opamp sine wave oscillator

[John Devereux]

[...]

Well, then maybe the circuit is not for you. You know, I am really
quite surprised at all the negativity in these responses. I figured
out something interesting, I wrote it up, and I posted it on the web
for you to enjoy. I'm not asking for anything from anyone other than
a plain, simple peer review and I've gotten some of that, thank you
very much. The circuit might prove very useful, it might not. I had
fun with the process of discovery so it was a pleasant, positive
experience for me. Why don't we all play nice for a change?
Thanks!

I thought your article itself was quite interesting, nicely laid out
too. I think some people are having a problem with your perceived
claim to "inventing" something new. These circuit arrangements are no
doubt obvous and well known for decades, to those here more
experienced. And for the stated goal of a sine wave oscillator there
are better solutions.

Well, I can't really be certain, but I kind of thought there was
something new here, yes. I thought that stuff about using the inverse
dual voltage divider on opposite sides of the circuit was new, the
fact that it's a whole family of oscillators not just one was new, and
features like the amplitude stabilization and amplitude modulation
were new as well. I think people are reminded of familiar circuits by
the new circuit, and then they reply: "that's been done before!", but
at this point I'm still thinking there is novelty here. If it's an
old familiar circuit, would someone please post a reference, perhaps
an existing web link?

Well you could be right - I am not an expert - but some of the people
commenting *are*.

I don't see that your amplitude stabilisation can work as
advertised. You initially claim to replace the "light bulb"
stabilisation by putting part of the network in the negative feedback
path. But all you are doing is rearranging the network, the overall
effect is the same as far as I can see. You are still left with a
system that requires a slight positive gain to startup reliably, which
will inevitably lead to clipping in the steady state condition.

Any feedback mechanism that operates at the level of a single cycle
will create distortion or require impossibly accurate matching. That
is why the classic "light bulb" is used, it has a relatively long
thermal time constant so as to *gradually* adjust the amplitude over
hundreds of cycles.

And in fact I see you show this with your test results.

 

[Inventor]

[...]
Well, then maybe the circuit is not for you. You know, I am really
quite surprised at all the negativity in these responses. I figured
out something interesting, I wrote it up, and I posted it on the web
for you to enjoy. I'm not asking for anything from anyone other than
a plain, simple peer review and I've gotten some of that, thank you
very much. The circuit might prove very useful, it might not. I had
fun with the process of discovery so it was a pleasant, positive
experience for me. Why don't we all play nice for a change?
Thanks!
I thought your article itself was quite interesting, nicely laid out
too. I think some people are having a problem with your perceived
claim to "inventing" something new. These circuit arrangements are no
doubt obvous and well known for decades, to those here more
experienced. And for the stated goal of a sine wave oscillator there
are better solutions.

Well, I can't really be certain, but I kind of thought there was
something new here, yes. I thought that stuff about using the inverse
dual voltage divider on opposite sides of the circuit was new, the
fact that it's a whole family of oscillators not just one was new, and
features like the amplitude stabilization and amplitude modulation
were new as well. I think people are reminded of familiar circuits by
the new circuit, and then they reply: "that's been done before!", but
at this point I'm still thinking there is novelty here. If it's an
old familiar circuit, would someone please post a reference, perhaps
an existing web link?

Well you could be right - I am not an expert - but some of the people
commenting *are*.

I don't see that your amplitude stabilisation can work as
advertised. You initially claim to replace the "light bulb"
stabilisation by putting part of the network in the negative feedback
path. But all you are doing is rearranging the network, the overall
effect is the same as far as I can see. You are still left with a
system that requires a slight positive gain to startup reliably, which
will inevitably lead to clipping in the steady state condition.

Any feedback mechanism that operates at the level of a single cycle
will create distortion or require impossibly accurate matching. That
is why the classic "light bulb" is used, it has a relatively long
thermal time constant so as to *gradually* adjust the amplitude over
hundreds of cycles.

And in fact I see you show this with your test results.

Yes, that is an interesting question. I also do not quite see how the
amplitude is stabilized by this circuit. I think it is simply
clipping at the peaks. There's not all that much I can do with that
old scope in the photo, but one thing I did was I looked at the peak
on all three waveforms: Vo, V-, and V+. I noticed that during the
distortion V- was not equal to V+, so as expected the circuit is not
in the linear mode during the peak distortion interval. Then, sure
enough, after the peak distortion interval, the circuit transitions
into the linear mode where V- = V+.

Another observation about this behavior is the distortion that occurs
while amplitude modulating. In this case, the distortion appears at
only one of the peaks, while the other peak is left smooth (and it is
more pronounced). I am thinking that this is because under amplitude
modulation, the amplitude clipping occurs on only one side of the
waveform.

But really, I think it will take one of those experts to explain it
properly.

 

[MooseFET]

Your circuit has way too many parts.


John

I bow to your greatness

 

[MooseFET]

As opposed to not enough parts?

Try it. It doesn't wack against the rails and it does oscillate quite
nicely.

 

[MooseFET]

The "alternatives" have more parts for reasons. If you just let an RC
oscillator build up amplitude until it clips, you get a lot of
distortion. If you clip just a tiny bit, it's marginal to oscillate at
all.

Hewlett's lightbulb was a nonlinear element with a long gain time
constant that rolled off the gain to get stable amplitude with low
distortion.

Today an optically varied resistor or a JFET can used in place of the
lightbulb.

An LED shining on a chunk on some strange sort of material that varies
its resistance over the period of seconds works very well but it takes
a long time to settle.

If you want to go with the JFET, you need one with a low gain so that
the change in drain voltage doesn't effect the resistance much.

 

[MooseFET]

[...]

Well, then maybe the circuit is not for you. You know, I am really
quite surprised at all the negativity in these responses. I figured
out something interesting, I wrote it up, and I posted it on the web
for you to enjoy. I'm not asking for anything from anyone other than
a plain, simple peer review and I've gotten some of that, thank you
very much. The circuit might prove very useful, it might not. I had
fun with the process of discovery so it was a pleasant, positive
experience for me. Why don't we all play nice for a change?
Thanks!

I thought your article itself was quite interesting, nicely laid out
too. I think some people are having a problem with your perceived
claim to "inventing" something new. These circuit arrangements are no
doubt obvous and well known for decades, to those here more
experienced. And for the stated goal of a sine wave oscillator there
are better solutions.

There is one interesting thing: A whole tread has gone nearly to its
end without a "use a PIC". To make a sine wave the micro from Cygnal
would be a better way to go. The total parts count not counting power
supply stuff comes out to be (drum roll) one.

 

[Fred Bloggs]

I read the PDF. I fail to observe any invention. Please describe the
improvement.

...Jim Thompson

Phase sensitivity wrt normalized resonant frequency is 3.5x that of the
conventional Wien, making it much more frequency stable, but that is
stable at whatever frequency it settles on...

 

[John Larkin]

I bow to your greatness

Or nod to my weirdness.

John

 

[John Popelish]

Try it. It doesn't wack against the rails and it does oscillate quite
nicely.

Does the definition of "nicely" include "producing a low
distortion sine wave"?

 

[Phil Hobbs]

What a bunch of guys. What's the THD of those ones? More than a bit of
clipping, I bet.

Today an optically varied resistor or a JFET can used in place of the
lightbulb.

An LED shining on a chunk on some strange sort of material that varies
its resistance over the period of seconds works very well but it takes
a long time to settle.

That's usually cad sulphide (CdS), cad selenide (CdSe), or an alloy of
the two. They're very sensitive, adequately slow for lots of things,
but unfortunately their photoresponse depends on their previous
history...would you believe 5:1 change in resistance for CdSe based on
how recently it's been in room light? For oscillators that tends to
turn into lots and lots of amplitude drift unless you're careful.

If you want to go with the JFET, you need one with a low gain so that
the change in drain voltage doesn't effect the resistance much.

And you need to linearize it really carefully---Jim Williams's chapter
on ALC oscillators is a classic. (It's in one of his books, but I don't
have it handy.) Residual nonlinearity in the JFET (even with the 1:1
voltage divider between gate and drain) turns out to be one of the big
problems, iirc.

Cheers,

Phil Hobbs

 

[John Larkin]

What a bunch of guys. What's the THD of those ones? More than a bit of
clipping, I bet.

That's usually cad sulphide (CdS), cad selenide (CdSe), or an alloy of
the two. They're very sensitive, adequately slow for lots of things,
but unfortunately their photoresponse depends on their previous
history...would you believe 5:1 change in resistance for CdSe based on
how recently it's been in room light? For oscillators that tends to
turn into lots and lots of amplitude drift unless you're careful.


And you need to linearize it really carefully---Jim Williams's chapter
on ALC oscillators is a classic. (It's in one of his books, but I don't
have it handy.) Residual nonlinearity in the JFET (even with the 1:1
voltage divider between gate and drain) turns out to be one of the big
problems, iirc.

Cheers,

Phil Hobbs

An analog multiplier is the obvious choice.

John

 

[Jim Thompson]

An analog multiplier is the obvious choice.

John

If the "Q" of the effective "tank" is high enough, all you need to do
is limit energy input.

Not the best example in the world, since it has ~1% distortion...

http://analog-innovations.com/SED/Sedra-Espinosa-Oscillator.pdf

But I have some AGC'd versions around here someplace that do several
orders of magnitude better.

I'll post when I find where I filed them away ;-)

...Jim Thompson

 

[Winfield Hill]

John Larkin wrote:

Today an optically varied resistor or a JFET can used in place of the
lightbulb. [snip]
If you want to go with the JFET, you need one with a low gain so that
the change in drain voltage doesn't effect the resistance much.

And you need to linearize it really carefully---Jim Williams's chapter
on ALC oscillators is a classic. (It's in one of his books, but I don't
have it handy.) Residual nonlinearity in the JFET (even with the 1:1
voltage divider between gate and drain) turns out to be one of the big
problems, iirc.

We also discussed it well in our book. WRT
an exact factor of two, 1:1 ratio, that can
be viewed as advantage, having a distortion-
adjustment point. Our Wien-bridge oscillator
design (for AoE 3rd ed) uses 1M and 976k+50k
pot for distortion trimming, 1:1 +/-2.5%, and
along with a gain trimpot (to minimize the
JFET's work), achieves 2ppm distortion. We
used an inexpensive Fairchild 2n5458 JFET.

 

[Tim Shoppa]

Well, there are a few.  For one, it is a single-opamp oscillator.  It
does not have multiple opamps, transistor feedback stages, light
bulbs, or other such stuff to make it work.  For another, it is
amplitude stabilized by the supply rail, which is nice.  Plus it can
operate as an amplitude modulator.  It is also a class of circuits,
not just a single circuit, based on some design principles.  Those are
some of the improvements

Your amplitude stabilization consists of setting the loop gain to just
a little bigger than unity and hitting the rails.

Changes in the rail voltages will show up as both amplitude changes
and frequency changes in the output.

Sounds like a very real step backwards from the Wien bridge with a
light bulb (or whatever).

Tim.

 

[Vladimir Vassilevsky]

There is one interesting thing: A whole tread has gone nearly to its
end without a "use a PIC".

OK, let's go digital. Make the HC86 gate as invertor and connect the
output to the input with the single RC lowpass


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

 

[Joerg]

Our new 2-opamp "AoE oscillator" (a modified
Wien bridge made with two inverting opamps and
G=2) has about 0.0002% (2ppm) distortion, IIRC.

Well, now you brought it upon yourself: When is that new edition going
to hit the store shelves?

 

[vincent.thiernesse]

OK, let's go digital. Make the HC86 gate as invertor and connect the
output to the input with the single RC lowpass

For the same purpose (sine wave), I have another use of HC86

http://cjoint.com/?cfwLDUyuMn

 

[MooseFET]

Does the definition of "nicely" include "producing a low
distortion sine wave"?

Not in this case unless we are speaking of very high values of "low".

If you use the inverting pin of the op-amp as the output, the 3rd
harmonic is down by about 20dB.

 

[MooseFET]

What a bunch of guys. What's the THD of those ones? More than a bit of
clipping, I bet.

Mine doesn't clip at all when run on normal supply voltages. It makes
a sawtoothy waveform that doesn't get tot he rails.


[...]

That's usually cad sulphide (CdS), cad selenide (CdSe), or an alloy of
the two. They're very sensitive, adequately slow for lots of things,
but unfortunately their photoresponse depends on their previous
history...would you believe 5:1 change in resistance for CdSe based on
how recently it's been in room light?

The ones I was speaking of are in a black housing. The no-light
resistance depends on what happened last week but the full light
resistance only really depends on the last minute or so.

For oscillators that tends to
turn into lots and lots of amplitude drift unless you're careful.

You just tell the user to leave it on for a week before use. A
nonlinear feedback gets it settled more quickly. For large errors,
you turn the LED to full power or off.


[....]

And you need to linearize it really carefully---Jim Williams's chapter
on ALC oscillators is a classic. (It's in one of his books, but I don't
have it handy.) Residual nonlinearity in the JFET (even with the 1:1
voltage divider between gate and drain) turns out to be one of the big
problems, iirc.

Yes, it adds distortion for even fairly modest swings. "deleveraging"
the JFET with capacitors and not resistors like:


!
=== C1
!
!------+
Control-->! !
!-- === C2
! !
GND GND

helps.

It divides down the amplitude the JFET sees and C2 also tends to short
out the harmonics.

 

[Robert Latest]

Well, now you brought it upon yourself: When is that new edition going
to hit the store shelves?

That's not even my main concern. I can wait. What I'm worried about is: How
is the community going to get their hands on what's finished of AoE3 when
Win dies of old age 20 or 30 years from now?

robert