Log Sweeper (III) Improve 

August/07 2016

If don't want know the detail, please JUMP TO CONCLUSION (Oct 2016): DBM Frequency mix generator

This is the 2nd one of a series build blogs: 

Signal: Prototype of Log Detector Sweeper

Signal: Improve The Log Detector Sweeper

Signal: Build The Log Detector Sweeper

Build VCO and Buffer

Module one: the VCO and the Wide band buffer.

changes:

    the joint of 15 k and Collector of 9018 should connect a 103 capacitor to ground.

Buffer

Changes:

2N5485 should connect to VCO via a DC blocker capacitor)

Problem(NOTE: while varactor voltage low than around 0.7V, it's obviously cloud not work correctly):

    hard to start oscillation on low frequency end (45 Mhz), solution: change the feedback capacitor from 6.8 pf to 10 pF.

    (while soldering the buffer, it's easy to start oscillation,  because measurement absorb too much energy.  it's not harmonics clean, obviously.

Another pitfall is that this oscillator get more feedback on high frequency end, so if the low end get enough amplitude, the high frequency end will distortion.  this is not a very big problem while use as a sweeper. and we mix the 60 Mhz-120 Mhz VCO with 60 Mhz get a  300 khz to 45 Mhz signal, in this scenario, the second harmonics exceeded the seep band very far, and will be filter out almost completely.

(this problem addressed later, refer to following sections)

Build Crystal Oscillator and buffer

Use the layout schematic: adding the emitter degeneration.

NOTE:

    buffer output need DC block capacitor and a serials a 47 ohm resistor to kill mixer distortion.

oscillator build note :

1. the oscillator stage quiescent current is too large, leading to waveform distortion, less quiescent reduce the harmonic 

2. final use a 18 pF Mica instead of 16.8 pF capacitor,  the Coil is tunable

3. should try replace 3.3k with 2.2k

buffer build note:

1. output 330 and emitter bypass: 1000pF

2. degeneration resistor:  51R

A problem,  the flatness is bad than prototype, I think there is 4 db change from 300 khz to 45 Mhz this time, but later i found my Power meter output 10 mV/dB. 

i try insert a parallel , 160R//14p, to the post mixer amplifier input (original is  input 51 R resistor).  got almost 1 db flatness，but only -8 dbm output.

Output contain strange frequency component

Possible improvement to the crystal oscillator:

 1:  43 pF present strong feedback, might be 143 pf or 200 pf. 

 2:  6T, 16.8 pF is not actually resonant at 60 Mhz,  it's actually resonant frequency is far more high. the oscillator is working because there is 22 pF serials 43 pF parallel with the tank.  capacitor around 27 pf resonant at 60 Mhz with this coil .

 3:  the 2 pF capacitor at 60 Mhz present only 800 ohm impendance .  while tune the filter, it's pull the oscillator too much, at the worst case, OSC stopped oscillating. the filter should  connected behind the buffer.

a better design is , suppose we want 0-45 Mhz,  Fix OSC is x Mhz,  2*x - (x+45) should at least high than 45 Mhz,  we get a better sweeper if the x>= 90 Mhz

so there is more point could be improved:

  remove the filter, added it after the buffer, might should transform the impendance high enough.

  use 1k-3k resistor coupled to the buffer, get signal from hot end of the 43pf capacitor, instead of connect to the tank.

  adjust feed back capacitor.

  use 90 Mhz fix frequency OSC. 

NOTE: 

later experiment proof the reason is not cause by 2nd harmonic, it's 3rd harmonic. 

design mistake

change the 43 pf to 100 pf,  and move the single tuned filter behind buffer. then what we got this time

 NOTE: 

the older waveform trace on the screen is FAT， this the most bad thing for a oscillator

after changes  trace is much more clear than before，even it's obviously distortion but still much more good than before）

The sweeper final output, 300 khz, and 30 Mhz: (screen trace is still fat on top and bottom, but much much better than before)

NOTE:

    move the filter position is far more important than change 43 pF to 100 pF,.  but that change still have visible effect.

    couple the buffer from the joint hot end 43 pF,  improved the oscillator stability, avoiding possible parasitic oscillation. 

problem:

    DBM  driven level decrease to around -8 dBm, not enough.

All modification showing in the following schematic:

Move the filter behind amplifier chain

The output power dropped, i want more power from IF port, i believe it's possible to increase the mixer output  if the LO get more power (currently, the LO is the crystal oscillator ) .

i observed if the LO driven is less than 0 dbm, the flatness got bad, bumping more than 4 dB from 400 khz to 45 Mhz. So it's why i want increase the LO driven level, so add one more amplifier, but got distortion waveform again...  tune filter to a better waveform,  the LO drive power then decrease again. Dead loop.

i reallize the amplifier will introduce more harmonics, the filter should put just before the DBM.

To reduce the amplifier harmonic, it need remove the degeneration's parallel capacitor from all amplifiers(following 103 capacitor near the text 50-100 ohm), that's capacitor introduce too much distortion.

 by remove the amplifier's capacitor, the filter now won't cause the output power drop too much, because there is less harmonics energy, and the Power Meter is based on AD8307 which report a fake high power if the signal is rich harmonics. 

What does such a  waveform trace mean on Oscilloscope?

most in-frustrating things is,  for several month i suffer from the oscilloscope waveform trace like hell, what does this really mean? what's this type waveform really mean?  i try hard to adjust the trigger，Finally i got a clean output (as following photo)

Wow, now reduce the sweep time, the true story revealed:

it's obviously, there is a lower frequency modulate this signal. AM period is 0.5 uS( 2 Mhz), the base frequency of this signal is 20 Mhz. 

further measurement data reveal the true root cause:

29.3 Mhz AM by  1.25 Mhz 0.8uS

29.8 Mhz AM by  400 khz 2.5 uS

29.98 Mhz almost zero AM                (VCO 89.55, LO:60.001)  ===>  3*60 - 2*89.55 = 0,   the zero beating happens.

30.3 Mhz AM by  2 Mhz  0.5 uS

crystal oscillator frequency drift a little bit, but  according 2*VCO-2*LO , zero beat always around 30 Mhz. this confirm it's the LO's 3rd harmonics beating VCO 2nd harmonics.

Target to reduce the 3rd harmonics 

insert new 68 Mhz LPF with image reject( bad elliptic)

NOTE the LO is still the crystal oscillator.

NOTE: filter inductor is 200 nH , and this is not really a elliptic design, don't use this filter.

by tune the filter, i got following result , mixer output and waveform pass through 45 Mhz LPF (-10 dbm, flatness: 2 db）

experimental Experience:

    * while output level less than -10 dBm, the flatness will no more keep in 2dB range. 

    * while output level is less then -10 dBm, waveform will get better. 

    * Wired : the LO energy is less then RF driven level: LO: around 0 dbm, RF: +4 dBm. this is really strange.

this time the TDK coil is totally useless. i had to wind a coil use 13 mm  Nylon Hex Nut, 

    120 nH coil:  5T on 1 3mm Nylon Hex Nut

with all of these painful experimental, what i learned is( for sweeper):

        *   High RF power leading to bad third harmonics distortion

        *   Less LO driven level cause the flatness became bad.

        *   So use  fixed oscillator as RF signal, output level should less than -15 dBm, corresponding RF driven level  < -10 dBm

        *   insert 5 dB pad before DBM after the RF elliptic filter, reduce it's power to less than -10 dBm,  better termination the filter.

        *   boost the 0-45 Mhz signal to -10 dBm or 0 dBm by wind band amplifier and compensate the flatness        

Use Crystal oscillator as RF,  VCO as LO(high power),  reduce RF level less than LO around 10 dB (normal use case need lower 20 dB).  

replace it's buffer with a tuned Class A amplifier , 3:1 output, 9T:3T on 13mm Hex Nut and a 13 pF - 50 pF variable capacitor, the  elliptic even don't needed with tuned class A amplifier.

The most wired thing also been explained: while the crystal connected to LO, if want to get clean output, LO need below the RF port -10 dB !!!!!

 20 Mhz and 300 khz waveform (300 khz contain 60 Mhz RF leaking)

Further Improve performance 

2 . amplifier VCO to 10 dbm

3.  insert 6-10 dB VCO pad

result:

        the flatness keep in 2dB

        output level -15 dBm

        slightly visible AM!!!

        low frequency(<  5 Mhz), LO leaking decreased 

Compare the result by switch in the LO 8 dB pad (LO now is the VCO, and test it at 1.6 Mhz)

another LO pad improvement is reduce the low frequency harmonics, about 5 dB

(4.4 Div/0.2 DIV, improved to 4.4 Div / 1.2 Div).

same symptom also confirmed with CLP-303.

Try SBL-1X

experiment confirmed the SBL-1X obviously beter than CB346M1A. several different behavior observed:

    * SBL-1X leaking obviously less LO energy VS the CB346M1A

    * SBL-1X produce relatively good  signal even out of the LO filter band.

    * SBL-1X , in whole sweep range, 300 khz to 45 Mhz,  almost same quality signal.  only at the LO filter's upper corner, there is a visible amplitude modulation in a very narrow frequency range.

    * SBL-1X  need only 3 dB isolation on LO port to achieve clean signal, the CB346M1A need at least 6dB. even 1 db Isolation pad for the SBL-1X LO port, there is visible improvements.  

NOTE: the photo could not present how good the left waveform is, but my eye could see the different..

output 150 mVpp, there is no visible amplitude modulation even use 5 mV/Div on Tek485.  And if  the 3 dB pad in the RF move to LO, no visible amplitude modulation neither(but still suggest insert pad to RF, leaving the LO driven level as 7 dBm, there is visible distortion if insert pad to LO, driven obviously not enough).

High Power Class A Amplifier

(  filter is hard to tune to good state without SA, and small inductor make it hard to construct)

60 Mhz reject reached -40 dB theoretically, after tune: output from 100 khz to 45 Mhz almost achieved same result with 7 order Chebyshev filter. AM rejection is bad than 7 order Chebyshev.

change 24 k resistor to get stable oscillator. too large quiescent current leading to parasitic oscillation.  feed buffer from crystal via 2~ 6 pF capacitor.

Mixer 2 crystal controlled OSC still get very strong non harmonics spur. add filter after the RF signal eliminated most of spur, this confirmed the LO/RF harmonics component generating the non harmonics spur on IF band. 

DBM RF signal should filter out the harmonics.

Simulation the Double balance mixer

Simulating get same result, RF harmonic contribute lots of non harmonics spur. but  filter out RF from IF output don't improved anything. might because the simulating DBM have a very good RF to IF isolation.

Wave trap for RF

The SBL-1X ( same as SBL-1) RF to IF isolation is much bad than LO IF isolation. this inspired me to adding a wave trap for the RF signal, a 60 Mhz series resonant tank to DBM IF output. Adjust it at several hundreds khz IF signal, I saw the leaking RF signal been perfectly eliminated. this help a lots to eliminated the none harmonics spur frequency.

12T on 13 mm Nylon nut, series with 4-20 pF capacitor trimmer works perfectly.

this somehow confirm, the IF band non-harmonics spur not only related to the LO, RF harmonics, but also related to the leaking RF signal. worth to try another trap for secondary harmonics of the RF signal.

AD8307 Detector: I killed it's speed

after i changed this cap to 100 pF, i finally got know what the  speed of response time mean to me.

Sweeping Things is fun

Key factor to get harmonic/non-harmonic free signal

1. RF level should be less than LO 15~20 dB ( reduce harmonics, reduce IMD, high RF control the DBM instead LO)

2. RF signal harmonics clean, 5th filter  (filter out the harmonics, reduce the non harmonics spur )

3. Output filter should eliminate the  RF leak very well  ( DBM RF isolation is weak than -30 dB, leaking RF generating many non harmonics spur ) 

    7 order chebyshev  + wave trap, RF located above 1.5 time the corner frequency ( 45 Mhz stop filter alter by the wave trap to about 38 Mhz, 60 Mhz RF)

    5 order elliptic+ 7 order chebyshev , 45 Mhz corner frequency, 60 Mhz RF, 1.3 times

4. wave trap for RF does help to kill non harmonics spur, even in a high order filter. 

5. LO's harmonic reduce is also help  to kill non harmonics spur, even with a 3 order filter.

6. LO's PAD is another good to have improvement.  experiment 13 dbm LO with a 6 db pad given a very valid improvement.

7. ***** the DBM wire to the coax connector is key factor: reduce any unnecessary lead, keep it as short as possible.

All of these hard work get a > 30dBc- 40dBc non-harmonics spur. 

Suggestion:

     LO&RF spur free is key factor.

     spur frequency should 2 times high than filter Conner frequency to ensure a good rejection. i.e, RF better set to 90 Mhz while we want 45 Mhz bandwidth.