# Help for negative resistance oscillator

Discussion in 'Electronic Design' started by Robert Baer, Apr 28, 2004.

1. ### Robert BaerGuest

What i have: an inductor of unknown value, but most likely in the
henry region, with resistance from 4K to 10K.
It is a grounded inductor, and has so many turns that it cannot stand
current thru the coil - as it would saturate.
I would like to make an oscillator that uses the inductor as one of
the frequency determining components, snd it seems a negative resistance
oscillator would do the trick.
However, i have no reference material that gives circuitry for
solid-state devices that could do this - not even op amps.

2. ### Kevin AylwardGuest

oscillator concept. Unless the oscillator is a *special* one that uses
an actual negative *resistance* region of a device, e.g. a tunnel diode,
it is still a conventional oscillator. The negative real part of an
impedance concept, is no more than a mathematically re-interpretation of
conventionally understood loop gain. For example, a Colpits oscillator
can be analysed either directly by loop gain, or by noting that its
input impedance at a node has a 1/s^2 term, i.e. a negative real part.

A true negative resistance oscillator has an explanation based on an
independent of frequency resistance. Very few oscillators have this
property.

It may be that such a negative resistance oscillator may have an
advantage for high resistance inductors, but you need to supply a solid
reason why, before departing from conventional approaches. What
frequency do you want the oscillator at?

Kevin Aylward

http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

3. ### Tim StinchcombeGuest

Maybe you are thinking of the "negative resistors" frequently involved in
chaotic oscillators? They are often referred to as "nonlinear resistors" or
"Chua's diode", and generally have an I-V characteristic with a negative
slope (usually with a 'breakpoint' in the middle). They are quite easily
made with an op-amp and a few resistors. Have a look around here for
starters: the resistors around the OP-27 give the desired effect:

http://www.cmp.caltech.edu/~mcc/chaos_new/Chua.html

Googling on the latter two terms above should give other hits using similar
arrangements.

4. ### John JardineGuest

(Sorry, I can't see the original post)

Have found the most reliable way to resonate high resistance big inductors
is to use a ('balanced' or 'long tail') pair of transistors. The L-C tuned
circuit is connected as a collector load and the oscillating voltage across
the tuned circuit is fed back to the other transistor's base.
Run the DC supply voltage in the 10-20V area and this allows many volts of
oscillation headroom before the circuit starts clipping.
(If you use PNP transistors then the oscillation will be centred on the 0V
rail)

Static DC coil resistance isn't an issue as the tuned circuit is being fed
purely by AC and no static voltage offset is developed. The coil resistance
just appears as an oscillating 'loss component' causing the resonant
frequency to drop a (calculable) amount.
As a balanced pair is naturally "wideband" then the arrangement is also
capable of resonating inductors out to many MHz.

A neat side benefit is that the oscillation voltage can be very smoothly
adjusted, just by altering the DC current in the commoned emitter
connection.
This 'tail current' ma's equates directly to the coils "Q" value at the
resonant frequency. Hence all the usual tuned circuit parameters can be
extracted from the info generated by this simple setup.

(I say this as just for fun, I'm playing with a similar design at the
moment. It's quite amusing to see a big inductor resonating with 100uF at
1Hz.

regards
john

5. ### Robert BaerGuest

It would be nice if it could oscillate in the tens of KHz region, but
the large inductance probably would not support that.
The internal resistance is so high that i have not found a way to
determine the "self-resonant" frequency (exciting with a zener in
negative resistance mode was of no help).

6. ### Robert BaerGuest

I take it that the resistor-diode part acts as an amplitude limiter.
Thanks for the idea; maybe i can get son=mething from it on an
experimental basis.
Since the final circuit canot have floating supplies, this exact
scheme seems to be unuseable with a grounded inductor.

7. ### Robert BaerGuest

However, there are two problems (which i had originally mentioned):
the inductor cannot support any DC current, as it will easily saturate,
and it is grounded on one end.

8. ### John JardineGuest

?
A PNP pair -will- oscillate a 'one end grounded' inductor. And Yes; it
generates a two terminal negative resistance.
The negative resistance thingy applies to any oscillator. It's prime
invocation is to cast fear and confusion upon people not of the
brotherhood. (a la dB's and hexadecimal
The inductor never sees a static-DC-current-offset hence no lopsided core
magnetism and assymetric saturation. The circuit is -resonant-, the ma's of
fluctuating collector current are just making up for the oscillating losses.
The coil's DC resistance turns up as a balanced term in both the positive
and negative half cycles.
regards
john

9. ### Fred BartoliGuest

So what ? Just use PNPs and active loads.

|
.-.
| |
| |
'-'
|
---+---
| |
|< >|
-| |-
|\ /|
| |
| GND
+---+--------------------.
| | |
| .-. |
--- | | || |/
--- | | .---||---+---|
| '-' | || | |>
| | | |\ | |
| `-----+-|-\ | |
.------+ | >---+ .-.
| .|. Ref---|+/ | |
--- |C| |/ | |
--- |C| '-'
| |C| |
| | |
GND GND GND
created by Andy´s ASCII-Circuit v1.22.310103 Beta www.tech-chat.de

I let you close the oscillator loop.

BTW 10kHz with a 1H inductor is 250p. And a multi kOhm DC resistance
inductor is probably much more than 1H, so don't count too much on the tens
of kHz.

Thanks,
Fred.

10. ### Fred BartoliGuest

???
I must have missed something. Could you post a schematics please ?

Thanks,
Fred.

11. ### Steve RobertsGuest

Do a google search for lambda diode and grid dip meter
its a P chanel and a N channel fet kinda back to back
and is used to replace tunnel diodes.

or see

http://www4.tpgi.com.au/ldbutler/NegativeResistance.htm

your inductor sounds like it is large, I dont know if a lambda pair will
oscillate that low but it may be worth a try.

Steve Roberts

12. ### Scott StephensGuest

It ought to be in the 10's of henry region with that resistance. IIRC an
auto ignition coil is 20 henry's and several K ohms.
Look into using a gyrator. Art of Electronics has one type, there are
others. I think you might get some gyrators to go unstable (whether you
want it or not, it seems).

I've seen several articles for negative resistance oscillator design for
microwave vco's, that might not be of much use.

Although you might apply the concept of using a grounded-base
oscillator. Make a series tuned circuit with your inductor and a
resonant cap, and ground the base of a transistor. Then apply your
positive-feedback from collector output-port to emitter input-port.

--
Scott

**********************************

DIY Piezo-Gyro, PCB Drill Bot & More Soon!

http://home.comcast.net/~scottxs/

**********************************

13. ### John JardineGuest

Maybe -I've- missed something
It's pretty much the same as your ascii art but no series blocking cap. I've
sent the circuit to A.B.S.E.

regards
john

14. ### Fred BartoliGuest

Ok, so at least I'm still good to something. But I still don't understand
how you can say there's *no* DC current through the inductor.
Do you intend to run it class C ?

Oh, and I'm sorry but my ISP just dropped the abse group. Could you please
send it directly to me ?

Thanks,
Fred.

15. ### John JardineGuest

You're correct.
Having just looked again at the diagram it's obvious that a DC current must
flow throught the L. I'll keep my mouth shut from now on.
(also said times before!).
The error in my thinking occurred when the large inductors I was resonating
were passing DC currents of a few uA and the meters weren't indicating it.
I was also doing the sums from an equiv' circuit that had no bias.
(if in error I blame the test gear
I'll send the diagram on so you can see this.
regards
john

16. ### Fred BartoliGuest

So will I. My comment about class C was obviously equally wrong as I
realised just after having it sent.

Thanks,
Fred.

17. ### John JardineGuest

[I'll try again. Second time sent.]

You're correct.
Having looked at the diagram it's obvious that a DC current must
flow throught the L. I'll keep my mouth shut.
(also said times before!).
The error in my thinking occurred when the large inductors I was resonating
were passing DC currents of a few uA and the meters weren't indicating it.
I was also doing the sums from an equiv' circuit that had no bias.
(if in error I blame the test gear
I'll send the diagram on so you can see this.
regards
john

18. ### Tony WilliamsGuest

[snip]

I wonder if it could be done with a variation of the
Baxandall sine wave oscillator. ie, Parallel-resonate
the coil and drive it with a switched constant-current
square-wave, with the polarity of the const-I being
swapped at each zero-crossing of the voltage across the
tank. Rough sketch below.

-+---+-----+- +Vs
| | R5
| R1 |
| | |/e
| +---|pnp C2
| | |\ +------||------+
| R2 | | |
0v--|\| | | C1 | RL L |
|S>---+ +--||--+--/\/\--////--+--0v
+-----+|/| | | |
| | R3 | | +------+
| | | |/ | | |
| | +---|npn | +-|\ |
| | | |\e | |O>--+
| | R4 | +----|/ |
| | | R6 |
| -+---+-----+- -Vs |
| |
+--------------------<--------------+

It's a pair of constant-current sources, alternately
switched by comparator/gate (S), which is driven by
the buffered sinewave across the resonant tank. C1
is a large dc-blocker and C2 is the resonating capacitor.

The Baxandall circuit requires a Q of roughly 5-10 for
best operation. So the design frequency will initially
be determined by Q = wL/R. With the rough numbers already
given this suggest an Fosc in the 6KHZ to 10KHZ region.

That then determines the value of C2, which will be around
500pF. C2 includes the Cstray of the inductor.

The impedance of the tank at Fres is Z = L/CR, so will
be up in the 300k region. With 15v supplies, and a few volts
across the tank, the required switched currents would then
be of the order of +/- 20-40 uA or so.

All numbers above winged on the fly, and to be confirmed.

An interesting possibility could be to use a single OTA as
the switching constant-current source. An OTA has a voltage
adjustable current output and this could be used to
stabilise the amplitude of the voltage across the tank.

C2
+------||------+
__ +Vs | |
0v--+---| \| C1 | RL L |
R1 |OTA>----||--+--/\/\--////--+--0v
+--+|__/| |
| | -Vs | +------+
+----|->--+ | | |
| +-------<---R2---|--+-|\ |
| | |O>--+
| C3 +----|/ |
+--R3--+---||--+ |
| | D |
| /|--+--+---R4---|<|--+
+--<O| |
\|--+ +---R5---|>|--+
| D |
0v-+- |
-+- -Vs

The same +1 buffer looks at the voltage on the tank
and now switches the polarity of the OTA's output
current via R2 and R1. R1 and R2 should probably be
sized for a 2V or so peak voltage across R1.

Actually I'm not quite certain of the last sentence.
An OTA has a linear (log) region of about 100mV or
so. So there may be a need to check that there is
enough loop gain for oscillation with R2:R1 ratios.

The additional opamp is an erroramp/integrator that
compares the half-wave rectified AC via R4 against
a reference current from R5, and sets the required OTA
output current via R3. R3 should be sized for about
100uA through it when the integrator is at pk +Vout,
(R3= 250k-ish with 15-0-15 supplies). R4 and R5 should
be sized so that the ratio of R5/R4 is about 4.7/1
(giving ACVpk= -Vs*2/3 approx).

Ummm... haven't a clue whether the above will work or
not. Suggest you SPICE it first.

19. ### Robert BaerGuest

I sort-of guess that i would be lucky to get a "self-resonance" in the
KHz region...

20. ### Robert BaerGuest

OOOOh! *VERY* nice!