Signal generator instability cause?

There appears to be very little to stabilise the output of this osc, similar osc designs have a lamp or other variable R device ( or sloped knee zener ) to provide a feedback effect to keep the output at a definite magnitude - this is not very well done for this ckt ...

Easy peasy said:

There appears to be very little to stabilise the output of this osc, similar osc designs have a lamp or other variable R device ( or sloped knee zener ) to provide a feedback effect to keep the output at a definite magnitude - this is not very well done for this ckt ...

Click to expand...

I guess you refer to the signal level and not the frequency drift. In that case, as I see it, the bottom bjt in the differential amplifier reduces the gain of the differential amplifier so that it is not saturated, thus giving a clean sinewave without the need for a LPF. He says at the article that the driving of the L with square waves ensures low distortion too. I have measured the harmonics to be of the order of -40dBc at all shortwave frequencies, which is not bad at all for such a simple circuit.
However, it is not very clear to me why he gets the feedback directly out of the oscillator bjt pair (through the jfet) and not from the output of the mosfet. This would ensure even more flat amplitude response. Maybe he does that in order to isolate the feedback loop from the output of the generator (SWR etc).
If I build an external amplifier for it, I am thinking of using one of the two mosfet gates as a variable attenuator, driven by a feedback loop taken directly from the output of this external amplifier, this would ensure stable amplitude level and it will be independent of the oscillator feedback. I am not sure if this is a good idea though.

- - - Updated - - -

betwixt said:

It is quite logical, if you show a signal on trace A and another signal on trace B then select ADD mode, it shows the waveform A+B.
If you do the same with one channel inverted (lets say channel B), you get A+(-B) which is A-B.
Think about what you measure normally with a ground and tip probe, its the voltage difference between them. If instead you use two (ideally x10) probes and only connect the tips, it will also show the voltage difference between them but now your scope is isolated through the probe divider networks.

There is a good explanation here: https://www.youtube.com/watch?v=BHY4o7Iknes

It uses Tek scopes but yours will I'm sure have the same functions.

Brian.

Click to expand...

Nice video I like it, thanks!
Now I need to buy another probe.....
So I actually need to use the DIFF function of my scope (which is like combining the ADD and INVERT) which measures the difference between the two probes right?

So I actually need to use the DIFF function of my scope (which is like combining the ADD and INVERT) which measures the difference between the two probes right?

Click to expand...

It is more commonly called "add & invert" which are two functions you combine but maybe your scope calls it something different. The only way to be sure is to check in its manual.

Brian.

betwixt said:

Increasing the transformer voltage or using a low drop-out voltage regulator will give better results than both those circuits.

Click to expand...

I will increase the transformer voltage then. But at the same time, because I like to build it with discrete components, can I use the zener stabilized capacitance multiplier PSU instead of the linear regulator IC? I mean, will it degrade the performance achieved by the increase in transformer voltage if I use this instead of the regulator IC?

- - - Updated - - -

Also, what if I use a center tapped transformer in one of these configurations (I like the second one for better mechanical stability, is it ok?)
Will it be enough to put a capacitor parallel to each diode?

The output of the two schematics will be identical and yes, a small capacitor across the diodes will help. The interference they create is very unpredictable and may vary even from one diode to the next in the same batch.

Your can get approx 0.3V DC more if you use Schottky rectifiers because of their lower Vf.

An IC LDO will still work better but you can use the capacitor 'multiplier' circuit. It is only an emitter follower current amplifier with the input held 'clean' by the capacitor. You don't need the 1N4007 in the base circuit or the 47K in the emitter if it will be permanently connected to the oscillator, it is only there to drain any leakage in the transistor but the oscillator will do that just as well.

Brian.

betwixt said:

The output of the two schematics will be identical and yes, a small capacitor across the diodes will help. The interference they create is very unpredictable and may vary even from one diode to the next in the same batch.

Your can get approx 0.3V DC more if you use Schottky rectifiers because of their lower Vf.

An IC LDO will still work better but you can use the capacitor 'multiplier' circuit. It is only an emitter follower current amplifier with the input held 'clean' by the capacitor. You don't need the 1N4007 in the base circuit or the 47K in the emitter if it will be permanently connected to the oscillator, it is only there to drain any leakage in the transistor but the oscillator will do that just as well.

Brian.

Click to expand...

I have also another issue with the oscillator, because of the feedback loop which controls the gain of the oscillator, I would expect the output level of the carrier to be dead flat without variations from frequency to frequency. It is adequately flat but there is something like 100mVpp from the start of the coverage to the end, at each inductor setting.

I was wondering why is this happening?
Maybe the silicon diodes in the feedback loop could be replaced with germanium ones so that they can conduct earlier and possibly make the output more flat (but also the signal level lower)?

I could try a second feedback loop from the output to one of the gates of the mosfet, but I would like to investigate why it cannot be done right with the existing feedback loop.

Amplitude variation over frequency is normal operation, I think. The ALC loop has little gain, respectively the input voltage will rise when the oscillator current has to be increased to compensate for Q drop.

FvM said:

Amplitude variation over frequency is normal operation, I think. The ALC loop has little gain, respectively the input voltage will rise when the oscillator current has to be increased to compensate for Q drop.

Click to expand...

So increasing the gain of the loop can be the solution? Maybe an additional amplifier in the loop or just replacing the loop diodes with germanium types?

An additional control amplifier, with proper frequency compensation to keep the loop stable.

FvM said:

An additional control amplifier, with proper frequency compensation to keep the loop stable.

Click to expand...

If it was related to the frequency wouldn't one expect different level at higher bands than that of the lower? Now as it behaves, at each band (coil setting) at it's lower frequency the level is a bit lower and at it's higher it is a bit higher

The question is how oscillator Q varies with L, C and ESR of the different coils. I think, you should be satisfied if the circuit is oscillating stable and amplitude variations are not too high.

betwixt said:

The output of the two schematics will be identical and yes, a small capacitor across the diodes will help. The interference they create is very unpredictable and may vary even from one diode to the next in the same batch.

Your can get approx 0.3V DC more if you use Schottky rectifiers because of their lower Vf.

An IC LDO will still work better but you can use the capacitor 'multiplier' circuit. It is only an emitter follower current amplifier with the input held 'clean' by the capacitor. You don't need the 1N4007 in the base circuit or the 47K in the emitter if it will be permanently connected to the oscillator, it is only there to drain any leakage in the transistor but the oscillator will do that just as well.

Brian.

Click to expand...

I used this PSU because I already had this transformer. I have added a 7812 regulator at the output plus another small (68uF) shunt capacitor after it.
Without the parallel capacitors to the diodes low frequency oscillation (noise, no clean tone in nearby SSB receiver) is high. With the parallel capacitors this noise goes down to 60-70%, impressive!

I am trying to remove these last pieces of noise, so the tone becomes clean as when powered with a battery. I have tried Pi filters after the regulator with 3.3mH or 10mH inductors but these did not help.
What else can I do? Maybe adding such a circuit https://www.qsl.net/yo5ofh/projects/voltage_regulator_noise/finesse voltage regulator noise.htm (the first)?
Or I can add some resistors aside of the capacitors in the rectifiers? I do not know why are these used but I have seen them in old magazines, I do not remember if these were parallel or in series to the diodes or capacitors.

The resistors in some old power supplies were to distribute the voltage more equally when several low PIV diodes were joined in series. If you add them to the present supply it will make matter worse by increasing the 'backward' current from the capacitor to the opposite polarity from the transformer.

You can try the noise reduction shunt but I wouldn't guarantee it will make any difference. All it does is sink a little current through the shunt to drop the peaks of voltage but of course ideally you wouldn't produce peaks in the first place. A more pragmatic solution would be to wire the power source so it's output impedance is a low as possible. Basically, you make the lowest impedance point the ground and wire all the current sources and loads to it. The best place is almost certainly the negative side of the 4700uF capacitor so you would need to move all the connections (in the PSU) so they physically meet at that point. It's called a 'star' ground because all the wires radiate from a central place. Make sure the output negative wire also goes to the same point.

Brian.

betwixt said:

The resistors in some old power supplies were to distribute the voltage more equally when several low PIV diodes were joined in series. If you add them to the present supply it will make matter worse by increasing the 'backward' current from the capacitor to the opposite polarity from the transformer.

You can try the noise reduction shunt but I wouldn't guarantee it will make any difference. All it does is sink a little current through the shunt to drop the peaks of voltage but of course ideally you wouldn't produce peaks in the first place. A more pragmatic solution would be to wire the power source so it's output impedance is a low as possible. Basically, you make the lowest impedance point the ground and wire all the current sources and loads to it. The best place is almost certainly the negative side of the 4700uF capacitor so you would need to move all the connections (in the PSU) so they physically meet at that point. It's called a 'star' ground because all the wires radiate from a central place. Make sure the output negative wire also goes to the same point.

Brian.

Click to expand...

Star ground right, even if the components are spaced 5cm or so apart in the current setup?
Can the problem be the long wires I use (about 1meter) which connect the PSU to the oscillator?
Remember the oscillator inductor is directly connected to the vcc.

Remember that all the current flowing into the capacitor is pulsed, the diodes only conduct when the transformer secondary voltage is higher than the capacitor voltage plus Vf of the diodes, that means it is likely to be anything but a clean sine wave. All the current causes voltage drops in the wiring, even if it only has a few milli-Ohms of resistance so the PSU is 'alive' with signals and voltages everywhere. It follows that where you think there is a clean DC voltage, it probably has several mV of strange waveforms on it. Keeping the ground tied together helps to minimize the voltage drop between them. In particular, consider that a voltage regulator tries to maintain a constant voltage between its output and ground pins so if the ground has signal on it, so will it's output.

The length of wires to the oscillator isn't too important but do remember to add some extra filtering at the oscillator end, at least one large (>10uF) and one ceramic (100nF) in parallel and close to where the power wires arrive.

Brian.

betwixt said:

Remember that all the current flowing into the capacitor is pulsed, the diodes only conduct when the transformer secondary voltage is higher than the capacitor voltage plus Vf of the diodes, that means it is likely to be anything but a clean sine wave. All the current causes voltage drops in the wiring, even if it only has a few milli-Ohms of resistance so the PSU is 'alive' with signals and voltages everywhere. It follows that where you think there is a clean DC voltage, it probably has several mV of strange waveforms on it. Keeping the ground tied together helps to minimize the voltage drop between them. In particular, consider that a voltage regulator tries to maintain a constant voltage between its output and ground pins so if the ground has signal on it, so will it's output.

The length of wires to the oscillator isn't too important but do remember to add some extra filtering at the oscillator end, at least one large (>10uF) and one ceramic (100nF) in parallel and close to where the power wires arrive.

Brian.

Click to expand...

I found what the problem was. You won't believe it. It was a digital clock I had near by. It was causing interference. When I switched this off, the tone was clean even with no capacitors parallel to the rectifier diodes. I also placed the inductor a bit further away from metals.

neazoi said:

I found what the problem was. You won't believe it. It was a digital clock I had near by. It was causing interference. When I switched this off, the tone was clean even with no capacitors parallel to the rectifier diodes. I also placed the inductor a bit further away from metals.

Click to expand...

Maybe the cause for the level instability is the setting of the variable capacitor? As I see it at higher capacitances more signal is shunted to the ground. That does not explain of course why the ALC cannot compensate for the loss.

I have updated the page for you to see the whole circuit I built **broken link removed**