Electronic tuning (high voltage varactors)
 

I'm looking for suggestions on how one might go about changing the effective
capacitance of a resonant circuit (it's part of a filter) in order to change
the center frequency. The tricky part is that there's ~30dBm (10Vpp) RF
running around, so with standard varactor diodes the RF becomes 'the bias'
and the tuning is destroyed. I'm told that there are 'high voltage'
varactor diodes out there; does anyone have a recommended source?

This is at 500MHz and the nominal component values are ~3pF. If I could
obtain a 2:1 tuning range, I'd be happy. I need perhaps 30 steps within
that range (3-6pF), and (doing the calculations) each step changes the
capacitance by little more than about 100fF to start with -- so I'm thinking
that switching physical capacitors into and out of the circuit is out of the
question here.

Thoughts?

Thanks in advance,
---Joel Kolstad

You might consider putting varactors or switched capacitors across only
part of the inductor, i.e., a tap. The voltage will be lower and the
required capacitance higher.

You might also consider ways to change the inductance instead of the
capacitance.

Q, temperature stability, repeatability, and perhaps other factors will
play a big role in determining which way is best, or acceptable.

Roy Lewallen, W7EL

Joel Kolstad wrote:

I'm looking for suggestions on how one might go about changing the effective
capacitance of a resonant circuit (it's part of a filter) in order to change
the center frequency. The tricky part is that there's ~30dBm (10Vpp) RF
running around, so with standard varactor diodes the RF becomes 'the bias'
and the tuning is destroyed. I'm told that there are 'high voltage'
varactor diodes out there; does anyone have a recommended source?

This is at 500MHz and the nominal component values are ~3pF. If I could
obtain a 2:1 tuning range, I'd be happy. I need perhaps 30 steps within
that range (3-6pF), and (doing the calculations) each step changes the
capacitance by little more than about 100fF to start with -- so I'm thinking
that switching physical capacitors into and out of the circuit is out of the
question here.

Thoughts?

Thanks in advance,
---Joel Kolstad

On Sat, 30 Oct 2004 01:22:51 -0700, "Joel Kolstad"
wrote:

I'm looking for suggestions on how one might go about changing the effective
capacitance of a resonant circuit (it's part of a filter) in order to change
the center frequency. The tricky part is that there's ~30dBm (10Vpp) RF
running around, so with standard varactor diodes the RF becomes 'the bias'
and the tuning is destroyed.

Perhaps not tuning range, but if there are strong unwanted signal, you
are going to get a lot of unwanted mixing products (this is basically
a varactor multiplier :-).

Is there a specific reason for runnig the filter at 50 ohms (as it
would appear from your figures). Why not design your filter for, say,
1-10 ohms and the voltage range would be reduced.

I'm told that there are 'high voltage'
varactor diodes out there; does anyone have a recommended source?

You might investigate any ordinary small signal signal diodes (such as
1N4148or SMD equivalents) with sufficient breakdown voltages and look
for their reverse voltage vs. capacitance curve.

This is at 500MHz and the nominal component values are ~3pF. If I could
obtain a 2:1 tuning range, I'd be happy.

That would require a 4:1 capacitance range.

I need perhaps 30 steps within
that range (3-6pF), and (doing the calculations) each step changes the
capacitance by little more than about 100fF to start with -- so I'm thinking
that switching physical capacitors into and out of the circuit is out of the
question here.

If you can switch in (using swithing diodes or reed relays) fixed
capacitances as base capacitances, this should not be a problem.

Put a small (perhaps 5 pF) fixed capacitor in series with a big
(perhaps even 5..50 pF) varactor and you can generate a moderate
tuning range. Using some back to back configuration with anodes at
ground potential and varactor varacitors at the tuning voltage, you
can also reduce the IMD products.

Paul OH3LWR

Hi Paul,

"Paul Keinanen" wrote in message
...
Perhaps not tuning range, but if there are strong unwanted signal, you
are going to get a lot of unwanted mixing products (this is basically
a varactor multiplier :-).

Very true!

Is there a specific reason for runnig the filter at 50 ohms (as it
would appear from your figures). Why not design your filter for, say,
1-10 ohms and the voltage range would be reduced.

No, a lower impedance would be fine. Still, even at 1 ohm, 30dBm is 1.43Vpp
so it's still far from 'small signal.' I do realize that going from 50
ohms - 1 ohm would get me capacitors that were 50 times as large, but on
the other hand the inductors would be 50 times as small!

You might investigate any ordinary small signal signal diodes (such as
1N4148or SMD equivalents) with sufficient breakdown voltages and look
for their reverse voltage vs. capacitance curve.

Thanks, will do.

This is at 500MHz and the nominal component values are ~3pF. If I could
obtain a 2:1 tuning range, I'd be happy.

That would require a 4:1 capacitance range.

Sorry, that was poorly worded. A 2:1 capacitance range giving a 1.414:1
tuning range would be fine. (Large tuning ranges are even better. :-) )

If you can switch in (using swithing diodes or reed relays) fixed
capacitances as base capacitances, this should not be a problem.

If you mean 'reasonable sized' (maybe 3.3pF on up) by 'base capacitors,'
that's what I was planning on doing. Initially I was planning on using reed
relays, but having looked at some Hittite switches, I'm now planning on
using a HMC221.

For ~100fF, I was thinking the problem is that you can't reliably switch in
something as small as 100fF in that it's pretty hard to build anything that
small in a repeatable manner. I mean, you can't buy a 100fF capacitor as a
chip cap -- you end up just using a pad of copper on the PCB above a ground
plane, right? Is that feasible? Perhaps if I stuck 4 such 'capacitors' in
series, each one would then be ~400fF. However, at that point I'd
presumably have to use a field solver to figure out the true capacitance
since all those copper pads are going to be pretty close together to one
another and have non-negligible coupling.

Put a small (perhaps 5 pF) fixed capacitor in series with a big
(perhaps even 5..50 pF) varactor and you can generate a moderate
tuning range. Using some back to back configuration with anodes at
ground potential and varactor varacitors at the tuning voltage, you
can also reduce the IMD products.

I'm not quite following this (I don't know what a 'varactor varacitor is'?).
Something like:

signal -- | | -- || -- gnd

Where | | is the 5pF capacitor and || is the varactor with anode facing
'left'? I'm not following how the 'back to back' configuration works.

Thanks a lot for the help; you've given me a great deal to think about!

---Joel

On Sat, 30 Oct 2004 09:59:15 -0700, "Joel Kolstad"
wrote:


I'm not quite following this (I don't know what a 'varactor varacitor is'?).
Something like:

signal -- | | -- || -- gnd

Where | | is the 5pF capacitor and || is the varactor with anode facing
'left'? I'm not following how the 'back to back' configuration works.

Something like this (fixed font):

Vt
+-+
| |
| R
| |
+----||------+---||---+
| | |
| R |
| | |
+----||---+------||---+
| |
| |
+--LLLLLLLL+LLLLLLLLL--+
|
---
Gnd

The cent er of the inductance L is grounded to get ground reference
for the tuning voltage Vt. With a small series capacitance and a large
varactor capacitance, most of the RF voltage is over the fixed
capacitor. The anti-parallel structure should reduce the distortion.
You need to bring the tuning voltages Vt to the varactors through
separate resistors R (or through inductor resistor combinations).

Paul OH3LWR

Thanks Paul, much appreciated!

"Paul Keinanen" wrote in message
...
On Sat, 30 Oct 2004 09:59:15 -0700, "Joel Kolstad"
wrote:


I'm not quite following this (I don't know what a 'varactor varacitor
is'?).
Something like:

signal -- | | -- || -- gnd

Where | | is the 5pF capacitor and || is the varactor with anode facing
'left'? I'm not following how the 'back to back' configuration works.

Something like this (fixed font):

Vt
+-+
| |
| R
| |
+----||------+---||---+
| | |
| R |
| | |
+----||---+------||---+
| |
| |
+--LLLLLLLL+LLLLLLLLL--+
|
---
Gnd

The cent er of the inductance L is grounded to get ground reference
for the tuning voltage Vt. With a small series capacitance and a large
varactor capacitance, most of the RF voltage is over the fixed
capacitor. The anti-parallel structure should reduce the distortion.
You need to bring the tuning voltages Vt to the varactors through
separate resistors R (or through inductor resistor combinations).

Paul OH3LWR


Paul,
The "back to back" configuration I am familiar with is like this:

+ Control voltage
|
R
|
+----||-----+---||---+
| |
| |
+--LLLLLLLL+LLLLLLLLL--+

There could also be a series cap in this configuration--except a resistor or
choke would be required on one of the varactors. (no different than the one
to the control line)
I don't know the advantages of either config. as the RF cycle effects each
diode the same way "on opposite half cycles" in either.

I also don't get this talk about the filter Z. Since you'd need to Z match
in/out of the filter, it seems to me the varactor voltages will be the same
for any Zin/out since this will be determined by how "tightly" they are
coupled into the resonant circuit and not the Zin/Zout, no? The Z match
will just change the Vin/out.

Steve K9DCI

On Mon, 1 Nov 2004 17:24:10 -0600, "Steve Nosko"
wrote:


The "back to back" configuration I am familiar with is like this:

+ Control voltage
|
R
|
+----||-----+---||---+
| |
| |
+--LLLLLLLL+LLLLLLLLL--+

There could also be a series cap in this configuration--except a resistor or
choke would be required on one of the varactors. (no different than the one
to the control line)

You would need fixed capacitors on _both_ anodes and also resistors to
ground from both anodes to get the DC bias. The small series
capacitors are essential, since they take the most part of the RF
voltage. Of course, the series resistor will reduce the tuning range.

You could get away with the series capacitor and put multiple (maybe
10) varactors in series. Thus, the RF voltage across each varactor
would be low. You will need a high tuning voltage, perhaps 100 V.

The worst problem is how to get the DC voltage distributed over the
varactors. The simple solution would be to put resistors across each
varactor to form a voltage divider. Apart from possible thermal noise
problems (if weak signals are also involved), the nasty thing is the
resistors will have a parasitic capacitance across the ends of the
resistor. This capacitance is in parallel with the varactor, forming a
significant base capacitance. Also the losses (and hence Q) of these
parallel parasitic capacitances may degrade the total Q of the
resonant circuit. Putting multiple varactors in series also increase
the total inductance, which would not be so nice in this case, since
the inductance levels are already low.

However, if the parasitic capacitance/diode is much less than the
varactor minimum capacitance, quite large tuning ranges could be
obtained.

I also don't get this talk about the filter Z. Since you'd need to Z match
in/out of the filter, it seems to me the varactor voltages will be the same
for any Zin/out since this will be determined by how "tightly" they are
coupled into the resonant circuit and not the Zin/Zout, no? The Z match
will just change the Vin/out.

Think about two resonant circuits coupled by a small capacitance at
the top. Connect the signal from the input line to the first resonator
using inductive coupling (transformer with untuned primary). By
selecting the number of turns on the primary, you can get any
impedance transformation ratio, so you can match the 50 ohm line to
any low impedance resonator. On the output side on the other
resonator, you can do the opposite with the other transformer and
restore the impedance to 50 ohms for the output line.

To reduce the Z in a resonator, you will have to reduce both the
inductive Xl and capacitive Xc reactance by reducing the inductance
and increasing the capacitance (e.g. by multiple varactors).

Rhode wrote an article in QST a few years ago about running the HF
varactor tuning front end at a lower impedance level to avoid the high
RF voltages on the varactors.

In VHF/UHF reducing the inductance to enable larger capacitances and
thus lower impedance and RF voltages is problematic, since the
inductance is already extremely small.

In principle it should be possible to connect several "coils" in
parallel (actual wire loops across the capacitor) and this is how many
text books explain how the cavity resonators are formed by adding
further and further wire loops surrounding the capacitor. However, I
have never seen parallel "coils" in any practical circuit, apparently
there are some parasitic capacitance problems.

However, I think that the OP should also study of making a shortened
1/4 (stripline or microstrip) resonator, with very wide resonators
(and thus low impedance levels) and do the impedance transformation at
the input and output coupling. If multiple stage filtering is needed,
look for interdigital filters and again design for low resonator
impedances to reduce the RF voltage across the tuning capacitors.
These might be more practical for the intended frequencies than
ordinary LC filters.

If the tuning speed is not very large, look for some mechanical tuning
at the end of the stripline resonator, such as moving the grounding
electrode closer to the resonator hot end by a piezoelectric crystal
etc., thus increasing the capacitance.

Paul OH3LWR

Hi Paul,

Thanks again for the suggestions; I need to do a few simulations to see how
viable some of the approaches are.

However, I think that the OP should also study of making a shortened
1/4 (stripline or microstrip) resonator, with very wide resonators
(and thus low impedance levels) and do the impedance transformation at
the input and output coupling.

The problem with this approach is that I've yet to see a
stripline/microstrip resonator design that -- by design -- doesn't have
re-entrant modes well before covering a 16.7:1 range (30-500MHz, in my
case). Otherwise I'd be all for it!

If the tuning speed is not very large

It's not, 'some low number of seconds' to re-tune is fine.

I don't suppose anyone makes motorized piston trimmer caps? Sure would be
nice...

---Joel

"Paul Keinanen" wrote in message
...
On Mon, 1 Nov 2004 17:24:10 -0600, "Steve Nosko"
wrote:
The "back to back" configuration I am familiar with is like this:
+ Control voltage
|
R
|
+----||-----+---||---+
| |
| |
+--LLLLLLLL+LLLLLLLLL--+

There could also be a series cap in this configuration--...

You would need fixed capacitors on _both_ anodes and also resistors to
ground from both anodes to get the DC bias. The small series
capacitors are essential, since they take the most part of the RF
voltage.

Hi Paul,

Why both? It is the total series cap which is of concern. One cap
would simply be half the value of each of the two. 2 x 5pf = 1 x 2.5pf,
no?

Of course, the series resistor will reduce the tuning range.

I don't see this. It would be just like the one shown. 100k or 1M. It
is for DC and is large enough to be neglegible at RF. (strays acknowledged)

I think there are many two diode configurations.
One option:
+ Control voltage
|
R
2.5pf |
+-||-+-||---+---||---+
| | |
| R |
| | |
| gnd |
| |
| |
+--LLLLLLLL+LLLLLLLLL--+ DC gnd down here

THough I don't think it is necessary, if you require symmetry.
Another option:
+----+ Control voltage
| |
R R
| |
+----||-+-||-+--||---+
| 2.5pf |
| |
| |
| |
| |
| |
+--LLLLLLLL+LLLLLLLLL--+

Another option:
+-----------+ Control voltage
| |
R R
| |
+-||-+-||-+-||-+--||-+
| 5pf | 5pf |
| | |
| gnd |
| |
| |
| |
+--LLLLLLLL+LLLLLLLLL--+


[[I am obviously ignoring parasitics of the diode or cap body to ground.
More "stuff" of any kind in tthe in the circuit = more paracitic capacitance
to fight. That is a mechanical RF layout issue and, of course, not to be
ignored.]]

You could get away with the series capacitor and put multiple (maybe
10) varactors in series. Thus, the RF voltage across each varactor
would be low. You will need a high tuning voltage, perhaps 100 V.

I agree, Clearly a complex arrangement. However, is there really an
advantage to having low voltage if you now have so many contributors to the
problems...


... I also don't get this talk about the filter Z. Since you'd need to Z
match
in/out of the filter, it seems to me the varactor voltages will be the
same
for any Zin/out since this will be determined by how "tightly" they are
coupled into the resonant circuit and not the Zin/Zout, no? The Z match
will just change the Vin/out.

[describes z matching in/out of a filter...]

That's my point. Once you select an inductor type, you have fixed a number
of future decisions. The transformation can get you to wherever you need to
be *IN* the resonator....see next

To reduce the Z in a resonator, you will have to reduce both the
inductive Xl and capacitive Xc reactance by reducing the inductance
and increasing the capacitance (e.g. by multiple varactors).

So now the varactor must be _more_ of the overall capacitance to get the
desired tuning range, no. If you want minimal side effects, then the
varactors need to be just barely inthe circuit, so to speak, which means
that they will have small voltages.

Rhode wrote an article in QST a few years ago about running the HF
varactor tuning front end at a lower impedance level to avoid the high
RF voltages on the varactors.

Probably necessary reading, but I guess that I am trying to point out that
this is one of those "no free lunch" situations. That a wide tuning range
AND good IM, temp performance are each pulling the design in opposite
directions? I mean:
Wide tuning - means - varactors need to be a larger fraction of the total
capacitance...
But this means that the non linearity and temp drift of the diode has more
effect.
If you put series caps to reduce voltage, although you are reducing the
effect of cap non linearity and temp drift because the diode is now
"decoupled" from the tuned circuit, you are reducing the potential tuning
range.

I have no significant comments on most of the rest...except...

...However, I
have never seen parallel "coils" in any practical circuit, apparently
there are some parasitic capacitance problems.

Side bar: My first "short wave" receiver did this.
Take one 50's tube AM radio. Parallel both the RF
and LO coils with outboard coils to get 75 Meter
phone band (it was mostly AM then). I think it was
in Popular Electronics or some such mag.

However, I think that the OP should also study of making a shortened
1/4 (stripline or microstrip) resonator, with very wide resonators

Been there, done that (not varactor tuned, though). The needed
direction.

(and thus low impedance levels) and do the impedance transformation at
the input and output coupling. If multiple stage filtering is needed,
look for interdigital filters

Sure, Interdigital or combline... 220, 221. Whatever it takes...


If the tuning speed is not very large, look for some mechanical tuning
at the end of the stripline resonator,

Something you should not reject is a "ranged" system. Tune a smaller
range with the varactors and switch in/out other caps for larger shifts.

73,
--
Steve N, K,9;d, c. i My email has no u's.