3 ms·

I remember when VoIP ATAs first came out that one of the cardinal rules was not to use them with dial-up modems, since the codecs used didn't do well with the e

by cbhl 3y ago

I remember when VoIP ATAs first came out that one of the cardinal rules was not to use them with dial-up modems, since the codecs used didn't do well with the encoding schemes used a 33.6kbps and 56kbps. What has changed since then that makes this work now? (Are networks fast enough that ATA just send PCM now?)

jaywalk 3y ago
Nothing has changed. The issue wasn't with ATAs themselves, but VoIP providers. They handle things like compression that are out of your control and can mess with the modem signaling. Without a VoIP provider, you can configure everything so that the modems will be able to connect at 33.6kbps. As the article mentions, 56kbps requires special equipment on the ISP's end which is why 33.6kbps is the limit.
- tjohns 3y ago
  The big problem isn't compression - that can at least be disabled (in theory). Modems expect a circuit-switched network which guarantees zero packet loss and zero jitter. Even if you connect using G.711 μ-law (uncompressed), you're still using a packet-switched network subject to packet loss and jitter. There's just no way around that. In that regard nothing has changed. This setup is inherently unreliable. You can do it for fun, but trying to get it to work outside of a lab environment is never going to be production quality.
gorkish 3y ago
Nothing actually has changed; the regular phone network has been digital TDM for ages. If you disabled the echo cancelling and use an uncompressed codec like ulaw it's always worked to run a modem over an RTP stream, provided you were able to guarantee zero packet loss and zero frame delay (no jitter). But if you are asking if this will really be reliable over the public Internet, the answer is still pretty much no, not well. Any network hiccups or delays will often just result in the connection dropping completely. The advice to steer clear of such solutions is still as valid as it ever was, but for a hobby retrocomputing project, yeah knock yourself out.

by cbhl 3y ago

jaywalk 3y ago

Nothing has changed. The issue wasn't with ATAs themselves, but VoIP providers. They handle things like compression that are out of your control and can mess with the modem signaling. Without a VoIP provider, you can configure everything so that the modems will be able to connect at 33.6kbps. As the article mentions, 56kbps requires special equipment on the ISP's end which is why 33.6kbps is the limit.

tjohns 3y ago
The big problem isn't compression - that can at least be disabled (in theory). Modems expect a circuit-switched network which guarantees zero packet loss and zero jitter. Even if you connect using G.711 μ-law (uncompressed), you're still using a packet-switched network subject to packet loss and jitter. There's just no way around that. In that regard nothing has changed. This setup is inherently unreliable. You can do it for fun, but trying to get it to work outside of a lab environment is never going to be production quality.

gorkish 3y ago

Nothing actually has changed; the regular phone network has been digital TDM for ages. If you disabled the echo cancelling and use an uncompressed codec like ulaw it's always worked to run a modem over an RTP stream, provided you were able to guarantee zero packet loss and zero frame delay (no jitter). But if you are asking if this will really be reliable over the public Internet, the answer is still pretty much no, not well. Any network hiccups or delays will often just result in the connection dropping completely. The advice to steer clear of such solutions is still as valid as it ever was, but for a hobby retrocomputing project, yeah knock yourself out.