The Utah VHF Society Using conventional analog test gear to evaluate and test
D-Star systems
Purpose of this page:
As new technologies come into use on the amateur
bands, there is an increasing challenge to be able to evaluate and
support these technologies. In the past, conventional test
equipment has been used to maintain and diagnose such systems, but
with these new technologies there is a challenge to be able to
provide a means of being able to support such systems in a
meaningful way.
An example of such a technology is D-Star. As this (and
similar) systems become more widespread, the challenge to be able
to design and maintain such systems increases. Using
conventional test gear, one is limited in exactly how much
diagnosis is possible - but there are still a few things that can
be done to determine important aspects of the system's
performance.
Important Notes:
This page deals only with the narrowband
D-Star modes as found on the VHF and UHF U.S. amateur bands,
and not the "high speed" modes that may be used on 23cm.
ONLY analysis of the
disruption of voice transmission was considered. If the
transmission of data is to be the primary concern rather than
digital voice, it is possible that even more protection may be
required to maximize performance.
A bit of background:
As transmitted on-air, a
Figure 1:
Comparative spectra of D-Star (left) and typical analog
(right) signals. In each case, vertical divisions are 10
dB and horizontal divisions are 2 kHz. An unmodulated
carrier has been overlaid atop both images and the green line
represents the level of the unmodulated carrier. Click on image for a larger
version
D-Star signal is simply FM: More
specifically, it uses Frequency-Shift Keying (FSK) to convey
data. By properly shaping the modulating waveform and
appropriately choosing the amount of deviation, the transmitted
spectrum can be adjusted to minimize the occupied bandwidth while
still maintaining reasonable power efficiency in terms of being
able to transmit data.
Figure 1 shows the typical transmit spectra of a D-Star
signal and compares it with a typical analog NBFM signal, showing
the "peak + average" power density. Because the data stream
is fairly consistent in its spectral content, the spectral makeup
of a transmitted D-Star signal is also very consistent.
Because the symbol rate is 4800 baud, the highest modulated
frequency using this scheme is 2400 Hz, with numerous subharmonics
resulting from the modulation of the data stream. Another
strong component of the D-Star voice signal, as can be seen in Figure
2, is that of the 50 Hz voice frame rate: It is this
that causes the characteristic "buzzing" sound that is heard when
a D-Star signal is monitored on an analog receiver.
As can be seen from Figure 1 the majority of the energy of
the transmitted D-Star is constrained to within a few kHz of the
carrier, with "nulls" at approximately +-3.6 kHz from the center
frequency and sidebands of decreasing energy beyond that. It
is through careful shaping of the modulated signal and the
appropriate amount of deviation that this transmitted spectral
shape is obtained.
Comment: Please take note of the resolution bandwidth of
these analyzer plots and its effect on the relative power
density of the modulated and unmodulated carriers.
D-Star baseband modulation:
The baseband modulation (that is, the signal being fed into the
modulator) of the D-Star signal consists of 0's and 1's being
modulated onto the carrier, but to simply throw a 0 or 1
(represented by a logic level) at the modulation would result in
an abrupt frequency/phase change, causing the transmitted signal
to occupy considerable bandwidth. It makes sense, then, to
slow down the rate of change that can occur during modulation -
but one can only go so far: With the 4800 bit-per-second
D-Star signal, we could send alternate 0's and 1's. Without
filtering, this would become a 2400 Hz square wave, but with
filtering, this could be turned into a 2400 Hz sine wave - a
signal that would take a fairly minimal amount of bandwidth to
modulate.
Filtering the original "square wave" data into something
resembling a sine wave is rather tricky. If all that you
wanted to do was to generate a 2400 Hz sine wave to transmit
alternating 0's and 1's then it would be easy, but with data, you
will have a combination of 0's and 1's - sometimes several of each
in a row. When trying to filter the original data, one must
make sure to minimize the filter's "memory": Suppose that
you had been sending a bunch of 0's - but then a single "1" comes
along, followed by a bunch of 0's. With a simple,
un-optimized filter, everything will settle out to a "0" state -
but when a "1" comes along, it would have to be able to fully
change to a 1 - and then fully change back to a 0.
Improper filtering will tend to cause the previous state to
"linger" and it can be more difficult to determine, upon decoding,
if and when, exactly, that "1" began and ended - and any
uncertainty on whether a "0" or "1" was received will
increase the likelihood that one would get it wrong!
To help solve this problem, our single "1" is turned into a smooth
pulse - one that can go from 0, to 1, and back again smoothly -
and it so-happens that the filtering used to do this is Gaussian
in nature - the name referring to a particular shape of the pulse
and its properties. It also so-happens that with this sort
of pulse filtering, if you were to alternate 0's and 1's, you
would, in fact, end up with a nice sine wave as can be seen in
parts of Figure 3. Because the fastest that one
could "change" the waveform with the 4800 baud D-Star signal is,
in fact, 2400 Hz, that is the maximum peak frequency that can be
modulated.
Figure 2:
Spectrum analysis of a baseband
D-Star signal. There is a null at 4800 Hz correlating
with the bit rate and there are strong spectral components at
intervals of 50 Hz that correlate with the 20ms voice
frame. Click on image for a larger
version
Because we aren't sending just a "01010101" all
of the time, this nice, continuous sine wave is constantly being
interrupted to form the data stream and in so-doing, multiple
spectral sidebands result - which is why, in Figure 2,
there is not just a peak at 2400 Hz. Instead, energy is
spread around 2400 Hz but very little of it goes above 4800 Hz.
There is another important aspect of the D-Star modulation:
The amount of deviation. For mathematical reasons, good
spectral and power efficiency for this type of modulation occurs,
with data, when one sets the total deviation to be one-half of the
baud rate: Because of the baud rate, the total deviation is
2400 Hz, or +- 1200 Hz, and this particular setting is referred to
as Minimum
Shift Keying. Because we have already pre-filtered our
data with a "Gaussian" filter, the combination is called Gaussian
Minimum Shift Keying, or GMSK.
Don't let all of this scare you: All one really needs to
remember is that the D-Star's baseband modulation consists of bits
of 2400 Hz sine waves.
Another important aspect of the D-Star's baseband modulation is
the limitation of low-frequency components. If too-many 0's
or 1's were transmitted sequentially, the low-frequency content of
the baseband would increase and the DC level representing a 0 or a
1 could become indistinct - particularly if capacitive coupling
were used. Another problem with low frequency content is
that the radio's synthesizer works by locking the "average"
frequency. Because this is frequency modulation, the
synthesizer was designed to avoid canceling out the modulation -
but this is typically done by preventing the synthesizer from
responding to any by the slowest changes in frequency. If
the baseband modulation has too much low-frequency content, the
synthesizer will attempt to track it and cancel it tout. As
it turns out, the data stream used for D-Star voice has some
fairly low-frequency components, most notably the 50 Hz "voice
frame" rate. Because of this, the frequency response of the
baseband must extend down well below 50 Hz to avoid distortion of
the waveform - and the need to pass these low frequencies results
in an inevitable "frequency wander" as we'll see below.
D-Star transmitter and receiver:
As it turns out, the D-Star transmitter is just an FM transmitter
- with a few special considerations given to assuring that it will
properly pass both low (<30 Hz) and high frequency (to 4800 Hz)
energy with a minimum of distortion. Likewise, for
reception, a D-Star receiver simply takes the demodulated signal
from the discriminator and passes it to the modem board.
Because D-Star is just FM, it follows that standard test gear
designed for use with FM communications gear may be useful in the
evaluation and diagnosis of D-Star equipment - although some of
the techniques for doing so are different for standard analog
voice.
Comment: For the purpose of this discussion we are
ignoring the fact that some Icom radios - such as Icom repeaters
and the ID-1 - generate their D-Star modulation through
baseband/quadrature methods rather than direct FM.
Tests using analog test gear:
There are a number of tests that one may do using normal analog
test gear to verify performance of a D-Star radio. To some
extent, these rely on the assumption that the codec in the radio
is working properly, but if there are other problems with the
system, one may be able to determine what they are.
Transmit power test:
Being that the D-Star transmitter is simply an FM receiver with a
digital codec, one can perform normal tests for forward and
reflected power: No surprise there. Like analog FM,
there is no amplitude component present and
amplifier linearity and measurements of PEP are irrelevant.
Frequency test:
It is possible to use ordinary means to determine whether or not a
D-Star transmitter is on-frequency: The modulation should
not skew readings of most test gear to any significant
degree. Don't forget that most D-Star radios are capable of
analog modes as well so one may simply switch the radio to an
analog mode to check to see if the radio is within specifications.
SINAD test:
One common test of receive system performance is the SINAD
test. For this test, a single, precise tone is generated -
usually 1 kHz - at a standard deviation - usually +-3 kHz in the
U.S. The level of this tone is then compared to the amount
of noise that is NOT at 1 kHz. For a
full-quieting signal, a SINAD reading of over 30 dB may be
expected for most radios, while a SINAD of just 12 dB sounds quite
noisy, but is still easily intelligible to most people.
Switching a D-Star capable radio to analog and running a SINAD
test is a convenient way to verify its performance. Note,
however, that for the D-Star digital modes, FM-Narrow mode is
used. Because the nominal peak deviation in "narrow" mode is
+-2.5 kHz, a deviation of +-1.5 kHz is often used instead of +-3
kHz for the 1 kHz test tone.
"Equivalent SINAD"
test:
It is possible to relate the SINAD in FM-Narrow mode to the
performance in D-Star digital voice mode. This is is
possible because the SINAD measurement tells us something about
the amount of extraneous noise in the receiver's baseband -
something that correlates very well with data errors! This
test is handy as it requires no special test gear at all, other
than what would ordinarily be used for SINAD measurements.
Figure 3:
Baseband waveform of a D-Star signal. In this image can
be seen a period of alternating 0's and 1's (toward the
right.) Also evident from this picture is a bit of DC
level (or low frequency) shift caused by the IC-91AD's
synthesizer attempting to track the data. Click on image for a larger
version
Comment:
For the measurements below, an Icom
IC-91AD was used. This represents a "typical" radio
used by a D-Star user.
For this test four levels of D-Star signal
disruption were investigated. Note that there was no attempt
to relate these conditions with bit error rates as the Icom gear
used at the time of writing had no provisions for doing so.
Instead, subjective measurements were used that could be easily
duplicated - with good repeatability - by anyone wishing to repeat
such a test:
"Clean" audio decoding: No bit
errors were observed over a period of 60 seconds.
"Mostly clean" decoding: One
"bloop" (an unrecoverable bit error) occurred every 10 seconds
or so.
"Ratty, but mostly copyable":
With this signal, it is possible for an experienced operator
to understand most of what is saying. With this
level of signal quality, it usually took 2-5 seconds to
achieve lock on the received signal.
Loss of D-Star sync: At
this error rate, not only has recovered speech become
unintelligible, but the receiver can no longer maintain
reliable synchronization on the D-Star signal.
For this test, two types of situations were
simulated using test equipment:
Weak signal degradation: For this
test, the signal level of a D-Star signal was reduced until
each of the 4 levels of D-Star signal disruption were
achieved.
D-Star adjacent channel degradation:
For this test, another D-Star signal was generated 8 kHz
offset from the one being received. With the test signal
set at -90 dBm, the level of the interfering signal was
increased until each of the three levels of D-Star signal
disruption were achieved.
When each of the four levels of disruption were
reached, the IC-91AD was switched to FM-Narrow mode while, at the
same time, the test generator was switched from generating a
D-Star signal to generating an FM signal modulated with a 1 kHz
tone at +-1.5 kHz deviation: At this point, an un-weighted
SINAD measurement was taken using the audio from the IC-91AD's
speaker connector.
Comment:
The behavior of the signal degradation,
whether it was due to weak signal or degradation due to
adjacent channel interference resulted in the same
"equivalent" SINAD. For this reason, these two types
of signal degradation are treated as one in the table below.
The correlating SINAD levels were:
Table 1:
Comparing
SINAD
of
an
analog signal (as received in FM-Narrow mode) to the perceived
quality of a D-Star signal.
Quality
of digital signal
SINAD
in "Narrow FM" mode
Additional
comments
"Clean"
D-Star decoding achieved
>17-18dB
SINAD
No audible decoding errors
of digital audio
"Mostly
clean" decoding
15.5-16dB
SINAD
Occasional "bloops" in
audio (approx. one every 10 seconds)
"Ratty,
but mostly copyable"
12dB
SINAD
Considerable degradation
in the digital signal, but mostly copyable by an
experienced operator. Synchronization to a received
signal typically took 2-5 seconds.
"Loss of
sync"
<9dB SINAD
The D-Star decoder would
not maintain reliable lock of signal and no intelligible
audio was recovered
Comment:
With the narrower bandwidth used for D-Star
recording, a theoretical 2-2.5 dB weak-signal gain should be
obtained due to the reduction in detection bandwidth, as
compared to the "normal" (+-5 kHz) FM mode. In reality,
however, this difference is closer to 1 dB owing to some S/N
gain in the wider bandwidth due to the wider deviation and the
commensurate increase in recoverable audio.
The above test requires that one has onhand
a D-Star receiver that is capable of analog (FM) reception in
the "FM-Narrow" mode. Most - if not all -
portable/mobile D-Star radios are capable of analog operation
in either the "narrow" (+-2.5kHz) or "wide" (+-5kHz) FM modes.
Comparison of Analog and Digital signals of equal
levels:
In this test we decided to see how D-Star signals of various
qualities correlated with conventional "wide" analog FM signal in
terms of copyability. This test can be useful in that, using
one's own experience as well as conventional test gear, get a
general idea as to how a D-Star system might perform under similar
circumstances.
It is important to remember that this test may not be entirely
valid in the presence of adjacent-channel interference as a
typical D-Star receiver has somewhat narrower bandwidth and it may be somewhat
less-susceptible such interference: If done over the air,
this sort of testing should be done while potential
adjacent-channel signal sources are not transmitting.
For this test, the following configuration was used:
An Icom IC-91AD was used to generate a
signal in FM-Wide (+-5kHz) mode.
Both analog and digital signals were
received using an Icom IC-2200H.
The signal/noise of the received was
reduced and SINAD measurements were taken using a 1 kHz tone
modulated to +-3kHz.
SINAD readings were measured using the
external speaker connector of the IC-2200H using both
"unweighted" (unfiltered) and CCITT weighting as noted.
At each "step" of SINAD readings, the 1 kHz
tone used for the SINAD measurements was sent, analog voice
was sent, and then both the IC-91AD and IC-2200H were switched
to D-Star voice (DV) mode.
For each test, the audio from the IC-2200H
was recorded.
Audio recordings made of the received signals
consist of three parts:
About 10 seconds of 1 kHz tone as received
in analog mode used for measuring the SINAD.
A voice recording transmitted and received
in analog with the peak deviation set to +-5 kHz, the standard
for analog FM use with the IC-2200H set for "Wide" FM mode
(e.g. standard for +-5kHz deviation.)
One or more repetitions of a voice
recording as transmitted and received in D-Star mode.
Table 2:
Comparing
SINAD
of
an
analog signal (as received
in "Wide" FM mode) with
Analog
signal is quite noisy: Copyable by experienced
operators with little or no difficulty and with only minor
difficulty by inexperienced listeners.
There was considerable
degradation of the digital stream resulting in
"recognizable but mostly uncopyable" speech. At this level, it takes 5-7 seconds before signal
lock is achieved.
Analog signal is very
noisy: Generally copyable by experienced listeners,
with some difficulty by inexperienced listeners.
The receiver would not
lock on digital signal: Signal was briefly boosted
10dB to force lock (during the "This is K7" portion) and
then reduced to the original level.
Comments:
Please note: The SINAD
readings below are those measured using "Wide" FM (e.g. +/-
5kHz deviation) and are not directly comparable to
those SINAD readings in Table 1, above.
Additionally, for those measurements taken in Table 2,
an Icom IC-2200H was used as the receiver whereas an IC-91AD
was used for the results shown in Table 1: It
was found that the IC-2200H could outperform the IC-91AD by
1-2dB under identical conditions.
The above recordings have been MP3
compressed to reduce file size and the fidelity of the analog
portions, especially in the presence of noise, may suffer
somewhat. (Uncompressed versions of the above files
may be obtained by changing the .mp3 suffix in the above
links to .wav).
At the 12dB SINAD level, it usually took
2-5 seconds for the digital voice stream to acquire
lock. Users should
keep this property in mind when making short transmissions,
or if important information is placed at the very beginning
of a transmission.
When a D-Star transmission begins, it is
preceded by a short preamble that is used by the D-Star
decoder to rapidly recognize and acquire lock onto the
signal. If this preamble is missed, as may be the case
when signals are weak and/or multipathy, it can take several
additional seconds for D-Star decoder to lock onto the signal
and produce audio.
At the 7dB SINAD level, the D-Star decoder
usually locked within 5-7 seconds, but only a few bits
of the speech were recognizable.
As mentioned above, the D-Star decoder
would not reliably lock onto the D-Star signal at the
3dB SINAD level: The D-Star signal was briefly boosted
(during "This is K7") by 10dB to allow the receiver to lock
onto the signal and then reduced again to the 3dB SINAD level.
At weaker signal levels (7-12dB SINAD)
slightly better results (1dB or so) were obtained with the
digital signal when the deviation was artificially boosted to
the 3-4 kHz range, well above the recommended 1.2kHz
setting. Note: This is not a
recommended practice as it causes the transmitted signal to
significantly exceed the design bandwidth of D-Star.
"Why
are your results different from those obtained by the ARRL?"
In the June, 2005 issue of QST, there was a review of the Icom
IC-V82 HT. Associated with this review was a brief overview
comparing D-Star and Analog FM signal performance. ARRL
members may read this article here: http://www.arrl.org/members-only/prodrev/pdf/pr0506.pdf
In this article the ARRL lab reports that a D-Star signal
maintained "...solid, virtually noise-free communication,
equivalent to 'full-quieting' at any analog SINAD above
6dB." Our results do not reflect this and we thought that
the discrepancy was likely a result of possibly different
methodologies used in measuring SINAD. Fortunately, the ARRL
has put their "Test Procedures Manual" (available online to ARRL members at this URL: http://www.arrl.org/members-only/prodrev/testproc.pdf
).
Having reviewed the ARRL's procedures for measuring SINAD and
determined that our methods are equivalent to theirs, we are at a
loss to explain the discrepancy between our readings and those
stated in the June 2005 article, or why the results obtained by
the ARRL lab do not correlate with Icom's own
specifications: If you conduct similar measurements, please
inform us of your results!
Comment:
It is suspected that some of the signal/noise readings
mentioned in the ARRL article were observed at the modem's
input (e.g. discriminator audio) rather
than the radio's audio (speaker) output! SINAD
measurements taken at this point would, in fact, reflect a much
lower reading that those obtained after de-emphasis - such as
those at the speaker terminal!
Checking the deviation of a D-Star transmitter:
To create an MSK signal, the deviation of a D-Star transmitter
should be set to +-1.2 kHz: As mentioned above, this value
is chosen so that the total amount of deviation (2.4 kHz) is equal
to half of the bit rate of 4.8 kbps to generate an optimum signal
in terms of occupied bandwidth and BER performance.
To verify that a D-Star transmitter is set up properly, one may
use the same methods used for setting the deviation of any FM
transmitter. An important note here: For this
test, one must make sure that the test equipment is measuring
"flat" FM rather than PM, or FM with some sort of filtering
switched in (e.g. CCITT, etc.)
"Excess" deviation due to "PLL Wander":
There is a caveat with this measurement,
however: Some of the Icom radios (such as the IC-91AD and
IC-2200H) tend to suffer from "PLL Wander" as can be seen by
observing the low-frequency shift in Figure 3. This
effect is caused by the radio's synthesizer trying to track
low-frequency components (such as the 50 Hz "voice frame rate") of
the D-Star waveform with the result of the transmitter wandering
up and down several hundred Hz about the center frequency.
The result of this is that the "deviation meter" on many pieces of
test equipment may read an amount of deviation higher
than that of the D-Star's modulation. If this occurs - and
the deviation is set to +-1.2 kHz, this could result in the actual
D-Star deviation being set a bit too low, causing a slight amount
of degradation of the signal.
The amount of "excess" deviation seems to vary from radio to radio
and it probably varies with operating frequency band (e.g. VHF or
UHF) and the temperature and age of the radio as well. In
our tests, the amount of deviation for the same radio also varied,
depending on which deviation meter we looked at and how it was
able to track the low-frequency components: Some deviation
meters were fast enough in responding that this "frequency wobble"
caused the meter to read only slightly high - that is, about
+-1.4 to 1.5 kHz for a signal modulated to +-1.2 kHz, while others
seemed to accurately read the total amount of
frequency swing, which caused readings as high as +-1.7 kHz.
There is a solution to this: The use of the
monitor scope. Many service monitors or communications test
sets include an oscilloscope (either analog or digital) that may
be read to determine the precise deviation of a signal being
received. On these scopes, one can see the "frequency
wobble" - but, if the scope is correctly adjusted, you can also
make out the peak-to-peak values of the individual bits in
modulation waveform itself and, apart from the "wobble", determine
the true amount of deviation of the data.
Some Icom transmitters (such as those used in various Icom
repeaters and in the ID-1) do not directly modulate
their synthesizer but, instead, perform quadrature modulation at
an IF: These radios have far less "wobble" in their carrier
frequency as the PLL itself is unmodulated.
Note that the "PLL Wander" of these radios also increases the
effective bandwidth used by the transmitter. While difficult
to quantify, it seems as though this "wander" was on the order of
+-300 to 500 Hz, potentially increasing the "occupied bandwidth"
of the D-Star transmitter by up to 1 kHz! In practical
terms, the width of the filter in the receiver will accommodate
such a (relatively) minor frequency error, but this artifact of
the radio's operations should be kept in mind.
Generating D-Star signals using analog test
gear:
Because a D-Star signal is simply a special case of an FM signal
generated by applying an appropriate baseband signal to an FM
transmitter, it would make sense that one could apply this same
type of baseband signal to a good-quality frequency modulator and
create a D-Star signal. Some intrepid hombrewers have done
this by adapting an Icom D-Star module for their own use and
interfacing it with their own transmitter: This method works
well, but it can be rather complicated and expensive.
There is another way: Using a "canned" D-Star transmissions.
Because the baseband is simply audio, it would make sense that one
could simply "record" this audio from a D-Star transmitter and
play it back later - and this is, in fact, true!
There are several caveats:
The source of the baseband audio must be
"flat." What this means is that "discriminator
audio" is required as this has no
audio
bandpass
filtering or de-emphasis. Many service
monitors or test sets have "demod" outputs, directly from the
discriminator that have excellent frequency response - from
near DC to well over 10 kHz. Note that many test sets
also have various audio filters (such as CCITT or some type of
equalization) that should be disabled. Finally, some
test sets have both an "FM" and "PM" mode, so make sure that
"FM" is used!
The source baseband audio must be
"clean." For a faithful recording to be made, it should
be as free of noise and distortion as possible. If one
is using a service monitor, this is easily accomplished by
connecting the transmitter directly to the service monitor (as
one would do to measure transmitter power) and make a
recording. The fact that the "receive" bandwidth of the
receiver in the service monitor is wider than that of a D-Star
receiver is of little consequence if the signal fed to the
service monitor in this way.
It has been reported that some radios
with 9600 baud packet capability (such as the FT-817) may be
used to demodulate a D-Star signal for recording, provided
that the received signal is strong enough to be noise and
interference-free.
If you are willing to "hack" into your
D-Star radio, you can obtain baseband waveforms directly
from the radio's modem itself, avoiding any possible
degradation from the equipment farther down the line.
The recording system must be capable of
frequency response from a few 10's of Hz to at least 10
kHz. Fortunately, most computer sound cards fit the bill
very nicely!
The playback system must be capable of
frequency response from a few 10's of Hz to at least 10
kHz. Again, most computer sound cards work well for
this.
The modulator must be a "flat" FM with no
pre-emphasis, filtering or equalization of any kind. It
must be capable of flat frequency response from a few 10's of
Hz to about 10 kHz. Examples of of suitable modulators
include:
Many service monitors have suitable
"External Modulation" inputs. Just make sure that this
equipment is configured for "flat" FM, without
any filtering or equalization (such as CCITT) or
pre-emphasis.
It has also been reported that some
radios capable of 9600 baud packet operation (such as the
Yaesu FT-817) may also be modulated with a D-Star baseband
with excellent results.
Again, if your test set has both "FM" and
"PM" modes, make sure that you use "FM".
Many computers (especially laptops!) DO
NOT have very accurate sampling rates on their
audio inputs! If you are always going to
use the same computer for record and playback, it is likely
(but not guaranteed!) that the record and playback sampling
rates will be the same. What this means, however, is
that if you make a recording and play it back on a portable
audio device or another computer, the sample rate (and, the
D-Star data rate) may not be close enough for reliable
operation! It is recommended that one records a tone of
known frequency (such as the standard tones transmitted by
WWV/WWVH as received in AM mode) and use a program such as Spectran
(findable via a Google search) to determine the frequency that
the computer thinks that it is. If the
frequency is in error by more than a couple of percent, it
may not be suitable for the task!
For our initial test we simply connected the
DEMOD output of a service monitor (a Schlumberger 4031, for the
majority of our tests) tuned to the transmit frequency of the
D-Star transmitter (an IC-91AD) to the Line Input of a
laptop computer of known accurate sampling
rate. Using a program such as Audacity,
we then recorded the audio from the D-Star transmission to a .WAV
file. We made sure to start the recording just before
the transmitter was keyed up and to stop the recording after
the transmitter was unkeyed to be sure to capture the "key" and
"unkey" portions of the D-Star transmission.
For playback, we simply connected to the Line Output of
the sound card to the external modulation input of the service
monitor. We then played back the D-Star waveform, adjusting
the
deviation to +/-1.2 kHz as described above.
The use of different sample rates and encoding of baseband
D-Star audio files:
For our initial recording, we set the sound card to a sample rate
of 44.1 kHz with 16 bit audio to generate an uncompressed .WAV
file. In later tests, we found that a sample rate of 22.05
kHz at 16 bits was also adequate with only a very slight (and
probably insignificant) degradation in the baseband waveform.
We also experimented with resampling of the
44.1kHz/16 bit waveform down to an 8 kHz/8bit waveform using an
audio editing program and found that, although the baseband
waveform became slightly "ringy" owing to a slight amount of
aliasing, there was little degradation in the ability of the
D-Star receiver to decode the signal under poor conditions.
Note that recording and then down-sampling to 8 kHz/8bit is likely
to yield better results than recording at 8kHz/8 bits owing to the
fact that the software resampling is likely to be of higher
quality than "capturing" a signal live at 8 kHz and relying on the
sound card's hardware and drivers to do the appropriate filtering
"on the fly."
Comments
about
sampling
rate
errors.
This
is important enough that we are mentioning it again:
The user should be aware that many sound
cards (especially USB-types) can have sample rates that may
differ from the nominal rate: Even the record and
playback rates may be different! These differences can
result in slightly different bit rate of the played-back
"canned" D-Star audio, causing receive bit errors. It is unknown how much bit-rate
difference the D-Star decoders can typically tolerate.
Even built-in sound cards have been noted
to have significant frequency errors due to inaccurate
sampling rates.
For example, it was noted that one
particular Dell (tm) laptop had a sampling rate that was
about 8% higher than nominal: This much error is too
great for reliable testing if this file is "shared" with
other users, or if it is used to play back a file recorded
on a computer that had accurate sample rates.
Note that if only the same
computer is used to record and play back "canned" test
files, there may be no problem - until those files are
played on other computers that do have accurate
sample rates!
Some operating systems (such as XP and
Vista) can cause slight sampling-rate errors, especially at
rates other that
48kHz. This is usually due to the fact that the hardware
is always operating at just one sample rate (such as 48 kHz)
and is converted to other rates, on-the-fly, as needed - but
this conversion may not be very precise. This, too, can
cause errors due to slight differences in the bit rate.
In some cases, by choosing the "base" rate of the sound card -
such as 48 kHz - one can get accurate results, but the trick
is that one may have no way of knowing what the sound card's
internal sample rate really is!
As mentioned above, it is best to record a
tone of known frequency (such as that from WWV or WWVH) and
use a program such as Spectran to determine the sample rate
error.
D-Star MP3 files:
Later, we took the 44.1 kHz 16 bit audio file and used WinLame - a
freeware program - to encode the original .WAV file to MP3.
Through experimentation, we observed that recoding this .WAV file
to 128kbit/second mono (with 44.1kHz sampling)
produced a fairly good replica of the original D-Star baseband
waveform and rates of lower than 64kbps (in mono) produced usable
(although somewhat degraded) results. If stereo coding is used, a
bitrate of 192 kbps or higher is recommended.
Note: Most MP3 encoding utilities do not
offer the users specific options for encoding, such as the
selection of sample rate and whether the result should be a stereo
or mono .MP3 file. If this is the case, simply select the
"highest" quality mode available until the quality of the playback
waveform can be closely analyzed.
We then loaded the .MP3 files into a number of different portable
audio players - some of them fairly expensive, and one of them extremely
cheap (e.g. <$20) and we found that they all worked fine - as
long
as
the equalization was disabled and any phase "inversion" (see
below) was accommodated!
Playing back "canned" D-Star baseband recordings:
When doing a playback of a "canned" D-Star recording, there are a
number of things to remember:
The "keyup" and "unkey" portions of the
transmissions should be preserved to allow the D-Star
receiver's codec to gracefully detect the beginning and ending
of the transmissions. This means that the recording
should be started before the transmitter
being recorded is keyed up and stopped after
the transmitter is unkeyed. In some cases, such as
long-duration test tones, etc. where the speed of acquisition
is unimportant, one may not need to worry about this.
When playing back, be absolutely
certain to disable all audio
equalization and special effects! You may have to delve
into the "mixer panel" or the sound card's own control program
to disable these features.
Many sound cards have treble, bass,
equalization, reverb, echo, and/or "3D" effects - all
of which should be set to zero or disabled
before playback as any one of these can wreck the D-Star
waveform!
Like sound cards, many portable audio
players also
have settings for equalization, and some may even have some
other fancy audio effects - all of which should be
disabled or set to a "flat" response.
You may need to do an audio phase inversion
in playback to be able to decode the D-Star waveform.
If you experiment with multiple audio
playback devices (e.g. different computers, portable audio
players, etc.) or different service monitors, you should
remember that each of these may or may not require a phase
inversion. Because most audio players do not
have a way to flip the audio phase, you must take this into
account! Remember: If you use a compressed audio
format, you may have to flip the phase before
compressing it from the original .WAV file as the MP3
conversion itself *may* cause its own phase inversion.
Most audio editing software packages
(like Audacity) can be used to "flip" the phase using the
"invert" function.
If you are using a battery-powered audio
player, you may be able to get away with reversing the audio
leads, although this can lead to pick up of noise and RF as
shielding is compromised when you do this. Note:
A good quality 1:1 audio isolation transformer,
properly terminated, may circumvent this.
You may build a simple transistor or
op-amp circuit that will allow you to reverse the phase with
the flip of a switch.
Any data input or callsigns programmed into
the transmitter at the time of the recording will be
maintained through the recording. If you plan to use the
"canned" D-Star recording as a test signal - especially when
going through a repeater or network, make sure that your
callsign and other configurations are set appropriately!
If this original recording is made using a
digital data mode, it may be possible to generate a
rudimentary BER test system by transmitting this "canned" file
through the system and analyzing the results.
What might be put on the "canned"
recordings?
Some obvious examples are:
Plain speech announcing the test.
This is a good test to see how things sound and to determine
if decoding errors are occurring.
Standard tones. Using the procedures
in the Icom service manuals, one can feed a standard tone into
the microphone connector at a known level. This is
particularly useful in a system where D-Star audio may be
converted to analog, as might be the case on an
D-Star<>Analog gateway. Note:
Under certain conditions, the D-Star audio codec may produce
unexpected results with a constant tone!
If the transmitter was in data mode, known
data may be transmitted for the purpose of analysis to
determine the magnitude of data corruption or loss.
What can you do with the "canned"
recordings?
Using a standard piece of analog test gear and a portable audio
recorder, it is possible to generate a standard D-Star test signal
to test the performance of a D-Star system in much the same way as
one can test an analog radio system. This can include tests
such as:
Receiver sensitivity. One can see if
the receiver is working as well as it should! It is also
possible to remotely check a repeater using test gear from a
remote location to see if it is performing as it has in the
past.
Desense. This is particularly
important in a repeater system to determine if its own
transmitter (or another transmitter) is reducing receive
system sensitivity.
Interference. One can verify the
performance of a system to see if other signals (adjacent,
on-channel, or those resulting from intermodulation
distortion) may be causing a problem.
Audio level tests - particularly if there
is an interface to the "analog" world somewhere.
BER tests: If the "canned" recording
includes known data, this may be analyzed to determine the
error rate of the system, otherwise one may simply listen to a
known-good recording and listen for errors.
Analyzing received transmissions with analog test
gear:
Unfortunately, receiving a D-Star signal and decoding it
back to audio with test gear is not so easy. Again, it may
be possible to interface an Icom D-Star module or a so-called
"D-Star Dongle" to a service monitor or test set - provided that a
means of generating a GMSK baseband signal is provided - but these
alternatives will likely require homebrewing and/or the necessity
to lug a laptop around.
In many cases, your D-Star radio may be able to serve as a piece
of test gear: With it, you can monitor the transmission to
see if it seems to decode properly, and you may even be able to
run rudimentary BER tests using the data mode.
One of the ways that an analog test set may be useful is to
demodulate the receive signal and analyze the baseband waveform
with the monitor scope. With it, one can see if the baseband
waveform appears to be correct and if the "eye" pattern looks
clean.
(More about analyzing the pattern on the monitor's scope will
be added later.)
Two simple tests that can be performed are:
Frequency measurement. A standard
test set should be able to accurately read the transmitted
frequency.
Deviation. As mentioned above, the
deviation meter - and especially the modulation scope - can be
used to see if the D-Star signal is being modulated as it
should.
Some caveats:
If an off-air signal is being monitored,
remember that the test set likely has a much wider bandwidth
than a typical D-Star receiver. This means that the
received signal may be being degraded by adjacent-channel
interference that would not bother a D-Star
receiver.
Many test sets and service monitors are not
particularly sensitive when used as receivers for off-air
signals. If this is the case - or even if there is
interference - additional noise may "fuzz up" the received
signal, making precise measurements difficult.
It has been reported that some radios, such
as the FT-817, can be used with reasonable results by taking
audio from the 9600 baud output and switching to "FM-Narrow"
mode.
BER Testing:
One aspect of the D-Star technology is that even though it is a
digital system, there are very few tools available to the D-Star
system designer that are available to the designer of almost any
other digital wireless system. It is somewhat alarming that
even the most basic of these tools - a Bit Error Rate (BER)
indication is sadly lacking: This is somewhat surprising, as
even the most inexpensive digital wireless devices - such as 802.x
wireless cards, most cell phones, and satellite receivers, just to
name a few, have available (albeit sometimes obscured) indicators
of bit error rate.
At the time of this writing Icom has published very little
pertaining to D-Star system design and provided minuscule
resources in the form of tools to allow the design, analysis,
maintenance and diagnosis of problems in their D-Star
products: It would have taken very little extra effort on
their part to provide even rudimentary tools to the D-Star user
and system designer!
Fortunately, all is not lost: It is possible that the
intrepid homebrewer can devise a means to glean this information
from the innards of their D-Star radio - provided that they aren't
afraid to do a bit of hacking.
The GMSK modem used in the ID-1 and IC-91AD and its "BER"
indicator:
The ID-1 uses a CMX589A GMSK modem chip for recovering data from
the GMSK baseband signal from the MC3356 demodulator in both the
low-speed (DV) and high-speed (DD) modes: The ID-1 uses a
separate modulator to generate I/Q signals for transmit, leaving
half of this chip unused. The ID-91AD, on the other hand,
uses this chip for both reception of and generating the GMSK
baseband waveforms.
The CMX589A is an integrated receiver/transmitter that is designed
to receive and generate GMSK baseband waveforms. (A data
sheet for this chip may be found here.)
Interestingly,
this
chip
has
an
"RX S/N" pin that outputs a signal that can be used to
approximately estimate the signal-noise ratio of the received
signal but alas, this connection (pin 23) is left disconnected in
the ID-1 and IC-91AD: This is a pity, as the use of this pin
might have proven helpful in determining optimal signal quality
when setting up D-Star links - not to mention in everyday use by
the casual user!
How would one use this signal? In the simplest form, simply
"listening" to it with an audio amplifier gives a rough indication
of the signal quality as it would get "noisier" (and the average
voltage would get lower) with higher bit error rates as this pin
outputs a low pulse every time a received data bit transition
occurs outside the expected time window. The use of a few
components (resistors, capacitors, etc.) can also be used to
develop a voltage from the pulse train on this pin that would
provide a repeatable, consistent indication of the received signal
quality and this information could then be translated into a form
that the system designer/maintainer or even the casual could use.
Comment:The ID-1 uses this chip only for receive while the
IC-91AD uses it for both receive and transmit. It is worth
noting that BT (the ratio of the transmit filter's -3dB
bandwidth and the bit rate) is set for 0.5 in the IC-91AD's
modulator, a reasonable compromise between occupied bandwidth
and ISI.
BER output indication from the
AMBE 2020 codec:
In perusing the datasheet of the AMBE 2020, one of the audio
codecs that can be used in D-Star, one can see that the codec
produces a word in its output stream (word 7, to be precise) that
can be used to determine the BER. This is the same bitstream
that is used by the radio to determine the status of the codec -
among other things - but it would seem that this BER data is
simply thrown away by the radio rather than being made available
to the user. It may be possible that, in the future, Icom
may make this data visible in some way, but in the meantime, one
could, theoretically, "eavesdrop" on this bitstream with another
microcontroller and bring this data out of the radio in a usable
form.
There is good news, however, for the would-be programmers of the
DV dongle: It is there in the code, available for the
developer. (e.g. "BitErrors" in the structure "tOutFrame").
(The datasheet for the AMBE
codec used in the Icom radios and the nature of this BER
indication can be found on DVSI's website.)
Conclusion:
Even with "conventional" gear such as a service monitor or a
communications test set, it is possible to use it to assess and
troubleshoot a D-Star radio system with little extra equipment.
This page is a work in progress and is often updated.
Disclaimers:
The above procedures have been tested
using available test gear and Icom D-Star radios and are
believed to be valid. It is likely that this
information will, in the future, be updated and techniques
refined.
It is up to you, the reader, to
verify that this information is, in fact, correct and
suitable for your needs. We cannot be held responsible
for the use of the above information!
If you find that the above information
is incorrect or incomplete, please contact the frequency
coordinator using the link below.
Observations
of the codec used for D-Star - How does the codec
used for D-Star respond if subjected to sounds other than
those of the human voice? We decided to find out.
D-Star repeater
installation - Experiences, problems encountered
with the installation of a D-Star Stack at a very busy
radio site - and solutions to these problems.
The following are FAQ's provided by the
Utah VHF society. Note that these may topically overlap
the links above:
The above
list is, by no means, exhaustive: Other information may be
found via web searches.
This matter is open for
discussion: If you have concerns or opinions one way
or another, please make them known to the frequency
coordinator at the email address below.
Questions, updates, or comments pertaining
to this web page may be directed to the frequency coordinator.
