SSB Phase Noise Measurements - Any Experts want to give some input?

[Hawaii596]

One of the challenges of being a "Jack-of-All-Trades" is that you have to get into SOOOOO many different disciplines.  I am a HUGE believer in not being afraid to look stupid for the sake of learning...

Over the years I've done a lot of lower to mid echelon signal generators (among way too many other things in two zillion metrological disciplines).  Going way back including such as 8640B's (those were nice generators for their time), to a little more recently 8656, 8657A/B, etc....  Now we're starting to see some of the E4400 series (E4420B, E4422, E4425B, etc..).

I'm theoretically basically familiar with what Phase Noise is (not at all expert).  And I understand that if you make a quasi-measurement on a good spectrum analyzer, the floor is limited by the phase noise quality of the spec an.  For example, checking an E4425B at 1 GHz with a spec of <-130dBc/Hz....

I recently inherited an HP 89441A spec an (very nice system).  In this indirect, "quasi" method, I follow a White Paper produced by Agilent for making SSB phase noise measurements on the HP 89441A, use markers to measure a power band of 1 KHz at 10 KHz offset from freq.  I use 200 Hz RBW and subtract 23 dB (to correct to 1 Hz equivalent RBW).  And on an E4425B get a displayed floor in my power band markers of about -89 dB (subtract 23 dB brings me to -112 dBc/Hz).  I understand that PROBABLY this incorrectly high reading is based on the quality of the spec. an's oscillator circuitry quality (or it's timebase quality = = short term phase stability).

So short of finding a resurrectable old HP 3048C (I think that's the model), or buying the new Agilent Phase Noise Measurement system (don't know the model off hand - but I saw one at the NCSLI conference - NICE!!! but expensive), what are my options?  Do I misunderstand anything in my descriptions above?

One thought I have is when I ship one of my better generators out for OEM accredited cal, using measured SSB Phase Noise measured values, measure it on my HP 89441A, and use the delta as a correction factor for the given frequency.  The two best generators I have are an HP 8665B and an Agilent 83623B.  I don't believe the 83623B is as good as the 8665B.  Most of what we have to cal is in the low GHz (mostly <= about 3 GHz).  So the 89441A has adequate bandwidth for most of it (2.65 GHz).

One other detail in the measurement is that the White Paper tells me to use Band Power marker function where it measures power level in a 1 KHz wide band at 10 KHz offset from the carrier; or 20 KHz when needed.  If I measure using the Band Power markers compared with normal marker at that same frequency, there is about a 7 dB difference between them (Band Power higher).  Can anyone shed light on that?

Guess that's enough for now.

[CalLabSolutions]

First of all I AM NOT AN EXPERT..

But I have helped so labs figure out how to make SSB measurements without a spending a lot of money.
The first question is always, how important is SSB to me and or my customers.  From my understanding if of the importance of this test is when the signal generator is being used to source an RF Signal where phase is critical. Usually in Radar system, Phase Modulated signals and especially in IQ modulated signals.
* Note  I find it hilariously entertaining when see a signal generator in a calibration lab with a full blown calibration including SSB and all they use it for is CW signals to calibrate bandwidth on scopes..

As I understand is SSB is like Rise Time.  Minus infinity is a perfectly clean RF Signal, just like 0.0s is a perfect rise time.  And in SSB measurements, just like in Rise Time measurements you have two factors you have to consider with making a measurement.
1)   Can the Standard measure the SSB signal that low.
2)   How accurate is the measurement..
Most O-Scopes can make accurate rise time measurements because the accuracy is based on the time base of the scope.   But their bandwidth limits how fast of a rise time signal they can measure.  And the same is true in SSB..  Your standards have to have the dynamic range to measure the signal, and the accuracy at that level to be creatable.
That said I do not think that sending a signal generator out for calibration, then measuring the SSB with your standard and using the delta value as a calibration factor is a good idea.
However, I believe if you send your signal generator out for cal, you can use it as a check standard for the 89441A.  The first question is can your 89441 measure the SSB at or below the levels provided.  (Note if not that is a good indication you have reached the Noise Floor of the 89441.  And If the 89441A is able to measure the values lower, that only means you have a lower noise floor.) 
* Again Keep in mind that Noise Floor is independent from uncertainty.
I have never tried to measure SSB with an 89441.  It is like using a sledge hammer to drive a nail, you can do it, but it is not the right tool. 
Typically what happens in newer analyzers is they digitize the signal.  Then they can manipulate the signal on the screen using math mathematical formulas. (Unlike the old days when the engineered something).  If you where to take a new Analyzer and press the marker button and note the reading, then press the maker noise button you would find the maker noise is (Noted Reading – 10*log(Resolution Bandwidth)).  This is why they are saying set the RBW to 200Hz and subtract 20 db..  (10*Log(200)=-23.0103).
This brings us to the can you measure SSB with an 89441.  As I see it the procedure is valid, the technique is sound, but it is based on some assumptions of digitization.  The problem is you are not actually measuring power down at that level.  And you can't without filtering the signal, amplifying the signal, filtering it some more and amplifying it again.  This is what the Agilent, system does.

[Hawaii596]

Thanks very much for that insightful input.  I'm still in process of reading everything I can on the topic.  The generators being cal'd in question are from our external customer.  I am fairly certain it isn't critical to this particular customer.  But as we expand (converting from an inhouse lab to a commercial one), I have to always presume every customer needs everything tested to spec unless we agree otherwise.  So since I recently got the 89441A system (prior to that I had/have an 8566B and a 3589A), I want to see what I can make it do.

I discovered last night after writing the first post that there is a function on the 89441A cal "C/N" -- "Carrier to Noise Ratio."  When I turn that function on (under MARKER FUNCTIONS), and look at 20 KHz offset from a 1 GHz carrier on the E4425B that I'm testing, it magically gives me approximately the spec'd dBc/Hz value (<-120 dBc/Hz).  Even though my measured value in dBc is about -96 dBc, it seems to approximately match up.  I have the analyzer set to Vector mode....

I feel like I'm getting old.  I wish I was still as young as the 10 year old avatar photo I use on here.  It helps me feel a little younger comparing that photo to my current one.  I'm going to "keep plugging" on this one.  I'll do some research about what that C/N marker function means.

I will say this though.  The HP 89441A is a really nice system.  It was transferred from another of our labs.

[CalLabSolutions]

I agree, the 89441 is great system.  For a while it was the only system that could test the digital (i.e. IQ) modulated Signal Generators.  Now you can get an Agilent N5531A or R&S FSMR with the digital modulation personalities.

[Hawaii596]

Among my numerous development projects is broadening and improving our RF area (and bringing the mothballed electrical standards lab online, and getting our MKS vacuum calibration system online, etc..).  Kind of the bottom line is that I'm trying to do so as cost effectively as possible.  Then as business upturns, bring in new revenues, and then buy some of the nicer newer stuff.  I just finished bringing an 8902A back from the grave, got the 89441A online, re-organized the RF area.

Yes, I would definitely love to get one of the new Agilent or R&S systems (I think our other lab has the R&S replacement for the 8902A).  But - no  money these days.  So I get to tinker on the old stuff and bring it online.  That's about as fun as learning to use the new stuff.

Regarding the 89441A, I need to do a little homework on the MARKER FUNCTION = "C/N" (Carrier-to-Noise Ratio).  even though the dBc from carrier to 20 KHz offset (averaged display at 200 Hz RBW) was about -96 dB, the C/N number was 120 dB, which is exactly what the SSB Phase Noise ratio spec is for that generator at that frequency.  If it is correct that it is telling me in direct readout the Phase Noise number - - - - NICE!!!!  Although I'm not yet sure.  Maybe I'll try it on a few other generators and compare to their SSB specs.

Besides how nicely the numbers seem to  match, I get this reading when I put the 89441A in VECTOR mode, which would be measuring phase relationships.  It just sounds too easy and too good to be true.

[RFCAL]

We use an Aeroflex PN9000 with it's own generator and down converters.This is SWEET!!!!you can have a plot in about 2 minutes with 3 clicks or less. The noise floor on this is about-165.This will handle all of your PSG's,ESG's,MXG's and SMR's.

[Duckbutta]

Aeroflex? That was a good one. You had me going for a minute there. I actually believed you were vouching for Aeroflex............Oh, wait.... You were.

[Wilk]

I'll vouch for Aeroflex when it comes to Phase Noise.  They were the best there ever was in this paticular area.  Problem is I said was.  I think they stoped making Phase Noise Stands.

If you don't think so Duck you either never saw one, or you could not figure out how to use it.  The crystal references that their high end stands used where easily 10-30dB/Hz better than any stand Agilent has ever built could measure.  You go into any lab, that deals with Satellites, or high bandwidth/frequency Comms in any way over the last 15 years, and racks of Aeroflex is all you will find.  There is a reason for that.  You don't see alot of it, cause they charged a premium for their gear.  Only the best could afford it.  Hell their old low end PN8000 which is a 10-15 year old design now will still hold its own with anything Agilent has to offer if you know how to use it.  And it was a toy compared to a what a fully loaded out PN9500 can bring to the table.

I am not a huge fan of alot of their other offerings, but when it comes to Phase Noise they used to be un-touchable. 

Nope, I do not work for them.

[Duckbutta]

Wilk,

You may be right about Aeroflex Phase Noise Standards but I have my doubts. I don't have any first hand experience with them but I do have a LOT of experience with their Comm. Analyzers and I can tell you that they are GARBAGE. If their Phase Noise system is anything like those, I would be less than impressed. Reputation is everything in the test equipment business and Aeroflex doesn't have a very good one. Just ask anyone who works in an Air Force cal lab and has the unfortunate experience of working on that slop. As for Phase Noise, I'll stick with my Agilent E5504B and sleep good at night.

Once upon a time Wavetek actually made a pretty accurate meter calibrator (4808) so that's proof that even a blind squirrel can get a nut. Your post, if true, is really just saying that.

[RFCAL]

I can vouche for the Phase Noise Systems.They work great and yes,they still make them.Call and ask for a Demo.

[mrrob007]

I can vouche for the Phase Noise Systems.They work great and yes,they still make them.Call and ask for a Demo.

Wow really? That makes me wanna run out an go get one today.....