R&S SMIQ as a replacement for a general signal generator?

[davorin]

As I'm thinking of getting a vector signal generator I saw in the SMIQ specs that it can also do general AM/FM modulation which I sometimes also need for testing AM/FM radios (mostly those old vacuum tube ones ;o)

Is a SMIQ03 basically just a SMT03 with additional QAM option (besides higher frequency start).

Or are there better alternatives in this price range the SMIQs pop up on Ebay?

[Wuerstchenhund]

As I'm thinking of getting a vector signal generator I saw in the SMIQ specs that it can also do general AM/FM modulation which I sometimes also need for testing AM/FM radios (mostly those old vacuum tube ones ;o)

Is a SMIQ03 basically just a SMT03 with additional QAM option (besides higher frequency start).

The R&S SMIQ is based on the SME (the SMT is a SME intended for VOR/ILS testing) and is a true vector signal generator which can create a range of complex modulation signals. It's essentially R&S counterpart of the Agilent ESG-D Series (the SME is the equivalent to the Agilent ESG-A), although the SMIQ supports more modulation variants than the ESG-D.

Both ESG-D and SMIQ can be equipped with a built-in AWG to generate I/Q signals, but the one for the SMIQ only offers 512k (1M for the ESG-D). However, the SMIQ offers some slightly better specs than the ESG-D, and the SMIQ's display is much better than the poor LCD in the ESG Series.

In terms of software, Keysight offers some old version of IntuiLink for download which allows you to transfer waveforms into the AWG option of an ESG-D, but it's really old and very simple (it works directly from MS Office, although not with newer versions). R&S offers some better tools (i.e. WinIQSim) for the SMIQ, and while older, they work fine even on Windows 8.1 x64.

Or are there better alternatives in this price range the SMIQs pop up on Ebay?

The Keysight ESG-D could be an alternative, but prices of SMIQ and ESG-D are pretty similar. The ESG Series has a common fault in the output stage (shows UNLVL on the display even when AGC is enabled) which isn't too difficult to repair. The SMIQ (like the SME) is pretty robust and doesn't show any similar common faults, however the metallic coating of the display glass tends to deteriorate over time which looks a bit like someone has put some glue on borderline of the glass. It's purely optical but replacement glass isn't cheap.

You should be able to get either one for a good price if you're patient. If I had to choose however I'd take an SMIQ over an ESG-D any day, mostly because of the better display and (with the exception of the AWG memory) slightly better specs.

[davorin]

Well...I would prefer the R&S SMIQ over the Agilent model....quick look on ebay shows me that the Agilent E4432B is over US$ 2k while the SME03 is under 1k...some of them even lower but faulty with error numbers 172 and the like..not sure if fixing would be easy then...

Doesn't look though that the SMIQ would accept analog AC input for AM/FM modulation directly...only 0 - 1V....

[Wuerstchenhund]

Well...I would prefer the R&S SMIQ over the Agilent model....quick look on ebay shows me that the Agilent E4432B is over US$ 2k while the SME03 is under 1k

The SME is not vector signal generator, so it's clear that prices are lower.

...some of them even lower but faulty with error numbers 172 and the like..not sure if fixing would be easy then...

Often its just a dead battery, but before buying I'd ask in the R&S Yahoo forum for opinions as there should be a few people who have experience fixing them.

Doesn't look though that the SMIQ would accept analog AC input for AM/FM modulation directly...only 0 - 1V....

Not sure what you mean. You can input an external AM/FM modulation signal but as you said the voltage should be 1Vpk (the limit is +15V).

It's the same with the Keysight ESG-D.

[davorin]

Did I write SME03? No should be SMIQ03 of course..my fault (o;

Yes I definitively need to get more information on those errors and how it could be fixed...yahoo was always a good source for that kind of information...
Is there an equivalent NNTP group?

Hmm...not many youtube videos of the SMIQ in use besides booting it up (o;

[KJDS]

Anritsu have an assortment of similar sig-gens that are worth considering.

[Wuerstchenhund]

Anritsu have an assortment of similar sig-gens that are worth considering.

True, but getting service/calibration manuals and (when required) spare parts for Anritsu kit can be challenging. It's often a lot easier for the western brands.

[G0HZU]

I would definitely recommend the ESGD if it has the highly desirable dual arb option fitted. I have an E4433B with this option and it's a very powerful sig gen.

However, neither the SMIQ or the ESGD are going to deliver anything like lab grade performance in terms of phase noise or spurious. These are very much economy grade generators in terms of their basic frequency synthesis but they offer lots and lots of features including a vector modulation capability.

They look pretty impressive on the outside but don't expect old school lab grade performance in terms of phase noise or spectral purity from these basic old workhorses. For this reason it's best to go for one with the best options fitted because that's where their true value comes from IMO.

I recently bought an ESGD E4433B for £1k with the dual arb option and several other options. This is probably half the normal used price but the price of these should start to drop soon because they offer limited capability in terms of dual AWG performance/bandwidth and also the Agilent support tools are pretty lame (obsolete). So many companies will be moving over to the later MXG type generators.

AFAIK the SMIQ is inferior to the early ESGD models in terms of AWG performance so I would avoid it unless very cheap. If you want the ARB capability I would consider it a no brainer to go for the Agilent unless there is some other unique feature that you need from the SMIQ.
If funds permit then consider the E4438C. This looks the same as the E4433B but it does have better performance in several key areas including its (now 16 bit) AWG and it is supported by the 89600 SW.

[Wuerstchenhund]

However, neither the SMIQ or the ESGD are going to deliver anything like lab grade performance in terms of phase noise or spurious. These are very much economy grade generators in terms of the basic frequency synthesis but they offer lots and lots of features including a vector modulation capability.

I disagree with the notion that ESG and SMIQ are not "lab grade" (which is a silly term anyways), unless you think "lab grade" means "best specs ever in all areas" (which it doesn't).

SMIQ and ESG-D are most certainly "lab-grade" equipment ('lab-grade' does *not* mean offering the best specs possible in all areas), and at new prices in excess of $30k I'd say calling it "economy grade" (a term usually reserved for entry level kit like the stuff from Hameg) is quite a stretch.

But that doesn't change that both SMIQ and ESG-D are pretty much "lab-grade" instruments, and the fact alone that they found use in a huge number of labs around the globe puts the notion that their performance isn't up to scratch for professional use clearly into the sphere of nonsense.

I recently bought an ESGD E4433B for £1k with the dual arb option and several other options. This is probably half the normal used price but the price of these should start to drop soon because they offer limited capability in terms of dual AWG performance/bandwidth and also the Agilent support tools are pretty lame (obsolete). So many companies will be moving over to the later MXG type generators.

That cycle has passed many years ago already, at least for the first generation ESG-D (B models). At the moment its the later ESG-D (E4438C) that are being replaced.

The same is true for the SMIQ, which has been replaced with the successor (SMU) a long time ago, and similarly this successor is now slowly being replaced with the current modell (SMW).

Most of the ESG-D B models and SMIQs on the market are probably coming from second or third hand users.

AFAIK the SMIQ is inferior to the early ESGD models in terms of AWG performance

Can you elaborate? The only difference between the AWG in the early ESG-D models and the SMIQ is the sample memory (512k/ch on the SMIQ vs 1M/ch on the ESG-D), and both values are pretty tiny by today's standards anyways.

The second generation ESG-D (which is more comparable to the R&S SMU200A) offers a much better AWG performance, but that clearly comes at a cost.

[G0HZU]

But that doesn't change that both SMIQ and ESG-D are pretty much "lab-grade" instruments, and the fact alone that they found use in a huge number of labs around the globe puts the notion that their performance isn't up to scratch for professional use clearly into the sphere of nonsense.

If you read what I said again I said they are not lab grade in terms of phase noise or spurious.

I know the ESGD pretty well and in terms of close in phase noise in the LF/HF/VHF bands the phase noise performance is very, very poor because the generator uses a very simplistic mixer based frequency plan to generate signals down here.

In terms of close in phase noise down in these bands the ESGD is a total dog compared to a high end sig gen. Easily 35-40dB worse than something decent. Even a novice could build a simple homebrew oscillator with phase noise that would be 30-40dB better than the ESGD down in these bands.

Can you elaborate? The only difference between the AWG in the early ESG-D models and the SMIQ is the sample memory (512k/ch on the SMIQ vs 1M/ch on the ESG-D), and both values are pretty tiny by today's standards anyways.

I've not used the SMIQ AWG but I think the SMIQ is only a 12 bit AWG and the E4433B ESGD is 14 bit? I've produced my own AWG analysis/support tools for my ESGD and it asks for 14bit data for the AWG in the manual.

[G0HZU]

and at new prices in excess of $30k I'd say calling it "economy grade" (a term usually reserved for entry level kit like the stuff from Hameg) is quite a stretch.

Back in the 1990s Hewlett Packard classed/marketed the first ESG signal generators as "Economy Signal Generators" and the A and D variants use the same basic frequency plan. So the close in phase noise performance for sig gen frequencies from LF through to lowish VHF frequencies is dire on both A and D versions because of the shared frequency plan.

[davorin]

Nice discussion I have started (o;

Well for sure I know it's not gonna be the Signalhound VSG25A....

Speaking of the Agilent model...can the VSA software control the E4406A and the E4433B at the same time?

[G0HZU]

Sadly no...

The E4433B isn't directly supported and it also uses a different scheme to download IQ data to it via GPIB. So it's a double no in this respect unless someone knows otherwise? Maybe an older version supports it?

I'm using the E4406A and the E4433B together as part of a research project so I've created some reasonably powerful Windows based tools to allow me to view/edit/process captured IQ data from the E4406A and it converts the format and sends it to the E4433B for playback. These tools do some specific things unique to my requirements and I'm afraid that I can't give away my software tools for various reasons. 

Note that I'm not a SW engineer and it really isn't difficult to create apps like this using something like VB anyway

[davorin]

Well I just started recently with C#...and was impressed how easy it is to develop a websocket communication app for Windows phone talking to a Raspberry Pi via Amazon server (o;

I'm still not sure which model I should choose...Agilent or R&S....I like both brands....I'm very happy with me new E4406A...but also like R&S...otherwise I wouldn't have two CRTU-RUs and a SMT03 now...kinda weird hobby collecting old equipment for studying and learning RF technology....

And since it's a private hobby I tend more to go for a device with lower price.....but currently most SMIQs on Ebay are listed as "for parts only" with a low price...or Agilent models starting at US$ 2500....

Hmm..speaking of self written software....is there a software API how to directly dump I/Q from a E4406A via socket communication?

[ivaylo]

I replied to your other thread. E4433 with UND can be found cheaper in the states and many sellers ship internationally.

Why are those E4406 so cheap (well, so to speak)? Limited spec or what?

[KJDS]

I replied to your other thread. E4433 with UND can be found cheaper in the states and many sellers ship internationally.

Why are those E4406 so cheap (well, so to speak)? Limited spec or what?

E4406A are cheap because they're no longer wanted my modern businesses and Nokia put about a thousand onto the second user market a few years ago. I've never seen more than ten signal generators for sale at the same time.

[trukresom]

Since the OP mentions the possible use of an SMIQ with AM and FM I would like to underline
that the basic SMIQ is not equipped to do FM. For FM and PM the SMIQ has to be fitted
with the Option B5.
A characteristic of the SMIQ which I find is very pleasant is the fact that when making small
changes in frequency with the knob, the output amplitude remains constant. This is not the
case for example with my HP8644B; even the smallest frequency change (0.01 Hz) causes
a drop of the output amplitude of almost 100% for a few milliseconds.

[G0HZU]

I'm still not sure which model I should choose...Agilent or R&S....I like both brands...

Looking at your wanted ad in the Buy/Sell section are you sure you really need to purchase a generator like this?

If you just want to learn stuff about vector modulation I'd recommend either making your own basic vector modulator or maybe just buy a basic vector modulator chip (or eval PCB) from one of the major chip manufacturers. eg Hittite or RFMD etc etc.

Build it and play with it and maybe use your C# skills to use a PC soundcard to drive it?

I really do think you will learn a lot more and learn it faster if you do it this way?

But if you really do want to buy a vector sig gen then either the Agilent or SMIQ will be fine for general research or learning.
However, you are going to end up with your head in their manuals a lot of the time doing stuff their way in terms of programming etc.

I would still repeat the warning about LF >> VHF phase noise performance though. Especially for the ESGD series from Agilent. If you expect this generator to 'replace' a decent shortwave band RF generator in terms of noise performance then you could hit problems if you wanted to do any critical receiver testing of something like a decent HF ham radio. The same applies if you wanted to use it as a temporary LO in a shortwave converter. The relatively high phase noise will markedly limit the performance of the converter in terms of reciprocal mixing compared to even a simple homebrew oscillator.

[G0HZU]

I disagree with the notion that ESG and SMIQ are not "lab grade" (which is a silly term anyways), unless you think "lab grade" means "best specs ever in all areas" (which it doesn't).

Generally speaking, a decent lab sig gen should produce phase noise performance in the same ballpark (or better) as a fairly basic oscillator designed for that frequency.

HP set the bar for this way back in the early 1970s with the HP8640. AFAIK HP designed this solid state generator to try to emulate the excellent phase noise performance of the vacuum tube based sig gens of the 1960s. They did it using a mechanically tuned octave bandwidth cavity oscillator (256-512MHz?) followed by a chain of dividers and roofing filters and a classic ALC levelling circuit.

In doing so they created the industry benchmark for phase noise (and spurious) performance for many years.

This old generator will produce very low close in phase noise when used in the MW and SW bands. The problem is that this type of generator is expensive and it also needs lots of switched LPFs to cover all the divider ranges. Also, using division means that accurate calibration of FM deviation has to be achieved at the cavity oscillator for every divider range.

Lots of other makes of sig gen have since adopted a similar oscillator + divider approach and a decent lab generator should at least try and get close to the phase noise performance of the old HP8640 across 'most' of its range.

(Modern generators often use several VCOs to replace the big old cavity oscillator in the HP8640 and this usually compromises performance. However modern generators can use modern synthesis techniques to clean up very close in phase noise so they can outperform the HP8640 in some areas)

Because it uses a cheap mixer based frequency plan to generate low frequencies (rather than using dividers and filters) the ESGD (AKA Economy Signal Generator Digital) generator fails this benchmark by about 40dB down in the lower parts of the SW band. eg for phase noise offsets of a few kHz at a CF around 10MHz or so. That's a pretty significant fail in terms of phase noise

[G0HZU]

In terms of vector modulator performance I can comment on the E4433 ESGD as I have a couple of them here. Both give very similar modulator performance but there are a few niggles...

The basic calibration of the IQ modulator is fairly good. I can run a full modulator calibration across the full range of the generator (takes a while to do this)  and then not bother to recalibrate it for quite a while and still get pretty decent carrier and image rejection on something like a multi carrier SSB plot. I generally don't bother to calibrate it for most tasks so it can go for weeks without a calibration.

However, for critical applications it does allow a User Cal that lets you calibrate the modulator across a narrow band of interest and this takes a few seconds. The performance of the IQ internal cal is quite good in terms of image rejection but if I had to fault it, the resolution of the manual trim is too coarse.

With a raw auto calibration it can deliver about 50dB carrier suppression and over 60dB image suppression for simple narrowband waveforms but the manual trim step size is too coarse to let you 'improve' the carrier null much beyond the auto calibration. The image suppression obviously degrades a bit for wider bandwidth signals that get close to the full modulation bandwidth of the modulator.

Overall, it's pretty good but I think it could be better.

Note: One of my E4433 generators is an older A model and it has a scary bug/feature during IQ calibraion in that it does the calibration with the generator set at full level with the output enabled!
This is not good if you have a circuit connected that could be damaged by a +20dBm signal!

The B version (sensibly) sets the attenuators to max during IQ modulator calibration so no signal leaks from the generator during IQ calibration.

It could be that my generator is faulty in this respect but I suspect that this is a firmware bug in this early model.

[G0HZU]

One (optional?) feature of the E4433 is that you get raw baseband IQ outputs at the rear of the instrument.

So if I set one (the E4433B) to give out a 4 tone SSBSC signal using the AWG and then feed the IQ out into the analogue IQ inputs of the E4433A I can look at how well the two perform when daisychained like this.

In the plot below you can see that the 4 image terms are now visible on the left side of the suppressed carrier. The suppressed carrier is the tone at the centre of the display. On the E4433B the image terms are much lower and the carrier is at about -50dBc rather than -45dBc in the image of the E4433A below.

But it's still pretty good if you consider that this is being driven from the baseband generator from one generator and fed into the IQ inputs of another. I did a User Cal on the E4433A IQ modulator prior to feeding in this signal and also the ALC is turned off as the ALC action degrades the performance slightly.

[PA0PBZ]

.... the 4 image terms are now visible on the left right side of the suppressed carrier.

you had me staring at the image there for a while 

[G0HZU]

Sorry for any confusion... by image terms I mean the 'unwanted' image terms on the left of the carrier as arrowed in blue below. If there was only 30dB image rejection there would be a set of 4 (unwanted) image tones on the left of the carrier that sit 30dB lower than the (wanted) terms on the right of the carrier. These would appear as per the dotted yellow lines below.

I only have to manually trim the IQ calibration slightly away from the auto cal settings to upset the modulator and make these unwanted terms begin to appear as per the yellow lines

[PA0PBZ]

Ah, I see, thanks for clarifying that, it's the unwanted sideband you are talking about.

[G0HZU]

For a bit of fun I tried capturing a similar 4 tone SSBSC signal with the E4406A and the VSA SW and I then sent the captured IQ data back to the E4433B as an ARB file for playback. So in an ideal world the signal should look the same as the raw signal from the E4433B.

However, to prevent glitches on an analyser that would appear at the loopback (start/stop) point of the captured segment my software can crop the data to make sure there is a fairly seamless crossover at this point. This removes a huge discontinuity at the loopback point and improves the playback.

Without this, the analyser would show jumpy glitches and something like an old school swept analyser would show a huge fan of spiky clicks across the spectrum.

This plot below is my second attempt BTW. The first one had lots more distortion terms. I managed to improve this by setting up the VSA interface in a better way on this second attempt.

The result is pretty good although I think it can be better with a bit more experience on my part. I'd imagine something like the SMIQ (or a decent Anritsu vector sig gen) would give similar performance if I used it instead of the E4433B.

But I guess the system performance here is set by the E4433B and also the combination of the linearity of the RF converter in the E4406A and the performance of its 14 bit ADC/digital IF and also the VSA SW and also the way I process the captured VSA data to make it compatible with the E4433B before sending it to the E4433B for playback via its 14 bit AWG.