How to avoid breaking the spectrum analyzer & what accessories to have

[Bicurico]

Hi,

As suggested in the Siglent SSA3kX thread, where I asked about avoiding grilling the RF input of the spectrum analyzer, I opened this new dedicated thread.

Please contribute:
- advise on how to avoid breaking the spectrum analyzer due to feeding the wrong signal
- advise on best practices
- what accessories to have (attenuators, DC blockers, etc.) and where to get them (will cheap Chinese products do, or should one buy the most expensive ones)?
- how to know how much attenuation is required or if the signal cable is unsafe?
- how to measure the total dBm of the input signal, prior connecting it to the spectrum analyzer?
- ...

Regards

[nctnico]

When it comes to accesoires you have to decide on which connector to standarise first. In general BNC is considered unsuitable for anything over 500MHz. This is probably debatable but in/on RF gear you'll only find the big N connectors, SMA (threaded) and SMB (push-on with latch). I choose the standarise on SMA connectors.

Over the time I have collected a whole bunch of adapters (BNC <-> SMA, SMA <-> N),  SMA Tees, SMA 90 degree male/female, SMA attenuators and recently some DC blockers. Furthermore a splitter (to split a signal without impedance mismatch), step attenuator, directional coupler/bridge are usefull accesoires.

You'll also need cables. These can be bought on Ebay for little money. RG316 coax cables are very cheap but only good to approx 1GHz. I use these for general purpose work and I'm not afraid to cut a few up. RG316 is quite thin and flexible so very easy to work with and it doesn't clutter your desk so much as the 'classic' thick RG58. For signals over 1GHz I have some Huber+Suhner RG223 cables (according to the specs useful up to 6GHz) with SMA connectors but even from Ebay these are in the $30 range (each!).

[Bicurico]

Thanks!

To sum up from the other thread:

1) Attenuator: -20dB (I imagine further values would never hurt)
2) DC Blocker (I wonder how "safe" they actually are, as you get them from 5 Euro to 500 Euro)
3) Cables (I was not aware about the frequency ratings - for my TV applications, the coaxial cable is used up to 2150MHz)
4) Connectors (Yes, I already have a ton of them lying around, but only BNC, "F" type and some SMA - I totally lack the "N" type)
5) I imagine that a white noise source would be great (I use them for measuring attenuation in cable networks, to measure amplifiers or filters. etc.)
6) Calibrated signal source (to measure active and passive components in CATV/SAT distribution with a given frequency)

Regards

[G0HZU]

If I'm working on high power RF or AF stuff I always fit a limiter to the front end of the analyser.
In the image below you can see a few scruffy homebrew examples of some of my RF limiters.

Limiter #1 works from AF through to about 25MHz and it is also DC blocked and fused so the RF fuse will blow if the overload becomes too severe for the limiter diodes to cope with.
 
Limiter #2 is also DC blocked and fused and it works up to 450MHz but it does have about 0.5dB insertion loss at 450MHz.

Limiter #3 is unfused and has no DC block and uses BNC connectors and I often use this to protect a sig gen from reverse power when testing transceivers etc.

Limiter #4 is a low loss limiter design (unfused and no DC block) and it provides reliable/effective limiting up to several Watts up to about 500MHz. it has very low VSWR and also very low loss of about 0.05dB at 500MHz. I often use this to protect my RF power meter sensor heads. Up to 500MHz it has negligible loss and very low VSWR meaning I can usually ignore any loss it causes.

For higher frequencies I use a PIN diode based limiter (not shown) but all of the limiters in the image below use a pair of back to back 1N4148 diodes. i.e. a total of 4 diodes in each limiter. These are cheap and easy to make and provide a useful level of protection. The fused limiter #2 is the one I use the most for HF and VHF stuff. The RF fuse will hopefully blow in a fraction of a second and hopefully this will be quick enough to not fry the limiter diodes if the analyser ever got zapped with a very high RF power.

I've been using spectrum analysers all my working life and I've never damaged one yet

Note that some of my colleagues at work can't make the same claim... the most common way to damage a spectrum analyser is via external DC if the analyser doesn't have a DC block built in. quite a few of ou r analysers at work have been repaired more than once for this type of damage

[pascal_sweden]

Isn't there an off the shelf kit that includes all these items in one kit?

The Spectrum Analyzer Emergency Kit

[Gyro]

For higher frequencies I use a PIN diode based limiter (not shown) but all of the limiters in the image below use a pair of back to back 1N4148 diodes. i.e. a total of 4 diodes in each limiter. These are cheap and easy to make and provide a useful level of protection. The fused limiter #2 is the one I use the most for HF and VHF stuff. The RF fuse will hopefully blow in a fraction of a second and hopefully this will be quick enough to not fry the limiter diodes if the analyser ever got zapped with a very high RF power.

Ah, fortuitous timing (for me anyway!). I'm just looking at putting together a prescaler for my counter, so the same requirements. Max input frequency is spec'd to 600MHz, hopefully good for a bit more in pratice. I was thinking that I would need to use 1N5711s but based on the above I'm now wondering whether I could get away with 1N4148s. It's probably entering boarderline territory, but obviously 4148s are more rugged.

[HighVoltage]

One main thing to avoid is cheap adapters.
I bought a few cheap Type N and SMA and SMB connectors and they did not even fit the tread correctly.
You can easily destroy a input connector by trying to get a badly made mating connector installed.

From my observation, you can buy from at least three different quality groups of adapters, connectors, cables and so on.

1) Cheap
Avoid by all means

2) Normal lab quality
It is good for most applications

3) Metrology grade
You only understand this, when you have installed your first real metrology grade adapter or cable
They fit so good, it is just amazing.

I learned this the hard way!

[pascal_sweden]

You can compare it with bearings for a skate board or inline skates.

They also come in different quality classes: ABEC-1, ABEC-3, ABEC-5 and ABEC-7

Note that the higher precision, doesn't mean that they last longer. On the contrary sometimes.

[joeqsmith]

Looks good.  What are you using for your RF fuses?   

If I'm working on high power RF or AF stuff I always fit a limiter to the front end of the analyser.
In the image below you can see a few scruffy homebrew examples of some of my RF limiters.

Limiter #1 works from AF through to about 25MHz and it is also DC blocked and fused so the RF fuse will blow if the overload becomes too severe for the limiter diodes to cope with.
 
Limiter #2 is also DC blocked and fused and it works up to 450MHz but it does have about 0.5dB insertion loss at 450MHz.

Limiter #3 is unfused and has no DC block and uses BNC connectors and I often use this to protect a sig gen from reverse power when testing transceivers etc.

Limiter #4 is a low loss limiter design (unfused and no DC block) and it provides reliable/effective limiting up to several Watts up to about 500MHz. it has very low VSWR and also very low loss of about 0.05dB at 500MHz. I often use this to protect my RF power meter sensor heads. Up to 500MHz it has negligible loss and very low VSWR meaning I can usually ignore any loss it causes.

For higher frequencies I use a PIN diode based limiter (not shown) but all of the limiters in the image below use a pair of back to back 1N4148 diodes. i.e. a total of 4 diodes in each limiter. These are cheap and easy to make and provide a useful level of protection. The fused limiter #2 is the one I use the most for HF and VHF stuff. The RF fuse will hopefully blow in a fraction of a second and hopefully this will be quick enough to not fry the limiter diodes if the analyser ever got zapped with a very high RF power.

I've been using spectrum analysers all my working life and I've never damaged one yet

Note that some of my colleagues at work can't make the same claim... the most common way to damage a spectrum analyser is via external DC if the analyser doesn't have a DC block built in. quite a few of ou r analysers at work have been repaired more than once for this type of damage

[G0HZU]

They are just very small 160mA quickblow fuses from Maplin.

They are a bit inductive and a bit lossy at RF but I absorb the inductance into a LPF to minimise losses up at UHF. So for limiter #2 the circuit appears to small signals like a 500MHz LPF.

In the case of limiter #2 I've tested the 'blow time' of the fuse when hit with 5W of RF and it's a fraction of a second because the RF current into the limiter diodes will be very high.

For small signals below 0dBm and below 50MHz the insertion loss is less than 0.15dB. It's about 0.3dB loss at 300MHz and 0.5dB at 450MHz. But above this it falls off rapidly due to the LPF topology. It also has a DC block that works with minimal loss down to 50kHz.

I feel quite safe when using this on my analyser when playing with LF/HF/VHF transceivers. Obviously, I put a big beefy attenuator ahead of it because the limiting action starts at around 0dBm. But the idea is that if I ever forget to fit the attenuator then the limiter diodes and/or the fuse or its internal DC block will protect the analyser.

[FuzzyOnion]

One main thing to avoid is cheap adapters.
I bought a few cheap Type N and SMA and SMB connectors and they did not even fit the tread correctly.
You can easily destroy a input connector by trying to get a badly made mating connector installed.

From my observation, you can buy from at least three different quality groups of adapters, connectors, cables and so on.

1) Cheap
Avoid by all means

2) Normal lab quality
It is good for most applications

3) Metrology grade
You only understand this, when you have installed your first real metrology grade adapter or cable
They fit so good, it is just amazing.

I learned this the hard way!

This is good advice for most any test equipment.  I've been down that road too.  Be very wary of cheap cables.  Those RG cable designations are generic.  Good quality coax does not come cheap - especially some of the small double shielded types.  If you're going to be working at frequencies above 20Mhz, determine the pedigree of the coax cable used to make any cables you buy.  Nothing can send you down the wrong road like flaky cables and connectors.

Good luck with your SA.

[Bicurico]

Can you please tell me if this line of thought is correct:

I have a noise source with the following specs: gaussian noise with -56dBm measured at 100kHz RBW, 75 Ohm

What is the total input power when connected directly on the IF of the SA? From what I read so far, I need to consider power at ALL frequencies, so:

-56dBm is roughly 0.000000002 W

Over 3.2GHz the 100kHz reference power specified (I understand that it is the avarage power for a 100kHz bandwidth) fits 32000 times.

So I multiply the 0.000000002W by 32000, which equals 0.00064W for the range of 1MHz to 3200MHz and convert back to dBm.

I get roughly -2 dBm over the whole bandwidth.

So it would be safe to connect this noise source as total power of -2 dBm is much less than the max 30dBm, even considering 75 ohm to 50 ohm conversion (noise source is 75 ohm so I should multiply by 1.5 to get power equivalent on 50 ohm, right?).

Is this roughly a correct assumption to not overload the IF of the SA?

Thanks,
Vitor

PS: Edit was done on a propper laptop to correct typos and formatting.

[Bicurico]

New question:

Would it make sense to build this circuit http://www.zen22142.zen.co.uk/Circuits/Testgear/rfprobe.htm? It uses the diode do convert RF AC to DC, so that a multimeter can measure the current/voltage and hence it allows determination of the RF power.

This seems usefull to test if the source can safely connected to the IF of the SA.

Question 2: Up to which frequency is this specified?

Question 3: I understand this requires a dummy load connected between the multimeter and the source, so that the current can be measured?

Thanks,
Vitor

[Performa01]

Can you please tell me if this line of thought is correct:

I have a noise source with the following specs: gaussian noise with -56dBm measured at 100kHz RBW, 75 Ohm

What is the total input power when connected directly on the IF of the SA? From what I read so far, I need to consider power at ALL frequencies, so:

-56dBm is roughly 0.000000002 W

Over 3.2GHz the 100kHz reference power specified (I understand that it is the avarage power for a 100kHz bandwidth) fits 32000 times.

So I multiply the 0.000000002W by 32000, which equals 0.00064W for the range of 1MHz to 3200MHz and convert back to dBm.

I get roughly -2 dBm over the whole bandwidth.

So it would be safe to connect this noise source as total power of -2 dBm is much less than the max 30dBm, even considering 75 ohm to 50 ohm conversion (noise source is 75 ohm so I should multiply by 1.5 to get power equivalent on 50 ohm, right?).

Is this roughly a correct assumption to not overload the IF of the SA?

Thanks,
Vitor

Let’s clarify your request – what do you mean by “connecting directly to the IF of the SA”?
You cannot access any IF of the SA without opening it up – it wouldn’t make much sense either. You can access the input of the first mixer almost directly though, if the preamp and attenuators are turned off.

It’s generally very unlikely that a noise source provides such a high output power that it could damage an SA. But let’s do the exercise nonetheless…

First of all, you should use the specified bandwidth of the noise source and not of the SA. Just because an SA is specified up to a certain frequency, that does not mean it cannot be affected – or even damaged – by frequencies higher than that. Usually there is some input lowpass filter, but you may not know for sure what it’s actual corner frequency is, how much power it can handle in the stop band and whether there still are some sensitive components before the filter. Unless you know all that, it’s actually best to stick with the bandwidth of the noise source.

Let’s assume the bandwidth of interest is actually 3.2GHz.
We have a noise source with 75 ohms output impedance. If we use that in a 50 ohms system, the output power will drop by only ~4% due to the impedance mismatch. The output voltage will drop to 80% of what it is with correct 75 ohms termination, but because of the lower load resistance, the total power will not change significantly. All in all, we can ignore the impedance mismatch when calculating the total output power, but depending on the internal circuitry of the noise source, a wideband (resistive) matching network might be in order to preserve the flat frequency response required for white noise.

3.2GHz is 32000 times 100kHz, so the total output power will be 45dB higher than the -56dBm/100kHz.
10*log(32000) = 45,05dB;
Without a calculator, a power ratio of 10000 = 40dB and a ratio of 3.2 = 5dB, hence the result is 40+5 = 45dB.

This results in a total output power for 3.2GHz bandwidth of -56 + 45 = -11dBm;

So yes, this would be totally safe.

[Performa01]

New question:

Would it make sense to build this circuit http://www.zen22142.zen.co.uk/Circuits/Testgear/rfprobe.htm? It uses the diode do convert RF AC to DC, so that a multimeter can measure the current/voltage and hence it allows determination of the RF power.

This seems usefull to test if the source can safely connected to the IF of the SA.

Question 2: Up to which frequency is this specified?

Question 3: I understand this requires a dummy load connected between the multimeter and the source, so that the current can be measured?

Thanks,
Vitor

This circuit will give you a rough estimation on the peak voltage; you’d divide the DVM reading by sqrt(2) in order to get the rms voltage.

Accuracy will not be very high, particularly for voltages <1V and higher frequencies.

Regarding bandwidth, it totally depends how this thing is built. There are RF demodulator probes commercially available, which are usually specified up to 500MHz. They are meant to connect on a scope and do not have any internal load resistor.

While it’s possible to construct reasonable dummy loads with ordinary film resistors as long as they are shielded in a way so they become a lossy transmission line of the desired characteristic impedance, the switch would be increasingly problematic for frequencies >100MHz and should thus be omitted. Build a separate detector for each desired input impedance instead.

The OA91 would not be easy to get nowadays and I very much doubt it will actually work up in the UHF range. Use an RF Schottky diode such as the 1N6263 instead. With that it might be possible to get a bandwidth >1GHz – maybe I’ll give it a try this weekend…

[Jester]

Am I correct to assume that this diode based detector scheme only works for voltages greater than Vf of the diode, which is okay if the goal is to simply determine if the signals level is to "hot" for the SA input?

[Bicurico]

@­Performa01

Thank you very much for your time and response.

"Let’s clarify your request – what do you mean by “connecting directly to the IF of the SA”?"

Sorry, my mistake: I meant the input connector of the SA.

"It’s generally very unlikely that a noise source provides such a high output power that it could damage an SA. But let’s do the exercise nonetheless…"

Yes, it was meant as a just an exercise, to see if I understood the concepts correctly.

"3.2GHz is 32000 times 100kHz, so the total output power will be 45dB higher than the -56dBm/100kHz.
10*log(32000) = 45,05dB;
Without a calculator, a power ratio of 10000 = 40dB and a ratio of 3.2 = 5dB, hence the result is 40+5 = 45dB.

This results in a total output power for 3.2GHz bandwidth of -56 + 45 = -11dBm;"

Great! That was exactly what I thought, which means I am on the right track.

@Jester:

"Am I correct to assume that this diode based detector scheme only works for voltages greater than Vf of the diode, which is okay if the goal is to simply determine if the signals level is to "hot" for the SA input?"

I have no idea about the diode, but the goal is indeed to first determine if a source is safe to be connected to the SA.

@all: In practical terms, I don't have any real problem. I have several TV field meters with spectrum analyzer and their input connector is rated roughly the same or even below the one of the SSA3021X SA, I just purchased. My sources are either signal generators, noise sources or live cables transporting TV, CATV or SAT signals. Apart from the obvious DC blocker (though DC will be 24V maximum, normally for SAT only 14V/18V) there is hardly any chance to overload the SA. The TV field meters have been designed to measure exactly these signals, so I assume that the SA will survive them, too.

However, I want to fully understand what I am doing and I want to have all required accessories to measure whatever signal I come across, without damaiging the SA.

So thank you for your help and patience.

Regards,
Vitor

[Performa01]

I promised to have a look at the diode detector, particularly what can be expected with a homebrew implementation. I don’t want to hijack this thread, hence would like to keep this post short, but there are several aspects that might be of interest for anyone having an SA, hence dealing with HF. So here it goes:

The diode detector consists of two parts – a dummy load and the detector itself. If the dummy load doesn’t work reasonably well, the detector cannot work satisfactory either. So I wanted to know what can be achieved here with very little effort.

I decided to use two 100 ohms / 2W metal oxide resistors in parallel, so the detector can handle powers up to 4W = +36dBm. No attempts have been made to preserve the characteristic impedance, so I had no high expectations. But surprisingly, the return loss is near perfect up to 770MHz and still decent enough up to some 2.4GHz (RL_MOX_2x100R_3000MHz_2)



Please note that the directional coupler used for this measurement is only specified up to 2GHz and even though I’ve calibrated for an “open” over the full frequency span, the results might not be totally reliable for frequencies >2GHz.

Here is a picture of the test assembly, exemplary for the peak-peak detector (Detector_1N6263_Vpp)




I have tested several configurations:

1: Germanium diode AA116 with 1nF capacitor, just as in the article posted in reply #12.
2. Small Signal Schottky diode 1N6263 with 1nF capacitor.
3. Small Signal Schottky diode 1N6263 with 27nF + 1nF capacitor.
4. Voltage doubler (for peak-peak detection) with 1N6263 and 68nF capacitors.

As a reference, I’ve also evaluated a demodulator probe from TESTEC, the TT-DE 112, which is specified up to 950MHz / 3dB. This is an oscilloscope probe and has no internal dummy load, hence I’ve just used to measure the e.m.f. of the signal generator and consequently the voltage output is 2 * Vpp.

First the transfer curves measured at 10MHz (DD_TF)



The TT-DE 112 looks way more sensitive, but this is because it actually sees twice the voltage. Nevertheless it’s quite impressive, as the error at the specified “turn-on” voltage of 200mV (which corresponds with 100mV in the diagram) is just -23.63%.

All the homebrew detector circuits generally show an error of some -10% even at higher voltage levels. The following diagram shows the error  for the various implementations and here the germanium diode certainly looks best (DD_TE)




Now for the frequency response at 1Vrms (2Vrms for the TT-DE 112) up to 1GHz  (DD_FE)



As expected, the germanium diode only works up to some 170MHz – it is just not fast enough. Other types might give better results, but then again, the AA116 is intended for detector circuits, albeit for low impedance ones. The high impedance detector AA119 would have been better suited for this application, but I don’t have one here in this lab and it appears to be heavily sought after and not easy to obtain, let alone for a reasonable price.

The 1N6263 is not happy with just 1nF, as it shows a resonance with +8dB at 800MHz. This was the reason why I added a 27nF capacitor right behind the diode and this looked way better, but still climbs up to +7.44dB at 1GHz.

Finally the voltage doubler configuration with two 1N6263 diodes and two 68nF caps, shown in the diagram as “1N6263-3”. The frequency response looks similar to the previous configuration, but it outputs twice the DC voltage.

The TESTEC TT-DE 112 excels here, not exceeding its specified +/-3dB error up to 1GHz.

As a conclusion, the TT-DE 112 together with a good quality pass-through terminator would be the best solution after all. Commercially available pass through terminators are usually rated to 2 watts so it almost doesn’t pay off to build your own using MOX resistors, just to gain a couple of watts. Rather use a power attenuator in front of the pass-through terminator if needed.

Right now I don’t have an accurate RF signal generator for frequencies >1GHz , but expect to get one within the next couple of weeks. Then I should be able to check the behavior of the TT-DE 112 for frequencies up to 2.5GHz.

[nctnico]

I doubt that setup will work at several GHz. A couple of years ago I made a detector using a HSMS-286B detector diode, SMD resistors as a 50 Ohm load on an SMA connector in dead-bug style. IIRC that worked OK up to several GHz.

[Performa01]

I doubt that setup will work at several GHz.

Yes, as the measurements clearly show, the bodged-together detectors work up to 700MHz at best. With a little more effort, it might be possible to reach 1GHz, but I certainly don’t expect anything more than that.

[Performa01]

So I was curious what would be possible with leaded components in my testbox.
I’ve rebuilt the “1N6263 27n+1n” configuration so to reduce lead lengths to an absolute minimum and added a choke between the two capacitors in order to reduce any possible influence of the cable connected to the DMM.

Now it looks like this (Detector_1N6263-2)




This was actually an improvement, even at 10MHz accuracy got slightly better. But most obvious is the vastly improved frequency response (DD_FE2)



Up to 900Mhz, the revised contraption now beats the TESTEC probe in terms of flatness, but gets worse again above some 940MHz. Still usable up to close to 1GHz now.

For anything better, we’d probably need SMT components and a connector that provides controlled impedance right up to the solder joint.

[rastro]

I always keep a DC block (N-type DC-18GHz) on the front of my SA.  A good quality new/used go for around $100.00.  It also saves the SA's input connector from wear/damage.

HP 859X Specific:
If you acquire an older HP 859X SA you should immediately copy down all corrections constants.  There may be over 100 values depending on the frequency span/model.  If the backup battery dies you'll lose these.  If you don't know the constants to re-enter, then the SA will need a lengthy calibration procedure to put into working order.
see page 201/210 in the service manual:
http://www.keysight.com/upload/cmc_upload/All/08590-90316.pdf

-rastro

[joeqsmith]

When I was attempting to increase the BW of my old current probe,  I made up a load with an attenuator built in so I could monitor the signal with my scope or SA.   Fairly flat to 150MHz. 

https://www.eevblog.com/forum/projects/home-made-pa-and-load/msg640180/#msg640180

[Bicurico]

Cool! 

It will take me a while until I feel comfortable in building something similar, but I have one question: where do you get the components? I know I could order them from RS or Farnell, but the shipping cost would be huge and I am not sure if they sell the small quantities ("1 diode please")...

This has always been a hurdle to me, when it comes to doing some hacks - I never have the right components at home!

Regards,
Vitor

[Ivan7enych]

Why not to use diodes specially designed for RF detection?
HSMS-286K Agilent Microwave Detector Diode - very cheap on ebay, works up to 6GHz, with bias and instrumentation amplifier it should give a very good sensitive detector.