best value for money spectrum analyzer

[fisafisa]

Hi guys
I am calling upon your experience
I am looking for a good instrument for work to perform the following analyses

1) Phase noise of clocks (10MHz)
2) Pre EMC qualifications

Budget bracket up to 10K

Suggestions?

Many thanks in advance

Filippo

[tautech]

With the current promo on the EMI option one of the SSA3000X Plus series would be a good fit:
https://www.siglenteu.com/spectrum-analyzers/ssa3000x-plus/

[pdenisowski]

The Rohde and Schwarz FPC is also a good choice for EMI in that price range:

https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_common_library/dl_brochures_and_datasheets/pdf_1/App_sheet__EMI_debugging_solution_V1.01.pdf

[fisafisa]

Hi
Our main focus is quality of clocks to be used on high frequency data acquisitions.
We are developing a PTP driven distributed clock data acquisition system and are worried about the quality of the clock.

So I would like to be able to measure at the clock wander before correction and the resulting clock jitter after correction.
Will this instruments be able to measure this?

Thanks
Filippo

[1Ghz]

So I would like to be able to measure at the clock wander before correction and the resulting clock jitter after correction.
Will this instruments be able to measure this?

R&S FPC1000 able to measure phase noise.
Noise measurement: https://www.rohde-schwarz.com/webhelp/fpc_html_usermanual_en/Content/8f2e5a689ce04097.htm

I'm not sure Siglent SSA3000X Plus has the same functionality.
I can't find similar noise measurement chapter on the SSA3000X Plus user manual.

[tautech]

SVA1032X = SSA3021/3032X Plus HW
FFT Max Hold, markers as per screenshot.
16 bit 10 MHz sinewave 0dB from SDG6022X
Neither instrument referenced.

[pdenisowski]

To make phase noise measurements with a spectrum analyzer, you need to be able to measure both the power of the carrier and the power in dBm/Hz at a given frequency offset from the carrier.  The carrier power can be measured with a standard spectrum analyzer marker, but the noise power requires a special "noise marker" function that calculates the result in dBm/Hz.  The R&S FPC does have this noise marker function, and it can be used with a normal marker to produce results in dBc/Hz, which is the unit in which phase noise is measured.

In mid- to higher-end spectrum analyzers (outside of your price range), there is often a special phase noise measurement routine/option which has additional features and which automates this process to produces nice plots of phase noise versus frequency offset, etc. but you can still get good results using the "manual" noise marker process. 

I wrote some introductory whitepapers on measuring phase noise:

https://www.rohde-schwarz.com/us/products/test-and-measurement/analyzers/signal-spectrum-analyzers/white-paper-understanding-phase-noise-fundamentals-register_255143.html

https://www.rohde-schwarz.com/us/products/test-and-measurement/analyzers/signal-spectrum-analyzers/white-paper-understanding-phase-noise-measurement-techniques-register_255398.html

This material is also available in video format on the R&S YouTube channel:




The next time I get an FPC in my lab, I'll do a short video on how to use the noise marker 

[TurboTom]

...
So I would like to be able to measure at the clock wander before correction and the resulting clock jitter after correction.
Will this instruments be able to measure this?
...

Filippo -

considering your application, and of course, depending on the frequency range that you're looking for, there may be other instruments that better suit your needs. The HP 53310A modulation domain analyzer may be such a device. These aren't produced today anymore but they usually can be found second hand and aren't too expensive since it's some kind of "niche product".  Search for it on the forum and you'll probably find enough information to decide if that would possibly be an option for you.

[HalFoster]

+1 on the 53310A - this is what they *do*

http://www.testequipmenthq.com/datasheets/Agilent-53310A-Datasheet.pdf

Hal

[Performa01]

Of course, an instrument that is capable of noise markers or even proper phase noise plots would be preferable, but ultimately, this is more a matter of convenience.
For occasional use, one can do without. It is just a matter of doing the math to convert conventional marker data into noise density, as long as the RBW is known.

Much more important is the phase noise of the analyzer itself. Is it sufficiently lower than what you want to measure? Many affordable instruments might be as bad as 90 dBc/Hz at 10 kHz distance from the carrier, which is pretty much useless for even very moderate demands.

Most of the more expensive instruments hardly exceed 100 or 105 dBc/Hz at 10 kHz.

So the phase noise of the analyzer is the first parameter you need to specify in order to find a suitable instrument.

Upper midrange oscilloscopes provide optional jitter measurement packages, which might be better suited for characterizing a clock signal.

[1Ghz]

If you prefer dedicated instruments and have enough space, old HP/Agilent 4352B VCO/PLL Signal Analyzer is good choice for the phase noise analysis.
It has nice phase noise performance such as -137 dBc/Hz at 10KHz offset typically.

Datasheet: https://www.atecorp.com/atecorp/media/pdfs/data-sheets/hp-agilent-4352b_specs.pdf

[nctnico]

Hi
Our main focus is quality of clocks to be used on high frequency data acquisitions.
We are developing a PTP driven distributed clock data acquisition system and are worried about the quality of the clock.

Define high frequency... if you need better than 10ns of time / frequency stability across your system, then forget about PTPv2 but look at White Rabbit. The latter will offer a time stability in the low tens of ps when looking at (relatively) long term (>10ms) time intervals on links that span tens of meters to mabe a couple of hundred meters.

Either way, the phase noise above 100Hz (the exact number depends on PTPv2 / White Rabbit control loop speed) will be dominated by the TCXO / OCXO that is used in combination with PTPv2 / White Rabbit.

For the longer term frequency stability a good (high resolution) time interval counter (I can recommend the Tektronix FCA3100 but the Keysight 53230A will also work) together with Timelab is good enough for making time / frequency stability measurements (Allan deviation) when using PTPv2. Note that many manufacturers for off-the-shelve PTPv2 equipment won't support accuracies below 100ns; they don't intend their equipment for such usage.

If you go the White Rabbit route, then a phase noise / frequency stability analyser is the tool to use in order to verify performance.

You'll likely want to use a synthesizer with a jitter cleaner to clock your ADC from.

All in all, a spectrum analyser is not the right tool for your purpose.

I strongly suggest to get solid answers to the following questions before going further:
- What is the frequency stability you need across the system between the acquisition nodes?
- What is the time accuracy (maximum difference in time) that can be allowed between the acquisition nodes?

PS: I have developed a distributed data acquisition a couple of years ago; White Rabbit was the way to go for that even with relatively low sampling rates of several tens of MHz. On top of that I have quite a bit of experience with developing PTPv2 / White Rabbit equipment.

[bicycleguy]

Is this correct for phase noise?
The SSA3021X+ can put a noise marker with the units (dBm/Hz) like pdenisowski info that is supposed to be normalized over 1 Hz.  Attached is a shot of the output of my homemade GPSDOCXO.  According to post by pdenisowski the phase noise should be (noise dBm/1Hz - Peak dBm).  So you have to do the math:
(-47.12-16.32)=-63.44dBm

[1Ghz]

If you go the White Rabbit route, then a phase noise / frequency stability analyser is the tool to use in order to verify performance.

Yes, you are right. They use Agilent E5052B Signal Source Analyzer, which is a successor to the 4352B VCO/PLL Signal Analyzer.
Recent White Rabbit presentation slide: https://ohwr.org/project/white-rabbit/wikis/uploads/c91f41a34bde79d0ce18dfddb2ec5c91/WR_OCP_2021-05-06.pdf

[Performa01]

Is this correct for phase noise?
The SSA3021X+ can put a noise marker with the units (dBm/Hz) like pdenisowski info that is supposed to be normalized over 1 Hz.  Attached is a shot of the output of my homemade GPSDOCXO.  According to post by pdenisowski the phase noise should be (noise dBm/1Hz - Peak dBm).  So you have to do the math:
(-47.12-16.32)=-63.44dBm

Yes, it is correct, except that you cannot get dBm when you subtract two absolute power levels: dBm - dBm = dB;
Since these are not random power levels, but an unwanted (noise) signal in relation to the single wanted signal (carrier), we would prefer to call it dBc (deciBell below Carrier) instead of just dB.

Furthermore, it would be unusual to specify the phase noise at a carrier distance at just 3 Hz of all things. 100 Hz, 1 kHz and 10 kHz are usually more relevant, but of course it depends on the application. Ideally, a complete noise density plot from 10 Hz up to at least 100 kHz distance from the carrier would be what we really want.

[1Ghz]

HP 8563E Spectrum Analyzer with the 85671A Phase Noise Utility.
250 MHz CW 0dB signal from the R&S SMJ100A Vector Signal Generator.

You can get good HP 856xE less than 3k USD.

[fisafisa]

Hi guys
Thanks for the help.
As usual a lot of reading and learning....

We do not need time stamping precision higher than 250ns,
but at the same time we want to acquire data from a few  thousands 2Mhz 18 bit ADCs located in about 20 cabinets spread  over a very large building.
The data acquired is sent over UDP network after downsampling (10kHz) and is afterwards recombined in some computers to be used in a distributed feedback system.
The system will operate continuously without resynchronisation.
What matters is that the ADCs produce the same rate of data otherwise the software will end up combining data belonging to very different time periods.(and diverging)
For this reason every source needs to operate at the same exact frequency: 2MHz / 200 and therefore require that the ADC clock be disciplined by a single central source (via PTP).
Data is also sent at full rate via a separate network to a large database, for higher frequency analysis. For this purpose we need to have decent time stamping.
At the same time we need a low enough jitter so that we achieve the ENOB of the ADCs.

So far the PTP packets jitter dos not seem to be a problem. The only issue appears the stability of the local oscillator of one of the systems (prototype) which appears to wander excessively. Tuning up the PTP algorithm to correct it will probably compromise the final clock jitter. We need to troubleshoot this.

Cheers

[nctnico]

What kind of local oscillator are you using? Realistically PTP only influences your phase noise in frequencies below 10Hz to 50Hz.

Does your hardware support SyncE? Using SyncE (which performs frequency transfer) would solve your problem. You could use PTP for a one-time time synchronisation and then let SyncE take care of frequency transfer; what SyncE does is to use the carrier for the ethernet signal (optical is the best way) to forward the clock to downstream devices. A downstream device can use a PLL to lock it's local oscillator to the ethernet carrier. The result is that you basically have the same clock across the entire system. However, temperature variations can cause slow drifts between the clocks but likely below 10ns.

[pdenisowski]

Yes, it is correct, except that you cannot get dBm when you subtract two absolute power levels: dBm - dBm = dB;
Since these are not random power levels, but an unwanted (noise) signal in relation to the single wanted signal (carrier), we would prefer to call it dBc (deciBell below Carrier) instead of just dB.

Yes, phase noise is always specified in dBc/Hz.

Furthermore, it would be unusual to specify the phase noise at a carrier distance at just 3 Hz of all things. 100 Hz, 1 kHz and 10 kHz are usually more relevant, but of course it depends on the application. Ideally, a complete noise density plot from 10 Hz up to at least 100 kHz distance from the carrier would be what we really want.

Uncommon, yes, but not unheard of    High-end instruments usually specify phase noise performance all the way down to 1 Hz, but as Performa01 says, the starting offset depends on the application and the measuring instrument.  Measuring close-in phase noise is both technically challenging and time consuming for a variety of reasons, and accurate measurements of close-in phase noise usually require more than an entry-level spectrum analyzer.

[cncjerry]

Until you spend a lot of money on a spectrum analyzer you aren't going to get accurate low level phase noise mes]asurement for even the most basic crystal oscillator.  Andrew Holme built an FPGA based system and using dual low noise references and coorelation that a number of us recreated that have noise floors approaching -185dB.  For instance, an HP8568b, a decent old SA with an 85685 preamp can maybe get to -122dB. A Wenzel ULN DOTCXO is under -165dB at 10k and under -175dB at 200k.  The old HP phase noise test system maybe hit -160dB noise floor.

If you are interested in measureing phase noise you need a dedicated system.  Agilent and others have papers that are readily available that discuss phase noise measurement. 

Jerry

[MegaVolt]

Andrew Holme built an FPGA based system and using dual low noise references and coorelation that a number of us recreated that have noise floors approaching -185dB.

Any link?

http://www.aholme.co.uk/PhaseNoise/Main.htm

[pdenisowski]

If you are interested in measureing phase noise you need a dedicated system.  Agilent and others have papers that are readily available that discuss phase noise measurement.

I would completely agree with that phase noise is one of those measurements (like EVM) that you really need a "good" instrument to measure.  You can measure a lot of things (IP3, P1dB, THD, OBW, NPR, etc.) using a relatively modest spec an and get reasonable results, but unless your DUT has absolutely terrible phase noise performance (and/or you're only measuring at large offsets), you can quickly find yourself in a situation where you're measuring the phase noise of your instrument.

Early in the thread I posted links to a few introductory videos on phase noise, but if you prefer whitepapers, here are links to the same content: 

https://www.rohde-schwarz.com/us/products/test-and-measurement/analyzers/signal-spectrum-analyzers/white-paper-understanding-phase-noise-fundamentals-register_255143.html

https://www.rohde-schwarz.com/us/products/test-and-measurement/analyzers/signal-spectrum-analyzers/white-paper-understanding-phase-noise-measurement-techniques-register_255398.html

[Joel_Dunsmore]

Value for money is tough to judge: but you did put a limit on the top end price so it lets out my favorite.  Not for nothin', and you didn't ask, but here's also a plot of phase noise measurements at 1 GHz using the PNA-X, which might also be the world's fastest spectrum analyzer"*  and you get a VNA too.
[ Specified attachment is not available ]

*for RBW<1kHz over spans of > 10 GHz or something like that.

[pdenisowski]

Well, since Joel is posting phase noise screenshots from Keysight's high-end spectrum analyzer, I'll post a link to my video showing you how to measure phase noise using the FSW --- Rohde and Schwarz's high-end spec an    It's got lots of screenshots in it ....



Did I mention that R&S has two dedicated phase noise analyzers as well?

https://www.rohde-schwarz.com/us/products/test-and-measurement/phase-noise-analyzers/rs-fswp-phase-noise-analyzer-and-vco-tester_63493-120512.html
https://www.rohde-schwarz.com/us/products/test-and-measurement/phase-noise-analyzers/rs-fspn-phase-noise-analyzer-and-vco-tester_63493-1087847.html

None of these are in the OP's price range, of course, but if we're going to "flex" (as the younger people say) .... 

[cncjerry]

Ha!  How much do those analyzers run, typical street price, not used?  I have about $1,500 in this system, maybe a little more, most of which in the reference oscillators.  In the attached chart (raw data, unsmoothed) the spike in the Orange (noise floor) is from the harmonic of the 26Mhz references and the 77.6Mhz ADC clock.  I have since purchased new reference oscillators that should move the harmonics out past the end of the chart.  The blue trace (real data) is the coorelated phase noise.  I used batteries to help eliminate the noise.  I was measuring a wenzel 10Mhz sprinter class oscillator (500 series).  I have others that go down another 5dB or so.

This is from Andrew's FPGA system.  I haven't added the reflectionless filters that seem to improve things a little and this is running the older code but still pretty impressive for the money.