Noise source for filter measurements

[neazoi]

Hello, can a noise source be used for filter responce measurements as an alternative to the sweeper method?
I am thinking of using a simple noise source like this **broken link removed** followed by the filter to be measured, followed by the spectrum analyzer with the video filter turned on. Do you think this setup will work?

Also why a symmetrical noise source is better than an asymetrical one? (zener diode based noise source)

 

[jiripolivka]

Hello, can a noise source be used for filter responce measurements as an alternative to the sweeper method?
I am thinking of using a simple noise source like this **broken link removed** followed by the filter to be measured, followed by the spectrum analyzer with the video filter turned on. Do you think this setup will work?

Also why a symmetrical noise source is better than an asymetrical one? (zener diode based noise source)

In principle the filter response can be measured by wideband noise source, and using a spectrum analyzer. I did it, too; the spectrum analyzer had to be operated with a memory and integration mode, so peaks of noise are accumulated to form the filter frequency response.

Similarly you can use a swept-frequency signal source instead of noise. While the dynamic range of spectrum analyzers can reach >60 dB, not many noise sources have enough output power. Some are also not flat enough.

 

[neazoi]

In principle the filter response can be measured by wideband noise source, and using a spectrum analyzer. I did it, too; the spectrum analyzer had to be operated with a memory and integration mode, so peaks of noise are accumulated to form the filter frequency response.

Similarly you can use a swept-frequency signal source instead of noise. While the dynamic range of spectrum analyzers can reach >60 dB, not many noise sources have enough output power. Some are also not flat enough.

I am glad it can be measured using the noise method

A noise source signal of this circuit is 150mVpp. What would you expect as an adequate output noise power is a spectrum analyzer is to be used?
The output may be further amplified by an mmic which is very wideband, but
I do not know about distortion since it amplifies a wide range of signals (noise). But this is noise, so who cares about extra distortion produced by the mmic?

Another point is that when talking about narrow band filters, flatness of the noise is not much of an issue, because it will be quite flat in this small filter region.

Another point also is that this method could be considered even better than the sweep oscillator method because in high sweep speeds the filter hysteresis has to be considered, whereas on the noise method there is not such consideration.

What do you thing about my thoughts?

 

[volker_muehlhaus]

not many noise sources have enough output power.

I agree. Even if you amplify the signal, the "signal" power in the given measurement bandwidth (spectrum analyzer BW) is only a very small fraction of the total noise power.

Let's assume you have a noise source followed by a nice power amplifier, with total output power of 1 Watt, over a 1 GHz bandwidth.
Now let's assume you want to measure a filter and use 10kHz spectrum analyzer bandwidth.

The noise "signal" in this 10kHz band is 1W * 10kHz/1GHz = 10µW
This low "signal" means that dynamic range of your measurement is bad.
And unlike measurement with a tracking generator, you can not improve the signal/noise ration because your "signal" will also decrease when you reduce the bandwidth.

In my opinion, this concept is inefficient and you will get very poor performance. It is much easier to use a spectrum analyzer with tracking oscillator, where the RF power of the source is exactly in the measurement bandwidth.

 

[neazoi]

I agree. Even if you amplify the signal, the "signal" power in the given measurement bandwidth (spectrum analyzer BW) is only a very small fraction of the total noise power.

Let's assume you have a noise source followed by a nice power amplifier, with total output power of 1 Watt, over a 1 GHz bandwidth.
Now let's assume you want to measure a filter and use 10kHz spectrum analyzer bandwidth.

The noise "signal" in this 10kHz band is 1W * 10kHz/1GHz = 10µW
This low "signal" means that dynamic range of your measurement is bad.
And unlike measurement with a tracking generator, you can not improve the signal/noise ration because your "signal" will also decrease when you reduce the bandwidth.

In my opinion, this concept is inefficient and you will get very poor performance. It is much easier to use a spectrum analyzer with tracking oscillator, where the RF power of the source is exactly in the measurement bandwidth.

Ok I know what you mean, If I understood well, in simple words, the power spreads in the full bandwidth in a noise generator whereas it is concentrated in a small point in an ordinary oscillator.

What is a limited bandwidth (filtered) amplifier is used at the output of the noise generator?
I know what you may say, making filters is much harder than sweeping

 

[volker_muehlhaus]

What is a limited bandwidth (filtered) amplifier is used at the output of the noise generator?

I think the problem is the ratio of "used" noise power (measurement bandwidth of the spectrum analyzer) and required total noise power (total noise bandwith that you generate). For selective, narrow band measurements this is a problem.

If you think of reducing the total noise bandwidth, that will work. In the end, by reducing the noise bandwidth further and further, you have a single tone that is swept over frequency.

 

[neazoi]

I think the problem is the ratio of "used" noise power (measurement bandwidth of the spectrum analyzer) and required total noise power (total noise bandwith that you generate). For selective, narrow band measurements this is a problem.

If you think of reducing the total noise bandwidth, that will work. In the end, by reducing the noise bandwidth further and further, you have a single tone that is swept over frequency.

Thank you, I can understand the constraints now.
A last thing, What if a quite powerful amplifier is used to amplify this noise, even though it is in wide spectrum? Will it make any difference?
I am thinking of these 5W mosfets followed by wide band transformers.

 

[volker_muehlhaus]

A last thing, What if a quite powerful amplifier is used to amplify this noise, even though it is in wide spectrum? Will it make any difference?

Sure, you can do this. In the example that I calculated above, you can use a 1W amplifier to generate 10µW "noise signal" in 10kHz measurement bandwidth.

 

[vfone]

For many years the Noise Power Ratio Measurement use a Noise Source, a BPF, and a Notch Filter for evaluating the non-linearity of Power Amplifiers.
The measurement arrives to a shaped noise pedestal (shape of the BPF) and a deep in the middle, which is function of the non-linearity of the PA.

**broken link removed**

 

[neazoi]

For many years the Noise Power Ratio Measurement use a Noise Source, a BPF, and a Notch Filter for evaluating the non-linearity of Power Amplifiers.
The measurement arrives to a shaped noise pedestal (shape of the BPF) and a deep in the middle, which is function of the non-linearity of the PA.

**broken link removed**

I would like to find more such papers that describe practical ways of measurement of various thiings in the lab. Is there any good source?