| |
ZL3AI > APRDIG 05.10.03 10:28l 95 Lines 5015 Bytes #999 (0) @ WW
BID : 2095-ZL3AI
Read: GUEST
Subj: TAPR Digest, Oct 02, 5/5
Path: DB0FHN<DB0RGB<OK0PPL<DB0RES<ON0AR<VE3FJB<W4JAX<WB0TAX<W7NTF<KD7HAH<
GB7BHX<ZL2BAU<ZL3VML
Sent: 031004/1258Z @:ZL3VML.#80.NZL.OC #:9738 [Chch-NZ] FBB7.00i $:2095-ZL3AI
From: ZL3AI@ZL3VML.#80.NZL.OC
To : APRDIG@WW
Subject: Re: Silk Purses
From: <BobDonnell@arkalmus.com>
Date: Thu, 2 Oct 2003 17:08:17 -0700
X-Message-Number: 35
A deviation meter/scope normally provides its own discriminator, and
connects directly to it. So it's the most accurate measure. However
optimizing for reproduction of 1200 baud AFSK packet are not as critical
- the biggest is that the audio make it from one TNC to the other with:
1) enough level to activate the receiving TNC
2) low distortion of the sine waves
3) low distortion of the balance in level between the 1200 Hz tone and
the 2200 Hz tone.
#1 is pretty simple - if your TX audio level is too low, stations set up
to copy other stations won't have enough audio to decode your signal.
The modem has a limited dynamic range. How limited depends on the
specific modem. Also, if your level is low, a weak signal will result
in a lower signal to noise ratio, making the job of the modem chip that
much harder.
#2 is where reducing the transmit audio level below that which causes
the deviation limiter in the transmitter to be activated. Distorted
tones that reach the modem are also seen as noise, and degrade the
modem's capability to decode the packet, under weak signal conditions.
Also, during the time when the instantaneous RF frequency of the
transmitter is resulting in a frequency that (through the superhetrodyne
process) ends up near the edges of the IF filters, that additional
distortion due to group-delay differences in signal propagation through
the filter, can result in adding distortion.
#3 is also affected by high drive levels. With the pre-emphasis of
audio tones through a conventional voice-grade FM transmitter, the 2200
Hz tone will be increased in level (prior to the deviation limiter) 3-6
dB relative to the 1200 Hz tone. On the receiver, (hopefully) matching
de-emphasis is applied. The purpose of this is to reduce high-frequency
noise content on scratchy signals by the amount of the de-emphasis.
Since for most situations where voice information is being sent, most of
the energy is at low frequencies, the pre-emphasis of high frequencies
results in increasing the audio energy at those higher frequencies in
the transmitter. A result of the pre-emphasis/de-emphasis process is
that when radios are used where the pre-emphasis and de-emphasis are
poorly matched, the receiving station may hear some stations that sound
muffled, if there's too little pre-emphasis, or too much de-emphasis.
If things are the other way 'round, with the transmitter having too much
pre-emphasis or the receiver having too little de-emphasis, the received
audio may sound thin or excessively crisp. So where does this come into
the discussion of packet? Well, we're using voice radios to send packet
data. For best effect, the audio tones created by the sending TNC
should be faithfully reproduced at the receiving TNC, both in respect to
distortion and audio level. If the sending TNC is driving the
transmitter into deviation limiting, the 1200 and 2200 Hz tones will
have effectively had the pre-emphasis undone, because the deviation
limiting will assure that neither tone results in more than 5 kHz of
deviation. So when the signal is received, and the de-emphasis is
applied, the 1200 Hz tone will be 3-6 dB less than the 2200 Hz tone.
This along with the addition of distortion by the limiting circuits and
the group delay in the receive IF filter all contribute to making that
signal that sounds so loud in the speaker much harder for the modem to
decode - and there goes another lost packet.
With regard to using a spectral waterfall or real-time analyzer, that's
an interesting thought. The things it might be able to show is harmonic
generation in the radio chain. 3rd harmonics are the most likely to be
generated by the limiting action, so they're going to be at 3600 and
6600 Hz. I don't think the waterfall displays of any of the typical
waterfall displays are going to cover that, though some of those
intended for general purpose audio frequency analysis might. Depending
on the resolution of the display system, it might be possible to see if
the two tones are making it through the chain of
TNC->transmitter->receiver->TNC in a balanced way, and it might as a
result be possible to see the effect as described in #3 above comes into
play. However too, if you're doing sampling with a PC sound card, the
general purpose sampling software will usually include a VU or record
level meter of some sort. It can be used in a similar manner to the
scope.
With regard to using a scope on receive audio output vs. discriminator
output, use what you have - it's got to be more accurate than listening
by ear, or ignoring the problem. If pre-emphasis and de-emphasis are
reasonably matched, the levels of the two tones will come out of the
speaker jack at pretty equal levels, where they won't if the scope is
connected to the discriminator.
Hope that helps!
73, Bob, KD7NM
---
END OF DIGEST
Read previous mail | Read next mail
| |
