Spectrum Analyzers comparison: Rigol DSA815 & Signal Hound SA44B. Input needed.

[Simon Loell]

I've searched a lot for a comparison on the Rigol DSA815 and Signal Hound SA44B spectrum analyzers, but haven't yet found any.
People only look at, and compare the specs and mostly they only have experience with the Rigol.

Now I think time is up for a comparison, a Signal Hound vs. Rigol.
So I purchased a Signal Hound for test, in mind I can return it for 30 days. http://www.signalhound.eu/SA44B.htm
I got my hands on a Rigol DSA815-TG as a demo. http://www.rigol.com/prodserv/DSA800/
So within a day or 2 I will have the possibility to test them side by side.
I might be able to have a 18GHz Rhode & Schwarz SA reference, if needed?

Since I'm not that much used to spectrum analyzers and my need for one is mainly from DC-converters, switch noise to 433MHz radios and a few 1.2GHz radios I need your help to test as many things as possible.

I will take pictures and if wanted make a short video from different setup.

What do you want me to test for you?

-Simon
www.nerdclub.dk

[videobruce]

Television signals. ATSC and/or QAM. Or whatever is available where you live.
Side by side using the same feed (w/ a splitter)

[olsenn]

Since I'm not that much used to spectrum analyzers and my need for one is mainly from DC-converters, switch noise to 433MHz radios and a few 1.2GHz radios I need your help to test as many things as possible.

This one just spells out Rigol. None of these applications requires frequency measurements above the 1.5GHz limit of the DSA815, and needing higher frequencies would be the only reason to prefer the SignalHound to the Rigol. The DSA815 is MUCH faster, has lower RBW, and is AC coupled to provide 50VDC input protection which may come in handy for your DC converters.

[KJDS]

Since I'm not that much used to spectrum analyzers and my need for one is mainly from DC-converters, switch noise to 433MHz radios and a few 1.2GHz radios I need your help to test as many things as possible.

This one just spells out Rigol. None of these applications requires frequency measurements above the 1.5GHz limit of the DSA815, and needing higher frequencies would be the only reason to prefer the SignalHound to the Rigol. The DSA815 is MUCH faster, has lower RBW, and is AC coupled to provide 50VDC input protection which may come in handy for your DC converters.

A quick look at the datasheets shows that the minimum RBW of the Signal hound is 0.1Hz and the smallest for the Rigol is 100Hz. Am I missing something?

[Simon Loell]

@olsenn:  The Signal hound price is still a bit better. Besides that I would like to measure harmonic from both 433MHz and 1G2 onto 2G4.

[olsenn]

A quick look at the datasheets shows that the minimum RBW of the Signal hound is 0.1Hz and the smallest for the Rigol is 100Hz. Am I missing something?

Sorry; I could have sworn when I looked up the Signal Hound its RBW was higher; the DSA815 is definitely 100Hz. Perhaps it was the noise floor that I was wasn't impressed with. I definitely know that the Signal Hound's sweep times are outrageous compared to the Rigol.

As for price, I think the DSA815-TG is cheap enough for what you get to justify the cost. To each their own though

[casinada]

There is no much dept on the review but it has other models I didn't know:
http://www.tmworld.com/design/characterization/4412164/2/EMC-troubleshooting-roundup--Low-cost-spectrum-analyzers

[Andy2]

Just dragging up an old thread to add my thoughts. I have a Signal hound SA44b+TG here and I find it has much lower close-in phase noise than my Rigol 1030. It's ideal for sweeping narrow filters down to very low levels, where the Rigol struggles with lots of clutter around the LO.
The only problem(s) with the Hound is the unstable software, especially when using it is 'high res' mode. It frequently crashes and requires a complete re-start, and of course it always does this while I'm in the middle of something important! Users were promised a 'complete re-write' a couple of years ago, but it's all gone quiet.

[videobruce]

That is a huge surprise. I wonder if the 1030A is better?

[Andy2]

Sorry Bruce, I should have said that mine is a 1030A-TG. The difference between the SH and the Rigol is huge. I suppose it's because the Rigol uses standard sweep techniques across its range of spans and RBW's, whereas the SH switches into SDR mode for narrow modes. At least I think I'm correct? It even has a high res mode for sweeping crystal filters etc, and in this mode it really does get down to the dirt. Trouble is it can take all morning and it crashes whenever the wind changes direction.

There's no doubt that the Rigol is the 'better' instrument for speed, ergonomics and reliability, but when the Hound is working properly it puts in a hell of a performance, especially for narrow sweeps.

Both photos show sigs from a Marconi 2024 sig gen at 100MHz, -10dBm.

[Mr Simpleton]

Having several spectrum analyzers including the SA44B, I find myself using the Hound more than I expected at first. It's size and portability is perfect, and it uses the USB for power.

Now one thing I still have problems with is that the RBW is somewhat related to FFT bins rather than 1:3:10 that we are used to.
But honestly it's more a mental thing than limiting the usage.

Sweep times are not too bad most of the time, but wide sweeps cand take longer time than one is used to. The Hound have a 250 kHz wide IF that is digitalized, and narrow sweeps are quite zippy

Phase noise is not too bad, and the phase noise measurement utility quite nice, and so is the harmonic measurement.

Sensistivity is great, but the attenuator seems to be distributed in the signal chain rather than at the front as sensistivity can go down with less attenuation !

Saving plots, changing colour of traces, having several delta cursors makes it really simple for documentation. No need to save trace first, just use the clip tool in windows and paste into your document.

[videobruce]

Sorry Bruce, I should have said that mine is a 1030A-TG. The difference between the SH and the Rigol is huge....Trouble is it can take all morning and it crashes whenever the wind changes direction.

The 2nd part is the deal breaker. Way too high a price to pay. I don't understand how something that small can outproform a full size scope. yes, I know the Laptop (or PC) is doing to 'computing', but the front end still is inside that small case.

I'm more interested in your 1030A. How long have you had it and what did you have before that? Can you make comparisons between this and anything else you have had experience with (other that the SH)?

[Andy2]

Before the SH I had an HP 8568b, but it had to go due to a major down-sizing period. It was also loud, big and hot. Oh, and the graticule was dim and getting dimmer... The SH was purchased as a replacement, and I found the phase noise of the SH was at least in the same ballpark as the HP. However, the longer I used the SH the more frustrated I grew with its poor ergonomics and tendency to crash, especially when I was using it in the slow HiRes mode.
If you have the patience of Job himself and good anger management, the HR mode can produce some truly impressive results when sweeping narrow filters (in the range of hundreds of Hertz or low kHz width). It does not display a trace until it has finished, and during that time (yawn) it gives no indication of what's happening. It may have crashed, it probably has, and any attempt to abort the sweep will 90% cause a crash. Some crash conditions can be recovered by simply closing and re-starting the SH program, others require an unplug of the USB leads. I finally snapped and spent the kids' inheritance on the 1030 when the promised 'complete re-write' of the Hound's software never came.
The 1030A-TG has been with me for about one year and generally I'm happy with it. It is ergonomically good and it has never done anything unexpected. It has many useful features and they are easy to access. Only that noisy LO lets it down.
Hope this helps!

[Mr Simpleton]

Andy,
you mentions that you had a HP 8568 before... and that you "somewhat frustrated" when using your Hound with the TG in high res mode. But did you use a TG for your 8568??

I find the Hound to operate OK most of the time if I only use spectrum analysis, and I do have a VNWA when I need to sweep filters up to 1 GHz, so very seldom use my TG for the SA44B.

As for phase noise, the HP 8568 is a lot better that the Hound, especially close to the carrier.. seeing the noise of the Rigols was the main factor for me to stay away and get a used HP8560E instead. The 8568B is my preferred SA in my work lab, where I do have the space.

Lets hope TEP do get their act together and fix the bugs and maybe polish up the UI... would really nice. As for now I cannot really justify a BB60C given their track record 

[Andy2]

No tracking gen with the 8568, I just let it run and manually stepped up (asynchronously) on a separate sig gen. This could take several minutes, but at least I could see the result building on the screen and nothing ever crashed. I would have loved to keep the HP, but there was only so much room in the new place. It went to a good home, BTW.
The SH could be really nice if the bugs were driven out and better use made of the widescreen format of modern laptops. By using the whole of the screen width, all the controls could be accessible at once without having to scroll up & down. This may sound a minor quibble, but constantly having to grab the mouse to find the TG controls quickly gets tedious.

[videobruce]

It was also loud, big and hot.

You forgot heavy & a power hog. 

It does not display a trace until it has finished, and during that time (yawn) it gives no indication of what's happening. It may have crashed, it probably has, and any attempt to abort the sweep will 90% cause a crash.

You can have it, all of that makes it useless AFAIC. 

[videobruce]

Mr Simpleton;
Of the recent SA's (10 years and newer) under $8k (used) which ones have you found acceptable regards to phase noise?

[Andy2]

OK, a final post on the phase noise issue from me (sorry I seem to have this particular bee in my bonnet at the moment). Below are shots from the Rigol and the Hound, both displaying two signals.
The sig on 10MHz is from the ref osc in my Racal frequency counter. The sig on 10.1MHz is from my Marconi sig gen.  On the Signal Hound it is obvious which is the cleaner device re phase noise, and it even shows the dip caused by the Marconi's PLL loop filtering. On the Rigol the analyser's own phase noise completely dominates by at least 20dB and the two traces look identical.
So the Hound seems to be far better for narrow stuff. If only TEP would sort it out!

[Mr Simpleton]

Mr Simpleton;
Of the recent SA's (10 years and newer) under $8k (used) which ones have you found acceptable regards to phase noise?

Well, I haven't really been through that many SA's At the salt mine, the 8568B has been my work horse, but we do have a few low end R&S wich are not as good, higher end R&S are supposed to offer even better pn than any new Agilent but the 8k will not be enough, 80k more likely.

New for less than 8k will not really give you any performance to brag about... look at the rigols! So I was thorn between used and new, I ended up with the Hound for portability and a 8560E at the home lab. Had my eyes on the 8568 but it is a bit too large for my little lab. Note that the 8566 offers better frequency coverage but the phase noise is not as good as the '68.

Have you looked at John Miles KE5FX "Phase Noise" program??  Along with the program there is data on several spectrum analyzers and you can compare! No need to have an instrument attached to the computer!

It seens the R&S FSIQ is the top dog from his plots.

[G0HZU]

You have to be a bit careful when comparing the sample phase noise plots in the KE5FX programs. Not all of the analysers get a fair test.

Eg if you look at the Advantest TR4172 phase noise plot it looks really poor at 100kHz offset. eg -114dBc/Hz. But this is because the person who tested it only put a -20dBm signal into the analyser with 10dB internal attenuation. The TR4172 is one of the cleanest analysers at 100kHz offset even today. Mine manages about -128dBc/Hz at 100kHz offset at 1GHz when tested with a higher drive level.

Because of the way the Signal Hound appears to work (looking at the block diagram) I would expect it to have really good close in phase noise at the bottom end of its range. eg down at a tuned frequency of a few MHz. This is because the LO runs very close to the tuned frequency. So it can achieve very low phase noise here because it can divide the (2-4GHz?) LO VCO many times in order to get down to an LO of few MHz.

But I would expect the phase noise to be similar to the Rigol when tested up at a few GHz. i.e. very noisy because the cleanup benefits of the division are not there anymore because the LO will be running at the fundamental VCO frequency.

[G0HZU]

On the Signal Hound it is obvious which is the cleaner device re phase noise, and it even shows the dip caused by the Marconi's PLL loop filtering.

Hi Andy.
Are you still using the 2024 here? If so then the 10.1MHz plot looks a bit confusing to me. The 2024 should achieve about -140dBc/Hz at 1kHz offset at 10MHz and maybe -150dBc/Hz at 10kHz offset. Note: This assumes a decent drive level into the phase noise tester.

Therefore, I would expect that your phase noise response plot is really dominated by the SH phase noise limitations and not the 2024.

[Mr Simpleton]

Looking at the frequency range of the Signal Hound and given that it is switching signal paths at 150 MHz, my bet is that they uses the Analog ADF4350 synthesizer chip, specified to run from 150 MHz to 4400 MHz.  And you are correct that the phase noise go up substancial when increasing the frequency.

How did Marconi acheive -140 dBc/Hz @ 1kHz?? Did they employ dividing scheme too?? The specs says -121 dBc/Hz@20 kHz fc=450 MHz. Measuring -140 dBc/Hz for carrier at 10 MHz could be done using a SDR with a good sample clock..

I did some measurements comparing the Signal Hound and HP8560E using a HP8648 generator. This generator is by no means a low noise type, still I measured lower phase noise on the 8560, i.e. the Signal Hound internal phase noise was higher.

[G0HZU]

Yes, on the 2024 I think it uses a single VCO at around 400-540MHz followed by a choice of multiply ratios and then a chain of dividers.

I can crudely predict the phase noise at 1kHz offset by measuring the phase noise at 1GHz on the TR4172 and then subtract 40dB for the divide by 100 to get 10MHz.

On my TR4172 my old 2024 shows about -99dBc/Hz at 1kHz offset at 1GHz. So this would improve to about -139dBc/Hz when the sig gen is set down to 10MHz. But this is just a crude approximation.

I also have access to an Agilent E5052A SSA at work and I measured a few sig gens a while back. The 2024 does achieve about -140dBc/Hz at 1kHz offset for a 10MHz signal as seen in the plot below. Note: This isn't a plot of my 2024. it's one from work that had an odd blip in the phase noise at a few kHz offset.

I agree that the SH is either using an ADF4350 or and ADF4351 for its LO generation

[G0HZU]

For completeness I dug out the E5052A phase noise plot of that dodgy works 2024 when tested at 1GHz.

You can see that the phase noise is generally about 40dB higher. It also shows about -99dBc/Hz at 1kHz offset which agrees nicely with the measurement taken of my 2024 using my TR4172 analyser

The noise floor bottoms out at about -150dBc/Hz on the earlier plot because of the finite divider noise floor of the chain of dividers used in the 2024.

Also, below 10MHz, the 2024 uses a BFO to synthesise frequencies from 9kHz to 10MHz rather than using yet more dividers (and filters). So the 20logN cleanup in noise doesn't apply below 10MHz.

[G0HZU]

Getting back on topic, it's worth understanding the (very unconventional) block diagram of the signalhound because it really does throw the design rulebook out of the window in order to achieve its operation.

i.e. it uses a choice of more than one low frequency first IFs in order to try and dodge its inherently poor image rejection and I think it tries to predict what responses are image terms (that never get allowed to be shown on to the display) and what terms are legit. So what you see on the display has been processed to erase image terms.

In reality, I suspect it would be possible to search for and reveal some serious limitations in the SH's performance because of this but I think the design relies on the fact that few people will bother to look for these limitations and for the vast majority of scenarios these problems won't affect 'real world' testing.

But it does mean that this analyser can never be treated as a reliable lab spectrum analyser where you can usually be fairly certain of decent spurious free performance in pretty much every scenario