K7ITM wrote:
On Mar 19, 9:23 am, "Anthony Fremont" wrote:
The waveform in a high Q tank that's lightly coupled to the amplifier
should be very nearly sinusoidal. If in addition, the amplifier
remains linear and represents a constant impedance over the whole
cycle of the waveform, then the waveforms should everywhere be
sinusoidal. If the amplifier+tank has barely enough loop gain to
sustain oscillation, then clipping will be minimal, but it's also
possible to detect the level and control the gain of the amplifier.
You could, for example, use a light bulb like HP did in their original
audio oscillator. Beware, though, that best oscillator performance in
other regards may not be achieved the same way you achieve lowest
harmonic distortion. Be careful that you optimize the right things
for your application.

After reading the other replies, it seems aparent that the shape of the
signal from the first stage is not that critical, it is stability and phase
noise that are most important. So, I should put things back where there is
clipping to be sure that the oscillator oscillates and then clean up the
signal in subsequent stages.

In the work I do, I need to measure distortion, and the generators I
use don't have low enough distortion in their outputs to be directly
useful. The distortion levels in the "raw" outputs are generally
about -40 to -50dBc. I use filters to make things better, and can get
to -140dBc distortion levels fairly easily. If it's low harmonic
distortion you want, I'd suggest that it may be better to just put a
filter on the output of the oscillator that has only moderately low
harmonic output, and not worry so much about that aspect of oscillator
performance. Filters work well when the oscillator frequency range is
about 1.5:1 or less. Much more than that and you'd need to switch in
different filters depending on the oscillator frequency.

Thanks. :-)

"Anthony Fremont" wrote in message
...
K7ITM wrote:
On Mar 19, 9:23 am, "Anthony Fremont" wrote:
The waveform in a high Q tank that's lightly coupled to the amplifier
should be very nearly sinusoidal. If in addition, the amplifier
remains linear and represents a constant impedance over the whole
cycle of the waveform, then the waveforms should everywhere be
sinusoidal. If the amplifier+tank has barely enough loop gain to
sustain oscillation, then clipping will be minimal, but it's also
possible to detect the level and control the gain of the amplifier.
You could, for example, use a light bulb like HP did in their original
audio oscillator. Beware, though, that best oscillator performance in
other regards may not be achieved the same way you achieve lowest
harmonic distortion. Be careful that you optimize the right things
for your application.

After reading the other replies, it seems aparent that the shape of the
signal from the first stage is not that critical, it is stability and
phase noise that are most important. So, I should put things back where
there is clipping to be sure that the oscillator oscillates and then clean
up the signal in subsequent stages.

I have never seen clipping. These things are supposed to limit in cutoff,
not saturation. As the signal build up, the conduction angle gets smaller
and smaller until the device runs out of gain. That is another way of saying
that the DC value of the gate voltage gets more negative the bigger the
amplitude. This works out automatically with a JFET. You need about 10K -
100K DC resistance from gate to ground. Using a bipolar transistor is not a
good idea.

Tam

Tam/WB2TT wrote:
"Anthony Fremont" wrote in message

I have never seen clipping. These things are supposed to limit in
cutoff, not saturation. As the signal build up, the conduction angle
gets smaller and smaller until the device runs out of gain. That is
another way of saying that the DC value of the gate voltage gets more
negative the bigger the amplitude. This works out automatically with
a JFET. You need about 10K - 100K DC resistance from gate to ground.
Using a bipolar transistor is not a good idea.

I was wondering about the load that a bipolar would present. I will see if
I can find a JFET in my junk pile, thank you. :-)

On Mon, 19 Mar 2007 16:25:26 -0500, "Anthony Fremont"
wrote:

Tam/WB2TT wrote:
"Anthony Fremont" wrote in message

I have never seen clipping. These things are supposed to limit in
cutoff, not saturation. As the signal build up, the conduction angle
gets smaller and smaller until the device runs out of gain. That is
another way of saying that the DC value of the gate voltage gets more
negative the bigger the amplitude. This works out automatically with
a JFET. You need about 10K - 100K DC resistance from gate to ground.
Using a bipolar transistor is not a good idea.

I was wondering about the load that a bipolar would present. I will see if
I can find a JFET in my junk pile, thank you. :-)

Without going to the purity levels that Tom requires, I've always found that
bipolars can be used to produce a fairly reasonable "visibly sinusoidal" (see
note) waveform. Follow the oscillator with an amplifier stage which drives a
limiter/clipper, and use that to control a gain element in the oscillator. It's
like the incandescent non-linearity arrangement except the oscillator stage
waveform remains fairly clean.

(Note: Harmonic distortion not readily discernible on a CRO)

Tam/WB2TT wrote:

I have never seen clipping. These things are supposed to limit in
cutoff, not saturation. As the signal build up, the conduction angle
gets smaller and smaller until the device runs out of gain. That is
another way of saying that the DC value of the gate voltage gets more
negative the bigger the amplitude. This works out automatically with
a JFET. You need about 10K - 100K DC resistance from gate to ground.
Using a bipolar transistor is not a good idea.

I have now changed it to an MPF102 that I've had laying around for many
years. It works great, thanks.
:-)

"Anthony Fremont" wrote in message
...
Tam/WB2TT wrote:

I have never seen clipping. These things are supposed to limit in
cutoff, not saturation. As the signal build up, the conduction angle
gets smaller and smaller until the device runs out of gain. That is
another way of saying that the DC value of the gate voltage gets more
negative the bigger the amplitude. This works out automatically with
a JFET. You need about 10K - 100K DC resistance from gate to ground.
Using a bipolar transistor is not a good idea.

I have now changed it to an MPF102 that I've had laying around for many
years. It works great, thanks.
:-)

Glad it worked out. By the way the feedback path is through the capacitive
network between source and gate. That would be more obvious in the
configuration that uses a tapped inductor, but works the same way. Leaving
out the diode was the right thing to do; it just adds to the noise. I don't
know what kind of stability and linearity you need, but if that is
important, do not use the common type of ceramic capacitors that are meant
for bypassing. They are lossy, and their value varies with applied voltage.
Use mica, NPO ceramic, or Mylar and similar for larger values.

Tam

Tam/WB2TT wrote:
"Anthony Fremont" wrote in message

I have now changed it to an MPF102 that I've had laying around for
many years. It works great, thanks.
:-)

Glad it worked out. By the way the feedback path is through the
capacitive network between source and gate. That would be more
obvious in the configuration that uses a tapped inductor, but works

I had to play around with this quite a bit to get it running near 10MHz. It
seemed to be more picky about the cap ratios than their actual values.

the same way. Leaving out the diode was the right thing to do; it
just adds to the noise. I don't know what kind of stability and

I figured that it would cause horrid clipping the way it was installed. I
thought maybe the designer intended it to be the other way around to protect
the transistor from reverse voltages on the gate, hmm.... always something
to keep you guessing. ;-)

linearity you need, but if that is important, do not use the common
type of ceramic capacitors that are meant for bypassing. They are
lossy, and their value varies with applied voltage. Use mica, NPO
ceramic, or Mylar and similar for larger values.

I don't really "need" anything in particular, it's just an exercise to try
and learn something. I just used the ceramic caps because they were handy.
I didn't have a pile of NPO caps laying around. ;-) I d

Time to add a buffer now since the scope probes load it down so badly now.
I'm getting 1.1Vpp into the apparent 5M load of two scope probes. It's
probably pulled a ton off frequency as well. Let's see.....(removes one
probe).....yep, 90KHz rise in frequency after taking one probe off.