Just another spectrum analyzer thread. Programmable filters

Discussion in 'The Projects Forum' started by MatheusLPS, Dec 3, 2013.

  1. MatheusLPS

    Thread Starter Member

    Jul 16, 2011
    Hello everyone

    Reading some application notes today , I came across something that exists for some time but had never crossed my mind.

    So , we have discussed here on the forum about spectrum analyzers. Several colleagues have already given hints solutions to prevent the construction of analog filters for an analyzer with a 12, 16 or even 32 bands, because the schematic would be huge.

    So far , the cleaner idea so far that I found isto use the MSGEQ. Although very cheap. I need to import this chip because here in my country I can not find it.

    I already read about a FFT but it is a hard to understand how this works and the processing power needed to run it.

    Thinking about this subject (spectrum analyzers) I found this today :




    I thought to myself that in this case I could create a variable bandpass filter using digital potentiometers. Probably it has been done in some corner of the internet. But I found the idea interesting.

    I made a block diagram like this:


    The text on the paper is in Portuguese and says it: Will use 1 micro controller or a LM3915.

    I researched and found an IC almost ready for it. With him, just by changing resistor values, we ​​can create a filter up to the 8th order. Would be the MAX264 and MAX265 's . Look about them here: http://datasheets.maximintegrated.com/en/ds/MAX263-MAX268.pdf

    The problem is that MAX is not found it easily on my country too . So at first, I would use some common operational amplifier and would change the pass band with the help of a micro-controller .

    The output of the filter would be injected into the input of an LM3915 or in same uC AD controlling the digital potentiometers . If only uses the uC, the same CI would be responsible for the switching matrix of LEDs or LED bars. For each passband filter, a column would be active, showing the amplitude of that frequency range.

    With this, we could have a very clean circuit. And the software would be so simple.

    I thought of using only positive sound source. Eliminate the negative part of the audio signal with a resistive divider .

    My concern is the timing. If I send to the digital pot that I want it to change his value, I do not know how much time the filter will require to settle.

    Considering that I am planning to do 32 passbands, probably the settle time will be too slow and I will will not make any quality measurement.

    Any thoughts?

  2. MatheusLPS

    Thread Starter Member

    Jul 16, 2011

  3. LvW

    Active Member

    Jun 13, 2013
    MatheusLPS - as far as I could understand your planning you want to build a spectrum analyzer based on a tunable bandpass filter, correct?
    This is the BASIC principle to visualize the frequency components within a certain frequency band.
    Now you could ask why commercial instruments do NOT use this simple principle, but instead the principle of superposition.
    Here are some comments from my side:
    You need a wide-range tunable bandpass, which fulfills the following requirements:
    1.) Very small bandwidth (for sufficient resolution)
    2.) constant bandwidth over the tuning range
    3.) rather short settling time for the filter.

    In particular, No 2 and No.3 are very hard to fulfill if - at the same time - No. 1 has to me met. Reading the referenced data sheet on filters you will notice that during filter tuning the quality factor (Fo/bandwidth) can be kept constant but not the bandwidth.
    Therefore, a spectrum analyzer unit based on the described principle will have a very limited application range.
  4. t06afre

    AAC Fanatic!

    May 11, 2009
    You have not said anything about the frequency range for this school project of yours. That is a very important design criteria