Symmetricom/Datum FTS-1050A Disciplined Frequency Standard
On Wed, 06 Apr 2022 03:30:35 -0400, time-nuts-request@lists.febo.com
wrote:
time-nuts Digest, Vol 216, Issue 10
Message: 4
Date: Sun, 3 Apr 2022 09:53:18 -0400
From: Bob kb8tq kb8tq@n1k.org
Subject: [time-nuts] Re: Low Phase Noise 10 MHz bench signal source
sought
To: ew ewkehren@aol.com, Discussion of precise time and frequency
measurement time-nuts@lists.febo.com
Message-ID: 11376923-062A-4011-A6D4-1D9CE3361466@n1k.org
Content-Type: text/plain; charset=utf-8
Hi
These days a PLL is going to either be analog or digital. If it’s
analog, you get into size constraints related to capacitors
as you go to lower crossover frequencies. With digital, you
get into all of the noise issues that any digital circuit will have.
(Yes, they can be addressed but it’s not easy at very low
offset frequencies).
All of the loop filters I've seen recently had nominal bandwidths in
the Hertz
to tens of Hertz, usually implemented in some kind of digital signal
processor.
10 Hz or higher is certainly do-able with analog loop components.
There are a lot of products out there that work that way.
About 30 years ago, there was a legacy 5 MHz disciplined
oscillator that could be set to a 100-second response time. I never
did find any real technical data or patents on it. I don't recall
its name, but it may come back to me. I think it was made by
Symmetricom.
I finally recalled the details, after all these years. It was from
Symmetricom, they having acquired Datum in 2002. It was model
FTS-1050A Disciplined Frequency Standard. Despite the implication of
the product name, it does appear to be a phase-lock loop design at
heart, from the users manual (my copy being dated 1999). This is the
one that I suspect was in fact a 3rd-order PLL design, because it
would become unstable if the the incoming reference were too faint,
being far more fussy than your usual PLL, which would happily lock
onto a pretty faint and ratty reference signal.
It has two switch-selectable integration periods, one second and one
hundred seconds. I assume that the integration is digital, but in
hardware versus a computer.
I can provide the documentation, if anybody wants a copy. Apparently
a number of folk were looking here, over the years. Maybe something
to add to Febo.com.
I wonder who the designers were. Hmm. I bet that Robert Lutwak,
William Riley, and Kenneth Lyon were involved, as these folk are the
inventors of patents assigned to DATUM TIMING TEST AND MEASUREMENT
Inc and Datum Inc in the day. I worked with Ken Lyon some time ago,
if I have the right Ken Lyon.
Joe Gwinn
Joe,
Yes, going back many years on time-nuts, two desirable military grade
vintage portable quartz frequency standards were AN/URQ-10 and
AN/URQ-23. The latter contained a FE-1050A oscillator which could be
disciplined by an external reference. The manual goes into great detail.
[1] See especially pages 2-10, 2-14, 3-2, 3-5, and 5-12.
Right, there is the front panel switch for short (1 s) / long (100 s)
time constant. In this instrument the integration is analog, not
digital. The text says it's a 1200 MΩ resistor; although the schematic
shows 2500 MΩ. Note also the use of a "memory circuit" to maintain
frequency when the reference input is removed. The manual is wonderful
old school.
Corby's photos match what's in the PDF. Let it us know if this is the
same instrument that you remember. I have a URQ 10 and 23 if you have
more questions. Let us know if your Symmetricom / FTS / Datum 1050A
looks like a clone of the FE-1050A.
/tvb
[1] ko4bb.com and search manuals for 1050A or URQ23
30,516,123 //
FrequencyElectronics_ANURQ-23_Frequency_Time_Standard_Service_Manual.pdf
On 4/18/2022 3:18 PM, Joseph Gwinn wrote:
It has two switch-selectable integration periods, one second and one
hundred seconds. I assume that the integration is digital, but in
hardware versus a computer.
I can provide the documentation, if anybody wants a copy. Apparently
a number of folk were looking here, over the years. Maybe something
to add to Febo.com.
That's an interesting old machine - very cool.
One thing though, is that unless I'm missing something, I believe the
two available loop time constants are in minutes, not seconds, or that
it should be in many more (maybe 100X) seconds, if stated that way.
Since the unit can synchronize to a 1 PPS reference, it would make sense
that the loop filtering goes way beyond 1 or 100 seconds.
If this is the case, then there's some typo errors in the manual.
As far as I know, time constant is still T=RC, or megohms X uF =
seconds, in the convenient short form I always remember. So, the long
time constant setting of 2500 megs by 10 uF gives 25,000 seconds - over
400 minutes. Now, I can picture it being defined also by the scaling of
the tuning range used. If you take the input divider 100 k/ 10 k, times
the amplifier gain a little less than 2, that gets it overall into the
100 minutes ballpark. The filter is not an integrator in the pure sense,
but an RC LPF, so the output is bounded to about 20% of the "stored"
tuning voltage from the DAC system.
Regardless of how you estimate, it seems like the times have to be in
minutes, not seconds.
Ed
It was model FTS-1050A Disciplined Frequency Standard.
The FTS-1050A was the second product of Frequency and Time Systems Inc
(FTS) and appeared in the market around 1980. The instrument architecture
was the product of Martin Levine (of Levine and Vessot Gravity Probe A) as
implemented by Jerry Welch. The 1050A employs an analog PLL. The heart of
the 1050A instrument was a 1000A quartz oscillator designed by Donald
Emmons.
Most of the FTS, Datum, Symmetricom products relied upon trade secrets for
protection of intellectual property which is why you'll find few patents or
detailed technical manuals.
The Datum 2110B, developed first at Austron (Austin, TX) was a similar (to
the 1050A) instrument which used a digital FLL. The 2110C (based upon the
2110B and developed in Beverly, MA), was a more sophisticated (though not as
low noise) version with a dual input FLL that would discipline to the
average of two reference inputs and, upon loss or severe degradation of one
input, would switch to the use of a single reference input. The primary
application was for robust, redundantly referenced timing sources for
telecom Central Office instruments. The design was by Peter Vlitas.
I was a scientist at FTS then CTO, retiring in 2011.
Mike Garvey
-----Original Message-----
From: Joseph Gwinn joegwinn@comcast.net
Sent: Monday, April 18, 2022 6:18 PM
To: time-nuts@lists.febo.com
Subject: [time-nuts] Symmetricom/Datum FTS-1050A Disciplined Frequency
Standard
On Wed, 06 Apr 2022 03:30:35 -0400, time-nuts-request@lists.febo.com
wrote:
time-nuts Digest, Vol 216, Issue 10
Message: 4
Date: Sun, 3 Apr 2022 09:53:18 -0400
From: Bob kb8tq kb8tq@n1k.org
Subject: [time-nuts] Re: Low Phase Noise 10 MHz bench signal source
sought
To: ew ewkehren@aol.com, Discussion of precise time and frequency
measurement time-nuts@lists.febo.com
Message-ID: 11376923-062A-4011-A6D4-1D9CE3361466@n1k.org
Content-Type: text/plain; charset=utf-8
Hi
These days a PLL is going to either be analog or digital. If it’s
analog, you get into size constraints related to capacitors as you
go to lower crossover frequencies. With digital, you get into all of
the noise issues that any digital circuit will have.
(Yes, they can be addressed but it’s not easy at very low offset
frequencies).
All of the loop filters I've seen recently had nominal bandwidths in
the Hertz to tens of Hertz, usually implemented in some kind of
digital signal processor.
10 Hz or higher is certainly do-able with analog loop components.
There are a lot of products out there that work that way.
About 30 years ago, there was a legacy 5 MHz disciplined oscillator
that could be set to a 100-second response time. I never did find
any real technical data or patents on it. I don't recall its name,
but it may come back to me. I think it was made by Symmetricom.
I finally recalled the details, after all these years. It was from
Symmetricom, they having acquired Datum in 2002. It was model FTS-1050A
Disciplined Frequency Standard. Despite the implication of the product
name, it does appear to be a phase-lock loop design at heart, from the users
manual (my copy being dated 1999). This is the one that I suspect was in
fact a 3rd-order PLL design, because it would become unstable if the the
incoming reference were too faint, being far more fussy than your usual PLL,
which would happily lock onto a pretty faint and ratty reference signal.
It has two switch-selectable integration periods, one second and one
hundred seconds. I assume that the integration is digital, but in hardware
versus a computer.
I can provide the documentation, if anybody wants a copy. Apparently a
number of folk were looking here, over the years. Maybe something to add to
Febo.com.
I wonder who the designers were. Hmm. I bet that Robert Lutwak, William
Riley, and Kenneth Lyon were involved, as these folk are the inventors of
patents assigned to DATUM TIMING TEST AND MEASUREMENT Inc and Datum Inc in
the day. I worked with Ken Lyon some time ago, if I have the right Ken
Lyon.
Joe Gwinn
time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send
an email to time-nuts-leave@lists.febo.com To unsubscribe, go to and follow
the instructions there.
I'm wondering if anyone has dissected enough common canned RF mixers, to
know how symmetric they are, internal construction-wise, or knows of
available info, especially on specific models.
I have taken some apart over the years, and I believe they generally are
made highly symmetric wrt the LO and RF ports. I typically use them
either way around, for units that have the same band specs on both. But,
the specs typically are with the given port assignments, so there may be
some questions, depending on application.
The particular models in this situation are the WJ M9D, and MCL SRA-1H,
which are high level +20 and +17 dBm, respectively. I have only one M9D,
and two SRA-1H - they're all I have in this class, with a favorable
pinout that I need. I want the IF port to be DC-isolated from ground
(but RF-shorted with C between commons).
Both units appear to be OK for the pinning I need, but with slightly
different arrangement. Having the option of L-R end and phase swapping
(along with IF pinning for best shielding), gives more hookup flexibility.
Since I only have three good candidate mixers, I need to be very careful
to not burn any out, as I'll be driving from an amp capable of over +30
dBm* (with lots of padding and maybe a limiter). Also, the signal input
will be fairly big, up to +6 dBm average, and possibly +20 dBm peak.
This is for that noise source down-converter project I mentioned before.
I'm trying to go as big as possible on signal levels, both to maximize
the output power after conversion and filter loss, and preserve fairly
high crest factor.
The above conditions are about the maximum - in reality, by the time all
the signals are properly padded, the levels will be more realistic. I'm
trying to minimize the padding of course, even looking at using a
diplexer at the IF to absorb the upper image power, to avoid padding the
reflection off the LPF.
*That's at normal 24 V supply. I'm going to try running at 15 V,
unspecified. The maximum Po should be greatly reduced at the lower supply.
Ed
The noise converter project based on the Scientific Atlanta 4647 is
moving along nicely. Still no luck in finding any more info about this
unit, but I did quite a bit of digging in the guts, and figured out
enough to make progress.
The noise generator, to my surprise, is not a typical noise-diode-based
type, but an all-amplifier deal, and apparently the fundamental noise
source is a 75 ohm resistor in conjunction with the input noise of a
2N5179 amplifier front end. The first few stages are broadband, followed
by maybe eight bandpass stages, to craft the power level and shape,
resulting in the 50-90 MHz noise signal, which gets passed to the noise
amplifier box.
The noise amplifier is broadband again, then feeding a CATV type hybrid
power amp for final output, which goes through a ferrite part, which is
either a splitter or directional coupler, for leveling, then on to a
decade step attenuator using Teledyne TO-5 style relays. The leveling
signal from the local detector is sent back to the noise generator box
where it somehow does the gain control. Altogether, a couple dozen or so
transistors are used in the gain stages.
The step attenuator output is sent to the last box, the "C+N amplifier,"
where the external carrier input is attenuated with a step attenuator,
then amplified up and leveled in similar fashion (including another CATV
hybrid PA), then through its own step attenuator, and added to the noise
through a reactive power combiner. So, the noise and carrier signals are
each at least 3 dB bigger than the spec output levels, to accommodate
the adding process.
I added a small board into the noise amp module, with an RF relay to
pass the signal as normal, or route it to the new converter. The maximum
PSD of the noise available there is about -70 dBm/Hz, versus the -73
dBm/Hz at the normal C+N output.
The rest of the action is all built into the 70 MHz oscillator/agc amp
module now. I sacrificed the agc amp function, and utilized the space
for the mixer and LPF, and added yet another CATV type PA in the
oscillator section, for the LO. More on this in the next installment.
Ed
Am 2022-05-11 7:56, schrieb ed breya:
The noise converter project based on the Scientific Atlanta 4647 is
moving along nicely. Still no luck in finding any more info about this
unit, but I did quite a bit of digging in the guts, and figured out
enough to make progress.
The noise generator, to my surprise, is not a typical
noise-diode-based type, but an all-amplifier deal, and apparently the
fundamental noise source is a 75 ohm resistor in conjunction with the
input noise of a 2N5179 amplifier front end. The first few stages are
broadband, followed by maybe eight bandpass stages, to craft the power
level and shape, resulting in the 50-90 MHz noise signal, which gets
passed to the noise amplifier box.
I have once made an informal "transfer normal" between job and home.
That was just a 500?? Ohm resistor for thermal noise and two
LMH6702 amplifiers for the positions HI and LO. I used MCL PSC2-1
power dividers to add the noise to an oscillator. In my case, that
was the FEI405 once distributed here. The spurious is the FEI.
The noise source was completely flat in the bandwidth of the LMH6702s.
Beware of compression effects in the amplifier. They change noise
statistics.
regards,
Gerhard DK4XP
Do you need a 4647?
Lester B Veenstra K1YCM MØYCM W8YCM 6Y6Y
lester@veenstras.com
452 Stable Ln (HC84 RFD USPS Mail)
Keyser WV 26726
GPS: 39.336826 N 78.982287 W (Google)
GPS: 39.33682 N 78.9823741 W (GPSDO)
Telephones:
Home: +1-304-289-6057
US cell +1-304-790-9192
Jamaica cell: +1-876-456-8898
-----Original Message-----
From: ed breya [mailto:eb@telight.com]
Sent: Wednesday, May 11, 2022 1:57 AM
To: time-nuts@lists.febo.com
Subject: [time-nuts] Noise down-converter project
The noise converter project based on the Scientific Atlanta 4647 is
moving along nicely. Still no luck in finding any more info about this
unit, but I did quite a bit of digging in the guts, and figured out
enough to make progress.
The noise generator, to my surprise, is not a typical noise-diode-based
type, but an all-amplifier deal, and apparently the fundamental noise
source is a 75 ohm resistor in conjunction with the input noise of a
2N5179 amplifier front end. The first few stages are broadband, followed
by maybe eight bandpass stages, to craft the power level and shape,
resulting in the 50-90 MHz noise signal, which gets passed to the noise
amplifier box.
The noise amplifier is broadband again, then feeding a CATV type hybrid
power amp for final output, which goes through a ferrite part, which is
either a splitter or directional coupler, for leveling, then on to a
decade step attenuator using Teledyne TO-5 style relays. The leveling
signal from the local detector is sent back to the noise generator box
where it somehow does the gain control. Altogether, a couple dozen or so
transistors are used in the gain stages.
The step attenuator output is sent to the last box, the "C+N amplifier,"
where the external carrier input is attenuated with a step attenuator,
then amplified up and leveled in similar fashion (including another CATV
hybrid PA), then through its own step attenuator, and added to the noise
through a reactive power combiner. So, the noise and carrier signals are
each at least 3 dB bigger than the spec output levels, to accommodate
the adding process.
I added a small board into the noise amp module, with an RF relay to
pass the signal as normal, or route it to the new converter. The maximum
PSD of the noise available there is about -70 dBm/Hz, versus the -73
dBm/Hz at the normal C+N output.
The rest of the action is all built into the 70 MHz oscillator/agc amp
module now. I sacrificed the agc amp function, and utilized the space
for the mixer and LPF, and added yet another CATV type PA in the
oscillator section, for the LO. More on this in the next installment.
Ed
time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an
email to time-nuts-leave@lists.febo.com
To unsubscribe, go to and follow the instructions there.
Continuing on, the 70 MHz for the LO is tapped off at a leveled low
impedance point, that feeds the normal 70 MHz 0 dBm output on the front
panel. The tap off point is probably around +3 dBm, and I added a higher
R attenuator to get about -10 dBm for the power amp. This CATV amp is
made for 24-30 V operation, but works OK on 15 V, with much less output
power available, and high distortion (obvious on a scope), but still
plenty of gain (35 dB). The output runs about 25 dBm, while the
saturated output power limit is about 28 dBm, which are just about right
for good drive level, but not too much fault power, to avoid mixer
damage if anything goes wrong. The output is already way into
compression, but that's OK. A 6 dB pad connects it to the mixer,
providing nominal drive around 19 dBm, or 22 dBm fault, which is the
mixer's maximum power rating.
That all is what was planned, but what actually shows is that the mixer
looks like a lower Z, well below 50 ohms. I set up the drive with a
built in monitor port that provides a -26 dB view, that showed about
right with a 50 ohm load in place of the mixer, but much lower with the
mixer - it looks like about 15 dBm. It seems to run fine, but is a
little odd. I don't want to push it too hard without more study, so it
is what is is for now.
The maximum noise power comes in at around -70 dBm/Hz from 75 ohms, and
it turns out that a min-loss 75-50 ohm broadband pad is just about right
to knock off 6 dB, putting the R input total power level around +1 dBm,
and peak up to +16 dBm due to crest factor. This is totally safe for the
mixer, and provides good power output. The crest factor will be degraded
somewhat due to running into the LO limit, but only at the highest power
settings. It should be preserved well at lower power.
The chosen mixer is the WJ M9D, which I've discussed previously. Since
this setup is a DSB down-conversion, the conversion loss is less (about
twice as good) than for SSB. I estimate it at around 4 dB, which seems
to agree with my measurements so far. Interestingly, the 50-90 MHz noise
power is not like a typical up-converted baseband signal. Each
"sideband" around the 70 MHz is not redundant to other - they are
independent and uncorrelated (I would think) noise, and simply add
together.
So anyway, ignoring the losses, half of the incident noise power is
converted to the 0 to about 25 MHz range, and the other half goes mostly
to the upper image centered at 140 MHz, and the higher order products.
The IF spectrum viewed on the SA is interesting. The DC-25 MHz portion
is the biggest, and dead-flat in the scale of things. The upper image
looks about 3 dB less, to account for all the rest of the power
contained in the higher products - they are quite large, and go out
quite a way.
That's all for now. Next up will be more mixer and filter stuff.
Ed
Continuing on, the mixer's output looks amazingly good. The filter's,
not so much. I have the IF now going directly to the SA input - no pads,
no filters, no nothing, except some SMB cable/adapter stuff, and about
20 feet of BNC cable. It looks great, letting the SA do the filtering.
The low end is a beautiful down-converted replica of the 50-90 MHz noise
signal.
I can't make high precision measurements here - most are eyeball
estimates from the SA screen, but everything is in the right ballpark,
and makes sense. The amplitude measurements depend on the SA's IF RBW
setting, which is 3 MHz maximum. The measured levels agree well with
different RBW settings. The video BW also affects it some, since extra
filtering is needed sometimes to smooth the curves.
The spec of the 4647 says the effective noise BW is 48.2 MHz. The IF
passes through the -3 dB point near 24 MHz, in close agreement. The
level is very flat (no discernible deviation), to around 20 MHz, where
it just visibly starts to curve into the band edge. The maximum PSD
appears to be around -80 to -83 dBm/Hz, estimated from the displayed
levels at different RBWs.
So, the desired signal is wonderful, if only it didn't include
everything else above. What I need is a very good LPF to get the job
done - the usual problem.
The actual filter I've been using does a good job on the higher
frequencies, but is poor on flatness. It has about 2-3 dB p-p passband
ripple, with periodicity around 5-7 MHz. I've tried various padding
arrangements at both ends, all of which tend to flatten it only a little
bit at best. Looking at it with the TG/SA setup, the character is
intrinsic to filter, and not due to just its reaction to the mixer and
cabling and such.
I hate building filters. Designing them in principle is easy, with all
sorts of available tools online, but actually rounding up the real parts
(and their parasitics) and physical implementation is a PITA. But, I
suppose I'll have to do it eventually for this project. I know how nice
it can be, with the right filter, but for now, I'll have to go with what
I have.
This particular filter is a packaged module type that I've had for a
long time, and used in many experimental setups. In fact, I had to
borrow it from its commitment to another project. Despite its
limitations, it can be very handy, and it is very simple inside, so I'd
like to replicate it for other uses. I plan to open a thread about this
as a separate issue.
In the mean time, it will be for this noise project, and I'll have some
more to report, so next up will be the low frequency/DC aspects.
Ed