Ryan - KJ7GIE
One thing I don't see anyone talk about, and because no one talks about it I find it causes confusion, is the difference between Modulation/Demodulation (MoDem) and the application which is riding over the modem. Here is a very quick summary which, if useful to expand on, I'm happy to talk more about it.
Bits - the 1s and 0s that make-up data communications
Serial communications - Most people are familiar with "RS-232" but often identify this as a 9 or 25 pin connector. Actually, RS-232 is a specification for how to represent bits using voltage changes referenced to ground. For this reason the two sides must share a stable ground in order to detect a voltage change referenced to whatever ground is. Now, ground voltage doesn't matter as long as its stable (although the spec only requires devices to handle +/- 25 volts it seems). There are various control pins for RS-232. Other serial standards exist with some supporting a single pin. Some used a lot more pins like my favorite v.35 which is a big clunky 34 pin plug but runs at significantly higher speed than RS-232. Serial can be, and often is, emulated over USB.
Modem - A modem (and give me tremendous leeway here in my description) on the TX side modulates the voltage changes on a serial interface into the audio spectrum and on the RX side demodulates that audio into actual voltage changes for the rx equipment. Any technology you can hold a voice conversation over you can attach a modem and "extend" a serial cable. This includes your home phone, your radio, sat-phone, your cell phone, tin cans and a string, etc.
Now, the part that I think gets confusing is that there are hundreds of ways to modulate voltage changes. And, as you can imagine, each one of them is designed to take advantage, and suppress disadvantages, of whatever "voice channel" is being used. Speed vs. reliability is the game.
Vara vs. "Packet Winlink".
Vara uses orthogonal frequency-division multiplexing (OFDM) to do the modulation / demodulation. You can look-up the specifics of that if you like. It also uses Forward Error Correction (FEC) which I'll touch on later.
"Packet Winlink" + Soundmodem, as described in Paul's link, is using a protocol derived from X.25. Now, I have a funny history with X.25 because it was a competing standard to Internet Protocol (IP) back in the ancient times. X.25 is ancient and I'm sorry to anyone who actually built these networks but it is. But, X.25 has 1 tremendous advantage going for it: it's silly reliable. And, that silly reliability is probably why it was chosen to develop AX.25 which became somewhat of a standard within the amateur radio community. That silly reliability comes with the tremendous cost of slow speed. Ultimately, that speed disadvantage is why it was scrapped as a viable commercial application and IP won. But, I digress. The other advantage, because its ancient, is that A/X.25 has tons of reference implementations and so is, somewhat, universally available on all platforms.
Winlink - Then, there is the winlink, or the application protocol, itself. I believe the current version of this is Open B2F. It specifies _which_ bits need to be exchanged and in what order to deliver and forward a message using Winlink. This "application protocol" does not change no matter which method you decide to initiate your winlink exchange (telnet, vera, winlink packet).
The actual "bits" transmitted using winlink don't change based on which modulation method you use. But, the way those bits get "modulated" do. No question Vara is faster than AX.25. But is it reliable? That's an exercise I've not done or looked into. Again, reliability vs. speed is always the battle. It's the same as a voice conversation. If I talk very slowly, repeat each word 5 times, you repeat back to me 3 times what you heard, I acknowledge you got it right, and you acknowledge that you received my acknowledgement before I speak the next word then we're going to have a _very_ reliable conversation that takes days to complete. As our radios get better, the expectation is that hearing becomes more reliable. Additionally, we're now using very sophisticated modulation methods that selectively use the limited audio spectrum to provide error correction _without_ having to retransmit bits (FEC). Vara using a mathod of FEC called Turbo Code. Using FEC becomes a huge speed advantage as it means you don't need to retransmit bits as the tx side already provided enough data for the rx end to reliably reconstruct any bits they may may have lost. You're transmitting just bit more data than the raw message but overall you're transmitting significantly less than repeating yourself.
Software - Winlink provides the application. You run it on your computer.
Modem? This gets tricky. We need _something_ to convert bits to audio tones. In the case of Vara this is also software. Vara, running on your computer, is a software modem. It's converting bits to audio tones and "playing" those audio tones out to a sound card. Soundmodem provides the same function. What's a TNC? A TNC is a modem+packet assembler/disassembler. Now, I didn't explain that part assembler part. You see, X.25 has very specific data its going to wrap around bits that make-up your winlink message. In it's most bare form Winlink will communicate the message bits, over serial, to the TNC. The TNC will "wrap" (or really assemble) this in an AX.25 packet, modulate it, key the radio, and play the modulated audio. The TNC will also "listen" for X.25, demodulate it, disassemble it, and flip the voltages on the serial to the PC where the software will display the message. So a TNC is lot more than a modem and may implement one or more of the network layers like AX.25, PACTOR, Winmor, APRS, etc. This lets the software on the computer feed the TNC raw bits without knowing what happens next. TNCs like this have fallen out of favor simply because our PCs are powerful enough to emulate everything a TNC does in software (except the sound card part, see below) cheaper and we can change from AX.25 to Vara in real time without swapping boxes (like Paul is).
Sound Card? - All right, so we've got a winlink set of bits, we've got the network formatted bits (Vara or AX.25) wrapped around it, and we've modulated those bits into the audio spectrum. How do we get that on the air? We use a sound card. Sound cards are responsible for converting digital audio into analog audio via use of a Digital Audio Convertor (DAC). This _cannot_ be emulated. The analog bit means we really need an analog signal just like you produce when talking into your microphone. If your radio does not have a built-in sound card then you need one between the radio and and the computer. The interface to the computer is digital (commonly USB) and the interface to the radio is analog (usually a phono plug or DB-9 connector). Now, the additional bit we need for the radio is the serial pin that keys the radio so that the radio knows to switch to TX mode when the sound card wants to tx sounds. That's why the DRA series and other "radio sound cards" are needed.
All bits deconstructed:
Software - Network - Modem - Sound Card - Key the Radio - On the Air
In my Vara set-up it looks like this:
Winlink(software) -> Vara(network, modem, key the radio), IC-7100(sound card, emulated PTT pin over USB, On the Air).
Anyway, this got a bit long. But I hope it helps. Some radios make this very easy (like the IC-7100). Others just need a sound card with the ability to key the radio (and a horribly cheap way to do this is with VoX. Don't use VoX). As time marches on I assume we'll see more packet implementations that emulate everything right up to the sound card. And, I suspect, it's not too long before more radios will embed sound cards to do packet; even HTs. This is all a side-effect of PCs getting really fast which allows for very sophisticated ways of hearing tones in the noise.