Armstrong, Hartley, Colpitts, Clapp, Wallin...

Fascinating circuits Vadim!

"Delay is introduced in mixer/limiter/current to voltage stage."  - Buggins

The diodes and 1 Ohm Rs seem problematic.  There to prevent current starving / open loop operation?  Probably a lot of added C, and a scary over voltage situation.

"As well, I'm trying to use two current mirrors - to provide drive signal close to supply voltage.  I think it's a nice idea to implement oscillator with two pairs of current mirrors.  Although, I have some troubles when trying to get it oscillating..."

Splitting the sense resistor is quite an interesting idea.  If the supply voltage were higher one could maybe do cascode to speed it up a little.

Start-up usually seems to come down to the issue of supplying some drive at crossover, the same sort of thing they run into with audio amplifiers and distortion.  My 8 xistor oscillator has many elements of an audio amplifier (differential input stage, push-pull output stage) which I guess isn't a surprise.

The diodes and 1 Ohm Rs seem problematic.  There to prevent current starving / open loop operation?  Probably a lot of added C, and a scary over voltage situation.

1 Ohm Rs are just replacement for wires - to avoid deletion, then adding back if needed. Forgot to cleanup the circuit

R at drive stage power supply can be reduced a bit. (Downto 1 Ohm - for tesla coil mode?)

I'm looking for some replacement of output stage.
Tried opamp - but it starts oscillating if powered via Rs.
CMOS push-pull stage looks as a good idea. Tried nmos+pmos - working pretty good w/o load.
Some powerful unbuffered inverter like lvc1gu04 might work. Found SN74LVC1GU04 spice model, but it does not work correctly.
As well, not sure if unbuffered inverter would work fine with small signals (oscillation startup).

Tried current mirror as output stage. Common emitter connection of current mirror allows almost rail-to-rail output.
As well, such drive stage could eliminate drive-to-voltage stage. Output of current mirror current sensing might feed current mirror drive directly.
There are two issues: hard to center output voltage (some feedback loop is needed), delay when saturated (about 15ns).

Once issues resolved, we could get really nice circuit, full of mirrors

Designed current sensing oscillator on LT6200-10 opamp as differential amplifier.
Back to classic Eric's sensing approach - with R_sense in series with LC tank.

LT6200-10 is a bit expensive, but looks like a good choice.
Most important achievement in new schematic is extra low phase error - 1.5ns between inductor current and drive signal.
It allows to track oscillation frequency very close to LC resonance.

E.g. for Eric's BJT diff amp current sensing oscillator, phase error is near 15ns.

For 2.7mH inductor with 120 Ohm serial resistance and 1.5pF parallel capacitance, it gives 310Vpp on antenna by feeding LC tank with 8mA via 22 Ohm R_sense.

Drive waveform and LC tank current:

antenna swing:

Increasing of gain may give more square-like drive signal, with higher voltage swing on antenna.
But when loop gain is too high, strange high frequency pinging appears in drive signal.
It looks like amp gain should be selected based on inductor parameters.

Cheaper opamp is used as output buffer - I like Eric's idea to use low slew rate signal as sensor output.
Alternatively, drive signal may be fed to MCU directly, or via some logic buffer.

LTSpice model link.

I've tried to simulate hand touch on antenna - with C_hand 330pF and R_hand 150K.

For lower R_hand simulation stops. Probably it makes sense to tune components for better touch resistance, but I'm unsure what is proper way to simulate touch.

Super interesting stuff Vadim!  I wish I could contribute more but the D-Lev kits are consuming most of my free time.  Still, I'm thinking about your oscillator a lot.  On a basic level I believe the coil requires voltage drive, not current drive, so that the current through it can be sensed.  Sensing the current via separate resistors in the drivers themselves is genius, but I don't have a ready topology to try.  And an inversion somewhere is necessary.  The starting problem seems to come down to crossover drive, and audio amplifiers might provide some solutions (like output current limiting).  Since the coil doesn't need much drive (~5mA) I wonder if that could somehow be made quiescent through both driver transistors, with the swing offsetting it one way or the other?  I think we're hampered in a way by the 3.3V supply, which makes it even more of a challenge.

[EDIT] I should be working on kits! ;-)

Your addition of a capacitor and removal of the input base resistor in the 4 xistor buffer/driver is a good move:

But it works better I think when the input collectors are shorted to the output emitters? [SPICE]

Current Mirror Oscillator

A blend of my output buffer and a variation of Vadim's innovative current mirror input:

Note the weird interconnectedness of the mirror.  No clue if it works in real life, but only needs a minor kick in LTspice to get going (1nA) which is usually a good sign, and the sim can take a fairly heavy damping load.  Drive waveform gets weird looking with smaller coils and lighter loads, but the resulting oscillations are still healthy (~180Vpp above).  4th order LPF output is centered and very close to a sine.  I like the symmetry!  [SPICE]

Nice strange oscillator, Eric 

Current sensing oscillator contest...

I've managed to implement weird 4-transistor current sensing oscillator, 8ns phase error, 220Vpp on antenna - for 2.7mH 120 Ohm, C_ant=10pF : drive is 2.4V 6.5mA
Only 4 resistors and one cap...

Will share design soon...

"Current sensing oscillator contest..."  - Buggins

I can't tell you how much I've been enjoying your oscillator explorations, thank you Vadim for sharing with us!

Diving deeper into oscillators exploration.

How can we improve recent Eric's oscillator?
Let's replace output driver buffer.
We can take opamp, connect power pins via Rs and measure current.
But fast rail to rail opamps are a bit expensive, and have overdrive issues.
What about using amplifier which does not have overdrive issues?
Unbuffered inverters (U04, actually a pair of mosfets) have no problem with overdive, can provide enough current, and LVC series propagation delay is low enough.
We need non-inverting buffer - two chained invertors can do their work.

Working good enough. Gives ~400Vpp for 2.7mH 120 Ohm inductor, and 8pF antenna.

Can we simplify it?

...

Here comes the magic.

...

Lets break mirrors and cut them off.

It's alive!

LTSpice model file is available on GitHub

SN74LVC2GU04 and other component models can be downloaded here

Unbuffered invertor looks very interesting, but has some issues, at least with simulation in LTSpice.
LVC2GU04 probably has incorrect (or unsuitable for LTSpice) spice model.
I've tried to add several inverters in parallel as drive, and additional inverter as sensor output amplifier.
Often signals disappear from simulation.

With opamp as driver, curent sensing via resistors on power rails does not look as a working idea.
In most cases, opamp self-oscillates at high frequency.

Trying classic drive resistor current sensing.

Yet another oscillator: cheap enough dual opamps oscillator.
Antenna voltage swing 400Vpp for 2.7mH  120 Ohm inductor and 10pF antenna.
Phase delay between inductor current and drive signal: 15ns

It looks like fully differential amplifier ad8132 has internal limiter which limits its output by 1Vpp.
It's very nice, because w/o limiter opamps enter into overdrive mode and introduce very high phase delay.

Additional opamp amplifies it to 3.1Vpp and used as tank driver.

Survives 22K + 330pF touch simulation.

LTSpice model link

ad8132 symbol and model are available on github.

I've designed current sensing oscillator with PLL.
Gives ~390Vpp on antenna with 2.7mH 120 Ohm inductor and 10pF antenna.
Can be tuned to have really low (<1ns) phase shift between LC drive and current.
Drive signal is sine-like, and produces little noise on antenna.

Fully differential amplifier on U1 measures current via R7 to VCO LC tank.

ad8132 has internal limiter which limits output swing by 1Vpp.
VCO drive current is limited additionally by big R7 value and R5/R6 divider to limit VCO voltage.
Inductor L1 and caps C3, C4 and varactors form VCO LC tank.
Component are chosen to have frequency range covering main LC tank frequency range.
U1 output is amplified by buffer on U2 opamp and fed to main LC tank via sensing resistor R_sense.
R_sense current is detected by U3 comparator adcmp600.
Other side of L1 VCO inductor is sensed by U6 inverter buffer, and delayed by R16/C8 to tune phase shift between main LC tank drive signal and its current at 0 when PLL is locked.
U4 is XOR based phase detector. It's output is integrated by U8, scaled a bit and used as VCO control.
U7 is just a buffer for sensor output signal.

LTSpice model link