[Scan-DC] Trunktracker III vs IV

[Dave Emery]

On Sat, Dec 04, 2004 at 11:25:27AM -0500, Doug K. wrote:
> FWIW, as you probably have seen, I just got a 796 about 3 weeks ago, and 
> haven't wrung it out much yet... but... living in, and listening to, 
> Montgomery County, things are pretty clear - but I have noticed that 
> Howard's fire dispatch is pretty garbled (although the apparatus seems to 
> be more clear).  All I can think of is that this involves distance, because 
> when I went over to a friend's in Columbia last Sat I threw the 796 in the 
> car, and Howard's transmissions seemed to become more clear as I went east 
> (ie toward Howard County).  But... from what I've read about digital 
> communications, if I understand it correctly (!), it's either there or it 
> ain't, and what you hear is what you get.


	To the first order it is sort of true that APCO-25 is a either
you hear it or you don't medium like most modern forward error corrected
digital communications.

	But in fact this is not entirely true.  The vocoder speech data
in APCO-25 is actually sorted into three categories - bits for which
robust error correction is provided, bits for which error detection and
more limited correction is provided and some bits for which basically only
error detection is provided.

	The intent of this is to provide a gradual degradation of speech
quality as bit error rate (indicative of a weak or garbled signal) goes
up without the using the extra bandwidth it would take to provide robust
forward error correction of the entire vocoder frame.

	The bits given the best protection were chosen as those most
necessary for the basic speech sounds to be (barely) intelligible, the
less protected ones have less impact on basic intelligibility and more on
naturalness and speaker recognition.

	And any particular implementation of IMBE and APCO-25 framing
may handle errors (something called error concealment) when they do
happen with more or less impact on speech intelligibility.   Basically any
vocoder attempts to patch over errors by repeating the sound  generated
during the last good frame or by using the speech parameters for the
last good frame if the more minor bits are in error.  But some do this
better than others.

	How effective this is highly dependent on the quality of the
vocoder implementation (eg how good the DSP firmware in the radio is).
Virtually any properly coded standard conforming IMBE firmware will
reproduce almost the same sound given error free data - but given
errored data there may be significant differences.   And the quality of
the hardware/firmware used to recover the C4FM or CQPSK data from the
radio channel obviously determines how many errors the vocoder sees -
which means a cheaper or poorer radio will produce more errors and less
clear sound than a better radio with a lower error rate even if they use
the same vocoder firmware.

	And it takes time, experimentation, and lots of intelligibility
testing to improve a vocoders performance with errored data.   To some
extent this is subjective and function of the language the speakers are
speaking and their gender and local dialect - also the fidelity and
frequency response of the microphones, speakers and so forth used.  So
it involves art and human choice and preferences as well as hard math.

	Most of the vocoder firmware for APCO 25 radios was created by
DVSI in Massachusetts - which owns the rights to the MIT patents the
IMBE vocoder is based on.   I am not sure if they offer various grades
of speech quality in the presence of errors in their products, but it
would not surprise me if some of the big players in the APCO 25 market
(like Motorola) have either improved this themselves or paid for extra
features not offered in the standard DVSI IMBE products.

	And it would not surprise me either if GRE and Uniden have not
paid for the best fidelity, best error performance vocoder firmware
possible for their APCO-25 scanners.   Scanners are very price sensitive
and paying even a few dollars in extra royalties (or for a faster DSP
chip or more RAM) for better audio in the presence of errors is far less
important than a low sales price.  After all, it is unusual for lives or
property to be at stake in scanner reception whilst this is completely
expected for PS radios themselves.

	In any case, what this all means is the APCO-25 audio very
definitely gets worse as the signal gets marginal and the bit error rate
becomes significant.   So it is not a myth that audio may be poorer from
some units or sites or that as one gets closer to a site the audio may
get better.   Nor is it other than a myth that just because the audio is
digital all radios will hear it exactly the same or not at all.


-- 
   Dave Emery N1PRE,  die at dieconsulting.com  DIE Consulting, Weston, Mass 02493