bez diskuze a zaveru

1 files changed, 185 insertions, 7 deletions

diff --git a/text/labina.tex b/text/labina.tex
index 30514b2..4664183 100644
--- a/text/labina.tex
+++ b/text/labina.tex
@@ -103,6 +103,7 @@ PDP context info 1:
 
 Another two commands are \texttt{link sysmode} for switching between NB-IoT and LTE-M and \texttt{link funmode} for turning the modem to and from flight-mode for save antenna and attenuators reconfiguration.
 
+
 \section{Power consumption - ampere meter}
 
 Power consumption can be accurately measured with negligible influenced by introduction of additional measurement equipment so measurement with external meter was chosen over the modem estimates or table values. For the power consumption evaluation of \ac{IoT} devices there are two different factors to consider. First is standby or sleep power that is usually long (orders of minutes or hours) and very small (orders of miliamps) and second are short rapid transmission spikes in orders of amperes. This requires ampere meter with very high sampling speed and very high dynamic range.
@@ -117,7 +118,6 @@ For measuring of power consumption was done using Nordic \ac{PPK II}, seen in pi
 	\label{fig:ppk}
 \end{figure}
 
-
 Its key features are \cite{ppkdatasheet}:
 \begin{itemize}
     \item 200nA to 1A current measurement range with a resolution that varies between 100nA and 1mA
@@ -127,16 +127,25 @@ Its key features are \cite{ppkdatasheet}:
     \item Supported through nRF Connect for Desktop's Power Profiler app
 \end{itemize}
 
-The measurement setup of nRF9161 DK with PPK II in EMC chamber and addition attenuators can be seen in figure \ref{fig:ppkmeas}. These development kits from Nordic enable a direct connection of serial ampere meter using special pins. This is them main reason why both boards were chosen for this measurement. 
+The measurement setup of nRF9161 DK with PPK II in EMC chamber and addition attenuators can be seen in figure \ref{fig:ppkmeas}. Antennas connected to the other end of coaxial cable can be seen in figure \ref{fig:emcant}. These development kits from Nordic enable a direct connection of serial ampere meter using special pins. This is them main reason why both boards were chosen for this measurement. 
 
 \begin{figure}[H]
 	\begin{center}
-		\includegraphics[width=0.6\textwidth]{obrazky/nrfppkemc.jpg}
+		\includegraphics[width=0.8\textwidth]{obrazky/nrfppkemc.jpg}
 	\end{center}
 	\caption[Nordic PPK II measuring setup]{Nordic Power Profiler Kit II measuring setup with nRF9161 in EMC chamber\cite{ppkdatasheet}}
 	\label{fig:ppkmeas}
 \end{figure}
 
+\begin{figure}[H]
+	\begin{center}
+		\includegraphics[width=0.6\textwidth]{obrazky/31_anthenna3.JPG}
+	\end{center}
+	\caption[Custom LTE antenna for bands around 800 MHz]{Custom LTE antenna for bands around 800 MHz}
+	\label{fig:emcant}
+\end{figure}
+
+
 The output of PPK II can be seen in figure \ref{fig:ltemcurr} for LTE-M and and in figure \ref{fig:nbiotcurr}.
 
 Yellow vertical lines mark a start of ping set and red vertical it end. These boundaries were set due to noise manually.
@@ -160,15 +169,184 @@ Green horizontal line indicates average draw in idle state where the modem does
 
 Another value that is calculated from these data and cannot be depicted in the chart is the raw charge needed for transmission of one packet. This is calculated by integration of the current over time set by the boundaries. This value expressed in coulombs or more commonly in mAh.
 
-\section{Results}
+\section{Cellular Results}
+
+Data described in sections above are aggregated into data tables using Python scripts. Example of such table can be seen in table \ref{tab:ltem61_1024ex}. Complete data for both LTE modes and all payload sizes are, for their large extent, placed in Appendix \ref{appendix:ltetables}.
+
+\begin{table}[!h]
+    \begin{center}
+        \small
+\hspace*{-2cm}\begin{tabular}{rrrrrrrrrrrr}
+\hline
+KU & RSRP & SNR & RT min & RT max & RT avg & Transmit & Max I & Ready I & Standby I & Charge & DR \\
+{[dB]} & {[dBm]} & {[dB]} & {[ms]} & {[ms]} & {[ms]} & {[s]} & {[mA]} & {[mA]} & {[mA]} & {[uAh]} & {[kB/s]} \\\hline
+
+0 & -75 & 1 & 1190 & 4500 & 3146 & 3.57 & 233.39 & 30.90 & 2.38 & 37.16 & 0.30 \\
+3 & -77 & 2 & 1039 & 3731 & 1869 & 1.95 & 222.60 & 31.11 & 2.21 & 19.20 & 0.56 \\
+6 & -82 & -1 & 983 & 2201 & 1618 & 1.74 & 220.29 & 31.56 & 2.71 & 17.59 & 0.63 \\
+10 & -84 & 2 & 1089 & 3103 & 1634 & 1.79 & 217.21 & 30.82 & 2.35 & 18.38 & 0.61 \\
+16 & -89 & 2 & 994 & 2888 & 1379 & 1.53 & 217.21 & 31.03 & 2.32 & 17.08 & 0.71 \\
+20 & -94 & 1 & 1089 & 2618 & 1660 & 1.72 & 222.60 & 31.56 & 5.71 & 19.04 & 0.63 \\
+21 & -92 & 3 & 1037 & 1542 & 1254 & 1.48 & 222.60 & 30.60 & 0.60 & 16.80 & 0.74 \\
+26 & -98 & 4 & 976 & 1364 & 1162 & 1.32 & 222.60 & 30.52 & 2.36 & 15.51 & 0.82 \\
+30 & -101 & 4 & 961 & 1670 & 1244 & 1.40 & 225.68 & 31.01 & 0.82 & 16.44 & 0.78 \\
+36 & -109 & 3 & 1127 & 1562 & 1327 & 1.52 & 221.83 & 30.58 & 2.02 & 17.68 & 0.72 \\
+40 & -112 & 4 & 1315 & 2352 & 1737 & 1.94 & 224.91 & 28.93 & 2.38 & 27.68 & 0.56 \\
+46 & -119 & 3 & 1370 & 2040 & 1717 & 1.92 & 248.06 & 28.84 & 2.22 & 34.42 & 0.57 \\
+50 & -124 & 1 & 1734 & 2917 & 2228 & 2.44 & 224.91 & 31.18 & 2.22 & 55.45 & 0.45 \\
+51 & -124 & 2 & 1645 & 2501 & 2072 & 2.30 & 224.91 & 30.13 & 2.22 & 53.41 & 0.47 \\
+56 & -131 & -3 & 3393 & 5091 & 4307 & 4.54 & 226.45 & 31.08 & 2.42 & 115.42 & 0.24 \\
+60 & -133 & -5 & 11772 & 13685 & 12833 & 11.91 & 222.60 & 29.30 & 2.22 & 435.51 & 0.09 \\
+\hline
+\end{tabular}
+
+    \end{center}
+    \caption[NB-IoT nRF9161 measurements with payload size 1024 B]{NB-IoT nRF9161 measurements with payload size 1024 B}
+    \label{tab:ltem61_1024ex}
+\end{table}
+
+These final data tables are used for generating more visually understandable plots. Lets have a closer look at some of them that would evaluate the performance of nRF916x as outlined in beginning of this chapter.
+All charts can be found in appendix \ref{appendix:ltecharts}.
+
+\subsection{Validity of attenuation and sensitivity}
+
+Firstly to ensure validity of this measurement it is important to check if the addition attenuation really corresponds to attenuation of input signal. This can be seen in figure \ref{fig:rssitoatt}. The sudden jump at 20 dB is due to transition of the physical attenuator form the inside of the EMC chamber to outside to prevent RF leakage. Th same thing happened at 50 dB. Other than that the slope of the course seems linear.
+
+Another key observation is that NB-IoT has by 3 dB better sensitivity over LTE-M (-133 dBm and -130 dB respectively). This is in contrast to datasheet values of -114 dBm and -108 dBm (see chapter \ref{chap:nrfdesc}) where the difference between the modes should be 6 dB and measured sensitivity is better by 19 dB and 22 dB.
+
+Alternative value provided by modem that could be use to guess the sensitivity is its reported \ac{DL} path loss. This was 159 dB for NB-IoT and 148 for LTE-M. If the same transmit power of 23 dBm is assumed for the \ac{BTS} it would give a sensitivity of -136 dBm for NB-IoT and -125 dBm for LTE-M. This method or reported DL path loss in general are used sporadically in subsequent analysis for the author's doubts about validity and accuracy of such numbers.
+
+Seemingly big difference is \ac{RSRP} at same attenuation between both modes. This has multiple possible explanation:
+\begin{itemize}
+    \item Slightly different bands - LTE-M was measured on LTE band 8 (UL: 880 - 915 MHz DL: 925 - 960 MHz) while NB-IoT was on LTE band 20 (UL: 832 - 862 MHz DL: 791 - 821 MHz )
+    \item Different settings on actual \ac{BTS} for both modes
+    \item Higher robustness of NB-IoT
+\end{itemize}
+For this reason subsequent analysis will be done using \ac{RSRP} supplied by modem.
+
+From these possibilities the advantage of NB-IoT compared to LTE-M for detection of very weak signals is 3 dB if RSRP is to be belived, 6 db according to datasheet, 11 dB according to reported DL pash loss and 14 dB according to attenuation.
+
+\begin{figure}[H]
+    \begin{center}
+        \includegraphics[width=1\textwidth]{"mereni/komora2/both61ReferenceSignalReceivedPowertoattenuation.png"}
+    \end{center}
+    \caption[Reference Signal Received Power to attenuation]{Reference Signal Received Power to attenuation}
+    \label{fig:rssitoatt}
+\end{figure}
+
+\subsection{Power consumption analysis}
+
+The value crucial for calculation of needed battery capacity in \ac{IoT} devices is electrical charge that is consumed by the transmission, together by constant current of ready and idle states and planned lifetime of such device.
+
+In figure \ref{fig:chargenrf} can seen a comparison of LTE-M and NB-IoT in regards to this needed charge per single transmission of given payload size dependent to the signal strength reported using \ac{RSRP} from the LTE modem.
+
+From this figure can be conclude following:
+\begin{enumerate}
+    \item LTE-M uses consistently less energy for every payload size compared to NB-IoT. If the total attenuation is takes as determinative value even than NB-IoT consumes less energy then LTE-M only at the limit of LTE-M abilities. See appendix figure \ref{fig:both61Chargepermessagetoattenuation} for this chart.
+    \item Rapid increase in used energy for LTE-M happens only at signal levels close to connection loss. Gradual mild increase can be better seen in appendix table \ref{fig:ltem61ChargepermessagetoRSRP}
+    \item NB-IoT is able to transmit messages at the cost of vastly higher energy consumption in worse conditions where LTE-M is unable to operate.
+\end{enumerate}
+
+Summed up this means that usage of NB-IoT is only advisable for situations where signal conditions are so bad that the better resilience is needed.
+
+\begin{figure}[H]
+    \begin{center}
+        \includegraphics[width=1\textwidth]{"mereni/komora2/both61ChargepermessagetoRSRP.png"}
+    \end{center}
+    \caption[nRF9161 LTE-M and NB-IoT Charge per message to RSRP]{nRF9161 LTE-M and NB-IoT Charge per message to RSRP}
+    \label{fig:chargenrf}
+\end{figure}
+
+
+\subsection{Transmission speed and comparison to LoRa}
+
+Difficulty of direct comparison of LTE modes to other technologies such as LoRa are described in depth the discussion chapter \ref{chap:drysty}. But since LoRa is the main "competitor" of LTE IoT protocols it should at least be tried with data that are available.
+
+As a representative of LoRa the Semtech SX1262 chip was chosen. Primarily for similar band (around 800 MHz) and usage in similar applications\cite{loradatasheet}.
+
+Values of data rate and formula for sensitivity of SX1262 were obtained from its datasheet \cite{loradatasheet}.
+
+The formula is following:
+$$
+S = (12 - SF) * 2.5 - 137
+$$
+
+S is sensitivity in dBm, SF is LoRa speeding factor number, 2.5 is lowering of sensitivity in dB per SF and -137 is sensitivity of SX1262 for SF 12 in dBm.
+
+These data were integrated into chart with measured data for NB-IoT and LTE-M as you can see in figure \ref{complora}.
+
+\begin{figure}[H]
+    \begin{center}
+        \includegraphics[width=1\textwidth]{mereni/complora.png}
+    \end{center}
+    \caption[Comparison of sensitivity and transmit speed of LTE to LoRa]{Comparison of sensitivity and transmit speed of LTE to LoRa}
+    \label{fig:complora}
+\end{figure}
+
+Getting conclusions from comparing measured data and datasheet data without any assurance of measurement method comparability is hard but lets at least try to find some meaning in this.
+
+NB-IoT seems to be slightly faster and so better than LoRa up to SF 10 at RSRP -132 dBm where we set a limit of NB-IoT abilities. Since attributes of fair measurement compression were not met, from current data and from knowledge of their limitation there can be made a vague statement that abilities of NB-IoT and LoRa would seem comparable.
+
+LTE-M is a considerable improvement in aspect of data rate over NB-IoT. Over LoRa even though the difference seems huge, in case of real data they share on the signal strength axis only one point. From this it seems that these two are not very comparable in their target ideal use-cases.
+
+\section{DECT NR+}
+
+An additional measurement was made to test the capability of nRF9161 to transmit and receive using DECT NR+.
+
+Measurements were done similarly to these of LTE but with other nRF9161 DK in place of LTE antenna. Measurement of current wa done on the node.
+
+The fixed slot timing property of DECT NR+ means that device can transmit only in fixed time window. This gets rid of need for integration and power consumption can be established from maximal current and number of time slots used.
+
+The two DKs do not communicate \ac{P2P} strictly speaking but one operates as a beacon that periodically sends information needed to establish communication with itself. The other operates as node, in 5G terminology named as \ac{PDU}. In implementations with more than 2 devices some devices can be beacon with established connection to other beacon and by this manner it is possible to create the mesh trees described in chapter \ref{dectdescrition}.
+
+In the analysis there is used a term of \ac{FSPL} distance. This represents the attenuation as distance in completely free space that would have the same attenuation (ignoring latency increase). Here is how it was calculated. This is equation for freespace path loss:
+
+$$ K_U = \left( \frac{4 \pi d}{\lambda} \right) ^2  $$
+
+When frequency of 1.9 GHz is substituted, the equation solved for distance and signal loss expressed in decibels it forms this equation that was used in later analysis:
+
+$$ d = 10^{\frac{K_U-98}{20}} $$
 
+%In this case a attenuation is no longer arbitrary values with loose connection to real signal degradation but real 
 
+\begin{table}[!h]
+    \begin{center}
+        \small
+\begin{tabular}{rrrrrrrr}
+\hline
+& \multicolumn{2}{c}{Beacon} & \multicolumn{2}{c}{PDU} & & & \\
+$K_U$ & RSSI & PWR & RSSI & PWR & FSPL d & Max I  & Att conf \\
+{[dB]} & {[dBm]} & {[dBm]} & {[dBm]} & {[dBm]} & {[km]} &  {[mA]} & {[dB]} \\
+\hline
+106 & -105 & 4 & -91 & 23 & 2.51  & 213.28  & 60+46 \\
+103 & -103 & 4 & -88 & 23 & 1.78  & 218.63  & 60+43 \\
+93 & -93 & 4 & -78 & 23 & 0.56    & 220.93  & 60+33 \\
+83 & -83 & 4 & -68 & 23 & 0.18    & 221.69  & 40+43 \\
+73 & -72 & 4 & -60 & 16 & 0.06    & 160.71  & 40+33 \\
+63 & -62 & 4 & -59 & 7 & 0.02     & 97.30   & 40+23 \\
+53 & -52 & 4 & -60 & -4 & 0.01    & 89.05   & 40+13 \\
+43 & -41 & 4 & -58 & -12 & 0.00   & 88.30   & 40+03 \\
+\hline
+\end{tabular}
+    \end{center}
+    \caption[Result table of DECT NR+ measurements]{Result table of DECT NR+ measurements}
+    \label{tab:dect}
+\end{table}
 
-\input{mereni/ltem61_1024_table}
+Results of these measurement can be seen in table \ref{tab:dect} and visualized in figures \ref{fig:dectPowerlevelstotoattenuation} and \ref{fig:dectRSSIlevelstotoattenuation}.
+Note to the data: maybe contrary to common sense beacon TX power is output power transmitted from the beacon but its RSSI is value measured from PDU of the beacon's signal. Same thing is true for the PDU and its values.
 
-This is 
+\input{"mereni/komoradect/dectRSSIlevelstotoattenuation"}
+\input{"mereni/komoradect/dectPowerlevelstotoattenuation"}
+From these can be conclude following:
+\begin{enumerate}
+    \item Beacon did not had automatic adjunction of power but PDU did. This can be clearly seen in figure \ref{fig:dectPowerlevelstotoattenuation}. This could be due to using experimental \ac{MoSh} for DECT NR+ that supports only the physical layer of DECT NR+.
+    \item Maximal current grows in discrete jumps with PDU power level.
+    \item Maximal theoretical range for measured power of 4 dBm is 2.5 km.
+    \item If maximal power of 23 dBm on both boards is assumed, the maximal possible attenuation is 125 dB and theoretical range is 22.4 km.
+    \item Minimal signal strength was -105 dBm. According to nRF9161 documentation its sensitivity for DECT NR+ is -103 dBm. 
+\end{enumerate}
 
-Complete data are for their large extent placed in Appendix \ref{appendix:ltetables}.