Need someone to design bandpass filters. Most want to drop in values. (see circuit). Need 16 filters, running from a common buss line. Each filter will have a 625Hz band width.
(1) Theoryical 0 to 625Hz
(2) 625 to 1250Hz
(3) 1250 to 1875Hz
(4) 1875 to 2.5kHz
(5) 2.5kHz to 3125Hz
(6) 3125 to 3750Hz
(7) 3750 to 4375Hz
(8) 4375 to 5kHz
(9) 5kHz to 5625Hz
(10) 5625 to 6250Hz
(11) 6250 to 6875Hz
(12) 6875 to 7.5kHz
(13) 7.5kHz to 8125Hz
(14) 8125 to 8750Hz
(15) 8750 to 9375Hz
(16) 9375 to 10kHz
Plus one highpass filter above 10kHz
If Theoryical 0 can't be obtained One lowpass filter to 25Hz, but the frequencys will have to be recalulated. Im thinking of LC instead of RC so the dropoff is steep. Perhaps the L can be a constant value, and easley obtained. Only the Cx's keeps changes.
the modern solution to this is to use a DSP and do an FFT though it's harder to get constant width bands. Which does pose the question of why you would want that...
When you say you want
625 to 1250Hz
for example, does this mean you want the center at 937.5Hz?
Also, when you spec this bandwidth does this mean you want
-3db gain at each band edge?
MrAl is right, the specs you have are not in a form that is easy to use as design criteria. Unless you DSP the whole signal then do the filtering in digital form before playing back through a D/A you'll never get exact bandpass cutoff frequencies. Need to know what gain and overlap you can stand.
I tried the numbers in the filter design wiz in MultiSim but I had to braoden the peaks because it required more than 10th order filter (the limit). Ultimately it spit out a filter that used 10 op amps, not terribly useful. The passive filter it designed was much more compact I didn't simulate any of the examples but I did notice they had some example circuits in the educational section of their website. I think they still have a limited free version for download. Your circuit is ripe for simulation since you'll probably have to do some trial and error to get what you need. With those frequencies, PCB layout shouldn't be too critical but i f you want sharp filtering, good qaulity caps and inductors will be needed (high Q)
Note that graphic equalisers are based on musical notes that increase in frequency exponentaily.
His requirements have the bands all with an equal passband bandwidth, with passband center frequencies progressing linearly, like AM radio allocated frequencies.
I didn't notice the frequency progression but with the chip in the link you can set the center frequency and Q using a couple capacitors and a resistor. See the description near figure 4 of the datasheet.
It also appears that you could use each filter section seperately instead of summing them at the output amplifier.
There might be other solutions than a filter bank, depending on what you are trying to do. The filter bank may be necessary if you need to separate and measure all 16 channels at the same time. A Digital Signal Processor (DSP) can probably do this with a FFT type of algorithim, since the frequencies are in the audio range. If you only need to know which channel a signal is in, a frequency measurement may need a lot less hardware.
Filters tend to have wide skirts, especially at the upper end of the frequency range. This will effect the type and complexity of the filters. Bandwidth is normally measured at 3 db down, the 20 db down bandwidth might cover several adjacent channels.
