The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Mixed-Signal Design >> sinusoida ADC output for FFT
https://designers-guide.org/forum/YaBB.pl?num=1325873498

Message started by singh on Jan 6^th, 2012, 10:11am

Title: sinusoida ADC output for FFT
Post by singh on Jan 6^th, 2012, 10:11am

i need to calculate FFT of the sinusoidal output of incremental ADC with 16 bit resolution

In order to achieve sinusoidal output vs input samples, i gave ramp input voltage amplitude from 0 to 3V for 20000 samples in sin function with fixed frequency freq =1 and fs = 60K with ADC cycles 257 and veried the phase from 0 to 2pi.

algorithm:

for vin = 0 to 3 with step input v =3/20000
for phase = 0 to 2pi
output = ADCout(vin,freq,phase,fs);
end
rmsout = sqrt(mean(output))
end

plot(rmsout)

ADCout is function which has sin wave input function = vin*sin(2pi*f/fs*i/n) where n are number of cycles =257 and i varies from 1 to n

but i dont get the sinusoidal output function wrt samples, so that i can take fft.

Am i doing it the right way?

Title: Re: sinusoida ADC output for FFT
Post by ywguo on Jan 9^th, 2012, 12:45am

I don't understand what you are doing. Just apply a sine wave input and measure the output.

Yawei

Title: Re: sinusoida ADC output for FFT
Post by singh on Jan 9^th, 2012, 1:58am

in order to plot 16000 samples Vs ADC output,do i need to apply sin wave with 16000 different frequencies?

eg

ADC out = adc(amplitude,freq,phase,OSR);

inside adc function
[
for i = 1 to OSR
input = amplitude* sin(2pi*f*i/OSR + phase)

do all ADC stuff
end

filter section
output
]

now when i feed my OSR of 128 ,amplitude =2.8V ,freq of 1KHz phase =0, the sin wave genertaed at input is shown in attached figure.but at the output of digital filter, i just get one value from the ADC output for the whole sin wave input and not the sin wave output :-/

Title: Re: sinusoida ADC output for FFT
Post by loose-electron on Jan 9^th, 2012, 11:52am

single sinusoid at one frequency.
Do an FFT of the output.

I believe thats what you are looking for?

Title: Re: sinusoida ADC output for FFT
Post by ywguo on Jan 9^th, 2012, 5:40pm

Singh,

Now that OSR = 128, you get one number at the ADC output when you apply an input that is 128×Fs.

If the clock frequency is Fclk, and Fclk = OSR × Fs. Just apply a sine wave at the input. The input signal frequency is
Fsig = Fclk × n / 2^m.

n is a odd number so that n and 2^m are coprime numbers. Then do 2^m-point FFT with the ADC output data.

Choose a big enough number for 2^m so that the frequency bin is small enough to identify signal, distortion, and noise.

You need to read a book named Understanding Delta-Sigma Data Converters.
http://www.designers-guide.org/Books/#Schreier-2004

Yawei

Title: Re: sinusoida ADC output for FFT
Post by singh on Jan 10^th, 2012, 2:06am

ywguo wrote on Jan 9^th, 2012, 5:40pm:

thanks yawei for detailed explanation :)

thank you loose electron

Title: Re: sinusoida ADC output for FFT
Post by singh on Jan 10^th, 2012, 7:34am

ok..i was able to plot output sinusoidal plot for ADC (fig 1 attached)

input f = 15
sampling f = 2^14
no of samples = 2^14
i tried to plot FFT for that but i get too many harmonics (fig 2 attached)..

is it fine becoz wen i calculate SNR of output it comes around 4.416 dB.sometime i think i am calculating it wrong becoz by seeing the picture it doesnt seem bad? please let me know the script to calculate power and SNR?
and also how to suppress the first bin before the fundamental as i had to put dc offset in my input sin function due to which thr is one peak before fundamental peak?

Title: Re: sinusoida ADC output for FFT
Post by loose-electron on Jan 10^th, 2012, 3:38pm

The noise floor is 120dB down from the carrier.

Why do you say that the harmonics are too large?

Title: Re: sinusoida ADC output for FFT
Post by Dan Clement on Jan 10^th, 2012, 6:47pm

Maybe if you zoom in you see some spectral leakage that looks like distortion? Maybe play around with different windows?

It's hard to tell since all the frequencies are there. Maybe include one with a zoomed in region near the fundamental.

Matlab has some built in functions for calculating SNR doesn't it? If not most circuit textbooks explain the theory.

Title: Re: sinusoida ADC output for FFT
Post by singh on Jan 11^th, 2012, 1:49am

loose-electron wrote on Jan 10^th, 2012, 3:38pm:

The noise floor is 120dB down from the carrier.

Why do you say that the harmonics are too large?

i was calculating SNR wrong (4.046 dB) which is weird seeing this fft plot....thats why i was wondering..now i have fixed that and got SNR around 98 dB for 16 bit with 255 OSR

now i have one doubt regarding measuring total harmonic distortion as from the definition given in (page 6) :
http://www.analog.com/static/imported-files/tutorials/MT-003.pdf

suppose my fundamental peak bin is 7 ,so do i need to measure sum of five highest harmonic (distortion D) power values after fundamental or i need to measure harmonics at bin 14,21,28.... and then get it divided by signal power?

Title: Re: sinusoida ADC output for FFT
Post by raja.cedt on Jan 12^th, 2012, 1:34pm

seems all your harmonics are in the same level (more or less)??

Title: Re: sinusoida ADC output for FFT
Post by ywguo on Jan 15^th, 2012, 10:22pm

Hi Sing,

You need to sum all harmonics at bin 14, 21, 28 ...

There are not any distinguishable harmonics in the FFT plot, I mean that all harmonics are below the noise floor. But it is correct to sum all harmonics which are integer multiples of signal frequency. Of course only the sum of the first few harmonis, saying 10 or 20 harmonics.

Yawei

Title: Re: sinusoida ADC output for FFT
Post by singh on Jan 17^th, 2012, 8:03am

Hello,

i was just wondering how i can see the power spectral density of the output of ADC...i mean signal and noiseshaping spectrum from the output of the ADC ( signal at low frequency and noise curve thrown at high frequency)

if i apply FFT tothe output code,i get the picture posted above.

how shud i do that?

Title: Re: sinusoida ADC output for FFT
Post by ywguo on Jan 18^th, 2012, 10:36pm

I thought there was a decimation filter in your ADC. Could you please show a block diagram of your ADC? -- Yawei

Title: Re: sinusoida ADC output for FFT
Post by singh on Jan 19^th, 2012, 6:24am

ywguo wrote on Jan 18^th, 2012, 10:36pm:

I thought there was a decimation filter in your ADC. Could you please show a block diagram of your ADC? -- Yawei

i am using ADC architecture attached in the link. yes there is filter (accumulator + counter). it is initialized before every conversion.
suppose y is filter output and xx(i) contains comparator outputs for n cycles
y = 0;
for i = 1 to n
y = y + xx(i);
end
output_ADC = y;

now how i check my output spectrum if i remove this filter function? if i remove my filter functionality, I will just have compartor output values and output voltages of the integrators? how i see the spectrum of both signal and noise together?