I have a photo-resistor captured light (serial data with bit of stop and bit of stop). Instead of reading it with a regular serial port on the host computer, I'm recording it using the pc’s sound card . Then I get the resulting sampled data in Matlab, and I need to reconstruct the original text stream. Has somebody already solved this problem?
Hi Kamel,
Do you have raw data as sampled by the sound card or do you already have cleaned up data?
If you have raw data, then you have to tidy up the data:
You will have a sampled values matrix (with sampled variable low and high values from the A/D converter in the sound card) at a certain sampling rate, so several (N) sequential sample values in your matrix will correspond to the same bit - how many will depend on the ratio between your serial stream's binary rate and the audio sampling rate you specified for recording the stream at your sound card (this is a 'design' chosen constant for your setup, for example 4800bps sampled at 48Khz would make it for N=48000 samples/s over 4800 bits/s = 10 samples/bit) - make sure your sampling rate has been at least a bit more than double the serial bit rate -
For easeness you may opt for filtering your data (select a threshold - if you are sampling at 8 bits your might take this to be 127 or 128 for example - and everything above the threshold will become a logical 1 and everything below the threshold will become a logical 0). At this time, you will have transformed your sample valued matrix in an all 0s and 1s matrix. We have 'squared' all samples in the matrix to convert them into a more bit-like (square wave) and logical sense (LOW(0)/HIGH(1)) convention - this will also ease, later on, the averaging and addition to recover the 8-bit value -.
Then you have to take the ratio N above and decimate your matrix (the way you decimate is up to you: e.g., you can easily decimate by taking an average of each group of N samples or by taking the central sample in each group). The resulting decimated values are your actual bit stream (including start and stop bits).
From that you can again decimate by extracting sensor values by reconstructing the 8-bit values by sweeping through your serial cleaned-up bit stream and summing up the 8 bits between each start and stop bit. For the sake of best practices, you could check for each byte, that its start and stop bits have the values (0 or 1) that they should have and if you would have some parity bit in your stream, usually just before the stop bit, you could also check that the resulting byte + parity bit coincides with the appropriate parity configured for the RS-232 stream.
Hope that helps. Good luck!
Edit:
NB: It might be obvious, but for the sake of completeness ... when I refer to summing up the 8 bits in the byte between the start and stop bits, please remember to multiply each bit value times its intended byte position weight - (2^b) with b = [0..7] - from LSB to MSB, of course...
- Badges
- Science topic
- Similar topics
- Engineering
- Computer Engineering
- Computer