
See my post #3 related to required stability. I do not know how hard will it be without an oven, given the response in post #4
And yes a square wave is not of a single frequency, but it will be used to switch a flop flop (discrete components as well), so it would be better to be a square wave I think.

KlausST said:

Hi,

HC4060 and a crystal and come Cs, maybe Rs.

Klaus

Click to expand...

This is the correct answer. The 4060 fits the bill perfectly.

If for some reason you're opposed to using a crystal+divider, linear technology sells some precision astable oscillators capable of very low frequency operation.

mtwieg said:

This is the correct answer. The 4060 fits the bill perfectly.

If for some reason you're opposed to using a crystal+divider, linear technology sells some precision astable oscillators capable of very low frequency operation.

Click to expand...

I mentioned discrete in the topic. If it can be done using discrete components.

A tolerance of 0.5% may appear to be tight, but it is 5000 PPM.
Even the cheapest crystal oscillator is an order of magnitude more stable.

If you want to build a crystal oscillator using only discrete components, why not use a classic Pierce crystal oscillator?

If you want a stable 30Hz as a reference to lock a higher frequency oscillator then why not make the higher frequency oscillator a crystal oscillator??

I suspect this is part of a 'huff n puff' stabilizer, perhaps Neazoi can confirm.

It is a method of locking a higher frequency oscillator to a low frequency reference without using a PLL. It uses the LF reference to open a gate, allowing the HF to reach a counter. Depending on whether it overflows or not, you raise or lower an analog voltage to 'pull' the HF oscillator one way or the other. Essentially it produces a modulo 30 count of the HF and through a long time constant network, locks it to a multiple of 30Hz.

It's a simple and effective frequency stabilizer and the resulting frequency has the same drift as the LF (30Hz in this case) oscillator rather than the drift being multiplied by the PLL divider. It's drawback is slow locking speed and dependency on the HF oscillator being stable or it drifts in 30Hz steps.

Brian.

betwixt said:

I suspect this is part of a 'huff n puff' stabilizer, perhaps Neazoi can confirm.
Brian.

Click to expand...

That is absolutely correct.
I need to investigate if the stability of the multivibrator circuits is enough for producing this audio reference.
Two crystal oscillators mixed down would be too difficult to keep unlocked when they operate 30hz apart.

This would be no more difficult than building a high stability VFO to operate at several Mhz, except its at 30Hz.

All the very same principles apply, including adding temperature compensation to the tuning components. Temperature compensation is conceptually simple, but the difficulty in achieving perfection will be the different thermal time constants of various components.

Extreme rigidity of construction, using parts with very high thermal stability, and excellent thermal insulation to slow down any rapid temperature changes.

It CAN be done, but its a very difficult time consuming exercise for the extreme masochist.

Better to just use a low cost quartz crystal and a digital frequency divider...

Long ago, I have seen a Caesium clock made with tubes and with "nixie" type display. Certainly qualifies for a discrete component machine- only problem that it occupied almost a cupboard.

I am still waiting to know what is meant and understood by high stability.

- - - Updated - - -

betwixt said:

It is a method of locking a higher frequency oscillator to a low frequency reference without using a PLL. It uses the LF reference to open a gate, allowing the HF to reach a counter. Depending on whether it overflows or not, you raise or lower an analog voltage to 'pull' the HF oscillator one way or the other. Essentially it produces a modulo 30 count of the HF and through a long time constant network, locks it to a multiple of 30Hz.

Click to expand...

In NMR spectrometers, we often need 500 MHz frequency that is stable to better than 0.01Hz (that is the typical resolution). We use the signal from the NMR itself (usually deuterium) to act as a lock. It appears that it is practically impossible to maintain high stability without some kind of external reference. The system has no way of correcting itself against drift, for example.

I suspect this is part of a 'huff n puff' stabilizer, perhaps Neazoi can confirm.

Click to expand...

neazoi said:

That is absolutely correct.
I need to investigate if the stability of the multivibrator circuits is enough for producing this audio reference.
Two crystal oscillators mixed down would be too difficult to keep unlocked when they operate 30hz apart.

Click to expand...

You are putting the cart before the horse here....
Trying to build a stable 30 Hz discrete component oscillator to stabilize a much higher frequency VFO is a complete waste of time.

Just build a stable VFO, it will not be any more difficult.

The very best basis for a home brew VFO is a Collins Permeability Tuned Oscillator (PTO).
These were built by Collins Radio in the 1950's when valves were the go, and well before the age of integrated circuits and digital synthesis.

Collins Radio were state of the art at that time, no expense spared, and the military and Government were about the only customers that could afford to buy their equipment.

They built a range of VFO s that were permeability tuned with a lead screw, and typically coverd 1 Mhz range with EXACTLY ten turns of the tuning knob per Mhz.
The original specification was +/- 300 Hz accuracy anywhere in the range to the mechanical dial setting. And most can still achieve that today after sixty years with a bit of tender loving care.

The military never repaired these VFOs they just fitted new ones, so millions of extra replacement VFOs were built and are now pretty readily obtainable on the surplus market very cheaply.

The trick with these is to junk the oscillator valve and buffer amplifier valve, and fit a home brew Jfet oscillator (no extra heat). These things are rock stable and will run for hours with only a few tens of Hz drift at several Mhz. The tuning components use the very best high grade military spec technology that you simply cannot reproduce. Some are even oven stabilised.

Various frequency ranges are available, and if you a looking for a nice large tuning knob calibrated with 1Khz divisions, a large combination safe dial is ideal.
Just google "Collins PTO" there is plenty on the web about these wonders.
And they come up on e-bay all the time for usually less than 50 Dollars.

https://www.ebay.com/sch/i.html?_fr...0.Xcollins+pto.TRS0&_nkw=collins+pto&_sacat=0

There are many different types, and various frequency ranges available. Most are specified for 1Mhz with ten turns, but will cover 1.2mHz with twelve turns of the lead screw.

All with just two Jfets and excellent phase noise.

Warpspeed said:

You are putting the cart before the horse here....
Trying to build a stable 30 Hz discrete component oscillator to stabilize a much higher frequency VFO is a complete waste of time.

Just build a stable VFO, it will not be any more difficult.

The very best basis for a home brew VFO is a Collins Permeability Tuned Oscillator (PTO).
These were built by Collins Radio in the 1950's when valves were the go, and well before the age of integrated circuits and digital synthesis.

Click to expand...

Thanks for the info Tony.
I have measured a very low distortion on the oscillator in post #1 at all hf bands at an output of 7dbm. All harmonics are kept below 40dbc using inductors and at about 50dbc using crystals. The oscillator needs only a resonator change to cover the whole hf, no other component change is required. The oscillator, prior to the buffer, is based on a very low phase noise bjt design, but the jfet version adds the broadband characteristic.
I am not an oscillator expert, but I think it is hard to beat this performance in such a low component count.
Not only that, but I have successfully tested an ALC circuit with it since the dynamic range that is achieved by the source resistor adjustment is very high.
For all these reasons, I tend to stick with this design. Crystal operation is excellent. It is just the coil (vfo) version that seems to depend on the coil characteristics for drift. Some coils are almost crystal stable, some others are bad and this is due to thermal drift.

I might end up in making a simple oven for it, it might be much easier than locking it to a low frequency oscillator and it will give me more fine tuning than 30Hz steps.

I just wanted to investigate if it is possible to make a simple puff n huff in discrete components.

Distortion will not be your problem, frequency stability will be.

Warpspeed said:

Distortion will not be your problem, frequency stability will be.

Click to expand...

Oh no, distortion is much more of a problem, especially in broadband designs where no output LPFs are used (the case here).
And whereas frequency drift can be cured by several techniques (puff-n-huff, pll, oven etc), distortion is generated by the oscillator (or buffers) and it is not always obvious how you can minimize it without the use of LPF.
My experiments have shown that the spectral content of the oscillator I am trying, is better than anything I have tried before. I have hard biased it's buffer into class-A and driving it with the low signal (~200mVpp) of the oscillator to preserve the levels of distortion.
I do all the measurements on the FFT, I do not have a real SA.

In this thread I investigate is a discrete method of frequency stabilization can be made.
One way is the slow puff-n-huff http://www.hanssummers.com/huffpuff/minimalist/2chip.html which I am trying to implement discrete. It only needs a reference audio oscillator, since the FF can be implemented in discrete components easily.

I wish you luck building an audio oscillator stable enough to lock a 30 Mhz oscillator to within 30 Hz.