The article below was from our paper publication originally published in July 24, 1998. It
was originally posted on the website on 12/10/98 and since has received some minor editing.
From the RF/SS Workbench -- Vol. 4. Num. 1
In one of our latest projects we have made use of the Stanford Telecom SREL -2000A-45
IC DS "Modem." This versatile chip can do many things in many different applications.
It was designed for burst mode communications and as such has separately programmable
preamble and data spreading codes. It is capable of half or full duplex DS
communications. It offers a host of features and options -- this makes it rather
complicated to use! See the sidebar on this page for more information about this chip.
16 MHz QPSK Spectrum
We have called this chip the result of a "PhD dissertation" -- because you need
almost a PhD to understand how the chip works and to be able to program its many
registers. This chip is not designed for the faint of heart or casual SS system
designer. Excellent results can be had using this chip, however. Evidence the
5 MChip/sec QPSK spectrum plot shown in the figure above. This spectrum was captured
on our Wandel & Galtermann TSA-1 Tracking Generator / Spectrum Analyzer. The
QPSK modulator design is shown schematically below. The 16 MHz center frequency
output of this modulator provides an example of what can be done with pre-modulation
filtering in a DS QPSK modulator. The STEL chip was programmed for an 11 bit
long Barker "PN" code (same code for data and preamble) which gave a baseband
data rate of 909.09 kBps. The DS QPSK symbol rate was 5 Mbits per second for these
tests. Thus a processing gain of 10.41 dB was achieved with this waveform. Note
that the output spectrum is 10 MHz wide, fist null first null and approximately
8 MHz wide at the -20 dB bandwidth. The 3 dB bandwidth of this signal is approximately
4 Mhz. Note that the first sidelobe peaks are approximately -40 dBc from the peak of
Many of you may think that a QPSK signal filtered this much may produce a lot of Bit
Errors or produce an unacceptably high implementation loss. However, in our tests
with a coherent QPSK demodulator (to be described in our next issue) we found the
implementation losses of this amount of filtering when used with a companion
STEL-2000A-45 IC to be less than 3 dB! Better results could certainly be attained
with FIR digital pre-modulation filtering or using non-linear filtering a la FQPSK.
For more information on FQPSK, see the June 1995 issue of RF Design magazine,
pages 26-34, "DSP Implementation of GFSK, GMSK and FQSK Modulated Wireless Systems."
But for a simple, "quick and dirty" QPSK Modem, this scheme provides very
acceptable performance with near minimum RF bandwidth. In fact, this overall modulation
and filtering scheme achives an RF to data rate bandwidth ratio of just 1.6 --
considering the -20 dB (or 99%) bandwidth.
16 MHz QPSK Modulator Schematic [click on image for larger picture]
The RF (or IF) components of the modulator follow a very straightforward design.
Individual I and Q 5th order Bessel passive low pass filters (with 2.75 MHz cutoff
frequency for BT=.55) drive standard Mini-Circuits discrete DBMs, Quadrature Hybrids
and Power Combiners. The tight low pass filter bandwidth (low BT product)of the pre-modulation
filters does cause some transmitted signal distortion, but this filtering is responsible
for the nice output spectrum shown on the cover spectrum plot. The Bessel filter response
was chosen over Butterworth to better approximate flat group delay in the filters passband.
Remember that no IF bandpass filtering was used with this modulator. MiniCircuits and
several other vendors offer packaged QPSK modulators in a single device -- but beware,
some of these designs will not achieve the RF bandwidth or flatness that this discrete
design is capable of. After all, this design provides a spectrum, at -60 dBc, of 16 Mhz
bandwidth (equal to center frequency or 100% bandwidth)
When IF loopback tests were done with this modulator looped back to the outboard
QPSK demodulator, both transmitted and received I and Q eye patterns showed approximately
10% Intersymbol Interference (ISI). This level of ISI causes some bit errors at low
Eb/No ratios and the differential encoding used with the STEL causes additional double
bit errors. Thus, this design is certainly not an "optimum" modulator or demodulator in
any sense. This modulator is an easily duplicated design, provided you can program and
control the STEL chip correctly.
Hints & Kinks:
Believe it or not, a QPSK modulator can just be treated like two independent BPSK modulators
or doubly balanced mixers (DBMs), for all practical purposes. Thus interfacing a QPSK
modulator is really just like interfacing two DBMs -- double pre-modulation filters are
needed AND you still need to impedance match / limit the current to each DBM IF port.
Have fun with QPSK!
Questions, comments or suggestions on anything covered in this column are
welcome. Please feel free to share any info or special knowledge you may have. Just drop us an email: