How to Make a Spectrum Analyzer: Electrical Engineering Guide

[rickmegens]

Im totally new to electrical engineering, and i would like to know how to make one (spectrum analyzer)?

 

[dlgoff]

Spectrum analyzers have complicated designs and good commercial ones are very expensive. Are you sure you want to take on something this hard?

Regards

 

[rickmegens]

Is it really that hard then?

 

[Guybrush Threepwood]

yep, pretty hard

 

[Cliff_J]

How new are you to EE?

The block diagram isn't too bad. You take a signal and run it into several parallels circuits. Each circuit has steep filters to narrow down the specific frequency range for that circuit. The amplitude of that resulting output on that range is then converted into a linear display format like a row of LEDs.

I own an AudioContol 3050A spectrum analyzer, and when opened its pretty much that, 31 circuits side by side to drive the display.

Nowadays, you can get PC software based analyzers that use FFTs to create an output with a whole lot less work...

If you really wanted to build it, you could start by looking at circuits for a VU meter, then how to add filters to it. Here's a couple commerical units one similar and one that's DSP based:

http://www.audiocontrolindustrial.com/
http://www.partsexpress.com/pe/pshowdetl.cfm?&PartNumber=390-805&DID=7

Cliff

 

[Averagesupernova]

Spectrum Analyzers

What frequency range are you talking about? Not all spectrums work the way Cliff explained. While I still agree it is rather difficult to build one yourself, I can explain how some of them work.

 

[dlgoff]

by Averagesupernova
Not all spectrums work the way Cliff explained.

Yes. My idea of a spectrum analyzer is not just for audio. I want to see the outputs on an oscilliscope and be able to look and measure rise-times etc.

Regards

 

[Averagesupernova]

Ok, I think you may be getting your test equipment mixed up. You want to use an oscilloscope to measure risetimes of various circuits in a spectrum analyzer? I'm a little confused as to what you want.

Explained briefly, an oscilloscope displays a voltage in the time domain.

X axis is time and Y axis is voltage.

A spectrum analyzer displays a voltage in the frequency domain.

X axis is frequency and Y axis is voltage.

Lets get this ironed out and I can explain a little more.

 

[rickmegens]

Originally posted by Averagesupernova
What frequency range are you talking about? Not all spectrums work the way Cliff explained. While I still agree it is rather difficult to build one yourself, I can explain how some of them work.

50Hz-10kHz, 10 bands or somewhat, don't needs to be 30 bands or so

 

[Averagesupernova]

Ok, I suspect you are talking about what I would call a 'cheap' spectrum analyzer. A group of LED bars. Each bar only responding to the band of frequencies it represents. The example Cliff gave is just this.

True spectrum analyzers don't work this way. I will give an example of how a spectrum that covers oh say from 1 Mhz to 1 Ghz would work.

A device like this uses the well known superheterodyne receiver scheme. I am going to assume you know the basics of RF components. The first thing that the input signal sees is a variable RF attenuator. How much it is able to attenuate is determined by the range of power that the spectrum analyzer is designed to be able to handle. Yes, power. Spectrum analyzers in the radio frequency range have an input impedance of 50 ohms and their display is calibrated in dBm. 0 dBm is 1 mWatt. The scale is a logrithmic scale. After the signal has been attenuated (or not, depending on how the user sets it) it travels through a low pass filter. The low pass filter will only pass signals from the high end (1 Ghz in this case) down. Then the signal goes into the first mixer. This is an interesting stage. You may know that a mixer has 3 ports. 2 inputs and one output. One input of the mixer is fed with the signal we have described. The other input of the mixer is fed with a swept local oscillator (LO) which is the interesting part. This LO is able to sweep a total range of about 1 Ghz since the input covers almost 0 to 1 Ghz. 'Sweep' means that the LO continously 'sweeps' through its range. The actual frequency of the LO would probably be from about 2 to 3 Ghz in this case. The third port on the mixer which is the output now has signals on it that are constantly varying in frequency because of the constantly sweeping LO. Their range would be from anywhere from 1 Ghz to 3 Ghz because of the mixing action of the mixer. The stage that this output port feeds is known as the intermediate frequency, or IF. There are often times more than one IF in a spectrum analyzer. Not in parallel but arranged in series with mixers and other local oscillators. The first IF in this case would be 2 Ghz. So, now we have frequencies coming out of the mixer from 1 to 3 Ghz, but the ONLY ones passed are the signals at 2 Ghz due to the filtering designed into the IF. This 2 Ghz signal is mixed down several more times with other stages of local oscillators and IF filtering stages. These IFs are designed to have their properties varied. By narrowing the IF so it only is several hundred hertz wide, you resolve what seems to be a band of jumbled signals into many separate signals on the display. After all the IF stages, the signal is finally detected with an AM detector. For everyone who thinks that AM is 'ancient' and of no use in todays world, guess again. After the signal is detected, it is run through a log convertor. This basically amplifies the small signals a lot and the large signals not so much. This is done because a spectrum analyzers display is read out in dB. Decibels are inherently logrithmic. Once the signal has gone through the log convertor, it is amplified linearly in order to drive the verticle deflection plates in the CRT display. The horizontal plates are driven with a ramp voltage that is synchronized with the swept local oscillator. Voila, you have a spectrum analyzer. When NO signal exists on the input, nothing exists on the output of the first mixer that will be able to pass through the IF. If there is a signal coming into the spectrum analyzer at say 500 Mhz, it will pass through the mixer and appear to be sweeping from 1.5 Ghz to 2.5 Ghz because of the mixing action of the mixer. It will pass through the IF only at ONE POINT during the sweep of the local oscillator. Do the math and you will find that point is when the first LO is at 2.5 Ghz which is in the exact middle of its range. Since the horizontal deflection of the CRT is synchronized with the sweeping first LO, can you guess where the trace will be when the detected signal appears at the verticle deflection plates? Right about in the middle, where 500 Mhz should show up. By changing the bandwidth of the IF, and only sweeping a small part of the first LO, you can narrow in on a small group of frequencies say in a range of a couple of Mhz. For instance, if you wanted to look at a group of signals in the FM broadcast band, you would tune the LO to sweep in such a way as to only let those signals through and only view a couple of Mhz of spectrum at a time. Now you may want to build the device you were talking about with a bunch of LEDs and active filtering, but I don't think you will be building the spectrum I described anytime soon. But as an engineering student, it didn't hurt you learn this little tidbit right?

 

[Cliff_J]

Dude, the only reason for RF is to get my tunes!

Good info, wondered what those big $$$ HP spectrum analyzer's were on ebay, looked like an oscope of a different flavor. Hey, I'm still rubbing two sticks together to make my 15MHz scope work, no reason to jump to GHz and guesstimated this guy was doing the same. No offense meant, I'm just trying to jest a bit.

While on this topic and explaining, is the type of spectrum analyzer you described (possibly more specialized) what is used to create those polar plots like some people use to see the distribution of signals within an NTSC signal? I've only seen them implemented in software like Cannopus's editing software or Final Cut Pro but I understand its a pretty standard tool in a TV repair shop.

Cliff

 

[Averagesupernova]

Cliff, a spectrum analyzer is not going to be standard tool in a TV repair shop. Are you thinking of a vectorscope? It shows the phases and amplitude of the color sub-carrier in an NTSC signal. A spectrum analyzer and vectorscope will be VERY standard tools in a TV station though.

Incidentally, a decent spectrum analyzer can show how the 3.58 Mhz color subcarrier content sort of 'lays between' the black and white or 'luminance' information. They are NOT in separate parts of the TV channels spectrum. They actually share the same bandwidth and are separated with what is known as a comb filter in the TV set.

 

[Njorl]

Originally posted by Cliff_J
Dude, the only reason for RF is to get my tunes!


While on this topic and explaining, is the type of spectrum analyzer you described (possibly more specialized) what is used to create those polar plots like some people use to see the distribution of signals within an NTSC signal? I've only seen them implemented in software like Cannopus's editing software or Final Cut Pro but I understand its a pretty standard tool in a TV repair shop.

Cliff

Those polar plots are usually the output of a vector network analyzer. They are essentially a combination of signal generator and spectrum analyzer, and some very good contacting cables and probes. They are used to characterize circuits. They are used to determine impedences, inductances and capacitance at a range of frequencies and signal levels. Many complicated RF devices change behavior depending on frequency. A good set up can easily run > $70,000.

Njorl

 

[Cliff_J]

Cool, I've never fully understood this stuff, although some of Joe Kane's stuff published in WideScreen Review a decade ago was quite revealing. Here's a couple questions maybe someone can shed light on:

I thought the B&W was effectively limited to 4MHz so while the 3.58MHz color sub carrier needed to be comb-filtered that the information above 3MHz was effectively not used much. If my memory hasn't cross-linked the wrong info, I recall it was just the stuff over 100IRE and chorma up there so as to be backward compatible.

Its odd you bring up the comb filter, its a pretty common sales tool and sometimes it seems to make a difference (a better one). But then sometimes in a Y/C versus composite comparison, on some equipment its night and day and other stuff its tough to guess which is which. So why am I messing with this silly DIN plug on my s-vid cable?

Am I remembering correctly that the chroma is G-R and G-B with one of the two signals phase shifted 90 degrees, and that green is effectively luminance since its on a 68/21/11% ratio or something like that and structured to be backward compatible? Then the horiz and vertical syncs added to that make a composite signal?

Hmm, now done thinking aloud, maybe need to find a library that would have WSR archives...or a good NTSC book to read to re-enlighten myself...

Cliff

 

[rickmegens]

sounds hard

do you guys have a circuit board or something, much easyer to me i think

 

[Averagesupernova]

Rick, it sounds hard because it IS hard for someone in your situation. Get a schematic for a color organ and enhance it.

Cliff, you are correct, there is very little luminance information around the color subcarrier, but it is still there. The comb filter works because every other horizontal line of video shifts the phase of the color subcarrier by 180 degrees. Since each line of video is not really that much different than the previous one, they simply delay the signal by one horizontal line, add the delayed signal with the non delayed signal and the color information cancels. All that is left is the luminance. If you want to keep the chroma, you do the same thing, but invert one of the signals before the summing. I wouldn't expect an off the air composite signal to look much different than and off the air Y/C signal. They come down the antenna cable as composite. The only difference is that your VCR tuner separates the luma from the chroma before it goes to the monitor. If you feed the monitor composite, then the monitors comb filter has to do the work. Now if one of these filters is much better than the other, then yes, there would be a difference. A VCR stores the luma and chroma in totally different way. The luma frequency modulates a carrier. The chroma I don't recall for sure, but it is at a frequency that is MUCH easier to separate from the luma as compared to composite video. It is entirely possible to notice the difference between Y/C and composite on a signal coming from a video cassette and especially a DVD.