Sweep: Build new topology signal source
project start@2014/8/03
band flatness response of a VCO is very important for sweep signal source. zoom oscillator suitable for such a application. i plan to build a array of VCO cover 300 khz to 50 Mhz range, use diodo as RF switching.
band switching control both the diode switch and the power of oscillator, so RF switch attenuation is not critical.
Negistor is been used to setting up oscillator. The Feedback capacitor C1, C2 should be adjust for each band.
Full oscillator:
for each band, need adjust the bias resistor: 10k to 100k, even < 10k or > 100k. basic policy is get enough gain from the feedback capacitor (larger one), then adjust bias until get flat response in whole band. typical flatness , amplitude change will < 10%, and even better.
300khz to 800khz
Q1, Q2: SC9014 SC9015
feedback C1, C2: 22pf
Bias: 240k at initial testing (without loading)
130k: when loaded by RF switch
tank: 390uH standard inductor combine with one 1sv149.
signal extract: directly from tank hot end
frequency cover: 476khz to 980khz
flatness to AD8307 power meter: Failure, stop oscillate, hard start to oscillation
First problem is, tank energy is too high, make varactor bias several detune by signal, output waveform is severely distortion. you can not lower the tank emerge by with reduce bias, because low bias make amplitude response not flatness. i made a wrong decision, limit the tank energy by 2 diode.
problem of this circuit is, the tank energy is too low and tank is very lossy, after finished all system. i found it need a buffer to driven the diode switch. and most Annoying thing is, after change to battery power, oscillation stopped, even with heavy bias, it still can not start oscillate, seem power rail need some noise to kick it up.
Failure reason check:
too compact layout, there are some short circuit!
1M to 4Mhz
ok, we learn something from first oscillator, the right way to setup a tank tolerant the high tank energy is use back to back varactor pair which you still can get eye candy waveform.
Q1, Q2: SC9014 SC9015
feedback C1, C2: 5.6x2
Bias: 43k
tank: 40uH standard inductor combine with back to back 1sv149 pair
signal extract: directly from tank hot end
frequency cover: 1.9Mhz to 4.6Mhz
flatness to AD8307 power meter: <10dB, -10dBm, almost square wave, before RF switch waveform is acceptable
another thing learn is a oscillator always want a buffer. so i pick up a wide recognized buffer introduce by ARRL:
but, unfortunately, the output from this buffer not be flat as VCO, from 1Mhz to 4Mhz amplitude reduced significantly. i run a spice simulation of this circuit, AC sweep show it Gain start roll down from 1Mhz, what a coincident! most of this response is BJT's miller capacitance and very high input couple resistor, which form a low pass RC filter.
by simulation and get ideas from MMIC, i got this buffer, which provide about up to 200Mhz flatness response, and > 1k input impedance@100Mhz. if use one transistor as buffer, input impedance will significantly small, most because bjt's input capacitance.
A simulation low input capacitance not mean this buffer will severely loading the tank make it stop, what really happen is the input capacitor will became a part of the tank, impact the frequency, and stability.
4M to 10Mhz
Q1, Q2: SC9014 SC9015
feedback C1, C2: 5.6x2
Bias: 1.6k (provide 50ohm Driven directly)
tank: 7uH FCZ style CAN combine with back to back 1sv149 pair
signal extract: secondary of the IF CAN, driven 50ohm directly
frequency cover: 4Mhz to 11Mhz
flatness to AD8307 power meter: Failure, stop oscillation
first time i use a easy to tapped tank, and i try connect the 50ohm load directly to the tank tap( 4:1 impedance ratio), amazing thing is, it could deliver 4V peak to peak voltage to 50ohm load.
so i decide keep this capability, left the bias 1.6k. but this decision still wrong. such a strong signal(with out loading), will make diode switch turn into a frequency doubler, not a switch. and if you connect 50ohm load after RF switch, the current flow through the diode switch will clipped by diode, cause the diode bias current is not that enough.
9-27mhz
Q1, Q2: 2n3906+2n2222
feedback C1, C2: 15.6pf
Bias: 4.7k bias
tank: 0.8uH FCZ style CAN combine with back to back 1sv149 pair
signal extract: secondary of the IF CAN, driven 50ohm directly
frequency cover: 10.5Mhz to 27Mhz
output: 600 mVpp
flatness to AD8307 power meter: <1db (50ohm load), almost square wave, before RF switch waveform is acceptable
20Mhz to 50Mhz
Q1, Q2: 2n3906+SC9018
feedback C1, C2: 15.6pf
Bias: 4.7k bias
tank: 10Mhz FCZ IF CAN toroid: 4T, with back to back 1sv149 pair
signal extract: secondary of the toroid inductor, 1T
frequency cover: 17.4 to 55Mhz
flatness to AD8307 power meter: <2db , almost square wave, before RF switch waveform is good.
the amplitude response is flat. and might because low input capacitance and higher Ft, tune range extend bigger than 3:1.
Buffer and switch
after whole system being integrated together, one problem raising. driven a 50 ohm Loading, oscillator can do , but RF switch can not. in another way, arrange the buffer before the RF switch, is meanless to the sweep. cause the oscillator could driven the RF switch, buffer provide 50ohm driven is useless before RF switch. (buffer for isolation is different thing.)
the wide band oscillator need a wideband buffer, normal buffer can not offer wide band flatness, so arrange buffer after the RF switch is a good choice. and isolation the oscillator should use tank tap provide light couple to oscillator, this will provide another benefit, because tap reduce signal level, then easily pass through RF switch without distortion.
Toward better design
in general, this construction is useless!
1. no frequency count output
2. bias of RF switch is bad, signal is too strong to pass through the diode
3. oscillator coupled to RF switch is not well controlled.
3. 50 ohm loading testing failure on 4-10Mhz VCO, other VCO is fine but pass after RF switch, no one work properly.
4. layout is too compact which prevent to do final adjust.
Better design:
1. use tapped inductor extract signal from VCO
2. extracted signal should less than 100 mVpp and even smaller, then it can pass through RF switch without over driven the diode switch, and easy to be buffered by use emitter follower.
3. VCO bias should take the load into account.
4. should not try get large signal from VCO, better way is use RF amplifier after the buffer to get strong signal
5. back to back varactor as tank variable capacitor, better strong signal response.
In one word, the experimental sweep signal source is failure, but found a way to achieve better result.
Band plan consideration:
455khz oscillator suffer with no suitable tank coil to construction, too big or too lossy. this band should drop wideband sweeping function, use a 455khz IF can, remove original embedded capacitor change to a smaller capacitor, and use varactor to form a VCO, could achieve 350khz to 550 khz range.
heterodyne signal source for low band sweep: 300khz to 3M, better cover should be heterodyne signal source, i.e, use 10Mhz crystal oscillator, and a 10 to 15Mhz oscillator, after mix, get 300khz to 3-5Mhz signal, cause 10mhz is a octave end to 5Mhz, 3:1 for 3Mhz, the signal is easy get very clean by a simple low pass filter.