Keysight EXA Signal Analyzer / Spectrum Analyzer Review, Teardown & Experiments

[Hugoneus]

(Warning: Very technical video!)

In this episode Shahriar reviews the Keysight EXA Signal Analyzer / Spectrum Analyzer (N9010A). The X-Series Signal Analyzers and Spectrum analyzers from Keysight are compared against each other. The EXA is a general purpose spectrum analyzer with extensive applications.

The full instrument block diagram is presented with focus on various signal paths and frequency planning. Several of internal modules (RF Front-End, Digital Baseband Processor and LO Synthesizer) are also presented and compared with the block diagram. In order to examine the instrument’s capabilities in a real-world situation, a complete superheterodyne wireless transceiver with a sliding-IF receiver is designed and examined. Each component in the system (LNA, mixer, PLL, etc.) are individually characterized by using the EXA Signal Analyzer. The complete system measurements are also presented.

Various chapters in the video can be found at the following time marks:

Introduction (0:08)
Model comparison (0:50)
Instrument block diagram (4:27)
Module teardown and examination (20:39)
Instrument front/back panel (37:05)
Measurement setup description (43:22)
Doubler characterization with EXG as tracking generator (45:31)
LNA and mixer with NF and gain with Noise Figure personality (57:31)
Signal Hound VSG25A I/Q modulator performance, OBW, ACPR, TOI (1:08:57)
PLL characterization with Phase Noise personality (1:19:49)
Full transmitter measurement with Keysight VSA (1:26:08)
Full wireless link characterization with Keysight VSA (1:30:28)
Concluding remarks (1:37:21)

You can watch the video here: [1 Hour & 39 Minutes]
youtu.be/ghMdDDlfiwI

More videos at The Signal Path:
http://www.TheSignalPath.com

[HighVoltage]

Always something really great to learn from your videos.
Thank you !

[awallin]

I recently compared phase-noise measurement noise floor of a high-end Rhode & Schwartz spectrum analyzer to the Microsemi 3120A phase noise test-set (a direct digitizing signal analyzer)
http://www.anderswallin.net/wp-content/uploads/2015/12/sa_vs_test-set_for_phasenoise.png

Any chance you have a good quality OCXO, BVA, or H-maser in the basement and could measure and compare this with the EXA?
The noise floor measurements are done by splitting the incoming signal with a 50-ohm splitter and feeding bot REF and DUT inputs with the same signal.

thanks,
Anders

[drws]

Nice video. Thanks Shahriar for this one and all your other vids this year. I really like the tear down with experiments format. 

[1design]

Nice to see a serious R&D level SA being reviewed. Have you tried to lover the input amplitude level on the SA when receiving the -28dBm signal from the receiver? As most of the impairment is SNR related(bloomed constellation points) maybe there is some room to improve the SNR?

BR

[Hugoneus]

I recently compared phase-noise measurement noise floor of a high-end Rhode & Schwartz spectrum analyzer to the Microsemi 3120A phase noise test-set (a direct digitizing signal analyzer)
http://www.anderswallin.net/wp-content/uploads/2015/12/sa_vs_test-set_for_phasenoise.png
Any chance you have a good quality OCXO, BVA, or H-maser in the basement and could measure and compare this with the EXA?
The noise floor measurements are done by splitting the incoming signal with a 50-ohm splitter and feeding bot REF and DUT inputs with the same signal.
thanks,
Anders

I have a 10MHz reference, I can give it a try.

[dr.diesel]

Wow that was superb, I really enjoy the in-depth reviews.   

[diyaudio]

this must have taken a very long time to prepare and shoot. appreciate  the level of effort you put in your work.

exotic lab instrumentation 

[Hugoneus]

this must have taken a very long time to prepare and shoot. appreciate  the level of effort you put in your work.
exotic lab instrumentation 

Thank you. Indeed it takes a tremendous effort to produce a video like this. I don't think there is any other way, you can't review a >$100k instrument unless you setup some elaborate experiments.

[plesa]

Wow, thanks for detailed video!! Big thumbs up to Keysight for providing the extra boards!!

[Hugoneus]

Wow, thanks for detailed video!! Big thumbs up to Keysight for providing the extra boards!!

Yes, thumbs up for that from me as well. I appreciate the transparency and commitment on their part. They did the same thing when I reviewed the S-Series scope. They sent the whole acquisition board AND loaned me a 20GHz ultra low phase noise PSG to test the ENOB.

[Neganur]

How did you like the Signal Hound vector signal gen?

[KE5FX]

Great review, I'm about a third of the way through it.    Are there any links to high-res still photos of the boards they sent you?

Also, PM me your address, and I'll have Amazon send you a late Christmas present. 

[G0HZU]

I skimmed to the bit with the PCB views and I think the exposed circuit shown at 24:31 is the first IF amplifier at 5.1GHz and not the first mixer.

I think the actual mixer is going to be further up on the PCB under one of the sealed sections.

In my image below I think the IF1 output signal from the mixer enters this section in the top left and there is a crude form of IF port diplexer here. The path shown in blue is the high frequency part of the diplexer and I think this uses those coupled lines to feed/dump the unwanted mixer output frequencies of (maybe) 7GHz and upwards into those two little resistors circled in blue. At a guess this is a low cost termination made from two 100R resistors in parallel. The wanted 5.1GHz IF signal passes through the LF stub section and then into the green section which looks like a highpass filter to me (maybe passes above 3GHz?) and then it feeds the 5.1GHz signal into the IF1 amplifier device shown as a red triangle. You can see the Vsupply and Vbias feeds for this amplifier shown in red.

At least that's how I see this circuit operating... Maybe someone from Keysight can comment   

[Hugoneus]

It's definitely possible that is only the IF amplifier. I think I mentioned that it might be. Although I do see three paths coming to this section which led me to conclude that it is the up convert mixer for the low-band. I'll ask the Keysight engineer to be sure and will mark the video with my findings. Remember that the output of this section is fed to the microwave band pass filter right after.

[G0HZU]

Just to clarify, I think the exposed section is just the first IF amplifier as highlighted in the green square in the image below.

I think that the actual mixer will be under the screened section further up

[Hugoneus]

Yes, I understood what you were referring to before. Thanks! I'll check.

[PA0PBZ]

I asked on YouTube how you can mix 34.4 to 44GHz down to 322.5MHz with an LO that has to pass a 14GHz low pass filter, and Shahriar answered "It is a sub-harmonic filter".
Since that does not explain a lot to me, does anyone want to explain this a bit more? I think that to mix 44GHz down to around 300MHz  you need an LO like 43.7GHz, and that doesn't pass a 14GHz low pass filter. So what is the trick here? I tried to search for "sub-harmonic filter" but it looks like that is more popular in the music industry.

[G0HZU]

One other comment I have is wrt 9:15 onwards. The mixer #2 that is used across 3.6-17GHz won't be a subharmonic mixer. This mixer is circled in red below. The other two mixers above it (#3 and #4) are harmonic mixers.

Mixer #2 does require 'double' the LO frequency but this doubling is done external to the mixer. So this mixer is a fundamental mixer rather than harmonic. This LO doubler can be seen below circled in dark blue.

You seemed a bit confused as to the role of the (seemingly unused?) 17GHz filter contained withiin the light blue circle below. These two lowpass filters are used to clean up the output spectrum of the doubler. The ideal filter to have here would be a tunable/LO_tracking filter but Keysight have used a pair of lowpass filters at Fc 14GHz and 17GHz. So although the block diagram doesn't show this, the analyser will dynamically swap between these two filters as the LO is swept within a range. You can't get adequate spurious suppression with just one filter so I think it will swap dynamically between these two filters as required

[G0HZU]

I asked on YouTube how you can mix 34.4 to 44GHz down to 322.5MHz with an LO that has to pass a 14GHz low pass filter, and Shahriar answered "It is a sub-harmonic filter".
Since that does not explain a lot to me, does anyone want to explain this a bit more? I think that to mix 44GHz down to around 300MHz  you need an LO like 43.7GHz, and that doesn't pass a 14GHz low pass filter. So what is the trick here? I tried to search for "sub-harmonic filter" but it looks like that is more popular in the music industry.

I'll see if I can dig out some notes on this but even a classic diode DBM like an SBL-1 from MiniCircuits can be used as a crude harmonic mixer within its normal operating range.

Take a look at the mixer charts for this mixer and you will see that the conversion loss for 1RF, 3LO is quite good! Not as good as an ideal harmonic mixer but it will be within about 11dB of the 'normal' 7dB conversion loss of this mixer.

The SBL-1 mixer won't work very well on even order multiples of the LO (i.e. it will be very poor indeed) but a real harmonic mixer can be designed to work well on more than just one multiple of the LO. Sometimes this requires external control of the mixer operating point and this can be done with a variable bias control.

So you could use the SBL1 as a harmonic mixer if you pumped it with an LO at a third of the normal frequency. Obviously, you would be a bit mad to actually do this because the conversion loss will degrade by maybe 11dB. But up at microwave frequencies this strategy can be the best compromise. It means you don't have to produce (and manage/amplify/distribute) LO signals up at 44GHz in the case of the EXA analyser. Note that the conversion loss of the harmonic mixers in the EXA will be minimised because the mixer design will be optimised to run in harmonic mode at a fixed LO multiple and this will improve the conversion efficiency.

[Christe4nM]

Wow, what a great video! First I though "well, lets see which parts to watch," but I ended up watching it all in one go from start to finish. I've never been able to use a SA myself, neither do I have any experience in RF design. Still it's on my "if I ever have the money for it"-wishlist to get into EMC pre-compliance stuff and of course general purpose measurements.

Still, I learned so much about the instrument's capabilities, use cases etc. It's amazing to see the amount of systems engineering that even goes into the firmware design to use the SA as part of an automated setup. Very well done on Keysight's part.

So a big thank you for this great teardown/tutorial/experiments video.

[PA0PBZ]

So you could use the SBL1 as a harmonic mixer if you pumped it with an LO at a third of the normal frequency. Obviously, you would be a bit mad to actually do this because the conversion loss will degrade by maybe 11dB. But up at microwave frequencies this strategy can be the best compromise. It means you don't have to produce (and manage/amplify/distribute) LO signals up at 44GHz in the case of the EXA analyser. Note that the conversion loss of the harmonic mixers in the EXA will be minimised because the mixer design will be optimised to run in harmonic mode at a fixed LO multiple and this will improve the conversion efficiency.

Yes thank you, I already realized that he probably ment to say sub-harmonic mixer and not sub-harmonic filter when I read your comment about the mixers. It is clear now, and maybe time for a bit of experimenting, I have a few SBL1's laying around.

[G0HZU]

I think the 3LO 1RF conversion efficiency for the SBL-1 might be a few dB better that I posted above as I have a simulator program here. A lot depends on what mixer you use and also how hard you drive the mixer LO port to 'square up' the LO swiching action.

But because the SBL-1 type of DBM mixer uses the LO to switch the diodes as a kind of ON/OFF switch you can predict the 3LO 1RF conversion from the equations for a simple switching mixer.

The even order conversion efficiency will be extremely poor though.

It looks like mixer #3 in the EXA is optimised for 2LO operation and mixer #4 is optimised for 3LO operation.

[G0HZU]

Yes, I understood what you were referring to before. Thanks! I'll check.

I would guess that the amplifier used for that 5.1GHz IF1 amp will be something like an Avago ATF-34143 PHEMT or one of its relatives.

http://www.avagotech.com/products/wireless/transistors/fet/atf-34143

You could try asking Keysight or maybe look at the markings on the part itself if you still have the PCB.

[AF6LJ]

Good video I loved it.