SIGLENT SSA3032X input related spurious

[nugglix]

To make clear that the spurs are relative to the input signals I traced 2 signals, 2 MHz apart.

-20dBm signal levels, 10dB attenuator setting.

One can clearly see the 2 spurs and they're at the same ~ -70dBc level as before.

Done w/ the same SDG2042X as above, just using the channel combination feature.


PS: No witches involved, this is a witch free household.


Edit: yes, played w/ signal levels a bit and the spur behaves completely linear to the input signal strength.
Edit2: remove - before attenuator setting.

[Performa01]

FOI ~145 MHz @ -30 dBm

It’s interesting to see the spectrum above the carrier for once. As it seems, there is another spur at almost exactly 200MHz above the carrier, at a similar level of ~-70dBc.

Did anyone else see such a spur on their instrument and what’s its level on the new units?

[RFDUK]

Here is the second (NOV manufacture) unit on a wider span.

All's well above the wanted to better than -70dBC.

BUT at -60MHz -50dBC!

[G0HZU]

I remember looking at the teardown for this analyser and I think the ADC samples at 40Msps. So it would make sense if the final IF (IF3) was centred on 30MHz especially if the 810MHz IF2 SAW filter is as selective as the datasheet suggests. The teardown showed two 810MHz SAW filters in series and so the image rejection at a 60MHz offset should be really good here according to the datasheet response.

But this assumes that the final IF really is 30MHz. It could be 50MHz or 70MHz or even higher. If it was 30MHz then I guess that the first IF would be 3930MHz, the second IF 810MHz. LO2 would be 3120MHz and LO3 would be 3120/4 = 780MHz.

This would give a strong image response at -60MHz if there was something wrong with the filtering at the two 810MHz SAW filters at IF2? But all this is a guess. I don't know what the correct frequency plan for this analyser is. However, it should be possible to go sniffing with another analyser for the LO1, LO2 and LO3 frequencies... maybe hold a tiny loop probe (5mm dia loop across the end of some coax?) near the analyser input or the case. You would probably need a preamp and a narrow span and narrow RBW on the sniffing analyser and you would have to set the Siglent to a narrow span of <<1MHz to keep the LO1 in one place.

[Performa01]

I remember looking at the teardown for this analyser and I think the ADC samples at 40Msps. So it would make sense if the final IF (IF3) was centred on 30MHz especially if the 810MHz IF2 SAW filter is as selective as the datasheet suggests. The teardown showed two 810MHz SAW filters in series and so the image rejection at a 60MHz offset should be really good here according to the datasheet response.

But this assumes that the final IF really is 30MHz. It could be 50MHz or 70MHz or even higher. If it was 30MHz then I guess that the first IF would be 3930MHz, the second IF 810MHz. LO2 would be 3120MHz and LO3 would be 3120/4 = 780MHz.

This would give a strong image response at -60MHz if there was something wrong with the filtering at the two 810MHz SAW filters at IF2? But all this is a guess. I don't know what the correct frequency plan for this analyser is. However, it should be possible to go sniffing with another analyser for the LO1, LO2 and LO3 frequencies... maybe hold a tiny loop probe (5mm dia loop across the end of some coax?) near the analyser input or the case. You would probably need a preamp and a narrow span and narrow RBW on the sniffing analyser and you would have to set the Siglent to a narrow span of <<1MHz to keep the LO1 in one place.

That’s what I would have liked to do – estimating the frequency plan and speculating about the source of this spurious response, but to be honest, I just didn’t feel like watching the teardown again and finding the threads where this has been discussed back then, in order to get the required information. So I’d like to say many thanks to you for your analysis!

Yes, one could verify your assumptions by doing the measurements as you’ve suggested. Not many will have another appropriate instrument to do this though, I’m afraid. Moreover, all the ones who have discovered that their recently purchased SSA3kX has this flaw will most likely be very reluctant to open it up, but rather plan to send it back for repair or replacement…

[RFDUK]

Thanks very much G0HZU, I went in search of architecture and frequency info while the first instrument was here but only found the one hi res photo of the PCB.

I did connect a high gain preamp to HP 8596E and looked at the leakage on the TG and RF in ports several weeks ago. Found the 1st LO leakage on the TG port, TG off, if I remember correctly. It was at a very low level, so full marks to Siglent there. I didn't manage to work anything out and gave up.

Thanks for your speculation on the frequency plan. I'll try your suggestion of using a pickup loop with the preamps.

Someone speculated they may have cost engineered one SAW filter out and looking at the SAW filter data the -50dBC spur performance I see would fit that. One SAW filter is probably only some 10s of pence so unlikey when the -65dBC specification is not going to be met on paper. Perhaps it's a change of SAW filter manufacturer on the BOM that's caught them out.

I'll post LO activity details seen with a probe.

[RFDUK]

There's lots of case radiation to see to several GHz with a probe, presumably PLL products.

I searched for the final LO specifically and the most prominent clean narrowband signal I found was 750.000MHz. This one is quite loud.

Looked at the conducted ports again. The RF input is the most fruitful. Two components observed as follows, SA zero span ...

(Centre f) ... Sig 1 ... Sig 2
(100MHz) ... 915 ... 2025
(200MHz) ... 1013 ... 2070
(1000MHz) ... 1815 ... 2475

The frequencies are approximate since recorded from a broad span.

If the observed 750MHz is the final LO, final IF is 60MHz. Does the -60MHz spur fit that? ... it's not immediately obvious to me but I'll have a ponder on the possible mechanism.

[Performa01]

There's lots of case radiation to see to several GHz with a probe, presumably PLL products.

I searched for the final LO specifically and the most prominent clean narrowband signal I found was 750.000MHz. This one is quite loud.

Looked at the conducted ports again. The RF input is the most fruitful. Two components observed as follows, SA zero span ...

(Centre f) ... Sig 1 ... Sig 2
(100MHz) ... 915 ... 2025
(200MHz) ... 1013 ... 2070
(1000MHz) ... 1815 ... 2475

The frequencies are approximate since recorded from a broad span.

If the observed 750MHz is the final LO, final IF is 60MHz. Does the -60MHz spur fit that? ... it's not immediately obvious to me but I'll have a ponder on the possible mechanism.

Here are my thoughts so far (bear in mind, I’m not very experienced in defining frequency plans for triple conversion receivers, but anyway):

We do not know the first IF, nor the 3rd, but we do know the 2nd to be 810MHz.

The first IF has to be above the highest input frequency, so I’d say it has to be at the very least 3.6GHz (hopefully a good deal higher than that).

To me it’s quite obvious the problem cannot be in the first mixer, if only because the spur has a fairly constant offset to the input signal (within 2MHz) and is present for the entire input frequency range. Any unwanted mixer product would only fall within the first IF for a very narrow range of input frequencies. Furthermore, I would rule out that it’s a mixing product with any harmonic of the input signal, as various different signal sources have been used with quite similar results, and the HP8642B used by rf-loop has provided a decent spur suppression even though this RF-generator of all things exhibits rather high harmonic distortion.

So let’s look at the 2nd mixer.
We have to make an assumption for the first IF and this is important, as this determines the 2nd LO frequency and the spurious mixer products in turn.

Since we obviously want the 3rd LO to be easily derived from the 2nd LO, it should indeed be a 4:1 ratio as suggested by G0HZU and we need to consider the 3rd IF at this point as well.

And here the troubles start. A 3rd IF of 30MHz could explain an unwanted image response, but that’s not what we’re seeing here. It rather looks like the spurious frequency moves twice as fast as the signal frequency, hence it has to be a 3rd (or 4th?) order intermodulation product, so the 3rd IF could be anything but 30MHz and 60MHz would be rather plausible.

Also, with the frequency plan posted by G0HZU, we’d never get out of the reverse frequency position of the IF signal, but that does not change with IF frequency, we would rather have to give up the 4:1 ratio for the 2nd and 3rd LO.

Having played with the numbers a bit (unfortunately I don’t have any fancy software for that), I just didn’t find a solution yet. Independent of your measurements, I’ve landed at 60MHz IF3 and 750MHz LO3 at some point, but still didn’t get anywhere. Because of the symptoms, I firmly believe that it has to be some unwanted mixing product with the 2nd harmonic of the 2nd IF signal. A 4th order mixing product looks good at first glance, but then I believe the spur would appear 30MHz off the carrier and not 60MHz.

In any case I am still puzzled and would not rule out that it’s not a problem with the 2nd IF filter, but maybe the 3rd mixer. Maybe they have used a double balanced mixer initially and thought they swap it out for a single diode – LOL. Just kidding…

[Bicurico]

Hi,

Out of curiosity I used my "VMA Simple Spectrum Analyzer" software with the 35MHz-4.4GHz SMA device and reproduced the same settings as in the first post of this thread.

Attached is the result. Mine is of the older SSA3021X.

Regards,
Vitor

[G0HZU]

Performa01: And here the troubles start. A 3rd IF of 30MHz could explain an unwanted image response, but that’s not what we’re seeing here. It rather looks like the spurious frequency moves twice as fast as the signal frequency, hence it has to be a 3rd (or 4th?) order intermodulation product

Don't forget that the analyser is running in an FFT mode so the span you see on the display isn't a linear sweep with a fixed IF of (say) 30MHz with a 1kHz 'digital' RBW. It's a patchwork of FFTs maybe 2MHz wide made up of thousands of FFT bins. The first LO will step in fixed steps as each FFT is taken across the sweep. This means that the image offset will be a function of the offset from the centre of each FFT. So the spur won't always appear 'exactly' at a 60MHz offset. However it will be close to 60MHz and can be expected to wrap around inside a window of a few MHz. A bit like TurboTom described in post #13?

We do not know the first IF, nor the 3rd, but we do know the 2nd to be 810MHz.

We don't even know if the IF2 is 810MHz for certain. It could be 812MHz or 814MHz for example. They might run the IF at a slight offset to the centre frequency of the 810MHz SAW filter. Also, I've only got a few images of the main board to work with and I'm guessing that the ADC samples at 40MHz. But it might sample at some other frequency and this would throw off all my guesstimates

(Centre f) Sig 2
(100MHz)  2025
(200MHz)  2070
(1000MHz) 2475

The LO1 synthesiser uses a doubler so double these numbers and you get:

(Centre f) Sig 2  (difference =IF1?)
(100MHz)  4050     3950MHz
(200MHz)  4140   3940MHz
(1000MHz) 4950   3950MHz

So the first IF could be close to 3950MHz. Can you zoom in on the 2025MHz signal (at CF 100MHz) and measure it more accurately?

[G0HZU]

(Centre f) ... Sig 1 ... Sig 2
(100MHz) ... 915 ... 2025
(200MHz) ... 1013 ... 2070
(1000MHz) ... 1815 ... 2475

(2025 * 2) - 915 = 3135MHz = LO2?
(2070 * 2) - 1013 = 3127MHz = LO2?
(2475 * 2) - 1815 = 3135MHz = LO2?


It's worth measuring all these frequencies accurately so we can get closer to the frequency plan. Note that they will probably shift the first IF in subtle ways to avoid major internal spurious terms.
So IF1 probably won't be fixed and this means that IF2 won't be fixed inside the 810MHz SAW filter. It could vary depending on the analyser frequency. They could be shifting/sidestepping the IF frequency several MHz to avoid internal (nLO1 +/- mLO2 +/- xLO3 = IF3) spurious terms on certain frequencies.

[rf-loop]

Hi,

Out of curiosity I used my "VMA Simple Spectrum Analyzer" software with the 35MHz-4.4GHz SMA device and reproduced the same settings as in the first post of this thread.

Attached is the result. Mine is of the older SSA3021X.

Regards,
Vitor

Where from this enormous shocking noise level is coming from.  I belive this around -60MHz spur is still there but your  noise covers it totally (because your spur level is inside specs. Also your test signal carrier level is bit low from -30dBm mixer level).

[RFDUK]

Thanks for all the input, very interesting.

I measured what must be the 1st LO more accurately ...
(100MHz) ... 2.015 x 2 = 4.03
(200MHz) ... 2.065 x 2 = 4.13
(1000MHz) ... 2.465 x 2 = 4.93

So confirmation of a 3930MHz 1st IF

[RFDUK]

I'm thinking the 750MHz I saw on the pick up loop wasn't 3rd LO after all ...

1st IF is 3930MHz.
2nd LO is 3120MHz.
3rd LO is 780MHz.
3rd IF is 30MHz.

[G0HZU]

If I had to guess what was 'wrong' with the analysers affected with this problem I would include the possibility that there is inconsistency in the way the big aluminium metalwork screen 'fits' to the PCB.
If the metalwork mating surface has highspots on it or maybe some contamination then this can mean the metalwork can become a leakage path rather than just a screen. Or maybe there needs to be some RF absorber here inside the metalwork pocket above the SAW filters?

If there is a leakage path that can bypass the SAW filters (via the metalwork) then the effective performance of the two cascaded SAW filters can be ruined very easily and you could easily get the problem as per the image below. The green trace might represent the full potential of the two filters but the red trace could represent what happens with a leakage path.

This would only have an IF3 image rejection of 55dB for example.

[RFDUK]

Yes, cascaded SAW filters looking for >70dB stop band could be a bit touchy.

I have a test that is quite convincing on the 3rd LO frequency determination ...

Using a sensitive comms receiver with SSB demod tuned to 780MHz, firstly with quarter wave whip placed around the case of the SSA. I heard a clean carrier but also noted carriers at 10MHz intervals above and below.

I connected the comms RX directly to the SSA RF input and only hear the 780MHz and at reasonable strength too.

[Performa01]

Where from this enormous shocking noise level is coming from.

Maybe from ref. level = 0dBm and attenuator = 20dB? (hint, hint, for Bicurico )

[Performa01]

Performa01: And here the troubles start. A 3rd IF of 30MHz could explain an unwanted image response, but that’s not what we’re seeing here. It rather looks like the spurious frequency moves twice as fast as the signal frequency, hence it has to be a 3rd (or 4th?) order intermodulation product

Don't forget that the analyser is running in an FFT mode so the span you see on the display isn't a linear sweep with a fixed IF of (say) 30MHz with a 1kHz 'digital' RBW. It's a patchwork of FFTs maybe 2MHz wide made up of thousands of FFT bins. The first LO will step in fixed steps as each FFT is taken across the sweep. This means that the image offset will be a function of the offset from the centre of each FFT. So the spur won't always appear 'exactly' at a 60MHz offset. However it will be close to 60MHz and can be expected to wrap around inside a window of a few MHz. A bit like TurboTom described in post #13?

We do not know the first IF, nor the 3rd, but we do know the 2nd to be 810MHz.

We don't even know if the IF2 is 810MHz for certain. It could be 812MHz or 814MHz for example. They might run the IF at a slight offset to the centre frequency of the 810MHz SAW filter. Also, I've only got a few images of the main board to work with and I'm guessing that the ADC samples at 40MHz. But it might sample at some other frequency and this would throw off all my guesstimates

I should have taken all this into account - but I got carried away quite obviously. Thank you for pointing this out and tossing me back on track!

With these modern processor controlled equipment it's really hard to tell what's actually going on inside...

[Performa01]

If there is a leakage path that can bypass the SAW filters (via the metalwork) then the effective performance of the two cascaded SAW filters can be ruined very easily and you could easily get the problem as per the image below. The green trace might represent the full potential of the two filters but the red trace could represent what happens with a leakage path.

It really is mazing to see what fancy tools RF professionals have at their disposal nowadays!

[G0HZU]

This may be a bit off topic, but when I looked at the teardown pictures I also spotted the IF3 bandpass filter on the PCB and this lumped LC filter appears to use a fairly classic 'symmetry preserving' topology. Dave's closeup images show the value of most of the inductors so I was able to reverse engineer the probable response of this filter. Obviously, I can't tell what the capacitor values are but I can get some idea from the colour of the dielectric which ones are below about 200pF. I also had to use fixed capacitor values and fixed inductors so this made it easier to guess the filter components.

This filter has to be symmetrical because it has alias terms to reject either side of it and that is why it uses this odd looking topology. See below for my first crude attempt at predicting the response assuming it really is a 30MHz IF at IF3. Because I know the inductor values and I have experience designing filters with this topology I can guess the values of the caps fairly well. But I could still be very wrong here!

Assuming I've got close to the actual design values you can see that the designer was aiming for over 80dB alias rejection ahead of the ADC. The 810MHz SAW filter effectively helps a bit here too. I've drawn the alias responses by hand in purple at about -80dBc.

Note that I've swapped around the order of the series L and C components in my CAD version but this is because this is the way they get produced in my synthesis/optimiser tool. But it will not affect the response.

[G0HZU]

If you look at Dave's image of the PCB in the post above you can see there are a lot of 'no fit' SMD parts on the PCB around this IF3 BPF. It looks like the designer considered having several switched IF3 bandpass filters. This could have meant several 30MHz filters with different bandwidths could have been fitted (selected in firmware?) or maybe an alternative filter at the 50MHz alias to act as a spurious cancelling option for certain types of internal spurious terms. But for whatever reason, these parts were not fitted to the PCB. Maybe this was for cost or maybe the selection switches had limited isolation and spoiled the performance in some way.

Just below the yellowed outline of IF3 in the PCB image are the two 810MHz SAW filters. They live in a tiny pocket (circled in blue below) but maybe the grounding to the metalwork is not good here? The lid/screen screws are a fair way away from this pocket and there is an isolation slot cut into the PCB here as well. So they were obviously trying for isolation here but maybe it isn't repeatable with the metalwork? So signals can bypass the SAW filters and get into the (ADE-2) third mixer?

[Bicurico]

Where from this enormous shocking noise level is coming from.

Maybe from ref. level = 0dBm and attenuator = 20dB? (hint, hint, for Bicurico )

Yes, I overlooked that...

Here the same screenshot, but with Attenuator = 0dB.

Regards,
Vitor

EDIT: Changed picture with exactly the same settings as RFLoop's first post.

[RFDUK]

Where from this enormous shocking noise level is coming from.

Maybe from ref. level = 0dBm and attenuator = 20dB? (hint, hint, for Bicurico )

Yes, I overlooked that...

Here the same screenshot, but with Attenuator = 0dB.

Regards,
Vitor

EDIT: Changed picture with exactly the same settings as RFLoop's first post.

Hello Vitor,

Yes that's fine, very similar to other older instruments and nearly 3dB inside Siglent specification.

Regards, Martyn.

[Performa01]

Here the same screenshot, but with Attenuator = 0dB.

Regards,
Vitor

EDIT: Changed picture with exactly the same settings as RFLoop's first post.

Almost there … this test was meant to be performed with a -30dBm mixer input level. So in order to have a completely comparable result, you should either lower the signal level to -30dB or set the attenuator to 10dB.

On the other hand, current hypothesis states that the spurious response 60MHz below the input signal is not a distortion product, but some sort of image response, so it might not make a difference.

Anyway, we can see lots of spurs here, but that might just be the signal source.
The relevant spur at 85MHz is only some -62dBc, so your instrument might  be slightly out of spec. as well.

To be sure about that, please add 10dB attenuation and repeat this very test.

Oh – you have changed settings while I wrote the above post. Still not completely there though, as the signal level should be -30dBm at the mixer input and that means you need a -20dBm signal with 10dB attenuator. Your signal level appears to be about -14dBm.

Still interesting to see that the spurious response is within specs now.

This would be a hint that it might be a distortion product after all!

EDIT: aah - and all the other spurs have vanished in the new screenshot, so the analyser was indeed overloaded at the first test.

[Bicurico]

Have patience with me... it was early in the morni by and I had little time to grab the pictures, before going to work...

Cheers,
Vitor