Using MATLAB to display streaming serial data

Question

I have written an m-function to capture and plot a 50Hz serial data stream. This works as expected, bar the fact that there is a large (approximately 10 second) delay before any changes in the data stream are reflected in the plot.

The m-function uses the serial data stream design pattern, and the plot is configured to resemble a strip chart. The MATLAB recommendations for optimizing graphics performance have been followed, with YData being updated using get/set methods rather than data being replotted.

Now, either there is something wrong with my code, or I have reached the limit of what MATLAB is capable of doing without using the real-time windows target. Can anyone spot what I am doing wrong? I am considering implementing a USB interface on the embedded system that is streaming the data that MATLAB is plotting, but I am not convinced this will fix the problem.

function [] = tiny_stream(file)
% TINY_STREAM Plot and record a live pulse oximeter data stream.
%   [] = TINY_STREAM(file) plots a live pulse oximeter data stream, and also
%   records it to file. The data stream must contain ten columns of data.
expected_column_count = 10;

%% Declare variables.
% The sample rate is currently set to 50Hz by the pulse oximeter firmware, and
% only the last 10 cycles are plotted. Declaring plot data beforehand makes it
% easier to configure the plot.
global MAIN_LOOP;
MAIN_LOOP = true;

sample_rate = 50;
cycle_count = 10;

red_lowpass_data = NaN(sample_rate * cycle_count,1);

%% Create and configure plot.
% The 'plot' command is not used to update the plot, as it is far too slow to
% handle data streaming at 50Hz. Instead, the existing plot data is updated.
close all;
strip_chart = figure('Renderer','painters');
set(strip_chart,'DoubleBuffer','on');
set(strip_chart,'KeyPressFcn',@stop_stream);

time = 1:(sample_rate * cycle_count);

% Configure red chart.
red_subplot = subplot(2,1,1);
red_lowpass_trace = plot(time,red_lowpass_data,'b');

set(red_subplot,'YTickLabel',...
    {'-10,000','0','10,000','20,000','30,000','40,000'},...
    'YTick',[-10000 0 10000 20000 30000 40000],...
    'YGrid','on');
ylim(red_subplot,[-10000 40000]);
ylabel('count');

set(red_subplot, 'XTickLabel',[],...
    'XTick',1:cycle_count,...
    'XGrid','on');
xlim(red_subplot,[1 cycle_count]);
xlabel('1 sec / div');

set(red_subplot,'OuterPosition',[0 0.5 1 0.5]);
box(red_subplot,'on');
title('Red Data');

set(red_subplot,'ALimMode','manual',...
    'CameraPositionMode','manual',...
    'CameraTargetMode','manual',...
    'CameraUpVectorMode','manual',...
    'CLimMode','manual',...
    'TickDirMode','manual',...
    'XLimMode','manual','YLimMode','manual','ZLimMode','manual',...
    'XTickMode','manual','YTickMode','manual','ZTickMode','manual',...
    'XTickLabelMode','manual','YTickLabelMode','manual',...
    'ZTickLabelMode','manual');

drawnow;

%% Create and configure the serial port object.
serial_object = serial('COM2');

serial_object.BaudRate = 57600;
serial_object.DataBits = 8;
serial_object.FlowControl = 'none';
serial_object.StopBits = 1;

%% Configure the data stream.
% Note the use of the callback function 'transfer_data'. This is called by
% MATLAB whenever it detects the specified terminator in the serial object
% buffer.
serial_object.Terminator = 'CR/LF';
serial_object.InputBufferSize = 2^18;
serial_object.BytesAvailableFcnMode = 'terminator';
serial_object.BytesAvailableFcn = {@transfer_data};

serial_object.UserData.string_data = [];
serial_object.UserData.is_new = false;

%% Open the serial port.
if strcmp(serial_object.Status,'closed')
    fopen(serial_object);
end

%% Open the file.
data_file = fopen(file,'w');

%% Main program loop.
% There may be more than one row of source data in the serial input buffer at
% the start of the main program loop. Any of these rows may be incomplete, so
% the first thing the main program does is to check that the data contains the
% expected number of entries. If it does not, then the entire data chunk is
% discarded.
while MAIN_LOOP == true
    if serial_object.UserData.is_new == true
        chunk_string = serial_object.UserData.string_data;
        serial_object.UserData.is_new = false;

        chunk_numeric = sscanf(chunk_string,'%d');
        chunk_length = length(chunk_numeric);

        if mod(chunk_length, expected_column_count) == 0
            data_column_count = chunk_length / expected_column_count;

            data = reshape(chunk_numeric,expected_column_count,...
                data_column_count);

            fprintf(data_file,...
                '%6d %6d %6d %6d %6d %6d %6d %6d %6d %6d
',data);

            % Update red subplot.
            red_lowpass_data = get(red_lowpass_trace,'YData');

            red_lowpass_data(1,1:end - data_column_count) =...
                red_lowpass_data(1,data_column_count + 1:end);

            red_lowpass_data(1,end - data_column_count + 1:end) =...
                data(4,:);

            set(red_lowpass_trace,'YData',red_lowpass_data);

            drawnow;
        end
    end

    pause(0.001);
end

fclose(data_file);
fclose(serial_object);
delete(serial_object);
clear serial_object;

return

%% Loop control.
function [] = stop_stream(source, event)
% STOP_STREAM Stop the pulse oximeter serial stream.
%   STOP_STREAM(source, event) sets the MAIN_LOOP global variable to false
%   whenever a key is pressed while plot has focus.
global MAIN_LOOP;

MAIN_LOOP = false;

return

%% Data transfer.
function [] = transfer_data(object, event)
% TRANSFER_DATA Transfer data between buffers.
%   TRANSFER_DATA(object, event) transfers data between the serial object
%   input buffer and the user data area of the serial object.
string_data = fgets(object);

if object.UserData.is_new == false
    object.UserData.string_data = string_data;
    object.UserData.is_new = true;
else
    object.UserData.string_data = [object.UserData.string_data string_data];
end

return

RBE · Accepted Answer

I posed the original question some time ago, and neglected to post the answer when I found it - sorry about that. For those who are interested, it turns out that the 10 second delay was being caused by an error in the way that I was setting the XLim and Xtick properties of my axes. Replacing the following code fragment:

set(red_subplot, 'XTickLabel',[],...
    'XTick',1:cycle_count,...
    'XGrid','on');
xlim(red_subplot,[1 cycle_count]);
xlabel('1 sec / div');

With this one:

set(red_subplot,'XTickLabel',[],...
    'XTick',(0:sample_rate:(sample_rate * cycle_count))',...
    'XGrid','on');
xlim(red_subplot,[0 sample_rate * cycle_count]);
xlabel('1 sec / div');

eliminated the 10 second delay. As is usually this case, solving one problem exposed another, more fundamental one. It appears that MATLAB is simply incapable of recording a 57600 serial data stream without dropping samples. This behavior was verified with code similar to that posted above (ie GUI code), as well as a cut-down, command-line version with no GUI at all. The number of samples dropped varied from run-to-run, but sat around 7 per minute, ie 7 per 3000 samples. I then wrote some LabVIEW code to do the same thing, and it did not drop a single sample, no matter how many hours I left it running for. The lesson I draw from all of this is that MATLAB is great for analyzing data, but is not so good for acquiring it. That is a task best left to LabVIEW, or dSPACE.

Using MATLAB to display streaming serial data

Answers (2)

Related Questions