Simple Sinusoidal Oscillators

[mawyatt]

We needed a low level simple floating (battery powered) sine wave oscillator with ~16KHz for testing. Grabbed a couple 2N3904 NPNs, a 470uH inductor, a 0.22uF cap and a 510 ohm resistor, and slapped together a Peltz Oscillator.

https://wiki.analog.com/university/courses/electronics/comms-lab-peltz-osc

Here's the output when powered by a AAA cell, and actually works down to 0.7VDC!!

This about as simple a low frequency sine wave oscillator as we've encountered.

Others have any simple examples of low frequency sine wave oscillators?

Best,

[Benta]

Cool!   

[mawyatt]

Even with a 0.8VDC supply the output doesn't look too bad!!

Best,

[fourfathom]

Here's a simple Twin-T oscillator, with a buffer.  In simulation it just barely starts up with a 1.2V supply.  You might be able to improve that by playing with the resistor values -- I just threw this together.

[Conrad Hoffman]

The Peltz isn't the lowest distortion thing I've ever seen but the performance vs. simplicity is fantastic!

[Benta]

The Peltz isn't the lowest distortion thing I've ever seen but the performance vs. simplicity is fantastic!

Amen to that. I'm impressed.

[mawyatt]

Add a single series resistor to the emitter junctions of Q1 and Q2 and you have a nice injection point for an Injection Locked Oscillator, works beautifully!! Also works inserting the resistor at the collector of Q1, and this is a nice zero volt (~ground) injection point!!

Best,

[T3sl4co1l]

Interestingly, the amplitude should have considerable PSRR: it's constrained within one Vbe peak.  Transconductance however will not, so it should become significantly flattened (with resulting FM) when driven with I_E >> Vbe/Ro (Ro being the LC equivalent parallel resistance).

A proper CCS of course addresses that.  As for intentional FM, you'll get lower distortion (and, maybe phase noise too?) holding it at optimal current, and using a varactor instead.

I once built an, er... broadly similar circuit, as an RC oscillator.  Similar, in that it's based on CML/ECL, but significantly more complex, as a.. I forget if I did it as a hysteresis comparator, or a 555, but it went at like 100MHz.

Hmm.. this might've been the circuit.  I can't remember the filename, thought I took a photo of the real thing...



which is probably all kinds of poorly matched, and I don't know Q12 is doing, something vestigial perhaps..

...Oh here it is,



never had a screenshot of it though.  Just dug it out; here's Q7 emitter (Q6 base = 2.78V, just fixed R's in the sim but obviously CV input; VCC = 4.82V):



Why only 52MHz?  It falls off pretty fast at higher bias. Hm, decreasing R4 seems to help.  Probably needs more current then, but then the collector resistors need to drop to keep bias in place... I'm not going to screw with it right now, but eh, it's a thing, I guess.

Tim

[mawyatt]

A proper CCS of course addresses that.  As for intentional FM, you'll get lower distortion (and, maybe phase noise too?) holding it at optimal current, and using a varactor instead.

Yes a CCS is better but adds complexity, the simple resistor works quite well indeed!!

The earlier post about the added injection resistor is not about FM, but for Injection Locking the oscillator to an outside Injection signal. Where the oscillator assumes the frequency of the Injection signal over a small frequency band centered around the oscillator free running frequency. This technique can even Lock around harmonic frequencies, for example 2X or 3X. We've used this Injection Locking technique in the past for many applications, including locking to a transmitted carrier (see patent 5603111, Synchronous Tracking AM Receiver), and resonate frequency dividers operating beyond 100GHz.

Edit: Injection locking is a fascinating subject, dated back to the discovery by Van der Pol when experimenting with Neon bulb relaxation oscillators. He noted that two Neon oscillators became locked (same flashing rate) when they had a common connection thru the power supply (slight coupling). Much later Adler in 1946 derived the fundamental equations describing injection locking. What we discovered and patented based upon Van der Pol and Adlers work was that the injection locking concept also produces a demodulation of the injected signal, that can be utilized to perform AM, FM or PM demodulation, and subsequently developed a single (silicon) chip Microwave receiver based upon this technique.


Here's some papers on the Injection Locking subject that are available, many others are behind IEEE fees unfortunately.

https://chic.caltech.edu/wp-content/uploads/2019/07/08753733.pdf

http://rfic.eecs.berkeley.edu/ee242/pdf/Module_7_4_IL.pdf

https://www.seas.ucla.edu/brweb/papers/Conferences/RCICC2003.pdf

Anyway, fun stuff!!

Best,

[moffy]

The earlier post about the added injection resistor is not about FM, but for Injection Locking the oscillator to an outside Injection signal. Where the oscillator assumes the frequency of the Injection signal over a small frequency band centered around the oscillator free running frequency. This technique can even Lock around harmonic frequencies, for example 2X or 3X. We've used this Injection Locking technique in the past for many applications, including locking to a transmitted carrier (see patent 5603111, Synchronous Tracking AM Receiver), and resonate frequency dividers operating beyond 100GHz.

Edit: Injection locking is a fascinating subject, dated back to the discovery by Van der Pol when experimenting with Neon bulb relaxation oscillators. He noted that two Neon oscillators became locked (same flashing rate) when they had a common connection thru the power supply (slight coupling). Much later Adler in 1946 derived the fundamental equations describing injection locking. What we discovered and patented based upon Van der Pol and Adlers work was that the injection locking concept also produces a demodulation of the injected signal, that can be utilized to perform AM, FM or PM demodulation, and subsequently developed a single (silicon) chip Microwave receiver based upon this technique.


Here's some papers on the Injection Locking subject that are available, many others are behind IEEE fees unfortunately.

https://chic.caltech.edu/wp-content/uploads/2019/07/08753733.pdf

http://rfic.eecs.berkeley.edu/ee242/pdf/Module_7_4_IL.pdf

https://www.seas.ucla.edu/brweb/papers/Conferences/RCICC2003.pdf

Anyway, fun stuff!!

Best,

It even happens in Ring Laser Gyroscopes, where there are two counter rotating laser beams that share the same cavity and mirrors. When they created the first RLGs they noticed that for low rotation rates they would get no ouput and then above a certain level they would get the output they expected, it took them  a while to work it out, because they didn't have any radio guys to explain what was happening. It was only when they stumbled across some old articles about injection locking that they realised that the counter rotating lasers were being injection locked by the back scatter from the common mirrors.

[RoGeorge]

...
Edit: Injection locking is a fascinating subject, dated back to the discovery by Van der Pol when experimenting with Neon bulb relaxation oscillators. He noted that two Neon oscillators became locked (same flashing rate) when they had a common connection thru the power supply (slight coupling). Much later Adler in 1946 derived the fundamental equations describing injection locking. What we discovered and patented based upon Van der Pol and Adlers work was that the injection locking concept also produces a demodulation of the injected signal, that can be utilized to perform AM, FM or PM demodulation, and subsequently developed a single (silicon) chip Microwave receiver based upon this technique.


Here's some papers on the Injection Locking subject that are available, many others are behind IEEE fees unfortunately.

https://chic.caltech.edu/wp-content/uploads/2019/07/08753733.pdf
...

Thank you for bringing in the injection locking aspect.  Fascinating subject indeed. 

I think "injection locking" is what leads to entanglement in physics (not necessarily electric injection locking, but a similar energy exchange happens in such a way that makes particles or macro-objects synchronize between them, which will also mean entanglement is caused by a hidden variable - the sync - which right now is considered as disproved in mainstream physics, demonstrated by the Bell theorem, which theorem and implications I disagree with because of Robert H. McEachern papers published on viXra).  Anyway, for now that's offtopic rambling and I'm not a physicist, so I might be totally wrong.

I don't want to derail the thread into offtopic speculations, posting here only to say the link to the part II:
https://chic.caltech.edu/wp-content/uploads/2019/07/08758318-small.pdf

[RoGeorge]

The earlier post about the added injection resistor is not about FM, but for Injection Locking the oscillator to an outside Injection signal.
...

It even happens in Ring Laser Gyroscopes, where there are two counter rotating laser beams that share the same cavity and mirrors. When they created the first RLGs they noticed that for low rotation rates they would get no ouput and then above a certain level they would get the output they expected
... [because] ...
the counter rotating lasers were being injection locked by the back scatter from the common mirrors.

Similar happens in chip-scale optical gyros

...thermal changes can modulate the phase of the
light and introduce a phase shift indistinguishable from that
produced by the Sagnac effect. In addition to thermal effects,
the difficulty of fabricating low-loss light paths in standard
processes at this scale makes lossy waveguides an inevitability.
Not only does loss reduce signal strength, it leads to back-
reflection. If there is only one signal path, as is the case in most
FOGs, this back-reflection couples to the reverse direction of
propagation

Source:  A Chip-Scale Nanophotonic Optical Gyroscope

[mawyatt]

It even happens in Ring Laser Gyroscopes, where there are two counter rotating laser beams that share the same cavity and mirrors. When they created the first RLGs they noticed that for low rotation rates they would get no ouput and then above a certain level they would get the output they expected, it took them  a while to work it out, because they didn't have any radio guys to explain what was happening. It was only when they stumbled across some old articles about injection locking that they realised that the counter rotating lasers were being injection locked by the back scatter from the common mirrors.

Much of my mid-career was with Honeywell, the masters of the RLG (and ESG). Although never worked on the RLG or the ESG, but did so on the Fiber Optic Gyro and granted a couple patents (Serrodyne Phase Modulator, Transimpedance Amplifier 5339055, 5216386).

Honeywell's original solution to injection locking was to dither one of the 3 mirrors forming the triangular path to unlock the HeNe laser self injection locking which created a "dead zone" in the output transfer function. The result was then integrated across the "dead zone" and created a highly linear overall transfer function. Very effective solution and enabled them to outpace the stiff competition from Sperry early on which elected to use an electro-optic solution by effectively dithering the mirror index of refraction and thus modulating the laser path length.

Fond memories from back then.

Best,

[fourfathom]

Honeywell's original solution to injection locking was to dither one of the 3 mirrors forming the triangular path to unlock the HeNe laser self injection locking which created a "dead zone" in the output transfer function. The result was then integrated across the "dead zone" and created a highly linear overall transfer

That's a nice physical analog to dithering in time that we did on a reference-clock tracking loop.  The designer had used an internal clock at the same nominal frequency as the reference clock, creating a one-period dead-band.  Rather than make him redesign the entire FPGA (prototype, to be converted to an ASIC), I had him just use a slightly slower unrelated clock to pre-sample the input reference.  All the sampling artifacts were way outside the filter bandwidth and we got a very nice phase-detector transfer curve.

[mawyatt]

Back to the Peltz oscillator. We changed the inductor to a wire wound around a 3mm diameter of 18 turns, this creates about 0.3uH, and changed the capacitor to 1nF film. Everything just plugged into a Proto-Board and should produce a frequency of ~9.19MHz based upon the LC product. Here's the results with a 1.5VDc supply.

This is such a simple little oscillator and works so well, hope others can tinker with it. You can use just about any small signal transistor (2n3904, 2n3906, 2n2222, 2n2907), any type inductor and capacitor, just make sure the impedances are reasonable for the L and C.

If we ever get access to a SiGe or InP process again, would like to roll one of these with a couple 500GHz transistors

Best,

[Benta]

@mawyatt, I'm slowly getting blown away here. This thread has been stored for future reference.

And perhaps even a "Sticky".

[LM21]

Is there a website for 1-2 transistor oscillators. I have seen plenty of low frequency RC oscillators, but I not in a single place.

[Kleinstein]

The Petz oscillator is well behaved and can work with relatively low Q resonators.

The classic sine oscillator for the audio range is the Wien bridge. This can can also be build with 1 or 2 transistors, though the THD may suffer a little.

[moffy]

That's nice, the higher frequency has about 20db lower distortion on the 3rd harmonic than the audio one. Less gain out of the transistor for the higher harmonics?

P.S. I think I remember the Sperry solution, they added magnetically changeable layers to one of the mirrors. Bad solution because it increased the back scatter problem as well as the stability issues.

[RoGeorge]

By looking at all 3 spectrum measurements, what intrigues me is the fact that in the last one, 2^nd, 3^rd and 4^th harmonics are about the same value, -70dBc.

In the first measurement there are big differences between the level of them, but with the next measurements, when the level of the fundamental is lower, all the harmonics tend to become equal.    Now, as explained in

https://wiki.analog.com/university/courses/electronics/comms-lab-peltz-osc

the amplitude on the LC tank is limited by the forward voltage drop on the BC junction of each transistor.  One transistor for each polarity of the oscillations, which means, if the two transistors have a different V_BC forward drop, then the generated sinusoid will be asymmetric in respect to X axis (in time domain), which means (in frequency domain) even order harmonics.

Also, the two V_BC junctions will limit the amplitude, turning the sinus into a more square"-ish" waveform.  This type of distorting will translate into odd order harmonics.  I don't know what leads to equal odd and even harmonics, maybe it's just a coincidence, though it doesn't look so to me, it looks like the lower the oscillation, the more equal harmonics.

First things I would like to experiment with would be:
- to try well paired transistor (for identical V_BC, thus symmetrical waveform, thus less even harmonics)
- try to put the output signal through an exponential amplifier.  The idea is to try to reverse (in time domain) the amplitude limitation/distortion introduced by the forward biased BC junctions by applying to the output signal the inverse function that produced the distortion.  The hope is to get lower distortions/harmonics.


About the optical gyros (I know nothing about the implementation details, only the principles), I wonder if using total reflection would have lowered the backskattering.  I imagine getting very flat glass surface would be easy by simply letting the melted glass to cool.  Using longer wavelength for the light, so the wavelength will be big in relation to the ruggedness of the glass surface, should help too.  But I guess these were considered already.

Were those gyro mirrors made out of glass, or metal?

[Kleinstein]

First things I would like to experiment with would be:
- to try well paired transistor (for identical V_BC, thus symmetrical waveform, thus lower less even harmonics)
- try to put the output signal through an exponential amplifier.  The idea is to try to reverse (in time domain) the amplitude limitation/distortion introduced by the forward biased BC junctions by applying to the output signal the inverse function that produced the distortion.  The hope is to get lower distortions/odd harmonics.

The nonlinearity is not so easy to remove later, as there is the resonator in between that keeps extra energy for a longer time.
This helps in reducing the hamonics quite a bit: quite often the transistors to all the way to switch on / off and thus a kind of rectangular excitation. The more linear range is much smaller - up to around a 25 mV in amplitude as the about linear range of a long tailed pair.

For the symmertry it is not only the VBE drop, but also the ohmic resistance of the inductor that produces a shift. So matching alone is no suffient. It would be more a point to trim the offset.

A way to reduce the harmonics would be by adding emitter resistors (maybe diodes) and adding a kind of amplitude stabilization loop, that reduces the current from the emitter side. So drive the resonantor only as hard as really needed and not too much into the limiting action of the BE junctions.

[moffy]

About the optical gyros (I know nothing about the implementation details, only the principles), I wonder if using total reflection would have lowered the backskattering.  I imagine getting very flat glass surface would be easy by simply letting the melted glass to cool.  Using longer wavelength for the light, so the wavelength will be big in relation to the ruggedness of the glass surface, should help too.  But I guess these were considered already.

Were those gyro mirrors made out of glass, or metal?

Getting them flat was only part of the problem with the mirrors. For the triangular laser gyro, the most common, two of the mirrors were flat and one was curved, the curved mirror was adjusted by hand at manufacture to maximise the lasers Q. They were a polished ceramic/glass called zerodure, a zero effective expansion material, below 1ppm/degC. They were a dielectric mirror with 14 incredibly thin dielectric layers deposited, the light was bent rather than reflected. It was believed that the curved mirror was the major contributor to backscatter, but the levels were incredibly small, it was atomic levels of uneveness.

[mawyatt]

By looking at all 3 spectrum measurements, what intrigues me is the fact that in the last one, 2^nd, 3^rd and 4^th harmonics are about the same value, -70dBc.

In the first measurement there are big differences between the level of them, but with the next measurements, when the level of the fundamental is lower, all the harmonics tend to become equal.    Now, as explained in

https://wiki.analog.com/university/courses/electronics/comms-lab-peltz-osc

the amplitude on the LC tank is limited by the forward voltage drop on the BC junction of each transistor.  One transistor for each polarity of the oscillations, which means, if the two transistors have a different V_BC forward drop, then the generated sinusoid will be asymmetric in respect to X axis (in time domain), which means (in frequency domain) even order harmonics.

Also, the two V_BC junctions will limit the amplitude, turning the sinus into a more square"-ish" waveform.  This type of distorting will translate into even order harmonics.  I don't know what leads to equal odd and even harmonics, maybe it's just a coincidence, though it doesn't look so to me, it looks like the lower the oscillation, the more equal harmonics.

First things I would like to experiment with would be:
- to try well paired transistor (for identical V_BC, thus symmetrical waveform, thus less even harmonics)
- try to put the output signal through an exponential amplifier.  The idea is to try to reverse (in time domain) the amplitude limitation/distortion introduced by the forward biased BC junctions by applying to the output signal the inverse function that produced the distortion.  The hope is to get lower distortions/odd harmonics.

Estimating/Explaining the harmonic content is difficult in the configuration we used since this was just assembled on a plug-in Proto-Board with long leads, so lots of distributed capacitance and inductance. Think of Q1 as a common base configuration and Q2 as a feedback emitter follower to Q1's emitter. This creates a + feedback loop, the DC bias current is set by the common emitters resistor and split between the transistors if they are matched.

Interested in what you find with your experiments, fun little circuit to play around with!!

Best

[mawyatt]

About the optical gyros (I know nothing about the implementation details, only the principles), I wonder if using total reflection would have lowered the backskattering.  I imagine getting very flat glass surface would be easy by simply letting the melted glass to cool.  Using longer wavelength for the light, so the wavelength will be big in relation to the ruggedness of the glass surface, should help too.  But I guess these were considered already.

Were those gyro mirrors made out of glass, or metal?

Getting them flat was only part of the problem with the mirrors. For the triangular laser gyro, the most common, two of the mirrors were flat and one was curved, the curved mirror was adjusted by hand at manufacture to maximise the lasers Q. They were a polished ceramic/glass called zerodure, a zero effective expansion material, below 1ppm/degC. They were a dielectric mirror with 14 incredibly thin dielectric layers deposited, the light was bent rather than reflected. It was believed that the curved mirror was the major contributor to backscatter, but the levels were incredibly small, it was atomic levels of uneveness.

Don't remember what type mirrors were used with the RLG, but at some time these could have been diamond turned aluminum since Honeywell had that capability in-house (we utilized this on our XM21 Remote Sensing Chemical Agent Detector based upon controversial Passive Spectral Radiometry....long story about solving the impossible if someone is interested).

Best,

[RoGeorge]

Detector based upon controversial Passive Spectral Radiometry....long story about solving the impossible if someone is interested

Interested?  You must be kidding, not only interested but rather eager to hear more, of course. 


About the Peltz osc, so far the easiest way to lower the harmonics seems to be by lowering the supply voltage, though it won't oscillate when powered with less than 0.6V, but have some other ideas to try.