Re: Speech output from a microcontroller

On Nov 30, 9:42 am, Vladimir Vassilevsky <antispam_bo...@xxxxxxxxxxx>
wrote:
Pertti Kellomäki wrote:
I am looking into generating speech output from
a small datalogger (approximately match box size).

The ATMega PWM output at 32kHz sample rate will produce the speech of
the reasonably good quality. Especially if you will employ the
preemphasis and the noise shaping. The waveforms can be stored in the
external flash as is; there is not much sense in using an audio compression.

Seconding this opinion. I used this approach in a vending machine
application, and also a device that needs to enunciate system status.
It sounds pretty good with an 8kHz sample rate, 32kHz PWM rate (i.e.
reload PWM compare register every 4th sample). 8 bit resolution in my
case, and this is stored as 5-bit delta compressed samples [lossless].

.