IMT2000DS Carrier-to-Noise Ratio

In the 2G (GSM) system the quality of the signal through the receiver processing chain
is determined primarily by the narrow bandwidth, that is, 200 kHz. This means that
the SNR of the recovered baseband signal is determined by the 200 kHz IF filter positioned
relatively early in the receive chain; little improvement in quality is available
after this filter. Consequently the noise performance resolution and accuracy of the
sampling ADC, which converts the CNR, must be sufficient to maintain this final quality
SNR. When the W-CDMA process is considered, a different situation is seen. The
sampled IF has a 5 MHz bandwidth and is very noisy—intentionally so. Because the CNR is poor, it does not require a high-resolution ADC at this point; large SNR
improvement through the processing gain comes after the ADC. A fundamental product
of the spreading/despreading process is the improvement in the CNR that can be
obtained prior to demodulation and base band processing—the processing gain.
In direct-sequence spread spectrum the randomized (digital) data to be transmitted
is multiplied together with a pseudorandom number (PN) binary sequence. The PN
code is at a much higher rate than the modulating data, and so the resultant occupied
bandwidth is defined by the PN code rate. The rate is referred to as the chip rate with
the PN symbols as chips. The resultant wideband signal is transmitted and hence
received by the spread spectrum receiver. The received wideband signal is multiplied
by the same PN sequence that was used in the transmitter to spread it.
For the process to recover the original pre-spread signal energy it is necessary that the
despreading multiplication be performed synchronously with the incoming signal. A
key advantage of this process is the way in which interfering signals are handled. Since
the despreading multiplication is synchronous with the transmitted signal, the modulation
energy is recovered. However, the despreading multiplication is not synchronous
with the interference, so spreads it out over the 5 MHz bandwidth. The result is that only
a small portion of the interference energy (noise) appears in the recovered bandwidth.
Processing or despreading gain is the ratio of chip rate to the data rate. That is, if a 32-
kbps data rate is spread with a chip rate of 3.84 Mcps, the processing gain is as follows:
The power of the processing gain can be seen by referring to the CNR required by
the demodulation process. An Eb/No of 10.5 dB is required to demodulate a QPSK signal
with a BER of 1 × 10-6. If a data rate of 960 kbps is transmitted with a chip cover of
3.84 Mcps, the processing gain is 6 dB. If a CNR of 10.5 dB is required at the demodulator
and an improvement of 6.0 dB can be realized, the receiver will achieve the
required performance with a CNR of just 4.5 dB in the RF/IF stages.
It must be considered at what point in the receiver chain this processing gain is
obtained. The wideband IF is digitized and the despreading performed as a digital function
after the ADC. Therefore, the ADC is working in a low-quality environment—4.5
dB CNR. The number of bits required to maintain compatibility with this signal is 6 or
even 4 bits. The process gain is applied to the total spread signal content of the channel.
If the ADC dynamic range is to be restricted to 4 or 6 bits, consideration must be
given to the incoming signal mean level dynamic range. Without some form of
received signal dynamic range control, a variation of over 100 dB is typical; this would
require at least an 18-bit ADC. To restrict the mean level variation within 4 or 6 bits, a
system of variable-gain IF amplification (VGA) is used, controlled by the Received Signal
Strength Indication (RSSI).
Prior to the change to 3.84 Mcps, the chip rate was at 4.096 Mcps, which when
applied to a filter with a roll-off factor ∝ of 1.22 gave a bandwidth of 5 MHz. Maintaining
the filter at 1.22 will give an improved adjacent channel performance. The
process gain is applied to the total spread signal content of the channel. For example, a
9.6-kbps speech signal is channel-coded up to a rate of 32 kbps. The process gain is
therefore 10 log(3.84/0.032) = 20.8 dB, not 10log(3.84/0.0096) = 26 dB, as may have
been anticipated (or hoped for).