VCO tug of war (A tale of two VCOs)

[colin]

Hi,
I have two 2.1ghz VCOs (smv2100l) wich need to be 455khz apart over a 200mhz
range.
I made a pcb to test how they worked with manual tuning and a diode type
mixer to observe the beat frequency,
They work fine as long as they are more than 2.5mhz apart, if I try and tune
them closer together than this they tend to pull each other to the same
frequency, 500khz is just about as close as they can get before this hapens,
but it is extremly jittery.

also the waveform is not at all sinewave shaped anymore, Im not sure if this
is due to harmonics in the VCO or distortion from the mixer or as I suspect
cuased by the pulling.

To try and reduce pulling I improved the layout considerably ... doubled the
number of vias and decoupling capacitors, used two 10db pads on each VCO,
put copper tape on the edges of pcb and over the VCOs and used a dual gate
mosfet as a mixer, however the pulling seems to be even worse certainly not
better, not being able to manualy get closer than 2mhz, I also have a lm2434
dual PLL wich manages to keep the VCOs apart but the phase noise is realy
bad indeed, especialy with a low compare frequecny to get the resolution.

Im wondering if its worth trying to improve the layout even more or PLL loop
etc or if I should move the two VCOs to seperate boards, or maybe just
better VCOs ? or .. should I use a different method like using a dual
quadrature mixer to add 455khz to just one VCO or maybe a dual dds and freq
multipliers ?

The frequencies are being used for hetrodyne LIDAR.

Colin =^.^=

 

[John Larkin]

Hi,
I have two 2.1ghz VCOs (smv2100l) wich need to be 455khz apart over a 200mhz
range.
I made a pcb to test how they worked with manual tuning and a diode type
mixer to observe the beat frequency,
They work fine as long as they are more than 2.5mhz apart, if I try and tune
them closer together than this they tend to pull each other to the same
frequency, 500khz is just about as close as they can get before this hapens,
but it is extremly jittery.

also the waveform is not at all sinewave shaped anymore, Im not sure if this
is due to harmonics in the VCO or distortion from the mixer or as I suspect
cuased by the pulling.

To try and reduce pulling I improved the layout considerably ... doubled the
number of vias and decoupling capacitors, used two 10db pads on each VCO,
put copper tape on the edges of pcb and over the VCOs and used a dual gate
mosfet as a mixer, however the pulling seems to be even worse certainly not
better, not being able to manualy get closer than 2mhz, I also have a lm2434
dual PLL wich manages to keep the VCOs apart but the phase noise is realy
bad indeed, especialy with a low compare frequecny to get the resolution.

Im wondering if its worth trying to improve the layout even more or PLL loop
etc or if I should move the two VCOs to seperate boards, or maybe just
better VCOs ? or .. should I use a different method like using a dual
quadrature mixer to add 455khz to just one VCO or maybe a dual dds and freq
multipliers ?

One vco and some sort of offset mixer might be easier. I've seen
well-shielded crystal oscillators, on separate boards in a card cage,
lock with enthusiasm, and precision xo's have been known to lock from
being in the same room.

The waveform distortion is probably caused by the variations in lock
gain around the phase range, sort of a wobbly lock.

John

 

[[email protected]]

One vco and some sort of offset mixer might be easier. I've seen
well-shielded crystal oscillators, on separate boards in a card cage,
lock with enthusiasm, and precision xo's have been known to lock from
being in the same room.

The waveform distortion is probably caused by the variations in lock
gain around the phase range, sort of a wobbly lock.

John

Christiaan Huygens noticed this first in 1665, except with much
slower oscillators. John Harrison solved it in his work by placing the
'oscillators' [1] in separate rooms.

Cheers,
James Arthur

[1] Pendulum clocks. Exquisitely fine, in his case.

 

[Mac]

Hi,
I have two 2.1ghz VCOs (smv2100l) wich need to be 455khz apart over a 200mhz
range.
I made a pcb to test how they worked with manual tuning and a diode type
mixer to observe the beat frequency,
They work fine as long as they are more than 2.5mhz apart, if I try and tune
them closer together than this they tend to pull each other to the same
frequency, 500khz is just about as close as they can get before this hapens,
but it is extremly jittery.

also the waveform is not at all sinewave shaped anymore, Im not sure if this
is due to harmonics in the VCO or distortion from the mixer or as I suspect
cuased by the pulling.

To try and reduce pulling I improved the layout considerably ... doubled the
number of vias and decoupling capacitors, used two 10db pads on each VCO,
put copper tape on the edges of pcb and over the VCOs and used a dual gate
mosfet as a mixer, however the pulling seems to be even worse certainly not
better, not being able to manualy get closer than 2mhz, I also have a lm2434
dual PLL wich manages to keep the VCOs apart but the phase noise is realy
bad indeed, especialy with a low compare frequecny to get the resolution.

Im wondering if its worth trying to improve the layout even more or PLL loop
etc or if I should move the two VCOs to seperate boards, or maybe just
better VCOs ? or .. should I use a different method like using a dual
quadrature mixer to add 455khz to just one VCO or maybe a dual dds and freq
multipliers ?

The frequencies are being used for hetrodyne LIDAR.

Colin =^.^=

I think the dual DDS and frequency multipliers approach should work. I'm
not an RF guy but I am aware of one application similar to yours where a
DDS (AD9858) is followed by a single mixing stage and then cascaded
multipliers.

Come to think of it, the 9858 has some mixers on it, too. You could
probably build a board around two of them that does most of what
you want, even without any external multipliers.

--Mac

 

[colin]

One vco and some sort of offset mixer might be easier. I've seen
well-shielded crystal oscillators, on separate boards in a card cage,
lock with enthusiasm, and precision xo's have been known to lock from
being in the same room.

yes that would be easier, and I wouldnt have the limit on resolution of the
IF frequency by using 2 high freq PLL.
However im not sure how to calculate what impact any residual of the
original frequency and other sidebands would have, probably limit
sensitivity.

The waveform distortion is probably caused by the variations in lock
gain around the phase range, sort of a wobbly lock.

Yes im inclined to agree, although the hytite mixer low frequency output is
bandwidth down to DC so shldnt cuase any distortion, I havnt experienced
using it in this range before.

Colin =^.^=

 

[colin]

I've seen offset PLLs where the second loop is driven
by the first with a mixer and the offset frequency
between. The design looked sufficiently braodband,
what cannot be told about another method, the SSB.
With SSB you can make a second frequency from the
first with a fixed offset, but only if you have a
90 degree phase shifter over the band.
The carrier could be expected 20 to 30 dB down.
If the 90 degrees are not 90 degrees, the carrier
and other sideband are supressed less.

Im thinking that even if I just use one vco and generate the other frequency
from it, it might still pull the VCO somehow, I think il try to make some
duaghterboards for the vcos and see if that helps. theres a nice mixer from
AD8346 wich does the 90' for you on the LO input and produces a ssb from I
and Q inputs, it looked quite expensive when i first coinsiderd it but if it
makes an easy job of this maybe worth it, but if it still pulls the first
vco it would make it noisy or distorted.

I have thought of having 2 vcos of very diferent lower frequencies and
multiplying them by diferent ratios to get what i need ie x2 and x3,
hopefully they shouldnt be able to lock on to such widley spaced frequencies
even if they have do harmonics that are close.

Colin =^.^=

 

[Joel Kolstad]

Christiaan Huygens noticed this first in 1665, except with much
slower oscillators. John Harrison solved it in his work by placing the
'oscillators' [1] in separate rooms.

What's the coupling mechanism with clocks? Just mechanical vibrations
transmitted through their frames and then through the floorboards or some
such?

Pheromones are supposedly the coupling mechanism in menstrual cycle
oscillations "locking"...

 

[[email protected]]

Christiaan Huygens noticed this first in 1665, except with much
slower oscillators. John Harrison solved it in his work by placing the
'oscillators' [1] in separate rooms.

What's the coupling mechanism with clocks? Just mechanical vibrations
transmitted through their frames and then through the floorboards or some
such?

Yes, that's it. Curiously, they're said to lock in anti-phase.

Pheromones are supposedly the coupling mechanism in menstrual cycle
oscillations "locking"...

Yes, I've heard and observed it. Certainly true in cats, though the
girlfriend, objecting to the idea of her enslavement to others'
pheromones, cited a snow storm of contrary studies sufficient to steal
my thunder. I yielded, lacking the inclination to sort through and
evaluate, and, given the incentive: landing myself in the doghouse if I
'won.'

Cheers,
James Arthur

 

[John Larkin]

Christiaan Huygens noticed this first in 1665, except with much
slower oscillators. John Harrison solved it in his work by placing the
'oscillators' [1] in separate rooms.

What's the coupling mechanism with clocks? Just mechanical vibrations
transmitted through their frames and then through the floorboards or some
such?

Yes, that's it. Curiously, they're said to lock in anti-phase.

Pheromones are supposedly the coupling mechanism in menstrual cycle
oscillations "locking"...

Yes, I've heard and observed it. Certainly true in cats, though the
girlfriend, objecting to the idea of her enslavement to others'
pheromones, cited a snow storm of contrary studies sufficient to steal
my thunder. I yielded, lacking the inclination to sort through and
evaluate, and, given the incentive: landing myself in the doghouse if I
'won.'

Cheers,
James Arthur

Nowadays, birth-control pills contribute a strong anti-locking effect.

John

 

[Joerg]

Hello John,

One vco and some sort of offset mixer might be easier. I've seen
well-shielded crystal oscillators, on separate boards in a card cage,
lock with enthusiasm, and precision xo's have been known to lock from
being in the same room.

Hmm, I have never had that happen. Maybe the oscillators I design are
lacking communication skills

So, is there a Yahoo Personals for oscillators?

 

[John Larkin]

Hello John,


Hmm, I have never had that happen. Maybe the oscillators I design are
lacking communication skills

Nah, your crystal oscillators just drift too much. The more accurate
they are, the fewer nanovolts it takes to lock them.

John

 

[Joerg]

Hello John,

Nah, your crystal oscillators just drift too much. The more accurate
they are, the fewer nanovolts it takes to lock them.

I did have to run some close together at times. The closest ones were a
two-tone test at 7.000MHz and 7.001MHz (pre-historic, before the advent
of DDS). And they were clean like whistles. Checked it by dividing down
and then listening to the harmonics beating with WWV.

 

[Paul Probert]

Hi,
I have two 2.1ghz VCOs (smv2100l) wich need to be 455khz apart over a 200mhz
range.
I made a pcb to test how they worked with manual tuning and a diode type
mixer to observe the beat frequency,
They work fine as long as they are more than 2.5mhz apart, if I try and tune
them closer together than this they tend to pull each other to the same
frequency, 500khz is just about as close as they can get before this hapens,
but it is extremly jittery.

....

Colin,
It seems to me the route for coupling is most likely through your
mixer that you're using to look at the beat frequency. Besides the 10 dB
pads, an amplifier with a good isolation might help. Also, you could
look at the beats using a hybrid combiner feeding a well matched
detector, rather than using the mixer.

Paul Probert
University of Wisconsin

 

[Mike Monett]

I did have to run some close together at times. The closest ones were a
two-tone test at 7.000MHz and 7.001MHz (pre-historic, before the advent
of DDS). And they were clean like whistles. Checked it by dividing down
and then listening to the harmonics beating with WWV.

Regards, Joerg

1KHz is way too much. Try 1Hz.

The term is "injection locking". Young hams learn about this when they
build their first regenerative receiver. It is difficult to get close to
zero beat without the oscillator dropping into lock. Later when you build
your first superhet, you have the same problem with the BFO for receiving
Morse code.

It is also a major problem with precision pll's. Multivibrator type vco's
can develop a large phase offset when crosstalk from the incoming signal
interferes with the switching point in the vco cycle. Severe crosstalk can
also force the vco into a limit cycle oscillation, which renders the pll
useless. Part of the problem is the multivibrator needs to have
widebandwidth in the comparator to give fast switching, so it responds to
the narrow switching transients from nearby logic.

An LC oscillator can reduce this effect by several orders of magnitude by
attenuating the crosstalk energy in the resonant tank. The difficulty is
now to initialize the phase of the oscillator to the incoming data in order
to reduce the lockup time. Ironically, injection locking can be used to
accomplish this, but it takes many cycles of driving the oscillator
directly with the incoming signal. An example of this is in US patent
3810234:

http://www3.sympatico.ca/add.automation/patents/3810234.htm

Regards,

Mike Monett

 

[Joerg]

Hello Mike,

1KHz is way too much. Try 1Hz.

Yeah, probably.

The term is "injection locking". Young hams learn about this when they
build their first regenerative receiver. It is difficult to get close to
zero beat without the oscillator dropping into lock. Later when you build
your first superhet, you have the same problem with the BFO for receiving
Morse code.

Actually mine was rather tolerant there. Before it locked the bandwidth
of that receiver was way below what is useful even for morse code. Wish
I still had it but my better half said it was too ugly. Believe it or
not, what brought it to grief was that the front panel rusted out (!).

When I was young and thus on a shoestring budget I made myself a
"synthesizer" that wasn't a real one. A free running oscillator that
would lock onto the harmonics of the 1MHz gate oscillator of my counter.
That was I could build a stable HF receiver because I had cut the whole
band into 1MHz slivers. Not pretty but cheap.

Regards, Joerg

 

[Mike Monett]

Actually mine was rather tolerant there. Before it locked the bandwidth
of that receiver was way below what is useful even for morse code. Wish
I still had it but my better half said it was too ugly. Believe it or
not, what brought it to grief was that the front panel rusted out (!).

When I was young and thus on a shoestring budget I made myself a
"synthesizer" that wasn't a real one. A free running oscillator that
would lock onto the harmonics of the 1MHz gate oscillator of my counter.
That was I could build a stable HF receiver because I had cut the whole
band into 1MHz slivers. Not pretty but cheap.

Regards, Joerg

You should have no problem receiving WWV, then

A popular type now is Direct to DC receivers where the local oscillator
sits on the carrier and all the gain is at AF. An example is the Tayloe
mixer, and some excellent designs such as the SM5BSZ and R2 Direct
Conversion receivers.

These need very low noise amplifiers, so much of the discussions here in
s.e.d. are very useful.

Regards,

Mike Monett

 

[Joerg]

Hello Mike,

You should have no problem receiving WWV, then

Actually I never had a problem with that, no matter where in the world.
Either 10MHz or 15MHz usually came through.

A popular type now is Direct to DC receivers where the local oscillator
sits on the carrier and all the gain is at AF. An example is the Tayloe
mixer, and some excellent designs such as the SM5BSZ and R2 Direct
Conversion receivers.

These have another nasty surprise in the bag. If just a wee bit of the
oscillator power leeks out and hits a power line that gets modulated
with all the crud on that line, courtesy of non-bypassed rectifiers in
cheap power supplies. Then the audio contains that "rat-tat-tat" and you
can't get rid of it.

On very leaky direct conversion sets you can get interesting and cool
sounds when swinging a big metal plate in front of them.

 

[Genome]

Hi,
I have two 2.1ghz VCOs (smv2100l) wich need to be 455khz apart over a
200mhz
range.
I made a pcb to test how they worked with manual tuning and a diode type
mixer to observe the beat frequency,
They work fine as long as they are more than 2.5mhz apart, if I try and
tune
them closer together than this they tend to pull each other to the same
frequency, 500khz is just about as close as they can get before this
hapens,
but it is extremly jittery.

also the waveform is not at all sinewave shaped anymore, Im not sure if
this
is due to harmonics in the VCO or distortion from the mixer or as I
suspect
cuased by the pulling.

To try and reduce pulling I improved the layout considerably ... doubled
the
number of vias and decoupling capacitors, used two 10db pads on each VCO,
put copper tape on the edges of pcb and over the VCOs and used a dual gate
mosfet as a mixer, however the pulling seems to be even worse certainly
not
better, not being able to manualy get closer than 2mhz, I also have a
lm2434
dual PLL wich manages to keep the VCOs apart but the phase noise is realy
bad indeed, especialy with a low compare frequecny to get the resolution.

Im wondering if its worth trying to improve the layout even more or PLL
loop
etc or if I should move the two VCOs to seperate boards, or maybe just
better VCOs ? or .. should I use a different method like using a dual
quadrature mixer to add 455khz to just one VCO or maybe a dual dds and
freq
multipliers ?

The frequencies are being used for hetrodyne LIDAR.

Colin =^.^=

I have a similar problem but, depending on the divider/multiplier settings,
it ends up demodulating encrypted stuff from the packet sniffer. Takes a bit
of finger waving but when you get it right you get plain text.

Strange.

DNA

 

[Jim]

Im wondering if its worth trying to improve the layout even more or PLL
loop
etc or if I should move the two VCOs to seperate boards, or maybe just
better VCOs ? or .. should I use a different method like using a dual
quadrature mixer to add 455khz to just one VCO or maybe a dual dds and freq
multipliers ?

===========================

It sounds as if there is not enough isolation between the two VCOs.
Considering that a VCO is essentually a regenretive amplifier, if there is
ANY amount of a second signal coupling into the VCO, assuming it is close to
the same frequency as to VCO is tuned to, the VCO regenratively amplifies it
and results in either a powerful spur, or possibly "pulling" the VCO to the
other frequency.

You probably need better than 30 dB isolation between the two VCOs. I'd
add attenuators and amplifiers to both VCOs to increase the isolation.
You can't just buffer and isolate one of them, the other one will be just as
happy to regenerate the spur as the first one.



Jim



--

19:15 Pacific Time Zone
May 26 2006

International Time
02:15 UTC
27.05.2006

 

[Mike Monett]

Hello Mike,

Actually I never had a problem with that, no matter where in the
world.

Either 10MHz or 15MHz usually came through.

Yes. A comb generator strong enough to injection lock the first LO
would feed other harmonics directly to the mixer. At 10MHz, a
frequency error of 1e-7 in the crystal would produce a 1Hz beat
note, interfering with the wwv signal. Of course, it also gives an
indication when the crystal is adjusted on frequency

These have another nasty surprise in the bag. If just a wee bit of
the oscillator power leeks out and hits a power line that gets
modulated with all the crud on that line, courtesy of non-bypassed
rectifiers in cheap power supplies. Then the audio contains that
"rat-tat-tat" and you can't get rid of it.

Nah. First, there should be very little leakage from the LO for the
reason described below. Second, power supply rectifiers are slooow.
Even if some LO signal did get to them, they wouldn't even see it.

Sounds more like snap recovery diodes in the power supply making a
120 Hz buzz. This has been discussed here several times.

On very leaky direct conversion sets you can get interesting and
cool sounds when swinging a big metal plate in front of them.

Self-Mixing due to LO feedthorugh. Produces a DC offset. This is a
well-known and understood problem with DC receivers. It also affects
the Binary Sampler when heterodyne mixing is used and the fequency
offset is set to zero:

http://www3.sympatico.ca/add.automation/sampler/intro.htm

There are other problems with DC receivers as well, but they are
fixable with clever design and good layout.

Regards,

Mike Monett