SSA3000X measures a tone, 10MHz external reference clock

[dpenev]

Hi All,

I am observing something interesting I think.
We all know that SSA3000X is not perfect in terms of local oscillator phase noise.
I have Anritsu MG3700A RF generator with oven controlled local oscillator with smaller phase noise compared to SSA3000X

I have decided to test if applying reference clock 10MHz from the RF generator to the SSA3000X reference input
will give me any improvements.

See picture of SSA3000X measures  1.1GHz sine wave from the RF generator (the same providing the OCXO 10MHz) measurements.
Trace B is when I plug the reference 10MHz cable. We seems to get some slight modulation in this case.
Also the phase noise skirt seems not affected in this frequency span.
So instead of improvements I am getting worse measurement.

Anyone having an idea why it is like this?

[TheSteve]

Some 60 Hz sneaking in with the external ref?

[dpenev]

50hz right, hmm. Should check how5to avoid this

Sent from my MI NOTE Pro using Tapatalk

[DaJMasta]

Am I reading this wrong or are the tones at 100Hz and 200Hz off the carrier?  If you're on a 60Hz power system, I'd expect to see 120Hz/240Hz instead of this, and you should be able to see that as a difference at this zoom level.  If the primary filter cap after the rectifier is not sufficient, though, these tones could certainly be getting through the regulation of the 10MHz source.

[edpalmer42]

Those tones are already < -50dBc.  Getting them lower could be really hard.  Here are some things to consider:

- Are both devices plugged in to the same power source?  If they happen to be on separate circuits, there could be common mode voltage between them.

- External reference inputs are typically on the back of the unit.  Is the BNC cable lying on or running parallel to the power cord?  Move it away.  If it has to pass over or under a power cord, position the cords so that they're at 90 degrees to each other.

- Have you tried a different cable (BNC, I presume) between the units?  Maybe the one you're using is defective.

- Are you using shielded power cords?

- Are you using double-shielded cables between the units?

- You might have to use an RF transformer between the units to break a ground loop between the units.

- You could try bolting a copper strap (a wire might work, but a strap would be better) between the chassis of the two units.  Keep it as short as possible.

This is why many spectrum analyzers look internally like a bank vault.  They need a ridiculous amount of shielding to keep the system free of artifacts like this.

Ed

[dpenev]

DaJMasta in Bulgaria we are on the 50Hz mains and indeed the spurs appears at 100Hz, 200Hz etc.

>Those tones are already < -50dBc.  Getting them lower could be really hard.  Here are some things to consider:
>- Are both devices plugged in to the same power source?  If they happen to be on separate circuits, there could be common mode voltage between them.

Yes they are on the same power source. 

>- External reference inputs are typically on the back of the unit.  Is the BNC cable lying on or running parallel to the power cord?  Move it away.  If it has to pass over or under a power cord, position the cords so that they're at 90 degrees to each other.

Yes it is BNC 1m cable. Back of the units indeed. There are no power cords near by.   

>- Have you tried a different cable (BNC, I presume) between the units?  Maybe the one you're using is defective.

I will try tomorrow and will report.   

>- Are you using shielded power cords?
No, I believe the power cords I have are not shielded.

>- Are you using double-shielded cables between the units?
I think my cable is single shielded RG58. 

> - You might have to use an RF transformer between the units to break a ground loop between the units.
This is not very easy to implement but I will think.

>- You could try bolting a copper strap (a wire might work, but a strap would be better) between the chassis of the two units.  Keep it as short as possible.
Will test this.
Already two coax cables connected between the two units so I don't expect miracles 

>This is why many spectrum analyzers look internally like a bank vault.  They need a ridiculous amount of shielding to keep the system free of artifacts like this.
Is this a common practice to use external 10MHz reference for the equipment like spectrum analyzers and signal generators?
I start thinking that this is probably not very easy to be done in general.

>Ed

Thank you Ed for your remarks.
Dimitar

[DaJMasta]

In that case, it seems like a perfectly reasonable culprit.  I would check the power supply line to the 10MHz source, but there are some less invasive things to try.  Fields can induce the hum in wiring, but the SA should be plenty shielded and the BNC to connect the two should be reasonably shielded as well... so my guess would be it's getting into the signal before it arrives at the SA.

If you can look at the 10MHz signal on a scope to verify (high resolution mode, FFT) that there are the same tones in it, then it would at least take the out of the list of potential culprits.  The SA could also do this (measure the 10MHz reference output without being referenced to it), but it's nice to have a neutral third party to measure, and -50dBc should be measurable on a modern scope FFT.

Then maybe check the DC resistance from the shield of the BNC for the reference to the chassis ground for the sig gen, and maybe check the resistance of the ground pin of your signal generator, if either is sort of half-floating because of a poor ground connection, this sort of hum can be picked up on a shield instead of kept out.

While there are some people that like isolation in their 10MHz reference distribution, most commercial distribution amps don't use it - since the signal is fairly large and fairly pure, just filtering it on the input side can keep out a lot of the potential noise, so elaborate grounding and shielding measures are usually not used or required.

[dpenev]

Hi All,

I have done few more experiments

- I have tested with another RG58 cable -> same 100Hz modulation spurs
- I have tested with double shielded RG142 cable -> same 100Hz modulation spurs
- I have tried to connect the chassis of both generator and the SSA3000X with no change in the observation   
- I have measured the 10 MHz reference clock from the SSA3000X (the analyzer is using its internal sync) -> no visible modulation 100Hz/200Hz  around the 10MHz tone.

[edpalmer42]

Have you checked your levels & impedances to make sure that the SSA3000X reference input isn't being overdriven?  Regardless of what the spec sheets say, try adding an attenuator to the reference link.  What happens when you reverse the connections and use the RefOut from the SSA3000X to the RefIn of the MG3700A?

Ed

[edpalmer42]

Something else you could try is to connect the MG3700A to the SSA3000X, but do it in a way that only connects the BNC shield, not the center conductor.  That would help to confirm whether the noise is coming over the shield.

Ed

[dpenev]

Hi Ed,

Thank for the suggestions.

Please see the attached measurement
Trace A - no sync between the generator and analyzer
Trace B - generator (with possibly more stable clock) provides clock to the analyzer
Trace C - analyzer provides clock to the generator
Both instruments detects the external clock on their displays.

We see that if sync is provided the tone is at the center (as it is expected)
We see also that in case of clock from the generator the phase noise 50 Hz around the tone is smaller but as disadvantage we see the 50 Hz tones at the edges.
On my previous measurements we see that the phase noise is practically equal (synced and not synced case) for frequency ranges starting from 200Hz (~ -75dBc@200Hz) further from the tone.

The 10MHz clock is a bit stronger for the SSA3000X (it is 12dBm and SSA expects up to 10dBm). Adding attenuation doesn't change the observations.

I have tried to connect only the cable shield with no change in the un-synchronized measurement.

Thanks
Dimitar

[TurboTom]

Have you possibly got a separate reference oscillator (or another device with 10MHz ref out)  that you could feed to both instruments? Any chance to install an isolation transformer in the 10MHz signal so the two instruments don't share common ground via the REF input?

[DaJMasta]

It definitely looks like noise on the Anritsu's oscillator from those, you can see the worse phase noise when the SA is using its own reference, but the tones at 100Hz are only present on the B trace (unless they're being perfectly covered up by the other two).

I still think looking at that reference signal with a scope's FFT is worthwhile, I think the 100Hz sidebands should show at least a little, though perhaps not the 200Hz, and that would just confirm that it's the Anritsu oscillator generating the noise.



From there.... well you'd probably have to be troubleshooting the Anritsu's reference board or power supply first and checking the power rails that go over to the oscillator for that frequency ripple on the DC lines.

[dpenev]

Hi DaJMasta,

Indeed I see the 50/100Hz disturbance only if Anritsu is providing sync to the SSA3000X

I have measured the Anritsu 10Mhz with SSA3000X using ~ 500Hz span  and I don't see those 50/100Hz disturbance.
However the 10MHz clock is about 40dB stronger (compared to my -30dBm 1.1GHz test sine) so I have to use different SSA attenuation settings.
So it is possible that that I just don't see the 50/100hz tones due to higher noise floor for this settings. 
 
I have check the Anritsu 10Mhz clock with oscilloscope and I don't see issue. However I don't expect I could see that small disturbers with my Rigol DS2102.

Hi TurboTom

I will try to get clock from another oscillator I have and will report

Thank you guys for the advice.

[rf-loop]

This image is not directly for previous messages but still it may give some info.

[edpalmer42]

Looking at the manuals, I see that the MG3700A RefOut is TTL and I can't find any specs on what the SSA3000X RefIn wants.  Dimitar, where did you find that number of +10 dBm?    to both of them!  It should be an AC-coupled sine wave with an impedance higher than 50 ohms.

Try putting a small capacitor in series with the cable between RefOut and RefIn.  Try something like 10 nf and see what happens.  The idea is to block the 50 Hz but have minimal effect on the 10 MHz.  An attenuator might also be helpful.  If the SSA3000X is expecting a sine wave at 0 dBm, it probably doesn't like a square wave at +18 dBm!  I know you just tested with the attenuator, but that TTL signal changes everything!  TTL includes trash that an analog input would choke on.

If the capacitor and attenuator don't work, I recommend that you spend some time determining what the parameters are for the SSA3000X RefIn.  Sine, Square, minimum input level, input impedance.  Now look at the RefOut of the MG3700A and see what, if anything, has to be done to that signal to make it truly compatible with the SSA3000X.  You might need a transformer and/or a high pass filter and/or a low pass filter.  Maybe even an amplifier! 

Ed

[dpenev]

Hi rf-loop,

Oh you have replicated my tests and you do see the 60Hz, 120Hz ... tones?!

For your trace B, your generator 10MHz ref out is connected to SSA ref in but the clock is not enabled. Is this what you mean?
What is the conclusion you have made based on your measurements?
   
Ed, SSA3000X datasheet says
10 MHz reference output 10 MHz, >0 dBm, 50 Ω, BNC-female
10 MHz reference input 10 MHz, -5dBm~+10dBm, 50 Ω, BNC-female

The MG3700A reference clock is 10MHz sine wave but with strong harmonics so definitely not pure sine wave.
(on the scope it looks like good sine wave) Its main tone is about 12dBm measured with the SSA. 
I will experiment with a capacitor.

[rf-loop]

Hi rf-loop,

Oh you have replicated my tests and you do see the 60Hz, 120Hz ... tones?!

For your trace B, your generator 10MHz ref out is connected to SSA ref in but the clock is not enabled. Is this what you mean?
What is the conclusion you have made based on your measurements?
   
Ed, SSA3000X datasheet says
10 MHz reference output 10 MHz, >0 dBm, 50 Ω, BNC-female
10 MHz reference input 10 MHz, -5dBm~+10dBm, 50 Ω, BNC-female

The MG3700A reference clock is 10MHz sine wave but with strong harmonics so definitely not pure sine wave.
(on the scope it looks like good sine wave) Its main tone is about 12dBm measured with the SSA. 
I will experiment with a capacitor.

Sidenote. In my country mains is 50Hz.
These small side peaks what ever frequency they have are generated by HP due to reason or other.

For your trace B, your generator 10MHz ref out is connected to SSA ref in but the clock is not enabled. Is this what you mean?

No.  Trace B   SSA run with its own internal reference and nothing - nothing - connected to SSA ref input. HP oven out is only connected to SSA RF input (via attenuator for adjust level so that all, trace A, B and D have been exactly same level)  And this image show small sidetones and they come from HP, with certainty. (sidenote: HP53131A is not generator. It is counter. But it have quite good 10MHz reference with quite excellent short time stability and not bad phasenoise.)

In D sidetones are same. Because HP oven out /Ref Out) is connected to SSA Ref In and locked. And it is also split  to SSA RF input via attenuator for adjust RF input level to match other traces.
But when compare all traces and look example phase noise shape. Using ext ref or internal ref did not change anything.  (I have seen other equipment what rise phase noise when external ref due to poor phase lock to ext ref. SSA do not show this issue or it is undetectable in this kind of totally "poormans" test setup what I did and can show only if there exist some catastrophal.

This my "test" is not for make further conclusions or to define the SSA3kX performance or features! (and also not against OP's findings.)
If want do some tests for meaningful conclusions it need totally different (higher) grade tests using different grade of equipments for testing if DUT is SSA3kX.  It is easy to do "Crap in = crap out" and this is just it in some meaning. But each one can make their own conclusions. There are information about signal,  settings and the image that was the result.

[dpenev]

Hi rf-loop,

I understand, you are measuring only the 10MHz reference clock.
Indeed this doesn't relate directly to what I am testing.

What I am observing looks like 10MHz from MG3700A is having some small 50Hz disturbance.
What puzzles me is that I can not see it if I measure the 10MHz clock with SSA3000X directly.
My current explanation is that since the clock is strong those disturbance is buried in the noise floor. 
Any idea how I can confirm this?
   

[TurboTom]

Maybe the SSA ref input's 50 Ohms impedance (supposedly) somehow loads down the Ref signal circuitry in the MG3700A, causing some rail to drop sufficiently so the 100Hz hum feeds throug and somehow causes the output signal to be slightly amplitude modulated. You may try the same thing with a terminator on the MG3700A's ref output. Peculiar thing, though...

Cheers,
Thomas

[rf-loop]

What I am observing looks like 10MHz from MG3700A is having some small 50Hz disturbance.
What puzzles me is that I can not see it if I measure the 10MHz clock with SSA3000X directly.
My current explanation is that since the clock is strong those disturbance is buried in the noise floor. 
Any idea how I can confirm this?

There are limits...

This image show near carrier 100Hz span. It is not perhaps absolutely best what it can do but but overall can say that  SSA3kX can not go more deep near carrier due to noise, phase noise. What is under noise level it can not see.

[dpenev]

Hi rf-loop,

Yes my measurements kind of match with yours.
But you are measuring at 10Mhz where the SSA phase noise is small.

Back on the original topic. I have set my Anritsu generator to generate 5dBm sine wave so similar strength as the 10MHz on its back panel.
Below 200MHz I can not see the 50Hz disturbance.
They start to appear at about 300MHz. Please check my measurements:
-A without external reference
-B with external reference.

We see the same I was observing in the past.
The external reference improves the phase noise 100Hz around the tone but introduce this 50Hz disturbance.
Here the disturbance is at -80dBc@50Hz. On my previous 1.1GHz measurements the disturbances are at around -67dBc@50Hz   
So this strange phenomenon become stronger as the generator tone frequency increases.

[rf-loop]

Hi rf-loop,

Yes my measurements kind of match with yours.
But you are measuring at 10Mhz where the SSA phase noise is small.

Back on the original topic. I have set my Anritsu generator to generate 5dBm sine wave so similar strength as the 10MHz on its back panel.
Below 200MHz I can not see the 50Hz disturbance.
They start to appear at about 300MHz. Please check my measurements:
-A without external reference
-B with external reference.

We see the same I was observing in the past.
The external reference improves the phase noise 100Hz around the tone but introduce this 50Hz disturbance.
Here the disturbance is at -80dBc@50Hz. On my previous 1.1GHz measurements the disturbances are at around -67dBc@50Hz   
So this strange phenomenon become stronger as the generator tone frequency increases.

I will later do some more tests when I can arrange test setup with good HP RF generator and then separate external reference for SSA. (I may try with "clean" and with "dirty" ext ref. and look if it rise some spurs with example over 300MHz input freq from quite clean source)


I'm interested to see, if possible, test so that your SSA run with its own reference (Nothing connected to extref in/out) and connect this 10MHz ref (is it from Anrizu reference output?) to SSA rf input.
So that can somehow analyze this external reference signal quality what you use for SSA. Including harmonics and possible non harmonic spurs (exaple 0 - 60MHz. Yes I know sweep time is long(!) with example 30Hz RBW and 3Hz VBW). (keep mixer level -30dBm)

Then also what it have near 10MHz carrier.

[dpenev]

Hi rf-loop,

A quick answer of your questions:

My setup includes
- Anritsu MG3700A RF signal generator
- SSA3000X analyzer

MG700A has 10Mhz reference output which is claimed to be TTL level. It is about 12dBm measured with the SSA and has lot of harmonics(So it is not pure sine wave)
I have checked it on the oscilloscope and it looks as good sine wave (but obviously it is not)

I have measured the MG3700A 10MHz ref output close around the peak and I see something as on your measurement without any 50Hz modulating tones.

Then the plots I have attached so far are Anritsu MG3700A sine wave output measured with SSA3000X with and without(nothing connected to SSA ref in) 10MHz reference from Anritsu to the Siglent.
     
50Hz disturbance I start seeing after I set Anritsu sine at 300MHz or more at 5dBm

[rf-loop]

Ok. Somehow I have jumped over msg #10 attached image. These white tiny attachment thumb size images with old eyes are sometimes...
I will try later do some tests using some ext references and with rf generator. 

[Performa01]

You should use a span of at least 250Hz (better 500Hz) for all marrowband measurements.

In your very first picture in the opening post, we can see that the 100Hz sidebands are dominant (probably from fullwave rectified mains), whereas the 50Hz spurs are almost negligible. Observing the 2nd and 3rd harmonics of the main frequency would be much more revealing.

[dpenev]

Hi Performa01,

Indeed you are right.

Note that my main goal was to see if using external reference will reduce the SSA phase noise.
I have got the Anritsu generator recently and it is with OCXO so it is suppose to have best clock quality in our lab.
After some measurements I came to a conclusion that external reference reduces the phase noise (at least for the case generator provides both reference and the measured signal)
in the аrеa 50hz around the tone. That's why I am using so short span.

For 50/100/etc Hz modulation your are perfectly right that we better refer to the stronger 100Hz spur.

[Performa01]

Yes, for lowest close-in phase noise an OCXO reference signal is mandatory - and of course this can only be verified by looking at a low phase noise input signal, so the only sensible setup is both the reference and the signal coming from a high quality synthesizer like the MG3700A.

I just thought this was already confirmed and now the only remaining question was where the mains harmonic modulation comes from - and how to minimize it.

One of the suggestions has been to observe the reference signal with the spectrum analyzer in order to see if the 100Hz sidebands can be seen - I think this was good avice and certainly the first thing I'd try myself in such a situation. But to do this, quite obviously a wider span than 100Hz is required to get a clear picture of the situation.

If it turns out that that interference actually exists at the MG3700A reference output, I'd check the PSU inside the MG3700A, replace old electrolytics and maybe add some additional supply rail filtering for the OCXO and the associated reference output buffer amplifier.

Best of luck and please keep us updated!

[edpalmer42]

Since the 10 MHz output from the MG3700A is labelled on the unit as being TTL, I'm skeptical of it's suitability for a low phase noise source.  When the TTL is at it's logical high level, it will leak all the trash on the +5V rail into the output.  Dpenev has stated that the signal looks mostly like a sine wave so they've obviously filtered the high frequencies, but it looks like there's some remaining low frequency trash.

Ed

[dpenev]

Hi performa01,

Observations so far:
-no 50hz and multiple in the Mg3700a 10MHz reference signal visible on my ssa3000x
-50hz and multiple with strong 100hz component visible if 1.1ghz(for example) sine wave from Mg3700a measured with ssa3000x using ext ref from Mg3700a. The 50hz and multiple not visible if internal Clock used with ssa3000x.
-funny enough 50hz and multiple are not visible on ssa3000x if Mg3700a generates sine with frequency below 300MHz and also provides Clock to ssa3000x.
-I don't know why the Mg3700a don't use pure sine wave for its 10MHz reference but I think that more square a like would have stepper edges and would be easier to lock to with less jitter. On a scope it looks as sine wave. Later i will provide measurement in bigger span so harmonics wisible.


Sent from my MI NOTE Pro using Tapatalk

[dpenev]

This is how the MG3700A 10MHz reference clock looks like

 

[rf-loop]

What ever happen using MG3700A but Siglent SSA3000X is extremely critical about external reference signal. Of course it need be stable and accurate enough but also it need be clean, as clean as SSA need reference clock.  It also is very sensitive for ExtRef amplitude modulation. Just: All need be clean.
Some other machine may work so that it only follow ext ref carrier. 

It looks somehow like that Siglent perhaps totally swap whole reference, internal or just external. And if this is case, dirty reference makes all results dirty. Just like garbage in - garbage out.

Here is three images one is not necessary but for better visibility.




Image 1.
100MHz CW from 30 years old HP Rf generator. In this image Siglent run with its own internal reference. (same as in image 3 trace D but other shot)




Image 2.
Signal used as External Reference for SSA in Image 3.
10MHz carrier AM modulated with 35Hz and modualtion depth 1% (35Hz sidebands -46dBc).   
Also this same signal is used with 10% modulation depth (35Hz sidebands -26dBc)




Image 3.
Trace D: as also in image 1. (SSA using internal reference)
Traces A and B:  SSA use ExtRef as image 2.

Also after done other many tests and my conclusion (until other real evidences)  is that do not use SSA ExtRef until you have  high quality very clean reference signal for this purpose.

[dpenev]

Hi rf-loop,

I don't understand your point.
You synchronize your SSA300X with 'dirty' 10 Mhz and obtain 'dirty' SSA3000X measurement.
In my opinion this is absolutely logical. Is it not the purpose of the external synchronization
to replace the whole internal clock system of the instrument based on the externally provided one?

Note that my original confusion is that my 10MHz reference appear clean but the SSA300X measurements
are dirty for signals above 300MHz.   

[TurboTom]

Did you try what I suggested here?

Since the REF output of your generator appears to be rather high level, maybe the load of the SSA's REF input (designed for a lower level) pulls it down so much that and internal power rail inside the generator gets unstable. Even more so since apparently at higher generator output frequencies, the intermodulation suppression of the power stage gets worse. It's a quick test to put a 50 Ohms terminator on the REF output and test again with the SSA running on it's internal reference.

It's just a thought that could very well be wrong but all your description of the problem maly also point into that direction...

[rf-loop]

Hi rf-loop,

I don't understand your point.
You synchronize your SSA300X with 'dirty' 10 Mhz and obtain 'dirty' SSA3000X measurement.
In my opinion this is absolutely logical. Is it not the purpose of the external synchronization
to replace the whole internal clock system of the instrument based on the externally provided one?   

This is perhaps bit sidepathway for your case. I do not know. I have not qualified your reference quality at all.

Do you think all instruments just swap whole internal reference with external reference if external is used directly without caring signal quality (example take only ref signal 10MHz carrier to adjust and lock internal freq to its freg).  I have "ton of" very accurate reference freq sources but not all are very pure if look AM sidetones or other impurities and I can use these with many equipments for frequency reference and no need care if there is some 50Hz or 100Hz mod or "ground loop hum".  But reference what run SA internals need be quite clean. Also, as can see my previous other test. I have used 10MHz quite clean reference oscillator and then injected just 50Hz via splitter (perhaps more high level than usually when just some ground loop). So it proof that it is also bit sensitive for  50Hz "hum" and not know where it produce 10M carrier modulation.


Example if look inside HP 859xE  SA.  External reference goes to External Reference PLL circuit. (if this HP SA have internal high accuracy reference or standard reference it is connected to external ref BNC using small coaxial bridge.)

This kind of system what HP have used is not so sensitive for external reference 1% or even 0.1% AM modulation, perhaps not even 10%. Or combined quite low level 50Hz "hum" + 10MHz strong carrier as reference. No it is not sensitive at all (this 50Hz is stopped by HPF. User can use dirty or clean reference as long as its frequency is stable (low phase noise etc).
But Siglent IS sensitive and this must be taken into account.
This is why I say: Use only very clean external reference for Siglent SSA. Period.

[SoundTech-LG]

this may/may not apply, but worth a read anyway...

http://literature.cdn.keysight.com/litweb/pdf/5990-9189EN.pdf

[dpenev]

Hi TurboTom,
Terminating the MG3700A ref output with 50 Ohm does not change the generated signal in terms of the 50Hz/100Hz modulating tones around the main tone.

Hi rf-loop,
I think I understand  your point now.
Amplitude modulation of the 10MHz reference signal should not affect the quality of the SSA3000X measurements.
As SSA3000X should contain PLL which should lock to the external reference, its phase(t) property but not to its amplitude(t) property.
Based on your measurements it is not the case and SSA3000X is sensitive to amplitude disturbance from the 10Mhz ref input. 
Good point!
It is interesting in my case that I measure the MG3700A 10MHz ref signal as a clean in the neighborhood or 10MHz