SSA2021X - 20 dB attenuator sufficient?

[fnaumann]

I recently bought an expensive toy - a Siglent SSA2021X spectrum analyzer. I don't want to harm this not only because of the high purchase price. Hence my question, for simple tests and measurements on a Breadboard and with a signal generator in the 5 to 12V DC range, is a 20 dB attenuator sufficient as protection?

SSA2021X RF input
50V DC max - I have as additional protection a DC blocker
+30dBm max - which is one Watt (if I remember right)

With -20 dB from my external Siglent attenuator (20dB@4GHz) the SSA2021X should not be damaged till +50dBm ? Or is this false?

With my Oszilloskope the matter is much more relaxed, with 1 MΩ Input (instead 50Ω on the SSA), and the 1:10 probe you are relatively safe here with voltages not exceeding 24 V. I have no practical experience with specific RF performance. I don't want to analyze transmitters either, but rather interference signals in circuits, RF noise in my home office (switching power supplies, lamps).

[Kim Christensen]

With -20 dB from my external Siglent attenuator (20dB@4GHz) the SSA2021X should not be damaged till +50dBm ? Or is this false?

Partially true. +50dbm is 100 watts, so your 20db attenuator would also have to be rated for 100W. If it's not, and it burns out when you put 100W into it, you might get more than +30dbm into your SA.

12VDC into 50 ohms is 2.9 watts, so that would damage the typical 1W attenuator, but your SA is rated for 50Vdc with the DC blocker so it should be OK in that scenario.

[fnaumann]

The situation is worse than expected.

"The maximum input values of +30 dB only apply if all attenuators are present in the device are raised to maximum."

https://www.microcontroller.net/topic/521714

I read on the internet that the direction of attenuators also matters and that you shouldn't buy cheap Chinese parts. Regardless of the fact that the expensive equipment sold in Europe is also produced in China, my "Siglent ATT-20 20 dB Attenuator" (https://www.siglent.eu/product/1141970/siglent-att-20-20-db-attenuator) has no label which direction is the correct one. Similar issue with my BLK-222+ DC blocker.

The Tracking generator of the Siglent SSA2021X

Output level -20 dBm~0 dBm
Output level resolution 1 dB
Output flatness +/-3 dB

should not harm when connect to the input ?

[Kim Christensen]

There should be no problem connecting the output of the SSA2021X's tracking gen to the SSA2021X's input. Usually this is how you calibrate it anyway.

I can see direction being an issue with some high power attenuators, but not with a 1W one or less.
For example, if the input is on the left you can see there'd be more power dissipated in R1 than in R3. So on some larger attenuators I can see them making R1 higher wattage than R3. But on a low power one, they are usually the same wattage and thus 100% bi-directional:

[fnaumann]

There should be no problem connecting the output of the SSA2021X's tracking gen to the SSA2021X's input. Usually this is how you calibrate it anyway.

That was my first thought, but the many warnings I found on the Internet about spectrum analyzers confused me.

I can see direction being an issue with some high power attenuators, but not with a 1W one or less.

Here (see attached picture) are two of my 20 dB, 50Ω  which covers a frequency range till 4GHz. One slim (with BNC connectors) for 1W (30 dBm ?) and a fat one for 10W. If I understand you right, I should not use the 10W attenuator as input protection for my SSA2021X?

based on https://www.siglenteu.com/wp-content/uploads/dlm_uploads/2017/10/SSA3000X_DataSheet_DS0703X-E04A.pdf

Maximum average RF power 30 dBm, 3 minutes, fc≥10 MHz, attenuation >20 dBm, preamp off
Maximum damage level 33 dBm, fc≥10 MHz, attenuation >20 dBm, preamp off

The output level of the Tracking Generator looks save

Output level -20 dBm~0 dBm


I know my questions are at Newbi level. Could we compare it with the DC area. Two 1Ω resistors, one with 5W and the other with 50W. If the voltage and therefore the current flow becomes too high, the resistor burns out, depending on its respective power value. When this burns down, the bond is dead and no damage can be done to the other side of the resistor. If the same applies to attenuator, what's the problem? Or does the attenuator then switch fully on without any attenuation?

[Kim Christensen]

Here (see attached picture) are two of my 20 dB, 50Ω  which covers a frequency range till 4GHz. One slim (with BNC connectors) for 1W (30 dBm ?) and a fat one for 10W. If I understand you right, I should not use the 10W attenuator as input protection for my SSA2021X?

The 10W probably doesn't matter which way you connect it. If it did matter, they would label the input and output. It's spec sheet would also tell you if it mattered.
It should be safe to use for situations where you expect less than 10W of RF power.

Here's an example of two attenuators that have a designated input and output:

[tautech]

There should be no problem connecting the output of the SSA2021X's tracking gen to the SSA2021X's input. Usually this is how you calibrate it anyway.

That was my first thought, but the many warnings I found on the Internet about spectrum analyzers confused me.

As you have the analyzer on front of you, you need take closer note of the info on the display.

Input attenuation is by default on Auto and normally 20dB or more depending on input settings.
The analyzer automatically deducts internal attenuation settings from settings to display the correct signal value even when the internal attenuation is set manually.

Add external attenuation and you need change the Reference level to account for it for the measurements to be correct.
Best advise is to have a play with the TG signal which of course you can set the level and then add external attenuation and change the reference level to get a correct result.
It will soon appear to all be quite simple and straightforward to get reliable results.

Once you go to just RF In signals you need be more careful as even a mobile phone nearby handshaking with a cell tower can send input levels beyond safe levels where you will be warned by the internal overload buzzer and OSD warnings. This is especially true when you have the Preamp set to ON which is when you need work with care.

Little steps and have fun getting to know your instrument.

[blackdog]

Hi,

And there are other things to consider when using analyzers.
I normally have -6dB or -10dB wide bandwidth attenuators mounted on the two connections of my spectrum analyzer.

This is for two reasons, that is some protection, I almost never measure on high power amplifiers and secondly
to make sure that the connections of the HF analyzer become nicer 50 Ohm.

But then if I do need less attenuation, I adjust the attenuation or remove the extra attenuator, if it cannot harm the instrument.

With HF analyzers, it is better to think first, then think again and can I do what I thought of?
That usually keeps your equipment in one piece. 

Kind regards,
Bram

[fnaumann]

Once you go to just RF In signals you need be more careful as even a mobile phone nearby handshaking with a cell tower can send input levels beyond safe levels where you will be warned by the internal overload buzzer and OSD warnings. This is especially true when you have the Preamp set to ON which is when you need work with care.

This means that if I want to use my Siglent SSA to measure the RF signals in my apartment (e.g. local WiFi and 5G signals), connecting an antenna directly to the SSA can destroy the spectrum analyzer if a smartphone is nearby? Is a 10 dB attenuator enough to avert this danger?

[blackdog]

Hi fnaumann, 

If a 10dB attenuator is enough to keep the signals you want to see from dropping into noise, then I would use the 10dB attenuator.

I don't know your environment, if you have a hobby room surrounded by your family who each have a cell phone then the power you get in at an unfavorable time can be quite high.
Also if you have radio amateurs nearby the power if you mount an antenna on your SA can be a problem.
Your 10dB attenuator attenuates the power 10x, with a 20dB attenuator it is 100x, but it may be if you were to use a 20dB attenuator that the signals you want to see disappear into the noise.
The point is, that you work "wisely", better to work with -20dB first, see if you can make sensible measurements, too much in the noise, then if you do not see very strong signals, use the -10dB attenuator.

Kind regards,
Bram

[fnaumann]

I don't know your environment, if you have a hobby room surrounded by your family who each have a cell phone then the power you get in at an unfavorable time can be quite high.
Also if you have radio amateurs nearby the power if you mount an antenna on your SA can be a problem.

A few iPhones are flying around near me. And the neighbors next to and below me will also use smartphones. No radio amateurs as far I know.

The point is, that you work "wisely", better to work with -20dB first, see if you can make sensible measurements, too much in the noise, then if you do not see very strong signals, use the -10dB attenuator.

This annoys me as a spectrum analyzer beginner - the thing is apparently so sensitive that you have to handle it with kid gloves. An oscilloscope is a much more robust part, especially if you use a differential probe (and do not need to cover a common ground). I'm honestly disappointed with the SSA use cases shown in YouTube videos. I can do a lot of antenna-related things much more relaxed with my NanoVNA-F V2, and other things with my SDR boxes. Above all, the FM and AM demodulation function on the SSA is a joke (completely unnecessary in my opinion). And then I now have a TinySA Ultra - which in one of the videos even offers a smaller noise floor. I know I can not compare a TinySA Ultra with a Siglent SSA2021X , but ....

[blackdog]

Hi fnaumann,

The point is to be careful, my spectrum analyzer has a limit of 100mW at its input.
If you put a pickup antenna next to a mobile phone, you can get over this limit.
The phone transmits not only when you make a call, it does this just when it feels like it. :-)

But don't get paranoid, it's all about being careful with your measuring equipment.
The 10 0f 20dB attenuator at the input you start with during a measurement session normally makes the SA indestructible.
( if you ar not doin measurement on tranmitters!)

The field of an antenna connected to a transmitter decreases quadratically with distance.
So if the distance to a phone is 4-meters, and you double this to 8 meters, then only 1/4 is left.

The manufacturers of this kind of measuring equipment have to perform tricks to get a good noise number, good distortion value and then also have a "decent protected" input, it is not easy.

Greetings,
Bram

[EggertEnjoyer123]

I don't know your environment, if you have a hobby room surrounded by your family who each have a cell phone then the power you get in at an unfavorable time can be quite high.
Also if you have radio amateurs nearby the power if you mount an antenna on your SA can be a problem.

A few iPhones are flying around near me. And the neighbors next to and below me will also use smartphones. No radio amateurs as far I know.

The point is, that you work "wisely", better to work with -20dB first, see if you can make sensible measurements, too much in the noise, then if you do not see very strong signals, use the -10dB attenuator.

This annoys me as a spectrum analyzer beginner - the thing is apparently so sensitive that you have to handle it with kid gloves. An oscilloscope is a much more robust part, especially if you use a differential probe (and do not need to cover a common ground). I'm honestly disappointed with the SSA use cases shown in YouTube videos. I can do a lot of antenna-related things much more relaxed with my NanoVNA-F V2, and other things with my SDR boxes. Above all, the FM and AM demodulation function on the SSA is a joke (completely unnecessary in my opinion). And then I now have a TinySA Ultra - which in one of the videos even offers a smaller noise floor. I know I can not compare a TinySA Ultra with a Siglent SSA2021X , but ....

You can get a lower noise floor by turning the resolution bandwidth down. (The tradeoff is that it'll take longer to sweep).
For low frequency stuff (like maybe under 50 MHz), a scope is usually better, but once you get to high frequencies scopes become prohibitively expensive. Also, high frequency scopes have 50 ohm inputs too and most of them can take even less power than your spectrum analyzer.

If you want to analyze noise around your house you should look at making a near field probe (There's a EEVBlog video on making your own out of coax). Connecting that directly to the spectrum analyzer should be perfectly fine (as long as you don't put it in the microwave or something).

[fnaumann]

You can get a lower noise floor by turning the resolution bandwidth down. (The tradeoff is that it'll take longer to sweep).

I know this. Watch this starting at minute 16.

[fnaumann]

I did my own testing with my SSA2021X and TinySA Ultra using a test signal from my SDG 1032X.

The signals were clearly visible on the TinySA Ultra out of the box. With the SSA2021X, the signal was also visible after various adjustments, but with significantly more noise. The signal/noise ratio seems scary to me as a layman. I tried to set the SSA2021X using various options so that I could see the test signal better on the display. As a test signal I used a simple 10 MHz sine wave signal with 1 to 5 dBm output level (SDG 50 ohm output mode).
To protect the SSA, I installed a 30 dB attenuator (10 W) plus a DC blocker. I reduced the internal attenuation to 0 dB (after the signal-to-noise ratio didn't improve much). Which doesn't make it much better. I also tried different bandwidth settings.

Then when I activated the internal preamp, the SSA beeped and warned of an overload situation. What crap. The SSA is specified as a maximum of 30 dBm and 50 V DC. Of course, these values are only given for special settings. But - I had used a maximum of a 5dBm signal with 30dB of external attenuation, which corresponds to a -25dBm input signal? How can a -25dBm signal produce an overload warning on the SSA? I only activated the PreAmp option under “Amplitude” without setting a manual value. Shouldn't the software prevent overload?

[tautech]

I did my own testing with my SSA2021X and TinySA Ultra using a test signal from my SDG 1032X.

The signals were clearly visible on the TinySA Ultra out of the box. With the SSA2021X, the signal was also visible after various adjustments, but with significantly more noise. The signal/noise ratio seems scary to me as a layman. I tried to set the SSA2021X using various options so that I could see the test signal better on the display. As a test signal I used a simple 10 MHz sine wave signal with 1 to 5 dBm output level (SDG 50 ohm output mode).
To protect the SSA, I installed a 30 dB attenuator (10 W) plus a DC blocker. I reduced the internal attenuation to 0 dB (after the signal-to-noise ratio didn't improve much). Which doesn't make it much better. Then when I activated the internal preamp, the SSA beeped and warned of an overload situation. What crap. The SSA is specified as a maximum of 30 dBm and 50 V DC. Of course, these values are only given for special settings.

BUT - I had used a maximum of a 5dBm signal with 30dB of external attenuation, which corresponds to a -25dBm signal. How can a -25dBm signal produce an overload warning on the SSA? I only activated the PreAmp option under “Amplitude” without setting a manual value. Shouldn't the software prevent overload?

TBH this is incorrect methodology.
Instead you should let the SA manage internal attenuation and add external attenuators when/if necessary.
External attenuation can then be nulled/corrected with use of the amplitude offset feature.

Post a screenshot up and lets compare settings.

[fnaumann]

Instead you should let the SA manage internal attenuation and add external attenuators when/if necessary.
External attenuation can then be nulled/corrected with use of the amplitude offset feature.

I assumed that the spectrum analyzer internal attenuation is one of the tunable parameters of a spectrum analyzer, which should be adjusted by the User if necessary, and the external attanuators are intended to protect the spectrum analyzer from power overload (worst case destruction). Since in many cases you don't know exactly how high the signal level to be measured is (dynamic range of the signal), I assumed that you always have to attenuate a little more generously in order to protect the device.

[tautech]

Instead you should let the SA manage internal attenuation and add external attenuators when/if necessary.
External attenuation can then be nulled/corrected with use of the amplitude offset feature.

I assumed that the spectrum analyzer internal attenuation is one of the tunable parameters of a spectrum analyzer, which should be adjusted by the User if necessary, and the external attanuators are intended to protect the spectrum analyzer from power overload (worst case destruction). Since in many cases you don't know exactly how high the signal level to be measured is (dynamic range of the signal), I assumed that you always have to attenuate a little more generously in order to protect the device.

Certainly so depending on what you are doing however don't underestimate how sensitive a SA is, easily equivalent to any radio.
My first a SSA3032X, scared me regarding this sensitivity and how careless use could set the overload buzzer off so now the cellphone is never nearby when sensitivity is increased or the Preamp is on.
Later I got a SVA1015X and then upgraded to SVA1032X which remains my fav RF instrument where its SA capability is identical to SSA models.

In the early days I was terrified of front end damage yet now on my 3rd none have ever been damaged however anything that might offer a strong signal is first assessed with a simple non-contact RF loop.
Apart from those early days until I found my feet I no longer use external attenuators but have a small selection of N, SMA and BNC types should the need for them arise.

[fnaumann]

Certainly so depending on what you are doing however don't underestimate how sensitive a SA is, easily equivalent to any radio.

A measuring device that declares a max input level of 30 dBm and then issues an overload warning when the signal is -55 dB (E -5) weaker is suspect to me. And if a internal preamplifier activation in a menu can kill the device - such an option should not be able to be activated or should turn off by itself if it get critical. Surely that should be possible with such an expensive measuring device?

And let's assume that I'm too stupid to be able to display the signal sensibly on the SSA2021X , then the automatic function "Auto Tune" - offered in a separate button - should be able to display a measurement signal reasonably sensibly (signal-to-noise ratio) and not as a small bump surrounded by a sea of noise. And why is this all possible out of the box for a device that only costs a fraction?

[tautech]

Some simple examples with 10dB attenuation on RF In using the default -20dB TG signal which of course will display ~-30dB sweep.



Then with an offset correction added and nothing else changed to now display a -20dB sweep....the TG level and with the 10dB attenuation corrected/nulled.



This is a simple and quick trick using the TG source to check you have set things correctly before proceeding to sniff anything of higher power.
When done this way Markers will display the correct amplitude values without any need for 'in the head' calculations.

[fnaumann]

Before I use newly purchased attenuator, I checked them with the TG. Especially their attenuation over the entire available frequency range. I don't do that before every measurement. And I know the meaning of Ref Offset.

The advantage of fixed attenuators is safety. With variable attenuators, I am always afraid that they will accidently change their attenuator value and then kill the SA during a measurement.

[tautech]

Certainly so depending on what you are doing however don't underestimate how sensitive a SA is, easily equivalent to any radio.

A measuring device that declares a max input level of 30 dBm and then issues an overload warning when the signal is -55 dB (E -5) weaker is suspect to me. And if a internal preamplifier activation in a menu can kill the device - such an option should not be able to be activated or should turn off by itself if it get critical. Surely that should be possible with such an expensive measuring device?

And let's assume that I'm too stupid to be able to display the signal sensibly on the SSA2021X , then the automatic function "Auto Tune" - offered in a separate button - should be able to display a measurement signal reasonably sensibly (signal-to-noise ratio) and not as a small bump surrounded by a sea of noise. And why is this all possible out of the box for a device that only costs a fraction?

As yet we have not seen any screenshots from you showing settings that might trigger warnings.
FYI max continuous signal is specified as: 30 dBm, 3 minutes, fc ≥10 MHz, att > 20 dBm, preamp off
This is of course without external attenuators.

As yet I prefer to control any instrument rather than let an Autotune process second guess my requirements therefore I have never used it, maybe I should get some experience with it. 

[fourfathom]

In many cases an "overload" indication doesn't mean a damaging-level input, but just that the input level with the current attenuation and gain settings results in a signal that is causing the detector to clip or distort.  It's not an input power problem, just a dynamic range problem that can be corrected by adjusting the attenuation or gain settings.  No harm done.

[pdenisowski]

Once you go to just RF In signals you need be more careful as even a mobile phone nearby handshaking with a cell tower can send input levels beyond safe levels where you will be warned by the internal overload buzzer and OSD warnings. This is especially true when you have the Preamp set to ON which is when you need work with care.

If I remember correctly, max UE transmit power in LTE and 5GNR is +23 dBm.  I lot of things would have to go wrong (so to speak) at the same time in order to see even +20 dBm at the analyzer input. 

And it's not just handshaking (I assume you mean RACH): theoretically the base station could power control the UE up to max transmit power at any time, although this is also very uncommon. I used to do cellular drive test measurements and I'm not sure I ever saw a UE power controlled up to max Tx power.

That said, a nearby UE can absolutely overload the front end of an analyzer and cause distortion / intermod:  I saw that many times while doing interference hunting in the field.  But I've never heard of an analyzer being damaged by OTA signals from a nearby UE.

Damage from overload when directly connected with cable?  Oh yes

[pdenisowski]

In many cases an "overload" indication doesn't mean a damaging-level input, but just that the input level with the current attenuation and gain settings results in a signal that is causing the detector to clip or distort.  It's not an input power problem, just a dynamic range problem that can be corrected by adjusting the attenuation or gain settings.  No harm done.

Yes, exactly this.  In fact, often I've seen what is clearly distortion caused by an overdriven input even BEFORE any kind of "overload" indication appears.

[fourfathom]

In many cases an "overload" indication doesn't mean a damaging-level input, but just that the input level with the current attenuation and gain settings results in a signal that is causing the detector to clip or distort.  It's not an input power problem, just a dynamic range problem that can be corrected by adjusting the attenuation or gain settings.  No harm done.

Yes, exactly this.  In fact, often I've seen what is clearly distortion caused by an overdriven input even BEFORE any kind of "overload" indication appears.

Yes.  The active stages of a spectrum analyzer will always distort, but it isn't until this distortion exceeds the measurement noise floor that you notice it.  There is compression, leading to clipping -- that is usually obvious and the analyzer may warn you about it.  But there is also intermodulation distortion that shows up as sum and difference mixing products.  And in newer gear there are sampling artifacts.  If you don't understand these you can reach some very wrong conclusions about what the analyzer is showing you.

[fnaumann]

As yet we have not seen any screenshots from you showing settings that might trigger warnings.
FYI max continuous signal is specified as: 30 dBm, 3 minutes, fc ≥10 MHz, att > 20 dBm, preamp off

I used a simple 10 MHz sine wave signal with 1 to 5 dBm output level (SDG 50 ohm output mode). 30 dB attenuator (10 W) plus DC blocker was connected externally upstream of the SSA. In the SSA, the internal attenuator was deactivated in the "Amplitude" menu (set manually to 0dB) and the trigger for the overload warnings was the activation of the internal preamplifier (Preamp = On). Of course, I didn't have time (and the motivation) to take a screenshot when the warning signal sounded. And I don't want to consciously reproduce these overload situations in order to produce screenshots for the forum.

As yet I prefer to control any instrument rather than let an Autotune process second guess my requirements therefore I have never used it, maybe I should get some experience with it. 

I also prefer to set my measuring instruments myself. Only when nothing works (no useful signal is shown) at all do I occasionally press "Autoset" on my Owon Oszilloskops to check whether there is even a displayable signal. On the other hand, I used the "Auto Tune" on the SSA 3021X Plus more often because the signals displayed look ugly via manual settings and didn't really get any better even after a lot of tweaking. With the TinySA Ultra, on the other hand, the user can see all the relevant signals without having to strain too much. And the TinySA Ultra also has complex setting options. Considering the large display and the very good operating concept of the Siglent SSA, the signals displayed are disappointing for me as a beginner. Or my expectations are too high and unrealistic. I'm going to look into my spectrum analyzer in more detail, simply because the thing was so expensive ...

[fnaumann]

Your 10dB attenuator attenuates the power 10x, with a 20dB attenuator it is 100x, but it may be if you were to use a 20dB attenuator that the signals you want to see disappear into the noise.
The point is, that you work "wisely", better to work with -20dB first, see if you can make sensible measurements, too much in the noise, then if you do not see very strong signals, use the -10dB attenuator.

I usually start with a 30 dB attenuator. I suspect that due to my overprotecting use of attenuators, the signals on the SSA2021X Pro are so weak that the signal-to-noise ratio is relatively low and therefore the display of the signals is suboptimal.

[tautech]

Your 10dB attenuator attenuates the power 10x, with a 20dB attenuator it is 100x, but it may be if you were to use a 20dB attenuator that the signals you want to see disappear into the noise.
The point is, that you work "wisely", better to work with -20dB first, see if you can make sensible measurements, too much in the noise, then if you do not see very strong signals, use the -10dB attenuator.

I usually start with a 30 dB attenuator. I suspect that due to my overprotecting use of attenuators, the signals on the SSA2021X Pro are so weak that the signal-to-noise ratio is relatively low and therefore the display of the signals is suboptimal.

Here is where you need to wind down the RBW and those faint signals will magically rise from the noise floor. 
Leave RBW/VBW at 1:1 for now.
The penalty is Sweeps will take longer however that can also be addressed by narrowing the sweep.

A good place to pick up some SSA tips/tricks is in the SSA3021X thread and posts from rf-loop starting from here:
https://www.eevblog.com/forum/testgear/siglent-ssa3000x-spectrum-analyzers/msg958295/#msg958295

[fnaumann]

compare TiniSA Ultra vs. SSA2021X Plus

Test signal: simple 10 MHz sine wave signal with 0 dBm output level (Siglent SDG1032X in 50 ohm output mode)
External attenuation: 20dB plus DC blocker

CF: 10MHz
Span: 1MHz
RBW: 10KHz (and below for SSA2021X Plus)

With the TinySA Ultra, the noise floor is below -70 dBm at 10 kHz RBW and there is a clear demarcation between signal and the noise floor. For the SSA2021X Plus with the same RBW value, there is (for me) no clear limit (from -75 dBm to -50 dBm) where the noise floor ends and the actual 10 MHz signal begins.

[markone]

I'm also interested to this topic being the owner of a Siglent SSA3021X-TG and I have a simple and clear question :

what is the input power limit that guarantees that the instrument will not be damaged with ANY possible instrument setting ?

In my opinion this number should be available on the instrument datasheet.

I'm in the process to select a set of  attenuators in order to analyze the power output of some HAM radios, of course trying to keep the balance between instrument safety and S/N ratio, but I'm now understanding that is not an easy task 

[RoV]

SSA3021X ? 

[hpw]

I liked to measured the freq-. response of my wide-band 20..40dB amplifier.

So the tracking gen., as a maximal of a maximal of 20dB attenuation supported, I did run into into an input level overdraft.
This is IMHO a very hard limitation/restriction.

So it required, to attenuate the tracking generator, to more than the allowed maximal of 20dB.

hp

[hfleming]

I recently bought an expensive toy - a Siglent SSA2021X spectrum analyzer. I don't want to harm this not only because of the high purchase price. Hence my question, for simple tests and measurements on a Breadboard and with a signal generator in the 5 to 12V DC range, is a 20 dB attenuator sufficient as protection?

SSA2021X RF input
50V DC max - I have as additional protection a DC blocker
+30dBm max - which is one Watt (if I remember right)

With -20 dB from my external Siglent attenuator (20dB@4GHz) the SSA2021X should not be damaged till +50dBm ? Or is this false?

With my Oszilloskope the matter is much more relaxed, with 1 MΩ Input (instead 50Ω on the SSA), and the 1:10 probe you are relatively safe here with voltages not exceeding 24 V. I have no practical experience with specific RF performance. I don't want to analyze transmitters either, but rather interference signals in circuits, RF noise in my home office (switching power supplies, lamps).

Siglent has an app note showing something similar we used in the lab to protect our analyzers. It is just a pic fuse and an HV-blocking capacitor. In ours, we also added a PIN-diode limiter. A blown pico-fuse was a lot cheaper to replace than blowing up our very expensive R&S or HP spectrum analyzers. One common thing that inexperienced users did was to ignore large signals that is off-screen (like an oscillation, or the fundamental) whilst they zoom in on a very small signal.
The app note is here, but I am sure R&S, Keysight and Textronics will also have some similar notes.
https://siglentna.com/application-note/diy-spectrum-analyzer-input-protection

[Wallace Gasiewicz]

I'm also interested to this topic being the owner of a Siglent SSA3021X-TG and I have a simple and clear question :

what is the input power limit that guarantees that the instrument will not be damaged with ANY possible instrument setting ?

In my opinion this number should be available on the instrument datasheet.

I'm in the process to select a set of  attenuators in order to analyze the power output of some HAM radios, of course trying to keep the balance between instrument safety and S/N ratio, but I'm now understanding that is not an easy task 

For Ham Radios and also adjusting the stand alone power meters I use a JFW 100 Watt 3 GHz 40 dB Attenuator.   
I also use an HP Power Sensor and use a SA for Harmonics.    I think these things were bought for the cell industry and have good specifications...At 100 Watt testing it does get warm after repeated testing over several minutes. 
When I test power meters, you might be surprised at the results.....

JFW 50FH 040 100 3N

I also have a much bigger oil filled Load that has a "take off" or Sampler connection that is not linear but can be easily calibrated out...( no bumps but a sagging signal at low end)   I THINK it works out to decreasing the signal by 60 dB....I would have to check.   It is a Military BIG Thing.  Great for Linear Amps.

[tautech]

FWIW
In all SSA/SVA models, Attenuator settings in the Amplitude menu are best left set to Auto and any additional external attenuation compensated for in the Ref Offset settings so to negate any need for level maths in the head on the fly and to have Marker level measurements autocorrect for external attenuation values.

A couple of no signal screenshots showing these settings, one with external 10dB attenuator added and Ref Offset changed. Note the marker level remains similar only reading the noise in the no signal sweep.

As levels get closer to 0dB and above, these analyzers will auto-change internal attenuation to protect themselves.
Use a Manual attenuator setting at your own risk.