packages of two WWVB watches: $20 at www.woot.com while they last <EOM>
Magnus,
This discussion is why I have allowed the ADA4899-1 to saturate in the ZCD
posted earlier. The transition times up & down in the second stage are 350
to
400 ns; thus I believe the <50ns overload recovery doesn't degrade the
measured jitter. With 500MHz 'scope bandwidth, the 1KHz squarewave
output has <1% over/under shoot & appears noise free at the zero crossing.
The output transitions are fast enough for my HP 5370A to show < 1ns
jitter over 1K or 10K samples.
Regards,
Pete Rawson
Pete wrote:
); SAEximRunCond expanded to false
Errors-To: time-nuts-bounces+bruce.griffiths=xtra.co.nz+bruce.griffiths=xtra.co.nz@febo.com
Magnus,
This discussion is why I have allowed the ADA4899-1 to saturate in the ZCD
posted earlier. The transition times up & down in the second stage are 350
to
400 ns; thus I believe the <50ns overload recovery doesn't degrade the
measured jitter. With 500MHz 'scope bandwidth, the 1KHz squarewave
output has <1% over/under shoot & appears noise free at the zero crossing.
The output transitions are fast enough for my HP 5370A to show < 1ns
jitter over 1K or 10K samples.
Regards,
Pete Rawson
Pete
What about the power supply noise that appears at the amplifier output
during saturation?
Power supply noise is almost always much larger than the opamp input noise.
A simple resistor + diode clamp in addition to the feedback clamp will
easily attenuate the output noise of an opamp with diode clamps across
its feedback resistor.
An npn plus a pnp resistor ccan also be used to clamp the output of an
opamp whilst still suppressing the noise gain to a very low value and
prevent opamp saturation.
Bruce
Operating an opamp in saturation for extended periods in a precision
circuit is a bad idea, as the changes in dissipation of the various
devices in the opamp will change the opamp input offset voltage when it
comes out of saturation. Even though modern opamps use cross coupled
quads and other layout symmetry techniques to greatly reduce the thermal
feedback on the chip, the effect is still finite. With input frequencies
above a few kHz the thermal feedback isn't significant, however with the
low frequencies typically used in this application thermal feedback is
significant. Thermal feedback is reduced by keeping the opamp out of
saturation and limiting the output swing to minimise the change in the
power dissipation in the output stage transistors.
Bruce
Pete wrote:
Magnus,
This discussion is why I have allowed the ADA4899-1 to saturate in the ZCD
posted earlier. The transition times up & down in the second stage are 350
to
400 ns; thus I believe the <50ns overload recovery doesn't degrade the
measured jitter. With 500MHz 'scope bandwidth, the 1KHz squarewave
output has <1% over/under shoot & appears noise free at the zero crossing.
The output transitions are fast enough for my HP 5370A to show < 1ns
jitter over 1K or 10K samples.
Regards,
Pete Rawson
Pete
Try measuring your circuits output noise spectrum.
Also measure the temperature dependance of its input output phase shift.
What does this tell you about how stable its temperature needs to be to
limit the phase shift change to a few nanosconds, or even a microsecond?
Bruce
Dr Bruce Griffiths wrote:
); SAEximRunCond expanded to false
Errors-To: time-nuts-bounces+cfharris=erols.com+cfharris=erols.com@febo.com
Pete wrote:
); SAEximRunCond expanded to false
Errors-To: time-nuts-bounces+bruce.griffiths=xtra.co.nz+bruce.griffiths=xtra.co.nz@febo.com
Magnus,
This discussion is why I have allowed the ADA4899-1 to saturate in the ZCD
posted earlier. The transition times up & down in the second stage are 350
to
400 ns; thus I believe the <50ns overload recovery doesn't degrade the
measured jitter. With 500MHz 'scope bandwidth, the 1KHz squarewave
output has <1% over/under shoot & appears noise free at the zero crossing.
The output transitions are fast enough for my HP 5370A to show < 1ns
jitter over 1K or 10K samples.
Regards,
Pete Rawson
Pete
What about the power supply noise that appears at the amplifier output
during saturation?
Power supply noise is almost always much larger than the opamp input noise.
A simple resistor + diode clamp in addition to the feedback clamp will
easily attenuate the output noise of an opamp with diode clamps across
its feedback resistor.
The output of an opamp is always referenced to its power supply,
regardless of the configuration. If the power supply is noisy,
the output is noisy.
-Chuck Harris
Chuck Harris wrote:
What about the power supply noise that appears at the amplifier output during saturation?
Power supply noise is almost always much larger than the opamp input noise.
A simple resistor + diode clamp in addition to the feedback clamp will
easily attenuate the output noise of an opamp with diode clamps across
its feedback resistor.
The output of an opamp is always referenced to its power supply,
regardless of the configuration. If the power supply is noisy,
the output is noisy.
-Chuck Harris
Only when that opamp output is saturated.
When in linear operation the power supply rejection, at least at low
frequency, is significant.
Bruce
The output of an opamp is always referenced to its power supply,
regardless of the configuration. If the power supply is noisy,
the output is noisy.
Beyond a few kHz this may be true, but at lower frequencies even a low-grade
opamp has excellent rejection of power supply noise. A good opamp data
sheet will characterize the PSRR versus frequency.
-- john, KE5FX
Pete
Try changing your input stage to the configuration depicted in the
attachment.
The input amplifier can then be something like an OP27 without affecting
performance.
The noise spectrum at the first opamp output will actually be reduced to
the opamp (and series resistor) input noise at frequencies outside the
tuned circuit bandpass rather than being significantly enhanced as in
you circuit. The first opamp will ouput also be dc stable with a zero
signal output of around zero volts unlike your circuit.
This will do nothing for the rather large phase shift tempco, but at
least its a more rational design.
In fact your circuit performance should improve if you just delete the
first tuned circuit.
You still need to think about a suitable limiter that doesn't switch the
power supply noise to the output during stauration.
Bruce
John Miles wrote:
The output of an opamp is always referenced to its power supply,
regardless of the configuration. If the power supply is noisy,
the output is noisy.
Beyond a few kHz this may be true, but at lower frequencies even a low-grade
opamp has excellent rejection of power supply noise. A good opamp data
sheet will characterize the PSRR versus frequency.
-- john, KE5FX
John
Only when the opamp is operating within its linear region, not when the
output is saturated.
Opamp saturation has been proposed as a good method of reducing the
output noise of an opamp when an amplifier is overdriven.
This is clearly not true when power supply noise is significant, as is
almost always the case.
Bruce
For those of you who are unfamiliar with the bipolar transistor feedback
limiter the attached schematic depicts it in use with both a
noninverting and an inverting amplifier.
The output voltage swing before clamping is limited by transistor Vbe
breakdown. However this can be oversome by adding a few diodes when a
larger output swing is required.
Unlike the simple shunt diode feedback clamp, the circuit noise gain is
less than 1 when the output is clamped. Off course the voltage noise at
one of the clamp transistor bases appears at the output when the output
is clamped. However this voltage is generally less than the power supply
and its noise can easily be made significantly lower.
Bruce