dipl/sections/DP/dac/main.tex

\subsection{Audio Digital-Analog-Converter}

A digital to analog converter takes a digital number and converts it to a
analog signal. The output of one such conversion is called a sample. With
enough samples per second various different waveforms can be produced, which,
when amplified and put onto a speaker, can be heared by the human ear as a tone.
With various tones in series a melody can be produced, which is what the DAC in
this implementation does.

\subsubsection{TLC 7528 Dual R2R Ladder DAC}

The TLC 7528 is a Dual output parallel input R2R Ladder DAC with a maximum
sample rate of 10MHz \cite{tlc7528}, and which (should be) is monotonic over the
entire D/A Conversion Range. The TLC-7528 was the only component chosen, where
availability was not a factor, but rather it's design. It is the cheapest
dual R2R Ladder dac which takes \textbf{PARALLEL} input, which is an important
feature, because the backbone of the project is its parallel bus. Further the
DAC was developed for audio aplications\cite{tlc7528} which made its use obvious
and the TLC-7528 was the only IC available as DIP
\footnote{DIP... Dual Inline Package}, of which the pinout can be seen in figure
\ref{fig:tlc7528_pinout}

\begin{figure}[H]
	\centering
	\includesvg[height=.3\textheight, angle=0]{pics/slas062e_pinout.svg}
	\caption{TLC-7528 Pinout\cite{tlc7528}}
	\label{fig:tlc7528_pinout}
\end{figure}

\subsubsection{IDT7201 CMOS FIFO Buffer}

The IDT7201 is an asychronous CMOS FIFO, which means that it can be read with
a completely independant speed from which it is written and vice versa. It has
9 bit words, which can be seen in figure \ref{fig:idt7201_pinout}, and can 
store up to 256 words\cite{idt7201}. It is used as a buffer
to store data describing the targeted waveform in order to free time on the
parallel bus for interaction with the 16550 UART. 

\begin{figure}[H]
	\centering
	\includesvg[height=.3\textheight, angle=0]{pics/idt7201_pinout.svg}
	\caption{IDT-7201 Pinout\cite{idt7201}}
	\label{fig:idt7201_pinout}
\end{figure}

\subsubsection{Theory verfication}

Before tests of the complete unit were conducted, the functionality of the 
device and the validity of the knowledge of operations were performed. For that
the DAC was directly connected to the ATMega without the FIFO in front of it.
A saw was generated on only the DACA channel, which was put into voltage mode
as described in the datasheet\cite{tlc7528} and seen in figure 
\ref{fig:tlc7528_volt}. 
After the result seen in figure \ref{fig:tlc7528_saw_nonlin}
was measured, a lot of effort was put in to determine the source of the heavy
noise, however no obvious conclusions can be made, execpt that it comes from the
DAC itself and is consistant over whatever frequency used. A damaged IC could be
the reason or a sloppy production progress. Filters can be used to reduce the
noise, however this was not done in this thesis, as the generated audio does
not seem to suffer from these non-linearities as badly as when measured 
standalone.

\begin{figure}[H]
	\centering
	\includesvg[height=.3\textheight, angle=0]{pics/slas062e_volt.svg}
	\caption{TLC-7528 in voltage modet\cite{tlc7528}}
	\label{fig:tlc7528_volt}
\end{figure}

\begin{figure}[H]
\begin{tikzpicture}
	\begin{axis}[
		ylabel=REFA Voltage,
		xlabel=Time,
		grid=both,
		minor tick num=5,
		xmin=-2.030599999999999e-05,
		xmax=0.000598364,
		width=\textwidth,
		height=0.4\textheight]

		\addplot table [x=t, y=c1, col sep=comma, mark=none] {meas/20200210_saw_nonlin.csv};
	\end{axis}
	\end{tikzpicture}
	\caption{Measurement of a generated SAW signal via the TLC7528}
	\label{fig:tlc7528_saw_nonlin}

\end{figure}

\subsubsection{Schematics}
Based on the descriptions in the datasheets the schematic in figure
\ref{fig:schem_dac} was developed.

\begin{figure}[H]
	\centering
	\includegraphics[height=.65\textheight, angle=-90]{schem_pdf/dac.pdf}
	\caption{The schematic of the DAC Module}
	\label{fig:schem_dac}
\end{figure}

\paragraph{Element Description}

Diodes D1 through D4 are used as OR-Gates in conjunction with R1 and R2 to
generate the $\lnot MODRD$ and $\lnot MODWR$ signals for the D Flip-Flop
\footnote{74HC374\cite{74hc374}} and FIFO respectively, by these formulas: 

$\lnot MODRD = \lnot RD \lor \lnot MS2$

$\lnot MODWR = \lnot WR \lor \lnot MS2$

On a read access the output enable of the D-Latch becomes low, which writes
the status bits of the FIFO onto the data bus. C1, C2 and C3 are for stability
reasons and are good practice similar to the UART module. 74HC00 is a quad 
NAND-Gate\cite{74hc00} which is only used for inversion, chosen, like the
74HC374, for availability reasons. The A part of the NAND-Gate inverts the $MR$
signal from the bus to a $\lnot MR$ signal, as the FIFOs reset is low active.
The B part of the NAND-Gate inverts the FIFO Empty flag. It's output is
connected to the $\lnot WR$ input of the DAC, which means that the DAC doesn't
convert the input anymore, if the FIFO Empty flag is set to low.

The NE555 generates the audio clock signal, which should be the double of
44.1kHz\footnote{Because we have 2 output channels} as 44.1kHz is the standard
samling rate of CD-Audio\cite{iec60908}. Resistors R9 and R10 togehter with C7
form the Oscillator part of the NE55. C4 is for stability reasons and doesn't
define the frequency of the oscillator.

The generated clock is used for the $\lnot RD$ of the FIFO and inverted on the
DAC, which makes the data available on the output before being stored into the
DAC as it receives the signal to store the data after the FIFO makes it
available on the bus.

The DAC is operated in voltage mode as described in \ref{fig:tlc7528_volt},
with it's voltage source beeing available at either $3.472V_{pp}$ for 
professional
audio or $0.894V_{pp}$ for consumer audio, as defined per convention.\cite{audiob}
The voltage source can be controlled via Jumper JP1.

C5 and C6 together with the load resistance on the audio jack form a high pass
with a cutoff frequency of 

$f_C = \frac{1}{2\pi R C} = \frac{1}{2\times \pi\times 10K\Omega\times  100\mu F} = 0.159154943Hz$ 

which should cover the hearable spectrum. The high pass was needed to generate
a positive and negative half of the wave form, as the DC-Offset with a frequency
of 0Hz is orders of magnitudes lower than the $f_C$ of the highpass gets 
filtered away.

R7 and R8 have been installed in order to unload the capacitors after device
poweroff.

\paragraph{NE55 Clock Source}

Though used as a clock source, the NE555 is a bad clock source, if a stable
frequency is needed, because it varies widely with temperature, preasure and 
ageing
elements. A better solution would have been a quartz, which is divided down to
the desired frequency, which was what CD-Drives used to do, but more commonly in
modern CD Drives, an ASIC
\footnote{ASIC...Application-specific integrated circuit} 
with an internal PLL is used, thus the required quartz can no longer be sourced
via conventional electronic resellers.

\subsubsection{Demonstration Software}

\paragraph{SAW Generator}

To prove that read and write access from the D Flip-Flop and the FIFO are 
working,
the same saw signal has been generated as in figure \ref{fig:tlc7528_saw_nonlin}
, however the signal was put into the FIFO and not the DAC directly. The
resulting saw wave can be seen in figure \ref{fig:tlc7528_saw_fifo} together
with the FIFO Empty flag. The FIFO Empty flag, as explained before, is inverted
and starts/ends the complete D/A conversion, until further data is received.

\begin{figure}[H]
\begin{tikzpicture}
	\begin{axis}[
		ylabel=REFA Voltage,
		xlabel=Time,
		grid=both,
		minor tick num=5,
		xmin=0.02023862769230769,
		xmax=0.02642198769230769,
		width=\textwidth,
		height=0.4\textheight]

		\addplot table [x=t, y=c1, col sep=comma, mark=none] {meas/20200308fifo_44_1_saw_withfifoempty.csv};
		\addplot table [x=t, y=c2, col sep=comma, mark=none] {meas/20200308fifo_44_1_saw_withfifoempty.csv};
		\legend{REFA,$\lnot EF$}
	\end{axis}
	\end{tikzpicture}
	\caption{Measurement of a generated SAW signal with the FIFO Empty flag}
	\label{fig:tlc7528_saw_fifo}

\end{figure}

The time difference betwen a store and complete write cycle can be seen in figure
\ref{fig:fifo_dac_store}, while figure \ref{fig:fifo_dac} shows the 
transmission between dac and fifo in more detail.
\begin{figure}[H]
	\centering
	\includegraphics[width=\textwidth, angle=0]{meas/20200308fifo_44_1_cnt.png}
	\caption{A transmission between the FIFO and the DAC}
	\label{fig:fifo_dac}
\end{figure}

\begin{figure}[H]
	\centering
	\includegraphics[width=\textwidth, angle=0]{meas/20200308fifo_44_1_saw.png}
	\caption{A fifo store operation in contrast to the load operation}
	\label{fig:fifo_dac_store}
\end{figure}

The initialisation routines and read/write operations for the DAC module are
basically the same as for the UART module, and have thus been ommitted. They
can be seen in listing \ref{lst:16550-transmit}.

\lstinputlisting[language=C,frame=trBL,
	breaklines=true, breakautoindent=true, formfeed=\newpage,
	label={lst:dac_saw}, caption={SAW Generation for the DAC with FIFO},
	style=cstyle, firstline=134, lastline=144]
		{code/dac/saw_fifo_backplane/src/main.c}


\paragraph{Sine Generator}

As a further example a sine was generated and played on the DAC. The ATMega
itself is not powerful enough to generate the sine on the fly, therefore a
lookup-table had to be generated, which can be seen in listing
\ref{lst:dac_sine_lut}. How the data is transmitted to the FIFO can be seen
in listing \ref{lst:dac_sine} and figure \ref{fig:fifo_sine_store}, and the
resulting sine on both output channels can be seen in figure
\ref{fig:sine_dacab}.

\lstinputlisting[language=C,frame=trBL,
	breaklines=true, breakautoindent=true, formfeed=\newpage,
	label={lst:dac_sine_lut}, caption={Sine LUT Generation},
	style=cstyle, firstline=118, lastline=123]
		{code/dac/sine_fifo_backplane/src/main.c}

The look-up table has a size of 256, which is the maximum value an 8 bit integer
can take. This size was chosen to make operation faster as it only takes
one cycle to load an array value into a register and another one to store it
into the GPIO register. The sine table in further examples was pre-genrated on
the compiling host to reduce startup time. The mothod shown in listing 
\ref{lst:dac_sine_lut} is not fast due to the lack of a floating point unit
on the AVR. \cite{atmega2560}

\lstinputlisting[language=C,frame=trBL,
	breaklines=true, breakautoindent=true, formfeed=\newpage,
	label={lst:dac_sine}, caption={DAC Sine Generation},
	style=cstyle, firstline=141, lastline=152]
		{code/dac/sine_fifo_backplane/src/main.c}

\begin{figure}[H]
	\centering
	\includegraphics[width=\textwidth, angle=0]{meas/20200310sine_dac.png}
	\caption{Storage and retrieval of a sine to and from the FIFO}
	\label{fig:fifo_sine_store}
\end{figure}

\begin{figure}[H]
\begin{tikzpicture}
	\begin{axis}[
		ylabel=Channel Voltage,
		xlabel=Time,
		grid=both,
		minor tick num=5,
		xmin=-0.001746707777777778,
		xmax=0.001774992222222222,
		width=\textwidth,
		height=0.4\textheight]

		\addplot table [x=t, y=c1, col sep=comma, mark=none] {meas/20200310sine_dac_osz.csv};
		\addplot table [x=t, y=c2, col sep=comma, mark=none] {meas/20200310sine_dac_osz.csv};
		\legend{DACA,DACB}
	\end{axis}
	\end{tikzpicture}
	\caption{Measuremet of the generated sine from the sine LUT on DACA and DACB}
	\label{fig:sine_dacab}

\end{figure}

\section{Addressing DACA and DACB}

The DAC used has 2 output channels which can be selected by the 
$\lnot DACA/DACB$ pin as seen in figure \ref{fig:tlc7528_pinout}. This pin was
mapped to bit 0 of the address bus in order to make use of it. Bit 8 on the fifo
was used to store the bit. It was not implemented with half the bus clock to
make both channels independent of each other. This however uses more time on the
backend because it means the fifo is used up at twice the speed. No current
example makes use of this, but it may be used in future examples and
implementations on this unit.

On the audio jack DACA is mapped to the right channel and DACB to the left
channel.