Low noise quartz crystal oscillator by Bruce Griffiths
Hi,
I've been trying to read up on low noise crystal oscillators and had
a closer look at the design by Bruce Griffiths[1]. There are explanations
to how the circuit works, but I have some questions on the details.
I would appreciate if someone could answer these questions.
[1] http://www.ko4bb.com/~bruce/CrystalOscillators.html
I will do a short recap how the circuit works, just to make sure I
haven't misunderstood it.
The oscillator core is the colpitts oscillator build around Q104,
C107/108 form the driving/feedback path to form a negative resistance
over the quartz crystal. The resistors R112 and R113 are there only
to keep the crystal bias free and prevent charges from building up.
The output is formed using the crystal as filter to get rid of
harmonics and noise outside the crystal bandwidth. The "ground" point
of the crystal is formed using the low input impedance of the common
base amplifier formed by Q102. The output is coupled using a transformer
to make it DC free and for impedance transformation.
Q103, LED102 and R116 form a constant current source for the collector
of Q104, using the base of Q104 as control input.
Q105 acts as a series voltage regulator, using multiple LM329's as
reference, which are averaged for lower noise and Q106 to compensate
for Q105's B-E voltage drop.
Q101 is the input power supply filter.
Now my questions:
Doesn't the non-zero input impedance of Q102 dampen the
crystal unnecessarily?
Why use a colpitts oscillator when using the crystal as output filter?
Wouldn't a Butler oscillator make more sense? Or is there some
disadvantage of Butler oscillator that I am not aware of?
Why are LEDs used as voltage references? Don't they have a horrible
temperature coefficient and bad aging characteristics?
My guess would be that LED101 is not that critical as it will only
result in a slight change of the collector current and thus only
a slight change in the input impedance common base amplifier Q102.
Does the constant current source (Q103, LED 102, R116) sufficiently
stabilize the power inside the crystal, and thus the output power?
My guess would be that changes in h_fe of Q104 will result in
different biasing of Q104 and thus in changes of the power within the
crystal, which then affects frequency and aging.
Can the noise induced by Q103 be further decreased by increasing C109?
Or is there a reason why C109 is just 10nF? Stability maybe?
If stability is the problem, how about using an RC low pass filter?
If one would want to make this circuit tunable, where would the
varicap get connected to? My guess would be on the right side of
the crystal, between the crystal and C105, going to ground
The bias voltage would be then applied directly at the crystal/C105/varicap node. Is this correct or is there a better way?
What are the criteria to choose the transistors?
Thanks in advance
Attila Kinali
--
Reading can seriously damage your ignorance.
-- unknown
Hi
On Oct 26, 2015, at 7:03 PM, Attila Kinali attila@kinali.ch wrote:
Hi,
I've been trying to read up on low noise crystal oscillators and had
a closer look at the design by Bruce Griffiths[1]. There are explanations
to how the circuit works, but I have some questions on the details.
I would appreciate if someone could answer these questions.
[1] http://www.ko4bb.com/~bruce/CrystalOscillators.html
I will do a short recap how the circuit works, just to make sure I
haven't misunderstood it.
The oscillator core is the colpitts oscillator build around Q104,
C107/108 form the driving/feedback path to form a negative resistance
over the quartz crystal. The resistors R112 and R113 are there only
to keep the crystal bias free and prevent charges from building up.
The output is formed using the crystal as filter to get rid of
harmonics and noise outside the crystal bandwidth. The "ground" point
of the crystal is formed using the low input impedance of the common
base amplifier formed by Q102. The output is coupled using a transformer
to make it DC free and for impedance transformation.
Q103, LED102 and R116 form a constant current source for the collector
of Q104, using the base of Q104 as control input.
Q105 acts as a series voltage regulator, using multiple LM329's as
reference, which are averaged for lower noise and Q106 to compensate
for Q105's B-E voltage drop.
Q101 is the input power supply filter.
Now my questions:
Doesn't the non-zero input impedance of Q102 dampen the
crystal unnecessarily?
Yes
Why use a colpitts oscillator when using the crystal as output filter?
Wouldn't a Butler oscillator make more sense?
No
Or is there some
disadvantage of Butler oscillator that I am not aware of?
Build it and see...
Why are LEDs used as voltage references? Don't they have a horrible
temperature coefficient and bad aging characteristics?
but the “right” tempco
My guess would be that LED101 is not that critical as it will only
result in a slight change of the collector current and thus only
a slight change in the input impedance common base amplifier Q102.
Does the constant current source (Q103, LED 102, R116) sufficiently
stabilize the power inside the crystal, and thus the output power?
No
My guess would be that changes in h_fe of Q104 will result in
different biasing of Q104 and thus in changes of the power within the
crystal, which then affects frequency and aging.
Except that Hfe is pretty stable over time
Can the noise induced by Q103 be further decreased by increasing C109?
Not really
Or is there a reason why C109 is just 10nF?
Look at the loop stability of the current source and the R/C between the 10K and C109
Stability maybe?
If stability is the problem, how about using an RC low pass filter?
If one would want to make this circuit tunable, where would the
varicap get connected to? My guess would be on the right side of
the crystal, between the crystal and C105, going to ground
In series with C105, with a blocking cap to the crystal and buffer sides.
The bias voltage would be then applied directly at the crystal/C105/varicap node. Is this correct or is there a better way?
Block the DC
What are the criteria to choose the transistors?
Low noise, more specifically the combination of low 1/F noise and low noise figure at RF and a low mixing
coefficient.
There are also a number of “tweaks” (added components) typically put into that circuit for best performance.
Bob
Thanks in advance
Attila Kinali
--
Reading can seriously damage your ignorance.
-- unknown
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
On Tuesday, October 27, 2015 12:03:49 AM Attila Kinali wrote:
Hi,
I've been trying to read up on low noise crystal oscillators and had
a closer look at the design by Bruce Griffiths[1]. There are explanations
to how the circuit works, but I have some questions on the details.
I would appreciate if someone could answer these questions.
[1] http://www.ko4bb.com/~bruce/CrystalOscillators.html
I will do a short recap how the circuit works, just to make sure I
haven't misunderstood it.
The oscillator core is the colpitts oscillator build around Q104,
C107/108 form the driving/feedback path to form a negative resistance
over the quartz crystal. The resistors R112 and R113 are there only
to keep the crystal bias free and prevent charges from building up.
The output is formed using the crystal as filter to get rid of
harmonics and noise outside the crystal bandwidth. The "ground" point
of the crystal is formed using the low input impedance of the common
base amplifier formed by Q102. The output is coupled using a
transformer
to make it DC free and for impedance transformation.
Q103, LED102 and R116 form a constant current source for the collector
of Q104, using the base of Q104 as control input.
Q105 acts as a series voltage regulator, using multiple LM329's as
reference, which are averaged for lower noise and Q106 to compensate
for Q105's B-E voltage drop.
Q101 is the input power supply filter.
Now my questions:
Doesn't the non-zero input impedance of Q102 dampen the
crystal unnecessarily?
The effect is relatively insignificant provided the crystal esr is significantly
larger than the CB stage input R.
With an overtone crystal this is readily achieved.
Why use a colpitts oscillator when using the crystal as output filter?
Wouldn't a Butler oscillator make more sense? Or is there some
disadvantage of Butler oscillator that I am not aware of?
Avoiding doubling resistance in series with the crystal due to the 2
transistors of the butler configuration.
Why are LEDs used as voltage references? Don't they have a horrible
temperature coefficient and bad aging characteristics?
My guess would be that LED101 is not that critical as it will only
result in a slight change of the collector current and thus only
a slight change in the input impedance common base amplifier Q102.
In both cases the LED forward voltage tempco is approximately matched
by the Vbe tempco of a transistor so that the resultant dc current is
nominally temperature independent.
LEDs have relatively low noise however they are somewhat photosensitive.
Using low noise dc bias circuits like these can significantly reduce the
close in phase noise of RF amplifiers significantly compared to a bias
circuit using a voltage divider from the power supply.
Does the constant current source (Q103, LED 102, R116) sufficiently
stabilize the power inside the crystal, and thus the output power?
My guess would be that changes in h_fe of Q104 will result in
different biasing of Q104 and thus in changes of the power within the
crystal, which then affects frequency and aging.
The colpitts oscillator transistor in this circuit operates in a discontinuous
mode.
Can the noise induced by Q103 be further decreased by increasing
C109?
Or is there a reason why C109 is just 10nF? Stability maybe?
If stability is the problem, how about using an RC low pass filter?
the noise contribution by Q103 isnt significant.
Yes bias loop stability is an issue you cant just insert arbitrary low pass RC
filters some design effort is required.
If one would want to make this circuit tunable, where would the
varicap get connected to? My guess would be on the right side of
the crystal, between the crystal and C105, going to ground
The bias voltage would be then applied directly at the
crystal/C105/varicap
node. Is this correct or is there a better way?
In series with C105 is a far better location.
You may then need to increase the value of C105.
What are the criteria to choose the transistors?
Low flicker noise and sufficient RF gain at the crystal frequency.
Thanks in advance
Attila Kinali
Bruce
Various versions of this oscillator circuit have been employed as high stability OCXOs eg:
http://ri.search.yahoo.com/_ylt=AwrTcaxMzS5WgJIAMwk3QIpQ;_ylu=X3oDMTBzbW1zYXBzBHNlYwNzcgRwb3MDMjEEY29sbwNncTEEdnRpZAM-/RV=2/RE=1445936589/RO=10/RU=https%3a%2f%2fescies.org%2fdownload%2fwebDocumentFile%3fid%3d60902/RK=0/RS=.Rmksavr9Ui3TZ8D1XyZ06TpeDY-
An AGC circuit can be employed to adjust the dc current of the oscillator transistor to stabilise the crystal current.
The circuit as given was merely intended to show an alternative to the corresponding Wenzel version which employs a high input impedance buffer. The Wenzel version has a relatively low oscillator transistor Vcb which is perhaps somewhat undesirable.
Driscoll developed various high frequency crystal oscillators employing MMICs RF splitters together with a crystal, various matching circuits and a diode limiter.
Bruce
On Tuesday, 27 October 2015 2:01 PM, Bruce Griffiths <bruce.griffiths@xtra.co.nz> wrote:
On Tuesday, October 27, 2015 12:03:49 AM Attila Kinali wrote:
Hi,
I've been trying to read up on low noise crystal oscillators and had
a closer look at the design by Bruce Griffiths[1]. There are explanations
to how the circuit works, but I have some questions on the details.
I would appreciate if someone could answer these questions.
[1] http://www.ko4bb.com/~bruce/CrystalOscillators.html
I will do a short recap how the circuit works, just to make sure I
haven't misunderstood it.
The oscillator core is the colpitts oscillator build around Q104,
C107/108 form the driving/feedback path to form a negative resistance
over the quartz crystal. The resistors R112 and R113 are there only
to keep the crystal bias free and prevent charges from building up.
The output is formed using the crystal as filter to get rid of
harmonics and noise outside the crystal bandwidth. The "ground" point
of the crystal is formed using the low input impedance of the common
base amplifier formed by Q102. The output is coupled using a
transformer
to make it DC free and for impedance transformation.
Q103, LED102 and R116 form a constant current source for the collector
of Q104, using the base of Q104 as control input.
Q105 acts as a series voltage regulator, using multiple LM329's as
reference, which are averaged for lower noise and Q106 to compensate
for Q105's B-E voltage drop.
Q101 is the input power supply filter.
Now my questions:
Doesn't the non-zero input impedance of Q102 dampen the
crystal unnecessarily?
The effect is relatively insignificant provided the crystal esr is significantly
larger than the CB stage input R.
With an overtone crystal this is readily achieved.
Why use a colpitts oscillator when using the crystal as output filter?
Wouldn't a Butler oscillator make more sense? Or is there some
disadvantage of Butler oscillator that I am not aware of?
Avoiding doubling resistance in series with the crystal due to the 2
transistors of the butler configuration.
Why are LEDs used as voltage references? Don't they have a horrible
temperature coefficient and bad aging characteristics?
My guess would be that LED101 is not that critical as it will only
result in a slight change of the collector current and thus only
a slight change in the input impedance common base amplifier Q102.
In both cases the LED forward voltage tempco is approximately matched
by the Vbe tempco of a transistor so that the resultant dc current is
nominally temperature independent.
LEDs have relatively low noise however they are somewhat photosensitive.
Using low noise dc bias circuits like these can significantly reduce the
close in phase noise of RF amplifiers significantly compared to a bias
circuit using a voltage divider from the power supply.
Does the constant current source (Q103, LED 102, R116) sufficiently
stabilize the power inside the crystal, and thus the output power?
My guess would be that changes in h_fe of Q104 will result in
different biasing of Q104 and thus in changes of the power within the
crystal, which then affects frequency and aging.
The colpitts oscillator transistor in this circuit operates in a discontinuous
mode.
Can the noise induced by Q103 be further decreased by increasing
C109?
Or is there a reason why C109 is just 10nF? Stability maybe?
If stability is the problem, how about using an RC low pass filter?
the noise contribution by Q103 isnt significant.
Yes bias loop stability is an issue you cant just insert arbitrary low pass RC
filters some design effort is required.
If one would want to make this circuit tunable, where would the
varicap get connected to? My guess would be on the right side of
the crystal, between the crystal and C105, going to ground
The bias voltage would be then applied directly at the
crystal/C105/varicap
node. Is this correct or is there a better way?
In series with C105 is a far better location.
You may then need to increase the value of C105.
What are the criteria to choose the transistors?
Low flicker noise and sufficient RF gain at the crystal frequency.
Thanks in advance
Attila Kinali
Bruce
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
The oscillator transistor and buffer amplifier are basically
the same as the HP 10811, except for the absence of mode
suppressors. The difference here is that the oscillator
self limits in the oscillator transistor, whereas the 10811
has ALC. The discontinuous operation of the transistor,
as explained by Driscoll some 45 years ago, is undesirable
because it increases the load resistance the crystal sees.
The 2 transistor "Driscoll oscillator" fixes this problem
by using an additional stage that limits instead of the
oscillator transistor. This has been widely used for
decades. It is interesting to note that the 10811 ALC
works by varying the DC bias current in the oscillator
transistor. This is in contrast to the elaborate DC
bias current stabilization here.
I have demonstrated that the close in phase noise in
the 10811 is entirely due to the flicker noise of the
crystal. The only place where the 10811 circuit comes
into play is beyond 1 kHz from the carrier, where the
Burgoon patent circuit (which apparently has prior art
from Ulrich Rhode) reduces the phase noise floor. I
have built two different oscillator circuits for 10811
crystals and have measured the flicker noise as being
the same as the intrinsic noise of the crystal.
Thus, obsessing over noise in oscillators circuits may
be overkill, unless you are planning to use a much
better crystal (BVA, etc). OTOH, it might be advantageous
to improve the reverse isolation by adding additional
grounded base buffer stages. There are various NBS/NIST
papers where several grounded base stages are cascaded.
I did this in the HP 10816 rubidium standard.
It is good to see time-nuts learning about oscillator
circuit by building them.
Rick Karlquist N6RK
The 10811A ocxo uses an oscillator of this type albeit with a lower crystal current, an overtone crystal. However the output stages spoil the PN floor..Cascaded transformer coupled CB stages are somewhat quieter.
Bruce
On Tuesday, 27 October 2015 2:31 PM, Bruce Griffiths <bruce.griffiths@xtra.co.nz> wrote:
Various versions of this oscillator circuit have been employed as high stability OCXOs eg:
http://ri.search.yahoo.com/_ylt=AwrTcaxMzS5WgJIAMwk3QIpQ;_ylu=X3oDMTBzbW1zYXBzBHNlYwNzcgRwb3MDMjEEY29sbwNncTEEdnRpZAM-/RV=2/RE=1445936589/RO=10/RU=https%3a%2f%2fescies.org%2fdownload%2fwebDocumentFile%3fid%3d60902/RK=0/RS=.Rmksavr9Ui3TZ8D1XyZ06TpeDY-
An AGC circuit can be employed to adjust the dc current of the oscillator transistor to stabilise the crystal current.
The circuit as given was merely intended to show an alternative to the corresponding Wenzel version which employs a high input impedance buffer. The Wenzel version has a relatively low oscillator transistor Vcb which is perhaps somewhat undesirable.
Driscoll developed various high frequency crystal oscillators employing MMICs RF splitters together with a crystal, various matching circuits and a diode limiter.
Bruce
On Tuesday, 27 October 2015 2:01 PM, Bruce Griffiths <bruce.griffiths@xtra.co.nz> wrote:
On Tuesday, October 27, 2015 12:03:49 AM Attila Kinali wrote:
Hi,
I've been trying to read up on low noise crystal oscillators and had
a closer look at the design by Bruce Griffiths[1]. There are explanations
to how the circuit works, but I have some questions on the details.
I would appreciate if someone could answer these questions.
[1] http://www.ko4bb.com/~bruce/CrystalOscillators.html
I will do a short recap how the circuit works, just to make sure I
haven't misunderstood it.
The oscillator core is the colpitts oscillator build around Q104,
C107/108 form the driving/feedback path to form a negative resistance
over the quartz crystal. The resistors R112 and R113 are there only
to keep the crystal bias free and prevent charges from building up.
The output is formed using the crystal as filter to get rid of
harmonics and noise outside the crystal bandwidth. The "ground" point
of the crystal is formed using the low input impedance of the common
base amplifier formed by Q102. The output is coupled using a
transformer
to make it DC free and for impedance transformation.
Q103, LED102 and R116 form a constant current source for the collector
of Q104, using the base of Q104 as control input.
Q105 acts as a series voltage regulator, using multiple LM329's as
reference, which are averaged for lower noise and Q106 to compensate
for Q105's B-E voltage drop.
Q101 is the input power supply filter.
Now my questions:
Doesn't the non-zero input impedance of Q102 dampen the
crystal unnecessarily?
The effect is relatively insignificant provided the crystal esr is significantly
larger than the CB stage input R.
With an overtone crystal this is readily achieved.
Why use a colpitts oscillator when using the crystal as output filter?
Wouldn't a Butler oscillator make more sense? Or is there some
disadvantage of Butler oscillator that I am not aware of?
Avoiding doubling resistance in series with the crystal due to the 2
transistors of the butler configuration.
Why are LEDs used as voltage references? Don't they have a horrible
temperature coefficient and bad aging characteristics?
My guess would be that LED101 is not that critical as it will only
result in a slight change of the collector current and thus only
a slight change in the input impedance common base amplifier Q102.
In both cases the LED forward voltage tempco is approximately matched
by the Vbe tempco of a transistor so that the resultant dc current is
nominally temperature independent.
LEDs have relatively low noise however they are somewhat photosensitive.
Using low noise dc bias circuits like these can significantly reduce the
close in phase noise of RF amplifiers significantly compared to a bias
circuit using a voltage divider from the power supply.
Does the constant current source (Q103, LED 102, R116) sufficiently
stabilize the power inside the crystal, and thus the output power?
My guess would be that changes in h_fe of Q104 will result in
different biasing of Q104 and thus in changes of the power within the
crystal, which then affects frequency and aging.
The colpitts oscillator transistor in this circuit operates in a discontinuous
mode.
Can the noise induced by Q103 be further decreased by increasing
C109?
Or is there a reason why C109 is just 10nF? Stability maybe?
If stability is the problem, how about using an RC low pass filter?
the noise contribution by Q103 isnt significant.
Yes bias loop stability is an issue you cant just insert arbitrary low pass RC
filters some design effort is required.
If one would want to make this circuit tunable, where would the
varicap get connected to? My guess would be on the right side of
the crystal, between the crystal and C105, going to ground
The bias voltage would be then applied directly at the
crystal/C105/varicap
node. Is this correct or is there a better way?
In series with C105 is a far better location.
You may then need to increase the value of C105.
What are the criteria to choose the transistors?
Low flicker noise and sufficient RF gain at the crystal frequency.
Thanks in advance
Attila Kinali
Bruce
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
As Rick has pointed out numerous times when the output signal is extracted via the crystal by a CB stage (or cascade thereof) the PN floor is determined by the ratio of the amplifier equivalent input noise current to the crystal current. That is the amplifier equivalent input noise current at frequencies for which the crystal impedance is high. If one neglects this crucial point one comes to the conclusion (e.g. see Eq 4.-1 page 274 of Ulrich Rohde's: Microwave and Wireless Synthesisers Theory and Design.) that with a crystal current of 1.4mA rms and a crystal esr of 50 ohms that the XO PN floor cannot be lower than -154dBc/Hz. Even the XO circuit in the ARRL handbook (attributed to Ulrich) using this method of signal extraction has a measured PN floor of -168dBc/Hz. Many other XO's (including the 10811A which uses a crystal current of 1mA ) have an actual PN significantly lower than this. One would have thought that this glaring discrepancy between "theory" and practice would have been noticed and corrected by now.
Bruce
On Tuesday, 27 October 2015 6:01 PM, Richard (Rick) Karlquist <richard@karlquist.com> wrote:
The oscillator transistor and buffer amplifier are basically
the same as the HP 10811, except for the absence of mode
suppressors. The difference here is that the oscillator
self limits in the oscillator transistor, whereas the 10811
has ALC. The discontinuous operation of the transistor,
as explained by Driscoll some 45 years ago, is undesirable
because it increases the load resistance the crystal sees.
The 2 transistor "Driscoll oscillator" fixes this problem
by using an additional stage that limits instead of the
oscillator transistor. This has been widely used for
decades. It is interesting to note that the 10811 ALC
works by varying the DC bias current in the oscillator
transistor. This is in contrast to the elaborate DC
bias current stabilization here.
I have demonstrated that the close in phase noise in
the 10811 is entirely due to the flicker noise of the
crystal. The only place where the 10811 circuit comes
into play is beyond 1 kHz from the carrier, where the
Burgoon patent circuit (which apparently has prior art
from Ulrich Rhode) reduces the phase noise floor. I
have built two different oscillator circuits for 10811
crystals and have measured the flicker noise as being
the same as the intrinsic noise of the crystal.
Thus, obsessing over noise in oscillators circuits may
be overkill, unless you are planning to use a much
better crystal (BVA, etc). OTOH, it might be advantageous
to improve the reverse isolation by adding additional
grounded base buffer stages. There are various NBS/NIST
papers where several grounded base stages are cascaded.
I did this in the HP 10816 rubidium standard.
It is good to see time-nuts learning about oscillator
circuit by building them.
Rick Karlquist N6RK
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hi
On Oct 27, 2015, at 12:15 AM, Bruce Griffiths bruce.griffiths@xtra.co.nz wrote:
The 10811A ocxo uses an oscillator of this type albeit with a lower crystal current, an overtone crystal. However the output stages spoil the PN floor..Cascaded transformer coupled CB stages are somewhat quieter.
….. and since this is the region that the circuit really is the issue, lowering noise there is well worth doing. The target is
basically signal to noise. That makes it much easier to analyze than some of the non-linear stuff that can impact the close in
noise.
Bob
Bruce
On Tuesday, 27 October 2015 2:31 PM, Bruce Griffiths <bruce.griffiths@xtra.co.nz> wrote:
Various versions of this oscillator circuit have been employed as high stability OCXOs eg:
http://ri.search.yahoo.com/_ylt=AwrTcaxMzS5WgJIAMwk3QIpQ;_ylu=X3oDMTBzbW1zYXBzBHNlYwNzcgRwb3MDMjEEY29sbwNncTEEdnRpZAM-/RV=2/RE=1445936589/RO=10/RU=https%3a%2f%2fescies.org%2fdownload%2fwebDocumentFile%3fid%3d60902/RK=0/RS=.Rmksavr9Ui3TZ8D1XyZ06TpeDY-
An AGC circuit can be employed to adjust the dc current of the oscillator transistor to stabilise the crystal current.
The circuit as given was merely intended to show an alternative to the corresponding Wenzel version which employs a high input impedance buffer. The Wenzel version has a relatively low oscillator transistor Vcb which is perhaps somewhat undesirable.
Driscoll developed various high frequency crystal oscillators employing MMICs RF splitters together with a crystal, various matching circuits and a diode limiter.
Bruce
On Tuesday, 27 October 2015 2:01 PM, Bruce Griffiths <bruce.griffiths@xtra.co.nz> wrote:
On Tuesday, October 27, 2015 12:03:49 AM Attila Kinali wrote:
Hi,
I've been trying to read up on low noise crystal oscillators and had
a closer look at the design by Bruce Griffiths[1]. There are explanations
to how the circuit works, but I have some questions on the details.
I would appreciate if someone could answer these questions.
[1] http://www.ko4bb.com/~bruce/CrystalOscillators.html
I will do a short recap how the circuit works, just to make sure I
haven't misunderstood it.
The oscillator core is the colpitts oscillator build around Q104,
C107/108 form the driving/feedback path to form a negative resistance
over the quartz crystal. The resistors R112 and R113 are there only
to keep the crystal bias free and prevent charges from building up.
The output is formed using the crystal as filter to get rid of
harmonics and noise outside the crystal bandwidth. The "ground" point
of the crystal is formed using the low input impedance of the common
base amplifier formed by Q102. The output is coupled using a
transformer
to make it DC free and for impedance transformation.
Q103, LED102 and R116 form a constant current source for the collector
of Q104, using the base of Q104 as control input.
Q105 acts as a series voltage regulator, using multiple LM329's as
reference, which are averaged for lower noise and Q106 to compensate
for Q105's B-E voltage drop.
Q101 is the input power supply filter.
Now my questions:
Doesn't the non-zero input impedance of Q102 dampen the
crystal unnecessarily?
The effect is relatively insignificant provided the crystal esr is significantly
larger than the CB stage input R.
With an overtone crystal this is readily achieved.
Why use a colpitts oscillator when using the crystal as output filter?
Wouldn't a Butler oscillator make more sense? Or is there some
disadvantage of Butler oscillator that I am not aware of?
Avoiding doubling resistance in series with the crystal due to the 2
transistors of the butler configuration.
Why are LEDs used as voltage references? Don't they have a horrible
temperature coefficient and bad aging characteristics?
My guess would be that LED101 is not that critical as it will only
result in a slight change of the collector current and thus only
a slight change in the input impedance common base amplifier Q102.
In both cases the LED forward voltage tempco is approximately matched
by the Vbe tempco of a transistor so that the resultant dc current is
nominally temperature independent.
LEDs have relatively low noise however they are somewhat photosensitive.
Using low noise dc bias circuits like these can significantly reduce the
close in phase noise of RF amplifiers significantly compared to a bias
circuit using a voltage divider from the power supply.
Does the constant current source (Q103, LED 102, R116) sufficiently
stabilize the power inside the crystal, and thus the output power?
My guess would be that changes in h_fe of Q104 will result in
different biasing of Q104 and thus in changes of the power within the
crystal, which then affects frequency and aging.
The colpitts oscillator transistor in this circuit operates in a discontinuous
mode.
Can the noise induced by Q103 be further decreased by increasing
C109?
Or is there a reason why C109 is just 10nF? Stability maybe?
If stability is the problem, how about using an RC low pass filter?
the noise contribution by Q103 isnt significant.
Yes bias loop stability is an issue you cant just insert arbitrary low pass RC
filters some design effort is required.
If one would want to make this circuit tunable, where would the
varicap get connected to? My guess would be on the right side of
the crystal, between the crystal and C105, going to ground
The bias voltage would be then applied directly at the
crystal/C105/varicap
node. Is this correct or is there a better way?
In series with C105 is a far better location.
You may then need to increase the value of C105.
What are the criteria to choose the transistors?
Low flicker noise and sufficient RF gain at the crystal frequency.
Thanks in advance
Attila Kinali
Bruce
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
On 10/26/2015 9:15 PM, Bruce Griffiths wrote:
The 10811A ocxo uses an oscillator of this type albeit with a lower crystal current, an overtone crystal. However the output stages spoil the PN floor..Cascaded transformer coupled CB stages are somewhat quieter.
Bruce
That's right. Burgoon (10811 designer) told me he had to meet other
requirements besides noise floor. He had a special
one-off version of the 10811 without these compromises
that he built to provide a reference source to use
for phase noise measurements. In the 10816 rubidium,
I used 3 common base transistors in cascade as
the output buffer and got similar results to
his special version.
Driscoll developed various high frequency crystal oscillators employing MMICs RF splitters together with a crystal, various matching circuits and a diode limiter.
Yes, he did, but long before that work he championed
his 2 transistor circuit that was extensively copied
by many other designers.
Rick
Salut,
Thanks everyone for answering my questions and englighting me on
the general topic.
Sorry for the late answer from my side. It took me some time to
read up on all the pointers and hints provided. I'll quickly add
some comments and questions to a few of the mails, that I find
noteworth, while i'm digging deeper into the topic.
On Tue, 27 Oct 2015 01:31:36 +0000 (UTC)
Bruce Griffiths bruce.griffiths@xtra.co.nz wrote:
The correct URL here is https://escies.org/download/webDocumentFile?id=60902
and referencing to the document
"Low Noise Master Oscillator LNMO" by Wagner and Desaules.
On Mon, 26 Oct 2015 20:19:35 -0700
"Richard (Rick) Karlquist" richard@karlquist.com wrote:
The oscillator transistor and buffer amplifier are basically
the same as the HP 10811, except for the absence of mode
suppressors. The difference here is that the oscillator
self limits in the oscillator transistor, whereas the 10811
has ALC. The discontinuous operation of the transistor,
as explained by Driscoll some 45 years ago, is undesirable
because it increases the load resistance the crystal sees.
Driscoll wrote a lot about oscillators over the years.
I couldn't find anything specific to discontinuous operation.
Do you have a titel of a paper related to this?
The 2 transistor "Driscoll oscillator" fixes this problem
by using an additional stage that limits instead of the
oscillator transistor.
A good introduction to Driscolls oscillator design can be found in:
"Notes on the Driscoll VHF Overtone Crystal Oscillator and
New Low-Noise VHF Crystal Oscillator Topology" by Chris Bartram GW4DGU, 2008
page 5 in "Scatterpoint"
http://www.microwavers.org/scatterpoint/2008/Scatterpoint_Apr_2008.pdf
I have demonstrated that the close in phase noise in
the 10811 is entirely due to the flicker noise of the
crystal. The only place where the 10811 circuit comes
into play is beyond 1 kHz from the carrier, where the
Burgoon patent circuit (which apparently has prior art
Is the Burgoon patent you are refering to US4283691
"Crystal oscillator having low noise signal extraction circuit" ?
On Wed, 28 Oct 2015 06:45:54 -0700
"Richard (Rick) Karlquist" richard@karlquist.com wrote:
Bruce has it exactly right. At offset frequencies beyond 1 kHz,
the source impedance for the grounded base is very high due to
the crystal impedance being very high. As Burgoon explains,
this condition suppresses base recombination noise, and the
only noise mechanism that is significant is the collector shot
noise. (To minimize shot noise, don't run more DC collector
current than necessary).
I read Ulrich Rohde's 1977 article showing this circuit,
before I started working at HP in 1979. When I got to HP,
The article in question is either
"Stable Crystal Oscillators", Ham Radio Magazine, June, 1975
or
"Effects of Noise in Receiving System", Ham Radio Magazine, November 1977
Unfortunately, there doesn't seem to be an online source for the
Ham Radio Magazine. If anyone has a digital version of these
articles, I would appreciate a copy.
Attila Kinali
--
Reading can seriously damage your ignorance.
-- unknown
Hi ,
Ham Radio Magazine used to be available on archive.org but has been
removed. 73 Magazine is still available.
CD's of it are available (expensive).
I'll look later and see which issue it was.
Cheers,
Will
On 09/11/15 11:19, Attila Kinali wrote:
Salut,
Thanks everyone for answering my questions and englighting me on
the general topic.
Sorry for the late answer from my side. It took me some time to
read up on all the pointers and hints provided. I'll quickly add
some comments and questions to a few of the mails, that I find
noteworth, while i'm digging deeper into the topic.
On Tue, 27 Oct 2015 01:31:36 +0000 (UTC)
Bruce Griffiths bruce.griffiths@xtra.co.nz wrote:
The correct URL here is https://escies.org/download/webDocumentFile?id=60902
and referencing to the document
"Low Noise Master Oscillator LNMO" by Wagner and Desaules.
On Mon, 26 Oct 2015 20:19:35 -0700
"Richard (Rick) Karlquist" richard@karlquist.com wrote:
The oscillator transistor and buffer amplifier are basically
the same as the HP 10811, except for the absence of mode
suppressors. The difference here is that the oscillator
self limits in the oscillator transistor, whereas the 10811
has ALC. The discontinuous operation of the transistor,
as explained by Driscoll some 45 years ago, is undesirable
because it increases the load resistance the crystal sees.
Driscoll wrote a lot about oscillators over the years.
I couldn't find anything specific to discontinuous operation.
Do you have a titel of a paper related to this?
The 2 transistor "Driscoll oscillator" fixes this problem
by using an additional stage that limits instead of the
oscillator transistor.
A good introduction to Driscolls oscillator design can be found in:
"Notes on the Driscoll VHF Overtone Crystal Oscillator and
New Low-Noise VHF Crystal Oscillator Topology" by Chris Bartram GW4DGU, 2008
page 5 in "Scatterpoint"
http://www.microwavers.org/scatterpoint/2008/Scatterpoint_Apr_2008.pdf
I have demonstrated that the close in phase noise in
the 10811 is entirely due to the flicker noise of the
crystal. The only place where the 10811 circuit comes
into play is beyond 1 kHz from the carrier, where the
Burgoon patent circuit (which apparently has prior art
Is the Burgoon patent you are refering to US4283691
"Crystal oscillator having low noise signal extraction circuit" ?
On Wed, 28 Oct 2015 06:45:54 -0700
"Richard (Rick) Karlquist" richard@karlquist.com wrote:
Bruce has it exactly right. At offset frequencies beyond 1 kHz,
the source impedance for the grounded base is very high due to
the crystal impedance being very high. As Burgoon explains,
this condition suppresses base recombination noise, and the
only noise mechanism that is significant is the collector shot
noise. (To minimize shot noise, don't run more DC collector
current than necessary).
I read Ulrich Rohde's 1977 article showing this circuit,
before I started working at HP in 1979. When I got to HP,
The article in question is either
"Stable Crystal Oscillators", Ham Radio Magazine, June, 1975
or
"Effects of Noise in Receiving System", Ham Radio Magazine, November 1977
Unfortunately, there doesn't seem to be an online source for the
Ham Radio Magazine. If anyone has a digital version of these
articles, I would appreciate a copy.
Attila Kinali
Hoi Will,
On Mon, 09 Nov 2015 14:09:27 +1300
Will ZL1TAO@gmx.com wrote:
Ham Radio Magazine used to be available on archive.org but has been
removed. 73 Magazine is still available.
CD's of it are available (expensive).
I'll look later and see which issue it was.
Thanks. I already got a copy of both articles.
Attila Kinali
--
It is upon moral qualities that a society is ultimately founded. All
the prosperity and technological sophistication in the world is of no
use without that foundation.
-- Miss Matheson, The Diamond Age, Neil Stephenson
Driscoll wrote a lot about oscillators over the years.
I couldn't find anything specific to discontinuous operation.
Do you have a titel of a paper related to this?
What Driscoll was talking about was self limiting in a
transistor. That is discontinuous operation, although
Driscoll doesn't call it that. His earliest papers on
this circuit go back to around 1972, and are in either
UFFC proceedings and/or FCS. Many later papers cite
these.
The 2 transistor "Driscoll oscillator" fixes this problem
by using an additional stage that limits instead of the
oscillator transistor.
Is the Burgoon patent you are refering to US4283691
"Crystal oscillator having low noise signal extraction circuit" ?
Yes.
Rick wrote:
What Driscoll was talking about was self limiting in a
transistor. That is discontinuous operation, although
Driscoll doesn't call it that.
A transistor amplifier can self-limit at either end of its output
swing -- by going into saturation at one end, or by running out of
current at the other end. "Discontinuous operation" refers to the
latter -- the transistor does not draw current through the full 360
degrees of each cycle. This is better than running the transistor
into saturation because when the transistor saturates, it presents a
low impedance to the resonator that spoils the resonator Q -- leading
to increased phase noise (as well as high distortion). With
discontinuous operation, when the transistor is off it presents a
high impedance to the resonator and avoids the phase noise
penalty. It also causes less distortion than saturation.
FET transconductance is a quadratic function of drain current, so a
FET amplifier can usually reach limiting gain (and, therefore, stable
oscillation) without actually cutting off the drain current. If a
FET is kept in its saturation region (drain-source voltage greater
than several volts throughout each cycle), it also has very high
drain resistance and presents a light load to the resonator at all
times. [NOTE that a FET's "saturation region" has nothing to do with
voltage saturation ("clipping") -- it designates the
constant-current, "pentode" region of the FET's characteristic curves.]
Very low PN oscillators can be made with JFETs that have low input
voltage noise. Presumably, the reasons we haven't seen more
commercial designs using low-noise JFETs are (i) the large spread of
FET parameters, which may require selecting FETs, and (ii) for
ovenized oscillators, the rapid rise of gate leakage current with
temperature -- both of which complicate the mass production of a
consistent product.
There is another transistor operating mode that minimizes oscillator
phase noise, namely "very Class C." In this mode, the transistor
draws current during only a very small portion of each
cycle. Although it loads the resonator while it conducts, that is
only for a very short portion of each cycle. For the rest of the
cycle, the transistor is not conducting and does not load the
resonator. The net effect is a lightly loaded resonator and good PN.
Best regards,
Charles