Low-distortion Audio-range Oscillator

[richiem]

@Samuel Groner -- I think that the lamp R was too small in any case to see the kind of issue that happened with the old HP200CD et al, which had lamps with *much* higher operating point resistances. Even so, I'm now not sure that any voltage coefficient effects could possibly be large enough, even at 10 or 20VRMS output to actually make a difference in stability. But it's an open question for now.

[KSTR]

I've built several lamp-based "thermal" compressors for musical instrument use, there I noticed the mentioned slope of distortion, too. In fact I was exactly after that as a welcomed side-effect, the smaller the lamp's thermal inertia the higher the distortion (I used the smallest lamps comercially availabe, 1.2V/10mA, as well as big ones with several watts). I biased the lamps with DC right at their resistance knee, just below a visible red glow, this gave raise to strong second harmonic at very low frequencies and low levels which is what I wanted.

I also noted that a hot lamp must be well shielded from vibration (notably with DC bias it literally is a microphone) and I've seen this done in commercial lamp-stabilized oscillators by Grundig.

[davada]

Originally Posted by KSTR

I've built several lamp-based "thermal" compressors for musical instrument use, there I noticed the mentioned slope of distortion, too. In fact I was exactly after that as a welcomed side-effect, the smaller the lamp's thermal inertia the higher the distortion (I used the smallest lamps comercially availabe, 1.2V/10mA, as well as big ones with several watts). I biased the lamps with DC right at their resistance knee, just below a visible red glow, this gave raise to strong second harmonic at very low frequencies and low levels which is what I wanted.

I also noted that a hot lamp must be well shielded from vibration (notably with DC bias it literally is a microphone) and I've seen this done in commercial lamp-stabilized oscillators by Grundig.

Hi KSTR,

Was this an optical coupling of the lamp or is the lamp in the amplifier circuit.
How did you isolated the DC bias from the amplifier? I would expect a strong dc offset in the amplifier from the bias.

[KSTR]

The lamp was the actual passive "gain-cell", lamp as series resistor, another R to GND to form a divider. The bias current was injected with a floating current source accross the lamp, and the whole thing was AC-coupled at in and out driven by small pwr stages, also there were two of these cells in series to handle the dynamic range.

I'm aware this introduced additional effects from the biasing alone working back on the capacitors. An alternative would have been to use HF bias but that seemed unneccesary for the application (bass guitar amplifier) because a little "thumping" was not a problem there.

Using no bias didn't sound as good because the time constants and dynamic response was not optimum.
EDIT: and no 2nd harmonics, of course.

[KSTR]

Originally Posted by coluke

Yeah - I was wondering about Victor's phase noise and THD+N. A full spectrum comparison betweem AP's and Victor's would be interesting.

I might do this measurement. With full spectrum you mean DC to xxkHz, linear view? With or without fundamental notch?

I might also scope the gate drive to see how much it fluctuates.

[davada]

Originally Posted by KSTR

The lamp was the actual passive "gain-cell", lamp as series resistor, another R to GND to form a divider. The bias current was injected with a floating current source accross the lamp, and the whole thing was AC-coupled at in and out driven by small pwr stages, also there were two of these cells in series to handle the dynamic range.

I'm aware this introduced additional effects from the biasing alone working back on the capacitors. An alternative would have been to use HF bias but that seemed unneccesary for the application (bass guitar amplifier) because a little "thumping" was not a problem there.

Using no bias didn't sound as good because the time constants and dynamic response was not optimum.
EDIT: and no 2nd harmonics, of course.

So the idea was to add some 2nd H to the base. I presume this fattened up the base sound.

[Samuel Groner]

Is the AP oscillator stabilized (disciplined)? It shows less close in phase noise than Victors.

No. The sidebands are also affected from the amplitude leveling loop and the plain noise floor of the filter and multiplier (close to the fundamental frequency, noise gain get's very high).

Our AP is about two years old, is that late enough?

No idea, I don't have any details.

I might do this measurement. With full spectrum you mean DC to xxkHz, linear view? With or without fundamental notch?

To evaluate sideband performance I'm using these settings:

* Direct FFT (without notch--it would cut most of the sideband).
* Linear frequency scale, range about +-10% (or even just +-5%) of the fundamental (e.g. 900-1100 Hz).
* Amplitude range set such that most of the fundamental is chopped off, and the sidebands good visible.
* A FFT window with very good side lobe behaviour is important; IIRC the "equiripple" design is the best for the SYS-2722.
* Use the highest possible FFT resolution (32k) and lowest possible sampling rate to get the most narrow main lobe.
* Experiment with different average numbers--I've seen that for some oscillators the sideband noise drops with higher averages. I'm not sure what the exact mechanism is, but there is surely more going on than just plain "phase noise" as we know it from HF oscillators (e.g. there's also amplitude noise).

Samuel

[scott wurcer]

Originally Posted by Samuel Groner

I'm not sure what the exact mechanism is, but there is surely more going on than just plain "phase noise" as we know it from HF oscillators (e.g. there's also amplitude noise).

Samuel

Another observation on my impractical thermally stabilized oscillator experiment is that when it was stable the side bands virtually dissapeared.

[Samuel Groner]

Yes, I've observed too that low phase margin in the leveling loop leads to pronounced sidebands (also funny stuff like peaking a few Hz from the fundamental). On the other hand, I've found that loop compensation for optimum settling time is, most unfortunately, not the one for best sideband performance.

Samuel

[coluke]

Originally Posted by Samuel Groner

Yes, I've observed too that low phase margin in the leveling loop leads to pronounced sidebands (also funny stuff like peaking a few Hz from the fundamental).

A kind of very low level (and very low freq) amplitude modulation? I've seen it too, and spent quite a few hours before figuring out what the hell was going on

L.