Store & Delay Samples — Giving DSP a Memory
Addition and multiplication can operate on the sample we have now. But many processors also need something that happened earlier. To do that, DSP must store a value and use it later.
1. Meet the little DSP diagram
current sample
previous sample
z−1 means: delay the signal by one sample. Nothing mystical happens inside the box. A value arrives, it is remembered, and one sample later it appears at the output.
A signal/value moves along this path.
Remember for one sample.
Delay by D samples.
2. Watch samples move through memory
Press Next sample. The highlighted input moves forward while the output shows the value from D samples earlier.
Changing D simply asks for a value further back in the stored history.
3. From samples to time
A sample delay is tiny. How tiny depends on the sample rate.
≈ 0.0208 ms
1 ms
10 ms
100 ms
500 ms
4. Hear memory become delay
A continuous tone is not ideal for demonstrating echo, so this section uses a short percussive pulse once per second. At very short times the delayed copy colours the sound; increase the delay and the second hit separates into a clear echo.
5. One box can become a delay line
Three one-sample memories in sequence give a three-sample delay. In practice, longer delays are implemented efficiently with buffers rather than drawing thousands of little boxes.
6. You can now read a basic DSP diagram
You already know every symbol here: delay/store an earlier signal, multiply it by 0.5, then add it somewhere. We will use exactly this small vocabulary to build familiar processors.
7. What memory unlocks
Use an older copy of the signal.
Return some output into stored history.
Continuously move a short delay time.
Combine current and previous samples.
Circulate energy through delay and loss.
Build networks containing many memories and feedback paths.