Multiple Feedback Filter With Independent Gain Failure

Discussion in 'The Projects Forum' started by S2L, Mar 11, 2016.

  1. S2L

    Thread Starter New Member

    Mar 10, 2016
    Multiple Feedback Filter With Independent Gain not working

    I'm a hobbyist maker and would like to share this circuit with you.
    This is a single supply design. An Electret mic is pre-amped then amplified in two more stages using 2 digital pots controlled by a microprocessor. The working signal, called audio, is then filtered into two bandpass signals, their center frequency not greater than 5Khz and a low Q 4-6, using a Multiple Feedback scheme. The filter outputs are gain adjusted in a separate stage producing the final output, biased at Vdd/2, which are sampled and processed by a micro-controller. The output range of the signals to the microprocessor are specified at 0-5v. These signals do not require tight audio performance such as lowest noise because of the manner in which they are processed by the micro-controller. 2 quad amps are used to build the circuit along with the dual digital pot.
    The manual gain, determined by the setting of the digital resistor, should be able to range the BP1 and BP2 outputs from near 0v to 5v. Optimally this resistor value should be able to clip the peaks at around 4/5 of it's maximum data value. This pot's digital range has 256 steps. The signal pics shown use a data value of 190 for the manual gain control.

    For signal measuring purposes a white noise input is used, frequency set at 1Khz.

    I'm having trouble getting the needed gain from the stages following the bandpass filters to satisfy the 0-5v specification. The amplifiers following the BP filters don't seem to be doing their job.
    Also the signals, as they develope throught the circuit look 'sloppy'.

    The last gain amplifier values are 100k / 10k for a gain of 10.
    For BP1 the input looks around 1v pp, the output around 1v pp.
    For BP2 the input looks around 0.5v pp, the output around 2v pp.

    The signal result of the AGC is specified at around 0.6v pp which the circuit accomplishes with a data value of 230 for it's digital pot having enough headroom to pump up even quieter ambient sound.

    The manual gain stage further pumps up the signal to it's desired operating level, in this case using a data value of 190. Plenty of room here with a max of 255.

    The BP filter seems working OK.
    And then the fail at the BP gain stages which have resistance values for a gain of 10.

    The circuit has been put on a 4-layer manufactured board. The board design may be adding to the failure. Until I can verify that the circuit design is working it's useless to scrutinize the board layout and connections.

    If anybody finds this circuit interesting and would like to help fix it or improve it, that would be amazing.


  2. AnalogKid

    AAC Fanatic!

    Aug 1, 2013
    C09 and R10 form a 1.6 kHz highpass filter. This will have a significant effect on the amplitude of a 1 kHz signal.
    C13 and R15 form a 3.2 kHz highpass filter. this will *really* affect the amplitude of a 1 kHz signal.
    Both of these filters also will affect all frequencies from 0 to around 7 kHz unequally.
    R01 is way too large to bias a standard electret element. Think 2.2K - 4.7K, and check the mic datasheet.
    What are the two corner frequencies of each bandpass filter supposed to be?
    What is the purpose of the overall system?

    Last edited: Mar 11, 2016
  3. S2L

    Thread Starter New Member

    Mar 10, 2016
    Hi ak,
    Good to hear from you.
    For the Bandpass calculations I used a friendly design tool that was found at:

    The values for BP1 were from a specification of:
    Center frequency= 852Hz
    Gain at Fo = -1
    Q= 4.0
    The design tool calculated damping ratio 0.1245

    The values of BP2
    Center frequency= 226Hz
    Gain at Fo = -0.9
    Q= 4.0
    The design tool calculated damping ratio 0.1255

    The tool provided a bunch of other information which was out of my realm of electronic knowledge.

    The circuit will be used in a new type of colororgan.
    As for the purpose of the circuit it is to provide a consistant and controllable signal that reflects the dynamic quality of a sound atmosphere. By consistant I mean that a sound environment (my view) is characteried by a datum pressure range that fluctuates within that range. There may be disturbances that take the temporal sound level outside the datum for short periods of time but this should not affect the gain setting the algorythmn has calculated. This allows the dynamic, movement, wavey motion, texture, whatever you want to call it, of any sound environment, to use the same method or scheme of interpretation or decoding. For instance the sound datum, say at 3am at night is much lower than at 5pm. An airplane flying overhead will bump the datum until the plane leaves the environment which it is influencing and then the datum will return to where it was. Controllable refers to the application user's ability to offset the datum providing a slightly different interpretation of the sound atmosphere.

    I'll be absorbing your recomendations tonight.

    The microphone resistor R01 at 47.5k seems hiiiigh to me also.
    Lowering it's value, will this also lower the db range it can handle. When the circuit is cleaned up I'll be doing a db test to determine what sound levels it can handle.

    Thank you very much for your response and suggestions.

  4. AnalogKid

    AAC Fanatic!

    Aug 1, 2013
    So, it's a long-time-constant AGC keyed to selective energy within two frequency bands?

  5. S2L

    Thread Starter New Member

    Mar 10, 2016
    Yes it's a long time constant could be 7-10 seconds.
    The frequency bands are not involved in calculating AgcSig. The AGC is keyed to the total sound pressure. The behavior of the AGC is determined by the sound level metric used which might be conventional (or one of a secret sauce), the rate which the digital pot is updated, and the strength of the update. Smaller datum ranges and quicker updates produce a flatter signal dynamic.
    For my application the audio signal is adjusted similiar to how our ear adapts to different sound pressure atmospheres.

    This is a PDF I'm looking at now to fix the microphone.

    47k for the bias resistor, wow, I must have slipped the decimal point. Good thing I'm not setting values for circuits controlling a nuclear reactor.

  6. S2L

    Thread Starter New Member

    Mar 10, 2016
    There was no particular rational reason why I chose this mic except that it operates with a 5v supply. It's 1K input impedance, for some reason, attracted my attention.

    Are there tradeoffs between a 1k and 2.2k impedance?
    There seems to be a larger variety of 2.2k electrets.

    The formula I used for calculating R01
    (Vcc – Vmic)/mic current consumption

    Out of curiosity I measured Vmic 71mV, way low.
    So 5 - .071/I = 47000
    the calculated current consumption becomes .0001ma , wholly cow!

    The specs of the mic AOM-6545P-R

    electret specifications.png

    I set up a spreadsheet and came up with this for replacing the bias resistor. It's interesting how the parameters interact with each other in the spreadsheet. I popped in the 5k, the closest one to 4.7k I had and kept the 1uF cap C01 ... for now ... measured Vmic and this was the result.

    Mic spreadsheet.png

    I must say the AgcSig and Audio signals are looking much better and much stronger.

    The heavy bass end is not of great importance to the application.

    Because the pre-amp is CMOS will this effect bias or coupling value choice, if any?

    Would a strategy for getting an even better result be to find the largest R01 resistance and smallest C01 capacitance while keeping reasonably within the electret specification?

    Thanks for your direction
    I'm feeling much better now.