Purpose of separate oven and electronics power supplies for OCXOs
Many OCXOs permit the use of separate power supplies for the oven and
electronics. Why is this beneficial?
The oven requires more power (at least during warm-up) and I presume is
less sensitive to supply noise translating into phase noise at the output.
That would permit the use of a higher-power possibly noisier supply for the
oven and a cleaner, lower-power supply for the electronics. Is this
correct? Are there other power supply factors affecting the output
frequency? If so, is their effect through the oven or electronics supply,
or both? For example, the FTS 1000B datasheet gives a 1% supply sensitivity
of <1e-11 fractional frequency change, but doesn't specify to which supply
this applies. The datasheet also gives an "EMI susceptibility", which is
basically the sensitivity of the output phase noise to supply noise. They
don't indicate the relevant supply for this either, but I imagine it's the
electronics supply.
Thanks,
Matt
Matt Huszagh via time-nuts writes:
Many OCXOs permit the use of separate power supplies for the oven and
electronics. Why is this beneficial?
For old-style OCXO's the logic is:
The OCXO is connected to the power supply via some wiring, and
we cannot ignore that when we have to use scientific notation
to report our stability.
The electronics draws constant current, which means that the voltage
over the resitance of the wiring is also constant, and so the supply
voltage inside the OCXO will be slightly lower than at the external
voltage regulator, but it will be constant.
The oven draws whatever current it needs to compensate for the
temperature difference between inside and outside, so the voltage
over the resistance of the wiring to the power supply will vary
proportionally.
QED: There should be separate pins, and maybe even separate
external regulators, for oven and electronics.
Modern OCXOs solve the problem by including a chip voltage regulator
for the electronics inside the oven, which makes it so stable
that an external voltage regulator would need to be in LM399
territory to compete.
Poul-Henning
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
On Sat, 18 Jan 2025 09:35:03 -0800
Matt Huszagh via time-nuts time-nuts@lists.febo.com wrote:
Many OCXOs permit the use of separate power supplies for the oven and
electronics. Why is this beneficial?
Because it decouples the oven power supply from the oscillator supply
The oven requires more power (at least during warm-up) and I presume is
less sensitive to supply noise translating into phase noise at the output.
That would permit the use of a higher-power possibly noisier supply for the
oven and a cleaner, lower-power supply for the electronics. Is this
correct? Are there other power supply factors affecting the output
frequency? If so, is their effect through the oven or electronics supply,
or both? For example, the FTS 1000B datasheet gives a 1% supply sensitivity
of <1e-11 fractional frequency change, but doesn't specify to which supply
this applies. The datasheet also gives an "EMI susceptibility", which is
basically the sensitivity of the output phase noise to supply noise. They
don't indicate the relevant supply for this either, but I imagine it's the
electronics supply.
You are on the right path. There are two issues 1) noise injection and
2) voltage modulation.
- Noise injection
With noise injection into the supply of the oscillator directly affects
the output. How much, is dependent on the quality of the internal regulator.
A decent regulator has 20-30dB of damping for the relevant frequencies, but
it usually drops at higher frequencies and can go to zero. A properly designed
power supply has input filters that filter out high frequency noise to
further insulate the device from these influences.
While the oven can create some noise, it is unlikely to do so. It is fairly
easy to design an oven control that does not generate any significant noise
above, let's say 10HHz... The bigger issue is the external power supply of
the OCXO. It s much easier to build a low noise power supply that only has to
supply 10mA than it is to build one that can supply the 500mA (sometimes 1A)
that are required for the initial warm-up. But even then, using a decent
LDO with a pre-filter is usually enough to give the OCXO a clean enough
power supply that it doesn't matter.
- Voltage Modulation
All OCXO have some supply voltage dependent frequency modulation. The amplifiers
are non-linear circuits (all transistors are non-linear, which means
all circuits built from transistors are non-linear) and have some dependency
of their phase and amplification on the supply voltage. I.e. the changes in
supply voltage slightly change the phase/amplification of the sustaining
amplifier of the oscillator, which in turn slightly shifts the stady-state
frequency point ([1] as a simple model how phase/amplitude affects oscillation
frequency). This is the "stability vs voltage" that you see in datasheets.
Now, if both the oven and the oscillator share the same package pin, then
there is a common impedance that both supplies share. I.e. if the oven
pulls more Amps, then there will be a voltage drop due to the impedance
of the power supply pin, which the oscillator will see as well. I.e.
you get a dependence of the oscillator frequency on the power the oven pulls.
Or in other words: Variantions in ambient temperature, will cause variations in
oven currents, which will cause variations in oscillator frequency. I.e. the
shared pin induces a slight temperature dependency of the oscillator.
Now, why don't all oscillators use seperate supply pins? Because in most
cases it's not worth the effort. Taking a run of the mill oscillator with
good temperature stability, like the Oscilloquartz 8663, we get a power
requirement of <2.5W steady state at 25°. At a 12V supply that's less
than 200mA supply current. Now, let us assume that the impedance is purely
resistive and 10mΩ, which is probably on the high side, that would mean
we would see a voltage variation of at most 2mV. The 8663 is specified
<3e-10 per 5%, or <5e-10/V (at 12V supply). So we get a worst case 1e-12
frequency dependency due to oven current. As the oven does never really
switch off completely, it's probably an order or two lower. Which means
it is insignificant compared to other influences that keep the 8663's
stability in the short term somewhere in the low 10^-12 range.
I hope, this answers your questions
Attila Kinali
[1] https://en.wikipedia.org/wiki/Barkhausen_stability_criterion
--
Science is made up of so many things that appear obvious
after they are explained. -- Pardot Kynes
Hi
The most basic answer is that ovens at start up pull a lot of current. Older
designs might pull 20W at turn on, numbers around 10 to 12W were more
typical.
Running that amount of power through the precision regulator(s) on an older
design was expensive and difficult. The answer was to run that side of things
off of the “bulk” side of the supply.
So: Simple answer was the cost of the gear the OCXO went into.
PHK earlier mentioned some other issues. To really be effective, those issues
needs to have a separate ground for the oven as well as power. You rarely
see those on OCXO’s (though the 10811 does have provision for one).
Even on OCXO designs that do have a separate oven ground it rarely gets
used properly. You can tear into a lot of gear and see them tying all the grounds
together …..
Why does the ground matter? Ground current changes the ground / reference
voltage at the OCXO. If one side of your EFC is tied to that point, that’s a problem.
How big a problem it is very much depends on what you are doing.
Now we’re off to a separate EFC return independent of the power ground. Sounds
fine in theory. When the customer looks at it the immediate question is “how do I deal
with that?”. Sit down and go over the options. Surprise …. we’re right back to tie
all the grounds together for cost reasons ….
Welcome to the real world :) :)
Bob
On Jan 18, 2025, at 12:35 PM, Matt Huszagh via time-nuts time-nuts@lists.febo.com wrote:
Many OCXOs permit the use of separate power supplies for the oven and
electronics. Why is this beneficial?
The oven requires more power (at least during warm-up) and I presume is
less sensitive to supply noise translating into phase noise at the output.
That would permit the use of a higher-power possibly noisier supply for the
oven and a cleaner, lower-power supply for the electronics. Is this
correct? Are there other power supply factors affecting the output
frequency? If so, is their effect through the oven or electronics supply,
or both? For example, the FTS 1000B datasheet gives a 1% supply sensitivity
of <1e-11 fractional frequency change, but doesn't specify to which supply
this applies. The datasheet also gives an "EMI susceptibility", which is
basically the sensitivity of the output phase noise to supply noise. They
don't indicate the relevant supply for this either, but I imagine it's the
electronics supply.
Thanks,
Matt
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
I hope I am not repeating another person’s answer on this.
All the responses to this question that I have seen are very true, but there is another more basic reason for the separate power supplies. In many high-end HP frequency counters and other frequency stabilized equipment back in the 80’s and earlier (using the 105, 10544, or 10811), The oven was always on whenever the unit was plugged into the mains regardless of whether or not the power switch was turned on. To keep frequency drift to a minimum the oven(s) and many times the entire oscillator was always powered up as long as the unit was plugged in. The stated frequency drift for the OCXO was with the oven operating at its designed operating temperature. When the OCXO oven powered down, the crystal could age at a different rate due to the “cut” of the crystal. . Many of these counters would use a section of the main supply for the OCXO. However, the rest of the main supply would be designed to energize from the power switch for the rest of the counter. As it has been pointed out, you would want to decouple the OCXO from any noise from the rest of the unit.
If the OXCO was in a much more complex piece of equipment such as the 5061 Cesium Beam Standard, the electronic frequency control could be very sensitive to other noise or return (ground) currents.
When I first started working on test equipment in the USAF, I was surprised that the 105 series oscillator had an outer oven which was controlled by a thermostatic switch as opposed to the inner oven’s sensitive voltage control (separate from the EFC). It appeared that the sharp on/off thermostatic switch noise with its large current did not affect the main oscillator or the short-term stability.
I hope this helps.
Jerry
Hi
….. then the energy crisis came along.
The rules for federal buildings changed and everything had to be turned off as folks left the work space. Somehow that didn’t apply to your Cesium standard. It very much did apply to “everything else”.
That all came in around 1980. Exactly when this or that agency / facility started following the rules likely varied a bit. I recall the guys at NIST and Ft. Monmouth having some “choice words” to say about these changes.
Once the government “showed the way”, a number of other folks followed suit. Gear with “alway on” ovens started to disappear from catalogs fairly soon after the change went in. The demand for “separate supply” OCXO’s was not significantly impacted by the change.
Bob
On Jan 19, 2025, at 1:44 PM, Jerry Chafee via time-nuts time-nuts@lists.febo.com wrote:
I hope I am not repeating another person’s answer on this.
All the responses to this question that I have seen are very true, but there is another more basic reason for the separate power supplies. In many high-end HP frequency counters and other frequency stabilized equipment back in the 80’s and earlier (using the 105, 10544, or 10811), The oven was always on whenever the unit was plugged into the mains regardless of whether or not the power switch was turned on. To keep frequency drift to a minimum the oven(s) and many times the entire oscillator was always powered up as long as the unit was plugged in. The stated frequency drift for the OCXO was with the oven operating at its designed operating temperature. When the OCXO oven powered down, the crystal could age at a different rate due to the “cut” of the crystal. . Many of these counters would use a section of the main supply for the OCXO. However, the rest of the main supply would be designed to energize from the power switch for the rest of the counter. As it has been pointed out, you would want to decouple the OCXO from any noise from the rest of the unit.
If the OXCO was in a much more complex piece of equipment such as the 5061 Cesium Beam Standard, the electronic frequency control could be very sensitive to other noise or return (ground) currents.
When I first started working on test equipment in the USAF, I was surprised that the 105 series oscillator had an outer oven which was controlled by a thermostatic switch as opposed to the inner oven’s sensitive voltage control (separate from the EFC). It appeared that the sharp on/off thermostatic switch noise with its large current did not affect the main oscillator or the short-term stability.
I hope this helps.
Jerry
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Hello,
This is a very interesting piece of information.
In the old OCXO I am reverse engineering and restoring, the output
signal is referenced to the positive supply.
Possibly as a noise mitigating measure, then.
Sebastien
On 19/01/2025 15:20, Bob Camp via time-nuts wrote:
Hi
The most basic answer is that ovens at start up pull a lot of current. Older
designs might pull 20W at turn on, numbers around 10 to 12W were more
typical.
Running that amount of power through the precision regulator(s) on an older
design was expensive and difficult. The answer was to run that side of things
off of the “bulk” side of the supply.
So: Simple answer was the cost of the gear the OCXO went into.
PHK earlier mentioned some other issues. To really be effective, those issues
needs to have a separate ground for the oven as well as power. You rarely
see those on OCXO’s (though the 10811 does have provision for one).
Even on OCXO designs that do have a separate oven ground it rarely gets
used properly. You can tear into a lot of gear and see them tying all the grounds
together …..
Why does the ground matter? Ground current changes the ground / reference
voltage at the OCXO. If one side of your EFC is tied to that point, that’s a problem.
How big a problem it is very much depends on what you are doing.
Now we’re off to a separate EFC return independent of the power ground. Sounds
fine in theory. When the customer looks at it the immediate question is “how do I deal
with that?”. Sit down and go over the options. Surprise …. we’re right back to tie
all the grounds together for cost reasons ….
Welcome to the real world :) :)
Bob
On Jan 18, 2025, at 12:35 PM, Matt Huszagh via time-nuts time-nuts@lists.febo.com wrote:
Many OCXOs permit the use of separate power supplies for the oven and
electronics. Why is this beneficial?The oven requires more power (at least during warm-up) and I presume is
less sensitive to supply noise translating into phase noise at the output.
That would permit the use of a higher-power possibly noisier supply for the
oven and a cleaner, lower-power supply for the electronics. Is this
correct? Are there other power supply factors affecting the output
frequency? If so, is their effect through the oven or electronics supply,
or both? For example, the FTS 1000B datasheet gives a 1% supply sensitivity
of <1e-11 fractional frequency change, but doesn't specify to which supply
this applies. The datasheet also gives an "EMI susceptibility", which is
basically the sensitivity of the output phase noise to supply noise. They
don't indicate the relevant supply for this either, but I imagine it's the
electronics supply.Thanks,
Matt
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Bob,
I agree with everything you say, but I was in the USAF working in PMEL at McChord AFB in Washington State and at Lowry AFB, Colorado as a PMEL instructor during the period of the 1980's. At McChord, as far as anything that was high precision (our main standards and very stable frequency sources), we were allowed to keep the items both plugged in and energized. They were exempt from the Federal Building energy rules. As a matter of fact, our PMEL lab had a very expensive heating/cooling system that kept the isolated lab at somewhere around 68 (+/- around 3) degrees 24/7. I remember because we could not verify calibration of other equipment if the lab was outside the temperature limits. I remember one hot summer when the heating portion of the system failed and kept the cooling section running at 100% for 2 days. It got so cold that it was more comfortable outside the lab. It took days to get the equipment where we could calibrate anything more accurate than a Simpson 260 multimeter.
As far as the precision frequency equipment, at McChord AFB we had to keep the items powered for quite some time, partially because our only way to calibrate very stable frequency standards (not good enough for a cesium or rubidium standard), was to use a Fluke 207 VLF Frequency Comparator Receiver tracking WWVB, and check the drift rate at the beginning and end of a 10 - 30 day period. Our lab did not have anything better than the stability of a 105 oscillator to test or work with. I happened to be the only person that understood the operation of the HP 5360A Computing Counter which typically was only used at SAC bases. When repair was needed, It was quite the troubleshooting nightmare. We did support and calibrate items for use at Hanford, WA, and Puget Sound Naval Shipyard. I guess they needed to use them with their Cesium beam standards and their Austron 2000 Loran receivers for precise time and frequency settings. I am not sure why they needed such accuracy, but nobody else seemed to understand the operation of the counters. HP wanted to keep away from the 5360A due to its unusual complexity (and it was getting old). It actually contained a built in calculator capability using standard 7400 series ICs. But I digress.
When stationed at Lowry, we did have to turn off all equipment when the school closed for the night or weekend. The only exceptions were in the Navy training section where they had 5062C Cesium frequency standards. These were allowed to be powered all of the time. I understand the 5062C was fraught with problems probably because of trying to make this a miniaturized tube design.
Thanks, Jerry