Hi, everyone. I have one question here. In my project, I need to analyze the spectrum of beat signal output from photo detector, the frequency would be around 20Mhz-30Mhz. I want to confirm that whether I could do this with Maple board coz I am not familiar with the sample frequency of this board. Also, could anyone give me one link to the DPS stuff on this forum. I tried to search but got nothing useful. Thank you
DSP on Maple r5
(4 posts) (2 voices)-
Anonymous
UnregisteredPosted 3 years ago # -
gbulmer
ModeratorRockyLi - The frequency is 20MHz to 30Mhz? What resolution do you need the measurements?
That frequency must be sampled at faster than 60MHz (Nyquist–Shannon http://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem). My knee-jerk reaction is Maple is not the tool.
Do you need analogue data, or will something digital, like zero-crossings of the signal, be enough?
(Full disclosure: I am not a member of LeafLabs staff.)
Posted 3 years ago # -
Anonymous
UnregisteredThanks for your reply. Gbulmer.
I know the Shannon theorem. But I am rookie at hardware thing with maple. Since Maple's clock frequency is 72MHz, I just wonder could I sample the signal at Clock Frequency. If so, this is JUST OK. I am thinking use a DSP chip or something which would be much more better. Do you have any advice about DSP chip to do the FFT? In my project, I need to do the FFT on the signal and display that value on LCDPosted 3 years ago # -
gbulmer
ModeratorRockyLi - I can't give you much more advice without some more details.
Since Maple's clock frequency is 72MHz, I just wonder could I sample the signal at Clock Frequency.
Maple's ADC's sample at 1Msps, under ideal conditions, so almost 2 orders of magnitude too slow.
Do you have to continuously calculate the FFT, or is it more like a batch operation, i.e. sample , then calculate the FFT, display it, then start again?
Have you coded the FFT and measured it? What is the ratio of calculation/sample? Once you know that, you'll be aqble to estimate how much computer power is needed.
I'd want as much computing headroom as I could get. I am not an expert but I'd want more than 10x, or more, computing vs samples; trying to make something work with not quite enough computer power is horrible.
TI makes a lot of DSP's, and some folks I know liked Analog Devices BlackFin. If the problem is straightforward FFT and display, then a 'proper' DSP might be effective.
Personally, I'd get the code for the FFT working, do some measurements, and then look for the right amount of computer power.
I could imagine something like a Raspberry-Pi might be okay, but there are some other great ARM boards based on TI.Posted 3 years ago #
Reply
You must log in to post.