Pierce ACMOS Oscillator Stability

[Rousty]

Recently I ran into a Pierce CMOS oscillator built froma 20MHz
fundamental mode quartz crystal, a 74AC04 active element, a 100K
linearizing resistor and the appropriate loading and phase shifting
elements (1.4K resisitor and 2 X 39pF capacitors) inthe traditional
configuration. The circuit has exhibited normal oscillation and a
relaxation mode oscillation at about 11MHz. While in the relaxation
mode, I can clearly see evidence of hysteresis in the 74AC04 which
allows the relaxation mode to exist. Has anyone noticed this? Is it
due to the low value of linearizing resistor (100K) since other designs
typically use greater than 1M?

 

[Jim Thompson]

Recently I ran into a Pierce CMOS oscillator built froma 20MHz
fundamental mode quartz crystal, a 74AC04 active element, a 100K
linearizing resistor and the appropriate loading and phase shifting
elements (1.4K resisitor and 2 X 39pF capacitors) inthe traditional
configuration. The circuit has exhibited normal oscillation and a
relaxation mode oscillation at about 11MHz. While in the relaxation
mode, I can clearly see evidence of hysteresis in the 74AC04 which
allows the relaxation mode to exist. Has anyone noticed this? Is it
due to the low value of linearizing resistor (100K) since other designs
typically use greater than 1M?

Why aren't you using 'HCU04 ??

"Linearizing" resistor ?? No. It's DC feedback for "Q-point"
control.

...Jim Thompson

 

[John Popelish]

Recently I ran into a Pierce CMOS oscillator built froma 20MHz
fundamental mode quartz crystal, a 74AC04 active element, a 100K
linearizing resistor and the appropriate loading and phase shifting
elements (1.4K resisitor and 2 X 39pF capacitors) inthe traditional
configuration. The circuit has exhibited normal oscillation and a
relaxation mode oscillation at about 11MHz. While in the relaxation
mode, I can clearly see evidence of hysteresis in the 74AC04 which
allows the relaxation mode to exist. Has anyone noticed this? Is it
due to the low value of linearizing resistor (100K) since other designs
typically use greater than 1M?

I don't think the 74HC04 has hysteresis. I think it has three gain
stages in series, so when DC biased with a single resistor , has
enough phase shift at the frequency where the gain is 1 to exhibit
positive feedback. I would try splitting the bias resistor into two
resistors and filtering the center node to ground, with a largish
capacitor, to keep the high frequency through this path low. You
might also have to divide the voltage into or out of the crystal with
a capacitive divider, to lower the total loop gain at resonance (and
at all other frequencies) to suppress the spurious oscillation.

See figure 4 of:
http://www.irf.com/product-info/datasheets/data/irf9z20.pdf

 

[Rousty]

I don't think the 74HC04 has hysteresis. I think it has three gain
stages in series, so when DC biased with a single resistor , has
enough phase shift at the frequency where the gain is 1 to exhibit
positive feedback. I would try splitting the bias resistor into two
resistors and filtering the center node to ground, with a largish
capacitor, to keep the high frequency through this path low. You
might also have to divide the voltage into or out of the crystal with
a capacitive divider, to lower the total loop gain at resonance (and
at all other frequencies) to suppress the spurious oscillation.

See figure 4 of:
http://www.irf.com/product-info/datasheets/data/irf9z20.pdf

John,

Your are right about the multistage aspect, there are three inverting
stages in each inverter circuit. The AC04 is not intended to exhibit
hysteresis (as does the AC14) however I have experimentally
demonstrated that static (DC) hysteresis on the order of 100mV does
exist in the 3-5V supply range. This must be due to some weak internal
positive feedback (possibly leakage). Originally I thought it might be
due to low voltage operation in the subthreshold region of the N and P
FETS but is does not diminish with increasing supply voltage. Probably
some sort of substrate leakage effect. I'm going to try some
temperature testing to establish a temperature effect. The unwanted
oscillation I have expereinced is clearly not due to delay/phase shift
combined with sufficient loop gain. It is a true relaxation
oscillation that is supported by the hysteresis of the device.

Thank you for your thoughts...

 

[Jim Thompson]

[snip]

The AC04 is not intended to exhibit
hysteresis (as does the AC14) however I have experimentally
demonstrated that static (DC) hysteresis on the order of 100mV does
exist in the 3-5V supply range. This must be due to some weak internal
positive feedback (possibly leakage). Originally I thought it might be
due to low voltage operation in the subthreshold region of the N and P
FETS but is does not diminish with increasing supply voltage. Probably
some sort of substrate leakage effect. I'm going to try some
temperature testing to establish a temperature effect. The unwanted
oscillation I have expereinced is clearly not due to delay/phase shift
combined with sufficient loop gain. It is a true relaxation
oscillation that is supported by the hysteresis of the device.

Thank you for your thoughts...

Carefully observe the power supply. It's probably bouncing.

...Jim Thompson

 

[Rousty]

Why aren't you using 'HCU04 ??

"Linearizing" resistor ?? No. It's DC feedback for "Q-point"
control.

...Jim Thompson

Jim,

The HCU04 would be a good choice due to its single stage architecture
however the final design is intended to be in a hi-rel space
application and that will require a radiation hardened device like the
JM38510R75701SDA (i.e. the radiation rardened version of the 54AC04).
Unfortunately, there is no domestic source for radiation hardened
HCU04's so the AC04 is all we have. The feedback resistor that
establishes quasi-linear operation and allows for oscillator start-up
has been refered to as a linearizing resistor by personnel in the
oscillator business. It has two purposes, during startup it provides
for quasilinear operation so low level linear feedback loop operation
can occur (note that outside of the quasi linear region, the gain of
the inverter as well as the composit loop gain is zero). Operation in
the quasilinear region supports the growth of the intended oscillation.
Once oscillation has saturated the output (i.e. the limiting aspect of
the circuit), the resistor serves to provide approximate DC bias to the
input of the inverter. Note that these are two distinct and different
modes of circuit operation (startup and operating modes).

Thank you for your thoughts.....

 

[Jim Thompson]

Jim,

The HCU04 would be a good choice due to its single stage architecture
however the final design is intended to be in a hi-rel space
application and that will require a radiation hardened device like the
JM38510R75701SDA (i.e. the radiation rardened version of the 54AC04).
Unfortunately, there is no domestic source for radiation hardened
HCU04's so the AC04 is all we have. The feedback resistor that
establishes quasi-linear operation and allows for oscillator start-up
has been refered to as a linearizing resistor by personnel in the
oscillator business. It has two purposes, during startup it provides
for quasilinear operation so low level linear feedback loop operation
can occur (note that outside of the quasi linear region, the gain of
the inverter as well as the composit loop gain is zero). Operation in
the quasilinear region supports the growth of the intended oscillation.
Once oscillation has saturated the output (i.e. the limiting aspect of
the circuit), the resistor serves to provide approximate DC bias to the
input of the inverter. Note that these are two distinct and different
modes of circuit operation (startup and operating modes).

Thank you for your thoughts.....

Sounds like too much book-learning and not enough hands-on. Why is it
you have the relaxation mode oscillation? Start-up my ass ;-)

http://analog-innovations.com/SED/WhenInDoubt.jpg

...Jim Thompson

 

[Genome]

Jim,

The HCU04 would be a good choice due to its single stage architecture
however the final design is intended to be in a hi-rel space
application and that will require a radiation hardened device like the
JM38510R75701SDA (i.e. the radiation rardened version of the 54AC04).
Unfortunately, there is no domestic source for radiation hardened
HCU04's so the AC04 is all we have. The feedback resistor that
establishes quasi-linear operation and allows for oscillator start-up
has been refered to as a linearizing resistor by personnel in the
oscillator business. It has two purposes, during startup it provides
for quasilinear operation so low level linear feedback loop operation
can occur (note that outside of the quasi linear region, the gain of
the inverter as well as the composit loop gain is zero). Operation in
the quasilinear region supports the growth of the intended oscillation.
Once oscillation has saturated the output (i.e. the limiting aspect of
the circuit), the resistor serves to provide approximate DC bias to the
input of the inverter. Note that these are two distinct and different
modes of circuit operation (startup and operating modes).

Thank you for your thoughts.....

Please don't be offended but you are seriously blowing it out of your arse.
It won't matter if your IC is RAD hard when it's in space and 'not'
oscillating. Before you try and bullshit your way through a bunch of rubbish
do yourself a favor and do a web search on crystal oscillators, in
particular have a look at what your crystal manufacturer has to say about
things.

You are using a digital IC in a linear manner and you are designing a
feedback loop to be nominally unstable at or close to the crystals
fundamental frequency. The only 'logic' IC's designed to work in this
application are unbuffered inverters with a single P/N channel stage. This
is because they will behave and can be designed to work in a 'predictable'
as a gain stage.

The only reason you may have seen it attempted with 'standard' gates is
because some asshole of a digital johnny didn't understand what U in the
part number meant and just saw the symbol and saw he had a spare inverter
knocking about the place so he used that.

Three months down the line all the product came back because the process
varied and the oscillators started not to. What is the UPS charge for
spacecraft recovery because you will be one seriously sick bunny if you try
it and it does not work. While you're at it do a bit of reasearch into
feedback loops and loop stability because, at the moment you are dribbling
about like some sort of loony.

If you want to make it RAD hard tear a piece off your tinfoil hat and make a
little house for it. I really dislike 'reasonable people'.

DNA

 

[Joerg]

Hello Jim,

Carefully observe the power supply. It's probably bouncing.

Yes. And hopefully there is a full ground plane. Else GND of the little
AC04 could be bouncing as well.

 

[Jim Thompson]

[snip]

Three months down the line all the product came back because the process
varied and the oscillators started not to. What is the UPS charge for
spacecraft recovery because you will be one seriously sick bunny if you try
it and it does not work. While you're at it do a bit of reasearch into
feedback loops and loop stability because, at the moment you are dribbling
about like some sort of loony.

If you want to make it RAD hard tear a piece off your tinfoil hat and make a
little house for it. I really dislike 'reasonable people'.

DNA

Bwahahahaha ;-)

Genome/DNA, I've decided to host a SED reunion just so I can meet you
;-)

...Jim Thompson

 

[Genome]

Bwahahahaha ;-)

Genome/DNA, I've decided to host a SED reunion just so I can meet you
;-)

...Jim Thompson

Thanks, but I am sure you would be severely disappointed and I'd be picked
up for vagrancy long before I got close.

In the meantime I'm looking closer at these things and they are not trivial.
What got up my nose was the blah de blah.....

Anyway, closest I can get to some concept of Mr Rousty's 'hysteretic'
problem is the two internal nodes (if there are three stages) are going ape
shit. The relatively lazy oscillation outside the IC biases them up and down
and they end up demodulating the signal, as the operating point and whatever
varies, to produce the DC offsets within the IC.

Something like that anyway. When you mention looking for the supply rails
'bouncing' I think you might find a sniff of current will show some really
high frequency stuff going on.......

That's my story anyway.

DNA

 

[Jim Thompson]

Thanks, but I am sure you would be severely disappointed and I'd be picked
up for vagrancy long before I got close.

In the meantime I'm looking closer at these things and they are not trivial.
What got up my nose was the blah de blah.....

Anyway, closest I can get to some concept of Mr Rousty's 'hysteretic'
problem is the two internal nodes (if there are three stages) are going ape
shit. The relatively lazy oscillation outside the IC biases them up and down
and they end up demodulating the signal, as the operating point and whatever
varies, to produce the DC offsets within the IC.

Something like that anyway. When you mention looking for the supply rails
'bouncing' I think you might find a sniff of current will show some really
high frequency stuff going on.......

That's my story anyway.

DNA

A very likely scenario.

...Jim Thompson