Post by thread:FFT with 16C84? Rather not

Hi Hendrik & David, > > We are thinking of building a simple audio-spectrum-analyzer. > > The output would consist of a LED-matrix (7 bars, 8 LEDs / bar). > > The short answer is no. The '84 does not have anywhere near enough > resources to perform real-time FTs. Even 'touch-tone' decoding > is a challenge. > > You are better off using 7 analogue band pass filters and sampling > the output peaks. This could be done with a MUX input ADC such as > the PIC16C74, or an external unit such as NSC's ADC8017 (16 input). > > If this seems like too much hardware, then use a DSP such as TI's > DSP320Cxxx series. I think some have inbuilt ADCs. The DSPs are a good engineering solution if you want to make it flexible. Analog Devices makes a Development systems for about US$ 100 I think. What I left Out from my previous posting on the EXAR parts was the part numbers. They are XR-1091 .. XR-1097. 1092 Has 12 bands and the MUX 20 pin IC 1097 has 7 bands and the MUX and ADC with serial control 14 pin IC The others have analog or matrix display outputs. Cheers -- Kalle Pihlajasaari spam_OUTkalleTakeThisOuTip.co.za http://www.ip.co.za/ip Interface Products P O Box 15775, DOORNFONTEIN, 2028, South Africa + 27 (11) 402-7750 Fax: 402-7751 http://www.ip.co.za/people/kalle DonTronics, Silicon Studio and Wirz Electronics uP Product Dealer

Kalle Pihlajasaari wrote: > > Hi Hendrik & David, > > > > We are thinking of building a simple audio-spectrum-analyzer. > > > The output would consist of a LED-matrix (7 bars, 8 LEDs / bar). > > > > The short answer is no. The '84 does not have anywhere near enough > > resources to perform real-time FTs. Even 'touch-tone' decoding > > is a challenge. > > > > You are better off using 7 analogue band pass filters and sampling > > the output peaks. This could be done with a MUX input ADC such as > > the PIC16C74, or an external unit such as NSC's ADC8017 (16 input). > > I think another method than FFT should apply at practice (in case of _simple_audio_spec_analyzer). I have use corellation method to solve problem like that Result of corellation may be used like FFT coeficients. Each corellation coeff computed as: A(k)=(A1(k))^2 + (A2(k)^2) where A1(k) = Sum [i=0..n-1] SIN(Xi)*Y(INPUTi) ; A2(k) = Sum [i=0..n-1] COS(Xi)*Y(INPUTi) ; [i=0..n-1] -> Xn = m*2*pi (full sine or cosine period - to minimize corellation sum deviation); To better result I calculate corellation with SIN(X), SIN(X+pi/4), COS(X), COS(X+pi/4) A(k)= (A1(k))^2 + (A2(k))^2 + (A3(k))^2 + (A4(k))^2 where A3(k) = Sum [i=0..n-1] SIN(Xi+pi/4)*Y(INPUTi) ; A4(k) = Sum [i=0..n-1] COS(Xi+pi/4)*Y(INPUTi) ; (in second case deviation A(k) resulting from initial phase of input signal much less) WBR Dmitry.