Low-voltage AC-powered oscillator

If the power supply voltage is high, say >=3V, it is pretty simple
to design an oscillator (of unimportant output frequency and shape),
which can be powered from an AC power source: just add a diode bridge
at the power input lines and use any suitable DC-powered oscillator
circuit. But what if the supply voltage is really low, say 0.2V?
JFET-based oscillators can go really low, allegedly down to 25mV,
so it should be possible to pair two of them anti-paralelly and
power them directly from AC. However, the JFET structure exhibits
a high degree of symmetry, i.e. the D and S pins are called so
by convention and the device can work in "inversion mode". Is
it possible to use this fact to build an AC-powered oscillator
with a single JFET?

	Best regards, Piotr

>JFET-based oscillators can go really low, allegedly down to 25mV, 
>so it should be possible to pair two of them anti-paralelly and 
>power them directly from AC. However, the JFET structure exhibits 
>a high degree of symmetry, i.e. the D and S pins are called so 
>by convention and the device can work in "inversion mode". Is 
>it possible to use this fact to build an AC-powered oscillator 
>with a single JFET?

Interesting question. The main problem is going to be that even a symmetric FET is still polarized, i.e. an NFET is different from a PFET. If you turn a common-source stage upside down, you get a source follower, so your voltage gain goes away. Another way of putting this is that in an NFET the drain current increases when the gate goes more positive. Thus at a given frequency inverting the supply turns positive feedback into negative, so there's not enough gain for it to oscillate there. 

There are common-drain oscillators (Colpitts for instance) as well as common source ones, and all the reactances are frequency dependent. 

If you don't mind the frequency being different on the two half-cycles, it's probably possible, I'd say. 

Cheers

Phil Hobbs

On Tuesday, 27 December 2016 10:54:04 UTC, Piotr Wyderski  wrote:

> If the power supply voltage is high, say >=3V, it is pretty simple
> to design an oscillator (of unimportant output frequency and shape),
> which can be powered from an AC power source: just add a diode bridge
> at the power input lines and use any suitable DC-powered oscillator
> circuit. But what if the supply voltage is really low, say 0.2V?
> JFET-based oscillators can go really low, allegedly down to 25mV,
> so it should be possible to pair two of them anti-paralelly and
> power them directly from AC. However, the JFET structure exhibits
> a high degree of symmetry, i.e. the D and S pins are called so
> by convention and the device can work in "inversion mode". Is
> it possible to use this fact to build an AC-powered oscillator
> with a single JFET?
> 
> 	Best regards, Piotr

Since the supply is ac there are times when the active element can't oscillate, so you're stuck with an oscillator that runs part of the time driving a tank that keeps resonating. Given that you must accept that why not just oscillate on +ve half cycles only. Spectral purity is maximised with high q - can you run a crystal at <0.2v?

Having said all that I can think of other approaches.


NT

Phil Hobbs wrote:

> There are common-drain oscillators (Colpitts for instance) as well as common source ones, and all the reactances are frequency dependent.

How about common gate? Seems to be symmetric enough.

> If you don't mind the frequency being different on the two half-cycles, it's probably possible, I'd say.

The question is out of pure curiosity, so different frequencies are fine.

	Best reagrds, Piotr

tabbypurr@gmail.com wrote:

> Since the supply is ac there are times when the active element can't oscillate

Which is exactly what I want. :-)
I was thinking about a weirdo isolated zero crossing detector.
Having 2xLND150 to cut the current down to 1.5mA the problem
is now to notify the other side that the current is flowing.
The AC-opto approach is too boring and a magamp too bulky. :-)

 > Having said all that I can think of other approaches.

Please share your ideas then. The assumption: the circuit
must be self-powered (i.e. no external power supplies are
allowed, but it can steal up to 1.5mA from the input).

	Best reagrds, Piotr

On 12/27/2016 09:37 AM, Piotr Wyderski wrote:
> Phil Hobbs wrote:
> 
>> There are common-drain oscillators (Colpitts for instance) as well as
>> common source ones, and all the reactances are frequency dependent.
> 
> How about common gate? Seems to be symmetric enough.

Doesn't fix the polarization of the FET.  It won't amplify if you come
in via the drain.


> 
>> If you don't mind the frequency being different on the two
>> half-cycles, it's probably possible, I'd say.
> 
> The question is out of pure curiosity, so different frequencies are fine.
> 
>     Best reagrds, Piotr

Cheers

Phil Hobbs

-- 
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510

hobbs at electrooptical dot net
http://electrooptical.net

On Tuesday, 27 December 2016 14:44:49 UTC, Piotr Wyderski  wrote:
> tabbypurr wrote:
> 
> > Since the supply is ac there are times when the active element can't oscillate
> 
> Which is exactly what I want. :-)
> I was thinking about a weirdo isolated zero crossing detector.
> Having 2xLND150 to cut the current down to 1.5mA the problem
> is now to notify the other side that the current is flowing.
> The AC-opto approach is too boring and a magamp too bulky. :-)

fairly easy... pass your oscillation through a small C

I expect your no-oscillate region will be too long to make a good enough zero cross detector, but some interpolation on the other side might get it passable for some things.

>  > Having said all that I can think of other approaches.
> 
> Please share your ideas then. The assumption: the circuit
> must be self-powered (i.e. no external power supplies are
> allowed, but it can steal up to 1.5mA from the input).
> 
> 	Best reagrds, Piotr

FWIW...
Put the 0.2v through a transformer and run a more conventional oscillator
Use L,C to create an out of phase supply line too, then give each supply line its own oscillator feeding the same tank so oscillation can be continuous
Mechanical rectification to get a bit more voltage.
Mechanically gated smpsu to get more volts out

Only the 1st has the faintest chance of meeting your requirements.


NT

tabbypurr@gmail.com wrote:

> fairly easy... pass your oscillation through a small C

Inductive coupling is far easier. If the potential difference
is high, the C will be nowhere close to any reasonable definition
of "small". An X2 cap would be bulky, but several inches of TIW
wire is cheap and solve the issue easily.

> I expect your no-oscillate region will be too long to make a good enough zero cross detector

Why? Do you mean the startup time?

> Only the 1st has the faintest chance of meeting your requirements.

Do not underestimate the capabilities of modern JFETs. :-)
It is pretty easy to make a blocking oscillator for 200mV
(google => Joule thief) and then couple two of them to oscillate
in the separate half-cycles. I'm just curious whether one
device would suffice, given its symmetry.

BTW, you don't need mechanical switching for boosting small
voltages. Just use regular low RDS_on MOSFETs and power
the controller from the switcher's output in the "perpetuum mobile" 
fashion ;-) and add a JFET-based Joule thief to kickstart the whole 
caboodle. BTDT.

	Best regards, Piotr

Phil Hobbs wrote:

> Doesn't fix the polarization of the FET.  It won't amplify if you come
> in via the drain.

I'm afraid I don't understand what you mean, Phil. The only polarization
in a JFET I know of is the polarization of the gate diode, which is a
P region somehow embedded within the N silicon bar which makes up D and 
S. Drain is the end with higher potential and source with the lower, but
it flips as the AC waveform changes. Is it correct?

Do you mean the problems with *external* gate polarization circuitry,
which would inject current through the gate if powered incorrectly?

	Best regards, Piotr

On Tue, 27 Dec 2016 17:47:46 +0100, Piotr Wyderski
<peter.pan@neverland.mil> wrote:

>Phil Hobbs wrote:
>
>> Doesn't fix the polarization of the FET.  It won't amplify if you come
>> in via the drain.
>
>I'm afraid I don't understand what you mean, Phil. The only polarization
>in a JFET I know of is the polarization of the gate diode, which is a
>P region somehow embedded within the N silicon bar which makes up D and 
>S. Drain is the end with higher potential and source with the lower, but
>it flips as the AC waveform changes. Is it correct?
>
>Do you mean the problems with *external* gate polarization circuitry,
>which would inject current through the gate if powered incorrectly?
>
>	Best regards, Piotr
>
>

Phil only thinks discrete devices which (usually) have back
gate/substrate tied to source.  In my I/C world it's trivial to have
both gates independent, so a _symmetrical_ grounded-gate situation is
quite possible.

(I am currently working with a major ASIC foundry to get a good JFET
structure added to their repertoire.)
		
                                        ...Jim Thompson
-- 
| James E.Thompson                                 |    mens     |
| Analog Innovations                               |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| STV, Queen Creek, AZ 85142    Skype: skypeanalog |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |