https://github.com/mattvenn/fpga-waterfall-badge
https://github.com/mattvenn/fpga-waterfall-badge
Last synced: about 2 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/mattvenn/fpga-waterfall-badge
- Owner: mattvenn
- Created: 2019-09-20T11:59:42.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2020-02-06T20:57:55.000Z (over 5 years ago)
- Last Synced: 2025-04-01T18:09:56.653Z (about 2 months ago)
- Language: Coq
- Size: 1.2 MB
- Stars: 3
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Resources
LCD datasheet: http://aitendo3.sakura.ne.jp/aitendo_data/product_img/lcd/tft/AT035DL01/LQ035NC111.pdf
19.5MHz
http://www.comm.toronto.edu/~dimitris/ece431/slidingdft.pdf
https://www.dsprelated.com/showarticle/776.php
PMOD mems ADC interface: https://github.com/ZipCPU/wbpmic/blob/master/rtl/smpladc.v
# Notes
## abs of complex fft out
using quick abs for complex number to avoid things getting too big
max error about 10%
but could also look at using the square and then a lookup table that converts the squared number to a log output for the gradient lookup
not a problem if it takes a long time? I have 480 fft cycles between bram updates to perform the conversion. each conversion takes 320 clocks. so can convert each bin in the same time as it takes to do an FFT and update them one by one.
use cordic?
## memory usage
* sdft.twiddle rom - the complex coefficients for the fft. need a pair of 8bit signed values for each bin
* sdft.frequency_bins_[real|imag] - output of the fft
* top.freq_bram - bram that stores absolute value of the fft bins: one for each bin at 8bit depth. Every screen update, the contents are copied to the frame buffer
* top.frame buffer - uses 3 x single port rams for 320 x 240 x 8bit frame buffer
* top.gradient rom - bram that maps 8 bit fft output value (stored in frame buffer) to a 24 bit colour. This is generated by the script in ./images
* top.scaling rom - generated with gen_scale.py, does a log mapping of input widht to output width for 12bit output of sdft to 8bit colour map
# Todo
* fix dsp/multiply issue - done
* buy mic pmod - done
* rgb driver for incoming amplitude, could also use WS2812 on a SAO
* work on colour scheme - done
* why does colour scheme change depending on SEED?
* auto gain control, or log output? stop the overflow/wrapping that is happening - done with log scaling LUT ROM
* fix missing first 2 x pix
* pipeline
* formally verifiy
* add some interactive button pressing stuff to change sample freq or sensitivity
* why is there that constant >0 value in the fft about half way through? changed the sample frequency and the line disappears, so maybe it's actually detected or something from the ADC or picking up some interference
# Want to learn
* feedback in general on code/structure - done
* idea about colour changing depending on seed
* pipeline conversation - done