extending range of 2ghz+ VCO

[colin]

Hi,
Im trying to cover a large frequency range with a PLL possibly with
multiple VCOs or using switched inductive element.

A YIG element would probably be ideal but id rather stick to using varicap
diodes. small 2ghz vco modules are available but tend to cover a narrow
range.

Below 200mhz im using a dds, above this the dds becomes the reference for a
PLL, a simple colpits vco can get a 3:1 frequency range from 200mhz to
600mhz using coil type inductor and varactors, however at increasing
frequencies it becomes much more dificult to get a wide range.
600mhz-1800mhz colpits vco using microstrip is just about working but im
finding it difficult to get it stable over the entire range. I'm using
bfp420 25ghz bjt with common collector, and bby39 varicap as the 2
capacitors, circuit is made as small as is possible.

With a longer microstripline its stable over a lower frequency range and
with a cap at the end of the longer line instead of a short, it is stable
over a higher but shorter range.

I was therefore thinking of putting a varactor at the end of the line wich I
believe turns it into a Clapp oscillator, and is similar to the circuit wich
I found here recently :-
http://www.eettaiwan.com/ARTICLES/2002SEP/A/2002SEP20_BD_MSD_HBM_MAT_AN.PDF
It could also be forward biased to turn it into a short or to switch to
progresivly longer transmision lines.

This led me to thinking about if it was possible to use several varactors
along the microstrip line with different voltages to effect a seemlessly
wide range. playing about with it in the simulator seems it might take some
time, I was wondering if anyone would know if it would be feasable or other
tips for a wide range ~ 2ghz vco ?

The signal is used to modulate a LASER diode for a hetrodyne LIDAR, I'm not
sure what the practical upper limit on modulation is but so far works at
1ghz.

thanks
Colin =^.^=

 

[Steveo]

Colin,

It looks like you are using off the shelf components to do your design,
but if you are working in a CMOS process and fabricating a chip, I
would look into the recently developed DCO's from TI. They have
written several papers on them, which use MOSFETs to 'digitize' the
frequency range by roughly doubling the width in the MOSFETs (which
roughly doubles the capacitance each gives). Using dithering (see
their journal papers) you can cover a wide frequency range at very
precise intervals.

However, this will probably not apply to you if you are using strictly
BJT's, which is more common in off the shelf components, which is what
I think you were mentioning here.

- zielstep

 

[John Larkin]

Hi,
Im trying to cover a large frequency range with a PLL possibly with
multiple VCOs or using switched inductive element.

A YIG element would probably be ideal but id rather stick to using varicap
diodes. small 2ghz vco modules are available but tend to cover a narrow
range.

Below 200mhz im using a dds, above this the dds becomes the reference for a
PLL, a simple colpits vco can get a 3:1 frequency range from 200mhz to
600mhz using coil type inductor and varactors, however at increasing
frequencies it becomes much more dificult to get a wide range.
600mhz-1800mhz colpits vco using microstrip is just about working but im
finding it difficult to get it stable over the entire range. I'm using
bfp420 25ghz bjt with common collector, and bby39 varicap as the 2
capacitors, circuit is made as small as is possible.

With a longer microstripline its stable over a lower frequency range and
with a cap at the end of the longer line instead of a short, it is stable
over a higher but shorter range.

I was therefore thinking of putting a varactor at the end of the line wich I
believe turns it into a Clapp oscillator, and is similar to the circuit wich
I found here recently :-
http://www.eettaiwan.com/ARTICLES/2002SEP/A/2002SEP20_BD_MSD_HBM_MAT_AN.PDF
It could also be forward biased to turn it into a short or to switch to
progresivly longer transmision lines.

This led me to thinking about if it was possible to use several varactors
along the microstrip line with different voltages to effect a seemlessly
wide range. playing about with it in the simulator seems it might take some
time, I was wondering if anyone would know if it would be feasable or other
tips for a wide range ~ 2ghz vco ?

The signal is used to modulate a LASER diode for a hetrodyne LIDAR, I'm not
sure what the practical upper limit on modulation is but so far works at
1ghz.

thanks
Colin =^.^=

Several people make stock 1-2 GHz VCOs, Emheiser for one.

http://www.emhiser.com/EmhiserMT/VCOs/VCODetail.asp?VCOID=192

What sort of laser are you using? And what's the modulation driver
circuit like?

I recently discovered a VCSEL laser that modulates great at high
frequencies but badly at *low* frequencies. It behaves like there's a
very slow PIN diode in series with the lasing junction.


John

 

[colin]

Several people make stock 1-2 GHz VCOs, Emheiser for one.

http://www.emhiser.com/EmhiserMT/VCOs/VCODetail.asp?VCOID=192

Thanks, that looks interesting, I didnt find any thing like that available
from usual supliers (RS, Farnell, ebay)
I wonder what circuit they use ... im also curious as to what they use in
wide range microwave signal generators, I gues they used to use yig in some.

What sort of laser are you using? And what's the modulation driver
circuit like?

Im using a cheap 5mw std red laser diode such as rlt6305g. the spec is very
limited. The mod/driver circuit originaly was designed for use at 10mhz but
worked fine at 100mhz, A common emitter bfg591 drives the laser, it has upto
about 3vpk to pk on the collector, it needs a pull up to drive it over
100mhz, the main problem seems to be the inductance of the leads combined
with the capacitance of the diode itself wich is probably 100pf.
I measure about .5v pk to pk at 1ghz on the case of the diode wich is
suposedly earthed, however at the moment the diode is socketed but I am
considering soldering it to the ground plane .. but im not sure if the diode
will survive.

I recently discovered a VCSEL laser that modulates great at high
frequencies but badly at *low* frequencies. It behaves like there's a
very slow PIN diode in series with the lasing junction.

strange ... I have heard that its best to avoid 100% modulation as the diode
takes a finite time to start lasing properly again.

The detector circuit is an APD with a bias modulated with a LO of similar
frequency.

I may be able to get away with a smaller relative bandwidth if I can get the
frequency high enough wich will alow for much easier tuned stages.


Colin =^.^=

 

[Mark]

It is difficult to make a VCO cover much over one octave with
continuous tuning.... it can be done with switching however... for an
octave in frequency you need a 4:1 cap change...look at the hyper
abrupt tuning diodes...

Mark

 

[John Larkin]

strange ... I have heard that its best to avoid 100% modulation as the diode
takes a finite time to start lasing properly again.

A good VCSEL can go from zero current to full-on with about a 100 ps
risetime and well under 10 ps RMS leading-edge jitter, if you drive it
with a clean, fast edge. The part I'm talking about here is apparently
*not* a good VCSEL. Unfortunately, we have a few bags full.

John

 

[colin]

It is difficult to make a VCO cover much over one octave with
continuous tuning.... it can be done with switching however... for an
octave in frequency you need a 4:1 cap change...look at the hyper
abrupt tuning diodes...

Yes as you point out frequency ratio is sqrt of capacitance ratio,
however some diodes such as the bb132 (now seemingly obsolete) has a cap
ratio of 26:1 wich is quite impresive,
wich makes tuning well over an octave quite easy, at least at lower
frequencies
however as it is listed as a vhf diode, presumably not designed to work so
well at 2ghz.

many diodes have a ratio of well over 4:1 usuualy 8:1 even for uhf/microwave
types,
the hyper abrupt diodes make it more convenient as they use lower voltage
<10v
however the high voltage diodes seem to perform as well if not better,
they just need a 30v supply.

I think the problem I am running into is the Q falls off at one end or the
other,
the impedance of the C/L falls at lower frequencies and with a very small
value inductor
the 1ohm diode resistance of my bby39 cuases a problem.

maybe i need to look for a better diode, however this had the lowest
capacitance.

Colin =^.^=

 

[colin]

A good VCSEL can go from zero current to full-on with about a 100 ps
risetime and well under 10 ps RMS leading-edge jitter, if you drive it
with a clean, fast edge. The part I'm talking about here is apparently
*not* a good VCSEL. Unfortunately, we have a few bags full.

What frequencies are we talking about ? and what does the spec say ?
maybe at high frequency the junction does not have enough time to lose all
of its carriers,
and therefore lases quicker,
at low frequencies the junction may become 'stone cold'
maybe its worth adding a small bias current ?

I had assumed I would need a better LASER as the spec does not mention diode
capacitance,
however it seems to acheive a high modulation at 1ghz.


Colin =^.^=

 

[vasile]

Could you define please exactly what modulation signal it's required
for the LASER first ?
RTL6305 can be driven at 1Ghz ? (it hasn't any modulation specification
on datasheet).
So how did you measured that it works ?
A complete datasheet of this laser diode will help us to understand
what do you mean (and satisfy also our curiosity)
Then check if the YIG is not too lazy (as it usualy is) for your
application.
Then try to see what phase noise is admitted for your signal, because
40MHz-1Ghz (2Ghz) range is the range covered by the syntesizer chip
from any UHF tuner (or satelitte tuner), and could be much easy
homebrewed for the RF out than any homemade VCO.

More than that, there are syntesizers up to 1Ghz without DDS (which
have a poor SFDR and required good designed and manufactured filters on
the output) in single chip. And there are cheap x2 multipliers (which
of course are adding noise).

And BTW, you need probably some output power too...

greetings,
Vasile

 

[colin]

Could you define please exactly what modulation signal it's required
for the LASER first ?

The modulation for the laser is for range finding (LIDAR), just a continous
but variable frequency, the returned phase delay is used to determine range,
a higher frequency allows better resolution, a variable frequency is needed
for range >1 wavelength.
It uses an optical hetrodyne receiver.

RTL6305 can be driven at 1Ghz ? (it hasn't any modulation specification
on datasheet).
So how did you measured that it works ?
A complete datasheet of this laser diode will help us to understand
what do you mean (and satisfy also our curiosity)

http://www.roithner-laser.com/All_Datasheets/Laserdiodes/RLT6305MG.pdf

Ive had the system running at 1ghz so far with this diode. It hasnt worked
so far at 2ghz but I think this is due to problems with layout, receiver and
diode capacitance etc. I need to sweep from 1ghz to 2ghz to see where it
starts to fall down.

Then check if the YIG is not too lazy (as it usualy is) for your
application.

I have a HP YIG from an old analyser, its rather large though and I would
need 2, I havnt tried it as yet though. if possible I would like to use
simpler varactors.

Then try to see what phase noise is admitted for your signal, because
40MHz-1Ghz (2Ghz) range is the range covered by the syntesizer chip
from any UHF tuner (or satelitte tuner), and could be much easy
homebrewed for the RF out than any homemade VCO.

Phase noise is a problem at the moment, I look at the beet frequency of 2
VCOs 1mhz apart and its not too pretty, this might be imrpoved with a new
layout with better supply/shielding etc.

My first step was to look at an old elc1043 tuner, however using a simple
colpits with good layout and 5ghz transistor its quite easy to get up to
1ghz with a 3:1 range, 1.5ghz+ is where it seems to be dificult to get the
parasitics small enough. maybe I need to switch from 0603 to 0402 smd
packages.

Satelitte tuners look more interesting.

I have a number of >1ghz oscillators but with very limited range, but im
looking to do it with as few as possible.

Like I said,.. idealy I would like to switch in different inductive elements
to just one VCO, perhaps seemlessly.

More than that, there are syntesizers up to 1Ghz without DDS (which
have a poor SFDR and required good designed and manufactured filters on
the output) in single chip. And there are cheap x2 multipliers (which
of course are adding noise).

wide range tracking filters for multipliers etc is quite a hassle.

And BTW, you need probably some output power too...

Well I have 2 stages of amplification and I just obtained a 45ghz 600ma
driver transistor for the laser driver.

Colin =^.^=

 

[vasile]

Hi Colin,

The modulation for the laser is for range finding (LIDAR), just a continous
but variable frequency, the returned phase delay is used to determine range,
a higher frequency allows better resolution, a variable frequency is needed
for range >1 wavelength.
It uses an optical hetrodyne receiver.

That sound nice, it's homebrewed too ?

http://www.roithner-laser.com/All_Datasheets/Laserdiodes/RLT6305MG.pdf

I've seen this datasheet but it's blind (for me). There is no frequency
modulation specification inside...

Ive had the system running at 1ghz so far with this diode.

and which is the distance range you could sense ?
what accuracy ?

It hasnt worked

so far at 2ghz but I think this is due to problems with layout, receiver and
diode capacitance etc. I need to sweep from 1ghz to 2ghz to see where it
starts to fall down.

Did you supplied the modulator with a microstrip or stripline and
adapted impedances on both ends or just a quick and dirty shielded
cable ?

What can you see on the PIN control photodiode ?

I have a HP YIG from an old analyser, its rather large though and I would
need 2, I havnt tried it as yet though. if possible I would like to use
simpler varactors.


Phase noise is a problem at the moment, I look at the beet frequency of 2
VCOs 1mhz apart and its not too pretty, this might be imrpoved with a new
layout with better supply/shielding etc.

Unfortunately the phase noise depends also by the VCO schematic.
I've bite deeply on this problem using a PLL clock distributing circuit
from TI
which distroyed the phase response from -140dBc/10Khz to about
-80dBc/10Khz .
I think your solution will be a fractionar/integer PLL .
Take a look here:
http://www.national.com/ds/LM/LMX2486.pdf#page=32

My first step was to look at an old elc1043 tuner, however using a simple
colpits with good layout and 5ghz transistor its quite easy to get up to
1ghz with a 3:1 range, 1.5ghz+ is where it seems to be dificult to get the
parasitics small enough. maybe I need to switch from 0603 to 0402 smd
packages.

Switching from 603 to 402 or 201 does not help too much as long they
aren't microwave components. Often suppliers (like Digikey) are mixing
microwave with general purpose components. The result is a mess because
you don't know which is ok and which is not without having a reference.
At least L and C should be verified SMD for RF purposes in Ghz range.
0402 has resonable dimension for homebrew under microscope.

Satelitte tuners look more interesting.

I have a number of >1ghz oscillators but with very limited range, but im
looking to do it with as few as possible.

Like I said,.. idealy I would like to switch in different inductive elements
to just one VCO, perhaps seemlessly.


wide range tracking filters for multipliers etc is quite a hassle.

yes that's true, but that depends if the images are bothering or
not.

 

[colin]

Hi Colin,


That sound nice, it's homebrewed too ?

Yes I made one some time ago wich worked at 10mhz rf/455khz if

I've seen this datasheet but it's blind (for me). There is no frequency
modulation specification inside...

Yes indeed, maybe they dont expect it to be modulated ..?

and which is the distance range you could sense ?
what accuracy ?

Well at 1ghz 90' phase shift is 37.5 mm, assumming a modest 1% resolution
this would be .375 mm,
I have no optics at the moment. 0.01 mm would be a challenging target.

It hasnt worked

Did you supplied the modulator with a microstrip or stripline and
adapted impedances on both ends or just a quick and dirty shielded
cable ?

Ive tried to avoid having to match impedances at the diode by avoiding long
signal lines.
the diodes are hardly well behaved 50ohm impedances so would be hard to
match.
Ive tried to compensate for the capacitance with a microstrip inductor.

What can you see on the PIN control photodiode ?

I just use it to monitor the average power,
I would gues any signal there is mostly due to crosstalk. it might be
usefull to look at it if it gives a more reliable phase relationship to the
output.

Unfortunately the phase noise depends also by the VCO schematic.
I've bite deeply on this problem using a PLL clock distributing circuit
from TI
which distroyed the phase response from -140dBc/10Khz to about
-80dBc/10Khz .
I think your solution will be a fractionar/integer PLL .
Take a look here:
http://www.national.com/ds/LM/LMX2486.pdf#page=32

Im using the lmx2434 as it can handle both VCOs, I tried the lmx2486 too.
I think I might go back to using single PLLs to avoid cross interference.
I hate those thin legless packages too.
Ive managed to avoid my pulling problem I had before by mounting the VCOs on
seperate boards,
with good cables/smc connectors 3db matching pads at each end etc.
Its all too easy to lose phase noise performance.

Switching from 603 to 402 or 201 does not help too much as long they
aren't microwave components. Often suppliers (like Digikey) are mixing
microwave with general purpose components. The result is a mess because
you don't know which is ok and which is not without having a reference.
At least L and C should be verified SMD for RF purposes in Ghz range.
0402 has resonable dimension for homebrew under microscope.

Actually I hadnt thought of looking for componets specd for microwave
performance,
I just assumed the parasitic LRC pretty much inevitable.

yes that's true, but that depends if the images are bothering or
not.

I think images could affect linearity/acuracy, but not sure to what extent.

Colin =^.^=