Design Article
Comment
pcsalex
R4 =33 ohm Figure 2, is that the correct value? text: "The lower speed PNP ...
anonymous user
Series-LC-tank VCO breaks tuning-range records
Louis Vlemincq, Belgacom, Evere, Belgium; Edited by Paul Rako and Fran Granville
10/20/2011 10:00 AM EDT
This Design Idea applies a novel
topology to an oscillator. It uses
a series-connected LC (inductive-capacitive)
tank circuit to give the circuit
a higher tuning range than circuits
that use a parallel-LC connection. The
architecture of the oscillator permits
wide frequency swings, well beyond the
capabilities of the best hyperabrupt
varactor. Engineers deem a VCO (voltage-controlled oscillator) capable
of covering one octave as state of the
art. This topology allows a 4-to-1 ratio
in output frequency. The LC tank alone
sets this frequency so that the parasitic
capacitances of other components do
not limit the output frequency. Unlike
standard oscillators, this circuit works
well at its frequency extremes.
At first glance, the central structure
of the oscillator resembles two transistors
that form a latching SCR (silicon-controlled-rectifier) structure (Figure
1). The structure is similar to that of
a thyristor, but you add degeneration
resistors that keep the circuit in a linear
mode of operation. The resistors make
the gain of this “SCR” smaller than one,
and it is dc-stable. The series-tuned
tank circuit increases the gain beyond
one at the resonant frequency, causing
the circuit to oscillate. No auxiliary components are necessary for oscillation,
and the node between the inductor
and the capacitor is free of other
connections, meaning that only the
varactor you use as the capacitor determines
the tuning range. The frequency
varies as the square root of the tuning
elements. To change the frequency by a
factor of two, you need a fourfold variation
of the tuning capacitance.
Unlike a parallel-LC tank, the resonant current passes through the active element and is, therefore, limited. This limit in turn means that the ac voltage appearing across the tuning components is small—typically, less than 100 mV. The small signal reduces the effects of circuit nonlinearity and the impact of the self-biasing effects of the signal on the varactor. You can use control voltages as small as 0.3V across the varactor. If you use a 1-μH inductor, the circuit still oscillates with capacitor values of 4.7 pF to 4.7 μF—a ratio of 106-to-1.
For the detailed design, move the
LC tank to the emitter of PNP transistor
Q2 (Figure 2). The lower speed
of the PNP creates greater phase difference
and encourages oscillation.
Connect L2 and C2 at a common power
point on the power rail, emphasizing the criticality of the layout in this part
of the circuit. The oscillator “senses”
the tuned circuit through C2 and C4,
and anything inside that loop adds
uncontrolled parasitics to L2. These
parasitics would compromise the AGC
(automatic-gain-control) action and
degrade the performance and accuracy
of the oscillator.
Q1 and associated components
implement the AGC. A parallel-LC
oscillator tolerates clipping of the signal,
but this series-LC circuit degenerates
into a multivibrator if you allow
the signal to grow so large that it clips.
The AGC servo action has the added
advantage of producing uniform output
amplitude. Use D5 to create a 0.6V dc
bias. R11 and R12 form a voltage ladder
that creates a dc-bias voltage close to
the forward-voltage drop of Schottky
diode D6. This bias allows D6 to work
as a more perfect rectifier of the small
output signal. C8 integrates the rectified
signal into a dc voltage proportional
to the amplitude of the circuit’s
output. Apply this dc signal to IC1,
the AGC amplifier, through a filter
comprising R15 and C8. 
The op amp
servo-controls the filtered dc signal
against the A-CTRL input-amplitude
signal you send to the circuit. This signal
allows you to set output amplitude
at 0 to 1V.
In this example, the output amplitude is 0.9V. The frequency range extends from 35 to 140 MHz, a 1-to-4 ratio—twice that of conventional high-performance VCOs—and requires a fourfold increase in the capacitance ratio. The overall capacitance ratio is 1-to-16, exactly that of the varactor itself. At the lowest (Figure 3) and highest (Figure 4) frequencies of the output range, the quality of the sine wave remains excellent, thanks to AGC action.
At first glance, the central structure
of the oscillator resembles two transistors
that form a latching SCR (silicon-controlled-rectifier) structure (Figure
1). The structure is similar to that of
a thyristor, but you add degeneration
resistors that keep the circuit in a linear
mode of operation. The resistors make
the gain of this “SCR” smaller than one,
and it is dc-stable. The series-tuned
tank circuit increases the gain beyond
one at the resonant frequency, causing
the circuit to oscillate. No auxiliary components are necessary for oscillation,
and the node between the inductor
and the capacitor is free of other
connections, meaning that only the
varactor you use as the capacitor determines
the tuning range. The frequency
varies as the square root of the tuning
elements. To change the frequency by a
factor of two, you need a fourfold variation
of the tuning capacitance.Unlike a parallel-LC tank, the resonant current passes through the active element and is, therefore, limited. This limit in turn means that the ac voltage appearing across the tuning components is small—typically, less than 100 mV. The small signal reduces the effects of circuit nonlinearity and the impact of the self-biasing effects of the signal on the varactor. You can use control voltages as small as 0.3V across the varactor. If you use a 1-μH inductor, the circuit still oscillates with capacitor values of 4.7 pF to 4.7 μF—a ratio of 106-to-1.
For the detailed design, move the
LC tank to the emitter of PNP transistor
Q2 (Figure 2). The lower speed
of the PNP creates greater phase difference
and encourages oscillation.
Connect L2 and C2 at a common power
point on the power rail, emphasizing the criticality of the layout in this part
of the circuit. The oscillator “senses”
the tuned circuit through C2 and C4,
and anything inside that loop adds
uncontrolled parasitics to L2. These
parasitics would compromise the AGC
(automatic-gain-control) action and
degrade the performance and accuracy
of the oscillator.
Q1 and associated components
implement the AGC. A parallel-LC
oscillator tolerates clipping of the signal,
but this series-LC circuit degenerates
into a multivibrator if you allow
the signal to grow so large that it clips.
The AGC servo action has the added
advantage of producing uniform output
amplitude. Use D5 to create a 0.6V dc
bias. R11 and R12 form a voltage ladder
that creates a dc-bias voltage close to
the forward-voltage drop of Schottky
diode D6. This bias allows D6 to work
as a more perfect rectifier of the small
output signal. C8 integrates the rectified
signal into a dc voltage proportional
to the amplitude of the circuit’s
output. Apply this dc signal to IC1,
the AGC amplifier, through a filter
comprising R15 and C8. The op amp
servo-controls the filtered dc signal
against the A-CTRL input-amplitude
signal you send to the circuit. This signal
allows you to set output amplitude
at 0 to 1V.In this example, the output amplitude is 0.9V. The frequency range extends from 35 to 140 MHz, a 1-to-4 ratio—twice that of conventional high-performance VCOs—and requires a fourfold increase in the capacitance ratio. The overall capacitance ratio is 1-to-16, exactly that of the varactor itself. At the lowest (Figure 3) and highest (Figure 4) frequencies of the output range, the quality of the sine wave remains excellent, thanks to AGC action.
|
|
|
|
|
Navigate to related information
WKetel
10/22/2011 10:07 PM EDT
This is a vey interesting design idea. And the included description of the operation is very good. Is thecircuit patented yet? It could become the basis for an inteestng all band receiver.
Sign in to Reply
bcarso
10/24/2011 1:42 PM EDT
Nice work. It would be helpful to see some frequency-domain plots as it's difficult to estimate spectral content from scope photos. One would suspect a fair amount of 2nd, at least, particularly at the lowest varactor bias voltages where the signal represents a larger fraction of the bias.
Having said that, it would be intriguing as well to produce a "balanced" version with intrinsic suppression of even-order harmonics.
Sign in to Reply
anonymous user
10/24/2011 3:44 PM EDT
These days, one might just subtract a frequency offset to create an arbitrarily large ratio. Or use digital techniques. But this is a great (!) circuit concept for when such typically power hungry digital techniques cannot be used.
Sign in to Reply
LostInSpace2010
10/24/2011 6:03 PM EDT
Looking at the scope pictures, there doesn't seem to be much 2nd or 3rd order distortion there (to answer another question). A nice rule of thumb is that when you can just start to see distortion on the scope - it's about 40 dBc (a rule of thumb, subject to squinting, your years of experience, etc, etc, etc.... So don't flame me for sharing... Ha, ha, ha, ha... But I find it about right)
Sign in to Reply
devassocx
10/24/2011 8:02 PM EDT
It looks to me that this circuit is dependent on
transistor leakage currents to even start...not exactly
a reliable approach.
Sign in to Reply
anonymous user
10/25/2011 6:33 AM EDT
>>This is a vey interesting design idea. And the included description of the operation is very good. Is thecircuit patented yet?
The circuit is NOT patented, you are free to use it.
>>One would suspect a fair amount of 2nd, at least, particularly at the lowest varactor bias voltages where the signal represents a larger fraction of the bias.
Remember that with series operation, the current through the reactive elements is constant, but their impedance vary with frequency, and hence, for a constant output voltage, the voltage across the varactor is not constant.
For example at the max. output amplitude, that rms voltage is ~0.1V (0.3Vpp)for 0.3V DC bias, against ~0.4V (1.1Vpp) for 15VDC.
There is therefore a partial compensation.
>>It looks to me that this circuit is dependent on
transistor leakage currents to even start...not exactly
a reliable approach.
It is a perfectly valid point, but if you examine the circuit more closely, you will see that R10 combined with D1 and D2 address this issue in a deterministic way.
Thanks for all positive comments
Sign in to Reply
anonymous user
10/25/2011 12:07 PM EDT
You state in the 1st circuit that the capacitor is 4.7pf to 4.7pf should that be 47 or 470pf
Sign in to Reply
bcarso
10/25/2011 12:21 PM EDT
Louis, thanks for your explication of the partial compensation. Have you examined the output on a spectrum analyzer yet? I agree with the one poster about being able to see of order 1% residuals, I'm just curious as to what the overall performance is.
Sign in to Reply
anonymous user
10/26/2011 3:03 AM EDT
What was the application that prompted this design?
Sign in to Reply
MITRONICS
10/26/2011 3:03 AM EDT
What was the application that prompted this design?
Sign in to Reply
anonymous user
10/26/2011 5:31 AM EDT
>>You state in the 1st circuit that the capacitor is 4.7pf to 4.7pf should that be 47 or 470pf
It is 4.7pf to 4.7µF (microfarad): maybe the symbol doesn't show correctly on your screen.
There is also a small typo in the text, the capacitance ratio is not 106 to 1, but 10 to the sixth power to 1.
>>Louis, thanks for your explication of the partial compensation. Have you examined the output on a spectrum analyzer yet? I agree with the one poster about being able to see of order 1% residuals, I'm just curious as to what the overall performance is.
Under the worst conditions (max. amplitude, min. tuning voltage), the 2nd harmonic is 37.5dB down wrt the fondamental, 3rd is ~50dB down.
A screenshot of the spectrum is visible at: www.cijoint.fr/cj201110/cijkBiQxnB.png
>>What was the application that prompted this design?
It was the swept generator for a prototype cable discontinuity analyser.
A frequency ratio >3 was required, and this was a direct, uncomplicated solution.
Sign in to Reply
bcarso
10/26/2011 2:09 PM EDT
Thanks! Those two numbers are plenty. The link you provided leads to an interesting website but lacking fluency in French I wasn't able to navigate to your photos. But those are very good results for 2nd and 3rd, especially as worst-case and given the large tuning range and circuit simplicity.
The "mu" character did reproduce for me at least --- I suspect the poster just couldn't believe that you were talking about microfarads.
Sign in to Reply
anonymous user
10/26/2011 3:20 PM EDT
>>The link you provided leads to an interesting website but lacking fluency in French I wasn't able to navigate to your photos.
Sorry about that, normally this site is the French equivalent of imageshack, etc, and no navigation should be necessary.
Here it is again, under imgur this time.
If it doesn't work, I'll keep trying other hosting sites!
(you have to add the usual h(tee)(tee)p plus double slash in front, as this is blocked in this comment)
i.imgur.com/cMeZt.png
Sign in to Reply
bcarso
10/26/2011 4:02 PM EDT
Ah that one on imgur comes right up for me --- very pretty!
Sign in to Reply
WKetel
10/29/2011 11:21 AM EDT
This is indeed quite a nice design, and the fact that it is amplitude-stable should indeed allow operation at a level much less likely to produce harmonics. Of course, the tuned-circuit "Q" will vary with frequency, and so there is a potential for a variation in distortion based on that. Adding a tuned buffer, controlled by the same tuning voltage, would probably provide the desired reduction in harmonics.
Sign in to Reply
anonymous user
11/2/2011 12:47 PM EDT
Just for the record, Edwards has been using a circuit
nearly identical to that of fig.1 in their film
thickness monitors since the late 70's
Sign in to Reply
pcsalex
11/24/2012 10:32 PM EST
R4 =33 ohm Figure 2, is that the correct value? text: "The lower speed PNP creates greater phase difference and encourages oscillation" AAccording the data sheets the PNP BF450 has fT =350MHz, the NPN BF240 has fT = 150MHz so the NPN is the slower...
Sign in to Reply