Being licensed for >27 years, my first steps into RTTY were during my 7QNL DXpedition in Malawi in May/June 2014.
According to ClubLog ‘digital’ modes from Malawi were (relatively) scarce, so I was obliged to give it a try.

It appeared that the built in sound card of my laptop did not work properly (with Windows 7?).
Fortunately I took a cheap USB sound dongle with me, which worked. The results were (imho) flabbergastingly good.
Signals with (very) low SNRs decoded properly.

I used the 300 Hz CW-filter in ‘DIG’ (i.e. ‘LSB’) mode on my FT-857, resulting in a ca. 1450 Hz mark pitch.
(It took me one day to discover that my dial frequency had to be tuned 1450 Hz up, anyway ….
I was on the right side of the pile up ;  -)

At home I received a few requests from my QSL manager to verify some QSOs people claimed,
as they could not find themselves in the online (Club)log.

Fortunately I recorded received audio of the whole operation for pile up research purposes.

Soon I discovered the disadvantage of not recording the audio of my own transmissions.

Before the DXpedition I thought about an ‘audio fork’, but limited time forced me towards other priorities.
Fortunately some audio of my CW transmissions ‘leaked’ into Audacity. While listening to my phone
transmissions it remained silent. This was also the case for my RTTY transmissions.

During such DXpeditions there are always some cases of doubt, due to pile up magnitudes, pile up spreads,
erroneous databases, my own typing errors, false copied calls, QRM, LIDs, etc.

There was a specific ‘RTTY case’ where I could not distinguish if I worked ‘station A’ or ‘station B’ as they
simultaneously showed up in my recorded audio (and were ‘almost simultaneously’ decoded by mmTTY
but with a slight offset in pitches).
Moreover.. the two stations had the same prefix and shared the first letter of their suffixes (iirc).

I logged ‘station A’ (which came back with CFM) but ‘station B’ claimed the QSO. Of course it was possible I made a typing error?
After intensive post processing/filtering of the recorded audio I still was not able to distinguish between ‘station A’ or ‘station B’.

The result was that ‘station A’ remained in the log and I had to disappoint ‘station B’.
If I had recorded my transmitted RTTY-AFSK, for 100% sure I was able to distinguish between station A and B (afterwards).

The aforementioned experience brought me to instant measures.

The 7QNL RTTY interface consisted of a bare USB sound dongle (U$ 0.99 incl. shipping on Ebay ; -) interfaced with
two 3.5mm stereo jacks into the dongle, and with a mini-DIN into my FT-857.

As the price of two 3.5mm jacks exceeds the price the USB sound dongle I now directly soldered the audio cable in/at the dongle.

RX- and TX-audio is levelled with two series resistors (12k7 each, your mileage may vary) and
in- and outputs were connected through a ~47k resistor so that transmitted audio runs into the recording port during TX.

Needless to say, this works for audio generated by an external device (such as a computer/laptop).
Theoretically it should work for ‘phone’ too, when e.g. the computer is used for sound processing purposes.
I have not figured this out (in N1MM) but perhaps somebody else may give this a try?

Next … is connecting the TX-mic to the interface (with a 3.5mm jack), and mute it (with a FET) while receiving.

Note: the contrapsion is used in mono (one sound channel, see solder drop on the green audio output (i.e. TX audio) connector).
Note2: Ground (GND) of the audio cable is connected to the metal casing of the USB connector.

Here and here two RTTY QSO’s. Replay them with e.g. mmTTY or fldigi.

Below pictures of the resulting interface (click on the images to enlarge in a new window).