There are ways of making low-bit-rate voice comm's compatible with moderate-rate (~48-56kb) dialup connections. Low-bit-rate voice comm's mostly involve various schemes, based in digital signal-processing, that reduce the amount of data that has to be sent over a voice channel in order to ensure that enough information arrives at the receiver to make the transmitter's speech recognizable and intelligible.

The biggest part, of course, lies in not transmitting non-information, i.e. silence. The resolution at which silence is recognized and eliminated from the bandwidth that is sent from one end of the link to the other determines, in part, how much bandwidth will be required. If you can discern and delete, then regenerate silent periods at the other end before they disrupt the intelligibility of digitally encoded speech, as much as 80% of the bandwidth can be spared. However, most low-bit-rate encoding schemes won't tolerate the elimination of very short bursts of non-signal, as they interpret those silent periods as part of the signal.

Each VoIP service provider has the option of devising his own proprietary scheme for minimizing bandwith requirement for his service. That's where the problems begin and end. Many years ago, the only encoding scheme used on telecom was Pulse Code Modulation PCM. As it happens, I was working in the VoIP arena twenty-five years ago when that was the new frontier and, consequently, had the opportunity to examine the various encoding schemes being considered. Now, I was interested in secure (encrypted) voice comm's, so my perspective was limited, but, I did note that some encoding schemes were much more adaptable to packetized comm's than others. I'm not aware of any "standard" way of reducing bandwidth between systems. The U.S. and Europe, for example, had used different encoding schemes even back when PCM was the standard. The U.S. military used a different scheme, with significantly lower bandwidth requirement, namely Continuously Variable Slope Delta modulation, and that was just as incompatible with the U.S. PCM scheme as the European one, though the U.S. and Europe both used PCM, though their compression schemes had a different modulus. As bandwidth became of greater concern, much attention was paid to Adaptive Differential Pulse Code Modulation, which relied on DSP reduction from the standard 64 kbps PCM to either 32- or 24-kbps which are rates often used with CVSD, and, not just coincidentally, which relied on transmitting the change rather than the absolute signal level on a given channel.

Now, consider what happens when one does things to the data density of a voice channel, which normally relies on a steady flow of data from the source, when bits are missing, as can happen when the system detects non-signal (silence) or when the data transport causes packets to be routed over different paths that take different amounts of time ...

That's why VoIP, and any other scheme for packetizing voice data streams, e.g. the cellular scheme, behaves differently than the old, reliable, land line to which we've all been exposed.

Since gadgets like this "magic jack" are done on a limited budget, unlike commercial telecom systems, it's clear that they're unlikely to have the best and most reliable systems for dealing with the complexity of voice comm's on a Packet-Switched Telecommunication Network. I hope that makes it easier to understand why devices such as this gadget sometimes do a very good job, and somtimes not.

RE