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

[G0HZU]

Just to show Davorin how poor the ESGD is when it comes to its phase noise performance in the shortwave bands I set up a very simple experiment/comparison.

I compared my E4433B against an old Marconi 2019 that dates back maybe 35 years. The Marconi 2019 was a regular sight in test and service departments throughout the 1980s and 1990s. We even had a couple of them in our design labs in the 1990s as the phase noise performance was pretty good across 2MHz-30MHz.

The performance of the 2019 isn't so impressive once you get up into the VHF/UHF ranges and they aren't very reliable either but even this old generator can be seen to totally outclass the ESGD for close in phase noise on the SW bands.The 2019 uses dividers to synthesise outputs down to about 2MHz so it achieves good (but not spectacular) noise performance in the SW bands.

I set each generator to a few kHz outside the passband of a 10.7MHz crystal filter and set the level to +4dBm to account for filter and cable loss. So the sideband phase noise that leaks through the filter in its passband can be read as -dBc/Hz. The leaked carrier can be seen on the right of the analyser plot and it's level is about 60dB lower than normal because it is being attenuated by the filter stopband.

If you look at the plot below there are two traces stored on this old HP8568B analyser. Can you guess which is which? Yes, the one that is over 30dB higher in noise level is the ESGD. When measured on a signal source analyser they typically produce about -100dBc/Hz at a few kHz offset down on the SW bands and maybe -108dBc/Hz at 10kHz.

The creaky old 2019 that I've owned for about 20 years manages about -133dBc/Hz at about 5kHz offset at 10MHz CF. The ESGD shows about -101dBc/Hz at the same offset in the plot below.

A Marconi 2024 should manage about -138dBc/Hz here and I'd expect the HP8640 to better this by several dB. I'm guessing maybe -145dBc/Hz but I've not had my hands on an 8640 for many years. But a decent 8640 should deliver about -115dBc/Hz at 5kHz offset up at its cavity oscillator at 320MHz and the dividers should clean this up by 30dB when dividing by 32 to get 10MHz. So it could (reasonably) be expected to achieve -145dBc/Hz at 5kHz ofset at 10MHz CF. Thats over 40dB better than the ESGD.

So anyone choosing to replace a regular sig gen for use on shortwave (AM bands) should be aware how poor the ESGD is for phase noise down here. Because it is at least 30dB off the pace here it can't be taken seriously as a lab sig gen for typical performance checks of a decent SW receiver for example. Even the 2019 will struggle here when testing a decent Ham radio for blocking or adjacent channel performance despite being >30dB better
than the ESGD.

[G0HZU]

I dug out an old E5052A library plot of a 2024 set to 10MHz and +13dBm.

This is one of the 2024 generators at work and it showed a curious blip in the phase noise at about 1.5kHz offset (green arrow)

But you can see that this generator manages about -140dBc/Hz at 5kHz offset at 10MHz. Apart from the blip, this generator was slightly better than a few other 2023 and 2024 generators I measured in comparison.

The red trace is hand drawn by me and represents the typical phase noise of an ESG4433 at 10MHz. It is much, much worse as you can see.

The blue trace is also hand drawn and it represents the typical phase noise of a simple homebrew LC oscillator at 10MHz. i.e. something you could make with a transistor, and a few resistors, caps and an inductor.

[analogNewbie]

I'm also looking for a signal generator these days.

Anyone has the experience with Agilent 8665A? This heavy guy takes about 6 minutes or more to do the cool boot calibration. Does anyone know what is this guy calibrating? I think the phase noise performance of 8665A is not bad, any comments  on this unit?

Thanks

[G0HZU]

The 8665A is an early PSG model. I've never used one because when these were being sold 20-25 years ago the company couldn't afford the PSG range. By the time the company grew enough to buy a £££ sig gen we bought the later PSG models. So I have zero experience of them.

However, I think that the early range of PSG models use a precision delay line based  discriminator as part of a PLL+FLL. This allowed a useful reduction in phase noise compared to a basic single loop PLL if the user selected this FLL mode. However, this complex/expensive FLL mode may be a factory option. I'm not sure.

This mode will be slower to tune and lock but gives better phase noise.

But (I think) the delay line in the FLL and the VCO tuning has to be calibrated and in some PSG models these delay lines are even heated by an oven to make them more stable so 'maybe' this is what takes all the time to stabilise after a cold boot.

But all of this is just a guess. Maybe someone on here has one and has serviced it etc.

In terms of phase noise, I think this model still resorts to using a mixer to get its lowest frequencies (eg LF through to VHF) so I would guess that its phase noise could be as bad as the ESGD down here unless its FLL 'noise cleanup' mode is enabled. In this mode it will be much better than the ESGD but I doubt it will be as good as something like a basic Marconi 2024 down in the SW bands. Note: the Marconi 2024 also uses a mixer to get its very lowest frequencies but this applies below about 10MHz and the phase noise is still much better than the ESGD below 10MHz despite this.

Up at UHF>>xGHz the 8665A phase noise performance will be extremely good. Much better than a 2024 because it has the FLL mode to exploit in order to achieve low phase noise across maybe 2kHz to 100kHz offsets and it uses a YIG oscillator.

Basically, the 8665A isn't a generator that would be on my wish list unless I really 'needed' its low phase noise up at UHF. It's very big, very heavy and offers limited modulation capability.

Maybe someone on here has one and can comment?

[analogNewbie]

Thanks for the detailed info about 8665A. The close in phase noise is not as good as 8662/8663A, but the 1k+ phase noise is pretty good.

I want to ge one with 4GHz capability and phase noise shoud be good. Now I found one 8665a and one R&S SMHU. The second hand ESG series are all over the place and the prices are much higher than these old heavy tanks.

Since I want to do some experiment to homebrew GHz spectrum analyzer and GPS recievers, which model do you suggest? I can live with poor phase noise performace under 10MHz range .

[G0HZU]

Since I want to do some experiment to homebrew GHz spectrum analyzer and GPS recievers, which model do you suggest? I can live with poor phase noise performace under 10MHz range .

They are both very good signal generators and assuming they are both healthy then I don't think there's a right one or a wrong one. So it may come down to simple human factors as below:

I had the pleasure of using the 2GHz R&S SMH throughout the 1990s as a general purpose sig gen. This generator was too expensive for the company to buy but it was donated to us by the customer on a research task. It is a really nice instrument to use, the whirly wheel has a high quality feel as do the buttons and it even sounds nice when the attenuators click etc. The display is pretty basic but it still looks good.

It hasn't been totally reliable since then (been back more than once for repairs) but we still have it and I think we first got it way back in the early 90s.

The phase noise is pretty good (but not spectacular). I have a library of phase noise plots for lots of our works generators here and here is the SMH at 1500MHz. The noise profile at 1500MHz shows a fairly high pedestal of noise out to 200kHz and you can see this isn't as good as the 8665A.

It's quite good down at lower frequencies although I didn't measure it below 10MHz. But there's a plot of it at 10MHz below.

Hope this is useful.

I would definitely find room for an SMH here, partly through nostalgia and partly because it is such a nice generator to use. However, I didn't ever use any of its modulation modes back then. It was just used as a low phase noise signal source in the tasks I used it for.

For me, the human factors of the 8665A make it a non starter. It is huge and it weighs about 35kg and unless I 'needed' the performance of the YIG + FLL mode up at UHF then I can't see why I would want to have this monster taking up space on my workbench. However, it is an amazing piece of technology and I don't want to appear disrespectful to HP with these views.

However, I just noticed that the 4GHz SMHU is bigger than I thought. The 2GHz SMH we have at work is only about the same size as a Marconi 2019 but the 4GHz SMHU looks to be very big and is going to be very heavy too.

[G0HZU]

Here's an old plot of that Marconi 2024 at 1296MHz as a comparison. It still has that odd pip in the noise inside the loop BW and other 2023/4 generators I've tested don't show it.

But you can see that the 2024 is only average up here. I'd expect the ESGD to be about the same here (1296MHz) too although I can't find a plot of one.

However. if I had to choose between the ESG, the 2024, the SMHu and the 8665 for 'just' doing stuff up at UHF then I'd probably pay the extra for the ESGD or even consider an ESGA version if the price was right.

But that would be with a severe warning about its awful phase noise down at LF/HF frequencies

[G0HZU]

I went through my library of phase noise plots to see which other generators have poor LF/HF/VHF phase noise like the ESGD.

Probably the most relevant is the Aeroflex/IFR341x series as this is a vector sig gen. In this case I measured a 3413 model. See below for a plot of the IFR3413 at 10MHz. I think I plotted this on the same day I measured the Marconi 2024 way back in 2012. You can see that down at 10MHz this generator has close in phase noise that is about 40dB worse than the 2024 making it just as bad as the ESGD on the SW bands.

This range of generators looks very much like a facelifted Marconi 2024 but in reality this is a different animal inside. These generators support IQ modulation so they have to adopt a different frequency plan to include the IQ modulator.

In this case they use a similar scheme as the ESGD with a downmixing plan from UHF to LF to get the low frequency ranges meaning that these generators also have awfully high phase noise down on the SW bands. There is a reason for this and that is that in order to include vector modulation they can't have dividers after the modulator. Also the IQ modulator will probably only work well over a limited range of frequencies and this is usually done up at UHF.

So the (cost/size/power/weight) driven solution here is to throw away phase noise performance and just mix down from approx 1GHz to the low frequencies and avoid the use of dividers.

So in the case of generators like the ESGD (and the IFR) you get 1GHz related phase noise directly mapped down to (say) 10MHz. If you compute this as a 20*logN ratio you get 40dB. So it's no surprise that these compromised generators are about 40dB noisier than they could be down on the shortwave bands because they will have the same phase noise as the mixer drive signal up at around 1GHz.

However, the phase noise performance of the ESGD and the IFR generators will be about the same as the 2024 when used up at high VHF through UHF because the frequency plans will be reasonably equivalent in terms of noise contribution.

Compare below the IFR3413 to the Marconi 2024 plot down at 10MHz to see the 40dB difference

[davorin]

How much equipment do you own? (o;

[G0HZU]

How much equipment do you own? (o;

I do have quite a few sig gens here, but (sadly) the E5052A signal source analyser used to take the plots isn't mine and none of the generators in the E5052A plots are mine either. I do have a Marconi 2024 amongst my stash of sig gens but it isn't the one used for the plots.

The sig gen plots were taken at work using works equipment, usually at lunchtimes on days when I was using the E5052A for design work or verification

[dkozel]

Thanks to Dave posting about the NVIDIA auction in the UK, this thread just became of the utmost interest to me.

I won the SMIQ 03B with options B1 (OCXO), B11 (Data Generator), B20 (Modulation Coder), and B60 (Arbitrary Waveform Generator). I have to figure out the pickup, but will post my impressions here when I have it on my bench.

[G0HZU]

Well done on winning the auction

I also tried bidding on a different item in that auction but I think they have discarded my (winning?) bid because there has been no conformation yet. Did they send you a confirmation email that you had won?

I think the close in phase noise on the SMIQ will be better than the ESGD when down on the SW bands although the phase noise performance may not be significant to you anyway. But the ESGD is especially poor down here.

[dkozel]

No confirmation yet, but I had an email from them that confirmed I was an approved bidder and confirmed my phone number. I won two other items and can send the money today, so hopefully they'll let us know soon.

I'm not keenly focused on the phase noise, but we'll see how it goes. I have a nice GPSDO 10MHz/1PPS unit so will have to compare the references.

[G0HZU]

I tried emailing them wrt the auction and their reply was a bit vague. I guess they are very busy just after an auction..

Hi J....

Once the auction is  over you will receive an email with the Final Auction Results.

If  you were the higher bidder  you will be receiving a commercial invoice 3 business  days  after the auction  has taken place.

Our only payment  method  available for this event  is  Wire  transfer , the  payment instructions  will be displayed on the bottom part of  your  Invoice.  Let me know  if you have any  additional  doubts.

Regards,

I'm pretty sure I was the high bidder because the status for that auction ended within my bid range and the status for the item was 'auction ended'  with the same price indicated.

I guess I'll just have to wait and hope but it seems odd that they will accept large bids based on a simple online registration.

I saw the SMIQ 03 in the auction and a few other nice RF items but I only bid once on one big item. One bid one win hopefully

[Robert763]

Really old thread but I've been doing some related work recently.
I've been wanting to play with a vector RF generator for a while. It's not a need so was low on the budget list.
However I spottted a bit of a "sleeper" generator by R&S. Its a sleeper because its not a signal generator its a "Television Test Transmitter". Model SFQ. While basically a early digital TV test generator, the RF specification is very similar to the SMIQ-03  . 300kHz to 3.3GHz -13dBm max output etc. More importantly it has an external IQ modulation input as standard. No AWG internally though. So I eventually bought a AMIQ. That was a learning curve. It was listed as working by a ebay business TE seller. On arrival it didn't work so a discount was negotiated. Repairs included replacing all the electrolytic capacitors on the internal PC motherboard. cleaning the floppy drive and replacing the HDD with a SSD (mSATA in PATA adaptor 2.5" enclosure). And the result is a working Vector Signal Generator with AWG. Total outlay under £200 
There is another sleeper in the R&S TV TE range. The older SFZ. This has less frequency range (70-2025 MHz) and no IQ input BUT it will generate up to three carrriers at once (plus 3 sound subcarriers at fixed spacing if you want). This has interesting possibilities for intermodulation and out of band signal testing.

Robert.