160/80M VFO - 'High Stab' - Class D, E digital drive

[W1VD]

Been running this VFO for a few months...exhibits <2Hz/hr drift after warm up. Oscillator stage might also make a good retrofit into older less stable VFOs...

http://www.w1vd.com/16080VFO.html

[WA1GFZ]

Jay,
I was reading an article by the master, Dr. Rohde this morning on oscillator phase noise. I saw a couple plots that showed a 20 dB drop in phase noise when the tank is a higher Q. I my dig up my old Ham radio article on U310 oscillators with a hunk of semi ridged coax as a resonator and see how it plays against the typical tank circuit.
It will be interesting to see your results when the air core inductor gets installed.

[steve_qix]

I like it    Karl KD3CN has also done some work with a Vaccor (?sp) VFO circuit with good results.  I think the Vaccor is more complicated.

I like tha analog approach.  No phase noise to contend with 

[WA1GFZ]

Steve,
All oscillators have phase noise. The higher the Q of the resonator the better it is. The best being a crystal. I agree analog is cleanest compated to DDS. That is why the R390A is still one of the cleanest LO systems.   

[w3jn]

Dr. Rohde also is a big proponent of eliminating the clipping diode at the gate of the oscillator FET.  This greatly reduces phase noise as well.  I don't have his synthesizer design book close at hand but IIRC it was on the order of 20 dB close in.  Dr. Rohde ran phase noise tests of a ARRL oscillator with and without the clipping diode; the PN plots were in the book.

Particularly with a parallel L-C oscillator, if you eliminate the clipping diode, you run the risk of frying the JFET due to excessive tank circuit voltage.  As your circuit uses a series L-C that shouldn't be an issue, but you can solve this problem if you run into it by very lightly coupling the tank circuit to the gate of the JFET.  I used a 10 PF variable in one of my designs and set it to the minimum capacity consistent with reliable oscillator starting.

[WA1GFZ]

Good catch John.
I told Jay about the problem but forgot the fix and source of info.
The Racal oscillator in the synthesizer taps the inductor in 3 places. bottom tap output, second tap the source through a resistor 47 ohms in the case of the 6830 with a u310 then third tap up goes directly to the gate. Bottom of the inductor to ground and top to the varactors so the tuning cap in this case. The series 47 ohm resistor looks like a source of noise for the tank though. gfz

[WA1QHQ]

I saw some experiments done a while back using a ceramic resonator for the VFO tank and pulling the resonant frequency with either an air variable or veractors, turns out the ceramic resonator can be pulled a lot further than a crystal and still maintains reasonably high Q. I think that for a free running oscillator this would be the best approach for lowest phase noise and best frequency stability. The down side is limited pulling range of the ceramic resonator and getting a resonator near the frequency of interest but I don't think these restrictions are all that difficult to work around, for instance there are standard resonators that are in the 75 meter band and a divide by 2 will get you into 160, to get a better tuning range start with a higher frequency resonator and then mix down to the desired band with a second crystal oscillator.

Mark WA1QHQ

[WA1GFZ]

Andersen Labs used active phase shifters to pull SAW resonators. Picture a quad hybrid with 2 ports for in and out. The two remaining had series tuned circuits with an inductor with a varactor to ground. We got about 1 % tuning range but it was pretty clean. Usually -100 dBc at 1 KHz or better. The oscillator loop was a SAW an MMIC amp an active phase shifter and a second quad to split power off from the loop. All were set up for enough phase shift to keep it oscillating. They were pretty cool devices....when the SAWs were right.
Mark, I wonder if you ever checked the phase noise of the PTO you solid stated?

[W1VD]

John, Frank

Frank mentioned the clamping diode the other day and I also recalled seeing mention of  that years back. Went home that night and ran another set of measurements - with and without the diode.   

The phase noise test setup I used is shown in the HP8640B manual page 4-14   (HP8640B#1 phase locked to oscillator under test, spectrum analyzer, atten., amplifier,  etc). 

Comparisons were made between an HP8640B#2, HP3325A and the VFO with and without diode. Tests were performed at 4 Mhz and I looked at spacings from 10 kHz - 100 kHz. The HP3325A was on average about 10 dB noisier than the 'benchmark' HP8640B#2 and the VFO was about 2 dB noisier than the HP8640B. Then I removed the diode...absolutely no difference!

Ran another test tonight to make sure I didn't miss anything. Instead of using the HP8640B/spectrum analyzer setup I used the R-390A, attenuator and outboard audio voltmeter - as you would for a receiver phase noise measurement. Comparisons were made of the 8640B, 3325A and VFO at 5, 10 and 20 kHz offset. Results were very close to those made with the other setup. At 5 kHz offset (closer in than I ran with the other setup) the 3325A was about 12 dB noisier than the 8640B and the VFO was 2 dB noisier than the 8640B.  Again, removing the diode made absolutely no difference.

So what am I missing here, guys?

BTW, both 8640Bs exceed their performance test spec for SSB phase noise and SSB broadband noise floor.







 

[w3jn]

I'd call it good, then 

[WA1GFZ]

My guess the voltage across the diode is small enough that it never conducts so it is waiting to act as a clamp if anything goes wrong. I've been looking at vCOs today and their phase noise spec.Also want to drag out my old wideband oscillator made from a bc 221 tuning cap and see how well it works. Most VCO specs I have seen that work in my RX have phase noise spec -85 to -90 dBc at 1 KHz. The wider the tuning range the more noise they make.
Some of my synthesizer modules have a connector with control signals going nowhere. I suspect they made a special oscillator board that had course tuning and used those signals. The closest I have seen to this configuration is all the control parts areinstalled but no oscillator board. I would love to drop my close in noise 10 dB. Now that i can see the noise real time with the flex software i will be able to play and see my results easily.
It will be interesting to see Jay's Phase noise when the air core inductor is installed. I take it the post man didn't deliver the package yet.

[W1VD]

No joy on the package Frank...maybe tomorrow.

Not anticipating that the Miniductor will improve the phase noise much -  expecting to see same ballpark Q as the -6 core. (Found an interesting paper on line correlating Q to phase noise but seem to have misplaced it). Where the Miniductor should help is frequency stability with wide temperature changes. While the -6 material isn't bad the Miniductor should be better.

Good luck on the hunt for a clean VCO.
   

[W1VD]

Some scope pix of the gate...



Gate without diode 5V/div dc coupled 





Gate with diode 2V/div dc coupled

You can't tell from the poor quality pix but the waveform with the diode is slightly cleaner - there's no sign of diode conduction.

[WA1GFZ]

Jay,
The gate swing is a lot greater without the diode. I wonder if the source current is near cut off or saturation. I bet the output is a lot higher. I threw an old oscillator I built years ago on the spectrum analyzer and saw the noise was pretty high. It tunes 40 to 80 mhz. The output coupling was fairly heavy and needs a buffer to make it more stable. Then I put the Racal on the spectrum analyzer. this is the firat time I looked at it since I nulled the phase noise. I could clearly see the loop is under damped. I went in and doubled the cap values and checked the phase noise. It now looks lower so thought I would walk away and check it again in a bit. It looks like I can further increase the cap values based on the spectrum analyzer picture.
I looked at a number of oscillator specs and they all seem to be about the same for frequency and tuning range. I need to did out an old Ham Radio article to see if using a hunk of semi ridged rather than aninductor makes a cleaner oscillator. This would be a problem at 4 MHz since I think you need about a quarter wave length of cable. I also saw running an oscillator at 2X frequency then using a divider to reduce frequency and phase noise doesn't buy you much. fc

[W1VD]

Frank

Wonder if you noticed the diode clamped the waveform to 0 VDC. Oscillator current measured .175 mA with diode and .630 mA without. That's about 7.8 mW vs. 22 mW. Frequency stability is significantly worse without the diode.

Package arrived today...2 days to go 20 miles. A 5 uH chunk of that Miniductor has a Q of 180. The -6 core measures 225. Will be interesting to try. Roger the Racal - would like to twist its knobs sometime. Could use a spec sheet with pinout on the SAWs when you get time. Thanks again for the goodies!

 

[WA1GFZ]

Jay,
I marked the bottom of the 1 GHz SAW ( the one not marked 600 MHz.)
You will need 12 to 15 volts to operate it as marked referenced to the case.
Internal 8 or 10 volt regulator.
Vt the tuning voltage usually ran 0 to 10 volts. You might go to 12 but linearity may take off. The corner pin marked Po is output and the pin next to it is the substrate ground at the output filter. I would love to phase lock a 1 gHz SAW and use it for a DDS clock. You should be able to get about plus 10 out of them.
Too bad I just found my bag of socket pins today or I would have sent you some. the only spec I remember on the 1 GHz SAW was it wasn't supposed to drift more than 50 KHz over the mil temp range. All of them should have phase noise about -100 dBC at 1 KHz. Tuning should be linear. Harmonics were usually 30 to 40 dB down.
Bummer on the wire inductor. Man I would think it more than 5 UH. maybe you should run it under some hot water and make sure it is clean. Q sounds crappy.
Maybe the thing to do is wind some silver plater wire on a form.
I would think the phase noise would run higher with more DC power going through the circuit. Either way the FET current is pretty low.
Well on phase noise I did two different cap configurations in the racal synthesizer and dropped the phase noise by 10 to 15 dB out past 1 KHz. Very close in I have a couple humps at 120 and 240 Hz. It looks like I can slow down the loop even more because it still looks underdamped and lock time is still as fast as I can press the enter button. I think a lot of these mil radios were interested in fast frequency change and didn't mind the extra noise. I dropped the loop corner by a factor of 4 and all it got was cleaner. fc

[w3jn]

If you lightly coupled the tank circuit to the gate, you'd enjoy some increased stability and you could get rid of the diode.  The diode's just yet another temperature instable component that under most circumstances will add drift.  It appears in your case, however, with the impressive stability everything is balanced and the drift caused by the diode is cancelled out by drift in other components.

[WA1GFZ]

John,
Maybe tapping down the coil would reduce the voltage swing allowing a higher operating Q. The Racal VCO does this.