Low distortion sine wave VCO's - state of the art?

[Tim Shoppa]

The 8038 sine wave VCO has been around for a good chunk of a century.
Probably about the same for the XR-2206. But their distortion is in the
half-percent range, and I'm desiring a low distortion sine wave VCO
with distortion in the 0.01% range (and will not complain if it's
better).

For my particular project, I need the center frequency to be somewhere
in the low audio range (20 Hz to 1kHz would be great.). I'm not picky
about the exact center frequency. I also have very undemanding needs
for the tuning range: +/- 10% would be enough, and maybe just a few
percent would be better.

DDS isn't particularly desirable because the control voltage is a
voltage, not a number. I know I could add an A/D to make it a number,
but my feeling is that this is going down the wrong road.

Wien bridge oscillators can certainly meet the distortion needs, but is
there a "good" way to control the frequency (even over a limited range)
with an external voltage? Again, don't get into digitally-controlled
pots. I'm thinking something more like photocells here. But wouldn't I
need a matched pair of photocells? If the variable R were only a small
part of the total R, matching wouldn't be all that important, right?

Since my frequency range is so narrow, is a 8038 or XR-2206 followed by
many poles of low-pass filtering out of the question?

Tim.

 

[Winfield Hill]

Tim Shoppa wrote...

The 8038 sine wave VCO has been around for a good chunk of a century.
Probably about the same for the XR-2206. But their distortion is in
the half-percent range, and I'm desiring a low distortion sine wave
VCO with distortion in the 0.01% range (and will not complain if it's
better).

For my particular project, I need the center frequency to be somewhere
in the low audio range (20 Hz to 1kHz would be great.). I'm not picky
about the exact center frequency. I also have very undemanding needs
for the tuning range: +/- 10% would be enough, and maybe just a few
percent would be better.

DDS isn't particularly desirable because the control voltage is a
voltage, not a number. I know I could add an A/D to make it a number,
but my feeling is that this is going down the wrong road.

Wien bridge oscillators can certainly meet the distortion needs, but is
there a "good" way to control the frequency (even over a limited range)
with an external voltage? Again, don't get into digitally-controlled
pots. I'm thinking something more like photocells here. But wouldn't I
need a matched pair of photocells? If the variable R were only a small
part of the total R, matching wouldn't be all that important, right?

Since my frequency range is so narrow, is a 8038 or XR-2206 followed
by many poles of low-pass filtering out of the question?

Yes, by all means, add a sharp low-distortion filter. Problem solved.

 

[martin griffith]

The 8038 sine wave VCO has been around for a good chunk of a century.
Probably about the same for the XR-2206. But their distortion is in the
half-percent range, and I'm desiring a low distortion sine wave VCO
with distortion in the 0.01% range (and will not complain if it's
better).

For my particular project, I need the center frequency to be somewhere
in the low audio range (20 Hz to 1kHz would be great.). I'm not picky
about the exact center frequency. I also have very undemanding needs
for the tuning range: +/- 10% would be enough, and maybe just a few
percent would be better.

DDS isn't particularly desirable because the control voltage is a
voltage, not a number. I know I could add an A/D to make it a number,
but my feeling is that this is going down the wrong road.

Wien bridge oscillators can certainly meet the distortion needs, but is
there a "good" way to control the frequency (even over a limited range)
with an external voltage? Again, don't get into digitally-controlled
pots. I'm thinking something more like photocells here. But wouldn't I
need a matched pair of photocells? If the variable R were only a small
part of the total R, matching wouldn't be all that important, right?

Since my frequency range is so narrow, is a 8038 or XR-2206 followed by
many poles of low-pass filtering out of the question?

Tim.

How about a state variable filter osc, with the two "tuning R's"
replaced with current in, current out VCAs, say those in the THAT semi
catalog, THAT2151 rings a bell


martin

 

[Allan Herriman]

The 8038 sine wave VCO has been around for a good chunk of a century.
Probably about the same for the XR-2206. But their distortion is in the
half-percent range, and I'm desiring a low distortion sine wave VCO
with distortion in the 0.01% range (and will not complain if it's
better).

For my particular project, I need the center frequency to be somewhere
in the low audio range (20 Hz to 1kHz would be great.). I'm not picky
about the exact center frequency. I also have very undemanding needs
for the tuning range: +/- 10% would be enough, and maybe just a few
percent would be better.

DDS isn't particularly desirable because the control voltage is a
voltage, not a number. I know I could add an A/D to make it a number,
but my feeling is that this is going down the wrong road.

Wien bridge oscillators can certainly meet the distortion needs, but is
there a "good" way to control the frequency (even over a limited range)
with an external voltage? Again, don't get into digitally-controlled
pots. I'm thinking something more like photocells here. But wouldn't I
need a matched pair of photocells? If the variable R were only a small
part of the total R, matching wouldn't be all that important, right?

Since my frequency range is so narrow, is a 8038 or XR-2206 followed by
many poles of low-pass filtering out of the question?

I'm curious. What application demands an analog tuning voltage?

Allan

 

[John Popelish]

Tim Shoppa wrote... (snip)


Yes, by all means, add a sharp low-distortion filter. Problem solved.

Does an 8038 0r XR-2206 have a low enough phase jitter to produce a
..01% pure frequency, even with reasonable filtering?

The kind of oscillator that came to my mind is the sine cosine
oscillator made with two integrators in a loop. But there are still
two resistors that have to change in tandem to swing the frequency.
But the automatic gain control will compensate for some mismatch, if
the frequency changes slowly enough. Or is there a variation that
allows the frequency to be changed with only one resistor?

http://www.interq.or.jp/japan/se-inoue/e_ckt20.htm

 

[Tom Bruhns]

Well, combining Win's idea with another, you could do it like the old
HP8640A/B does: make an RF vco using a varicap diode for tuning, and
run it through a divide-by-a-lot IC. The square wave output has very
low second harmonic (theoretically zero), so with only a +/- 10% tuning
range, it's straightforward to filter it. If you don't need super low
spurious, you can use a Maxim or Linear Technology switched-capacitor
filter, likely running the clock from a tap on the divide-by-a-lot IC.
Might have to use one following stage of non-switching filtering to get
things clean enough. Given varicap diodes that cover something like
50-500pF, the VCO frequency can be fairly (few hundred kHz or less) low
so "a-lot" isn't all that big a number.

You could also make an RC phase shift oscillator, where the C's are
large varicaps like that, and do it directly. There are probably other
configurations of RC sinewave oscillator that could be made to
work...possibly a Wien bridge type...using some sort of AGC to keep the
wave from clipping in each case. Varicap diodes with large capacitance
could get you to very low frequencies; I believe the old HP200CD uses a
couple of about 1200pF sections for the low frequencies, way lower than
you need. You'd probably want to run varicap diodes in back-to-back
series pairs to lower the distortion from the signal varying the
varicap bias, and even then might need a clean-up filter on the output.

BTW, Win, if you see this...did you get my recent email request? If
not, please drop me an email...

Cheers,
Tom

 

[Winfield Hill]

John Popelish wrote...

Does an 8038 0r XR-2206 have a low enough phase jitter to produce
a .01% pure frequency, even with reasonable filtering?

I'd guess, yes, at low frequencies. That's only 100ppm.

 

[Fred Bloggs]

Well, combining Win's idea with another, you could do it like the old
HP8640A/B does: make an RF vco using a varicap diode for tuning, and
run it through a divide-by-a-lot IC. The square wave output has very
low second harmonic (theoretically zero), so with only a +/- 10% tuning
range, it's straightforward to filter it. If you don't need super low
spurious, you can use a Maxim or Linear Technology switched-capacitor
filter, likely running the clock from a tap on the divide-by-a-lot IC.
Might have to use one following stage of non-switching filtering to get
things clean enough. Given varicap diodes that cover something like
50-500pF, the VCO frequency can be fairly (few hundred kHz or less) low
so "a-lot" isn't all that big a number.

You could also make an RC phase shift oscillator, where the C's are
large varicaps like that, and do it directly. There are probably other
configurations of RC sinewave oscillator that could be made to
work...possibly a Wien bridge type...using some sort of AGC to keep the
wave from clipping in each case. Varicap diodes with large capacitance
could get you to very low frequencies; I believe the old HP200CD uses a
couple of about 1200pF sections for the low frequencies, way lower than
you need. You'd probably want to run varicap diodes in back-to-back
series pairs to lower the distortion from the signal varying the
varicap bias, and even then might need a clean-up filter on the output.

Okay- you're getting close but still antiquated. State of the art is
achieved with DDS, no question about that, and there is that nasty
problem of spurious frequency content relatively close to the
fundamental which is a complicated function of the number theoretic
relations between the CLK and fundamental. This is not likely to support
0.01% THD in general. However, there are fundamental to CLK ratios that
will support 0.01% THD, so the obvious answer is fix that ratio, and
then tune your RF CLK with the varicap. Output filtering can then be
fixed and trivial.

 

[John Popelish]

John Popelish wrote...


I'd guess, yes, at low frequencies. That's only 100ppm.

You need pretty clean supply rails to achieve that, since the XR-2206
has a typical supply frequency sensitivity of .01% per volt (100ppm
per volt), but the worst case is 10 times that (and that spec applys
only to an optimum set of conditions). I didn't look up the 8038.

I would worry that any multivibrator oscillator risks more jitter than
one based on a a high Q resonance or a continuous time process that
depends on at least a whole previous cycle. Multivibrators make
irrevocable decisions based on instantaneous voltages.

 

[Rene Tschaggelar]

Well, combining Win's idea with another, you could do it like the old
HP8640A/B does: make an RF vco using a varicap diode for tuning, and
run it through a divide-by-a-lot IC. The square wave output has very
low second harmonic (theoretically zero), so with only a +/- 10% tuning
range, it's straightforward to filter it. If you don't need super low
spurious, you can use a Maxim or Linear Technology switched-capacitor
filter, likely running the clock from a tap on the divide-by-a-lot IC.
Might have to use one following stage of non-switching filtering to get
things clean enough. Given varicap diodes that cover something like
50-500pF, the VCO frequency can be fairly (few hundred kHz or less) low
so "a-lot" isn't all that big a number.

It may have little 2nd harmonic, but still plenty
of the odd harmonics. Better start with a 8038.

Rene

 

[John Popelish]

The kind of oscillator that came to my mind is the sine cosine
oscillator made with two integrators in a loop. But there are still two
resistors that have to change in tandem to swing the frequency. But the
automatic gain control will compensate for some mismatch, if the
frequency changes slowly enough. Or is there a variation that allows
the frequency to be changed with only one resistor?

http://www.interq.or.jp/japan/se-inoue/e_ckt20.htm

I just realized that if you replace the two integrator input resistors
with operational transconductance amplifiers, you turn this into a VCO
(Well, a current controlled oscillator anyway. Adding a couple
resistors to the gain control inputs would fix that.) A pair of OTA's
from the same pack should track well enough for the AGC to keep the
output amplitude quite constant. With low distortion opamps, I think
the second integrator (from the variable feedback point ) might well
be able to hit the .01% distortion spec.

 

[Winfield Hill]

John Popelish wrote...

You need pretty clean supply rails to achieve that, since the XR-2206
has a typical supply frequency sensitivity of .01% per volt (100ppm
per volt), but the worst case is 10 times that (and that spec applys
only to an optimum set of conditions). I didn't look up the 8038.

OK, let's say 10x worse max: this means he'd need 0.1-volt quiet
and stability, which is easily and routinely exceeded.

I would worry that any multivibrator oscillator risks more jitter
than one based on a a high Q resonance or a continuous time process
that depends on at least a whole previous cycle. Multivibrators
make irrevocable decisions based on instantaneous voltages.

Agreed, but that's why we'll say 100ppm, rather than 10x better.

 

[Ken Smith]

Well, combining Win's idea with another, you could do it like the old
HP8640A/B does: make an RF vco using a varicap diode for tuning, and
run it through a divide-by-a-lot IC. The square wave output has very
low second harmonic (theoretically zero), so with only a +/- 10% tuning
range, it's straightforward to filter it.

A couple of ideas:

(1)Once you have this VCO, you could use it as the clock on a DDS chip.


(2) if you use a moderately long Johnson counter as the last section of
the divider, a set of weighted resistors can get the first few odd
harmonics down under 1%. This makes the following filter a lot easier to
make.

 

[Rich Grise]

Does an 8038 0r XR-2206 have a low enough phase jitter to produce a .01%
pure frequency, even with reasonable filtering?

FWIW, I slapped together an 8038 circuit right out of the data book, and
no matter how much I twiddled R12 and R13, I couldn't get rid of that
little tittie on the top of the sine output. I didn't have a THD meter
(who does? ?:-\ ), but that little tittie had to have introduced some
distortion.

I haven't tried the eXaR chip, but it could be worth a shot.

Otherwise, i'd probably look into the Wein(Wien?)-bridge or a phase-shift
oscillator - with CMOS parts these days, you should probably be able to
make a very nice sine wave.

Good Luck!
Rich

 

[Tim Shoppa]

Well, combining Win's idea with another, you could do it like the old
HP8640A/B does: make an RF vco using a varicap diode for tuning, and
run it through a divide-by-a-lot IC. The square wave output has very
low second harmonic (theoretically zero), so with only a +/- 10% tuning
range, it's straightforward to filter it.

More likely (as elsewhere suggested in this thread, and in AOE as well)
I'd use a ring counter so I don't have to worry much until some high
harmonic (AOE says IIRC that with a 8-stage ring counter, 15th harmonic
is the first assuming perfect resistor balance).

And at that point no need for varicap stuff, the VCO from a HC4046
would be more than good enough.

You could also make an RC phase shift oscillator, where the C's are
large varicaps like that, and do it directly. There are probably other
configurations of RC sinewave oscillator that could be made to
work...possibly a Wien bridge type...using some sort of AGC to keep the
wave from clipping in each case. Varicap diodes with large capacitance
could get you to very low frequencies; I believe the old HP200CD uses a
couple of about 1200pF sections for the low frequencies, way lower than
you need. You'd probably want to run varicap diodes in back-to-back
series pairs to lower the distortion from the signal varying the
varicap bias, and even then might need a clean-up filter on the output.

I'm reluctant to put nonlinear components like varicaps in the
oscillator because I'd rather start with low distortion rather than put
it in. As you point out enough back-bias and small enough oscillator
amplitude might make this not so bad.

My first inclination for a "controllable low-distortion tuning" was
some Silonex photocell-coupled-to-LED thingies in a Wien bridge. I had
very good success with these in low distortion (0.01%) stuff in the
past and while I don't remember the guarantees about matching, the ones
I've used happened to be very well matched.

Tim.

 

[Fred Bloggs]

Any worthwhile electronics engineer could duplicate the performance of
this in the home lab:

http://www.thinksrs.com/products/DS360.htm

 

[John Fields]

My first inclination for a "controllable low-distortion tuning" was
some Silonex photocell-coupled-to-LED thingies in a Wien bridge. I had
very good success with these in low distortion (0.01%) stuff in the
past and while I don't remember the guarantees about matching, the ones
I've used happened to be very well matched.

 

[[email protected]]

More likely (as elsewhere suggested in this thread, and in AOE as well)
I'd use a ring counter so I don't have to worry much until some high
harmonic (AOE says IIRC that with a 8-stage ring counter, 15th harmonic
is the first assuming perfect resistor balance).

And at that point no need for varicap stuff, the VCO from a HC4046
would be more than good enough.


I'm reluctant to put nonlinear components like varicaps in the
oscillator because I'd rather start with low distortion rather than put
it in. As you point out enough back-bias and small enough oscillator
amplitude might make this not so bad.

My first inclination for a "controllable low-distortion tuning" was
some Silonex photocell-coupled-to-LED thingies in a Wien bridge. I had
very good success with these in low distortion (0.01%) stuff in the
past and while I don't remember the guarantees about matching, the ones
I've used happened to be very well matched.

Tim.

John P raised a useful point re possible jitter sources on the triangle
shaping chips. Extreme curiosity forced examination of my cheap
function generator, a Xtal based sine synthesiser and a GP Wein bridge
oscillator. The Topward function genny is identical in working to the
XR and ICL internals yet gave no noticable jitter, even at the 1MHz
mark (1kc THD 0.6%). Biggest problem was 50Hz power supply related,
carrier sidebands at -50dB. The digital synth offered a cleaner sine
but throws out wideband, low level noise at the -80dB level, ( THD
0.1%). Wien bridge, no jitter, clean -90dB noise floor and THD of 0.2%.

(-10dBm refs and not counting slow frequency drift as jitter)

I'd go with the guys. Start with a '8038 etc and filter down to 0.01%
john

[Have a pair of 8 year old Silonex devices. Originally used as Wien
bridge arms, with a HF pilot tone running through to allow dynamic
balancing. They still match within 10%, pretty much the same as when
bought!.]

 

[Tony Williams]

I just realized that if you replace the two integrator input
resistors with operational transconductance amplifiers, you turn
this into a VCO (Well, a current controlled oscillator anyway.
Adding a couple resistors to the gain control inputs would fix
that.) A pair of OTA's from the same pack should track well
enough for the AGC to keep the output amplitude quite constant.
With low distortion opamps, I think the second integrator (from
the variable feedback point ) might well be able to hit the .01%
distortion spec.

Yes you are spot on. I have a 1980-ish design doing
exactly that...... fast frequency slewing, with no
bounce. 5x multipliers instead of OTA's though.

2x multipliers in series with the integration R's
for the voltage control of frequency, as above.
2x multipliers squaring Sin and Cos, outputs added
together to produce a fast-response rectification without
the need for a smoothing cap. Vref was subtracted at the
same time to produce the error voltage for the
multiplier that increased/decreased the oscillation
gain.

I got about 0.005% THD in the 1980 prototype, which
was at the limit of the 400Hz notch filter I had at
that time. That was with 2x low THD multipliers in
the signal path (AD634?), using the correct low THD
X or Y input for the AC, and Vpk held down to no
more than about 7v (4-5Vrms).

 

[Tony Williams]

Wien bridge oscillators can certainly meet the distortion needs,
but is there a "good" way to control the frequency (even over a
limited range) with an external voltage?

BTDT with trying to voltage control a Wien Bridge
oscillator. The initial problem to solve is that
one of the resistive elements has a CMV on it, but
the real problem is moving the frequency without
transient amplitude changes/bounces.

The underlying problem there is that rectification
of a single phase (to get a low ripple dc for the agc)
cannot be made fast enough. I have a vague memory in
one of those reader Design Ideas at that time, of
someone resorting to a complicated s/hold on Vpk to
get a fast response dc for the agc.