[SOLVED] vco exist spur near dozens of khz

Hello，
We recently designed several kinds of vco.Its structure is similar to Colpitts.Its frequency is about 3-7GHz. We have a problem that the output of the vcos have a spurious at dozens of khz when the temperature below about -40 ℃.
We have made several experiments,and found that
1.when the source voltage reduce to 4V from 5V，the spurious disappeared；
2.when we change the tuning voltage，the frequency of spurious would change；
3.we use the battery as power supply ，but the spurious is not changed.

We don't know how and where the spurious produce.
Could you help me?

We look forward to the help of experts.
Thank you !

Hi,

about impossible to help without schematic and additional detailed informatins.

1.when the source voltage reduce to 4V from 5V，the spurious disappeared

Click to expand...

Is there a linear or switching voltage regualtor involved? Maybe it´s ringing..
or a switch mode regulator is in discontinous operating mode.


Klaus

1.when the source voltage reduce to 4V from 5V，the spurious disappeared；
2.when we change the tuning voltage，the frequency of spurious would change；
3.we use the battery as power supply ，but the spurious is not changed.

Click to expand...

Suggests loop instability. Redesign loop filter.

it's not a pll,just a vco

- - - Updated - - -

this is the schematic .

and we have used the battery as power supply，but it did not work

Do you see strong spurs (due to unstable or even intermitted oscillations) or only "light" ones? How many dB below carrier? Do you see related variation of DC operation point, e.g. R1 voltage?

Do the spurs vary when changing output amplifier bias? Also observed with output amplifier powered down?

Sure there's instability in Vt generation? E.g. operational amplifier that doesn't like capacitive load?

the spur is about -20dbc.
when the output amplifier powered down ，the spurs didn't disappear，but stronger a little
the Vt is generated by E5052b，but also produce a fixed voltage using the battery

Is tuning voltage sufficiently clean ?? I think it's coming from your variable voltage supply. ( it's probably Switch Mode )

BTW, why there are 2 tuning mechanism in your circuit?? Another possibility is that one tank circuit may pull the frequency,other one may pull through opposite side and that creates a perturbation and the effect of this chaos and perturbation may create instantaneous spurs.

I‘m not sure tuning the voltage sufficiently clean，and adding the filter circuit can improve the phase noise，but the spurs do not change at all.
2 tuning mechanism is to increase bandwidth.We have a try that fix one tank circuit ，the spur is the same.

mwvco1 said:

I‘m not sure tuning the voltage sufficiently clean，and adding the filter circuit can improve the phase noise，but the spurs do not change at all.
2 tuning mechanism is to increase bandwidth.We have a try that fix one tank circuit ，the spur is the same.

Click to expand...

I suggest you to filter your tuning voltage using with Ferrite Bead ( or equivalent Ferrite based Filter-see Murata) or a small EMI filter which is used in SMPS or similar sources.
And...
Apply tuning voltage absolutely through Coaxial Cable with a preferably SMA connector.This is important..

Have you put the tuning voltage through a bias-T,
to the spectrum analyzer?

Lots of power supplies these days are switcher
based, maybe there's just enough ripple on the
E5052 output to cause trouble. I have a rack of
cheapo (Dr. Meter) supplies that are switcher
based and they look clean down to 5mV or so,
but in RF you can care about a whole lot lower
than that. -20dBc doesn't tell me what that is,
since I don't know the carrier amplitude.

You could try using a battery on Vtune as well
(maybe with a trim-pot to adjust, and a solid cap
to RF ground on the wiper terminal). Make it all
batteries, no connection to anything but the
spectrum analyzer, and then start adding noise-
makers back in until you see the spurs again.

And if they didn't go away, then you know it's in
your board.

When you go to all-battery be sure to shut down
all the line powered equipment on the bench, in
case it's radiated susceptibility that's the problem.
Might consider trying to find a screened room or
make a little RF cage for your testing.

the output power is about 10dbm,so the spur is about -10dbm.
thank you for your reply.We will have a try according to your suggestion.
Now I can't understand why this phenomenon occurs only at low temperature，and the spurs become strong as the temperature decreases？

I can't understand why this phenomenon occurs only at low temperature

Click to expand...

Me neither. Are you sure that the transistor is operated safely within maximum voltage ratings? Same question about varactor diodes.

What's the transistor fT? Can it be that you have additional parasitic oscillations not observed by your SA?

One thing that changes a lot with temperature is BJT
saturation storage / recovery time. If you are driving
the BJT into saturation (either the oscillator or the
buffer) you might see weird self-modulation beat
behaviors (amplitude drops when saturated, rises
when out of saturation, and this could be cyclic /
bursty or coupled back from buffer to oscillator in
an "outer loop").

BJT gain also drops with temperature. Maybe changing
the relation of "outer loop" to "inner loop" (osc) or the
amount of bias current you are throwing to get the
amplitude or frequency you're after, or something.

Freeze spray or heat gun locally applied device
by device might show you location(s) involved.

hi，
today we have some experiment.
1.the BJT (both the oscillator and the buffer) work in the saturation
2.BJT gain rises as temperature decreases
3.the output S11 is bad, the oscillating part is only -3db and the buffer is about -10db
someone said that is probably a self-mixing,I couldn't understand the mechanism，do you know this？

RF circuits do not want saturated BJTs. BJT saturation
is bad (unlike MOS saturation, which is almost its opposite).
I'd guess you're throwing current at the BJTs trying for
more gain@frequency, but this can only go so far. More
collector voltage headroom is a thing to try. Less base
current is another.

BJT's drop in hFE with cold and run slower unlikemCMOS which runs faster when cold towards -50'C from experience and text books. Raising Vce also increases hFE so your DC bias may be resonating.

Try 1pF increments across Vce and

if layout looks like schematic, oh oh. L's are not R's They are L's with DCR and SRF.

- run simulation with all parameters for each part wher eL looks like RLC and C looks like RLC etc. Improvemyour decouplng as ESR can be a problem at high current.

While forward-active-region hFE rises w/ temp, the gain
(what's left of it) in saturation "could" go the other way
if (say) the Cjc capacitance is the at-frequency gain killer.

Whether this is the case, wants further poking (and just
get it out of saturation already, for starters). Presuming
that the designer isn't conflating MOS saturation region
with BJT saturation region, since BJT circuit design is
getting short shrift in most universities it seems.

Getting it into a decent schematic capture & simulation
tool (presuming good models can be had) would sure
ease and speed the debug - good models (and board
parasitics fidelity) being big "ifs".

If the oscillator load is really an inductor, like I make
it out to be, it's hard to see how the core BJT could
saturate at DC. Although clipping into saturation at
one end of the signal swing, certainly could happen
(and there's your harmonic distortion and possibly
LF self-modulation).