exercise 10.18
ZIP with MATLAB scripts and note:
exercise 10.18 note:
k_boltz=1.380649e-23; % [J/K] Boltzmann constant
T0=290 % [Kelvin]
NF_dB=6
NF=10^(NF_dB/10)
OIP1dB_dBm=21
G_dB=30
G=10^(G_dB/10)
OIP3_dBm=33
Nin_dBm=-105 % input noise power
Nin=10^((Nin_dBm-30)/10)
SNR_target_dB=8
B=2e7; % [Hz]
%%
Nin=10^((Nin_dBm-30)/10) % [W] input noise
Te=T0*(NF-1)
Nout=G*(Nin+k_boltz*Te*B)
Nout_dBm=10*log10(Nout)+30
%% LDR
LDR_dB=OIP1dB_dBm-Nout_dBm
%% SFDR
SFDR_dB=2/3*(OIP3_dBm-Nout_dBm)-SNR_target_dB
T0 =
290
NF_dB = 6
NF = 3.981071705534972
OIP1dB_dBm = 21
G_dB = 30
G = 1000
OIP3_dBm =
33
Nin_dBm = -105
Nin = 3.162277660168380e-14
SNR_target_dB =
8
Nin =
3.162277660168380e-14
Te =
8.645107946051420e+02
Nout = 2.703399694138427e-10
Nout_dBm = -65.680897396782711
LDR_dB = 86.680897396782711
SFDR_dB = 57.787264931188474
Comment 1
The example Distortion Measurements comes with f0 range of kHz. I tried *1e3 on all input frequencies as well as on Fs and it still works, but to simulate f=[2.1 2.3]*1e9 again the default upper limit prevents the script from starting.
This is why advanced MATLAB users add a powerful GPU module and lift up the default upper limit:
-
GPU computing with MATLAB ,
-
MATLAB GPU Computing Support for NVIDIA CUDA-Enabled GPUs ,
-
GPU Enables Obsession with Fractals ,
-
and check how good is your GPU with gpuBench
Comment 2
What Distortion Measurements example calls SFDR is actually SINAD for the [0 25] MHz bandwidth defined:
The point of distortion and spurs regulation is to make sure that if receivers meet certain quality, no spurs below certain levels can cripple such receivers.
The spurs to worry about, from the receiver point of view are the ones that:
1.- throw up right on band needed (by receiver)
2.- are close enough in frequency to, again, throw
up on allocated frequency channel.
This example generates a spur close enough to the carrier to cause problem, but such close spur is well down in power level, so then the script hops up over 3 octaves to next interference and calculates SFDR with such spur.
As I have shown in solution to example 10.4 the closer that input frequencies are, the closer in turn that output frequencies turn out, clustered around harmonics. Such closeness in frequency is what causes nearby spurs throwing up on the narrow frequency channel that the receiver needs clean enough for BER to be obtained below certain threshold, during all the time.
BER test procedure usually implies:
-
Plug 2 BER meters one each side of the wireless link, or if possible configure one side of the wireless link to loop back and use 1 BER meter only on the other side
-
Go do something else for something like the next 24 hours
-
Retrieve tester and analyse data
Mobile network operators usually contract other companies to deploy Radio Access, and although the operator measures BER, for instance compliance and acceptance of sites is done upon such measurements, not carried by the same operator, but by an outsourced contractor, so that if any problem with a regulator collecting spectrum problems from users or another operator, the non-compliance fine can be 'compartmentalized'.
Well, the place is full of such further away wireless carriers. Cellular network operators reuse frequencies all over their networks, even trading them with competing operators: here I let you use this frequency, there you let me use your mast and some room for my Indoor Unit (IDU).
Such 'further away' carriers are always present, but the intermodulation products do not 'throw up' on the frequency channel that the receiver needs clean to process inbound data. SINAD is defined for certain Bandwidth B. If SINAD is defined for signal Bandwidth within channel, the same way that signal power is calculated with signal within channel only, then for this case the signal Bandwidth wouldn't comprise neither of the spurs present.
Receivers may have input filters, but the better is a filter the more expensive it is too. Also, when filtering with coding, the CPU cannot allocate too much time to remove further away in frequency carriers, it has to focus processing time on the frequency band where signal is expected, along with securing a smooth user interface experience.
Actually, mobile phones with higher processing capacity implement soft handovers, listen to neighbouring base stations (those higher carriers further away in frequency), and perform faster pull up/down LO power level (GSM roughly twice per second, 3G approximately 1500/s).
Some 3G mobile terminals, and all 4G and 5G mobile terminals even download from additional neighbouring base stations (precisely the further away carriers, the so called 'fingers') to increase data rates, offer higher QoS and enable applications that require PC-like capacities.
For £1200/piece of smart phone, they'd better outperform PCs or they would end up back on sales shelves. They don't go back to store, so they do outperform average PCs, and are mobile. And the traffic demand keeps increasing, marketing saying data traffic demand upper boundary is not O(n) but at least O(n^2), don't believe the O(e^n) sales team exaggeration.