Maker Pro
Maker Pro

how to speed up an oscillator's wake up time?

F

frank

Jan 1, 1970
0
Hello,

I was recently awarded the task of re-designing an existing 50MHz
oscillator. One the parameters I was asked to reduce is the wake-up.
It currently stamds at ~5ms. Does anyone know of any tricks to speed
up the wake-up time of a pierce archicture oscillator ? The way I see
it, is, that the equivalent negative resistance of the oscillator
circuit as seen from the xtal, has to be greatly increased. I
calculated it to be in the neighborhood of -50ohms (does that sound
about rigth? ); is there any way to increase it to, say about -2kohms
?

-Frank
 
K

Kevin Aylward

Jan 1, 1970
0
frank said:
Hello,

I was recently awarded the task of re-designing an existing 50MHz
oscillator. One the parameters I was asked to reduce is the wake-up.
It currently stamds at ~5ms. Does anyone know of any tricks to speed
up the wake-up time of a pierce archicture oscillator ? The way I see
it, is, that the equivalent negative resistance of the oscillator
circuit as seen from the xtal, has to be greatly increased. I
calculated it to be in the neighborhood of -50ohms (does that sound
about rigth? ); is there any way to increase it to, say about -2kohms
?

Ho hummm... red herring. The "negative" resistance, which is totally
equivalent, in this example, to the amount of positive feedback, but is
an attempt to sound more impressive, has nothing (essentially) to do
with the start up time. Assuming that the basic circuit does not have
things like supply caps to charge, the start-up time is determined by
the "Q" of the circuit. Anything you do to reduce the Q to speed up the
start-up time, will near as damm it, reduce the frequency stability of
the oscillator.


Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.
 
J

Jim Backus

Jan 1, 1970
0
..... Assuming that the basic circuit does not have
things like supply caps to charge, the start-up time is determined by
the "Q" of the circuit. Anything you do to reduce the Q to speed up the
start-up time, will near as damm it, reduce the frequency stability of
the oscillator.

If start up time is more important than spectral purity, reducing the
Q would be the way to go.

Lower Q will give higher noise sidebands, which could show up as
jitter, reciprocal mixing and countless other effects depending on the
application.

--
Jim Backus OS/2 user
bona fide replies to jimb-thecirclethingy-jita-dp-demon-dp-co-dp-uk
or remove "NOT" from address
remove dashes and make the obvious substitutions for valid email
address
 
F

frank

Jan 1, 1970
0
Ho hummm... red herring. The "negative" resistance, which is totally
equivalent, in this example, to the amount of positive feedback, but is
an attempt to sound more impressive, has nothing (essentially) to do
with the start up time. Assuming that the basic circuit does not have
things like supply caps to charge, the start-up time is determined by
the "Q" of the circuit. Anything you do to reduce the Q to speed up the
start-up time, will near as damm it, reduce the frequency stability of
the oscillator.

I thougth that the start-time constant of the oscillator circuit was
responsible fro the start-up time of the oscillator. For a pierce
configuration,

tau = -2*L / [ Re(Zc) + Re(Zxtal) ]
where,
L=inductance of xtal
Zc= impedance of oscillator circuit as seen from xtal
Zxtal = impedance of xtal

Hence, increasing the negative equivalent resistance of Zc will lower
the start-up time constant.

-Frank
 
J

Jim Thompson

Jan 1, 1970
0
In PSpice?

Yep, but with max timestep set at 200ps.

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| Jim-T@analog_innovations.com Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |

For proper E-mail replies SWAP "-" and "_"

Why is it that Democrats can't debate politely?
And are only rude and interruptive.
Lack of mental capacity?
 
J

John Popelish

Jan 1, 1970
0
frank said:
Hello,

I was recently awarded the task of re-designing an existing 50MHz
oscillator. One the parameters I was asked to reduce is the wake-up.
It currently stamds at ~5ms. Does anyone know of any tricks to speed
up the wake-up time of a pierce archicture oscillator ? The way I see
it, is, that the equivalent negative resistance of the oscillator
circuit as seen from the xtal, has to be greatly increased. I
calculated it to be in the neighborhood of -50ohms (does that sound
about rigth? ); is there any way to increase it to, say about -2kohms

Does your oscillator have a capacitive divider at the input of the
crystal? If you can use a low capacitance switch to alter this
divider briefly (say, open the ground connection on the divider for
100 us) you should be able to provide a lot more positive feedback
(higher negative resistance) and get an oscillation built up much
faster. The frequency will be all over the place during this time,
but should settle down quickly when the divider is restored.
 
S

Spehro Pefhany

Jan 1, 1970
0
Hello,

I was recently awarded the task of re-designing an existing 50MHz
oscillator. One the parameters I was asked to reduce is the wake-up.
It currently stamds at ~5ms. Does anyone know of any tricks to speed
up the wake-up time of a pierce archicture oscillator ?

One way is to replace the crystal with a ceramic resonator (which will
have lower Q). Accuracy and stability will also be reduced, but are
still good enough for many purposes.

Best regards,
Spehro Pefhany
 
T

Tim Shoppa

Jan 1, 1970
0
Hello,

I was recently awarded the task of re-designing an existing 50MHz
oscillator. One the parameters I was asked to reduce is the wake-up.
It currently stamds at ~5ms. Does anyone know of any tricks to speed
up the wake-up time of a pierce archicture oscillator ? The way I see
it, is, that the equivalent negative resistance of the oscillator
circuit as seen from the xtal, has to be greatly increased. I
calculated it to be in the neighborhood of -50ohms (does that sound
about rigth? ); is there any way to increase it to, say about -2kohms
?

If you're in for a complete re-design, I've seen some applications use
a 32.768kHz oscillator that never stops running, current drain from that
oscillator is in the microamp range. The high frequency (10's of MHz)
is done by a microprocessor-controlled PLL, where at "startup" the
microprocessor just uses the previous VCO control voltage from the last
time the high-frequency end was shut down. Long-term the VCO is
locked to the 32.768kHz oscillator via the microcontroller, a digital
loop filter, and D/A converters. The VCO is just a relatively
low-Q LC (C partly from a varactor) so it starts up much faster than a
high-Q crystal. The advertised start time was in the tens of
microseconds.

I think Dallas was the vendor pushing this architecture (although I may
be misremembering).

Tim.
 
R

R.Legg

Jan 1, 1970
0
Hello,

I was recently awarded the task of re-designing an existing 50MHz
oscillator. One the parameters I was asked to reduce is the wake-up.
It currently stamds at ~5ms. Does anyone know of any tricks to speed
up the wake-up time of a pierce archicture oscillator ? The way I see
it, is, that the equivalent negative resistance of the oscillator
circuit as seen from the xtal, has to be greatly increased. I
calculated it to be in the neighborhood of -50ohms (does that sound
about rigth? ); is there any way to increase it to, say about -2kohms
?

-Frank

Pierce oscillators are used with microprocessors, and there are plenty
of app notes for these. 5mSec sounds like a typical start-up time with
recommended feedback swamping resistor values.

One trick to speed start-up is to split the capacitor value that is
present on the gate input of the oscillator gain cell into two parts;
one goes to the 'starting' supply rail, the other to normal ground.

If correctly these have appropriate ratios, they will bias the gate
input into it's switching threshold faster than just waiting for the
usual single grounded capacitor to charge up from zero.

RL
 
R

Robert Baer

Jan 1, 1970
0
Kevin said:
Ho hummm... red herring. The "negative" resistance, which is totally
equivalent, in this example, to the amount of positive feedback, but is
an attempt to sound more impressive, has nothing (essentially) to do
with the start up time. Assuming that the basic circuit does not have
things like supply caps to charge, the start-up time is determined by
the "Q" of the circuit. Anything you do to reduce the Q to speed up the
start-up time, will near as damm it, reduce the frequency stability of
the oscillator.

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

Absolutely, to some degree.
If one allows a square wave, start-up of an oscillator can be as fast
as one cycle (rather high gain in the feedback).
Then the oscillation rate is determined by the delays in the logic
circuit (a gated oscillator made with logic circuits).
To some degree, that rate is relatively stable.
So, if one was clever, have a variable feedback gain, where the gain
is rather high for fast start-up (and high distortion), and rapidly
decrease that feedback gain to "nominal" Colpitts, TPTG, or Pierce
feedback gain (depending on original circuit).
Frequency stability should not be a massive issue for the fitst 1,000
cycles (to 10,000 cycles or so) of operation, where this "trick" is
being applied.
 
R

Robert Baer

Jan 1, 1970
0
Tim said:
If you're in for a complete re-design, I've seen some applications use
a 32.768kHz oscillator that never stops running, current drain from that
oscillator is in the microamp range. The high frequency (10's of MHz)
is done by a microprocessor-controlled PLL, where at "startup" the
microprocessor just uses the previous VCO control voltage from the last
time the high-frequency end was shut down. Long-term the VCO is
locked to the 32.768kHz oscillator via the microcontroller, a digital
loop filter, and D/A converters. The VCO is just a relatively
low-Q LC (C partly from a varactor) so it starts up much faster than a
high-Q crystal. The advertised start time was in the tens of
microseconds.

I think Dallas was the vendor pushing this architecture (although I may
be misremembering).

Tim.

Just do not buy any %^@#$ from Dallas, even if you are lucky to find
someone that might have some on the shelves.
Dallas refuses to honour their "warranty" on parts.
Be advised that Maxim is the same crap company; there are scads of
advertised ICs by Maxim, but almost nobody carries them, samples are
impossible to get, the factory refuses to communicate.
Be prepared to pay for many thousands of dollars worth of one line
item and wait 6-14 months for them to get off their unmentionable to
actually get the production line started.
 
R

RP Henry

Jan 1, 1970
0
Just do not buy any %^@#$ from Dallas, even if you are lucky to find
someone that might have some on the shelves.
Dallas refuses to honour their "warranty" on parts.
Be advised that Maxim is the same crap company; there are scads of
advertised ICs by Maxim, but almost nobody carries them, samples are
impossible to get, the factory refuses to communicate.
Be prepared to pay for many thousands of dollars worth of one line
item and wait 6-14 months for them to get off their unmentionable to
actually get the production line started.

Haven't had that problem.

Rule #1 (just promulgated this week in a design review meeting, so it's the
Newest Rule in the Book): don't design in anything unless it is in stock at
Digikey (or substitute your favorite distributor).
 
S

Sir Charles W. Shults III

Jan 1, 1970
0
Here is a thought- I don't know if anyone thought of doing this.
Place a supercap and diode on the main power line to the device. Keep the
supercap charged but when main power is removed, disable the oscillator circuit.
Now, when main power reappears, enable it and it can run immediately because
there is power in the supercap. This should be okay for a couple of weeks of
zero power- perhaps more.

Cheers!

Chip Shults
My robotics, space and CGI web page - http://home.cfl.rr.com/aichip
 
J

Jim Backus

Jan 1, 1970
0
What about varying the Q? Start low, and increase quickly? A CMOS switch
might do wonders.

BTW 50 MHz and 5 ms is a quarter million cycles. Not that I know about
the Q of a crystal oscillator, but is it that high?

The model for start up of an oscillator is that noise in the circuit
is amplified, the frequency selective component filters it and the
signal is fed back. If the amplifier is low noise, there might be very
little filtered signal fed back. To achieve low noise in an
oscillator, which might be very important for a local oscillator in a
receiver or transmitter, one normally starts with a low noise
amplifier and adds just enough feedback to maintain oscillation. Under
these conditions start up could be very slow. Of course the definition
of start up could do with clarification - perhaps measured up to the
point where the amplitude was stable - possibly until it had settled
within a specified offset from the final frequency.

Many years ago a design I was working on needed an oscillator that
would start instantly. Having examined crystal oscillators and
discovered the slow start up problem, I ended up with an LC clock.
Very good start up time and with high quality components it provided
more than enough stability for the application.

--
Jim Backus OS/2 user
bona fide replies to jimb-thecirclethingy-jita-dp-demon-dp-co-dp-uk
or remove "NOT" from address
remove dashes and make the obvious substitutions for valid email
address
 
I

Ian Buckner

Jan 1, 1970
0
I see
Absolutely, to some degree.
If one allows a square wave, start-up of an oscillator can be as fast
as one cycle (rather high gain in the feedback).
Then the oscillation rate is determined by the delays in the logic
circuit (a gated oscillator made with logic circuits).
To some degree, that rate is relatively stable.
<snip>

As others have said, it depends what constraints you have on frequency
accuracy and phase noise. You can do an instantly startable oscillator
with
reasonable accuracy and jitter using a delay line. These have been
used in
digital oscilloscopes for time stamping samples.

Regards
Ian
 
K

Kevin Aylward

Jan 1, 1970
0
frank said:
Ho hummm... red herring. The "negative" resistance, which is totally
equivalent, in this example, to the amount of positive feedback, but
is an attempt to sound more impressive, has nothing (essentially) to
do with the start up time. Assuming that the basic circuit does not
have things like supply caps to charge, the start-up time is
determined by the "Q" of the circuit. Anything you do to reduce the
Q to speed up the start-up time, will near as damm it, reduce the
frequency stability of the oscillator.

I thougth that the start-time constant of the oscillator circuit was
responsible fro the start-up time of the oscillator. For a pierce
configuration,

tau = -2*L / [ Re(Zc) + Re(Zxtal) ]
where,
L=inductance of xtal
Zc= impedance of oscillator circuit as seen from xtal
Zxtal = impedance of xtal

Hence, increasing the negative equivalent resistance of Zc will lower
the start-up time constant.

Not really. The negative resistance approach to *standard* oscillators
is a bit of a misnomer. I agree, that the resistance of the xtal
determines the Q, however, its not directly related to the "negative"
resistance of the circuit. This is not a pure resistance as it varies
with frequency, i.e its from a 1/s^2 term. The equation derived from
considering either loop gain or this somewhat artificial negative
resistance, are identical. Would you expect the loop gain to have a
dramatic effect on the start up time if it varied from say 10 to 100?.
That is all varying the negative resistance will do. The negative
resistance is really only a more obtuse way of looking at the loop gain.
Its *identically* equivalent, for these normal oscillators, and so can
not contain any new information. Of course, there are oscillators that
are more aptly called negative resistance oscillators, such as tunnel
diode oscillators.


Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.
 
S

Steven Swift

Jan 1, 1970
0
Robert Baer said:
So, if one was clever, have a variable feedback gain, where the gain
is rather high for fast start-up (and high distortion), and rapidly
decrease that feedback gain to "nominal" Colpitts, TPTG, or Pierce
feedback gain (depending on original circuit).

Philips has a pretty cool patent on this.
 
Top