Zero span spectrum analyzer .FFT, sweep, filters and RBW.

[G0HZU]

A lot depends on what you want (or expect) from your spectrum analyser. You might not need a high performance analyser for your task as I'm not convinced you 'need' a 1Hz digital RBW filter. The Siglent might be fine for what you want to do.

However, if it turns out that you really do 'need' lots of spurious free dynamic range (SFDR) whilst using a narrow digital RBW then I think that you will find that the performance of the Siglent is going to be fairly poor compared to a proper lab spectrum analyser.

The Siglent is going to lose out in both the performance of the downconverter (phase noise, IIP3 and DANL) and I'd also expect it to lose out even further if it can't provide a 1Hz (or maybe even a 10Hz) digital RBW filter when in swept or zero span mode.

It looks like it is limited to a minimum of a 30Hz digital RBW. This is still quite reasonable, but it does limit the SFDR in swept mode if you want to look for tiny signals near very large ones for example.

FFT mode can offer great speed advantages, but it comes at a price in terms of how it can limit the spurious free dynamic range. Swept mode is slower but it usually offers the best SFDR.

Having said all that, most users won't need the extra performance and they may be perfectly happy with the Siglent analyser. They may not be aware of its limitations compared to a proper lab analyser but this probably doesn't matter.

[G0HZU]

The other thing to be wary of with the cheaper analysers (like the Siglent) is that they are unlikely to offer much (if anything) in the way of analogue preselection ahead of the ADC when narrow spans are selected.

A decent lab analyser will have a suite of adjustable preselectors ahead of the ADC and the bandwidth of the preselector is usually adjusted to be a bit wider than the bandwidth of the digital RBW filter (down to maybe 1kHz).

This means the ADC is only exposed to a limited spectrum of signals when set to narrow spans or narrow RBW settings.

I'm not sure what the Siglent does here, but I expect it might only have one (or maybe two) IF preselection bandwidths. This limits the spurious free dynamic range and also makes the analyser very prone to ADC overload if there is a very large signal maybe 100kHz or maybe even several MHz away.

Many users will never realise this limitation because they probably don't ever explore the limits of the analyser during typical use. However, having a wide bandwidth digital IF can sometimes be a disadvantage if there is no means to reduce the bandwidth that the ADC is exposed to.
 

[G0HZU]

If this all sounds a bit confusing, then consider the case of FFT vs swept for a span of 500kHz.

In the swept case, the digital RBW could be 1kHz and there could also be a 3kHz wide preselector ahead of the ADC. So the ADC is only ever exposed to a spectrum 3kHz wide even though it is sweeping through a span of 500kHz. This protects the ADC from overload from large, nearby signals.

In FFT mode, the analyser may well present the whole digital IF bandwidth to the ADC and this could be several MHz wide. So there is much more chance of ADC overload either from a single large signal or from the voltage summing of all the signals in the IF bandwidth.

If the analyser does support an adjustable IF preselector, then in FFT mode, the preselector could be set to about 2MHz bandwidth or the analyser could select a narrower preselector and do several smaller (narrower) FFTs whilst stepping or hopping the LO each time. So the analyser could use a narrower preselector bandwidth and provide more protection of the ADC. This would sacrifice some of the speed advantage of the FFT mode though.

I don't know if the Siglent can do this multiple FFT trick, but it would ideally need to have a suite of IF preselectors fitted in order to make it truly worthwhile.

[G0HZU]

Note that for a 500kHz span in FFT mode, you can't choose a preselector that is also about 500kHz wide as there will be droop in the preselector band edges and this will spoil the flatness of the analyser across a 500kHz span. Usually, the preselector is selected to be several times wider than the span in FFT mode in order to preserve flatness.

If multiple FFTs are available, the the preselector bandwidth could be reduced significantly and maybe ten or more smaller FFTs could be done whilst retuning the LO for each FFT. Building up the spectrum in chunks like this obviously takes longer than doing a single FFT but it means the analyser could auto adjust the gain to prevent overload for some of the smaller FFT chunks if they contain huge signals in their passband.

[ftg]

Yeah I have hit the dynamic range and phase noise limits of SSA3032X on occasion.
And indeed it only offers 1Hz RBW in swept mode and 30Hz in Zero Span mode, which makes sense I guess, as 1Hz wide Zero Span would require rather good frequency stability on higher frequencies.

What measurement requires 1Hz RBW on Zero Span?
Especially on UHF and up?

At least on my SSA3032X the VBW does go down to 1Hz, so for long term monitoring of some signals output amplitude the RBW can be wider and then the VBW setting can be used to filter out all quick variations.

[G0HZU]

On narrow spans, the dynamic range of the Siglent analyser will be phase noise limited. So I guess one could conjure up scenarios where a large adjacent signal can spoil zero span mode (with a 30Hz digital RBW) through reciprocal mixing (noise).

However, this is going to be a rare event.

Overall, the difference between the Siglent analyser and a high end lab analyser is that the lab analyser is designed to be able to measure very small signals that are close to 'very' large signals. The Siglent analyser can't compete here. It can easily have 20 to 30dB less dynamic range compared to the lab analyser in some tests.

My interpretation of the Siglent datasheet is that it can't offer a digital RBW less than 30Hz in swept mode. Therefore, this doesn't just impact zero span mode.

[xugmu]

I deduce from your fantastic comments and images that it is not possible, not even with high-end analyzers, to get a stable "analog" RBW (sweep) in the 1Hz range. Therefore, everything that is done in that resolution range is done with digital technology (FFT) so it does not matter if, in that range, a manufacturer announces sweep or FFT.

The target to analyze are pilot OFDM carriers with a bandwidth of approximately 1kHz where there is quite a difference (especially in the waterfall) between 1Hz and 10Hz.

Getting a high SFDR or many preselector input filters is totally beyond my economic possibilities and, as you have said, buying a measurement device must depend on the needs.

Best  regards

[hpw]

As I see, you will get any better results using a PN noise measurement system. Additional using an accurate down mixer as here posted. Not to forget the AM/AN spurs as looking PM/FM O;)

The link here to the product, will show only single band.

https://qsl.net/bg6khc/pn2060c_phase_noise_analyzer.htm

My measurements as you like to do, as simple using a 10MHz reference 1Hz@-120dBc and comparing the SSA30xx showed clearly the side PN/AM of the internal synthesizer.

just my 2 cents

hp

[G0HZU]

I deduce from your fantastic comments and images that it is not possible, not even with high-end analyzers, to get a stable "analog" RBW (sweep) in the 1Hz range. Therefore, everything that is done in that resolution range is done with digital technology (FFT) so it does not matter if, in that range, a manufacturer announces sweep or FFT.

The target to analyze are pilot OFDM carriers with a bandwidth of approximately 1kHz where there is quite a difference (especially in the waterfall) between 1Hz and 10Hz.

Getting a high SFDR or many preselector input filters is totally beyond my economic possibilities and, as you have said, buying a measurement device must depend on the needs.

Best  regards

I'm still a bit unsure what you are trying to say here.

If you go back over 30 years, almost all spectrum analysers had analogue RBW filters built using LC resonators or with crystals. These have poor shape factor and the sweep speed has to be really slow for an analogue 10Hz RBW filter. A 1Hz RBW filter would be extremely slow. I've used analysers with a 3Hz analogue RBW filter and the sweep time is painfully slow.

In the 1990s, analysers started including digital RBW filter technology for the narrowest RBW filters. These offered much improved shape factor and faster sweep times for a given RBW.

The digital RBW filter is generated in the digital domain rather than the analogue domain. It's possible to have a digital RBW down to about 1Hz BW on a decent lab analyser. A digital RBW filter is usually fixed at the IF frequency and it works in conjunction with a swept LO, just like the analogue RBW filters except the digital RBW filter is synthesised in the digital domain.

This process doesn't involve FFTs. Digital RBW technology is different to the FFT method.

Later analysers had more advanced ADC and DSP technology and this allowed fairly large FFTs to be run over a fairly wide bandwidth.
In this case, the RBW is defined by the sample rate and the FFT size and the window function. This is very different to the way a digital RBW filter is generated in the digital domain.

I have very little experience with OFDM stuff. You may be better off with a decent SDR or RTSA although I'm still not sure what you are trying to do.

A digital RBW filter with 1Hz RBW will have a very slow response time to any changes in signal level and this can cause some confusion. For example, if the signal was instantly gated off, then the analyser might take several seconds to actually show that the signal has completely gone. Similar things (even more confusing?) will happen if you try and use FFT mode.

So I'm not sure what you want to see if you select zero span with a 1Hz RBW. The system will be very sluggish to report any changes in signal level to the point where I'm not sure how much use it would be in your case of an OFDM signal.

[G0HZU]

I think I can generate OFDM signals here using an old Agilent signal generator and some Agilent SW. This generates 802.11x test signals for wireless LAN. Is this similar to what you are looking at?

I've got various spectrum analysers here but rarely use them for looking at stuff like WLAN. I don't know much about WLAN. However, I could try and look for the same thing you want to see?

[xugmu]

It is logical that you do not understand my answer since my answer was wrong. I confused digital filter with fft.

The only way I have right now to get close to 1 Hz resolution is with an SDR. In fact, I opened the thread with that intention, to know what I see on the screen using 1 Hz RBW using fft technology or scanning technology. The same could be applied to the topic of zero span, the doubt was between owon and siglent, owon only offers fft technology at those resolutions and siglent thus offers two options.

Bringing the zero span to 1 Hz resolution has also been the result of my lack of awareness. In fact, seeing the images that have been posted in this thread, I have doubts about the use of a variable RBW when using the zero span and how that influences what is seen on the screen.

The signals that interest me the most are DVB-T signals (COFDM), specifically the pilot carriers, which are simpler. They interest me in two ways: interference can be appreciated relatively well since they have quite a bit of "free space" and on the other hand, observing the pilots in the time domain (a simple sinusoidal signal) may perhaps provide information (an almost abandoned objective would be to obtain the impulse response (my television analyzers only reach 400us))

Although I have demodulated  signals complete with gnuradio I have been trying to do the same for some time with an individual data carrier  with absolutely no success . The field of quadrature signals seems very interesting to me.

[G0HZU]

I don't know if this helps, but if we confine things to just the ancient analogue RBW filters and digital RBW filters (and we ignore FFT for now) then it's possible to predict the sweep time of the analyser for a given span setting and RBW.

The sweep time goes down as the RBW is increased and there should typically be a RBW^2 relationship. So if you increased the RBW by a factor of ten from 100Hz to 1kHz then the sweep time would be 10*10 = 100 times faster.

For an analogue RBW filter the sweep time in seconds is typically = Span/(0.33*RBW*RBW) where Span and RBW are in Hz.

For a digital RBW filter the sweep time in seconds is typically = Span/(0.8*RBW*RBW) where Span and RBW are in Hz.

This shows about a 2.5 times improvement in sweep time for a digital filter.

However, if you want to explore very narrow RBW filters then the sweep time goes up very rapidly because of the RBW^2 term in the equation.

A 1kHz Span and 10Hz RBW would have a sweep time of 12.5 seconds for a digital RBW filter. If you reduced the RBW to 1Hz the sweep time would go up to 1250 seconds. This is obviously impractical. If the span was reduced to just 100Hz then the sweep time would be 125 seconds and this is still really slow, but it is tolerable. But you only get to see a 100Hz span and not a 1kHz span.

The modern solution for this slow sweep time is to select FFT mode.

i.e. to improve on the 1250 second sweep time, a modern analyser can digitise the whole IF spectrum using a suitable sample rate and do a large FFT on the sampled data.

The display update rate would hugely improve and if you wanted a 1kHz span and a 1Hz RBW and you selected a typical window function (eg Blackmann-Harris window), the screen would update about once every two seconds with the FFT mode selected.


This is a massive improvement. However, the FFT mode comes with it's own compromises in terms of spurious free dynamic range.

Most analysers will be set up to automatically use digital RBW filters above a certain span setting and then swap across to FFT on narrower spans. This give the best spurious free dynamic range in wider spans (using digital RBW) and it gives the best sweep time on narrower spans (using FFT mode) where the sweep time of a digital RBW filter would be too slow.

[G0HZU]

Note that the really ancient HP 8568B and HP 8566B spectrum analysers could also support FFT analysis but only on the post detection data. I think they supported several window functions including flat top and hanning. It allowed modulation analysis at frequencies 'very' close to the carrier.

However, hardly anybody will have ever used this FFT feature in these ancient analysers.

[G0HZU]

Note that I don't recommend that you buy an HP 8568B or HP 8566B for your task. These analysers are too big and heavy and they produce lots of fan noise. They have old school analogue RBW filters and they don't support digital RBW.
The FFT mode they support is impressive for the age of the instrument, but the FFT performance will be fairly grim by modern standards.

You are probably best off using an SDR although the Siglent analyser might be OK for some of the things you want to do.