[GreenKeys] ITTY Poll

[Harold Hallikainen]

On Thu, September 23, 2021 12:12 pm, steve bennett wrote:
>  Harold,
> "Depending on the desired frequency, each successive sample would be
> somenumber of steps through a sine table from the previous sample."
> I think that would work as long as the sin wave table was large enough to
> avoid aliasing issues.Also would need to be done on something like an
> Arduino NOT on a multi-tasking OS like RPi.Interrupts would interfere
> with the timing.

I think that as long as the system can work ahead, putting samples into an
output buffer or file, and that it can compute faster than the output
sample rate, it could be done on a multitasking system. In the system I
designed, which streamed live audio from a DSP using a microcontroller, I
did need to use interrupts. There was a ping pong DMA input buffer driven
by I2S from the DSP. That was then processed to a ping pong output buffer.
When an HTTP request for the stream arrived, the WAV header was sent, then
the output buffer not currently being written to. Then it would flip over
to the other buffer and flip back and forth until the connection was
dropped.

Here, though, If there is incoming text, it could go to a FIFO, then be
pulled from the FIFO and converted to FSK audio samples that are written
to an output FIFO or file. The HTTP response would be pulled from that
file. So, as long as the output file did not become empty, I don't think
synchronous processing or interrupts are required to create the FSK audio,
whether the FSK is generated by direct digital synthesis or by splicing
audio files.


> By "compression" I think you are referring to data compression
> (bandwidth) not audio compression. (volume dynamics)If I understand
> correctly formats like MP3 take less bandwidth by using some sort of
> compression algorithm and other formats like WAV are just strait up PCM
> resulting in larger files.Same with phone codecs.Some are better than
> others. Not all will pass FAX data.I think the codec on my phone is set to
> pass every other word. lol
> -Steve

Yes, I was referring to data compression and not dynamic range
compression. I think 64 kbps is pretty easy to serve, and 8 bit samples at
8 kHz should be more than adequate for voice grade FSK modems. And it's
very easy to convert audio samples to a WAV file that a browser can play.


Harold




-- 
http://w6iwi.org