Thanks for all your comments. Since speculation has started here is
what I know about the capacitors.

Those used in the 2.5 GHz source are surplus from a company that makes
high quality stuff. They were probably procured to a military or
space specification but I am not sure. The 10.5 GHz source was
manufactured by MA/COM about 20 years ago. All of them are the usual
deep maroon (is that the right word ?) to brown colour.

Keep in mind I am being pretty picky. I consider short term frequency
jumps of much over 100 Hz to be unsatisfactory - that is 10-40 parts
per billion depending on which source is considered. The largest
observed jumps are about ten times this. Of course, since these are
crystal oscillators, the corresponding capacitance jumps must be much
larger, since the crystal should dominate the oscillator stability. I
would not consider them "crappy", just not as good as one might be led
to expect. I have used capacitors from the same provenance as those
in the 2.5 GHz source in LC oscillators at a few MHz with no observed
problems. The smooth portion of the warm-up drift is reasonably
normal in both cases...only the jumpiness is unusual.

Steve

In article ,
(Steve Kavanagh) writes:

Thanks for all your comments. Since speculation has started here is
what I know about the capacitors.

Those used in the 2.5 GHz source are surplus from a company that makes
high quality stuff. They were probably procured to a military or
space specification but I am not sure. The 10.5 GHz source was
manufactured by MA/COM about 20 years ago. All of them are the usual
deep maroon (is that the right word ?) to brown colour.

Keep in mind I am being pretty picky. I consider short term frequency
jumps of much over 100 Hz to be unsatisfactory - that is 10-40 parts
per billion depending on which source is considered. The largest
observed jumps are about ten times this. Of course, since these are
crystal oscillators, the corresponding capacitance jumps must be much
larger, since the crystal should dominate the oscillator stability. I
would not consider them "crappy", just not as good as one might be led
to expect. I have used capacitors from the same provenance as those
in the 2.5 GHz source in LC oscillators at a few MHz with no observed
problems. The smooth portion of the warm-up drift is reasonably
normal in both cases...only the jumpiness is unusual.

Coming in at the end of this discussion, I'll have to question the
observation of the so-called "frequency jump." At 100 PPB we are
talking quite serious test equipment hook-ups and a number of
different techniques very much uncommon in home workshop
practice.

If you are locking the X-band source to a crystal reference, then
there is a great deal of this frequency-control subsystem which can
be a cause for the "jumps." That can be the sampler or prescaler,
the phase detector (assuming it is a form of PLL), the loop filter, and
even the power supply rail voltage (affecting the voltage control of the
presumed voltage-controlled frequency adjuster circuit). ANY of
those can be the culprit in small frequency "jumps." That would
include whatever it is you are using to heterodyne with the X-band
source to enable frequency measurement.

I'm going to question all those others' claims about "jumpy silver
mica capacitors" after about 54 years of having hands-on
experience in RF and pulse circuitry. I'm talking primarily the
"dipped" coating DMs with some excursions into the molded
plastic case axial lead models. I've never had one either open or
shorted and never "jumpy" in value and that includes the full-on
military environment testing of temperature, altitude, shock,
vibration, etc. A very few were found not quite within the capacity
value tolerance and not a single one experienced any "jumping"
of value. I've not heard of any such stories from contemporaries
in the industry...and I HAVE heard lots of urban-myth stories on
other things within all of electronics.

Len Anderson
retired (from regular hours) electronic engineer person

(Avery Fineman) wrote in message

Coming in at the end of this discussion, I'll have to question the
observation of the so-called "frequency jump." At 100 PPB we are
talking quite serious test equipment

Yes, but the equipment here is up to the task. For the 10.5 GHz case
I am using this source as a receiver LO. A second receiver is
available and both can receive a third source. Both receivers
heterodyne the signal down to audio frequencies where any frequency
steps can be easily detected by ear. The receiver with the M/ACOM LO
jumps (frequency steps) with respect to the third source while the
second receiver does not. A similar scheme was used at 2.5 GHz.

If you are locking the X-band source to a crystal reference, then
there is a great deal of this frequency-control subsystem which can
be a cause for the "jumps." That can be the sampler or prescaler,
the phase detector (assuming it is a form of PLL), the loop filter, and
even the power supply rail voltage (affecting the voltage control of the
presumed voltage-controlled frequency adjuster circuit). ANY of
those can be the culprit in small frequency "jumps."

And from qrk:

Crystals can also jump.

Yes, I have not yet narrowed down the search in the M/ACOM 10.5 GHz
source, but was wondering about the likelihood of the silver mica caps
being the culprit. It has a measured supply voltage sensitivity of
about 50kHz/V (post regulator). The plug-in (not soldered) crystal in
an oven would also seem to be a likely source, particularly as the
jumping seems to improve after warmup (but long after the oven reaches
temperature).

I'm going to question all those others' claims about "jumpy silver
mica capacitors" after about 54 years of having hands-on
experience in RF and pulse circuitry.... I've never had one either open
or shorted and never "jumpy" in value and that includes the full-on
military environment testing of temperature, altitude, shock,
vibration, etc. A very few were found not quite within the capacity
value tolerance and not a single one experienced any "jumping"
of value. I've not heard of any such stories from contemporaries
in the industry...

Well, Tom Bruhns is the first one I have run across who has also noted
this, so it can't be a very commonly experienced effect. But I am
pretty sure about the source of the jumping in the 2.5 GHz source.
Each silver mica capacitor replaced made it better (unless the PC
board just needed the thermal cycling resulting from a few extra
soldering operations).

Steve

In article ,
(Steve Kavanagh) writes:

(Avery Fineman) wrote in message

Coming in at the end of this discussion, I'll have to question the
observation of the so-called "frequency jump." At 100 PPB we are
talking quite serious test equipment

Yes, but the equipment here is up to the task. For the 10.5 GHz case
I am using this source as a receiver LO. A second receiver is
available and both can receive a third source. Both receivers
heterodyne the signal down to audio frequencies where any frequency
steps can be easily detected by ear. The receiver with the M/ACOM LO
jumps (frequency steps) with respect to the third source while the
second receiver does not. A similar scheme was used at 2.5 GHz.

Heterodyning down to an easier-to-use frequency range is an old,
established technique. The problem I see is one of "isolating"
the "jumping," of identifying the "jumps" in terms of time and
occurance (random or with some vague periodicity?).

A substitution method of isolation is necessary to pare down the
number of possibilities of the cause. To get into the parts-per-
billion range of stability, some elaborate measurement systems
need to be used...such as what the NTIS does to verify stability
of frequency sources. Lots of good material on their website for
this sort of thing but the site organization requires some digging
to reach it.

And from qrk:

Crystals can also jump.

Yes, I have not yet narrowed down the search in the M/ACOM 10.5 GHz
source, but was wondering about the likelihood of the silver mica caps
being the culprit. It has a measured supply voltage sensitivity of
about 50kHz/V (post regulator). The plug-in (not soldered) crystal in
an oven would also seem to be a likely source, particularly as the
jumping seems to improve after warmup (but long after the oven reaches
temperature).

If the supply voltage sensitivity is that value, then a 100 Hz to 1 KHz
deviation would be in the neighborhood of 2 to 20 mV supply rail
change. If, for any reason, the supply rail changes, that could cause
the frequency jumps.

I'm going to question all those others' claims about "jumpy silver
mica capacitors" after about 54 years of having hands-on
experience in RF and pulse circuitry.... I've never had one either
open
or shorted and never "jumpy" in value and that includes the full-on
military environment testing of temperature, altitude, shock,
vibration, etc. A very few were found not quite within the capacity
value tolerance and not a single one experienced any "jumping"
of value. I've not heard of any such stories from contemporaries
in the industry...

Well, Tom Bruhns is the first one I have run across who has also noted
this, so it can't be a very commonly experienced effect. But I am
pretty sure about the source of the jumping in the 2.5 GHz source.
Each silver mica capacitor replaced made it better (unless the PC
board just needed the thermal cycling resulting from a few extra
soldering operations).

The extra stress of capacitor replacement might have affected
some other "jump" cause, thus masking that by what you thought
might be the culprit.

I'm not doing any personal accusation thing here, just trying to get
some isolation on exactly where the possible cause might be other
than "jumpy silver micas." I have encountered such weird things as
exposed copper PCB traces gathering oxides over time, such oxides
actually bridging the supposed isulating gap between traces and
causing very minute fluctuations down in the 1 mV region with a DC
bias between conductors. On another case, of using "stiff" wire
jumpers (which were NOT entirely copper but plated nickel) in a
rotary switch assembly, the resulting temperature-sensitive voltage
(like a thermocouple junction) caused an out-of-scale drift in a
supposedly-simple meter circuit. Visual inspection would have
passed it no problem but a magnet showed an attraction to the
jumper wires indicating it was not copper. Even if the jumpers were
soldered in with good visible junctions, the thermal EMF from
dissimilar metals was enough to throw off calibration over the mild
industrial-grade temperature rating.

I'm not saying either of those conditions might be present with your
problem, just trying to illustrate weirdness that can happen. For
safety sake (of mind), I coat circuit traces around sensitive areas
such as oscillators, multivibrators, low-level DC and AC circuits with
some petroleum-based varnish (such as McCloskey's "Gym Seal")
after doing some simple, non-plated circuit board assemblies having
bare etched copper traces.

Having done a fair amount of L-, S-, and C-band RF work (that's 1 to
to 8 GHz for civilians who never bothered with the "new" military
alphabetic bands and preferred the ancient letters), and using the
heterodyning technique to bring them down to measureable regions,
I've encountered lots of different causes for oscillator instability that
ranges from supposedly-stable DC supply rails to the oxide problem
mentioned to odd things such as magnetic coupling from AC cords
to various ferrites used for external frequency control. Anything is
suspect and, at the tolerances you need, aren't given as standard
in texts.

I'd start with giving the supply rails a workup down to a gnat's
eyelash on stability, both at DC and low AC. Then I'd try out the
local environment, the bench, even to trying to raise the oscillator
by supporting it on a small piece of plywood (might be some power
wiring under the bench surface). Then it is small spritzes of cooling
spray, very small in small areas. Even banging on the support with
a tack hammer or equivalent weight for minor shock and vibration.
Then it would evolve into going "split personality" and trying to look
at the assembly critically as if you were another person just finding
out about a problem and not pre-judging what "might" be happening.

Now, it MIGHT be a faulty component made by someone else,
including the crystal unit, but pre-judging a "fault" is not a good way
to start. I've had contemporaries relate lots of urban myth stories
(like electrolytics "always go bad with time" but don't) about "cause"
but those aren't always truisms. The approach has to be analytical,
one of isolating all the factors that test good in order to narrow the
source area to a very small one. Easy that ain't. Perseverance
is an absolute must when pushing the envelope.

I have yet to discover a "jumpy silver mica capacitor" but I might find
one even with over a half century of playing RF games with parts and
going from the kind of micas found in old ARC-5 radios to sampling
PLL loop stabilized sources using DM30s in the loop filter (but had
one unidentifiable-cause varactor that was replaced but never
examined in detail later).

Len Anderson
retired (from regular hours) electronic engineer person

On 20 May 2004 06:12:49 -0700, (Steve
Kavanagh) wrote:

[snippage]
Well, Tom Bruhns is the first one I have run across who has also noted
this, so it can't be a very commonly experienced effect.
[snippage]

Steve

I doubt that many people would notice ppb changes and jumps. This
takes a bit of patience and ruling out bad test equipment/setup to
observe this phenomena.

Mark

qrk wrote:
On 20 May 2004 06:12:49 -0700, (Steve
Kavanagh) wrote:

[snippage]
Well, Tom Bruhns is the first one I have run across who has also noted
this, so it can't be a very commonly experienced effect.
[snippage]

Steve

I doubt that many people would notice ppb changes and jumps. This
takes a bit of patience and ruling out bad test equipment/setup to
observe this phenomena.

It has certainly been noticed by many more people than Tom.

The word "scintillation" rang a very faint bell, and Google found a
reference at:
http://www.seas.gwu.edu/~ecelabs/app...data/page2.pdf

These scanned pages from an unknown reference book define:
"Scintillation: minute and rapid fluctuations of capacitance, formerly
exhibited by silvered mica and silvered ceramic types [of capacitors]
but overcome by modern manufacturing techniques."

Well, maybe not *totally* overcome...

This explains why we only tend to hear about the problem in very old
capacitors (probably WW2 era) or in critical applications such as
precision oscillators.

The reference to silvered-ceramic capacitors is interesting. Evidently
that scintillation problem was "overcome" more completely than for
silvered-mica, which is why NP0 ceramic are now the capacitors of choice
for oscillator applications.


--
73 from Ian G3SEK 'In Practice' columnist for RadCom (RSGB)
http://www.ifwtech.co.uk/g3sek

In article , Ian White, G3SEK
writes
qrk wrote:
On 20 May 2004 06:12:49 -0700, (Steve
Kavanagh) wrote:

[snippage]
Well, Tom Bruhns is the first one I have run across who has also noted
this, so it can't be a very commonly experienced effect.
[snippage]

Steve

I doubt that many people would notice ppb changes and jumps. This
takes a bit of patience and ruling out bad test equipment/setup to
observe this phenomena.

It has certainly been noticed by many more people than Tom.

The word "scintillation" rang a very faint bell, and Google found a
reference at:
http://www.seas.gwu.edu/~ecelabs/app...data/page2.pdf

These scanned pages from an unknown reference book define:
"Scintillation: minute and rapid fluctuations of capacitance, formerly
exhibited by silvered mica and silvered ceramic types [of capacitors]
but overcome by modern manufacturing techniques."

Well, maybe not *totally* overcome...

This explains why we only tend to hear about the problem in very old
capacitors (probably WW2 era) or in critical applications such as
precision oscillators.

The reference to silvered-ceramic capacitors is interesting. Evidently
that scintillation problem was "overcome" more completely than for
silvered-mica, which is why NP0 ceramic are now the capacitors of choice
for oscillator applications.


in the crystal oscillator business silver mica capacitors were known for
scintillation . The potting compound of silvered mica capacitors was
often the cause of temperature coefficient drift. Scintilation was
probably due to delamination of the mica.
There are 2 types of mica caps cleaved mica and compressed? which
powdered the mica and then re-formed. Im not sure about their relative
scintillation .
Modern NPO ceramic are probably better particularly un-encapsulated
surface mount.
Note that I would design an overtone crystal oscillator with only enough
reactance to remove the manufacturing tolerance. This reactance does not
have to be capacitative could be inductive capacitative reactance could
alternatively be a varicap then the problem would be a clean varicap
supply.
Note crystals can do strange things, the jumps described are too small
for unwanted modes but there is the well known (to TCXO designers) band
breaks these are small frequency jumps that occur at exact temperatures
and are due to minor modes passing through the major mode frequency at a
particular temperature. This is what limits TCXO performance. OCXO
makers make sure that the set temperature is not on a bandbreak.

--
ddwyer

In article , Ian White, G3SEK
writes
qrk wrote:
On 20 May 2004 06:12:49 -0700, (Steve
Kavanagh) wrote:

[snippage]
Well, Tom Bruhns is the first one I have run across who has also noted
this, so it can't be a very commonly experienced effect.
[snippage]

Steve

I doubt that many people would notice ppb changes and jumps. This
takes a bit of patience and ruling out bad test equipment/setup to
observe this phenomena.

It has certainly been noticed by many more people than Tom.

The word "scintillation" rang a very faint bell, and Google found a
reference at:
http://www.seas.gwu.edu/~ecelabs/app...data/page2.pdf

These scanned pages from an unknown reference book define:
"Scintillation: minute and rapid fluctuations of capacitance, formerly
exhibited by silvered mica and silvered ceramic types [of capacitors]
but overcome by modern manufacturing techniques."

Well, maybe not *totally* overcome...

This explains why we only tend to hear about the problem in very old
capacitors (probably WW2 era) or in critical applications such as
precision oscillators.

The reference to silvered-ceramic capacitors is interesting. Evidently
that scintillation problem was "overcome" more completely than for
silvered-mica, which is why NP0 ceramic are now the capacitors of choice
for oscillator applications.


in the crystal oscillator business silver mica capacitors were known for
scintillation . The potting compound of silvered mica capacitors was
often the cause of temperature coefficient drift. Scintilation was
probably due to delamination of the mica.
There are 2 types of mica caps cleaved mica and compressed? which
powdered the mica and then re-formed. Im not sure about their relative
scintillation .
Modern NPO ceramic are probably better particularly un-encapsulated
surface mount.
Note that I would design an overtone crystal oscillator with only enough
reactance to remove the manufacturing tolerance. This reactance does not
have to be capacitative could be inductive capacitative reactance could
alternatively be a varicap then the problem would be a clean varicap
supply.
Note crystals can do strange things, the jumps described are too small
for unwanted modes but there is the well known (to TCXO designers) band
breaks these are small frequency jumps that occur at exact temperatures
and are due to minor modes passing through the major mode frequency at a
particular temperature. This is what limits TCXO performance. OCXO
makers make sure that the set temperature is not on a bandbreak.

--
ddwyer

qrk wrote:
On 20 May 2004 06:12:49 -0700, (Steve
Kavanagh) wrote:

[snippage]
Well, Tom Bruhns is the first one I have run across who has also noted
this, so it can't be a very commonly experienced effect.
[snippage]

Steve

I doubt that many people would notice ppb changes and jumps. This
takes a bit of patience and ruling out bad test equipment/setup to
observe this phenomena.

It has certainly been noticed by many more people than Tom.

The word "scintillation" rang a very faint bell, and Google found a
reference at:
http://www.seas.gwu.edu/~ecelabs/app...data/page2.pdf

These scanned pages from an unknown reference book define:
"Scintillation: minute and rapid fluctuations of capacitance, formerly
exhibited by silvered mica and silvered ceramic types [of capacitors]
but overcome by modern manufacturing techniques."

Well, maybe not *totally* overcome...

This explains why we only tend to hear about the problem in very old
capacitors (probably WW2 era) or in critical applications such as
precision oscillators.

The reference to silvered-ceramic capacitors is interesting. Evidently
that scintillation problem was "overcome" more completely than for
silvered-mica, which is why NP0 ceramic are now the capacitors of choice
for oscillator applications.


--
73 from Ian G3SEK 'In Practice' columnist for RadCom (RSGB)
http://www.ifwtech.co.uk/g3sek

In article ,
(Steve Kavanagh) writes:

(Avery Fineman) wrote in message

Coming in at the end of this discussion, I'll have to question the
observation of the so-called "frequency jump." At 100 PPB we are
talking quite serious test equipment

Yes, but the equipment here is up to the task. For the 10.5 GHz case
I am using this source as a receiver LO. A second receiver is
available and both can receive a third source. Both receivers
heterodyne the signal down to audio frequencies where any frequency
steps can be easily detected by ear. The receiver with the M/ACOM LO
jumps (frequency steps) with respect to the third source while the
second receiver does not. A similar scheme was used at 2.5 GHz.

Heterodyning down to an easier-to-use frequency range is an old,
established technique. The problem I see is one of "isolating"
the "jumping," of identifying the "jumps" in terms of time and
occurance (random or with some vague periodicity?).

A substitution method of isolation is necessary to pare down the
number of possibilities of the cause. To get into the parts-per-
billion range of stability, some elaborate measurement systems
need to be used...such as what the NTIS does to verify stability
of frequency sources. Lots of good material on their website for
this sort of thing but the site organization requires some digging
to reach it.

And from qrk:

Crystals can also jump.

Yes, I have not yet narrowed down the search in the M/ACOM 10.5 GHz
source, but was wondering about the likelihood of the silver mica caps
being the culprit. It has a measured supply voltage sensitivity of
about 50kHz/V (post regulator). The plug-in (not soldered) crystal in
an oven would also seem to be a likely source, particularly as the
jumping seems to improve after warmup (but long after the oven reaches
temperature).

If the supply voltage sensitivity is that value, then a 100 Hz to 1 KHz
deviation would be in the neighborhood of 2 to 20 mV supply rail
change. If, for any reason, the supply rail changes, that could cause
the frequency jumps.

I'm going to question all those others' claims about "jumpy silver
mica capacitors" after about 54 years of having hands-on
experience in RF and pulse circuitry.... I've never had one either
open
or shorted and never "jumpy" in value and that includes the full-on
military environment testing of temperature, altitude, shock,
vibration, etc. A very few were found not quite within the capacity
value tolerance and not a single one experienced any "jumping"
of value. I've not heard of any such stories from contemporaries
in the industry...

Well, Tom Bruhns is the first one I have run across who has also noted
this, so it can't be a very commonly experienced effect. But I am
pretty sure about the source of the jumping in the 2.5 GHz source.
Each silver mica capacitor replaced made it better (unless the PC
board just needed the thermal cycling resulting from a few extra
soldering operations).

The extra stress of capacitor replacement might have affected
some other "jump" cause, thus masking that by what you thought
might be the culprit.

I'm not doing any personal accusation thing here, just trying to get
some isolation on exactly where the possible cause might be other
than "jumpy silver micas." I have encountered such weird things as
exposed copper PCB traces gathering oxides over time, such oxides
actually bridging the supposed isulating gap between traces and
causing very minute fluctuations down in the 1 mV region with a DC
bias between conductors. On another case, of using "stiff" wire
jumpers (which were NOT entirely copper but plated nickel) in a
rotary switch assembly, the resulting temperature-sensitive voltage
(like a thermocouple junction) caused an out-of-scale drift in a
supposedly-simple meter circuit. Visual inspection would have
passed it no problem but a magnet showed an attraction to the
jumper wires indicating it was not copper. Even if the jumpers were
soldered in with good visible junctions, the thermal EMF from
dissimilar metals was enough to throw off calibration over the mild
industrial-grade temperature rating.

I'm not saying either of those conditions might be present with your
problem, just trying to illustrate weirdness that can happen. For
safety sake (of mind), I coat circuit traces around sensitive areas
such as oscillators, multivibrators, low-level DC and AC circuits with
some petroleum-based varnish (such as McCloskey's "Gym Seal")
after doing some simple, non-plated circuit board assemblies having
bare etched copper traces.

Having done a fair amount of L-, S-, and C-band RF work (that's 1 to
to 8 GHz for civilians who never bothered with the "new" military
alphabetic bands and preferred the ancient letters), and using the
heterodyning technique to bring them down to measureable regions,
I've encountered lots of different causes for oscillator instability that
ranges from supposedly-stable DC supply rails to the oxide problem
mentioned to odd things such as magnetic coupling from AC cords
to various ferrites used for external frequency control. Anything is
suspect and, at the tolerances you need, aren't given as standard
in texts.

I'd start with giving the supply rails a workup down to a gnat's
eyelash on stability, both at DC and low AC. Then I'd try out the
local environment, the bench, even to trying to raise the oscillator
by supporting it on a small piece of plywood (might be some power
wiring under the bench surface). Then it is small spritzes of cooling
spray, very small in small areas. Even banging on the support with
a tack hammer or equivalent weight for minor shock and vibration.
Then it would evolve into going "split personality" and trying to look
at the assembly critically as if you were another person just finding
out about a problem and not pre-judging what "might" be happening.

Now, it MIGHT be a faulty component made by someone else,
including the crystal unit, but pre-judging a "fault" is not a good way
to start. I've had contemporaries relate lots of urban myth stories
(like electrolytics "always go bad with time" but don't) about "cause"
but those aren't always truisms. The approach has to be analytical,
one of isolating all the factors that test good in order to narrow the
source area to a very small one. Easy that ain't. Perseverance
is an absolute must when pushing the envelope.

I have yet to discover a "jumpy silver mica capacitor" but I might find
one even with over a half century of playing RF games with parts and
going from the kind of micas found in old ARC-5 radios to sampling
PLL loop stabilized sources using DM30s in the loop filter (but had
one unidentifiable-cause varactor that was replaced but never
examined in detail later).

Len Anderson
retired (from regular hours) electronic engineer person