rdmseed.m

function varargout = rdmseed(varargin)
%RDMSEED Read miniSEED format file.
%	X = RDMSEED(F) reads file F and returns a M-by-1 structure X containing
%	M blocks ("data records") of a miniSEED file with headers, blockettes, 
%	and data in dedicated fields, in particular, for each data block X(i):
%		         t: time vector (DATENUM format)
%		         d: data vector (double)
%		BLOCKETTES: existing blockettes (substructures)
%
%	Known blockettes are 100, 500, 1000, 1001 and 2000. Others will be
%	ignored with a warning message.
%
%	X = RDMSEED(F,ENCODINGFORMAT,WORDORDER,RECORDLENGTH), when file F does 
%	not include the Blockette 1000 (like Seismic Handler outputs), specifies:
%		- ENCODINGFORMAT: FDSN code (see below); default is 10 = Steim-1;
%		- WORDORDER: 1 = big-endian (default), 0 = little-endian;
%		- RECORDLENGTH: must be a power of 2, at least 256 (default is 4096).
%	If the file contains Blockette 1000 (which is mandatory in the SEED 
%	convention...), these 3 arguments are ignored except with 'force' option.
%
%	X = RDMSEED without input argument opens user interface to select the 
%	file from disk.
%
%	[X,I] = RDMSEED(...) returns a N-by-1 structure I with N the detected 
%	number of different channels, and the following fields:
%	    ChannelFullName: channel name,
%	        XBlockIndex: channel's vector index into X,
%	         ClockDrift: vector of time interval errors, in seconds,
%	                     between each data block (relative to sampling
%	                     period). This can be compared to "Max Clock Drift"
%	                     value of a Blockette 52.
%	                        = 0 in perfect case
%	                        < 0 tends to overlapping
%	                        > 0 tends to gapping
%	  OverlapBlockIndex: index of blocks (into X) having a significant 
%	                     overlap with previous block (less than 0.5
%	                     sampling period).
%	        OverlapTime: time vector of overlapped blocks (DATENUM format).
%	      GapBlockIndex: index of blocks (into X) having a significant gap
%	                     with next block (more than 0.5 sampling period).
%	            GapTime: time vector of gapped blocks (DATENUM format).
%
%	RDMSEED(...) without output arguments plots the imported signal by 
%	concatenating all the data records, in one single plot if single channel
%	is detected, or subplots for multiplexed file (limited to 10 channels).
%	Gaps are shown with red stars, overlaps with green circles.
%
%	[...] = RDMSEED(F,...,'be') forces big-endian reading (overwrites the
%	automatic detection of endianness coding, which fails in some cases).
%
%	[...] = RDMSEED(F,...,'notc') disable time correction.
%
%	[...] = RDMSEED(F,...,'plot') forces the plot with output arguments.
%
%	[...] = RDMSEED(F,...,'v') uses verbose mode (displays additional 
%	information and warnings when necessary). Use 'vv' for extras, 'vvv'
%	for debuging.
%
%	Some instructions for usage of the returned structure:
%	
%	- to get concatenated time and data vectors from a single-channel file:
%		X = rdmseed(f,'plot');
%		t = cat(1,X.t);
%		d = cat(1,X.d);
%
%	- to get the list of channels in a multiplexed file:
%		[X,I] = rdmseed(f);
%		char(I.ChannelFullName)
%
%	- to extract the station component n from a multiplexed file:
%		[X,I] = rdmseed(f);
%		k = I(n).XBlockIndex;
%		plot(cat(1,X(k).t),cat(1,X(k).d))
%		datetick('x')
%		title(I(n).ChannelFullName)
%
%	Known encoding formats are the following FDSN codes:
%		 0: ASCII
%		 1: 16-bit integer
%		 2: 24-bit integer
%		 3: 32-bit integer
%		 4: IEEE float32
%		 5: IEEE float64
%		10: Steim-1
%		11: Steim-2
%		12: GEOSCOPE 24-bit (untested)
%		13: GEOSCOPE 16/3-bit gain ranged
%		14: GEOSCOPE 16/4-bit gain ranged
%		19: Steim-3 (alpha and untested)
%
%	See also MKMSEED to export data in miniSEED format.
%
%
%	Author: François Beauducel <beauducel@ipgp.fr>
%		Institut de Physique du Globe de Paris
%	Created: 2010-09-17
%	Updated: 2015-01-05
%
%	Acknowledgments:
%		Ljupco Jordanovski, Jean-Marie Saurel, Mohamed Boubacar, Jonathan Berger,
%		Shahid Ullah, Wayne Crawford, Constanza Pardo, Sylvie Barbier,
%		Robert Chase, Arnaud Lemarchand, Alexandre Nercessian.
%
%	References:
%		IRIS (2010), SEED Reference Manual: SEED Format Version 2.4, May 2010,
%		  IFDSN/IRIS/USGS, http://www.iris.edu
%		Trabant C. (2010), libmseed: the Mini-SEED library, IRIS DMC.
%		Steim J.M. (1994), 'Steim' Compression, Quanterra Inc.

%	History:
%		[2015-01-05]
%			- fixes a bug when a data block has 0 sample declared in header
%			  but some data in the record (STEIM-1/2 coding).
%		[2014-06-29]
%			- 24-bit uncompressed format tested (bug correction), thanks to
%			  Arnaud Lemarchand.
%		[2014-05-31]
%			- applies the time correction to StartTime and X.t (if needed).
%			- new option 'notc' to disable time correction.
%			- Geoscope 16/4 format passed real data archive tests.
%			- fixes a problem when plotting multiplexed channels (thanks to 
%			  Robert Chase).
%		[2014-03-14]
%			- Improved endianness automatic detection (see comments).
%			- Accepts mixed little/big endian encoding in a single file.
%			- minor fixes.
%		[2013-10-25]
%			- Due to obsolete syntax of bitcmp(0,N) in R2013b, replaces all
%			  by: 2^N-1 (which is much faster...)
%		[2013-02-15]
%			- Tests also DayOfYear in header to determine automatically 
%			  little-endian coding of the file.
%			- Adds option 'be' to force big-endian reading (overwrites
%			  automatic detection).
%		[2012-12-21]
%			- Adds a verbose mode
%		[2012-04-21]
%			- Correct bug with Steim + little-endian coding
%			  (thanks to Shahid Ullah)
%		[2012-03-21]
%			- Adds IDs for warning messages
%		[2011-11-10]
%			- Correct bug with multiple channel name length (thanks to
%			  Jonathan Berger)
%		[2011-10-27]
%			- Add LocationIdentifier to X.ChannelFullName
%		[2011-10-24]
%			- Validation of IEEE double encoding (with PQL)
%			- Import/plot data even with file integrity problem (like PQL)
%		[2011-07-21]
%			- Validation of ASCII encoding format (logs)
%			- Blockettes are now stored in substructures below a single
%			  field X.BLOCKETTES
%			- Add import of blockettes 500 and 2000
%			- Accept multi-channel files with various data coding
%		[2010-10-16]
%			- Alpha-version of Steim-3 decoding...
%			- Extend output parameters with channel detection
%			- Add gaps and overlaps on plots
%			- Add possibility to force the plot
%		[2010-10-02]
%			- Add the input formats for GEOSCOPE multiplexed old data files
%			- Additional output argument with gap and overlap analysis
%			- Create a plot when no output argument are specified
%			- Optimize script coding (30 times faster STEIM decoding!)
%		[2010-09-28]
%			- Correction of a problem with STEIM-1 nibble 3 decoding (one 
%			  32-bit difference)
%			- Add reading of files without blockette 1000 with additional
%			  input arguments (like Seismic Handler output files).
%			- Uses warning() function instead of fprintf().
%
%	Copyright (c) 2014, François Beauducel, covered by BSD License.
%	All rights reserved.
%
%	Redistribution and use in source and binary forms, with or without 
%	modification, are permitted provided that the following conditions are 
%	met:
%
%	   * Redistributions of source code must retain the above copyright 
%	     notice, this list of conditions and the following disclaimer.
%	   * Redistributions in binary form must reproduce the above copyright 
%	     notice, this list of conditions and the following disclaimer in 
%	     the documentation and/or other materials provided with the distribution
%	                           
%	THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
%	AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
%	IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 
%	ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 
%	LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 
%	CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 
%	SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 
%	INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 
%	CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 
%	ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 
%	POSSIBILITY OF SUCH DAMAGE.

if nargin > 6
	error('Too many input arguments.')
end

% global variables shared with sub-functions
global f fid offset le ef wo rl forcebe verbose notc force

% default input arguments
makeplot = 0;	% make plot flag
verbose = 0;	% verbose flag/level 
forcebe = 0;	% force big-endian
ef = 10;		% encoding format default
wo = 1;			% word order default
rl = 2^12;		% record length default
force = 0;		% force input argument over blockette 1000 (UNDOCUMENTED)
notc = 0;		% force no time correction (over ActivityFlags)

if nargin < 1
	[filename,pathname] = uigetfile('*','Please select a miniSEED file...');
	f = fullfile(pathname,filename);
else
	f = varargin{1};
end

if ~ischar(f) || ~exist(f,'file')
	error('File %s does not exist.',f);
end

if nargin > 1
	verbose = any(strcmpi(varargin,'v')) + 2*any(strcmpi(varargin,'vv')) ...
	          + 3*any(strcmpi(varargin,'vvv'));
	makeplot = any(strcmpi(varargin,'plot'));
	forcebe = any(strcmpi(varargin,'be'));
	notc = any(strcmpi(varargin,'notc'));
	force = any(strcmpi(varargin,'force'));
end
nargs = (makeplot>0) + (verbose>0) + (forcebe>0) + (notc>0) + (force>0);


if nargin > (1 + nargs)
	ef = varargin{2};
	if ~isnumeric(ef) || ~any(ef==[0:5,10:19,30:33])
		error('Argument ENCODINGFORMAT must be a valid FDSN code value.');
	end
end

if nargin > (2 + nargs)
	wo = varargin{3};
	if ~isnumeric(wo) || (wo ~= 0 && wo ~= 1)
		error('Argument WORDORDER must be 0 or 1.');
	end
end

if nargin > (3 + nargs)
	rl = varargin{4};
	if ~isnumeric(rl) || rl < 256 || rem(log(rl)/log(2),1) ~= 0
		error('Argument RECORDLENGTH must be a power of 2 and greater or equal to 256.');
	end
end

if nargout == 0
	makeplot = 1;
end

% sensible limits for multiplexed files
max_channels = 20;	% absolute max number of channels to plot
max_channel_label = 6;	% max. number of channels for y-labels

% file is opened in Big-Endian encoding (this is encouraged by SEED)
fid = fopen(f,'rb','ieee-be');
le = 0;

% --- tests if the header is mini-SEED
% the 7th character must be one of the "data header/quality indicator", usually 'D'
header = fread(fid,20,'*char');
if ~ismember(header(7),'DRMQ')
	if ismember(header(7),'VAST')
		s = ' (seems to be a SEED Volume)';
	else
		s = '';
	end
	error('File is not in mini-SEED format%s. Cannot read it.',s);
end

i = 1;
offset = 0;

while offset >= 0
	X(i) = read_data_record;
	i = i + 1;
end

fclose(fid);

if nargout > 0
	varargout{1} = X;
end

% --- analyses data
if makeplot || nargout > 1

	% test if the file is multiplexed or a single channel
	un = unique(cellstr(char(X.ChannelFullName)));
	nc = numel(un);
	for i = 1:nc
		k = find(strcmp(cellstr(char(X.ChannelFullName)),un{i}));
		I(i).ChannelFullName = X(k(1)).ChannelFullName;
		I(i).XBlockIndex = k;
		I(i).ClockDrift = ([diff(cat(1,X(k).RecordStartTimeMATLAB));NaN]*86400 - cat(1,X(k).NumberSamples)./cat(1,X(k).SampleRate))./cat(1,X(k).NumberSamples);
		I(i).OverlapBlockIndex = k(find(I(i).ClockDrift.*cat(1,X(k).NumberSamples).*cat(1,X(k).SampleRate) < -.5) + 1);
		I(i).OverlapTime = cat(1,X(I(i).OverlapBlockIndex).RecordStartTimeMATLAB);
		I(i).GapBlockIndex = k(find(I(i).ClockDrift.*cat(1,X(k).NumberSamples).*cat(1,X(k).SampleRate) > .5) + 1);
		I(i).GapTime = cat(1,X(I(i).GapBlockIndex).RecordStartTimeMATLAB);
	end
end
if nargout > 1
	varargout{2} = I;
end

% --- plots the data
if makeplot

	figure
	
	xlim = [min(cat(1,X.t)),max(cat(1,X.t))];

	% test if all data records have the same length
	rl = unique(cat(1,X.DataRecordSize));
	if numel(rl) == 1
		rl_text = sprintf('%d bytes',rl);
	else
		rl_text = sprintf('%d-%d bytes',min(rl),max(rl));
	end
	
	% test if all data records have the same sampling rate
	sr = unique(cat(1,X.SampleRate));
	if numel(sr) == 1
		sr_text = sprintf('%g Hz',sr);
	else
		sr_text = sprintf('%d # samp. rates',numel(sr));
	end
	
	% test if all data records have the same encoding format
	ef = unique(cellstr(cat(1,X.EncodingFormatName)));
	if numel(ef) == 1
		ef_text = sprintf('%s',ef{:});
	else
		ef_text = sprintf('%d different encod. formats',numel(ef));
	end
			
	if nc == 1
		plot(cat(1,X.t),cat(1,X.d))
		hold on
		for i = 1:length(I.GapBlockIndex)
			plot(I.GapTime(i),X(I.GapBlockIndex(i)).d(1),'*r')
		end
		for i = 1:length(I.OverlapBlockIndex)
			plot(I.OverlapTime(i),X(I.OverlapBlockIndex(i)).d(1),'og')
		end
		hold off
		set(gca,'XLim',xlim)
		datetick('x','keeplimits')
		grid on
		xlabel(sprintf('Time\n(%s to %s)',datestr(xlim(1)),datestr(xlim(2))))
		ylabel('Counts')
		title(sprintf('mini-SEED file "%s"\n%s (%d rec. @ %s - %g samp. @ %s - %s)', ...
			f,un{1},length(X),rl_text,numel(cat(1,X.d)),sr_text,ef_text),'Interpreter','none')
	else
		% plot is done only for real data channels...
		if nc > max_channels
			warning('Plot has been limited to %d channels (over %d). See help to manage multiplexed file.', ...
				max_channels,nc);
			nc = max_channels;
		end
		for i = 1:nc
			subplot(nc*2,1,i*2 + (-1:0))
			k = I(i).XBlockIndex;
			if ~any(strcmp('ASCII',cellstr(cat(1,X(k).EncodingFormatName))))
				plot(cat(1,X(k).t),cat(1,X(k).d))
				hold on
				for ii = 1:length(I(i).GapBlockIndex)
					if ~isempty(X(I(i).GapBlockIndex(ii)).d)
						plot(I(i).GapTime(ii),X(I(i).GapBlockIndex(ii)).d,'r')
					else
						plot(repmat(I(i).GapTime(ii),1,2),ylim,'r')
					end
				end
				for ii = 1:length(I(i).OverlapBlockIndex)
					if ~isempty(X(I(i).OverlapBlockIndex(ii)).d)
						plot(I(i).OverlapTime(ii),X(I(i).OverlapBlockIndex(ii)).d,'g')
					else
						plot(repmat(I(i).OverlapTime(ii),1,2),ylim,'g')
					end
				end
				hold off
			end
			set(gca,'XLim',xlim,'FontSize',8)
			h = ylabel(un{i},'Interpreter','none');
			if nc > max_channel_label
				set(gca,'YTick',[])
				set(h,'Rotation',0,'HorizontalAlignment','right','FontSize',8)
			end
			datetick('x','keeplimits')
			set(gca,'XTickLabel',[])
			grid on
			if i == 1
				title(sprintf('mini-SEED file "%s"\n%d channels (%d rec. @ %s - %g data - %s - %s)', ...
					f,length(un),length(X),rl_text,numel(cat(1,X(k).d)),sr_text,ef_text),'Interpreter','none')
			end
			if i == nc
				datetick('x','keeplimits')
				xlabel(sprintf('Time\n(%s to %s)',datestr(xlim(1)),datestr(xlim(2))))
			end
		end
		v = version;
		if str2double(v(1))>=7
			linkaxes(findobj(gcf,'type','axes'),'x')
		end
	end
end


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function D = read_data_record
% read_data_record uses global variables f, fid, offset, le, ef, wo, rl, 
%	and verbose. It reads a data record and returns a structure D.

global f fid offset le ef wo rl verbose notc force

fseek(fid,offset,'bof');

% --- read fixed section of Data Header (48 bytes)
D.SequenceNumber        = fread(fid,6,'*char')';
D.DataQualityIndicator  = fread(fid,1,'*char');
D.ReservedByte          = fread(fid,1,'*char');
D.StationIdentifierCode = fread(fid,5,'*char')';
D.LocationIdentifier    = fread(fid,2,'*char')';
D.ChannelIdentifier	    = fread(fid,3,'*char')';
D.NetworkCode           = fread(fid,2,'*char')';
D.ChannelFullName = sprintf('%s:%s:%s:%s',deblank(D.NetworkCode), ...
	deblank(D.StationIdentifierCode),deblank(D.LocationIdentifier), ...
	deblank(D.ChannelIdentifier));

% Start Time decoding
[D.RecordStartTime,swapflag] = readbtime;
D.RecordStartTimeISO = sprintf('%4d-%03d %02d:%02d:%07.4f',D.RecordStartTime);

if swapflag
	if le
		machinefmt = 'ieee-be';
		le = 0;
	else
		machinefmt = 'ieee-le';
		le = 1;
	end
	position = ftell(fid);
	fclose(fid);
	fid = fopen(f,'rb',machinefmt);
	fseek(fid,position,'bof');
	if verbose > 0
		warning('RDMSEED:DataIntegrity', ...
			'Sequence # %s: need to switch file encoding to %s...\n', ...
			D.SequenceNumber,machinefmt);
	end
end

D.NumberSamples	        = fread(fid,1,'uint16');

% Sample Rate decoding
SampleRateFactor        = fread(fid,1,'int16');
SampleRateMultiplier    = fread(fid,1,'int16');
if SampleRateFactor > 0
	if SampleRateMultiplier >= 0
		D.SampleRate = SampleRateFactor*SampleRateMultiplier;
	else
		D.SampleRate = -1*SampleRateFactor/SampleRateMultiplier;
	end
else
	if SampleRateMultiplier >= 0
		D.SampleRate = -1*SampleRateMultiplier/SampleRateFactor;
	else
		D.SampleRate = 1/(SampleRateFactor*SampleRateMultiplier);
	end
end

D.ActivityFlags          = fread(fid,1,'uint8');
D.IOFlags                = fread(fid,1,'uint8');
D.DataQualityFlags       = fread(fid,1,'uint8');
D.NumberBlockettesFollow = fread(fid,1,'uint8');
D.TimeCorrection         = fread(fid,1,'int32');	% Time correction in 0.0001 s
D.OffsetBeginData        = fread(fid,1,'uint16');
D.OffsetFirstBlockette   = fread(fid,1,'uint16');

% --- read the blockettes
OffsetNextBlockette = D.OffsetFirstBlockette;

D.BLOCKETTES = [];
b2000 = 0;	% Number of Blockette 2000

for i = 1:D.NumberBlockettesFollow
	fseek(fid,offset + OffsetNextBlockette,'bof');
	BlocketteType = fread(fid,1,'uint16');
	
	switch BlocketteType
		
		case 1000
			% BLOCKETTE 1000 = Data Only SEED (8 bytes)
			OffsetNextBlockette = fread(fid,1,'uint16');
			D.BLOCKETTES.B1000.EncodingFormat = fread(fid,1,'uint8');
			D.BLOCKETTES.B1000.WordOrder = fread(fid,1,'uint8');
			D.BLOCKETTES.B1000.DataRecordLength = fread(fid,1,'uint8');
			D.BLOCKETTES.B1000.Reserved = fread(fid,1,'uint8');
			
		case 1001
			% BLOCKETTE 1001 = Data Extension (8 bytes)
			OffsetNextBlockette = fread(fid,1,'uint16');
			D.BLOCKETTES.B1001.TimingQuality = fread(fid,1,'uint8');
			D.BLOCKETTES.B1001.Micro_sec = fread(fid,1,'int8');
			D.BLOCKETTES.B1001.Reserved = fread(fid,1,'uint8');
			D.BLOCKETTES.B1001.FrameCount = fread(fid,1,'uint8');
			
		case 100
			% BLOCKETTE 100 = Sample Rate (12 bytes)
			OffsetNextBlockette = fread(fid,1,'uint16');
			D.BLOCKETTES.B100.ActualSampleRate = fread(fid,1,'float32');
			D.BLOCKETTES.B100.Flags = fread(fid,1,'uint8');
			D.BLOCKETTES.B100.Reserved = fread(fid,1,'uint8');
		
		case 500
			% BLOCKETTE 500 = Timing (200 bytes)
			OffsetNextBlockette = fread(fid,1,'uint16');
			D.BLOCKETTES.B500.VCOCorrection = fread(fid,1,'float32');
			D.BLOCKETTES.B500.TimeOfException = readbtime;
			D.BLOCKETTES.B500.MicroSec = fread(fid,1,'int8');
			D.BLOCKETTES.B500.ReceptionQuality = fread(fid,1,'uint8');
			D.BLOCKETTES.B500.ExceptionCount = fread(fid,1,'uint16');
			D.BLOCKETTES.B500.ExceptionType = fread(fid,16,'*char')';
			D.BLOCKETTES.B500.ClockModel = fread(fid,32,'*char')';
			D.BLOCKETTES.B500.ClockStatus = fread(fid,128,'*char')';
		
		case 2000
			% BLOCKETTE 2000 = Opaque Data (variable length)
			b2000 = b2000 + 1;
			OffsetNextBlockette = fread(fid,1,'uint16');
			BlocketteLength = fread(fid,1,'uint16');
			OffsetOpaqueData = fread(fid,1,'uint16');
			D.BLOCKETTES.B2000(b2000).RecordNumber = fread(fid,1,'uint32');
			D.BLOCKETTES.B2000(b2000).DataWordOrder = fread(fid,1,'uint8');
			D.BLOCKETTES.B2000(b2000).Flags = fread(fid,1,'uint8');
			NumberHeaderFields = fread(fid,1,'uint8');
			HeaderFields = splitfield(fread(fid,OffsetOpaqueData-15,'*char')','~');
			D.BLOCKETTES.B2000(b2000).HeaderFields = HeaderFields(1:NumberHeaderFields);
			% Opaque data are stored as a single char string, but must be
			% decoded using appropriate format (e.g., Quanterra Q330)
			D.BLOCKETTES.B2000(b2000).OpaqueData = fread(fid,BlocketteLength-OffsetOpaqueData,'*char')';
		
		otherwise
			OffsetNextBlockette = fread(fid,1,'uint16');

			if verbose > 0
				warning('RDMSEED:UnknownBlockette', ...
					'Unknown Blockette number %d (%s)!\n', ...
					BlocketteType,D.ChannelFullName);
			end
	end
end

% --- read the data stream
fseek(fid,offset + D.OffsetBeginData,'bof');

if ~force && isfield(D.BLOCKETTES,'B1000')
	EncodingFormat = D.BLOCKETTES.B1000.EncodingFormat;
	WordOrder = D.BLOCKETTES.B1000.WordOrder;
	D.DataRecordSize = 2^D.BLOCKETTES.B1000.DataRecordLength;
else
	EncodingFormat = ef;
	WordOrder = wo;
	D.DataRecordSize = rl;
end

uncoded = 0;

D.d = NaN;
D.t = NaN;

switch EncodingFormat
	
	case 0
		% --- decoding format: ASCII text
		D.EncodingFormatName = {'ASCII'};
		D.d = fread(fid,D.DataRecordSize - D.OffsetBeginData,'*char')';

	case 1
		% --- decoding format: 16-bit integers
		D.EncodingFormatName = {'INT16'};
		dd = fread(fid,ceil((D.DataRecordSize - D.OffsetBeginData)/2),'*int16');
		if xor(~WordOrder,le)
			dd = swapbytes(dd);
		end
		D.d = dd(1:D.NumberSamples);
		
	case 2
		% --- decoding format: 24-bit integers
		D.EncodingFormatName = {'INT24'};
		dd = fread(fid,ceil((D.DataRecordSize - D.OffsetBeginData)/3),'bit24=>int32');
		if xor(~WordOrder,le)
			dd = swapbytes(dd);
		end
		D.d = dd(1:D.NumberSamples);
		
	case 3
		% --- decoding format: 32-bit integers
		D.EncodingFormatName = {'INT32'};
		dd = fread(fid,ceil((D.DataRecordSize - D.OffsetBeginData)/4),'*int32');
		if xor(~WordOrder,le)
			dd = swapbytes(dd);
		end
		D.d = dd(1:D.NumberSamples);
		
	case 4
		% --- decoding format: IEEE floating point
		D.EncodingFormatName = {'FLOAT32'};
		dd = fread(fid,ceil((D.DataRecordSize - D.OffsetBeginData)/4),'*float');
		if xor(~WordOrder,le)
			dd = swapbytes(dd);
		end
		D.d = dd(1:D.NumberSamples);
		
	case 5
		% --- decoding format: IEEE double precision floating point
		D.EncodingFormatName = {'FLOAT64'};
		dd = fread(fid,ceil((D.DataRecordSize - D.OffsetBeginData)/8),'*double');
		if xor(~WordOrder,le)
			dd = swapbytes(dd);
		end
		D.d = dd(1:D.NumberSamples);

	case {10,11,19}
		% --- decoding formats: STEIM-1 and STEIM-2 compression
		%    (c) Joseph M. Steim, Quanterra Inc., 1994
		steim = find(EncodingFormat==[10,11,19]);
		D.EncodingFormatName = {sprintf('STEIM%d',steim)};
		
		% Steim compression decoding strategy optimized for Matlab
		% -- by F. Beauducel, October 2010 --
		%
		%	1. loads all data into a single 16xM uint32 array
		%	2. gets all nibbles from the first row splitted into 2-bit values
		%	3. for each possible nibble value, selects (find) and decodes
		%	   (bitsplit) all the corresponding words, and stores results
		%	   in a 4xN (STEIM1) or 7xN (STEIM2) array previously filled with
		%	   NaN's. For STEIM2 with nibbles 2 or 3, decodes also dnib values
		%	   (first 2-bit of the word)
		%	5. reduces this array with non-NaN values only
		%	6. integrates with cumsum
		%
		% This method is about 30 times faster than a 'C-like' loops coding...
		
		frame32 = fread(fid,[16,(D.DataRecordSize - D.OffsetBeginData)/64],'*uint32');
		if xor(~WordOrder,le)
			frame32 = swapbytes(frame32);
		end
		
		% specific processes for STEIM-3
		if steim == 3
			% first bit = 1 means second differences
			SecondDiff = bitshift(frame32(1,:),-31);
			% checks for "squeezed flag"... and replaces frame32(1,:)
			squeezed = bitand(bitshift(frame32(1,:),-24),127);
			k = find(bitget(squeezed,7));
			if ~isempty(k)
				moredata24 = bitand(frame32(1,k),16777215);
				k = find(squeezed == 80);	% upper nibble 8-bit = 0x50
				if ~isempty(k)
					frame32(1,k) = hex2dec('15555555');
				end
				k = find(squeezed == 96);	% upper nibble 8-bit = 0x60
				if ~isempty(k)
					frame32(1,k) = hex2dec('2aaaaaaa');
				end
				k = find(squeezed == 112);	% upper nibble 8-bit = 0x70
				if ~isempty(k)
					frame32(1,k) = hex2dec('3fffffff');
				end
			end
		end

		% nibbles is an array of the same size as frame32...
		nibbles = bitand(bitshift(repmat(frame32(1,:),16,1),repmat(-30:2:0,size(frame32,2),1)'),3);
		x0 = bitsign(frame32(2,1),32);	% forward integration constant
		xn = bitsign(frame32(3,1),32);	% reverse integration constant
		
		switch steim
			
		case 1
			% STEIM-1: 3 cases following the nibbles
			ddd = NaN*ones(4,numel(frame32));	% initiates array with NaN
			k = find(nibbles == 1);			% nibble = 1 : four 8-bit differences
			if ~isempty(k)
				ddd(1:4,k) = bitsplit(frame32(k),32,8);
			end
			k = find(nibbles == 2);			% nibble = 2 : two 16-bit differences
			if ~isempty(k)
				ddd(1:2,k) = bitsplit(frame32(k),32,16);
			end
			k = find(nibbles == 3);			% nibble = 3 : one 32-bit difference
			if ~isempty(k)
				ddd(1,k) = bitsign(frame32(k),32);
			end

		case 2	
			% STEIM-2: 7 cases following the nibbles and dnib
			ddd = NaN*ones(7,numel(frame32));	% initiates array with NaN
			k = find(nibbles == 1);			% nibble = 1 : four 8-bit differences
			if ~isempty(k)
				ddd(1:4,k) = bitsplit(frame32(k),32,8);
			end
			k = find(nibbles == 2);			% nibble = 2 : must look in dnib
			if ~isempty(k)
				dnib = bitshift(frame32(k),-30);
				kk = k(dnib == 1);		% dnib = 1 : one 30-bit difference
				if ~isempty(kk)
					ddd(1,kk) = bitsign(frame32(kk),30);
				end
				kk = k(dnib == 2);		% dnib = 2 : two 15-bit differences
				if ~isempty(kk)
					ddd(1:2,kk) = bitsplit(frame32(kk),30,15);
				end
				kk = k(dnib == 3);		% dnib = 3 : three 10-bit differences
				if ~isempty(kk)
					ddd(1:3,kk) = bitsplit(frame32(kk),30,10);
				end
			end
			k = find(nibbles == 3);				% nibble = 3 : must look in dnib
			if ~isempty(k)
				dnib = bitshift(frame32(k),-30);
				kk = k(dnib == 0);		% dnib = 0 : five 6-bit difference
				if ~isempty(kk)
					ddd(1:5,kk) = bitsplit(frame32(kk),30,6);
				end
				kk = k(dnib == 1);		% dnib = 1 : six 5-bit differences
				if ~isempty(kk)
					ddd(1:6,kk) = bitsplit(frame32(kk),30,5);
				end
				kk = k(dnib == 2);		% dnib = 2 : seven 4-bit differences (28 bits!)
				if ~isempty(kk)
					ddd(1:7,kk) = bitsplit(frame32(kk),28,4);
				end
			end
			
		case 3	% *** STEIM-3 DECODING IS ALPHA AND UNTESTED ***
			% STEIM-3: 7 cases following the nibbles
			ddd = NaN*ones(9,numel(frame32));	% initiates array with NaN
			k = find(nibbles == 0);				% nibble = 0 : two 16-bit differences
			if ~isempty(k)
				ddd(1:2,k) = bitsplit(frame32(k),32,16);
			end
			k = find(nibbles == 1);				% nibble = 1 : four 8-bit differences
			if ~isempty(k)
				ddd(1:4,k) = bitsplit(frame32(k),32,8);
			end
			k = find(nibbles == 2);				% nibble = 2 : must look even dnib
			if ~isempty(k)
				dnib2 = bitshift(frame32(k(2:2:end)),-30);
				w60 = bitand(frame32(k(2:2:end)),1073741823) ...
					+ bitshift(bitand(frame32(k(1:2:end)),1073741823),30);	% concatenates two 30-bit words
				kk = find(dnib2 == 0);		% dnib = 0: five 12-bit differences (60 bits)
				if ~isempty(kk)
					ddd(1:5,k(2*kk)) = bitsplit(w60,60,12);
				end
				kk = find(dnib2 == 1);		% dnib = 1: three 20-bit differences (60 bits)
				if ~isempty(kk)
					ddd(1:3,k(2*kk)) = bitsplit(w60,60,20);
				end
			end
			k = find(nibbles == 3);				% nibble = 3 : must look 3rd bit
			if ~isempty(k)
				dnib = bitshift(frame32(k),-27);
				kk = k(dnib == 24);		% dnib = 11000 : nine 3-bit differences (27 bits)
				if ~isempty(kk)
					ddd(1:9,kk) = bitsplit(frame32(kk),27,3);
				end
				kk = k(dnib == 25);		% dnib = 11001 : Not A Difference
				if ~isempty(kk)
					ddd(1,kk) = bitsign(frame32(kk),27);
				end
				kk = k(dnib > 27);		% dnib = 111.. : 29-bit sample (29 bits)
				if ~isempty(kk)
					ddd(1,kk) = bitsign(frame32(kk),29);
				end
			end
		end
		
		% Little-endian coding: needs to swap bytes
		if ~WordOrder
			ddd = flipud(ddd);
		end
		dd = ddd(~isnan(ddd));		% reduces initial array ddd: dd is non-NaN values of ddd
		
		% controls the number of samples
		if numel(dd) ~= D.NumberSamples
			if verbose > 1
				warning('RDMSEED:DataIntegrity','Problem in %s sequence # %s [%d-%03d %02d:%02d:%07.4f]: number of samples in header (%d) does not equal data (%d).\n', ...
					D.EncodingFormatName{:},D.SequenceNumber,D.RecordStartTimeISO,D.NumberSamples,numel(dd));
			end
			if numel(dd) < D.NumberSamples
				D.NumberSamples = numel(dd);
			end
		end

		% rebuilds the data vector by integrating the differences
		D.d = cumsum([x0;dd(2:D.NumberSamples)]);
		
		% controls data integrity...
		if D.d(end) ~= xn
			warning('RDMSEED:DataIntegrity','Problem in %s sequence # %s [%s]: data integrity check failed, last_data=%d, Xn=%d.\n', ...
				D.EncodingFormatName{:},D.SequenceNumber,D.RecordStartTimeISO,D.d(end),xn);
		end

		if D.NumberSamples == 0
			D.d = nan(0,1);
		end
		
		% for debug purpose...
		if verbose > 2
			D.dd = dd;
			D.nibbles = nibbles;
			D.x0 = x0;
			D.xn = xn;
		end

	case 12
		% --- decoding format: GEOSCOPE multiplexed 24-bit integer
		D.EncodingFormatName = {'GEOSCOPE24'};
		dd = fread(fid,(D.DataRecordSize - D.OffsetBeginData)/3,'bit24=>double');
		if xor(~WordOrder,le)
			dd = swapbytes(dd);
		end
		D.d = dd(1:D.NumberSamples);
		
	case {13,14}
		% --- decoding format: GEOSCOPE multiplexed 16/3 and 16/4 bit gain ranged
		%	(13): 16/3-bit (bit 15 is unused)
		%	(14): 16/4-bit
		%	bits 15-12 = 3 or 4-bit gain exponent (positive) 
		%	bits 11-0 = 12-bit mantissa (positive)
		%	=> data = (mantissa - 2048) / 2^gain
		geoscope = 7 + 8*(EncodingFormat==14); % mask for gain exponent
		D.EncodingFormatName = {sprintf('GEOSCOPE16-%d',EncodingFormat-10)};
		dd = fread(fid,(D.DataRecordSize - D.OffsetBeginData)/2,'*uint16');
		if xor(~WordOrder,le)
			dd = swapbytes(dd);
		end
		dd = (double(bitand(dd,2^12-1))-2^11)./2.^double(bitand(bitshift(dd,-12),geoscope));
		D.d = dd(1:D.NumberSamples);
		
	case 15
		% --- decoding format: US National Network compression
		D.EncodingFormatName = {'USNN'};
		uncoded = 1;
		
	case 16
		% --- decoding format: CDSN 16-bit gain ranged
		D.EncodingFormatName = {'CDSN'};
		uncoded = 1;
		
	case 17
		% --- decoding format: Graefenberg 16-bit gain ranged
		D.EncodingFormatName = {'GRAEFENBERG'};
		uncoded = 1;
		
	case 18
		% --- decoding format: IPG - Strasbourg 16-bit gain ranged
		D.EncodingFormatName = {'IPGS'};
		uncoded = 1;
		
	case 30
		% --- decoding format: SRO format
		D.EncodingFormatName = {'SRO'};
		uncoded = 1;
		
	case 31
		% --- decoding format: HGLP format
		D.EncodingFormatName = {'HGLP'};
		uncoded = 1;
		
	case 32
		% --- decoding format: DWWSSN gain ranged format
		D.EncodingFormatName = {'DWWSSN'};
		uncoded = 1;
		
	case 33
		% --- decoding format: RSTN 16-bit gain ranged
		D.EncodingFormatName = {'RSTN'};
		uncoded = 1;
		
	otherwise
		D.EncodingFormatName = {sprintf('** Unknown (%d) **',EncodingFormat)};
		uncoded = 1;
		
end

if uncoded
	error('Sorry, the encoding format "%s" is not yet implemented.',D.EncodingFormatName);
end

% Applies time correction (if needed)
D.RecordStartTimeMATLAB = datenum(double([D.RecordStartTime(1),0,D.RecordStartTime(2:5)])) ...
	+ (~notc & bitand(D.ActivityFlags,2) == 0)*D.TimeCorrection/1e4/86400;
tv = datevec(D.RecordStartTimeMATLAB);
doy = datenum(tv(1:3)) - datenum(tv(1),1,0);
D.RecordStartTime = [tv(1),doy,tv(4:5),round(tv(6)*1e4)/1e4];
D.RecordStartTimeISO = sprintf('%4d-%03d %02d:%02d:%07.4f',D.RecordStartTime);

D.t = D.RecordStartTimeMATLAB;

% makes the time vector and applies time correction (if needed)
if EncodingFormat > 0
	D.t = D.t + (0:(D.NumberSamples-1))'/(D.SampleRate*86400);
end


offset = ftell(fid);
fread(fid,1,'char');	% this is to force EOF=1 on last record.
if feof(fid)
	offset = -1;
end



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function c = splitfield(s,d)
% splitfield(S) splits string S of D-character separated field names
C = textscan(s,'%s','Delimiter',d);
c = C{1};


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [d,swapflag] = readbtime
% readbtime reads BTIME structure from current opened file and returns
%	D = [YEAR,DAY,HOUR,MINUTE,SECONDS]

global fid forcebe

Year		= fread(fid,1,'*uint16');
DayOfYear	= fread(fid,1,'*uint16');
Hours		= fread(fid,1,'uint8');
Minutes		= fread(fid,1,'uint8');
Seconds		= fread(fid,1,'uint8');
fseek(fid,1,0);	% skip 1 byte (unused)
Seconds0001	= fread(fid,1,'*uint16');

% Automatic detection of little/big-endian encoding
% -- by F. Beauducel, March 2014 --
%
% If the 2-byte day is >= 512, the file is not opened in the correct
% endianness. If the day is 1 or 256, there is a possible byte-swap and we
% need to check also the year; but we need to consider what is a valid year:
% - years from 1801 to 2047 are OK (swapbytes >= 2312)
% - years from 2048 to 2055 are OK (swapbytes <= 1800)
% - year 2056 is ambiguous (swapbytes = 2056)
% - years from 2057 to 2311 are OK (swapbytes >= 2312)
% - year 1799 is ambiguous (swapbytes = 1799)
% - year 1800 is suspicious (swapbytes = 2055)
%
% Thus, the only cases for which we are 'sure' there is a byte-swap, are:
% - day >= 512
% - (day == 1 or day == 256) and (year < 1799 or year > 2311)
%
% Note: in IRIS libmseed, the test is only year>2050 or year<1920.
if ~forcebe && (DayOfYear >= 512 || (ismember(DayOfYear,[1,256]) && (Year > 2311 || Year < 1799)))
	swapflag = 1;
	Year = swapbytes(Year);
	DayOfYear = swapbytes(DayOfYear);
	Seconds0001 = swapbytes(Seconds0001);
else
	swapflag = 0;
end
d = [double(Year),double(DayOfYear),Hours,Minutes,Seconds + double(Seconds0001)/1e4];


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function d = bitsplit(x,b,n)
% bitsplit(X,B,N) splits the B-bit number X into signed N-bit array
%	X must be unsigned integer class
%	N ranges from 1 to B
%	B is a multiple of N

sign = repmat((b:-n:n)',1,size(x,1));
x = repmat(x',b/n,1);
d = double(bitand(bitshift(x,flipud(sign-b)),2^n-1)) ...
	- double(bitget(x,sign))*2^n;



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function d = bitsign(x,n)
% bitsign(X,N) returns signed double value from unsigned N-bit number X.
% This is equivalent to bitsplit(X,N,N), but the formula is simplified so
% it is much more efficient

d = double(bitand(x,2^n-1)) - double(bitget(x,n)).*2^n;