frequency reference for portable operation
Hi
With most modern lightweight Rb's the OCXO is integrated into the same heater block as the physics package. That makes it a bit tough to heat one without heating the other. …
Bob
On Mar 11, 2013, at 7:03 AM, Magnus Danielson magnus@rubidium.dyndns.org wrote:
On 10/03/13 16:17, Bob Camp wrote:
Hi
Remember - OCXO's are going to be acceleration sensitive. As you bump about on back roads, the oscillator is likely moving around by a few ppb. If you are after a hertz at 10 GHz, that's a lot.
Your GPS will be off by a fairly predictable amount based on it's idea of it's location. If you have a 3M position accuracy, then you will likely have 10 ns of time error. With a 10 second loop, you are right back to a ppb.
A good Rb should be able to hit a sub ppb accuracy at 10 minutes.
If you keep the OCXO on the Rb heated, you should get quicker on the mark. Then it is mostly the heatup of the physical package which consumes power and time then.
Then again, pre-heating with the car cuts down warm-up times.
Cheers,
Magnus
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
On 11/03/13 12:42, Bob Camp wrote:
Hi
With most modern lightweight Rb's the OCXO is integrated into the same heater block as the physics package. That makes it a bit tough to heat one without heating the other. …
On LPROs the OCXO sits on the opposite side of the board.
Cheers,
Magnus
Hi
So all you have to do is figure out when the physics package is hot enough
for the line spreading to work well enough to get a lock...
Bob
-----Original Message-----
From: time-nuts-bounces@febo.com [mailto:time-nuts-bounces@febo.com] On
Behalf Of Magnus Danielson
Sent: Monday, March 11, 2013 7:52 AM
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] frequency reference for portable operation
On 11/03/13 12:42, Bob Camp wrote:
Hi
With most modern lightweight Rb's the OCXO is integrated into the same
heater block as the physics package. That makes it a bit tough to heat one
without heating the other. .
On LPROs the OCXO sits on the opposite side of the board.
Cheers,
Magnus
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
With most modern lightweight Rb's the OCXO is integrated into the same
heater block as the physics package. That makes it a bit tough to heat one
without heating the other. .
On LPROs the OCXO sits on the opposite side of the board.
Pretty sure that's not an OCXO. If it is, it's a tiny one that will warm up
very quickly. I doubt it's even a TCXO, though.
Of the Rb standards I've seen, only the PRS-10 and HP 5065A use OCXOs. I'd
assume that the old-school Tracor and R&S models also used OCXOs, but for
the cheap telecom modules it would have been seen as a waste of money.
Also, the HP 5065A has separate ovens for the lamp and resonance cell, but I
believe the smaller units use a single heater for both. They won't lock
until the heater is up to temperature.
So, I have bought a lot of 10 MHz OCXOs from eBay over the last few
years. The best phase noise baseline reference I have found so far is my
Z3805. I have lots of OCXOs in the 2x2x1.5 inch size. Many had good
specs pointed at by the listings or word of mouth. When I used them with
my board and SA most were pretty crappy compared to the Z3805. A couple
of the ones I bought were Morion MV89As. - supposedly good, but what I
saw didn't look very great. One of the best ones I have is a small
2x2x.75 inch Wenzel I bought a few years back. It has a custom part
number of 500-11935. But don't buy by name. I recently picked up a 1x1x3
Wenzel 10 MHz with sma output connector and its phase noise looks pretty
horrible.
The MV-89A DOCXOs are good below 10 Hz but they have a high white PN floor
beyond 1 kHz, at about -150 dBc/Hz, with AM noise at a similar level. They
aren't actually 10 MHz OCXOs, but doubled 5 MHz units, so there's a lot of
extra circuitry to add noise between the oscillator stage and the output
jack. On the upside, this makes them relatively insensitive to load pulling
and injection locking.
Bob is right in that any scheme that involves multiplying 10 MHz directly to
X band is going to result in a relatively noisy output carrier, of course.
In the 8566B they multiply 10 MHz to 100 MHz, then to somewhere around 400
MHz, and only then does the sampler loop multiply that to reach 2-6 GHz.
-- john, KE5FX
Miles Design LLC
Hi
A few numbers on the 10 (or 5) MHz to 100 MHz stuff:
The most likely frequency range is 100 Hz to 100 KHz. That covers CW operation/filters as well as reasonable adjacent channel issues on SSB.
At 10 KHz and beyond you likely will be at floor on your 100 MHz oscillator. With a good design that will be in the -165 to -175 dbc/ Hz range. The offset between 10 MHz and 100 MHz would be 20 db in this range. That would require a -185 to -195 dbc / Hz 10 MHz OCXO to be "as good as". The normal DOCXO's run in the -160 range. The 100 MHz likely wins by > 20 db.
At 100 Hz, you are doing pretty well to hit -125 dbc / Hz. Some luck will get you to -135 dbc / Hz. At 100 Hz, you are likely to get some noise from your multiplier. A 23 db offset is probably a good guess here. That puts your 10 MHz OCXO in the -148 to -158 dbc / Hz range. The TBolt OCXO will hit these sort of numbers. Most DOCXO's run out around -145. At best you are same / same with the multiplier. Most likely you are 10 db better with the 100 MHz.
Between 100Hz and 10 KHz it's going to be a "that depends" sort of thing. The 100 MHz will always win. It's a question of ho much it'll win by.
The simple approach is to set up a PLL with a 3db (not Wn) loop bandwidth in the 10 to 20 Hz range. That should keep the 10 MHz well below the 100 MHz by the time you get to a 100 Hz offset.
Next step is to run up from 100 to something and then from there to 10+ GHz. If you get to the next step by multiplying, then the -175 dbc phase noise might be useful. If you are running into some sort of phase detector or most SRD's then you may not get the full benefit of noise that far down. Without knowing what the next step is going to be, you really can't fully spec out the 100 MHz oscillator.
Bob
On Mar 11, 2013, at 4:20 PM, John Miles jmiles@pop.net wrote:
With most modern lightweight Rb's the OCXO is integrated into the same
heater block as the physics package. That makes it a bit tough to heat one
without heating the other. .
On LPROs the OCXO sits on the opposite side of the board.
Pretty sure that's not an OCXO. If it is, it's a tiny one that will warm up
very quickly. I doubt it's even a TCXO, though.
Of the Rb standards I've seen, only the PRS-10 and HP 5065A use OCXOs. I'd
assume that the old-school Tracor and R&S models also used OCXOs, but for
the cheap telecom modules it would have been seen as a waste of money.
Also, the HP 5065A has separate ovens for the lamp and resonance cell, but I
believe the smaller units use a single heater for both. They won't lock
until the heater is up to temperature.
So, I have bought a lot of 10 MHz OCXOs from eBay over the last few
years. The best phase noise baseline reference I have found so far is my
Z3805. I have lots of OCXOs in the 2x2x1.5 inch size. Many had good
specs pointed at by the listings or word of mouth. When I used them with
my board and SA most were pretty crappy compared to the Z3805. A couple
of the ones I bought were Morion MV89As. - supposedly good, but what I
saw didn't look very great. One of the best ones I have is a small
2x2x.75 inch Wenzel I bought a few years back. It has a custom part
number of 500-11935. But don't buy by name. I recently picked up a 1x1x3
Wenzel 10 MHz with sma output connector and its phase noise looks pretty
horrible.
The MV-89A DOCXOs are good below 10 Hz but they have a high white PN floor
beyond 1 kHz, at about -150 dBc/Hz, with AM noise at a similar level. They
aren't actually 10 MHz OCXOs, but doubled 5 MHz units, so there's a lot of
extra circuitry to add noise between the oscillator stage and the output
jack. On the upside, this makes them relatively insensitive to load pulling
and injection locking.
Bob is right in that any scheme that involves multiplying 10 MHz directly to
X band is going to result in a relatively noisy output carrier, of course.
In the 8566B they multiply 10 MHz to 100 MHz, then to somewhere around 400
MHz, and only then does the sampler loop multiply that to reach 2-6 GHz.
-- john, KE5FX
Miles Design LLC
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
http://www.evrytania.com/lte-tools/lte-cell-scanner
claims the LTE reference is good to 0.05PPM.
I can verify that 739MHz and 751MHZ are the only LTE frequencies I found
in use in the 700MHz band in the San Francisco Bay Area.
If you want to use the software, I suggest trying 739MHz and 751MHZ
first. Get the calibration factor, then start the search with a reduced
"p" parameter. The initial search assumes a ppm error in the SDR of
120PPM. However, once you get the calibration factor, you can used the
reduced PPM parameter which substantially improves the search speed.
I haven't found a program to actually use the LTE frequency reference
for say NTP. Maybe satcom Dave has time on his hands. ;-)
Hi
CDMA should be good to ~0.0001 ppm when running normally. The driver there is holdover rather than transmit accuracy. The carrier comes from the same source, so the accuracy is "free". CDMA is all over the place….
A cheap GPS receiver sounds like a much better solution for NTP. Far more trustworthy.
Bob
On Mar 11, 2013, at 7:54 PM, gary lists@lazygranch.com wrote:
http://www.evrytania.com/lte-tools/lte-cell-scanner
claims the LTE reference is good to 0.05PPM.
I can verify that 739MHz and 751MHZ are the only LTE frequencies I found in use in the 700MHz band in the San Francisco Bay Area.
If you want to use the software, I suggest trying 739MHz and 751MHZ first. Get the calibration factor, then start the search with a reduced "p" parameter. The initial search assumes a ppm error in the SDR of 120PPM. However, once you get the calibration factor, you can used the reduced PPM parameter which substantially improves the search speed.
I haven't found a program to actually use the LTE frequency reference for say NTP. Maybe satcom Dave has time on his hands. ;-)
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
On 11/03/13 13:29, Bob Camp wrote:
Hi
So all you have to do is figure out when the physics package is hot enough
for the line spreading to work well enough to get a lock...
Should not be too hard to monitor. Isn't oven temperature part of the
BITE fail condition?
Cheers,
Magnus
On 11/03/13 21:20, John Miles wrote:
With most modern lightweight Rb's the OCXO is integrated into the same
heater block as the physics package. That makes it a bit tough to heat one
without heating the other. .
On LPROs the OCXO sits on the opposite side of the board.
Pretty sure that's not an OCXO. If it is, it's a tiny one that will warm up
very quickly. I doubt it's even a TCXO, though.
Considering that the sweeping ranges +/- 15 ppm, it needs to be pretty
good at least. TCXO is doable and the OCXO does not need to be very
advanced. There is a heat-sink on the crystal assembly. It would make
sense if it is an OCXO as the surrounding would be on the hot side while
operating, so increase heat dissipation in order to keep it in balance
would make sense.
The fact that we don't know the details is a side-effect of not
reverse-engineering them properly.
Of the Rb standards I've seen, only the PRS-10 and HP 5065A use OCXOs. I'd
assume that the old-school Tracor and R&S models also used OCXOs, but for
the cheap telecom modules it would have been seen as a waste of money.
The R&S XSRM has an OCXO, so does my russian one.
Also, the HP 5065A has separate ovens for the lamp and resonance cell, but I
believe the smaller units use a single heater for both. They won't lock
until the heater is up to temperature.
You need to keep two different temperatures of the lamp (~70) and cell
(~110), unless you tweaked it with buffer gas, which might be what they
have done.
Cheers,
Magnus
Hi
I'm not sure that the cheap ones really report back the temperature as part of the BITE. I suspect they report something more along the lines of "oven is not running all the time".
Bob
On Mar 12, 2013, at 6:41 AM, Magnus Danielson magnus@rubidium.dyndns.org wrote:
On 11/03/13 13:29, Bob Camp wrote:
Hi
So all you have to do is figure out when the physics package is hot enough
for the line spreading to work well enough to get a lock...
Should not be too hard to monitor. Isn't oven temperature part of the BITE fail condition?
Cheers,
Magnus
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.