How Can Accurate dBc Measurements Be Achieved with Limited Resolution Bandwidth?

[Rudibot]

Just some background - I am trying to measure spurious radio signals around a carrier in dBc units. It seems easy until you consider the resoultion bandwidth you are going to set your specan up with. If you start with a large bandwidth the spurious signals can get mixed up with the carrier and you can't see them. As you decrease the resolution bandwidth, spurious that are close to the carrier begin to appear.

So my question is how can I make meaningful dBc meausrements when I can never get down to an infinetly small bandwith? There is always a posibilty of spurious signal being contained in a bandwith smaller than 1Hz. For instance consider two spurious signals that are 1Hz apart with equal power. With my specan set to bandwith 1Hz I will see power coming from both of them as one signal and the magnitude of the dBc will be smaller. If I was able to nail them down to subhertz range I would see each contains half the power and magnitude of the dBc would be larger..So really you can only meausre the worst case dBc- not the real dBc.

Thanks,
Rudi

 

[skeptic2]

There is always a posibilty of spurious signal being contained in a bandwith smaller than 1Hz. For instance consider two spurious signals that are 1Hz apart with equal power. With my specan set to bandwith 1Hz I will see power coming from both of them as one signal and the magnitude of the dBc will be smaller. If I was able to nail them down to subhertz range I would see each contains half the power and magnitude of the dBc would be larger..So really you can only meausre the worst case dBc- not the real dBc.

Thanks,
Rudi

With the bandwidth set to 1 Hz, two signals of equal magnitude 1 Hz apart would appear to have a wider, flatter top than just a single spur. With more experience the you should be able to recognize when you're looking at more than one spur.

 

[Rudibot]

I would have thought it had a very sharp top? There would only be an instant when both signals are inside the filter bandwidth - the rest of the time it would encompass only one signal on each side and be flat.
Regardless, I will change the question to 10 equal power spurious within 1Hz. I would still be unable to discern the number of spurious inside the 1Hz and so would not be able to assign the appropriate power to each (ie 1/10 of the peak power seen on the specan). Again, I would only be able to guarantee the worst case spurious, in this case 10 times the actual spurious!

 

[yungman]

One way is to down convert first so you get it into lower frequency so you can see it better. Just a thought. 1Hz at higher freq is going to be hard to separate.

 

[Rudibot]

I'm not sure downconversion would expand the signal - just shift it downfrequency in which case I woud have the same experiment just at a lower frequency - same bandwidth.

 

[f95toli]

Remember that your signal also has a BW, and spurious signals are -by definition- signals outside that BW meaning this sets a lower limit on the frequency range.

If you want to do this properly you will probably need to know things like the phase noise of your whole system; this will tell you what the background looks like.

 

[dillonjerry]

One way is to down convert first so you get it into lower frequency so you can see it better. Just a thought. 1Hz at higher freq is going to be hard to separate.

There is logic behind this, but it is backwards. What you want to do is to "up convert" the frequency. This will increase the the number of cycles between the spurious signals and effectively give you more "resolution" between the signals.

 

[Rudibot]

Thanks f95toli, that will give me a direction to go with.

dillonjerry, I am sure this is not true.
Bandwidth of signals - and in this example the separation of the spurious signals is conserved in an up or downconversion process.
If this were not the case I would be able to compress the bandwidth of a signal by simple mixing and violate the Shannon-Hartley limit.

 

[Averagesupernova]

I think there is some confusion here. Down converting or upconverting does not change the distance between two carriers. Also, what determines the shape of the birdie on the display of a spectrum analyzer is actually the IF filters in the spec-an. One of the reasons signals are downcoverted in a spec-an is to allow tighter bandwidth. It is a lot easier to make an analog filter that is a couple of hundred hertz wide at several hundred kilohertz rather than several hundred hertz wide at many megahertz. Typically spectrum analyzers up convert the whole spectrum they are able to view to a much higher IF. This allows for easy image rejection by using a low pass filter on the input. Then this is converted to lower intermediate frequencies. Of course getting down to several hertz of resolution bandwidth requires digital filtering and a VERY slow sweep speed. Thinking spectrally is a bit of a mind-bender. A true unmodulated sine-wave has NO bandwidth. How close carriers can be squeezed together in a spectrum is limited to bandwidth of the detecting apparatus. A signal can have bandwidth if modulated but of course this is no longer a true carrier.
-
So to answer the original question, you are limited to the capabilities of the receiving apparatus which in your case is the spec-an. One thing that is often overlooked on spectrum analyzers is the video averaging. It can be a useful function. This lowers the noise floor over several sweeps and allows actual signals which are not random like noise to show up eaiser. Probably doesn't apply to your situation but I thought I would throw it in.

 

[Rudibot]

OK I am feeling more comfortable with this now. While it is fine to talk about infinitley small bandwidths in reality no equipment or receiver can achive this so in effect it does not exist - it is just a mind experiment.

So to take an accurate spurious measurement you need to consider the frequency sensitivity of the receiver, that is its abiility to distiguish between two frequency components. I am still investigating how to calculate this parameter from the receiver specs.

This problem came up when I was told to test if an oscillator module was up to manufactuter spec and found it failed the spurious. I know it is not effecting system performance so I was trying to find away around the spec! Interesingly it now puts into doubt other systems I have tested in the past:

Consider a receiver that cannot distinguish between say 100kHz and 110kHz. If we have two spurious at these positions they will appear on a specan as two separate spurious and pass the specification while to the reciver thet appear as one with their combined power and will degrade performance!


Rudi.

 

[Averagesupernova]

Interesingly it now puts into doubt other systems I have tested in the past:

Consider a receiver that cannot distinguish between say 100kHz and 110kHz. If we have two spurious at these positions they will appear on a specan as two separate spurious and pass the specification while to the reciver thet appear as one with their combined power and will degrade performance!


Rudi.

Ummmm, I think you are confusing something. A receiver that cannot select out one of the signals (100 Khz and 110 Khz) is just a receiver that is lacking in selectivity (narrow filtering). It has nothing to do with a spec-an.
-
Consider this scenario. Suppose we have a receiver tuned to 100 Mhz. It is just a standard FM broadcast band receiver. Since FM has a deviation of + and - 75 Khz it needs an IF that is about 150 Khz wide. We also have a transmitter that is broadcasting at exactly 100 Mhz with a spurious signal at 100.01 Mhz. In the receiver this is downconverted to 10.7 MHz by tuning the local oscillator to 110.7 Mhz. Of course the spurious is also downconverted to 10.69 Mhz. This spurious falls within the bandwidth of the IF in the FM receiver. So no matter how hard you try, you can fiddle with the tuning and you will never be able to tell that there are two signals there by the s-meter. The signal strength meter will come up only once when tuning through the spectrum. THIS is the performance test for the receiver. Testing an IF with a spec-an will tell you the shape of the filter and how many signals are in the passband. That is certainly not to say it is not a useful test. I would say you are misapplying the test in the above scenario.
-
Do you know how a spec-an displays what it does? Do you know how a spec-an actually works? I'm suspecting not. Once you know it makes the spec-an a much more useful tool. My experience anyway. Nothing wrong with that, I'm not trying to make you out as dumb or anything like that. I am often acused of it though. Haha.

 

[Rudibot]

"Ummmm, I think you are confusing something. A receiver that cannot select out one of the signals (100 Khz and 110 Khz) is just a receiver that is lacking in selectivity (narrow filtering). It has nothing to do with a spec-an."

Of course. My point being that what you see on the specan may not be relevant to the receiver.

"This spurious falls within the bandwidth of the IF in the FM receiver. So no matter how hard you try, you can fiddle with the tuning and you will never be able to tell that there are two signals there by the s-meter."

I think this is where I am confused. It is clear that any spurious inside the IF bandwidth will be processed. But what is the point of specifying maximum spurious, wouldn't total spurious, or just total noise be more appropriate? And my original question still stands, how do you make a meaningfull dBc measurements?

"I'm not trying to make you out as dumb or anything like that. I am often acused of it though. Haha."
Yep, I second that accusation :)

 

[sophiecentaur]

I would have thought it had a very sharp top? There would only be an instant when both signals are inside the filter bandwidth - the rest of the time it would encompass only one signal on each side and be flat.

No: a second signal, near the wanted signal will always give you an apparently broader peaked response. As the frequency separation approaches the 3dB bandwidth of the bandwidth defining filter, the broad peak resolves itself into two definite peaks with a dip in the middle.

 

[Averagesupernova]

I'm not sure I understand your last post Rudi. I mean the part about total vs. max spurious. As for the original question, I thought we had put that to rest.