FFT Spectrum Analysis Reviewed

[Performa01]

Therefore the noise is coming from the TG and not the SA - is this correct?

No, quite – your test doesn’t tell the whole story.

If you lower the signal level by means of external attenuation, the SNR of the signal source remains unchanged. Now you decrease the reference level of the SA which means nothing else than increasing the IF gain by the same amount, thus pronouncing the SA mixer noise.

As we see almost the exact same picture as before, that means the noise from the SA frontend is negligible

The point is, when evaluating the noise at 3GHz, we were suspecting the LO and not the mixer to be the noise source. The LO signal is getting multiplied by the input signal, hence LO phase noise will not be visible with no input signal present. Consequently, you get a fairly clean noise floor with no input.

That also means the two test scenarios (with and without attenuation) don't change anything in terms of LO noise, because its visibility is closely related to the IF level, which remains the same.

[Performa01]

Without a proper 3GHz signal source with known (good) qualities, it’s hard to evaluate the SA44 phase noise performance at that frequency.

The only idea to get a little closer to the answer is to analyze the harmonics of a 1GHz signal. First the fundamental at 1GHz (SA44_RBW200Hz_REF-10dBm_1GHz_-20dBm_H1_avg8)



Ignoring the weak sidebands, the phase noise is almost down -80dBc and we would expect to be the harmonics of similar quality, just the frequency spans multiplied by the order of the harmonic.



We can see the 3rd harmonic at some -75dBm and noise floor is <-110dBm. That is pretty quiet even at 3GHz, despite the mixer still gets the -20dBm 1GHz signal, which once again confirms that the mixer is working very well. But we still don’t know about the LO.

The earlier tests show pretty strong phase noise up to some -48dBc, and this would still be hidden in the picture above. So let’s just increase the signal level by 20dB to 0dBm, thus supposedly being near the compression point of the mixer (SA44_RBW200Hz_REF-10dBm_1GHz_0dBm_H3_avg8)



Now that is revealing! The 3rd harmonic is now -40dBm, so it has increased by 35dB, that is 15 dB more than the fundamental, clearly hinting that we have indeed hit or exceeded the CP, thus slightly overloading the mixer and let it generate some harmonics by itself.

But then, it doesn’t really matter where the signal comes from. What we see on the display is the IF, which includes the LO signal whose level hasn’t changed, but is now more pronounced as the mixer product has increased.

And we can see indeed the phase noise starting to emerge. Now lowering the reference level to -20dBm should give an even more clear picture (SA44_RBW200Hz_REF-20dBm_1GHz_0dBm_H3_avg8)



Phase noise starts at about -70dBc, so it is much better than what we got from the earlier tests – and it is only 10dB worse than at the 1GHz fundamental. So it’s not too bad…

[uncle_bob]

Hi

One of the most confusing things about looking at phase noise on a spectrum analyzer is bandwidth normalization. Phase noise has a "nose density" in dbc / Hz. Change the bandwidth and you change the dbc. In some cases *knowing* the actual noise bandwidth of the filter used is the first challenge. You then get into some cute conversion issues in addition to the basic "10 * log(square root of the bandwidth)" calculation.

The net result is that I run a test on my analyzer and get -90 dbc. You run a test on your analyzer and get "only" -80 dbc. Bob obviously wins the race and you buy the beer Well, not so much. If Bob fiddled the bandwidth so his is narrower, he just cheated you out of a beer ...

Bob

(Thanks for the beer)

[ADT123]

One tip that can help when using scopes that offer large numbers of FFT points is to use the maximum number of points and a higher bandwidth than you want to view.  Next zoom in on the area of interest.  In these examples the signals I want to see are below 1MHz but by doing an FFT to 5MHz the higher frequency noise is not folded back into the area of interest so lowering the noise floor.

That's some excellent suggestion!
That's also a reason why it's nice to be able to use a high number of sample points for the FFT, so we can avoid aliasing and still preserve a low RBW.

PicoScope 4262 16 bit.  93dB

Well, your signal source appears to be very clean indeed - but are you really sure it is that good? You are measuring the 3rd harmonic of the test signal, which in theory could be generated by the ADC nonlinearity, but this is unlikely, since it is still within the genuine dynamic range of the 16-bit ADC and Pico Technologies specify <-95dB THD for this scope. So I would rather suspect it's coming from the signal source...

EDIT: Btw, I've just noticed, you have the internal signal generator on. I've found that this also produces some spurs, even when the siggen output is not connected. So for high sensitivity measurements it should be turned off.

I tried using the internal signal generator on the PicoScope 4262 and also a Blackstar low distortion audio signal generator.  Both gave very similar results and to be honest I can not remember which I used! 

It is not easy making such sensitive measurements - I dont know if the signal source or scope is the limit here. Good tip about making sure its turned off when not in use.   As mentioned even the quality of the BNC cable matters when you are measuring these levels.  Suspect if a cat farts in the next town it will get detected!

[Performa01]

(Thanks for the beer)

I’d gladly send you some good Austrian beer – even though you have cheated – but I’m afraid it would be pretty warm by the time it arrives at your place…

Well, the original topic of this thread was FFT, now we’ve finally arrived at phase noise measurement on various signal sources. It shouldn’t be too difficult if the phase noise specifications of the equipment involved is known. Sadly the specs of the SA44 are rather incomplete, to say the least.

But my signal generator is well specified - and I was able to measure it quite nicely with the 16bit scope, so let’s evaluate that. The 2nd screenshot in my 16bit FFT post shows that signal generator at 2MHz and we can see some -85dBc of phase noise for frequencies up to 150kHz from the carrier, at a resolution bandwidth of ~17kHz. So the phase noise reads some 42dB worse than with 1Hz RBW and we get a total phase noise of -127dBc/Hz.

And what does the specification say? It states -126dBc/Hz at any distance 20-500kHz from the carrier. Bingo! We now know (and have verified) the performance of the signal source.

The neat thing is, according to the specifications, this generator maintains the phase noise over its entire bandwidth (within a few dB tolerance), so we can expect a similar phase noise at 1GHz.

But we can stay at 2MHz anyway, as the first screenshot with the SA44 shows. At only 200Hz RBW we get a phase noise of some -85dBc again. Because of the 200Hz RBW we can add 23dB and end up with some -108dBc/Hz, which is 18dB worse than the generator signal alone.

But then, it’s not so uncommon that low end SAs (even from reputable brands) have such a high phase noise that they are pretty much useless for phase noise measurements.

[G0HZU]

The other option is to simply make a basic test oscillator at 4.4GHz to test the phase noise. The noise performance at 100kHz offset would only have to be 'fairly' decent to be better than the LO phase noise on all but the best spectrum analysers.

eg you should easily be able to make something that could deliver -120dBc/Hz phase noise at 100kHz offset at 4.4GHz using a basic free running oscillator with a printed resonator. Maybe half a dozen parts on the PCB in total?

[Ivan7enych]

The other option is to simply make a basic test oscillator at 4.4GHz to test the phase noise. The noise performance at 100kHz offset would only have to be 'fairly' decent to be better than the LO phase noise on all but the best spectrum analysers.

eg you should easily be able to make something that could deliver -120dBc/Hz phase noise at 100kHz offset at 4.4GHz using a basic free running oscillator with a printed resonator. Maybe half a dozen parts on the PCB in total?

Can you tell a bit more how to make this basic test oscillator ?

[G0HZU]

You can throw some basic physics at it and design for a resonator power of 10dBm and a device noise figure of 6dB and a loaded Q of 20. At 100kHz offset you should see about -120dBc/Hz phase noise if you put these numbers into a basic version of Leeson's equation.

In reality, the results may vary by about 5dB and a lot depends on if you can control/set the zero phase point around the loop at 4.4GHz to coincide with the peak in group delay in the resonator. Get this wrong and the phase noise will degrade a lot.

But the circuit would just be a cheap MMIC (50 ohm in/out) amplifier from MiniCircuits on a tiny piece of decent PCB material (Rogers 4003 or 4350) with a simple printed resonator shape between the input and output of the MMIC amplifier and a couple of coupling caps tapped into the resonator and also a SMD choke and resistor network to provide the supply voltage. The art is in getting the phasing correct and in preserving the loaded Q. The best way to do this is to model the circuit on a PC.

But the noise performance of a basic oscillator like this is quite easy to predict. The phase noise at 10kHz offset will be about 20dB higher than at 100kHz but there will be some additional flicker noise from the active device. So you might see -97dBc/Hz at 10kHz offset and -120dBc/Hz at 100kHz offset from such an oscillator at 4.4GHz. But you would need a very good spectrum analyser to be able to measure the noise performance at 100kHz offset.

[hendorog]

In the spirit of uncle_bob, lets cheat:

I have one of these MLSL-0406IC which I got from eBay a couple of years ago.
http://www.admiral-microwaves.co.uk/pdf/micro-lambda/MLSL_IC_SeriesSheet_2GHz.pdf

The specs seem to match up with those in G0HZU's post.
10kHz: -98 dBc/Hz vs -97
100kHz: -122 dBc/Hz vs -120

Here are some plots at about 6GHz from it - (I don't have anything setup to change the frequency of it right now so its stuck at ~6Ghz)

So I guess either the SH cannot display the PN that the synthesizer is capable of, or the synth isn't working to spec?

[uncle_bob]

Hi

Ok, so if spectrum analyzers sometimes do and sometimes do not show you the real phase noise .... how do you settle the bet? After all, there *is* beer being drunk and *somebody* must pay for it all

One fairly easy way is to do a quadrature phase noise setup. You take two identical oscillators and run them into a mixer. Lock them up so the DC output of the mixer is zero volts. Net result... that messy carrier is now gone (actually it's DC, but that's not part of the bet). The noise on the oscillators has been folded together so you have both sidebands of both oscillators coming out the IF port of the mixer.

Next step is to calibrate the gizmo. It's a phase detector in this case, break lock and you have a nice predictable sine wave with 360 degrees phase coming out. Work out the degrees per volt around zero volts with a scope. You have your calibration factor.

Lock it back up and take a look at the output with a *low frequency* spectrum analyzer. You no longer need super duper dynamic range. You just need a really good noise floor.

Now you can prove what the noise is and get that slacker Bob guy to buy all the beer ...

Bob

(quietly leaving the bar by a side door ... burp ...)

[tautech]

Just love threads like this......thanks for sharing the knowledge guys. 
Subscribed.

[G0HZU]

At work I'm lucky to have access to some nice RF test gear and I grabbed one of the E5052A SSAs today and measured a homebrew version of the 4.4GHz oscillator outlined above.

http://www.keysight.com/en/pd-409739-pn-E5052A/signal-source-analyzer-10-mhz-to-7-265-or-110-ghz?cc=US&lc=eng


I used a MiniCircuits GALI -49 as the amplifier and the PCB was built on a tiny bit of Rogers 4003C 0.02" thick. It's not quite on 4.4GHz so would have to be trimmed down a bit to get inside the 4.4GHz limit of the Signalhound analyser but see below for a phase noise plot of this oscillator.

You can see that it agrees very well with theory! However, I did have to bias the GALI 49 at 80mA to get the phase noise this good. It was at -117dBc/Hz at 100kHz offset with 50mA bias. There are often hidden factors like resonator losses that degrade the phase noise a bit. So usually the result comes out slightly worse than the prediction using a very simple version of Leeson's equation. Close enough though!

I can post up pics of the PCB and the initial design simulation if anyone wants to see them?

[Performa01]

Uncle Bob (the one who likes beer so much) mentioned noise bandwidth of the filters – and now I have to admit that I was a bit sloppy about that when evaluating the phase noise of my signal generator. Back then, I just took the -60dB bandwidth, which clearly is much wider than the actual noise bandwidth.

So how do we know what the noise bandwidth in our FFT analysis is?

FFT can be viewed as being equivalent to a number of bandpass filters, that is half the number of sample points. These filters obviously are pretty steep and from the textbook we know that in this case the noise bandwidth will roughly be the same as the -3dB bandwidth.

So what’s the actual -3dB bandwidth in our FFT then?

This depends on the window function we’ve chosen. Pico Technology was kind enough to show a table of all available window functions in their user manual, so I’ll repeat the relevant information here:

Window	Rel. BW	Side Lobe
Blackman	1.68	-58dB
Gaussian	1.33 ~ 1.79	-42 ~ -69dB
Triangular (Barlett, Fejer)	1.28	-27dB
Hamming	1.30	-41.9dB
von Hann	1.20 ~ 1.86	-23 ~ -47dB
Blackman-Harris	1.90	-92dB
Flat-top	2.94	-44dB
Rectangular	0.89	-13.2dB

As can be seen from the table above, Blackman-Harris is the general purpose window, particularly if a high dynamic range is required, where it is the only window function with a reasonable side lobe suppression (apart from the Kaiser-Bessel window with high values for the Bessel parameter [beta], which is not available in the Pico software).

Most of these window functions have their specific uses because of specific merits, but due to their poor side lobe suppression cannot be recommended for any applications where dynamic range is of importance. This is particularly true for rectangular and triangular, as they create a lot of noise even in an 8 bit system. Rectangular has the narrowest bandwidth and maximal sharpness of all window functions though.

The performance of Gaussian and von Hann windows depends on window-specific parameters, which aren’t specified in the Pico documentation. This is most likely true with other scope manufacturers as well, where we can select these window functions from a menu but don’t actually know what we get. The table above just lists the performance for the usual range of these parameters. Consequently, these window functions should be avoided unless someone actually evaluates their specific properties on the scope in question beforehand.

It is clear, we should use Blackman-Harris for the phase noise measurement, but first let’s check whether the -3dB bandwidth of 1.9 times the bin width as specified in the table is plausible. The following screenshot shows both the -3dB and -60dB bandwidth by cursor measurement on the high frequency side of the filter curve. This is the zoomed picture from a 64k FFT at 5MHz bandwidth and 152.6Hz bin-width (FFT16_64k_2MHz_+3dBm_avg_zoom)



The automatic measurements confirm that the signal is indeed 2MHz at +3dBm. The -3dB point is at 2.000152MHz, hinting on a total -3dB bandwidth of 2 x 152Hz = 304Hz. Using the factor 1.9 from the table, we get 152.6 x 1.9 = 290Hz. Well, close enough. The difference is most likely due to a slight asymmetry in the top of the filter curve.
The -60dB point is at 2.0005631MHz, hence -60dB bandwidth is 2 x 563Hz = 1126Hz. That’s pretty steep indeed, but is still some 3.75 times the noise bandwidth, thus introducing almost 6dB error if this is used for noise calculation.

I will use 300Hz as the noise bandwidth for the following measurement.

Here’s the 64k point 16 bit FFT of a 2MHz signal from the analog synthesizer at +4dBm, slightly zoomed to show the range 1MHz to 3MHz (FFT16_64k_2MHz_+4dBm_avg)



I’m measuring the point of strongest phase noise here, which is at about 180kHz distance from the carrier. At only 20kHz distance, it would be some 4dB better.

Worst case phase noise at that frequency is -96.2dBc for 300Hz bandwidth. Normalized to 1Hz we can add 10 log(300) = 24.8dB, so we get a phase noise of -121dBc/Hz.

The specification of the synthesizer says -120 to -126dBc for the entire frequency range and the phase noise curve is only given for 480MHz, where it is indeed a flat line at -125dBc from 20kHz to some 500kHz distance from the carrier. So at least at the lower end of the frequency range, phase noise is 4dB higher than that, which is still low – certainly lower than what most ‘real’ spectrum analyzers can measure.

[G0HZU]

The specification of the synthesizer says -120 to -126dBc for the entire frequency range and the phase noise curve is only given for 480MHz, where it is indeed a flat line at -125dBc from 20kHz to some 500kHz distance from the carrier. So at least at the lower end of the frequency range, phase noise is 4dB higher than that, which is still low – certainly lower than what most ‘real’ spectrum analyzers can measure.

At a wanted carrier frequency of 2MHz that level of noise performance is very poor for a signal generator. There are much better generators available than that (for 2MHz carrier frequency) or you could make something using junk box parts.

I'd expect a fairly basic/cheap homemade 2MHz LC oscillator to achieve:

>-130dBc/Hz at 1kHz offset
>-155dBc/Hz at 10kHz offset
>-175dBc/Hz at 100kHz offset

Or you could try testing a 10MHz crystal oscillator divided by 5 to get 2MHz. This should give better than -150dBc/Hz at 10-100kHz offsets. eg use a 74HC390 to do the divide by 5.

certainly lower than what most ‘real’ spectrum analyzers can measure

Down at these test frequencies of a few MHz you can use a regular lab (microwave) spectrum analyser and a crystal filter to allow simple phase noise measurements down at -140dBc/Hz at just a few kHz offset.

Again, it is fairly simple to design/build a 2MHz crystal filter that is a few kHz wide

[G0HZU]

When I first bought my E4406A analyser I experimented with direct injection at the 14 bit ADC stage and was able to get -136dBc/Hz phase noise at just a few kHz from the carrier.

https://www.eevblog.com/forum/testgear/agilent-e4406a-vector-signal-analyser/

See the image below.
This is set at 15dB/div and not 10dB/div. Just to clarify, I have bypassed the RF converter stage in the E4406A here and I am just sending a signal into the ADC stage. So the phase noise performance will presumably be limited to about -136dBc/Hz by the 14bit ADC and the DSP after it.

Note that the screen says Err in red (Error) because I've broken (bypassed) the RF signal path inside the analyser and so it has failed one of its automated signal path health checks in the RF converter section. The error goes away once I restore the full RF signal path to the ADC.

[uncle_bob]

Hi

Another way to come up with the noise bandwidth is to put in a known level of noise into the analyzer. There an large number of ways to do this. They fall into a couple of groups:

1) Use something like a noise diode and possibly an amplifier with known gain.

2) Modulate a signal with a noise source

Both have their limitations.

The first one relies on the noise diode being correct (calibrated) and the amp (if used) having known gain when noise loaded. You still have to calibrate all that against a reference into the analyzer. You then will start asking questions about the flatness of your SA's response. I know this because I use this approach to check how far out some of my gear is. It's always a disappointing thing to check. A couple of db is not at all unusual and that messes this up for really wide measurements.

The second one appears to have all the problems of the first plus some of it's own. If you do it with analog gear, that pretty much is the case. You need all the calibrated noise stuff with flat response *plus* your generator needs to have a really good modulator. The way that's normally gotten around is to generate the signal digitally. That gets you right back to the same issues we have been looking at on ADC's except on DAC's. In most cases you can do a pretty good job. You then do a sweep run over the full bandwidth and ... yes indeed ... there's a pesky db or three hiding in the signal chain.

Could you go crazy an take out the flatness? Sure you could. It's normally easier to use this as a narrow(er) band approach and avoid the issue. Because of all the details, it's still not a great way to settle a beer bet ... (I had to get that in here somehow ..). The bet in this case would be:

Does my analyzer "throw away" noise with it's FFT windowing process or does it all show up in a bucket somewhere somehow?

Hint: This is why Matlab exists ...

Bob

[Performa01]

The other option is to simply make a basic test oscillator at 4.4GHz to test the phase noise. The noise performance at 100kHz offset would only have to be 'fairly' decent to be better than the LO phase noise on all but the best spectrum analysers.

eg you should easily be able to make something that could deliver -120dBc/Hz phase noise at 100kHz offset at 4.4GHz using a basic free running oscillator with a printed resonator. Maybe half a dozen parts on the PCB in total?

Well, that’s a good suggestion and I’d certainly give it a try one day.

Last time I’ve produced my own homemade PCBs has been more than 20 years ago and I guess the equipment is crusty by now and my stock of photoresist boards would be long dead. They are just FR4 anyway. So I’d have to order new photoresist boards and chemistry first at least.

In short, right now it would be far too much hassle tinkering a PCB - and ordering a professional made PCB just for prototyping a circuit with a hand full of components appears odd (and expensive).

The awful aspect of GHz circuits is the fact that we cannot just botch something together on a veroboard – all the more so if we want to use stripline resonators.

But then, for evaluating an SA, I don’t need a super clean oscillator anyway and I’m pretty sure I can get away with what I have on hand. Might post some more results later…

[Performa01]

At work I'm lucky to have access to some nice RF test gear and I grabbed one of the E5052A SSAs today and measured a homebrew version of the 4.4GHz oscillator outlined above.

...

I can post up pics of the PCB and the initial design simulation if anyone wants to see them?

That’s an impressive plot from an impressive instrument!

Sure I’d be interested in your design and particularly the comparison of simulation vs the real deal would be very instructive for everyone.

[Performa01]

At a wanted carrier frequency of 2MHz that level of noise performance is very poor for a signal generator. There are much better generators available than that (for 2MHz carrier frequency) or you could make something using junk box parts.

Well, you’re certainly right that -121dBc at 2MHz is poor and there are indeed better ones available – and worse ones as well. The -125dBc at 500MHz or 1GHz is a different story though.

The whole point is, as the SA fails to show a phase noise even remotely close to these figures, the signal generator is certainly good enough for these tests.

Down at these test frequencies of a few MHz you can use a regular lab (microwave) spectrum analyser and a crystal filter to allow simple phase noise measurements down at -140dBc/Hz at just a few kHz offset.

Again, it is fairly simple to design/build a 2MHz crystal filter that is a few kHz wide

Yes, of course that’s always an option. It’s basically what we have to do every now and then (way too often) – use some external devices or even tinker something together in order to overcome some limitation in a test instrument.

[Performa01]

2MHz Phase Noise SA44
I’d like to show some closer inspection on the phase noise of the SA44.

There are strong hints on local oscillator PLL phase noise limiting the dynamic range in closer proximity of the carrier. But the SA also uses the flat-top FFT window, which gives the best amplitude accuracy but isn’t particularly great in terms of side lobe suppression.

So let’s start over again with the 2MHz signal at -15dBm, whose phase noise never exceeds -121dBc/Hz at distances >10kHz from the carrier, as we have verified by now. Span is 500kHz and RBW 100Hz and we get the familiar picture (SA44_RBW100Hz_2MHz_-15dBm_avg4)



Phase noise is some -88dBc at 100Hz RBW and this would roughly mean -108dBc/Hz. Once again the abrupt step some 130kHz to the right from the carrier is striking. Of course this exists to the left as well, just not visible in that image above.

Now this SA has a nice utility function to produce a phase noise plot: In order to use this, the span has to be lowered to a maximum of 10kHz first. So I set a 4kHz span and 6.5Hz RBW (SA44_RBW6Hz_2MHz_-15dBm_avg8)



At that setting, we now clearly see all the sidebands from residual AM/FM, but also the phase noise which drops below -100dBc at about 1.3kHz distance from the carrier. So at least with very narrow resolution bandwidths, this SA is capable of providing >100dB dynamic range, but the phase noise from the signal source drops accordingly, so this doesn’t help with phase noise measurements

Now let’s start the plot (SA44_PhaseNoisePlot_2MHz_-15dBm_avg8)



The phase noise plot shows nothing we could not already estimate from the first measurement at 500kHz span and 100Hz RBW – the phase noise at 200kHz distance from the carrier is displayed at about -109dBc/Hz, which is some 12dB worse than what the real value would be.

Once again there’s a striking step at 200kHz distance from the carrier, where the displayed phase noise suddenly drops below -120dBc/Hz. It looks like we could indeed get some meaningful measurements at distances >200kHz from the carrier. But usually we are particularly interested in the range <100kHz, so this just means we might at least be able to measure the general wideband noise level of any signal source with reasonable accuracy.

Below 3kHz, the phase noise increases a lot and it would be interesting to know whether it’s the signal generator or the SA44 dominating here. So looking at some FFT16 results again in order to determine the respective phase noise figures (FFT16_512k_2MHz_+3dBm_avg_zoom)



Assuming 36.23Hz noise bandwidth from the 256k points FFT, we have to add 15.6dB to the measurements to get the phase noise normalized. The following table compares the normalized noise levels from the FFT16 measurements and the SA44 noise plot:

Frequency	FFT16 Phase Noise	SA44 Phase Noise
100Hz	-61dBc/Hz	-56dBc/Hz
200Hz	-71dBc/Hz	-67dBc/Hz
500Hz	-88dBc/Hz	-88dBc/Hz
1kHz	-103dBc/Hz	-101dBc/Hz
2kHz	-114dBc/Hz	-110dBc/Hz
5kHz	-120dBc/Hz	-113dBc/Hz

It appears the SA44 is not so bad here, giving results only slightly worse than the FFT16 measurements and being even in agreement at 500Hz.

[Performa01]

1-4GHz Phase Noise SA44
I’ve already shown the spectrum of a 1GHz signal, just for completeness here it is once again, this time at a level of -40dBm with preamplifier on and 100Hz RBW (SA44_RBW100Hz_1GHz_-40dBm_P_avg8)



Worst case phase noise is about -80dBc at 100Hz RBW or -100dBc/Hz.

The phase noise plot looks slightly worse than the picture above, but close enough in my book (SA44_PhaseNoisePlot_1GHz_-30dBm_P_avg8)



Now the same measurements at 3GHz (SA44_RBW100Hz_3GHz_-40dBm_P_avg8)



Phase noise plot at 3GHz (SA44_PhaseNoisePlot_3GHz_-40dBm_P_avg8)



Phase noise at 3GHz is about 10dB worse than at 1GHz. So let’s finally look at 4GHz (SA44_RBW100Hz_4GHz_-40dBm_P_avg8)



…and the phase noise plot (SA44_PhaseNoisePlot_4GHz_-40dBm_P_avg8)



Once again, the phase noise got a little worse, but only by some 2dB.

When comparing this with the noise plot shown in reply #36 by G0HZU, it appears that my test setup performs equally well between 300Hz and 3kHz – and even better below 300Hz. Above 3kHz, the combination of free-running oscillator and Agilent signal analyzer pulls away…

I guess all this finally confirms that it’s the LO noise we’re seeing, as FFT noise wouldn’t depend on the signal frequency. Still not too bad and very usable in my book – especially for the price.

Particularly, if compared to the Agilent E7495B as seen in this video at 17:23:



I tried to resemble the settings: 1.35GHz signal at -10dBm, 100MHz span, 100kHz RBW (SA44_RBW100Hz_1350MHz_-10dBm_P_avg8)



The SNR is about 73dB, whereas it is only some 65dB on the E7495B. Of course, following uncle Bob’s best practice, I’ve cheated a bit. The reference level is set at 0dBm because there really is no need for 30dB headroom above the signal. This means less internal attenuation (which is unknown on the E7495B because it doesn’t show) and lower noise floor. Also the vertical scale was left at 10dB/div, which doesn’t change the SNR but makes it look better visually.

[G0HZU]

See below for a screenshot of the basic simulation for the 4.4GHz oscillator. This is just an 'old school' small signal open loop analysis and there are better ways to do this.

But the goal here was just to make something that worked. I know from experience that the oscillation frequency is always a few percent lower then the small signal prediction so it was design centred for 4.6GHz. In reality, it oscillates at 4.45GHz.

The simulation is done using Agilent Genesys and the Sonnet EM engine. The PCB layout is simulated along with the components and the simulation helps optimise the layout to make sure the zero phase point aligns with the peak in loaded Q (or peak in group delay)

See also a link to a youtube video showing the short term stability on my old HP8566B spectrum analyser. The phase noise on this analyser is fairly similar to the oscillator so it can't measure the phase noise as accurately as the E5052A analyser at work. Also I turned down the bias to 60mA to improve the short term stability. The oscillator needs to be kept away from changing air currents and vibration so I've put a plastic cover over it during the stability test.

But you can see how stable it is. The CF is 4.455GHz and the span is only 100kHz and the oscillator only drifts a few kHz in a minute. The expensive PCB material helps here and I also used ATC 600S caps.

The video is a bit dull as it just shows the analyser display for about a minute but then the camera pans and zooms to show the oscillator at the end of the video. You can see the oscillator jump around in frequency as I operate the pan and zoom on the camera. It only takes the slightest bit of vibration or changes in air current/temperature to cause the oscillator to jump around. It really needs to be in a metal/screened box...

https://youtu.be/PMZyMLC2swI

[uncle_bob]

The video is a bit dull as it just shows the analyser display for about a minute but then the camera pans and zooms to show the oscillator at the end of the video. You can see the oscillator jump around in frequency as I operate the pan and zoom on the camera. It only takes the slightest bit of vibration or changes in air current/temperature to cause the oscillator to jump around. It really needs to be in a metal/screened box...

https://youtu.be/PMZyMLC2swI

Hi

A (dry) bar towel works wonders for slowing down the air induced wander in an oscillator breadboard ....If none is available, a spare ESD lab smock can be substituted.

Bob

[KE5FX]

See below for a screenshot of the basic simulation for the 4.4GHz oscillator. This is just an 'old school' small signal open loop analysis and there are better ways to do this.

This is extremely cool.  I wouldn't have guessed you could get -96 at 10 kHz out of a free-running oscillator at 4 GHz like this without going to a lot more trouble.  For instance, I'm surprised that the surface roughness of the solder on your resonator isn't degrading it quite a bit.

[G0HZU]

See below for a screenshot of the basic simulation for the 4.4GHz oscillator. This is just an 'old school' small signal open loop analysis and there are better ways to do this.

This is extremely cool.  I wouldn't have guessed you could get -96 at 10 kHz out of a free-running oscillator at 4 GHz like this without going to a lot more trouble.  For instance, I'm surprised that the surface roughness of the solder on your resonator isn't degrading it quite a bit.

Yes, you can see that the PCB has seen a bit of action with the soldering iron. When it was at work in front of the E5052A I couldn't resist experimenting with different caps and also I rotated the caps slightly to try with different tap points into the resonator. So what you see is the aftermath of this. You can also see in places how flat the solder 'was' before I did all this because it was originally flattened with solder wick.

But you can still see that the noise on the HP8566B is still in the right ballpark. The HP8566B has similar/worse phase noise than the oscillator up here.

For instance, I'm surprised that the surface roughness of the solder on your resonator isn't degrading it quite a bit.

You can see in my image that I've used desoldering braid to keep it looking tidy but this was before I messed with the caps at work.
What I can say from experience is that any attempt to use sticky copper tape instead of the original PCB copper for the resonator does not work well at all.

I've got a similar oscillator design somewhere here that runs at 6.7GHz and it had similar phase noise performance to the one at 4.4GHz. I've built quite a few oscillators like this and these were used (for example) to test the noise performance of the first few ranges on my HP8566B analyser. The main aim was to make something quick, simple and stable with similar phase noise to the HP8566B.