About HP10544A
Since retrofitting an improved oscillator circuit isn't really an option
you will need to filter the output to reduce the harmonic content.
Try a bandpass filter driven by the buffer and terminated in 50 ohms.
Well I have to decide what to do, get another oscillator or try the
filtering. The difference to another oscillators (like tbolt ocxo or
LPRO) is so huge that I do not know how hard it would filter the 10544A
to the same level, which parts to use and how much it will cost.
Should you retrofit an improved oscillator circuit you may as well
replace the oven controller to eliminate the oven switching frequency
related sidebands.
Infact the switcher sidebands are now gone:
http://www.amigazone.fi/files/gpsdo/544-13.png
http://www.amigazone.fi/files/gpsdo/544-14.png
It was like thios earlier:
http://www.amigazone.fi/files/gpsdo/544-1.PNG
http://www.amigazone.fi/files/gpsdo/544-2.PNG
It was easier than I expected... When I took the output directly from
OSC pins without using the coax connector like you suggested the
switcher peaks was gone!
So I had a closer look to the PCB today and noticed that it has only 2
layers and the ground net is too thin and goes around PCB totally wrong
way so that switcher current seems to flow via the signal flow. It's PCB
layout design fault, I think that with correct layout design this could
be done even with 2-layer PCB correctly.
But this is not a problem of course because I will design my own PCB for
the final system anyway, having SMA connectors for 10 MHz and EFC. I'll
also create separate power supply for ovens and signalling stages.
--
73s!
Esa
OH4KJU
Esa Heikkinen wrote:
Since retrofitting an improved oscillator circuit isn't really an option
you will need to filter the output to reduce the harmonic content.
Try a bandpass filter driven by the buffer and terminated in 50 ohms.
Well I have to decide what to do, get another oscillator or try the
filtering. The difference to another oscillators (like tbolt ocxo or
LPRO) is so huge that I do not know how hard it would filter the 10544A
to the same level, which parts to use and how much it will cost.
Should you retrofit an improved oscillator circuit you may as well
replace the oven controller to eliminate the oven switching frequency
related sidebands.
Infact the switcher sidebands are now gone:
http://www.amigazone.fi/files/gpsdo/544-13.png
http://www.amigazone.fi/files/gpsdo/544-14.png
It was like thios earlier:
http://www.amigazone.fi/files/gpsdo/544-1.PNG
http://www.amigazone.fi/files/gpsdo/544-2.PNG
It was easier than I expected... When I took the output directly from
OSC pins without using the coax connector like you suggested the
switcher peaks was gone!
So I had a closer look to the PCB today and noticed that it has only 2
layers and the ground net is too thin and goes around PCB totally wrong
way so that switcher current seems to flow via the signal flow. It's PCB
layout design fault, I think that with correct layout design this could
be done even with 2-layer PCB correctly.
But this is not a problem of course because I will design my own PCB for
the final system anyway, having SMA connectors for 10 MHz and EFC. I'll
also create separate power supply for ovens and signalling stages.
Esa
The switcher sidebands will still be there, they are just buried in the
spectrum analyser noise floor.
Does the board use the recommended LC filters and regulator for the
oscillator supply as depicted in Figure 3 on the 10544A data sheet?
Reducing the harmonics by 40 dB shouldn't be too difficult.
However any amplifier after the filter will need to be carefully
designed to keep the distortion below -60dBc.
Its better to use a multi section LC filter rather than trying to do it
with a single LC filter.
The filter phase stability will be better with a multi section LC filter.
Alternatively you could use a bandpass filter combined with some series
tuned shunt LC circuits to short out the 2nd and 3rd harmonic components.
The required parts shouldn't be too expensive, however you may need to
wind your own inductors for the series tuned LC circuits.
Air core or powdered iron core inductors should be OK as long as you use
shields between filter sections etc.
Bruce
Bruce Griffiths wrote:
Esa Heikkinen wrote:
Since retrofitting an improved oscillator circuit isn't really an option
you will need to filter the output to reduce the harmonic content.
Try a bandpass filter driven by the buffer and terminated in 50 ohms.
Well I have to decide what to do, get another oscillator or try the
filtering. The difference to another oscillators (like tbolt ocxo or
LPRO) is so huge that I do not know how hard it would filter the 10544A
to the same level, which parts to use and how much it will cost.
Should you retrofit an improved oscillator circuit you may as well
replace the oven controller to eliminate the oven switching frequency
related sidebands.
Infact the switcher sidebands are now gone:
http://www.amigazone.fi/files/gpsdo/544-13.png
http://www.amigazone.fi/files/gpsdo/544-14.png
It was like thios earlier:
http://www.amigazone.fi/files/gpsdo/544-1.PNG
http://www.amigazone.fi/files/gpsdo/544-2.PNG
It was easier than I expected... When I took the output directly from
OSC pins without using the coax connector like you suggested the
switcher peaks was gone!
So I had a closer look to the PCB today and noticed that it has only 2
layers and the ground net is too thin and goes around PCB totally wrong
way so that switcher current seems to flow via the signal flow. It's PCB
layout design fault, I think that with correct layout design this could
be done even with 2-layer PCB correctly.
But this is not a problem of course because I will design my own PCB for
the final system anyway, having SMA connectors for 10 MHz and EFC. I'll
also create separate power supply for ovens and signalling stages.
Esa
The switcher sidebands will still be there, they are just buried in the
spectrum analyser noise floor.
Does the board use the recommended LC filters and regulator for the
oscillator supply as depicted in Figure 3 on the 10544A data sheet?
Reducing the harmonics by 40 dB shouldn't be too difficult.
However any amplifier after the filter will need to be carefully
designed to keep the distortion below -60dBc.
Its better to use a multi section LC filter rather than trying to do it
with a single LC filter.
The filter phase stability will be better with a multi section LC filter.
Alternatively you could use a bandpass filter combined with some series
tuned shunt LC circuits to short out the 2nd and 3rd harmonic components.
The required parts shouldn't be too expensive, however you may need to
wind your own inductors for the series tuned LC circuits.
Air core or powdered iron core inductors should be OK as long as you use
shields between filter sections etc.
Bruce
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.
Esa
Another option (which I used several decades ago with a 10544A) was to
use a double tuned critically coupled RF transformer driven by a unity
gain buffer with another unity gain amplifier buffering the output
across the secondary tuning capacitor. NB the Q of the primary and
secondary tuned circuits should be the same (need to insert a resistor
in the secondary to set the Q). The inductors were hand wound and
trimmer caps were used for tuning.
Bruce
Esa Heikkinen wrote:
Hi Bruce,
White emitter follower circuit attached.
Here are the results. First, the circuit draws 33.3 mA at 12 vols. And
it's gain is 1:1 as expected with emitter follower - checked that with
RF generator as an input just to make sure that I built it correct.
With HP10544A the output level is below +8 dBm:
http://www.amigazone.fi/files/gpsdo/544-10.png
(with 2 meters of RG174 and SMA-connectors between the circuit and SA)
Now the harmonics look like this:
http://www.amigazone.fi/files/gpsdo/544-11.png
And 2nd harmonic level:
http://www.amigazone.fi/files/gpsdo/544-12.png
So the spec told that 2nd harmonic should be more than 25 dB down, it is
just in spec...
So the final result looks that the 10544A is OK but it really has this
kind of output spectrum?
Esa
You can reduce the circuit current by increasing the value of R10.
Try 47 ohms instead of 39 ohms.
The value depends on the difference in forward voltage drop of the LED
and the Vbe of Q5.
Should still get a low distortion output even when the load is a short
to ground.
Bruce
Esa,
I am back home now and can't access the amigazone files from my home IP.
Can you provide access?
Thanks,
Joe
-----Original Message-----
From: time-nuts-bounces@febo.com [mailto:time-nuts-bounces@febo.com] On
Behalf Of Esa Heikkinen
Sent: Monday, March 09, 2009 4:51 PM
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] About HP10544A
Since retrofitting an improved oscillator circuit isn't really an option
you will need to filter the output to reduce the harmonic content.
Try a bandpass filter driven by the buffer and terminated in 50 ohms.
Well I have to decide what to do, get another oscillator or try the
filtering. The difference to another oscillators (like tbolt ocxo or
LPRO) is so huge that I do not know how hard it would filter the 10544A
to the same level, which parts to use and how much it will cost.
Should you retrofit an improved oscillator circuit you may as well
replace the oven controller to eliminate the oven switching frequency
related sidebands.
Infact the switcher sidebands are now gone:
http://www.amigazone.fi/files/gpsdo/544-13.png
http://www.amigazone.fi/files/gpsdo/544-14.png
It was like thios earlier:
http://www.amigazone.fi/files/gpsdo/544-1.PNG
http://www.amigazone.fi/files/gpsdo/544-2.PNG
It was easier than I expected... When I took the output directly from
OSC pins without using the coax connector like you suggested the
switcher peaks was gone!
So I had a closer look to the PCB today and noticed that it has only 2
layers and the ground net is too thin and goes around PCB totally wrong
way so that switcher current seems to flow via the signal flow. It's PCB
layout design fault, I think that with correct layout design this could
be done even with 2-layer PCB correctly.
But this is not a problem of course because I will design my own PCB for
the final system anyway, having SMA connectors for 10 MHz and EFC. I'll
also create separate power supply for ovens and signalling stages.
--
73s!
Esa
OH4KJU
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.
Hello!
I am back home now and can't access the amigazone files from my home IP.
Can you provide access?
You should now have access.
--
73s!
Esa
OH4KJU
Hi Bruce...
Try increasing C2 and C8 in the white emitter follower circuit schematic
to 100nF.
Doesn't seems to change anything:
http://www.amigazone.fi/files/gpsdo/544-15.png
http://www.amigazone.fi/files/gpsdo/544-16.png
When doing this measurement I noticed that the 3rd and higher harmonics
level are changing! First I thought that the capacitor change was some
effect on harmonics but then those peaks come back...
Time domain analysis about 3rd harmonic level gave some explanation:
http://www.amigazone.fi/files/gpsdo/544-17.png
The EF circuit was turned on at start of sweep. Something is heating and
raising the harmonic level. The bump and the end of sweep is a test
where I momentary switch off the 12V feed to the EF circuit to see how
it reacts.
Temperature sensitivity was also verified with cold spray:
http://www.amigazone.fi/files/gpsdo/544-18.png
Well, that's not an issue but makes measurements harder because even
without any changes the results can differ. However the temperature
effect on the 2nd harmonic frequency was very small.
The switcher sidebands will still be there, they are just buried in
the spectrum analyser noise floor.
Yes you're right. Just changed some settings and there they are again:
http://www.amigazone.fi/files/gpsdo/544-19.png
About 90 dB below carrier.. I would say that it's good enough! The LPRO
for example, gives higher spuriouses, but far away from fundamental.
Does the board use the recommended LC filters and regulator for the
oscillator supply as depicted in Figure 3 on the 10544A data sheet?
I haven't analyzed it fully but it seems to be just the datasheet
circuit having LM723, 10 uH coil etc. But it's layout is totally wrong
because oven switcher current runs via wrong trace. It would also been
possible to have other side as grouding copper but this was not done.
The PCB is manufactured by Cubic western data.
The required parts shouldn't be too expensive, however you may need
to wind your own inductors for the series tuned LC circuits.
Air core or powdered iron core inductors should be OK as long as you
use shields between filter sections etc.
Sounds like hard. How it's possible that lower grade ocxo's (like in
thunderbolt) output so much better spectrum? Is it all about ocxo output
driver circuit? Would it be easier to modify the 10544A itself than
trying to clean the distortion? Has anyone tried that kind of modification?
--
73s!
Esa
OH4KJU
Esa Heikkinen wrote:
Hi Bruce...
The required parts shouldn't be too expensive, however you may need
to wind your own inductors for the series tuned LC circuits.
Air core or powdered iron core inductors should be OK as long as you
use shields between filter sections etc.
Sounds like hard. How it's possible that lower grade ocxo's (like in
thunderbolt) output so much better spectrum? Is it all about ocxo output
driver circuit? Would it be easier to modify the 10544A itself than
trying to clean the distortion? Has anyone tried that kind of modification?
Esa
Attached circuit schematic illustrates how you could filter the harmonic
content of the 10544A below -50dBc.
If the output stage is elaborated somewhat its possible to get below
-60dBc or so.
Input stage is a simple cascode which has lower distortion that the 10544A.
The double tuned transformer has a primary inductance of 4uH, a
secondary inductance of 8uH and a coupling coefficient of 0.1.
The output stage consists of a common base stage plus a common collector
stage capable of driving an open circuit load, a 50 ohm load or a short
circuit load.
The output impedance of the output stage is nominally 50 ohms, its input
impedance is also nominally 50 ohms.
Bruce
Esa Heikkinen wrote:
Hi Bruce...
Try increasing C2 and C8 in the white emitter follower circuit schematic
to 100nF.
Doesn't seems to change anything:
http://www.amigazone.fi/files/gpsdo/544-15.png
http://www.amigazone.fi/files/gpsdo/544-16.png
When doing this measurement I noticed that the 3rd and higher harmonics
level are changing! First I thought that the capacitor change was some
effect on harmonics but then those peaks come back...
Time domain analysis about 3rd harmonic level gave some explanation:
http://www.amigazone.fi/files/gpsdo/544-17.png
The EF circuit was turned on at start of sweep. Something is heating and
raising the harmonic level. The bump and the end of sweep is a test
where I momentary switch off the 12V feed to the EF circuit to see how
it reacts.
Temperature sensitivity was also verified with cold spray:
http://www.amigazone.fi/files/gpsdo/544-18.png
Well, that's not an issue but makes measurements harder because even
without any changes the results can differ. However the temperature
effect on the 2nd harmonic frequency was very small.
The switcher sidebands will still be there, they are just buried in
the spectrum analyser noise floor.
Yes you're right. Just changed some settings and there they are again:
http://www.amigazone.fi/files/gpsdo/544-19.png
About 90 dB below carrier.. I would say that it's good enough! The LPRO
for example, gives higher spuriouses, but far away from fundamental.
Does the board use the recommended LC filters and regulator for the
oscillator supply as depicted in Figure 3 on the 10544A data sheet?
I haven't analyzed it fully but it seems to be just the datasheet
circuit having LM723, 10 uH coil etc. But it's layout is totally wrong
because oven switcher current runs via wrong trace. It would also been
possible to have other side as grouding copper but this was not done.
The PCB is manufactured by Cubic western data.
The required parts shouldn't be too expensive, however you may need
to wind your own inductors for the series tuned LC circuits.
Air core or powdered iron core inductors should be OK as long as you
use shields between filter sections etc.
Sounds like hard. How it's possible that lower grade ocxo's (like in
thunderbolt) output so much better spectrum? Is it all about ocxo output
driver circuit? Would it be easier to modify the 10544A itself than
trying to clean the distortion? Has anyone tried that kind of modification?
Esa
The circuit that is used to extract and buffer the signal from the
crystal oscillator can have a significant effect on the distortion.
In the case of the 10544A the cascode buffer contributes most of the
distortion.
The emitter follower contribution to distortion with a 1k load is much
smaller, however it is sensitive to slew rate distortion with
significant capacitive loading.
Using a common base stage as in the 10811A can have lower distortion, as
the emitter current is filtered by the crystal.
If one uses RF transformers between stages then one can cascade common
base stages and increase the output signal current.
An output stage with a 50 ohm output impedance is useful in that the
connection to external circuitry can then be a 50 ohm transmission line.
Some of the early OCXOs even used a crystal filter to clean up the
oscillator output.
Modifying the 10544A circuit would require a new circuit board and the
mechanical space available would restrict what you can fit in unless you
you use some surface mount components.
You would need to reuse the trimmer capacitor.
Bruce
Esa Heikkinen wrote:
Hi Bruce...
Try increasing C2 and C8 in the white emitter follower circuit schematic
to 100nF.
Doesn't seems to change anything:
http://www.amigazone.fi/files/gpsdo/544-15.png
http://www.amigazone.fi/files/gpsdo/544-16.png
When doing this measurement I noticed that the 3rd and higher harmonics
level are changing! First I thought that the capacitor change was some
effect on harmonics but then those peaks come back...
Time domain analysis about 3rd harmonic level gave some explanation:
http://www.amigazone.fi/files/gpsdo/544-17.png
The EF circuit was turned on at start of sweep. Something is heating and
raising the harmonic level. The bump and the end of sweep is a test
where I momentary switch off the 12V feed to the EF circuit to see how
it reacts.
Temperature sensitivity was also verified with cold spray:
http://www.amigazone.fi/files/gpsdo/544-18.png
Well, that's not an issue but makes measurements harder because even
without any changes the results can differ. However the temperature
effect on the 2nd harmonic frequency was very small.
The switcher sidebands will still be there, they are just buried in
the spectrum analyser noise floor.
Yes you're right. Just changed some settings and there they are again:
http://www.amigazone.fi/files/gpsdo/544-19.png
About 90 dB below carrier.. I would say that it's good enough! The LPRO
for example, gives higher spuriouses, but far away from fundamental.
Does the board use the recommended LC filters and regulator for the
oscillator supply as depicted in Figure 3 on the 10544A data sheet?
I haven't analyzed it fully but it seems to be just the datasheet
circuit having LM723, 10 uH coil etc. But it's layout is totally wrong
because oven switcher current runs via wrong trace. It would also been
possible to have other side as grouding copper but this was not done.
The PCB is manufactured by Cubic western data.
The required parts shouldn't be too expensive, however you may need
to wind your own inductors for the series tuned LC circuits.
Air core or powdered iron core inductors should be OK as long as you
use shields between filter sections etc.
Sounds like hard. How it's possible that lower grade ocxo's (like in
thunderbolt) output so much better spectrum? Is it all about ocxo output
driver circuit? Would it be easier to modify the 10544A itself than
trying to clean the distortion? Has anyone tried that kind of modification?
Esa
The attached circuit schematic is for a slightly modified white emitter
follower that reduces the temperature dependence of the input transistor
collector current.
It also increases the loop gain slightly.
Bruce
