>  Is track 2 the best way to test a TNC for typical audio, or would data recorded from an actual radio speaker jack be a better?  I'm interested in learning more about the audio that most users would present to a TNC, which would not be from the discriminator.
>


The majority of users take de-emphasized receive audio from the speaker of a radio for use with their TNC because it's the easy thing to do, but this has many dis-advantages.

1)   The level varies, obviously,  depending on the volume control setting.

2)   The audio de-emphasis curve in many radios is implemented incorrectly and  varies wildly from the EIA standard.  (Some radio  designs apply excessive amounts of audio high-frequency rolloff (de-emphasis) in the RX audio power amp in an effort to attenuate popping noise and make the radio sound "quieter".)  A TNC connected to such "mangled" audio will perform less than ideally.

3)   The speaker audio is usually squelched which can add many milliseconds of delay to the start of decoding, forcing other users to increase the TXD time to compensate.

4)    A de-emphasized speaker-audio-type  connection absolutely WON'T WORK for 9600 baud operation.  


Many users have been inhibited from using the direct discriminator connection due to the need to make connections inside the radio.  However, many modern radios now offer a 6-pin mini-DIN "data" or "packet" connector which provides easy access to the raw discriminator audio. In this case, the discriminator connection will overcome the problems above.   

Details on the signals available at the 6-pin mini-DIN connector, which is standardized by different radio manuafacturers are in this PDF on my website:

     http://wa8lmf.net/msicinfo/MiniDIN6-Packet.pdf


VARIOUS FACTS


    * Traditionally, 1200 baud packet has been assumed to have  TX pre-emphasis due to users inserting their TX audio tones into the mic jack, and de-emphasis at the other end due to being taken off the speaker,   In other words the 2200 Hz high tone has nearly twice the deviation on-the-air than the 1200 Hz low tone, due to the transmitter mic audio pre-emphasis. 

      (You can see this by putting a TT or KPC3 in the CAL mode and selecting one tone at a time while observing the resulting deviation on a service monitor. Typically the high tone will be set to yield about 3.5 to 4.0 KHz dev. The low tone will then produce about 2.5 to  3.0 Khz or so.). 

    * 9600 baud operation is intrinsically transmitted "flat" since it uses direct FSK of the carrier (rather than audio subcarrier tones),  achieved by applying DC logic-level signals directly to the transmitter modulator, bypassing the mic amp and pre-emphasis entirely..

    * The Kantronics TNCs provide jumper-selectable equalization (de-emphasis) inside the box.  You can disable the de-emphasis if you use already-demphasized audio from the speaker, or enable de-emphasis if you use raw discriminator audio. 

    * The Kenwood D700 and TH-D7 transmit the two tones at EXACTLY THE SAME deviation. (Their internal TNCs are connected DIRECTLY to their TX modulator and RX discriminator, bypassing the TX mic channel with it's pre-emph and RX audio channel with it's DE-emphasis.) 

    * PLL-based decoders (such as the TNC2 and clones)  are very intolerant of the low tone being higher level than the high tone.  They must see the high tone at a level equal to or higher than the low tone.  When you have reverse "twist", i.e. low tone at higher level than the high tone, they often fail to lock up and decode.     Tuned-filter and zero-crossing-detect TNC demodulators are much less affected by this de-emphasis issue than PLL types. However, the de-emphasis (or lack therof) does also create phase shift that will cause the zero-crossings to be slightly advanced or retarded from their correct point in time.

    * If the RX de-emphasis curve on another radio is a little bit excessive,  packets received from Kenwoods will arrive at the attached TNC with the low tone far LOUDER than the high tone. This can cause the attached TNC to fail to decode bursts from Kenwoods. 



The bottom line of this intractable mess is that some users (mostly Kenwood owners)  transmit "FLAT" while others transmit "PRE-emphasized". Ideally the receiving TNC has to somehow accommodate both.   The best compromise is to split the difference. Take the receive audio from the discriminator,  and apply just a very slight amount of  de-emphasis at the TNC input.  

Or to accommodate users that connect to the radio speaker output, the TNC design should provide for a slight amount of optional (jumper-selectable) high-frequency BOOST at the TNC input.  This is easily done by using a smallish audio input coupling capacitor on the TNC, paralleled with a much larger one disconnectable with a jumper. When the small cap only is connected, the lower audio frequencies will be attenuated, creating the same net effect as a high-frequency boost.)



On the CD, track one emulates receive via raw discriminator, while track two emulates de-emphasized receive via speaker or earphone jack.   TNC developers are encourage to use both tracks alternately to optimize their TNC performance to tolerate BOTH flat AND de-emphasized audio.


To make the other tracks (recorded "flat" straight off the discriminator) simulate speaker audio,  you will have to connect a de-emphasis network between the CD player and the TNC. 

Assuming the CD player has a low output impedance  ( 1K or less, which is typical of consumer audio devices) and the TNC under test  has a high input impedance (disable any jumpers that select a 600 ohm input Z), then a simple RC network can do the job.  Connect a 100K resistor in series between the CD player out and the TNC input.  Shunt the TNC input to ground with a  750 pF capacitor.   Or use a 47K resistor and a .0015 uF capacitor.  Approximate this value by paralleling a .001 and a .0005 uF cap.)  Either of these two networks will approximate the standard 75 microsecond de-emphasis network.