Shop OBEX P1 Docs P2 Docs Learn Events
Frequency splitting — Parallax Forums

Frequency splitting

ArchiverArchiver Posts: 46,084
edited 2000-10-18 03:13 in General Discussion
I'd like to design a circuit which acted something like an EQ with a
"trigger point."

Essentially, I'd like to take an audio source (say a stereo input) and
have a series of triggers in each frequency band, which if the excitation
in that frequency rose above some point, it would send a stamp line high,
otherwise it would be low.

A friend of mine suggested using a ADC for the entire wave and doing FFT ;
this seems a little CPU-intensive for the stamp. Something like a series
of filter or something seems more like what I'm looking for, though I'm
not really sure thow to construct it. Perhaps filtering out a band with a
band-pass filter and then doing ADC on separate channels?

I'd appreciate any hints.


Sean T. Lamont, CTO / Chief NetNerd, Abstract Software, Inc. (ServNet)
Seattle - Bellingham - Vancouver - Portland - Everett - Tacoma - Bremerton
email: lamont@a... WWW: http://www.serv.net
"...There's no moral, it's just a lot of stuff that happens". - H. Simpson

Comments

  • ArchiverArchiver Posts: 46,084
    edited 2000-10-17 05:23
    I was sort of indicating that the filtration should be non-DSP based.
    I've written 560001 code, so I know a *little* about DSP stuff. I was
    wondering about the possibilities of pre-filtering (with an actual filter
    IC or analog filter) before sending the ADC output to the stamp, which
    could just look at a TTL input stream or a high/low pin state. This seems
    feasible to me, but I'm not quite sure how to go about it.

    On Mon, 16 Oct 2000, Dennis P. O'Leary wrote:

    > What you're describing is called a comb filter, in DSP jargon. But it's
    > pretty computationally intensive for a stamp, unless Al Williams has already
    > designed a PAK that does this [noparse]:)[/noparse].
    > Dennis
    >
    >
    Original Message
    > From: Sean T. Lamont .lost. [noparse]/noparse]mailto:[url=http://forums.parallaxinc.com/group/basicstamps/post?postID=dIR5envyfYbRttKgMXQV8UYUWxZgce9huSvYqolz_WJG66Ot0ZThtwIl8Mdb_gbnXW8Ht9OquPfmFrjL5w]lamont@a...[/url
    > Sent: Monday, October 16, 2000 6:06 PM
    > To: basicstamps@egroups.com
    > Subject: [noparse][[/noparse]basicstamps] Frequency splitting
    >
    >
    >
    > I'd like to design a circuit which acted something like an EQ with a
    > "trigger point."
    >
    > Essentially, I'd like to take an audio source (say a stereo input) and
    > have a series of triggers in each frequency band, which if the excitation
    > in that frequency rose above some point, it would send a stamp line high,
    > otherwise it would be low.
    >
    > A friend of mine suggested using a ADC for the entire wave and doing FFT ;
    > this seems a little CPU-intensive for the stamp. Something like a series
    > of filter or something seems more like what I'm looking for, though I'm
    > not really sure thow to construct it. Perhaps filtering out a band with a
    > band-pass filter and then doing ADC on separate channels?
    >
    > I'd appreciate any hints.
    >
    >
    > Sean T. Lamont, CTO / Chief NetNerd, Abstract Software, Inc. (ServNet)
    > Seattle - Bellingham - Vancouver - Portland - Everett - Tacoma - Bremerton
    > email: lamont@a... WWW: http://www.serv.net
    > "...There's no moral, it's just a lot of stuff that happens". - H. Simpson
    >
    >
    >
    >
    >
    >
    >
    >

    Sean T. Lamont, CTO / Chief NetNerd, Abstract Software, Inc. (ServNet)
    Seattle - Bellingham - Vancouver - Portland - Everett - Tacoma - Bremerton
    email: lamont@a... WWW: http://www.serv.net
    "...There's no moral, it's just a lot of stuff that happens". - H. Simpson
  • ArchiverArchiver Posts: 46,084
    edited 2000-10-17 05:31
    What you're describing is called a comb filter, in DSP jargon. But it's
    pretty computationally intensive for a stamp, unless Al Williams has already
    designed a PAK that does this [noparse]:)[/noparse].
    Dennis

    Original Message
    From: Sean T. Lamont .lost. [noparse]/noparse]mailto:[url=http://forums.parallaxinc.com/group/basicstamps/post?postID=_cMrctlh4mKX8tUOIxJQRO5cxEqgc_HLdMjfG9NgGo5Sw-2Vxd_V6o63a7P5SAQk9dBUT_a9ftoS1ZURq-Mc7O8l]lamont@a...[/url
    Sent: Monday, October 16, 2000 6:06 PM
    To: basicstamps@egroups.com
    Subject: [noparse][[/noparse]basicstamps] Frequency splitting



    I'd like to design a circuit which acted something like an EQ with a
    "trigger point."

    Essentially, I'd like to take an audio source (say a stereo input) and
    have a series of triggers in each frequency band, which if the excitation
    in that frequency rose above some point, it would send a stamp line high,
    otherwise it would be low.

    A friend of mine suggested using a ADC for the entire wave and doing FFT ;
    this seems a little CPU-intensive for the stamp. Something like a series
    of filter or something seems more like what I'm looking for, though I'm
    not really sure thow to construct it. Perhaps filtering out a band with a
    band-pass filter and then doing ADC on separate channels?

    I'd appreciate any hints.


    Sean T. Lamont, CTO / Chief NetNerd, Abstract Software, Inc. (ServNet)
    Seattle - Bellingham - Vancouver - Portland - Everett - Tacoma - Bremerton
    email: lamont@a... WWW: http://www.serv.net
    "...There's no moral, it's just a lot of stuff that happens". - H. Simpson
  • ArchiverArchiver Posts: 46,084
    edited 2000-10-17 06:17
    If the bands are fairly wide, say a few hundred Hz or more, you could analog
    pre-filter with a bank of Burr-Brown UAF42 chips, configured as pass-band
    filters. They are configured by choosing values for 4 resistors to set the
    high and low cut-off frequencies for the pass bands. Configuration equations
    are a bit hairy, but the Burr-Brown ap notes give good example descriptions.
    These are active filters, which don't require crytal clocks. They are only
    2-pole filters, which are probably robust enough for your application. They
    cost about $16/chip at DigiKey. If you need steeper filter roll-offs, try
    the Maxim 8th order filter chips, which do require crystal clocks.
    --Dennis

    Original Message
    From: Sean T. Lamont .lost. [noparse]/noparse]mailto:[url=http://forums.parallaxinc.com/group/basicstamps/post?postID=pJbxHRxR29lYJJiy9C5EEx0a6INqZisT9g-YaRydDfI6UXOhwlBQsy1nGEEDU92D1_Ahl2A3p0JtcGv15TU]lamont@a...[/url
    Sent: Monday, October 16, 2000 9:24 PM
    To: basicstamps@egroups.com.
    Subject: RE: [noparse][[/noparse]basicstamps] Frequency splitting



    I was sort of indicating that the filtration should be non-DSP based.
    I've written 560001 code, so I know a *little* about DSP stuff. I was
    wondering about the possibilities of pre-filtering (with an actual filter
    IC or analog filter) before sending the ADC output to the stamp, which
    could just look at a TTL input stream or a high/low pin state. This seems
    feasible to me, but I'm not quite sure how to go about it.

    On Mon, 16 Oct 2000, Dennis P. O'Leary wrote:

    > What you're describing is called a comb filter, in DSP jargon. But it's
    > pretty computationally intensive for a stamp, unless Al Williams has
    already
    > designed a PAK that does this [noparse]:)[/noparse].
    > Dennis
    >
    >
    Original Message
    > From: Sean T. Lamont .lost. [noparse]/noparse]mailto:[url=http://forums.parallaxinc.com/group/basicstamps/post?postID=pJbxHRxR29lYJJiy9C5EEx0a6INqZisT9g-YaRydDfI6UXOhwlBQsy1nGEEDU92D1_Ahl2A3p0JtcGv15TU]lamont@a...[/url
    > Sent: Monday, October 16, 2000 6:06 PM
    > To: basicstamps@egroups.com
    > Subject: [noparse][[/noparse]basicstamps] Frequency splitting
    >
    >
    >
    > I'd like to design a circuit which acted something like an EQ with a
    > "trigger point."
    >
    > Essentially, I'd like to take an audio source (say a stereo input) and
    > have a series of triggers in each frequency band, which if the excitation
    > in that frequency rose above some point, it would send a stamp line high,
    > otherwise it would be low.
    >
    > A friend of mine suggested using a ADC for the entire wave and doing FFT ;
    > this seems a little CPU-intensive for the stamp. Something like a series
    > of filter or something seems more like what I'm looking for, though I'm
    > not really sure thow to construct it. Perhaps filtering out a band with a
    > band-pass filter and then doing ADC on separate channels?
    >
    > I'd appreciate any hints.
  • ArchiverArchiver Posts: 46,084
    edited 2000-10-17 17:14
    You might take a look at the NJU7508, "band pass filter for audio spectrum
    analyzer display".

    http://www.njr.com/products/ic/audio.htm

    -- Tracy
  • ArchiverArchiver Posts: 46,084
    edited 2000-10-18 03:13
    How 'bout a 567 tone decoder? They have an adjustable bandwidth (I think
    2%-14% of center freq.)
    If it detects a signal in its detection band, the chip pulls pin 8 low. It
    requires 3 capacitors and 1 resistor.


    Barry

    BTW, I'm working on a stand alone FFT chip using the SX28 CPU. I'm
    considering providing a TTL output for X user definable frequencies (I don't
    know how many can be supported in realtime yet). I'm making it for my own
    project, but if it would be useful for anyone else, let me know.


    > Message: 12
    > Date: Mon, 16 Oct 2000 18:06:02 -0700 (PDT)
    > From: "Sean T. Lamont .lost." <lamont@a...>
    > Subject: Frequency splitting
    >
    >
    > I'd like to design a circuit which acted something like an EQ with a
    > "trigger point."
    >
    > Essentially, I'd like to take an audio source (say a stereo input) and
    > have a series of triggers in each frequency band, which if the excitation
    > in that frequency rose above some point, it would send a stamp line high,
    > otherwise it would be low.
    >
    > A friend of mine suggested using a ADC for the entire wave and doing FFT ;
    > this seems a little CPU-intensive for the stamp. Something like a series
    > of filter or something seems more like what I'm looking for, though I'm
    > not really sure thow to construct it. Perhaps filtering out a band with a
    > band-pass filter and then doing ADC on separate channels?
    >
    > I'd appreciate any hints.
    >
    >
    > Sean T. Lamont, CTO / Chief NetNerd, Abstract Software, Inc. (ServNet)
    > Seattle - Bellingham - Vancouver - Portland - Everett - Tacoma - Bremerton
    > email: lamont@a... WWW: http://www.serv.net
    > "...There's no moral, it's just a lot of stuff that happens". - H. Simpson
Sign In or Register to comment.