Message	Date	From	Subject
1082	2007-10-10 20:12:16	bkopski	Linearization of Varactor Tuned VFO
Hi All, I recently replaced a ratty air variable / vernier dial combination in a VFO with varactor tuning. The new 10-turn pot is such a refreshing pleasure to use! But like all such tuning schemes I've seen, this one suffers the classic "crowded low end" and "expanded top end" tuning non linearity. This was expected of course and is not a show-stopper, but I wanted to "straighten this out" - for the fun and utility of it. The actual tuning curve is shown in the files section and is titled "VFO Frequency vs. Voltage" on the VFO Transfer Curves graphic set. Some associated photos are in the photos section. I reran the same data this time adjusting the tuning voltage as needed to produce even increments of frequency and the resultant plot is shown as "VFO Varactor Voltage vs. Linear Frequency". The curve shape shown here is what is needed to "bend" a linearly changing tuning pot voltage to produce a corresponding linear VFO frequency change - for my particular VFO. But every VFO I've seen has a transfer curve of this same basic shape despite values particulars varying among them, so I believe what follows should be applicable in principle to all such VFOs. The circuit approach is shown in the "VFO Linearizer" schematic, and the result of using it is shown in the graphic "Measured Linearized VFO Transfer Characteristic". The circuit design began with some assumptions and decisions. For example, I decided to have a total tuning pot voltage range of 5 volts. (I had to pick something.) And I decided upon an approach to setting the low frequency point - the "start point". The former voltage is established by R2 and the latter by variable R3. Both resistors have 5 mA flowing in them by the action of the LM317LZ current source - which is also the reference / regulator for everything else in the circuit. Both voltages are buffered by op amp voltage followers resulting in the "V-LO" and "V-HI" busses. And the tuning pot has 5 volts across it as planned. Tuning pot wiper voltage is buffered by a third op amp which outputs drive to segment resistor R10, and a diode pedestaled voltage to drive segments 2 thru 5. The diode very nearly tracks out any temperature related motions of the transistor emitter diodes. Between this op amp and the final varactor-driving op amp output section lie 5 parallel paths each incrementally contributing to what will become the "bent" output voltage transfer function. Operation and tune up begins by first setting all segment resistors (R5 - R10) to maximum value and the tuning pot to the low end. R3 is then adjusted to produce the beginning or start value varactor voltage and the corresponding lowest frequency - as dictated by the graphic data. (In this case that voltage is about 2.55 volts and the frequency is 4.2 mHz.) The tuning pot is then adjusted upward from zero to result in 1 volt between the pot wiper and V-LO. The resulting frequency will fall short of the desired value so R10 is adjusted to set the frequency correctly. The first portion (segment) of the transfer curve is now properly set. Noting that the subsequent segment forming transistors have base voltages set at one volt increments, the tuning pot is now set to produce 2 volts between the wiper and V-LO. (Up to this point no transistor emitter are forward biased, so none of these paths was conducting. At this point however the first such path (second segment) initiates transistor conduction and the associated collector current flows into R16 incrementing the voltage already there from R10 current.) Once again the associated frequency will be low and so R9 is adjusted to make it the correct value. The remaining steps are more of the same 1-volt increments until finally R6 is adjusted, the circuit output voltage is 7 volts, and the upper end of the dial has a frequency of 4.36 mHz. The graphic "Measured Linearized VFO Transfer Characteristic" pretty much tells the success story. Note that the circuit details and specifics here are appropriate for my particular VFO. Since all VFO V-F curves I've seen follow much the same general shape, I believe this approach will be generally applicable - suitably customized to the needed specifics. I can imagine where fewer segments and a more-coarse approximation would be just fine in some cases. Similarly, I can imagine even more than these 5 segments being used to attain an even smoother curve. Then there are the very wide range of low-to-high tuning voltage ranges in use - all accommodated I believe with proper choice of circuit constants. Whatever the end design detail, I do think this approach to a piece wise approximation of the non-linear voltage transfer characteristic needed for most any varactor tuned oscillator can end that "crowded low end" and "expanded high end" lament once and for all! Cordially and 73, Bob K3NHI
1084	2007-10-11 07:19:43	Allison Parent	Re: Linearization of Varactor Tuned VFO

1085	2007-10-11 17:47:00	jr_dakota	Re: Linearization of Varactor Tuned VFO
That essentially has been my experience also, varactors are naturally drifty because they are a semiconductor junction that varies with temperature ... adding more junctions makes it worse but since they are usually used in the VCO of a PLL the drift isn't much of a problem True antilog pots are nearly impossible to find at reasonable prices (so are true log pots) so the 'trick' of adding an additional resistor between the wiper and one end is the best route ... In fact I've also made log pots that way when I needed them to be accurate, for example a precision attenuator calibrated in db's or calibrated mike preamp which doesn't work with most off the shelf log pots because most log pots aren't true log, they are really a series of linear sections that roughly approximates an antilog scale I have several resources on this but most aren't on the web but a good one that covers most of the bases can be found at http://www.elby-designs.com/documents/tailoringpotentionometers.pdf And here: http://sound.westhost.com/pots.htm (This is also a good site for those who want to sharpen up the audio frequency backend of their radio designs which I've noticed are often neglected or low priority (Every time I see a LM324 or a LM358 used in an audio design I want to tear my hair out ) Note also that varactors aren't truly antilog either so you have to tweak the value of the additional resistor from the calculated value but you can get pretty close to having linear tuning ... I usually use a diode and resistor in series on the ground end of the pot to keep the voltage from going too low as well as for temperature compensation if you can mount the diode close to the varactor ,,, LED's are also a good solution with their higher voltage drop which puts you right in the 2-3 volt range with a single junction (I'm not sure about how they work as temperature compensati
1086	2007-10-12 07:01:43	Allison Parent	Re: Linearization of Varactor Tuned VFO

Hi All,

I recently replaced a ratty air variable / vernier dial combination
in a VFO with varactor tuning. The new 10-turn pot is such a
refreshing pleasure to use! But like all such tuning schemes I've
seen, this one suffers the classic "crowded low end" and "expanded
top end" tuning non linearity. This was expected of course and is
not a show-stopper, but I wanted to "straighten this out" - for the
fun and utility of it. The actual tuning curve is shown in the files
section and is titled "VFO Frequency vs. Voltage" on the VFO Transfer
Curves graphic set. Some associated photos are in the photos section.

I reran the same data this time adjusting the tuning voltage as
needed to produce even increments of frequency and the resultant plot
is shown as "VFO Varactor Voltage vs. Linear Frequency". The curve
shape shown here is what is needed to "bend" a linearly changing
tuning pot voltage to produce a corresponding linear VFO frequency
change - for my particular VFO. But every VFO I've seen has a
transfer curve of this same basic shape despite values particulars
varying among them, so I believe what follows should be applicable in
principle to all such VFOs.

The circuit approach is shown in the "VFO Linearizer" schematic, and
the result of using it is shown in the graphic "Measured Linearized
VFO Transfer Characteristic".

The circuit design began with some assumptions and decisions. For
example, I decided to have a total tuning pot voltage range of 5
volts. (I had to pick something.) And I decided upon an approach to
setting the low frequency point - the "start point". The former
voltage is established by R2 and the latter by variable R3. Both
resistors have 5 mA flowing in them by the action of the LM317LZ
current source - which is also the reference / regulator for
everything else in the circuit. Both voltages are buffered by op amp
voltage followers resulting in the "V-LO" and "V-HI" busses. And the
tuning pot has 5 volts across it as planned.

Tuning pot wiper voltage is buffered by a third op amp which outputs
drive to segment resistor R10, and a diode pedestaled voltage to
drive segments 2 thru 5. The diode very nearly tracks out any
temperature related motions of the transistor emitter diodes.
Between this op amp and the final varactor-driving op amp output
section lie 5 parallel paths each incrementally contributing to what
will become the "bent" output voltage transfer function.

Operation and tune up begins by first setting all segment resistors
(R5 - R10) to maximum value and the tuning pot to the low end. R3 is
then adjusted to produce the beginning or start value varactor
voltage and the corresponding lowest frequency - as dictated by the
graphic data. (In this case that voltage is about 2.55 volts and the
frequency is 4.2 mHz.) The tuning pot is then adjusted upward from
zero to result in 1 volt between the pot wiper and V-LO. The
resulting frequency will fall short of the desired value so R10 is
adjusted to set the frequency correctly. The first portion (segment)
of the transfer curve is now properly set.

Noting that the subsequent segment forming transistors have base
voltages set at one volt increments, the tuning pot is now set to
produce 2 volts between the wiper and V-LO. (Up to this point no
transistor emitter are forward biased, so none of these paths was
conducting. At this point however the first such path (second
segment) initiates transistor conduction and the associated collector
current flows into R16 incrementing the voltage already there from
R10 current.) Once again the associated frequency will be low and so
R9 is adjusted to make it the correct value. The remaining steps are
more of the same 1-volt increments until finally R6 is adjusted, the
circuit output voltage is 7 volts, and the upper end of the dial has
a frequency of 4.36 mHz. The graphic "Measured Linearized VFO
Transfer Characteristic" pretty much tells the success story.

Note that the circuit details and specifics here are appropriate for
my particular VFO. Since all VFO V-F curves I've seen follow much
the same general shape, I believe this approach will be generally
applicable - suitably customized to the needed specifics. I can
imagine where fewer segments and a more-coarse approximation would be
just fine in some cases. Similarly, I can imagine even more than
these 5 segments being used to attain an even smoother curve. Then
there are the very wide range of low-to-high tuning voltage ranges in
use - all accommodated I believe with proper choice of circuit
constants. Whatever the end design detail, I do think this approach
to a piece wise approximation of the non-linear voltage transfer
characteristic needed for most any varactor tuned oscillator can end
that "crowded low end" and "expanded high end" lament once and for
all!

Cordially and 73,
Bob K3NHI