Maker Pro
Maker Pro

Yo! RF dudes!

M

Mike Monett

Jan 1, 1970
0
I've used them before. As transmission lines, they have very low
impedances, ballpark 10 ohms, so to make an instantly-triggerable
oscillator you wind up needing a lot of supply current. And they're
expensive in moderate quantities and tricky to solder down. The lowest
they go is about 600 MHz, and that's physically huge, so you're
talking ECL drivers and prescalers. The whole product uses only 2.5
watts.
An LC oscillator is ideal for what I'm doing. If I could get some 18
pF N1000 caps, the osc TC would be about zero. As is, I spent most of
the day taking temperature data and came up with a polynomial
temperature curve I can use to tweak the varicap bias... all the
hardware is already there... temperature sensor, tweak dac, varicap
stuff. So I guess we'll go with a software fix.
The other thing that would help would be a zero TC inductor!

I don't dispute the cost, but that is probably insignificant compared to
the selling price and to the time you spend fixing a problem that you don't
know if it will recur.

Physically huge is a problem. Perhaps the helical versions would help solve
that.

Tricky to solder is a process problem. You have experts that can figure
that out.

I'm not sure about the lower frequency limit. Seems I saw some that went to
300MHz, but I'll have to check. Another divider stage would solve that, and
give lower jitter at 50MHz.

I don't think low impedance is an issue. The 50MHz Colpitts you use now has
an inductive reactance of 50 ohms.

The ceramic resonator acts as a high-Q inductor, so it should behave
exactly the same, except at a higher frequency. The high Q gives less phase
noise (jitter), and the frequency stability means you don't need to add
much varicap control, which helps maintain the low noise.

The divider might be a problem. ECL would definitely burn your power
budget. Is there any reason it has to be so low, such as portable battery-
powered operation?

There should be a way around that problem. I have a simple battery-powered
frequency counter from Radio Shack that goes up to 2GHz. I haven't checked
the power drain, but it goes a long time on a set of alkalines.

Regards,

Mike Monett
 
M

Mike Monett

Jan 1, 1970
0
John,

Come to think of it, the higher Q with a ceramic means the oscillator would
run at lower power, which would help offset the cost of the divider. It
would also make it easier to start, so the start logic would need less
power.

The LTspice circuit I posted could use a simple CMOS logic gate as the VCC
source. I uderstand they now have risetimes around 500ps, which would
provide a very precise starting point for the oscillation. This would be
needed to take advantage of the low noise that comes with the higher Q.

Regards,

Mike Monett
 
J

John Larkin

Jan 1, 1970
0
John,

Come to think of it, the higher Q with a ceramic means the oscillator would
run at lower power, which would help offset the cost of the divider. It
would also make it easier to start, so the start logic would need less
power.

The low impedance makes it need a lot of idle current. The best way to
instant-start an lc (or coaxial resonator) oscillator at full
amplitude is to run quiescent current into the L while it's off, and
drop the current (quickly! like in 1/4 cycle, preferably less) to kick
off the oscillation. That's the "initial condition". The only other
initial condition that works would be zero inductor current and full
capacitor voltage, but that implies a switch between the L and the C,
which makes gobs of problems on its own.

If Il = 0 and Vc = 0 when off, it will take a long time to build up
oscillation amplitude. I need it to start instantly.

Anyway, it works now with the computed temperature compensation.

John
 
M

Mike Monett

Jan 1, 1970
0
The low impedance makes it need a lot of idle current. The best way to
instant-start an lc (or coaxial resonator) oscillator at full
amplitude is to run quiescent current into the L while it's off, and
drop the current (quickly! like in 1/4 cycle, preferably less) to kick
off the oscillation. That's the "initial condition". The only other
initial condition that works would be zero inductor current and full
capacitor voltage, but that implies a switch between the L and the C,
which makes gobs of problems on its own.
If Il = 0 and Vc = 0 when off, it will take a long time to build up
oscillation amplitude. I need it to start instantly.
Anyway, it works now with the computed temperature compensation.

John

John, check the LTspice circuit I posted. A great deal depends on the
loaded Q of the tank. That raises the impedance at resonance. And that
determines the quiescent current you have to use.

The circuit I posted improves the loaded Q by using a current source in the
emitter instead of a resistor.

It passes (2.5-0.6) / 3.3e3 = 0.000575A through the inductor when off. That
hardly qualifies as gobs of current.

It starts instantly. The first transition is positive, about 3.6V. The
amplitude after settling is 3.9V. So it reaches 3.6/3.9 = 92% of full
amplitude on the first half cycle.

There is a small "bump" in amplitude just after starting, but the zero
crossings are not affected. The phase error during settling is smaller than
I can measure in LTspice. That is measured in the second ASC file I posted,
using two identical oscillators, with one starting after the other to
compare the phase difference.

Since the coaxial resonator has 10X or higher Q, the tank impedance at
resonance should be higher than obtainable with the 1206 smd inductor.

This means the power in the circuit is much lower, which means the tank
current is lower. So it shouldn't need as much current when it is off, and
it should outperform the 1206 smd version in every aspect.

The circuit I posted was just proof-of-concept. I didn't do any
optimization before posting, especially in choice of transistors. I just
grabbed whatever was available in LTspice.

So the performance could be much better than described here.

But if you are happy with the current 50MHz oscillator the way it is,
that's all that counts.

Regards,

Mike Monett
 
J

John Larkin

Jan 1, 1970
0
John, check the LTspice circuit I posted. A great deal depends on the
loaded Q of the tank. That raises the impedance at resonance. And that
determines the quiescent current you have to use.

The circuit I posted improves the loaded Q by using a current source in the
emitter instead of a resistor.

It passes (2.5-0.6) / 3.3e3 = 0.000575A through the inductor when off. That
hardly qualifies as gobs of current.

It starts instantly. The first transition is positive, about 3.6V. The
amplitude after settling is 3.9V. So it reaches 3.6/3.9 = 92% of full
amplitude on the first half cycle.

There is a small "bump" in amplitude just after starting, but the zero
crossings are not affected. The phase error during settling is smaller than
I can measure in LTspice. That is measured in the second ASC file I posted,
using two identical oscillators, with one starting after the other to
compare the phase difference.

Since the coaxial resonator has 10X or higher Q, the tank impedance at
resonance should be higher than obtainable with the 1206 smd inductor.

This means the power in the circuit is much lower, which means the tank
current is lower. So it shouldn't need as much current when it is off, and
it should outperform the 1206 smd version in every aspect.

The circuit I posted was just proof-of-concept. I didn't do any
optimization before posting, especially in choice of transistors. I just
grabbed whatever was available in LTspice.

So the performance could be much better than described here.

But if you are happy with the current 50MHz oscillator the way it is,
that's all that counts.

Regards,

Mike Monett

I can't seem to view your LT spice file. Do you have a pic somewhere?

John
 
P

Phil Hobbs

Jan 1, 1970
0
John said:
I've used them before. As transmission lines, they have very low
impedances, ballpark 10 ohms, so to make an instantly-triggerable
oscillator you wind up needing a lot of supply current. And they're
expensive in moderate quantities and tricky to solder down. The lowest
they go is about 600 MHz, and that's physically huge, so you're
talking ECL drivers and prescalers. The whole product uses only 2.5
watts.

An LC oscillator is ideal for what I'm doing. If I could get some 18
pF N1000 caps, the osc TC would be about zero. As is, I spent most of
the day taking temperature data and came up with a polynomial
temperature curve I can use to tweak the varicap bias... all the
hardware is already there... temperature sensor, tweak dac, varicap
stuff. So I guess we'll go with a software fix.

The other thing that would help would be a zero TC inductor!

John

You can make zero TC inductors. Air core inductor TCs come from
dimensional changes due to thermal expansion, but the axial and radial
expansions fight each other. From the usual simple formula for
single-layer coils,

A**2 N**2 A=mean radius in inches,
L(uH) = ----------- B=length in inches
9A+10B

you can compute dL/dT in terms of the CTEs of A and B. If you choose
the ratio A/B correctly, and wind the coil on a mechanically weak
plastic mount (so that A has the CTE of copper and B has the CTE of
plastic), you ought to be able to make whatever TC you want, at least in
the +- 100 ppm range.

Cheers,

Phil Hobbs
 
J

John Larkin

Jan 1, 1970
0
You can make zero TC inductors. Air core inductor TCs come from
dimensional changes due to thermal expansion, but the axial and radial
expansions fight each other. From the usual simple formula for
single-layer coils,

A**2 N**2 A=mean radius in inches,
L(uH) = ----------- B=length in inches
9A+10B

you can compute dL/dT in terms of the CTEs of A and B. If you choose
the ratio A/B correctly, and wind the coil on a mechanically weak
plastic mount (so that A has the CTE of copper and B has the CTE of
plastic), you ought to be able to make whatever TC you want, at least in
the +- 100 ppm range.

Cheers,

Phil Hobbs

The danger being that the plastic will cold-flow from the winding
tension, and the frequency will creep over time. Lots of Coilcraft
parts do that. Baking helps.

John
 
M

Mike Monett

Jan 1, 1970
0
John Larkin said:
I can't seem to view your LT spice file. Do you have a pic somewhere?

Is there a problem with the ASC file? Would you like me to archive and
upload it to abse?

I could put a brief description with schematic and waveforms on my web
site, but that might take some time. I have a very heavy schedule this
weekend.

Also, I goofed in calculating the amplitude of the first positive peak. I
forgot to subtract the 2.5V reference voltage, so the 92% figure is wrong.
That is very easy to fix by changing the quiescent current. You can easily
set the amplitude of the first peak to pretty much whatever you want.

I posted the actual concept here a long time ago. The base-collector
junction of the transistor is forward biased when the oscillator is off.
The current from a resistor betwen the collector and ground flows through
the b-c junction and through the tank inductor.

Raising the collector voltage to VCC turns the transistor on and dumps the
stored charge in the base-collector junction into the tank. The inductor
back EMF forces the tank voltage to go positive at the same time the stored
charge from the junction is forcing the tank positive. These two increase
the energy of the first transition and reduce the quiescent current needed
when the oscillator is turned off.

To turn the oscillator off, the VCC is removed and the base-collector
juntion becomes forward-biased, applying the resistance across the tank.
The transistor is turned off, so there is no power added and the tank
energy dissipates through the collector resistance.

If you hit it at just the right point in the cycle, the energy dissipates
very quickly, perhaps in 1/2 cycle. But other stopping points may add
energy to the tank causing it to ring for 1/2 dozen cycles or so.

If you still would like to see the operation, please let me know how best
to proceed.

Regards,

Mike Monett
 

Similar threads

Top