So, let's do a little math.

Your RS232 link, you mention in the comments, has 38400 baud, which in the context of RS232 is typically (one RS232 symbol is only worth one bit) the same as saying 38.4kbit/s, assuming this is really the sustainable, and not a best-case, overhead-ignoring, perfect signal conditions number. If I were you, I'd treat the following calculation with a factor of 0.5.

So, let's assume we have 8bit audio samples.

That's pretty baseline. That leaves us with

\$ \frac{38.4\frac{\text{kbit}}{\text{s}}}{8\frac{\text{bit}}{\text{sample}}} = 4.8 \frac{\text{ksamples}}{\text{s}}\text{.}\$

The Nyquist-Shannon theorem states that with such a \$4.8 \frac{\text{ksamples}}{\text{s}}\$ sampling rate, the highest frequency is \$\frac{4.8 \frac{\text{ksamples}}{\text{s}}}2=2.4\text{kHz}\$. That is usually too little for clear communication.

Which means you will have to compress your audio. Now, typical audio compression schemes like those used in GSM can, in fact, make communication over such low-bitrate channels possible, but please be aware that you will have to implement compression/decompression yourself, or use existing ICs to do it for you. In effect, it will be much much easier not to use your existing link but to use a radio that is actually meant to do speech communication out of the box.

So, let's do a little math.

Your RS232 link, you mention in the comments, has 38400 baud, which in the context of RS232 is typically (one RS232 symbol is only worth one bit) the same as saying 38.4kbit/s(UPDATE: you know suddenly figure out the data rate over the air is 19.2kbit/s MAX, so I'll have to halve the following), assuming this is really the sustainable, and not a best-case, overhead-ignoring, perfect signal conditions number. If I were you, I'd treat the following calculation with a factor of 0.5.

So, let's assume we have 8bit audio samples.

That's pretty baseline. That leaves us with

\$ \frac{38.4\frac{\text{kbit}}{\text{s}}}{8\frac{\text{bit}}{\text{sample}}} = 4.8 \frac{\text{ksamples}}{\text{s}}\text{.}\$

UPDATE: with 19.2kbit/s over the air, the maximum sample rate is half of that, i.e. 2.4ksamples/s, maximum

The Nyquist-Shannon theorem states that with such a \$4.8 \frac{\text{ksamples}}{\text{s}}\$ sampling rate, the highest frequency is \$\frac{4.8 \frac{\text{ksamples}}{\text{s}}}2=2.4\text{kHz}\$. That is usually too little for clear communication.

UPDATE: 1.2kHz, see above.

Which means you will have to compress your audio. Now, typical audio compression schemes like those used in GSM can, in fact, make communication over such low-bitrate channels possible, but please be aware that you will have to implement compression/decompression yourself, or use existing ICs to do it for you. In effect, it will be much much easier not to use your existing link but to use a radio that is actually meant to do speech communication out of the box.

So, let's do a little math.

Your RS232 link, you mention in the comments, has 38400 baud, which in the context of RS232 is typically (one RS232 symbol is only worth one bit) the same as saying 38.4kbit/s.

So, let's assume we have 8bit audio samples.

That's pretty baseline. That leaves us with

\$ \frac{38.4\frac{\text{kbit}}{\text{s}}}{8\frac{\text{bit}}{\text{sample}}} = 4.8 \frac{\text{ksamples}}{\text{s}}\text{.}\$

The Nyquist-Shannon theorem states that with such a \$4.8 \frac{\text{ksamples}}{\text{s}}\$ sampling rate, the highest frequency is \$\frac{4.8 \frac{\text{ksamples}}{\text{s}}}2=2.4\text{kHz}\$. That is usually too little for clear communication.

Which means you will have to compress your audio. Now, typical audio compression schemes like those used in GSM can, in fact, make communication over such low-bitrate channels possible, but please be aware that you will have to implement compression/decompression yourself, or use existing ICs to do it for you. In effect, it will be much much easier not to use your existing link but to use a radio that is actually meant to do speech communication out of the box.

So, let's do a little math.

Your RS232 link, you mention in the comments, has 38400 baud, which in the context of RS232 is typically (one RS232 symbol is only worth one bit) the same as saying 38.4kbit/s, assuming this is really the sustainable, and not a best-case, overhead-ignoring, perfect signal conditions number. If I were you, I'd treat the following calculation with a factor of 0.5.

So, let's assume we have 8bit audio samples.

That's pretty baseline. That leaves us with

\$ \frac{38.4\frac{\text{kbit}}{\text{s}}}{8\frac{\text{bit}}{\text{sample}}} = 4.8 \frac{\text{ksamples}}{\text{s}}\text{.}\$

The Nyquist-Shannon theorem states that with such a \$4.8 \frac{\text{ksamples}}{\text{s}}\$ sampling rate, the highest frequency is \$\frac{4.8 \frac{\text{ksamples}}{\text{s}}}2=2.4\text{kHz}\$. That is usually too little for clear communication.

Which means you will have to compress your audio. Now, typical audio compression schemes like those used in GSM can, in fact, make communication over such low-bitrate channels possible, but please be aware that you will have to implement compression/decompression yourself, or use existing ICs to do it for you. In effect, it will be much much easier not to use your existing link but to use a radio that is actually meant to do speech communication out of the box.