Hello, I with this circuit, an RF oscillator

39 replies · 8K views · last reply
GTrax said:
It does require the crystal be substituted by the crystal equivalent circuit.
I designed the crystal using spice capacitor parameters. This models the fundamental, but does not model overtones. The code needed to compute crystal equivalent spice capacitor parameters is as follows;

' Crystal specifications wanted.
Fp = 7.040e6 ' crystal parallel resonant frequency, Hz
CL = 20e-12 ' at load capacitance, farads
Q = 20e3 ' crystal Q factor
R1 = 60 ' crystal series resistance, ohms
C0 = 5e-12 ' crystal shunt capacitance, farads

' variables needed for initial temp approximation and final value.
Fs_tmp, Fs ' series resonant frequency, Hz
Wo_tmp, Wo ' angular frequency, rad/sec
C1_tmp, C1 ' crystal motional capacitance
L1 ' crystal motional inductance

' initial computation.
Fs_tmp = Fp
Wo_tmp = Fp * TwoPi
C1_tmp = 1 / ( Q * Wo_tmp * R1 )
Fs = Fp / ( 1 + C1_tmp / ( 2 * ( C0 + CL ) ) )
Wo = TwoPi * Fs

' Refine to final values.
C1 = 1 / ( Q * Wo_tmp * R1 )
L1 = 1 / ( Wo^2 * C1 )

' It generated these capacitor parameters;
Basic capacitance = 18.8393 f F = 0.0188393 pF.
Series inductance = 27.149 mH.
Series resistance = 60.0 ohm.
Parallel capacitance = 5.0 pF.

For a parallel frequency of 7.040 MHz @ Load capacitance = 20 pF
The series frequency is 7.037348 MHz
Here is the crystal model and simple output response.
LTspice schematic of a crystal model: AC voltage source V1 with Rser=0.001 feeds a series-RLC crystal branch with Rser=60, Lser=27m14913, Cpar=5p and a 1.883936e-14 basic capacitance, to output node out1, with R1 10R to ground; simulation directive .ac lin 10000 7.03Meg 7.06Meg.

SPICE AC response plot of V(out1) from 7.030 MHz to 7.060 MHz: magnitude in dB (solid green, left axis) peaks near 7.037 MHz then dips sharply near 7.049 MHz; phase in degrees (dashed green, right axis) swings from about +90 to -90 across the same band.


Once simulating the circuit, change the tuning capacitor of the tank, to sweep across the synthetic crystal. Watch the initial transient ringing to see it decay, which implies it has loss not gain.

GTrax said:
I have information (PDFs) on most aspects of crystal oscillator design and simulation, and when I can find it, I may come back to this thread and post again.
I have searched many of the texts on oscillator design and have not found any similar design, ECO with a tank, or other similar topology.

I find it hard to believe this is an electron coupled oscillator with 390pF between the base emitter terminals.

There is a problem with crystal series traps in that, even when not resonant at the oscillator frequency, it will pass all the distortion and noise to the output.
 
Attachments
  • LTspice schematic of a crystal model: AC voltage source V1 with Rser=0.001 feeds a series-RLC crystal branch with Rser=60, Lser=27m14913, Cpar=5p and a 1.883936e-14 basic capacitance, to output node out1, with R1 10R to ground; simulation directive .ac lin 10000 7.03Meg 7.06Meg.
    xtal_model_1.png
    16.9 KB · Views: 1,149
  • SPICE AC response plot of V(out1) from 7.030 MHz to 7.060 MHz: magnitude in dB (solid green, left axis) peaks near 7.037 MHz then dips sharply near 7.049 MHz; phase in degrees (dashed green, right axis) swings from about +90 to -90 across the same band.
    Xtal_resp_1.png
    9.5 KB · Views: 1,149
  • Like
Likes   Reactions: Merlin3189 and dlgoff
@Baluncore
Can you get your transistor simulation to show the negative input impedance as mentioned by:
Merlin3189 said:
But, I do see references around to instability in emitter followers with a capacitive load. I can't find a detailed treatment of this (and I'm really not up to doing one myself !) but a TI application note for a wideband buffer amp IC makes a fairly unambiguous comment re.emitter follower cct,
I suspect that is really needed for the circuit to function.

(Looks like you got the crystal simulatoin nailed.)
 
GTrax said:
When converting a fundamental oscillator type (like Colpitts) into one mediated by a crystal, they sometimes get developed to exploit the crystal as a narrow filter to deliver an output with very low phase noise, which is a critical feature when the output is to be used in receiver mixer stages.

You are right in that the defining feature of a Colpitts sort is the tuning tank of 2 capacitors in series and an inductor across them.
The variant you have is not quite like that, but that it also has the equivalent circuit of a crystal in series with 220pf and 12 Ohms, all across the resonant part, as in parallel as well.

I can recognize that you are also using LTSpice simulation tool from Linear Technology. I have used this tool before to simulate crystal oscillators. It does require the crystal be substituted by the crystal equivalent circuit.

My first impression of how the circuit operates is that the shunt impedance of the crystal with the 220pf and 12 Ohms, put on the transistor base, will allow the feedback to be effective for oscillation only at the crystal frequency. The very small sample in series with the crystal, taken off the 12Ohms/220pf combination, driving the matching transformer amounts to minimal loading. The signal at this point will be from the resonant ringing in the crystal quartz, in effect a very narrow filter, and the waveform should have very low phase noise.

I am not completely sure of the inductor connection on the variable and 390pF capacitors. Not quite what one would call a "Colpitts" sort.

I have information (PDFs) on most aspects of crystal oscillator design and simulation, and when I can find it, I may come back to this thread and post again.
Thnx MAN:partytime:
 
Baluncore said:
I designed the crystal using spice capacitor parameters. This models the fundamental, but does not model overtones. The code needed to compute crystal equivalent spice capacitor parameters is as follows;

' Crystal specifications wanted.
Fp = 7.040e6 ' crystal parallel resonant frequency, Hz
CL = 20e-12 ' at load capacitance, farads
Q = 20e3 ' crystal Q factor
R1 = 60 ' crystal series resistance, ohms
C0 = 5e-12 ' crystal shunt capacitance, farads

' variables needed for initial temp approximation and final value.
Fs_tmp, Fs ' series resonant frequency, Hz
Wo_tmp, Wo ' angular frequency, rad/sec
C1_tmp, C1 ' crystal motional capacitance
L1 ' crystal motional inductance

' initial computation.
Fs_tmp = Fp
Wo_tmp = Fp * TwoPi
C1_tmp = 1 / ( Q * Wo_tmp * R1 )
Fs = Fp / ( 1 + C1_tmp / ( 2 * ( C0 + CL ) ) )
Wo = TwoPi * Fs

' Refine to final values.
C1 = 1 / ( Q * Wo_tmp * R1 )
L1 = 1 / ( Wo^2 * C1 )

' It generated these capacitor parameters;
Basic capacitance = 18.8393 f F = 0.0188393 pF.
Series inductance = 27.149 mH.
Series resistance = 60.0 ohm.
Parallel capacitance = 5.0 pF.

For a parallel frequency of 7.040 MHz @ Load capacitance = 20 pF
The series frequency is 7.037348 MHz
Here is the crystal model and simple output response.
View attachment 227696
View attachment 227697

Once simulating the circuit, change the tuning capacitor of the tank, to sweep across the synthetic crystal. Watch the initial transient ringing to see it decay, which implies it has loss not gain.I have searched many of the texts on oscillator design and have not found any similar design, ECO with a tank, or other similar topology.

I find it hard to believe this is an electron coupled oscillator with 390pF between the base emitter terminals.

There is a problem with crystal series traps in that, even when not resonant at the oscillator frequency, it will pass all the distortion and noise to the output.
Thank You Very Much, for helping me, and loosing time for me...:partytime:
 
First, apologies to Baluncore for confusing him with Anorlunda.
Well done on all the simulation work. I really ought to try to learn to use this Spice stuff myself.

Second, I think I can now see the Colpitts configuration, with the crystal and 390pF & VC across it. If the 4.3uH is bundled in with the VC, there will be a setting where, at the crystal frequency the combination appears as a capacitor of the right value to get the unit+ loop gain. Quite how the inductor helps, I'm not sure yet.

What would be very helpful, would be someone with a copy of the ARRL "Experimental Methods in RF Design" which seems to be the source of this cct.

And since my interest has been piqued in the idea of emitter follower negative resistance, I'd be very pleased if anyone could point me to a detailed analysis of the emitter follower with a complex load, rather than a predominantly real one. Otherwise I guess I'll have to make a pathetic attempt at doing it myself and posting a thread to ask for people to sort it out.
 
  • Like
Likes   Reactions: Tom.G and michael1978
Merlin3189 said:
First, apologies to Baluncore for confusing him with Anorlunda.
Well done on all the simulation work. I really ought to try to learn to use this Spice stuff myself.

Second, I think I can now see the Colpitts configuration, with the crystal and 390pF & VC across it. If the 4.3uH is bundled in with the VC, there will be a setting where, at the crystal frequency the combination appears as a capacitor of the right value to get the unit+ loop gain. Quite how the inductor helps, I'm not sure yet.

What would be very helpful, would be someone with a copy of the ARRL "Experimental Methods in RF Design" which seems to be the source of this cct.

And since my interest has been piqued in the idea of emitter follower negative resistance, I'd be very pleased if anyone could point me to a detailed analysis of the emitter follower with a complex load, rather than a predominantly real one. Otherwise I guess I'll have to make a pathetic attempt at doing it myself and posting a thread to ask for people to sort it out.
Hi Merlin this book is from
Merlin3189 said:
First, apologies to Baluncore for confusing him with Anorlunda.
Well done on all the simulation work. I really ought to try to learn to use this Spice stuff myself.

Second, I think I can now see the Colpitts configuration, with the crystal and 390pF & VC across it. If the 4.3uH is bundled in with the VC, there will be a setting where, at the crystal frequency the combination appears as a capacitor of the right value to get the unit+ loop gain. Quite how the inductor helps, I'm not sure yet.

What would be very helpful, would be someone with a copy of the ARRL "Experimental Methods in RF Design" which seems to be the source of this cct.

And since my interest has been piqued in the idea of emitter follower negative resistance, I'd be very pleased if anyone could point me to a detailed analysis of the emitter follower with a complex load, rather than a predominantly real one. Otherwise I guess I'll have to make a pathetic attempt at doing it myself and posting a thread to ask for people to sort it out.
You can take the book from library genesis.com only the book but the cd of book i don't know where to find.
 
Merlin3189 said:
What would be very helpful, would be someone with a copy of the ARRL "Experimental Methods in RF Design" which seems to be the source of this cct.
I have a copy here, and I see fig 4.24 shows a series output crystal claimed to reduce harmonics.
Schematic of a crystal oscillator output circuit with a transistor Q1, a crystal in series with capacitor C1 (330) feeding the Output, 10K and 470 bias resistors and capacitors, 2.2K emitter resistor, and a +12V supply through 47 ohms with a 0.1 decoupling capacitor.

But it is a Colpitts oscillator with the inductor replaced by a crystal to make a crystal oscillator. It has no tank, that is done by the crystal and capacitive divider. The output voltage appears across the series capacitor. This shows that the frequency of the oscillator is primarily set by the crystal.
Has anyone found any circuit in any reference with the tank circuit and the series crystal output.

I have looked for, but have not found negative resistance in the tank based circuit.
 
Attachments
  • Schematic of a crystal oscillator output circuit with a transistor Q1, a crystal in series with capacitor C1 (330) feeding the Output, 10K and 470 bias resistors and capacitors, 2.2K emitter resistor, and a +12V supply through 47 ohms with a 0.1 decoupling capacitor.
    Colpitts-XtalOsc.png
    45.5 KB · Views: 910
  • Like
Likes   Reactions: Merlin3189
Thanks to everbody for help, and nice wekend for everybody.:partytime:
 
Starting with the circuit from Fig 4.24 of ARRL (2003) “Experimental Methods in RF Design”.

LTspice schematic of a Colpitts crystal oscillator with a 7.040MHz crystal (Q = 20k), two 10k bias resistors, two 470p capacitors, a 330p output capacitor, a 2k2 emitter resistor, and a 2N3904 NPN transistor, labelled Fig 4.24.


A 7MHz crystal with a Q of 20k has a BW of about 7MHz / 20k = 350Hz. That should give an oscillation rise-time of about 1 / 350Hz = 2.86ms. That same crystal model is used here in all simulations.
Here is the output rise-time envelope, which can be seen is slower than expected to start.
Simulated transient plot titled V(out1), green waveform: output voltage rises from 0 mV near 5 ms to a steady ±200 mV envelope by about 12 ms, axes in ms and mV.


Now we get to the OP's “QRP Homebuilder circuit” which is really a Colpitts crystal oscillator, where the crystal replaces the tuning inductor. The series combination of crystal and output capacitor are an integral part, in parallel with the capacitive divider of the oscillator. Here it is drawn in a way that shows that arrangement more clearly.
Schematic of a crystal oscillator with a 7.040MHz crystal, Q = 20k, in a Colpitts-style arrangement with an NPN transistor Q1 (2N3904), inductor L1 4μH, and capacitors C1 390p, C2 180p, C3 0µ 1, C4 0µ 1, C5 220p; beside it is a table of C2 values with output amplitude and harmonic level in dB.


The table of C2 shows half Vpp output amplitude as ±Vp, and the highest harmonic relative to the fundamental. There is a region with good output amplitude and good harmonic suppression.

The simple Colpitts crystal oscillator without an inductor or RFC is slow to start. The addition of the inductance can significantly reduce the rise-time of the oscillation, but that requires the bias current be set separately from the capacitive divider, hence the unusual circuit.

It is important that what we have been calling the 'tank' is NOT made resonant near the crystal frequency. The tank must be tuned on the capacitive, lower frequency side of the crystal frequency. That way it looks like a capacitor at the crystal frequency, which is what reduces the circuit to a crystal oscillator with the Colpitts capacitive divider.

The capacitor output technique can reduce the harmonic content to be 50dB below the fundamental, but don't expect anything more than about 50dB down. Here is the clean output waveform, with C2 = 180 pF.
Simulated time-domain plot of V(out1) in green on black, showing a sinusoid from 0 to about 980 ns with amplitude near 2.3 V and a slow envelope rise at the start.


Risetime with C2 = 470pF is;
Simulated transient plot of V(out1) from 0 to about 15 ms, showing a sinusoidal output that stays at zero for roughly 6 ms, then grows to about 330 mV peak in a slow envelope.


Risetime with C2 = 195pF is;
Simulated transient plot titled V(out1) showing a green oscillation envelope rising from 0V at about 1.3ms to a steady amplitude near ±1.9V by about 8ms, axes from 0ms to 15ms and -2.0V to 2.0V.


Risetime with C2 = 180pF is; ( The saw artefact is a beat between the simulation timestep and the crystal frequency. )
Simulated transient plot of V(out1) in volts against time from 0 to 15 ms, showing a green oscillation envelope rising from 0 V after about 1 ms to about plus and minus 2.2 V by 5 ms and staying roughly steady to 15 ms.


Risetime with C2 = 130pF is; ( Here the artefact shows up earlier in the rise, but then later as whiskers. )
Simulated transient plot titled V(out1) showing a sinusoidal output starting from 0 V near 1.5 ms, growing over roughly 2 ms to a steady amplitude of about ±2.5 V, with x-axis from 0 ms to 14 ms.

It can be seen that this rises significantly faster than was expected.
 
Attachments
  • LTspice schematic of a Colpitts crystal oscillator with a 7.040MHz crystal (Q = 20k), two 10k bias resistors, two 470p capacitors, a 330p output capacitor, a 2k2 emitter resistor, and a 2N3904 NPN transistor, labelled Fig 4.24.
    fig4.24circuit.png
    8.8 KB · Views: 1,109
  • Simulated transient plot titled V(out1), green waveform: output voltage rises from 0 mV near 5 ms to a steady ±200 mV envelope by about 12 ms, axes in ms and mV.
    fig4.24risetime.png
    5.8 KB · Views: 1,010
  • Schematic of a crystal oscillator with a 7.040MHz crystal, Q = 20k, in a Colpitts-style arrangement with an NPN transistor Q1 (2N3904), inductor L1 4μH, and capacitors C1 390p, C2 180p, C3 0µ 1, C4 0µ 1, C5 220p; beside it is a table of C2 values with output amplitude and harmonic level in dB.
    CXOckt.png
    33.1 KB · Views: 1,040
  • Simulated time-domain plot of V(out1) in green on black, showing a sinusoid from 0 to about 980 ns with amplitude near 2.3 V and a slow envelope rise at the start.
    CXOwave180pF.png
    15.8 KB · Views: 1,043
  • Simulated transient plot of V(out1) from 0 to about 15 ms, showing a sinusoidal output that stays at zero for roughly 6 ms, then grows to about 330 mV peak in a slow envelope.
    CXOrise470pF.png
    5.7 KB · Views: 1,013
  • Simulated transient plot titled V(out1) showing a green oscillation envelope rising from 0V at about 1.3ms to a steady amplitude near ±1.9V by about 8ms, axes from 0ms to 15ms and -2.0V to 2.0V.
    CXOrise195pF.png
    5.1 KB · Views: 1,021
  • Simulated transient plot of V(out1) in volts against time from 0 to 15 ms, showing a green oscillation envelope rising from 0 V after about 1 ms to about plus and minus 2.2 V by 5 ms and staying roughly steady to 15 ms.
    CXOrise180pF.png
    7.2 KB · Views: 1,039
  • Simulated transient plot titled V(out1) showing a sinusoidal output starting from 0 V near 1.5 ms, growing over roughly 2 ms to a steady amplitude of about ±2.5 V, with x-axis from 0 ms to 14 ms.
    CXOrise130pF.png
    6.3 KB · Views: 1,006
  • Like
Likes   Reactions: michael1978, Merlin3189 and Tom.G
Baluncore said:
Starting with the circuit from Fig 4.24 of ARRL (2003) “Experimental Methods in RF Design”.

View attachment 227805

A 7MHz crystal with a Q of 20k has a BW of about 7MHz / 20k = 350Hz. That should give an oscillation rise-time of about 1 / 350Hz = 2.86ms. That same crystal model is used here in all simulations.
Here is the output rise-time envelope, which can be seen is slower than expected to start.
View attachment 227806

Now we get to the OP's “QRP Homebuilder circuit” which is really a Colpitts crystal oscillator, where the crystal replaces the tuning inductor. The series combination of crystal and output capacitor are an integral part, in parallel with the capacitive divider of the oscillator. Here it is drawn in a way that shows that arrangement more clearly.
View attachment 227807

The table of C2 shows half Vpp output amplitude as ±Vp, and the highest harmonic relative to the fundamental. There is a region with good output amplitude and good harmonic suppression.

The simple Colpitts crystal oscillator without an inductor or RFC is slow to start. The addition of the inductance can significantly reduce the rise-time of the oscillation, but that requires the bias current be set separately from the capacitive divider, hence the unusual circuit.

It is important that what we have been calling the 'tank' is NOT made resonant near the crystal frequency. The tank must be tuned on the capacitive, lower frequency side of the crystal frequency. That way it looks like a capacitor at the crystal frequency, which is what reduces the circuit to a crystal oscillator with the Colpitts capacitive divider.

The capacitor output technique can reduce the harmonic content to be 50dB below the fundamental, but don't expect anything more than about 50dB down. Here is the clean output waveform, with C2 = 180 pF.
View attachment 227808

Risetime with C2 = 470pF is;
View attachment 227809

Risetime with C2 = 195pF is;
View attachment 227811

Risetime with C2 = 180pF is; ( The saw artefact is a beat between the simulation timestep and the crystal frequency. )
View attachment 227812

Risetime with C2 = 130pF is; ( Here the artefact shows up earlier in the rise, but then later as whiskers. )
View attachment 227815
It can be seen that this rises significantly faster than was expected.
Thnx baluncore you are the best:partytime: