Sensing small inductances

[Steve Wilson]

jlarkin@highlandsniptechnology.com wrote:

Jeroen's sim, run in time domain, clearly shows the negative
resistance that an emitter follower can present.

Jeroen's sym does not reflect reality. How do you control the amplitude if an
oscillator is running on negative resistance.

 

On Thu, 29 Aug 2019 16:04:04 GMT, Steve Wilson <no@spam.com> wrote:

Jeroen Belleman <jeroen@nospam.please> wrote:

Steve Wilson wrote:
Jeroen Belleman <jeroen@nospam.please> wrote:

Steve Wilson wrote:
John Larkin <jlarkin@highland_atwork_technology.com> wrote:

On Wed, 28 Aug 2019 22:35:44 GMT, Steve Wilson <no@spam.com> wrote:
That's his circcuit. Try a Colpitts.

Any passive resonator that oscillates is seeing a negative
resistance. Energy is conserved, and the resonator loss has to be
made up for.

Where did you get that? Try measuring the input impedance of a
Colpitts. It is always positive. The negative resistance is a
complete misnomer. It has nothing to do with the voltage increasing
and the current decreasing. It is only the resistance needed to stop
oscillations.

The Barkhausen criteria states the phase shift has to be zero or
multiples of 360 degrees, and the loop gain has to be equal to or
greater than 1. No negative resistance is needed.

The Barkhausen criterion and the negative resistance view are
really just two ways of looking at the problem, completely equivalent.

You have to get rid of this "voltage increasing and the current
decreasing" idea. Impedance is a complex function of frequency.
If its real part goes negative in some range of frequency, it's
completely valid to talk of negative resistance and it will
happily supply energy to an LC resonator if it happens to be
tuned within that range. This often happens by accident in emitter
followers.

It's really trivial to turn my LTspice emitter follower example
into a CC Colpitts oscillator. Just replace the AC source by a
200nH coil and increase the series resistor to 500 Ohms or so, to
limit the rate of growth of the oscillator amplitude to something
reasonable. Note that the equivalent coil Q is really, really bad! Set
an initial condition ".ic v(in)=1" to get it going. Do a .tran
simulation of the first 50ns or so and you'll see a healthy
exponentially growing sinusoid, in spite of the poor Q.

Of course, there is nothing in this AC equivalent circuit to limit the
amplitude, so it just keeps growing.

Jeroen Belleman

The exponential increase in amplitude as oscillations start is a normal
function of oscillators. The amplitude increases until the energy
delivered to the tank matches the energy lost in the tank.

If the idea of negative resistance were true, there would be nothing to
limit the amplitude of oscillations, and no way to control it.

Isn't that what I wrote, just a few lines higher?

That was in your simulation. In a real oscillator, the amplitude of
oscillations increases until the energy delivered to the tank matches the
energy lost.

No linear analysis will predict the oscillation amplitude. If the loop
gain is 0.9999, oscillation won't happen. If it's 1.0001, the gadget
will explode and kill us all.

In a cc Colpitts, the energy delivered to the tank is controlled by the
emitter current, which is adjusted by changing the emitter resistor.

This proves the idea of negative resistance is false.

Jeroen's sim, run in time domain, clearly shows the negative
resistance that an emitter follower can present.

 

[piglet]

On 30/08/2019 05:32, bitrex wrote:

On 8/29/19 1:41 AM, piglet wrote:
On 28/08/2019 16:03, bitrex wrote:
Well the issue is the inductances I wanna measure, in the single
digital microhenries, to within a nanohenry, say, have a Q of about
0.3 in the low MHz.

when you can get them to work as part of a standard oscillator tank
that oscillates at all, an octave and a half below their
self-resonant frequency, the stability is poor. if you have say a 4uH
resonating with a 10n cap to get 5MHz a 1 nH difference in the L is
only a few hundred Hz shift. But the oscillator is drifting around by
several kHz over 20 minutes

Huh? My calculator makes 4uH 10nF come to 800kHz? For 5MHz at 4uH it
needs C=250pF

Right, SOMEBODY forgot the two pi while writing that

Are these nH changes in your low-Q uH inductor that you want to track
happening fast or slowly? Are you after absolute values or just
watching deltas?

piglet


Not much faster than 10s of Hz. And absolute values in this case but
it's an inductive transducer would like to know how much it is strained
compared to its relaxed state, so absolute accuracy with respect to some
standard is less important than precision/repeatability.

Ok so I figure you don't have time to measure free running oscillator
frequency to very high precision but since the inductance is pretty much
confined to a small range you might be able to use very simple old
technique like the 3 voltmeter method. No phase sensitive detectors
required.

<https://www.dropbox.com/s/qiwnh4r7jra0tok/IMG_1472.JPG?raw=1>

Select R (an ordinary positive resistor) and frequency for best ranging.
Controller will have to do some trig to extract L value (and r loss
resistance value of the coil) but if your range is limited lookup tables
may be enough.

piglet

 

On Fri, 30 Aug 2019 10:44:00 GMT, Steve Wilson <no@spam.com> wrote:

jlarkin@highlandsniptechnology.com wrote:


Jeroen's sim, run in time domain, clearly shows the negative
resistance that an emitter follower can present.

Jeroen's sym does not reflect reality. How do you control the amplitude if an
oscillator is running on negative resistance.

Any stable oscillator, with any gain mechanism, has to have a
nonlinear mechanism to set the oscillation amplitude. You can think of
a negative resistance as being nonlinear on sinewave amplitude, or you
can imagine it to be a theoretically perfect negative resistor in
parallel with a tank, in parallel with some sort of hard or soft
clipper that adds lossy shunt resistance as oscillation amplitude
increases. Something has to servo the gain to exactly 1 at the
operating amplitude. Barkhausen magically assumes the gain is
1.000000.

So it's difficult to use one negative resistance element to make two
tanks oscillate; if the gain is enough for one to oscillate, it's not
enough for the other, lower-Q one.

Transistors used to be expensive, so the original TouchTone phone
managed to oscillate at two frequences simultaneously using one
transistor.

https://www.dropbox.com/s/br6namqyxas1f02/Bell_DTMF.jpg?raw=1

Sneaky.

 

On Fri, 30 Aug 2019 09:34:23 GMT, Steve Wilson <no@spam.com> wrote:

Steve Wilson <no@spam.com> wrote:

George Herold <gherold@teachspin.com> wrote:

AS JL said, I think you're just having a terminology disagreement.
All the oscillators I've built grow until the amplitude hits the power
rails.. or is limited by some other non-linear element that is in the
loop/ circuit.

There is no excuse for railing an oscillator.

Post one of your designs and I'll fix it for you.

Many oscillators inherently rail, for example common CMOS XOs. There
is no other amplitude limiting mechanism. Nothing wrong with that.

Some oscillators self-rectify at the gate or base and bias themselves
off at some not-hard-railed amplitude. I invented this one when I was
a kid; it flew on the C5A.

https://www.dropbox.com/s/gisv3uqrm5wb61m/LC_osc.JPG?raw=1

The p-p amplitude is almost exactly 2x V+, and it has a tiny flat on
the negative swing of the sine wave. At the negative swing peak, it
steals its own base bias. It even has a near-zero amplitude tempco.

 

On Fri, 30 Aug 2019 08:07:55 -0700 (PDT), George Herold
<gherold@teachspin.com> wrote:

On Friday, August 30, 2019 at 10:32:53 AM UTC-4, jla...@highlandsniptechnology.com wrote:
On Fri, 30 Aug 2019 10:44:00 GMT, Steve Wilson <no@spam.com> wrote:

jlarkin@highlandsniptechnology.com wrote:


Jeroen's sim, run in time domain, clearly shows the negative
resistance that an emitter follower can present.

Jeroen's sym does not reflect reality. How do you control the amplitude if an
oscillator is running on negative resistance.

Any stable oscillator, with any gain mechanism, has to have a
nonlinear mechanism to set the oscillation amplitude. You can think of
a negative resistance as being nonlinear on sinewave amplitude, or you
can imagine it to be a theoretically perfect negative resistor in
parallel with a tank, in parallel with some sort of hard or soft
clipper that adds lossy shunt resistance as oscillation amplitude
increases. Something has to servo the gain to exactly 1 at the
operating amplitude. Barkhausen magically assumes the gain is
1.000000.

So it's difficult to use one negative resistance element to make two
tanks oscillate; if the gain is enough for one to oscillate, it's not
enough for the other, lower-Q one.

Transistors used to be expensive, so the original TouchTone phone
managed to oscillate at two frequences simultaneously using one
transistor.

https://www.dropbox.com/s/br6namqyxas1f02/Bell_DTMF.jpg?raw=1
Huh, neat. (I have a hard time following what the transformers are doing.)

Are RV1 and RV2 the non-linear gain control do-thingies?

George H.

Sneaky.

Yes. Each parallel LC tank sees the series-shared negative resistance
but is paralleled by its own local varistor, so it oscillates but
limits its own amplitude.

Seems weird now, but switch matrices and multi-winding transformers
and varistors were cheaper than transistors back then.

 

[George Herold]

On Friday, August 30, 2019 at 5:34:28 AM UTC-4, Steve Wilson wrote:

Steve Wilson <no@spam.com> wrote:

George Herold <gherold@teachspin.com> wrote:

AS JL said, I think you're just having a terminology disagreement.
All the oscillators I've built grow until the amplitude hits the power
rails.. or is limited by some other non-linear element that is in the
loop/ circuit.

There is no excuse for railing an oscillator.

Post one of your designs and I'll fix it for you.

Irrelevant, I can fix it.. the fact remains that it does rail.
(w/o 'fix')
I guess an excuse could be that watching it rail, helps you
understand the oscillator.

George H.

 

[George Herold]

On Friday, August 30, 2019 at 12:00:10 AM UTC-4, Steve Wilson wrote:

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

I think Steve is stuck thinking of negative resistance in the time
domain, i.e.

dI_1/dV_1 = -R.

In the frequency domain, an input impedance like that gives an S11
greater than unity at all frequencies.

But in a real oscillator, S11 only has to have a negative real part near
the frequency of the resonator--since only that frequency band is
relevant, it's at least as good as having a negative input resistance at
all frequencies. In fact it's generally better, since the (slow) bias
network doesn't have to cope with the negative resistance at all.

The oscillator responds on a cycle-by-cycle basis to the input signal. See "A
General Theory of Phase Noise in Electrical Oscillators", Ali Hajimiri and
Thomas H. Lee:

https://authors.library.caltech.edu/4917/1/HAJieeejssc98.pdf

You are thinking of a 1-port device reflecting energy back to the source.

A real oscillator has two ports. For example, in a cc Colpitts, you can say
the base is the input and the emitter is the output.

The negative resistance is measured by inserting a resistor at the base. It
is the value needed to shut down oscillations.

However, if the oscillator ran on negative resistance, it would be impossible
to control the amplitude of oscillations. They would increase until the
oscillator railed.

Railing the oscillator increases the noise, which is bad in precision
oscillators.

Crystal oscillators must have careful control of the amplitude. If it is too
low, the crystal may not start. If it is too high, the crystal may fracture.

It is easy to control the amplitude of oscillation in a cc Colpitts. The
emitter resistor controls the emitter current and thus the energy delivered
to the tank. You set the current by changing the value of the emitter
resistor so that the energy delivered to the tank equals the energy lost in
the tank. This also works the same way with crystal oscillators.

This satisfies the Barkhausen criteria by setting the loop gain to unity.

Again, you cannot do this if the oscillator ran on negative resistance. It
has no concept of amplitude.

So the concepts of negative resistance and Barkhausen criteria are not
equivalent, and are not two ways of saying the same thing.

Negative resistance in oscillators does not exist.

Hey Steve, is negative resistance the same as having a pole in the right half
plane? Or are those two different things?

I wonder if you've ever built a Wien-bridge oscillator?

George h.

Cheers

Phil Hobbs

 

[Bill Sloman]

On Saturday, August 31, 2019 at 12:32:53 AM UTC+10, jla...@highlandsniptechnology.com wrote:

On Fri, 30 Aug 2019 10:44:00 GMT, Steve Wilson <no@spam.com> wrote:

jlarkin@highlandsniptechnology.com wrote:

Any stable oscillator, with any gain mechanism, has to have a
nonlinear mechanism to set the oscillation amplitude.

If you want a clean sine wave, you need a controllable gain element.

Jim Williams application notes on Wien Bridge oscillators used a FET as a controlled resistor.

My current mirror variation on the Baxandall class-D oscillator used an asymmetrical current mirror,and trimmed the asymmetry to get exactly the gain needed to sustain oscillation.

I've simulated a Wien bridge oscillator that used an AD734 multiplier to adjust the gain in the same sort of way.

A controllable gain element is intrinsically non-linear, but some at least can be set to give a linear (if variable) gain.

You've got to rectify the output of the oscillator to find out what amplitude output it is producing, and filter the rectified signal before you use to set the gain, but that's doable, if you wanted to be really cute , you could cancel out the residual ripple that would otherwise get into the gain-controlling signal, or use a sample and hold scheme to clean it up.

You can think of a negative resistance as being nonlinear on sinewave amplitude, or you can imagine it to be a theoretically perfect negative resistor in parallel with a tank, in parallel with some sort of hard or soft clipper that adds lossy shunt resistance as oscillation amplitude
increases.

You can think of lots of things, but most of design involves discarding the less productive thoughts.

Something has to servo the gain to exactly 1 at the

operating amplitude. Barkhausen magically assumes the gain is
1.000000.

That's what feedback is for.

<snipped a less productive thought>

--
Bill Sloman, Sydney

 

[George Herold]

On Friday, August 30, 2019 at 10:32:53 AM UTC-4, jla...@highlandsniptechnology.com wrote:

On Fri, 30 Aug 2019 10:44:00 GMT, Steve Wilson <no@spam.com> wrote:

jlarkin@highlandsniptechnology.com wrote:


Jeroen's sim, run in time domain, clearly shows the negative
resistance that an emitter follower can present.

Jeroen's sym does not reflect reality. How do you control the amplitude if an
oscillator is running on negative resistance.

Any stable oscillator, with any gain mechanism, has to have a
nonlinear mechanism to set the oscillation amplitude. You can think of
a negative resistance as being nonlinear on sinewave amplitude, or you
can imagine it to be a theoretically perfect negative resistor in
parallel with a tank, in parallel with some sort of hard or soft
clipper that adds lossy shunt resistance as oscillation amplitude
increases. Something has to servo the gain to exactly 1 at the
operating amplitude. Barkhausen magically assumes the gain is
1.000000.

So it's difficult to use one negative resistance element to make two
tanks oscillate; if the gain is enough for one to oscillate, it's not
enough for the other, lower-Q one.

Transistors used to be expensive, so the original TouchTone phone
managed to oscillate at two frequences simultaneously using one
transistor.

https://www.dropbox.com/s/br6namqyxas1f02/Bell_DTMF.jpg?raw=1
Huh, neat. (I have a hard time following what the transformers are doing.)

Are RV1 and RV2 the non-linear gain control do-thingies?

George H.

Sneaky.

 

[Joseph Gwinn]

On Aug 29, 2019, jlarkin@highlandsniptechnology.com wrote
(in article<c05hmelgktf15aj83l3akrpp53o2jjqp4u@4ax.com&gt:

On Thu, 29 Aug 2019 21:45:29 -0400, Joseph Gwinn
joegwinn@comcast.net> wrote:

On Aug 29, 2019, jlarkin@highlandsniptechnology.com wrote
(in article<hqofmet497f174cr4i1ddbrr64ipf7pao5@4ax.com&gt:

On Thu, 29 Aug 2019 09:47:27 +0200, Jeroen Belleman
jeroen@nospam.please> wrote:

Steve Wilson wrote:
John Larkin<jlarkin@highland_atwork_technology.com> wrote:

On Wed, 28 Aug 2019 22:35:44 GMT, Steve Wilson <no@spam.com> wrote:
That's his circcuit. Try a Colpitts.

Any passive resonator that oscillates is seeing a negative resistance.
Energy is conserved, and the resonator loss has to be made up for.

Where did you get that? Try measuring the input impedance of a Colpitts.
I

is always positive. The negative resistance is a complete misnomer. It
has
nothing to do with the voltage increasing and the current decreasing. It
i

only the resistance needed to stop oscillations.

The Barkhausen criteria states the phase shift has to be zero or
multiples
of
360 degrees, and the loop gain has to be equal to or greater than 1. No
negative resistance is needed.


The Barkhausen criterion and the negative resistance view are
really just two ways of looking at the problem, completely
equivalent.

You have to get rid of this "voltage increasing and the current
decreasing" idea. Impedance is a complex function of frequency.
If its real part goes negative in some range of frequency, it's
completely valid to talk of negative resistance and it will
happily supply energy to an LC resonator if it happens to be
tuned within that range. This often happens by accident in
emitter followers.

It's really trivial to turn my LTspice emitter follower example
into a CC Colpitts oscillator. Just replace the AC source by a
200nH coil and increase the series resistor to 500 Ohms or so,
to limit the rate of growth of the oscillator amplitude to
something reasonable. Note that the equivalent coil Q is really,
really bad! Set an initial condition ".ic v(in)=1" to get it
going. Do a .tran simulation of the first 50ns or so and you'll
see a healthy exponentially growing sinusoid, in spite of the
poor Q.

Of course, there is nothing in this AC equivalent circuit to limit
the amplitude, so it just keeps growing.

Jeroen Belleman

It's a terminology thing. The mechanization of the Barkhausen
criteria, namely provide a feedback in phase with the resonance and a
multiple of 360 degrees, with power gain avove 1, is in fact negative
resistance.

The oscillation criterion that net loop gain be exactly 1 is met by
something going nonlinear before it goes nuclear.

I do have a question. One can build an oscillator using an EFDA amplifier
and a length of optical fiber in a ring connecting input to output. Where is the
negative resistance here?

It's not a 1-port electrical resonator, so I don't know.

OK. I agree that a one-port amp must present a negative effective resistance
to the resonant part of the circuit to support oscillation. But not all
oscillators are one-port.

One can also build an oscillator with an amplifier and a length of
twisted-pair transmission line connecting input to output. Someone got a
patent for putting a half-twist in the line, to get the 180 degree phase
shift. Nice and small, fits into an IC.

Joe Gwinn

 

[bitrex]

On 8/30/19 7:11 AM, piglet wrote:

On 30/08/2019 05:32, bitrex wrote:
On 8/29/19 1:41 AM, piglet wrote:
On 28/08/2019 16:03, bitrex wrote:
Well the issue is the inductances I wanna measure, in the single
digital microhenries, to within a nanohenry, say, have a Q of about
0.3 in the low MHz.

when you can get them to work as part of a standard oscillator tank
that oscillates at all, an octave and a half below their
self-resonant frequency, the stability is poor. if you have say a
4uH resonating with a 10n cap to get 5MHz a 1 nH difference in the L
is only a few hundred Hz shift. But the oscillator is drifting
around by several kHz over 20 minutes

Huh? My calculator makes 4uH 10nF come to 800kHz? For 5MHz at 4uH it
needs C=250pF

Right, SOMEBODY forgot the two pi while writing that

Are these nH changes in your low-Q uH inductor that you want to track
happening fast or slowly? Are you after absolute values or just
watching deltas?

piglet


Not much faster than 10s of Hz. And absolute values in this case but
it's an inductive transducer would like to know how much it is
strained compared to its relaxed state, so absolute accuracy with
respect to some standard is less important than precision/repeatability.

Ok so I figure you don't have time to measure free running oscillator
frequency to very high precision but since the inductance is pretty much
confined to a small range you might be able to use very simple old
technique like the 3 voltmeter method. No phase sensitive detectors
required.

https://www.dropbox.com/s/qiwnh4r7jra0tok/IMG_1472.JPG?raw=1

Select R (an ordinary positive resistor) and frequency for best ranging.
Controller will have to do some trig to extract L value (and r loss
resistance value of the coil) but if your range is limited lookup tables
may be enough.

piglet

That seems plausible, thanks! yeah the original problem is that when DUT
is in a regular tank, when you can get it to oscillate stability at all
(at a frequency that's not say 150 MHz and kills the power budget on the
oscillator and uP clock speed requirement, both), the deviation with
respect to say a 1 nH change is too few Hz for the uP to distinguish
reliably from noise. The oscillators love to drift.

 

[Jeroen Belleman]

On 2019-08-30 03:45, Joseph Gwinn wrote:
[Snip!]

I do have a question. One can build an oscillator using an EFDA amplifier and
a length of optical fiber in a ring connecting input to output. Where is the
negative resistance here?

This doesn´t use fiber, but the principle is the
same:<https://www.findlight.net/blog/2017/07/18/optical-parametric-
oscillators/>.

Joe Gwinn

I like to think of a laser as a distributed negative-loss
transmission line for optical frequencies.

Jeroen Belleman

 

[Jeroen Belleman]

On 2019-08-30 18:26, Joseph Gwinn wrote:

On Aug 29, 2019, jlarkin@highlandsniptechnology.com wrote
(in article<c05hmelgktf15aj83l3akrpp53o2jjqp4u@4ax.com&gt:

On Thu, 29 Aug 2019 21:45:29 -0400, Joseph Gwinn
joegwinn@comcast.net> wrote:

On Aug 29, 2019, jlarkin@highlandsniptechnology.com wrote
(in article<hqofmet497f174cr4i1ddbrr64ipf7pao5@4ax.com&gt:
something going nonlinear before it goes nuclear.

I do have a question. One can build an oscillator using an EFDA amplifier
and a length of optical fiber in a ring connecting input to output. Where is the
negative resistance here?

It's not a 1-port electrical resonator, so I don't know.

OK. I agree that a one-port amp must present a negative effective resistance
to the resonant part of the circuit to support oscillation. But not all
oscillators are one-port.

One can also build an oscillator with an amplifier and a length of
twisted-pair transmission line connecting input to output. Someone got a
patent for putting a half-twist in the line, to get the 180 degree phase
shift. Nice and small, fits into an IC.

Joe Gwinn

The idea of negative resistance does not fit for such
oscillators, I agree. I've done that with ECL logic gates
as the amplifier. Very nice for snappy stop/start behaviour.
Gosh, that was in the late 1980's, over 30 years ago.

Jeroen Belleman

 

[Steve Wilson]

George Herold <gherold@teachspin.com> wrote:

On Friday, August 30, 2019 at 5:34:28 AM UTC-4, Steve Wilson wrote:
Steve Wilson <no@spam.com> wrote:

George Herold <gherold@teachspin.com> wrote:

AS JL said, I think you're just having a terminology disagreement.
All the oscillators I've built grow until the amplitude hits the power
rails.. or is limited by some other non-linear element that is in the
loop/ circuit.

There is no excuse for railing an oscillator.

Post one of your designs and I'll fix it for you.

Irrelevant, I can fix it.. the fact remains that it does rail.
(w/o 'fix')
I guess an excuse could be that watching it rail, helps you
understand the oscillator.

George H.

LOL!

 

[Steve Wilson]

George Herold <gherold@teachspin.com> wrote:

On Friday, August 30, 2019 at 12:00:10 AM UTC-4, Steve Wilson wrote:
So the concepts of negative resistance and Barkhausen criteria are not
equivalent, and are not two ways of saying the same thing.

Negative resistance in oscillators does not exist.

Hey Steve, is negative resistance the same as having a pole in the right
half plane? Or are those two different things?

I have no idea. I do know that negative resistance has no effect on the
operation of an oscillator.

> I wonder if you've ever built a Wien-bridge oscillator?

No, but I worked on plenty in NATO after WWII as an instrument repair
specialist. They were pretty reliable compared to some of the other stuff.

Of course, you'd never use a Wien bridge these days due to the phase noise
and drift. But if you'd like to try an ultra low distortion audio
oscillator, try this one. Cheap and simple, and it will beat any other
oscillator for total harmonic distortion. All you need is a ganged pot from
an old stereo amplifier marked as R1 and R2 in the schematic.

Version 4
SHEET 1 1204 680
WIRE 784 -128 0 -128
WIRE 928 -128 784 -128
WIRE 1088 -128 928 -128
WIRE 1120 -128 1088 -128
WIRE 0 -96 0 -128
WIRE 976 -96 848 -96
WIRE 992 -96 976 -96
WIRE 848 -64 848 -96
WIRE 784 -48 784 -128
WIRE 816 -48 784 -48
WIRE 144 -32 80 -32
WIRE 928 -32 928 -128
WIRE 928 -32 880 -32
WIRE 240 -16 208 -16
WIRE 272 -16 240 -16
WIRE 448 -16 272 -16
WIRE 512 -16 448 -16
WIRE 624 -16 592 -16
WIRE 800 -16 624 -16
WIRE 816 -16 800 -16
WIRE 0 0 0 -32
WIRE 112 0 0 0
WIRE 144 0 112 0
WIRE 624 0 624 -16
WIRE 0 16 0 0
WIRE 240 16 240 -16
WIRE 848 16 848 0
WIRE 448 32 448 -16
WIRE 848 32 848 16
WIRE 992 32 992 -96
WIRE 624 80 624 64
WIRE 0 112 0 96
WIRE 80 112 80 -32
WIRE 176 112 80 112
WIRE 240 112 240 96
WIRE 240 112 176 112
WIRE 240 128 240 112
WIRE 848 128 848 112
WIRE 992 128 992 112
WIRE 448 144 448 96
WIRE 544 144 448 144
WIRE 576 144 544 144
WIRE 624 144 576 144
WIRE 544 160 544 144
WIRE 624 176 624 144
WIRE 240 224 240 208
WIRE 432 224 240 224
WIRE 448 224 432 224
WIRE 448 240 448 224
WIRE 544 240 544 224
WIRE 240 256 240 224
WIRE 544 304 496 304
WIRE 624 304 624 256
WIRE 624 304 544 304
WIRE 624 320 624 304
WIRE 240 352 240 336
WIRE 448 352 448 336
WIRE 624 416 624 400
FLAG 624 80 0
FLAG 0 112 0
FLAG 992 128 0
FLAG 848 128 0
FLAG 976 -96 +12
FLAG 176 -48 +12
FLAG 848 16 -12
FLAG 176 16 -12
FLAG 112 0 U1P
FLAG 272 -16 U1O
FLAG 176 112 U1N
FLAG 800 -16 U2P
FLAG 1088 -128 VOut
FLAG 240 352 0
FLAG 448 352 0
FLAG 432 224 J1D
FLAG 544 304 J1G
FLAG 624 416 0
FLAG 544 240 0
FLAG 576 144 C3R6
SYMBOL res 608 -32 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R1
SYMATTR Value {RT}
SYMBOL res 16 112 R180
WINDOW 0 36 76 Left 2
WINDOW 3 36 40 Left 2
SYMATTR InstName R2
SYMATTR Value {RT}
SYMBOL cap 608 0 R0
SYMATTR InstName C1
SYMATTR Value 1.5n
SYMBOL cap -16 -96 R0
SYMATTR InstName C2
SYMATTR Value 1.5n
SYMBOL res 224 0 R0
SYMATTR InstName R3
SYMATTR Value 12k
SYMBOL res 224 112 R0
SYMATTR InstName R4
SYMATTR Value 11K
SYMBOL voltage 848 16 R0
WINDOW 123 0 0 Left 2
WINDOW 39 0 0 Left 2
SYMATTR InstName V1
SYMATTR Value -12v
SYMBOL voltage 992 16 R0
WINDOW 123 0 0 Left 2
WINDOW 39 0 0 Left 2
SYMATTR InstName V2
SYMATTR Value +12v
SYMBOL opamps\\lt1001a 176 -80 R0
WINDOW 3 -32 -18 Left 2
SYMATTR InstName U1
SYMBOL opamps\\lt1001a 848 -96 R0
SYMATTR InstName U2
SYMBOL res 224 240 R0
SYMATTR InstName R5
SYMATTR Value 1K
SYMBOL njf 496 240 M0
WINDOW 0 -11 31 Left 2
WINDOW 3 -29 93 Left 2
SYMATTR InstName J1
SYMATTR Value 2N4416
SYMBOL diode 432 96 M180
WINDOW 0 24 64 Left 2
WINDOW 3 24 0 Left 2
SYMATTR InstName D1
SYMATTR Value 1N4148
SYMBOL res 608 160 R0
SYMATTR InstName R6
SYMATTR Value 470k
SYMBOL res 608 304 R0
SYMATTR InstName R7
SYMATTR Value 470k
SYMBOL cap 528 160 R0
SYMATTR InstName C3
SYMATTR Value 1u
TEXT 296 -304 Left 2 !.tran 0 {ST} 0 {MT} startup
TEXT 376 -336 Left 2 ;'Ultra Low Distortion Audio Oscillator
TEXT 832 -256 Left 2 !.param RT = 2e4
TEXT 832 -304 Left 2 !.param ST = RT / 1.5e6
TEXT 832 -280 Left 2 !.param MT = ST / 10000
TEXT 288 -184 Left 2 ;Original Version:
http://www.qsl.net/zl1an/Downloads/Twostage.pdf


> George h.

 

[Steve Wilson]

jlarkin@highlandsniptechnology.com wrote:

On Fri, 30 Aug 2019 10:44:00 GMT, Steve Wilson <no@spam.com> wrote:

jlarkin@highlandsniptechnology.com wrote:

Jeroen's sim, run in time domain, clearly shows the negative
resistance that an emitter follower can present.

Jeroen's sym does not reflect reality. How do you control the amplitude
if an oscillator is running on negative resistance.

Any stable oscillator, with any gain mechanism, has to have a
nonlinear mechanism to set the oscillation amplitude.

Not true.

In a cc Colpitts, you set the amplitude by changing the current through the
emitter follower. This changes the energy delivered to the tank so it is
equal to the loss in the tank. This is a perfectly linear operation.

However, this presupposes the oscillator already meets the Barkhausen
criteria, which states the loop gain must be equal to or greater than 1, and
the phase shift must be zero or multiples of 360 degrees.

So the oscillator is not running on negative resistance. It is running on
Barkhausen. It will oscillate regardless of negative resistance, so negative
resistance really does not exist.

You can think of
a negative resistance as being nonlinear on sinewave amplitude, or you
can imagine it to be a theoretically perfect negative resistor in
parallel with a tank, in parallel with some sort of hard or soft
clipper that adds lossy shunt resistance as oscillation amplitude
increases.

False. You do not want to add any lossy shunt resistance to the tank. This
reduces the Q and increases the phase noise.

There is no clipping in a properly designed oscillator. Clipping increases
the noise, which is critical in precision oscillators. You also have to be
able to set the amplitude correctly for crystal oscillators. If it is too
low, the crystal may not start. If it is too high, the crystal may fracture.

Something has to servo the gain to exactly 1 at the
operating amplitude. Barkhausen magically assumes the gain is
1.000000.

False. Barkhausen requires the loop gain to be equal to or greater than 1 for
oscillations to start.

You set the loop gain by adjusting the energy into the tank to be equal to
the loss. This sets the loop gain to be exactly 1.

Bruce Griffiths and Ulrich Rohde have described feedback methods to reduce
the noise in oscillators. These also happen to set the loop gain to 1, but at
a slightly higher level. The oscillator must already meet the Barkhausen
critera and be oscillating before the feedback can operate.

Your statements prove that oscillators do not run on negative resistance, and
it has no effect on a running oscillator.

Therefore, it does not exist.

 

[Steve Wilson]

jlarkin@highlandsniptechnology.com wrote:

Many oscillators inherently rail, for example common CMOS XOs. There
is no other amplitude limiting mechanism. Nothing wrong with that.

I already mentioned that regarding Pierce oscillators. You adjust the
oscillator amplitude by changing the resistor between the CMOS output and the
input to the tank.

Many Pierce crystal oscillators badly overdrive the crystal. Some don't even
have a resistor. I assume they rely on weak drive capability of the CMOS.

 

[Steve Wilson]

Steve Wilson <no@spam.com> wrote:

Bruce Griffiths and Ulrich Rohde have described feedback methods to
reduce the noise in oscillators. These also happen to set the loop gain
to 1, but at a slightly higher level. The oscillator must already meet
the Barkhausen critera and be oscillating before the feedback can
operate.

I forgot to mention - Bruce Griffiths method is shown in file # 07.asc in
Oscillators.zip

Ulrich Rohde's method was too simple to be much challenge for my fast start
method.

The Pierce crystal oscillator is shown in files # 08.asc and # 09.asc

https://drive.google.com/open?id=1ZsbpkV0aaKS5LURIb1dfu_ndshsSaYtf

 

On Sat, 31 Aug 2019 00:03:59 GMT, Steve Wilson <no@spam.com> wrote:

jlarkin@highlandsniptechnology.com wrote:

On Fri, 30 Aug 2019 10:44:00 GMT, Steve Wilson <no@spam.com> wrote:

jlarkin@highlandsniptechnology.com wrote:

Jeroen's sim, run in time domain, clearly shows the negative
resistance that an emitter follower can present.

Jeroen's sym does not reflect reality. How do you control the amplitude
if an oscillator is running on negative resistance.

Any stable oscillator, with any gain mechanism, has to have a
nonlinear mechanism to set the oscillation amplitude.

Not true.

In a cc Colpitts, you set the amplitude by changing the current through the
emitter follower. This changes the energy delivered to the tank so it is
equal to the loss in the tank. This is a perfectly linear operation.

What determines the amplitude of the voltage in the tank?

If the gain element delivers constant output power regardless of
input, it's nonlinear.

If it's linear, how does it know exactly how much gain to have to keep
the amplitude constant?