Low-distortion Audio-range Oscillator

[Terry McGee]

Whew, at least I'm not alone in not knowing why 2kHz!

Zigzagflux, your struggle makes this oscillator look attractive:

eBay Australia: Buy new & used fashion, electronics & home d?r

"Frequency of modulation is only a Hz or two. It is a little inconsistent; it holds for a bit, then bounces around and settles, then starts up again. Checked the power supplies; rock steady."

This seems to suggest it's not a straightforward loop instability issue - you would expect that to be cyclical. I'd also check the stability of the reference voltage.

Might be worth checking any connections or components you have any reason to have doubts about.

The approach has a lot in common to one of the references at the end of the article:

http://cds.linear.com/docs/Article/u...0Ad%20Free.pdf

(Also 2kHz, hmmmmm....) Perhaps worth contemplating the differences?

Terry

[davada]

Originally Posted by Terry McGee

Whew, at least I'm not alone in not knowing why 2kHz!

Zigzagflux, your struggle makes this oscillator look attractive:

eBay Australia: Buy new & used fashion, electronics & home d?r

Terry

Has anyone actually bought one of these and tested it out?

There a lot of talk about them but I haven't heard of any reports.

David.

[1audio]

I just bought a set of three. I'll be checking them out when they get here in a week or so. The builder has tested them pretty thoroughly and I have confidence in them but I will test and report.

[Samuel Groner]

Annoying is the fact the amplitude modulation is still occurring, at +/- 12 mV. Maybe this is the best I can expect, not sure.

To check the stability of the level control loop I suggest to modulate the reference voltage with a low-frequency (perhaps 10 Hz, or even 1 Hz) square wave. The response observed at the oscillator output or at the integrator output should be reasonably well-damped. If not, try setting the integrator zero at lower frequency.

Can you post the exact schematic you're using?

Samuel

[KSTR]

Hi,

I tried simulating the circuit and it seems it is really very picky in the AGC behaviour. The LDR coupler was modelled with a controlled resistor, running along an 1/Iset curve around 80Ohms @ 10mA which matched the published number and graphs quite well (but not the response time). I used a Zener to clamp the top integrator output because of ideal opamps used. Startup is triggered by injecting one single period of the osc. freq. at the beginning. Find LTspice simulation file attached (rename to .asc).
The noted asymetric behaviour of the rectifier is confirmed in the sim, too (depends on the diode load resistor and the comp cap).


I finally managed to get to a stable operating point, playing with the series and parallel resistor around the LDR.

Injecting a step at the AGC reference, as per Samuels suggestion, shows a quite well damped response. See screenshot.

But whenever missing the correct parameter range even by a slight amount the osc either died or showed some sort of "gain hiccuping" with 1s order of magnitude period times, just what ZigZag is seeing.

I would think the circuit is very sensitive and might easily be working "just on the edge" where small amounts of noise and other disturbances or drifts could upset stability.

I'd like to see the outcome of this and even try it myself, the game is : will it beat my trusty Tek SG505 (which I think is a masterpiece of an adjustable osc)?

[PChi]

The PCB layout looks OK.

It might be possible to clean up the full wave rectifier output a little.
I have built some full wave rectifiers myself a long time ago and found that 'Applications of Operational Amplifiers' by Jerald G. Graeme helpful. I think that the alternating DC values are due to the offset voltage of the first operational amplifier stage TL082D. The switching around 0 V is also affected by the slew rate and high frequency small signal gain of the TL082. So a lower offset, higher bandwidth, greater slew rate operaational amplifier should help. It might also be worth using schottky diodes for their lower forward voltage but with the disadvantage of higher capacitance and leakage.

I agree with KSTR the circuit amplitude will be working on an edge. It's got two frequency roll off stages in series in the control loop with the integrator and the filter on the transistor base along with the delay in amplitude change in the Wien bridge.

[DDB]

Whew, at least I'm not alone in not knowing why 2kHz!

Zigzagflux, your struggle makes this oscillator look attractive:

eBay Australia: Buy new & used fashion, electronics & home d?r

After studying the ebay pcb pix, I thought it appeared similar to a JanasCard inverting Wien bridge but with a jfet AGC. The seller said it was similar but used his own AGC jfet design with 1/2 D-S fed to G for second harmonic reduction.

The seller also indicated he was comtemplating posting the schematic here or on his ebay ad.

A couple of these with a good summer and output circuit could make a nice IM generator.

[zigzagflux]

Originally Posted by richiem

Just seems like the AGC control range is too small. I'd try making the 18R resistor larger, say 27R. I think I'd make the 10uF at the base of the transistor larger too.

Good call. Increased R18 from 15 to 22, to 47, and finally 63. Improvement each time with gain control. Now getting about +/- 0.5mV of wobble, and its pretty consistent and doesn't get out of control. Also notice that stability is best when the current through the LED is low, around 4-5 mA. May not be best for linearity, as I can see my 3rd harmonic is a little higher, but we are stable. Starts up and settles very quickly, too.

The cap at transistor base was about the same when increased by 10uF, but got unstable with an extra 22uF.

At this point I have determined that my simple D/A through transformer works better. If (and that's a big if) the author really did get distortion measurements that low, the design is way too marginal and sensitive to luck that it's not worth pursuing anymore.

I will keep watch over the test results from the ebay oscillator. It was worth a try. Thanks for the help.

[richiem]

@zigzag -- You're not seeing the real performance of the oscillator, just the residual of your ADC. I suggest you build a pre-filter, hopefully an active twin-tee notch filter to reduce the fundamental level while leaving the harmonics unattenuated. This will reveal the harmonics accurately. For one way, see my page Active Twin-T notch filter.

[zigzagflux]

Yes, I have thought about that; never got a true measurement of the accuracy of my soundcard, but I have obtained better results with the DCX; consistent, too. I'll think about building a notch filter in the next few weeks.