Ultra low noise, low distortion,
general purpose LF/MF/HF oscillator

This is my try to build a general purpose, ultra low noise and low distortion oscillator for the LF/MF/HF. I have previously designed low noise oscillators using BJTs, but in order to make them work in different frequencies, their components had to be changed (apart from the resonant circuit), so special multi-switches had to be used. This raises the cost of the system, not to mention that these switches are more rare to find.

Recently I have found a design similar to my low noise oscillator, that uses a FET instead of a BJT. The author claims that it can be used from 1-30MHz without any changes in components, other than the resonant circuit. I have tried to build this circuit first, to evaluate it.

The original design oscillator, with the values shown, starts to oscillate reliably at 500KHz. Using an old type 455KHz crystal, the oscillator starts a few ms after connecting the crystal and using a 400KHz ceramic resonator, the oscillation starts reliably again, but there is some distortion on the signal with this resonator, probably because of the nature of the resonator. But anyway, you get the feeling of the lower frequency limit, which is 400-500KHz. The higher frequency limit, oscillating all times reliably with all kinds of crystals, is about 15-16MHz or so. Above that, some crystals oscillate but some others not. For example I had a 20MHz crystal that oscillated reliably, whereas another 20MHz crystal from another brand/type did not oscillate at all.

To correct the problem, I changed the 250pF to 150pF. Now the oscillator starts reliably up to 24MHz and at the lower end of 500KHz still oscillates reliably. However for 27MHz both gain loop caps need to be changed so something like 47pF and 22pF, but the distortion will be very noticeable at lower frequencies (and higher too) with these caps. Two 150pF for the gain loop capacitors, seem to be a good tradeoff.

I have also found that a 100K from the gate to the drain of the buffer amplifier, distorted the higher end of the sinewave, so I replaced this with a 82K. The coupling variable capacitor must be small because in higher capacities the output sinewave distorts again.

Accidentally, I have found that the circuit oscillates even if replacing the crystal with a single coil, giving a very nice sinewave at different frequencies (using different coils). That's basically a Colpitts configuration, where the series LC (coil and variable capacitor) looks inductive at the resonant frequency and the two larger caps become "taps" into the tank. The inductor and capacitor to ground act like a low-pass filter, helping to reduce the harmonics. Although the coil does not have a very high Q (like the crystal), it still gives low distortion oscillation. However, I have not made any phase noise measurements with the coil in place.

100nF 100uF 2n4401 12-15v
100nF LC 100 100uF 100nF
XTAL 100nF
82k 1-10pF
100k 100pF 47pF 220k 150pF
150pF 1k
100nF 180 LM

The design of my oscillator is shown above. The results are very satisfying and the oscillator can be used with any configuration of resonant circuit, without a noticeable visual sinewave distortion on the scope. I have tested it with many quartz crystals ranging from 450KHz to 24MHz and the oscillation was always reliable, although output levels varied. However, my 27MHz crystals could not make it oscillate. A quick solution, is to change the gain loop capacitors as explained above, but I believe that this is more depended on the specific crystal characteristics rather than the gain loop capacitors.

I have also tested the oscillator with parallel LC resonant circuits. This allows a VFO to be made, but take care of the setting of the parallel variable capacitor, as in extreme values it can distort the otherwise clean sinewave. Also, note that the frequency is changed with the parallel variable capacitor, but with the series coupling capacitor too. When using crystals, this is not of much importance, but for LC this is. The series capacitor also changes the output level.

As mentioned previously, another test I did was to connect just a single coil as a resonator. Initially, I tried different molded chokes and they all gave reliable low visual distortion sinewave oscillation from 300KHz to 24MHz. Then I tried T50-2, T37-2, T50-6 and T37-6 cores with different number of turns. Again the oscillation was reliable and of low visual sinewave distortion, with the higher end near 45MHz, without changing any other components in the circuit.

Another interesting configuration I have tried, was a single coil (cored coil, not air) like previously, but with the addition of a small axially magnetized neodymium rod magnet, mechanically approaching close to it (LM), causing the magnetic permeability of it's core to change, which leads to frequency changing. In this case, the sinewave was again very undistorted at all frequencies (better than the parallel LC configuration). For small tuning range, use toroidal cores for the coil, but for wide tuning range, you have to use molded chokes (coaxial) and approach the magnet near the ends of the coil core. I was able to cover the entire 300KHz to 24MHz using only five coils with this configuration. However the output level varied with frequency. This method can be useful when cost and size of the oscillator has to be kept minimum, as the use of a variable capacitor is mostly an expensive and bulky solution, especially when bigger capacities are required.

There is nothing to stop you from using an electromagnet to control the oscillator frequency, instead of a permanent magnet. This way, leads to a low distortion voltage controlled oscillator that can be electronically controlled and stabilized further.

Do be continued...

Back to main site