Colpitts Oscillator Just WON'T run!

[KI4YAN]

First, let me start by saying I have NEVER built a Colpitts oscillator that worked. Ever. Hartley? Fine. Grid-leak? Fine. Colpitts? Nope.

That includes the series tuned Clapp, as well, which is what I have now. It's been four days of futzing around and it is still absolutely stable in ALL conditions. The grid tank is resonant over the band of interest, the cathode follower buffer after is perfectly stable and works just fine injecting the signal from the function gennie, and if I inject the signal into the grid tank circuit, the proper amplitude signal appears at the output.

But the dang thing just won't oscillate, even if I grab a 12AT7 or 12AV7 to replace the myriad of 12AU7's I've tried.

I've tried playing with the capacitive voltage divider between the grid, cathode, and ground, varying it from 750pf each cap, to 750pf top/910pf bottom, 910pf top/750pf bottom, 1000pf/1000pf (what the schematic calls for), 350pf/350pf, etc, and it doesn't seem to matter one bit.

Circuit is a series tuned tank, Ltank=55uH, Ctank=5.9-50pf air variable, Ctrim=120pf air trimmer.

Tank couples to grid via 100pf silver mica cap, new production.
56k resistor from grid to ground.

Capacitive voltage divider from tank side of grid cap to cathode; from cathode to ground. Circuit calls for 1000pf from tank side of grid cap to cathode, and 1000pf from cathode to ground. I have used the above mentioned values.

Cathode is loaded with an 2.5mh RF choke

"Grounded Plate" oscillator configuration, cathode is directly coupled to the grid of the cathode follower stage.

Plate voltage 140v. Plates of the two triodes are tied together, fed from a common plate resistor of 1Kohm.

Schematic is from RCA Ham Tips, Volume 13, No. 2, June-July 1953.

What am I doing wrong, guys?

[N2DTS]

Sounds like the circuit I use in my homebrew RX for the local osc.

In my case, I used a 6C4 triode with an RF choke from the cathode directly to ground.
Between the grid and ground I have two 200 PF NPO caps in series, with the center between both going directly to the cathode.
In series with the grid to the caps is a 50 ohm resistor, then a 15k ohm resistor with a 50pf cap across it.

Across the two 200 PF caps on the other side is my coil (B+W coil stock) and a 20 PF variable cap (tuning) in parallel, not series.

RF comes off the grid of the 6C4 directly to the mixer grid, the plate of the 6c4 goes directly to 100 volts with a .01 uf to ground.

I tried building all sorts of circuits for an LO and could not get anything else to work.
This circuit works down to about 5 volts on the plate of the 6C4!

I used the same circuit for two receivers and my VFO/exciter.
I tap the coil and short it to go between 160, 80 and 40 meters, with no other changes needed.
Its very stable!

[KI4YAN]

Here's the current circuit...exactly as constructed by the RCA Ham Tips article, except for the tank circuit values...event the operating range is the same! I used the best air variable I had for the tank cap...it was a 5.9-50pf instead of the 150pf-ish one called for, so I adjusted the trimmer capacitor to fix it.





The article calls out the cathode choke as a 0.5mH National R-50. Is there something special here I don't know about?

EDIT:

goofed the schemo, forgot the bypass caps from the plate of the 12AU7's to ground, 0.001uF, one above the 1K plate resistor, and one below.

[WD5JKO]

  I attach a scanned page from the 1970 Bill Orr Radio Handbook.

Bill nails the issue in good detail. The caps, C2, C5 at .001uf are too big. Better yet go to a higher Gm tube

My recommendation would be to replace that 12AU7 with a 6U8A, and let the oscillator run with a sharp cutoff

pentode.

Jim
Wd5JKO

[KI4YAN]

I've considered it, but in the past getting a 6BL8 (my favorite oscillator/buffer tube) tempco'ed takes a LOT of work.

If I can't futz around with this and get it running, I'll set it up with the 6BL8.

[WD5JKO]

I've considered it, but in the past getting a 6BL8 (my favorite oscillator/buffer tube) tempco'ed takes a LOT of work. If I can't futz around with this and get it running, I'll set it up with the 6BL8.

    One issue I have found with a Clapp oscillator is the output amplitude drops as the tuning frequency goes up. In my case with a Knight kit V44 VFO, the oscillator would fail to start when keyed at the upper end of the range. It would key lower, and then tune all the way up. One possibility was that the silver mica caps in there had diminished 'Q' from age. Another was as Bill Orr points out that, the Clapp needs a high Gm tube, or the tapped Colpitts capacitors need to be reduced in value. The V44 used 780 pf. The tube used was a 6BH6.

  So looking for another low filament draw tube, I was digging for a 6AK5. I never found one, but as luck would have it, I found a 6AS6. This tube is VERY special in that G3 can be used as an input, and the characteristics are unique.

Look at the fourth schematic posted here:

http://theradioboard.com/rb/viewtopic.php?f=3&t=3714&sid=1d92f0e17add1d67444a2b1e90795b51

This is a 6AS6 based Vackar that runs class A, and uses oscillator AGC with G3 as an input for amplitude stabilization. Running the oscillator class A should yield low harmonic output, and low phase noise.

Anyway, in my case with my V44, using the 6AS6 made all the difference. The oscillator keying issue went completely away, and quite unexpectedly, the harmonic output went way down. The output to the buffer is now a near perfect sine wave. The output still drops going up in frequency, so I was considering adding the G3 AGC input from the oscillator output rectified into AGC bias.

My V44 using frequency stabilization with a Cumbria design X-Lock PIC based stabilizer. That is another story however..

Scroll down to my Knight V44 listing:
http://www.cumbriadesigns.co.uk/x-lock_solutions.htm

Jim
WD5JKO

    

[N2DTS]

Its amazing how many slightly different designs there are of this one basic circuit.

I spent a LOT of time trying various circuits, and the one I use was the only one I could get working.
Output levels were very stable over frequency, with the ability to cover 160 to 40 meters directly, and it would work down to a very low plate voltage, something like 5 volts on the plate and it would start and run fine!

I would start with a design out of something that works, like an old rig (dx100, etc) since you KNOW it works.
Project books, you really do not know if it would really work.
I do not think I ever saw that design in a working rig....

It might just be something to get people to buy a bunch of parts....or someones untested idea...
 

[w4bfs]

take a look at TMC GPT750 vfo .... I have redone mine as a 12at7 colpittts osc / cathode follower .... similar to Ham tips ckt ... very stable and works down to 35 v b+

[KA2DZT]

The 1000pfd caps are not right, too big.  May be some error in the schematic.  The 1000pfd across the cathode choke is just grounding any feedback.

Fred

[KI4YAN]

Most of the Colpitts circuits I have found call for caps that size...I bet they don't run either.

I cut the caps down to 350pf and it took right off...plenty of 2nd harmonic and a little 3rd, which is fine...it's a multiplier type VFO, with the grid tank on 1.7-2.0mc and the first multiplier stage makes the 3.5-4.0mc output and starts the multiplier chain to get 80, 40, 20, 15, and 10.

It's the multiplication that makes it so critical to get the tempco worked out.

[N2DTS]

If you have an sdr, you can check the frequency, range, stability and output of the design, and can change values and see the effects.

In all my tests, the best coil ended up being B+W coil stock, as big and heavy as you can make it.

[ka4koe]

Try decreasing the value of the 100 PF next to the grid leak resistor. I had similar issues with the Hartley oscillators in my theremin not starting reliably. Refer to Termen, Radio Engineer's Handbook, 1943 Ed, Page 483. Those two components form an RC time constant. If tau is too large the oscillator may not start. Reduce the time constant by lowering the value of C, as lowering the resistor will change the biasing point.

[WD5JKO]

Most of the Colpitts circuits I have found call for caps that size...I bet they don't run either.

I cut the caps down to 350pf and it took right off...plenty of 2nd harmonic and a little 3rd, which is fine...it's a multiplier type VFO, with the grid tank on 1.7-2.0mc and the first multiplier stage makes the 3.5-4.0mc output and starts the multiplier chain to get 80, 40, 20, 15, and 10.

It's the multiplication that makes it so critical to get the tempco worked out.

   My Knight V44 oscillator runs on 160M, and the VFO out is doubled to 80m before it leaves the box. From there frequency multiplication in the transmitter is required to get 40-10m. This situation is similar to your Colpitts.

  Some considerations for a VFO that works this way:

* Drift is multiplied along with the frequency...4x on 40m, 8X on 20m, 12X on 15m, and 16X on 10m! So a measly 300hz drift at the oscillator is 5 Khz on 10m!

* Any oscillator jitter is multiplied as well making any sideband crud either side of the the carrier get wider band by band going up.

* Any keyed oscillator has some amount of RF heating upon key up causing a shift. With a stock V44, that shift is about 400hz on 20m (50 hz at the oscillator). This key up shift takes about 5 seconds to move (up in my case) and then the normal drift continues even when warmed up, about 1 khz / hr on 20M (8X).

  It would be fun to try some of the ideas presented by Brett (big oscillator coil), or Phil (play with the C-R grid leak time constant). Also any ripple on the oscillator B+ will amplitude modulate the VFO. So bypassing the heck out of the oscillator B+ cannot be over done. Also regulating the osc/buffer tube filament would not hurt either.

   With my V44 (stock version), the oscillator 3rd harmonic is strong. Then in my Gonset G-76, multiplying up, I see some spurious emission on the scope since the Gonset tuned circuits are too broad. I see distortion in the RF sine wave at a fast sweep, or at a slow sweep, the carrier has two levels, say 4 cm, and 3.8cm on the scope. This varies with the Gonset final grid tuning capacitor. On my heavily modified V44, the VFO output is much cleaner, and the Gonset output is making a carrier that is pure. My point here is to look at the osc 3rd harmonic as you multiply up to 40M and above. You are likely going to be transmitting crud
at a surprisingly high level.

  One way to deal with that 3rd harmonic is to use it! For example, for 40-10m operation, run your VFO at 1/3 of 7-7.3Mhz instead of 1/4th. This will lower the multiplication factors, and the final transmitted output will be cleaner. The feel of the VFO tune knob will also be more precise since the tuning will not be so coarse.

   Fun stuff!

Jim
Wd5JKO
 

[N2DTS]

Things that can be done with keying drift:

Leave the vfo tube run all the time, if its not on frequency, it will not bother anything.
Since I wanted a digital freq readout, I had to have it run all the time...

Run low voltage on the vfo tube.
I run mine at 100 volts, they work down to under 10 volts on the plate.

Use a small cap to couple the vfo to the buffer.

There was also a design in the old QST's that had a circuit that changed the bias slightly for keying, the tube drew plate current all the time, but something was changed (slightly) on keying.


I get a slight change on key down, from the overall voltage sag I guess as the rig draws a lot of power off the 120vac line.
Or maybe its the buffer getting voltage.
Its not enough for me to worry about it..

[KI4YAN]

I won't be keying the oscillator...it's brilliant as it is.

Still don't have it on frequency quite yet...tunes exactly 1.7999Mc to 2.1010Mc, and after a 15 minute warmup, it's been sitting on the bench at 2.0001Mc for four hours...and has drifted down from 2.0005Mc.

I'm quite afraid to sneak it down that last hundred kilocycles...but we'll have a go at it anyway.

Update: since I typed this, it's drifted back up to 2.0003Mc, and the counter updates every second. Power-On drift is less than 40Hz!

[KI4YAN]

Ugh, I hate it when I read the counter wrong.


Drift is about -100Hz/hr, power on drift is about -400Hz...

So far, all silver-mica fixed caps, but the oscillator wants to drop out at the top of the band. working on that part first, as I work my way down in frequency.

[WD5JKO]

Ugh, I hate it when I read the counter wrong.


Drift is about -100Hz/hr, power on drift is about -400Hz...

So far, all silver-mica fixed caps, but the oscillator wants to drop out at the top of the band. working on that part first, as I work my way down in frequency.

Yep, I know.....

Back a few years ago when I stabilized a LakeShore Band Hopper VFO (5-5.6 Mhz), I quickly discovered that the time base in my frequency counter was drifting almost as much as my VFO. Then I noticed that the 9 Mhz crystal in my Central Electronics 20A was also moving around a fair amount..The point being, that as you stabilize this VFO of yours, the drift in your test equipment might become a significant percentage of the total drift indicated.

So as a suggestion, perhaps turn your counter (or rcvr) on an hour or so before doing drift tests.

I have an old Fluke counter with LED display. It used a simple 10 Mhz crystal as the timebase. It drifted over 100 hz the first hour. My Icom R-75 receiver was better then that, but some drift was there too. So I ended up buying a 10 Mhz TCXO for my Fluke counter, and that pretty much nailed it. I am usually within a 1hz beat note of WWV at 10 Mhz.

As far as the VFO not always starting at the high end of the range, that Bill Orr stuff I posted talks about that. The Colpitts capacitors either need to come down a bit, or the Gm of the oscillator tube must come up..Do you have a 6DJ8 to try? This is pin for pin with a 12AU7 except for the filament where 4.5 are the 6.3V filament pins, and pin 9 is a shield.

Jim
Wd5JKO

[KI4YAN]

Swapped out for a 12AV7 it runs fine now. Double the gm of 12au7. Drift is better after a little work.

My frequency counter is a tek dc503a with option 1, the TCXO timebase. It doesn't always gate properly but I am saving up to get it repaired and cal'ed.

[WU2D]

This trick has been mentioned many times, but it is still a good "fast" method of taming a VFO with stable drift, that is, one heading south by so many Hz per minute. All you need is a small differential trimmer of say 30 pf and two ceramic caps, for instance, one a 22 pf NPO and one a 22 pf N750. The idea is to dial in compensation with the trimmer till you hit the sweet spot. In the extremes you have pure NPO and in the other N750, but in all cases you have the "same" equivalent capacitance. Obviously these three components are located inside the VFO compartment.

[WU2D]

This may help.

[Tom WA3KLR]

In most cases, the majority of the drift is caused by the inductor, so try to get the N750 capacitor by the VFO coil.

[KI4YAN]

Thanks for the tips, guys, I'll be working on this again soon. been fighting with the old car...blew the intake clean off of it earlier in the month.