Please note that iconv('UTF-8', 'ASCII//TRANSLIT', ...) doesn't work properly when locale category LC_CTYPE is set to C or POSIX. You must choose another locale otherwise all non-ASCII characters will be replaced with question marks. This is at least true with glibc 2.5.

Example:
<?php
setlocale(LC_CTYPE, 'POSIX');
echo iconv('UTF-8', 'ASCII//TRANSLIT', "Žluťoučký kůň\n");
// ?lu?ou?k? k??

setlocale(LC_CTYPE, 'cs_CZ');
echo iconv('UTF-8', 'ASCII//TRANSLIT', "Žluťoučký kůň\n");
// Zlutoucky kun
?>

The "//ignore" option doesn't work with recent versions of the iconv library.  So if you're having trouble with that option, you aren't alone.  

That means you can't currently use this function to filter invalid characters.  Instead it silently fails and returns an empty string (or you'll get a notice but only if you have E_NOTICE enabled).

This has been a known bug with a known solution for at least since 2009 years but no one seems to be willing to fix it (PHP must pass the -c option to iconv).  It's still broken as of the latest release 5.4.3. 

https://bugs.php.net/bug.php?id=48147
https://bugs.php.net/bug.php?id=52211
https://bugs.php.net/bug.php?id=61484

[UPDATE 15-JUN-2012]
Here's a workaround...

  ini_set('mbstring.substitute_character', "none");
  $text= mb_convert_encoding($text, 'UTF-8', 'UTF-8');

That will strip invalid characters from UTF-8 strings (so that you can insert it into a database, etc.).  Instead of "none" you can also use the value 32 if you want it to insert spaces in place of the invalid characters.

Interestingly, setting different target locales results in different, yet appropriate, transliterations. For example:

<?php
//some German
$utf8_sentence = 'Weiß, Goldmann, Göbel, Weiss, Göthe, Goethe und Götz';

//UK
setlocale(LC_ALL, 'en_GB');

//transliterate
$trans_sentence = iconv('UTF-8', 'ASCII//TRANSLIT', $utf8_sentence);

//gives [Weiss, Goldmann, Gobel, Weiss, Gothe, Goethe und Gotz]
//which is our original string flattened into 7-bit ASCII as
//an English speaker would do it (ie. simply remove the umlauts)
echo $trans_sentence . PHP_EOL;

//Germany
setlocale(LC_ALL, 'de_DE');

$trans_sentence = iconv('UTF-8', 'ASCII//TRANSLIT', $utf8_sentence);

//gives [Weiss, Goldmann, Goebel, Weiss, Goethe, Goethe und Goetz]
//which is exactly how a German would transliterate those
//umlauted characters if forced to use 7-bit ASCII!
//(because really ä = ae, ö = oe and ü = ue)
echo $trans_sentence . PHP_EOL;

?>

to test different combinations of convertions between charsets (when we don't know the source charset and what is the convenient destination charset) this is an example :

<?php
$tab = array("UTF-8", "ASCII", "Windows-1252", "ISO-8859-15", "ISO-8859-1", "ISO-8859-6", "CP1256");
$chain = "";
foreach ($tab as $i)
    {
        foreach ($tab as $j)
        {
            $chain .= " $i$j ".iconv($i, $j, "$my_string");
        }
    }

echo $chain;
?>

then after displaying, you use the $i$j that shows good displaying.
NB: you can add other charsets to $tab  to test other cases.

If you want to convert to a Unicode encoding without the byte order mark (BOM), add the endianness to the encoding, e.g. instead of "UTF-16" which will add a BOM to the start of the string, use "UTF-16BE" which will convert the string without adding a BOM.

i.e.

<?php
iconv('CP1252', 'UTF-16', $text); // with BOM
iconv('CP1252', 'UTF-16BE', $text); // without BOM

Here is how to convert UCS-2 numbers to UTF-8 numbers in hex:

<?php
function ucs2toutf8($str)
{
        for ($i=0;$i<strlen($str);$i+=4)
        {
                $substring1 = $str[$i].$str[$i+1];
                $substring2 = $str[$i+2].$str[$i+3];
 
                if ($substring1 == "00")
                {
                        $byte1 = "";
                        $byte2 = $substring2;
                }
                else
                {
                        $substring = $substring1.$substring2;
                        $byte1 = dechex(192+(hexdec($substring)/64));
                        $byte2 = dechex(128+(hexdec($substring)%64));
                }
                $utf8 .= $byte1.$byte2;
        }
        return $utf8;
}
 
echo strtoupper(ucs2toutf8("06450631062D0020"));

?>

Input:
06450631062D
Output:
D985D8B1D8AD

regards,
Ziyad

Like many other people, I have encountered massive problems when using iconv() to convert between encodings (from UTF-8 to ISO-8859-15 in my case), especially on large strings.

The main problem here is that when your string contains illegal UTF-8 characters, there is no really straight forward way to handle those. iconv() simply (and silently!) terminates the string when encountering the problematic characters (also if using //IGNORE), returning a clipped string. The

<?php

$newstring = html_entity_decode(htmlentities($oldstring, ENT_QUOTES, 'UTF-8'), ENT_QUOTES , 'ISO-8859-15');

?>

workaround suggested here and elsewhere will also break when encountering illegal characters, at least dropping a useful note ("htmlentities(): Invalid multibyte sequence in argument in...")

I have found a lot of hints, suggestions and alternative methods (it's scary and in my opinion no good sign how many ways PHP natively provides to convert the encoding of strings), but none of them really worked, except for this one:

<?php

$newstring = mb_convert_encoding($oldstring, 'ISO-8859-15', 'UTF-8');

?>

iconv with //IGNORE works as expected: it will skip the character if this one does not exist in the $out_charset encoding.

If a character is missing from the $in_charset encoding (eg byte \x81 from CP1252 encoding), then iconv will return an error, whether with //IGNORE or not.

If you are getting question-marks in your iconv output when transliterating, be sure to 'setlocale' to something your system supports.

Some PHP CMS's will default setlocale to 'C', this can be a problem.

use the "locale" command to find out a list..

$ locale -a
C
en_AU.utf8
POSIX

<?php
  setlocale(LC_CTYPE, 'en_AU.utf8');
  $str = iconv('UTF-8', 'ASCII//TRANSLIT', "Côte d'Ivoire");
?>

Be aware that iconv in PHP uses system implementations of locales and languages, what works under linux, normally doesn't in windows.

Also, you may notice that recent versions of linux (debian, ubuntu, centos, etc) the //TRANSLIT option doesn't work. since most distros doesn't include the intl packages (example: php5-intl and icuxx (where xx is a number) in debian) by default. And this because the intl package conflicts with another package needed for international DNS resolution.

Problem is that configuration is dependent of the sysadmin of the machine where you're hosted, so iconv is pretty much useless by default,  depending on what configuration is used by your distro or the machine's admin.

function detectUTF8($string)
{
        return preg_match('%(?:
        [\xC2-\xDF][\x80-\xBF]        # non-overlong 2-byte
        |\xE0[\xA0-\xBF][\x80-\xBF]               # excluding overlongs
        |[\xE1-\xEC\xEE\xEF][\x80-\xBF]{2}      # straight 3-byte
        |\xED[\x80-\x9F][\x80-\xBF]               # excluding surrogates
        |\xF0[\x90-\xBF][\x80-\xBF]{2}    # planes 1-3
        |[\xF1-\xF3][\x80-\xBF]{3}                  # planes 4-15
        |\xF4[\x80-\x8F][\x80-\xBF]{2}    # plane 16
        )+%xs', $string);
}

function cp1251_utf8( $sInput )
{
    $sOutput = "";

    for ( $i = 0; $i < strlen( $sInput ); $i++ )
    {
        $iAscii = ord( $sInput[$i] );

        if ( $iAscii >= 192 && $iAscii <= 255 )
            $sOutput .=  "&#".( 1040 + ( $iAscii - 192 ) ).";";
        else if ( $iAscii == 168 )
            $sOutput .= "&#".( 1025 ).";";
        else if ( $iAscii == 184 )
            $sOutput .= "&#".( 1105 ).";";
        else
            $sOutput .= $sInput[$i];
    }
    
    return $sOutput;
}

function encoding($string){
    if (function_exists('iconv')) {    
        if (@!iconv('utf-8', 'cp1251', $string)) {
            $string = iconv('cp1251', 'utf-8', $string);
        }
        return $string;
    } else {
        if (detectUTF8($string)) {
            return $string;        
        } else {
            return cp1251_utf8($string);
        }
    }
}
echo encoding($string);

Here is an example how to convert windows-1251 (windows) or cp1251(Linux/Unix) encoded string to UTF-8 encoding.

<?php
function cp1251_utf8( $sInput )
{
    $sOutput = "";

    for ( $i = 0; $i < strlen( $sInput ); $i++ )
    {
        $iAscii = ord( $sInput[$i] );

        if ( $iAscii >= 192 && $iAscii <= 255 )
            $sOutput .=  "&#".( 1040 + ( $iAscii - 192 ) ).";";
        else if ( $iAscii == 168 )
            $sOutput .= "&#".( 1025 ).";";
        else if ( $iAscii == 184 )
            $sOutput .= "&#".( 1105 ).";";
        else
            $sOutput .= $sInput[$i];
    }
    
    return $sOutput;
}
?>

In my case, I had to change:
<?php
setlocale(LC_CTYPE, 'cs_CZ');
?>
to 
<?php
setlocale(LC_CTYPE, 'cs_CZ.UTF-8');
?>
Otherwise it returns question marks.

When I asked my linux for locale (by locale command) it returns "cs_CZ.UTF-8", so there is maybe correlation between it.

iconv (GNU libc) 2.6.1
glibc 2.3.6

There may be situations when a new version of a web site, all in UTF-8, has to display some old data remaining in the database with ISO-8859-1 accents. The problem is iconv("ISO-8859-1", "UTF-8", $string) should not be applied if $string is already UTF-8 encoded.

I use this function that does'nt need any extension :

function convert_utf8( $string ) { 
    if ( strlen(utf8_decode($string)) == strlen($string) ) {   
        // $string is not UTF-8
        return iconv("ISO-8859-1", "UTF-8", $string);
    } else {
        // already UTF-8
        return $string;
    }
}

 I have not tested it extensively, hope it may help.

I have used iconv to convert from cp1251 into UTF-8. I spent a day to investigate why a string with Russian capital 'Р' (sounds similar to 'r') at the end cannot be inserted into a database.

The problem is not in iconv. But 'Р' in cp1251 is chr(208) and 'Р' in UTF-8 is chr(208).chr(106). chr(106) is one of the space symbol which match '\s' in regex. So, it can be taken by a greedy '+' or '*' operator. In that case, you loose 'Р' in your string.

For example, 'ГР   ' (Russian, UTF-8). Function preg_match. Regex is '(.+?)[\s]*'. Then '(.+?)' matches 'Г'.chr(208) and '[\s]*' matches chr(106).'   '.

Although, it is not a bug of iconv, but it looks like it very much. That's why I put this comment here.

ANSI = Windows-1252 = CP1252
So UTF-8 -> ANSI:

<?php
      $string = "Winkel γ=200 für 1€"; //"γ"=HTML:&gamma;
      $result = iconv('UTF-8', 'CP1252//IGNORE', $string);
      echo $result;
?>

Note1
<?php
      $string = "Winkel γ=200 für 1€";
      $result = iconv('UTF-8', 'CP1252', $string);
      echo $result; //"conv(): Detected an illegal character in input string"
?>

Note2 (ANSI is better than decode in ISO 8859-1 (ISO-8859-1==Latin-1)
<?php
      $string = "Winkel γ=200 für 1€";
      $result = utf8_decode($string);
      echo $result; //"Winkel ?=200 für 1?"
?>

Note3 of used languages on Websites:
93.0% = UTF-8;
3.5% = Latin-1; 
0.6% = ANSI <----- you shoud use (or utf-8 if your page is in Chinese or has Maths)

Didn't know its a feature or not but its works for me (PHP 5.0.4)

iconv('', 'UTF-8', $str)

test it to convert from windows-1251 (stored in DB) to UTF-8 (which i use for web pages).
BTW i convert each array i fetch from DB with array_walk_recursive...

Provided that there is no invalid code point in the character chain for the input encoding, the //IGNORE option works as expected. No bug here.

As orrd101 said, there is a bug with //IGNORE in recent PHP versions (we use 5.6.5) where we couldn't convert some strings (i.e. "∙" from UTF8 to CP1251 with //IGNORE). 
But we have found a workaround and now we use both //TRANSLIT and //IGNORE flags:
$text="∙";
iconv("UTF8", "CP1251//TRANSLIT//IGNORE", $text);

Here is a code to convert ISO 8859-1 to UTF-8 and vice versa without using iconv.

<?php
//Logic from http://twiki.org/cgi-bin/view/Codev/InternationalisationUTF8
$str_iso8859_1 = 'foo in ISO 8859-1';
//ISO 8859-1 to UTF-8
$str_utf8 = preg_replace("/([\x80-\xFF])/e", 
            "chr(0xC0|ord('\\1')>>6).chr(0x80|ord('\\1')&0x3F)",
             $str_iso8859_1);
//UTF-8 to ISO 8859-1
$str_iso8859_1 = preg_replace("/([\xC2\xC3])([\x80-\xBF])/e", 
                "chr(ord('\\1')<<6&0xC0|ord('\\2')&0x3F)",
                 $str_utf8);
?>

HTH,
R. Rajesh Jeba Anbiah

Here is how to convert UTF-8 numbers to UCS-2 numbers in hex:

<?php
 
function utf8toucs2($str)
{
       for ($i=0;$i<strlen($str);$i+=2)
       {
                $substring1 = $str[$i].$str[$i+1];  
                $substring2 = $str[$i+2].$str[$i+3]; 
               
                if (hexdec($substring1) < 127)
                        $results = "00".$str[$i].$str[$i+1];
                else
                {
                        $results = dechex((hexdec($substring1)-192)*64 + (hexdec($substring2)-128));
                        if ($results < 1000) $results = "0".$results; 
                        $i+=2;
                }
                $ucs2 .= $results;
        }
        return $ucs2;
}
 
echo strtoupper(utf8toucs2("D985D8B1D8AD"))."\n";
echo strtoupper(utf8toucs2("456725"))."\n";
 
?>

Input:
D985D8B1D8AD
Output:
06450631062D

Input:
456725
Output:
004500670025

Note an important difference between iconv() and mb_convert_encoding() - if you're working with strings, as opposed to files, you most likely want mb_convert_encoding() and not iconv(), because iconv() will add a byte-order marker to the beginning of (for example) a UTF-32 string when converting from e.g. ISO-8859-1, which can throw off all your subsequent calculations and operations on the resulting string.

In other words, iconv() appears to be intended for use when converting the contents of files - whereas mb_convert_encoding() is intended for use when juggling strings internally, e.g. strings that aren't being read/written to/from files, but exchanged with some other media.

On some systems there may be no such function as iconv(); this is due to the following reason: a constant is defined named `iconv` with the value `libiconv`. So, the string PHP_FUNCTION(iconv) transforms to PHP_FUNCTION(libiconv), and you have to call libiconv() function instead of iconv().
I had seen this on FreeBSD, but I am sure that was a rather special build.
If you'd want not to be dependent on this behaviour, add the following to your script:
<?php
if (!function_exists('iconv') && function_exists('libiconv')) {
    function iconv($input_encoding, $output_encoding, $string) {
        return libiconv($input_encoding, $output_encoding, $string);
    }
}
?>
Thanks to tony2001 at phpclub.net for explaining this behaviour.

To strip bogus characters from your input (such as data from an unsanitized or other source which you can't trust to necessarily give you strings encoded according to their advertised encoding set), use the same character set as both the input and the output, with //IGNORE on the output charcter set.
<?php
// assuming '†' is actually UTF8, htmlentities will assume it's iso-8859  
// since we did not specify in the 3rd argument of htmlentities.
// This generates "&acirc;[bad utf-8 character]"
// If passed to any libxml, it will generate a fatal error.
$badUTF8 = htmlentities('†');

// iconv() can ignore characters which cannot be encoded in the target character set
$goodUTF8 = iconv("utf-8", "utf-8//IGNORE", $badUTF8);
?>
The result of the example does not give you back the dagger character which was the original input (it got lost when htmlentities was misused to encode it incorrectly, though this is common from people not accustomed to dealing with extended character sets), but it does at least give you data which is sane in your target character set.

I just found out today that the Windows and *NIX versions of PHP use different iconv libraries and are not very consistent with each other.

Here is a repost of my earlier code that now works on more systems. It converts as much as possible and replaces the rest with question marks:

<?php
if (!function_exists('utf8_to_ascii')) {
  setlocale(LC_CTYPE, 'en_AU.utf8');
  if (@iconv("UTF-8", "ASCII//IGNORE//TRANSLIT", 'é') === false) {
    // PHP is probably using the glibc library (*NIX)
    function utf8_to_ascii($text) {
      return iconv("UTF-8", "ASCII//TRANSLIT", $text);
    }
  }
  else {
    // PHP is probably using the libiconv library (Windows)
    function utf8_to_ascii($text) {
      if (is_string($text)) {
        // Includes combinations of characters that present as a single glyph
        $text = preg_replace_callback('/\X/u', __FUNCTION__, $text);
      }
      elseif (is_array($text) && count($text) == 1 && is_string($text[0])) {
        // IGNORE characters that can't be TRANSLITerated to ASCII
        $text = iconv("UTF-8", "ASCII//IGNORE//TRANSLIT", $text[0]);
        // The documentation says that iconv() returns false on failure but it returns ''
        if ($text === '' || !is_string($text)) {
          $text = '?';
        }
        elseif (preg_match('/\w/', $text)) {        // If the text contains any letters...
          $text = preg_replace('/\W+/', '', $text); // ...then remove all non-letters
        }
      }
      else {  // $text was not a string
        $text = '';
      }
      return $text;
    }
  }
}

<?php
//script from http://zizi.kxup.com/
//javascript unesape
function unescape($str) {
  $str = rawurldecode($str);
  preg_match_all("/(?:%u.{4})|&#x.{4};|&#\d+;|.+/U",$str,$r);
  $ar = $r[0];
print_r($ar);
  foreach($ar as $k=>$v) {
    if(substr($v,0,2) == "%u")
      $ar[$k] = iconv("UCS-2","UTF-8",pack("H4",substr($v,-4)));
    elseif(substr($v,0,3) == "&#x")
      $ar[$k] = iconv("UCS-2","UTF-8",pack("H4",substr($v,3,-1)));
    elseif(substr($v,0,2) == "&#") {
echo substr($v,2,-1)."<br>";
      $ar[$k] = iconv("UCS-2","UTF-8",pack("n",substr($v,2,-1)));
    }
  }
  return join("",$ar);
}
?>

You can use native iconv in Linux via passthru if all else failed.
Use the -c parameter to suppress error messages.

When doing transliteration, you have to make sure that your LC_COLLATE is properly set, otherwise the default POSIX will be used.

To transform "rené" into "rene" we could use the following code snippet:

<?php

setlocale(LC_CTYPE, 'nl_BE.utf8');

$string = 'rené';
$string = iconv('UTF-8', 'ASCII//TRANSLIT', $string);

echo $string; // outputs rene

?>

mirek code, dated 16-May-2008 10:17, added the characters `^~'" to the output.
This function will strip out these extra characters:
<?php
setlocale(LC_ALL, 'en_US.UTF8');
function clearUTF($s)
{
    $r = '';
    $s1 = @iconv('UTF-8', 'ASCII//TRANSLIT', $s);
    $j = 0;
    for ($i = 0; $i < strlen($s1); $i++) {
        $ch1 = $s1[$i];
        $ch2 = @mb_substr($s, $j++, 1, 'UTF-8');
        if (strstr('`^~\'"', $ch1) !== false) {
            if ($ch1 <> $ch2) {
                --$j;
                continue;
            }
        }
        $r .= ($ch1=='?') ? $ch2 : $ch1;
    }
    return $r;
}
?>

You can use 'CP1252' instead of 'Windows-1252':
<?php
// These two lines are equivalent
$result = iconv('Windows-1252', 'UTF-8', $string);
$result = iconv('CP1252', 'UTF-8', $string);
?>
Note: The following code points are not valid in CP1252 and will cause errors.
129 (0x81)
141 (0x8D)
143 (0x8F)
144 (0x90)
157 (0x9D)
Use the following instead:
<?php
// Remove invalid code points, convert everything else
$result = iconv('CP1252', 'UTF-8//IGNORE', $string);
?>

iconv also support CP850.
I used iconv("CP850", "UTF-8//TRANSLIT", $var);
to convert from SQL_Latin1_General_CP850_CI_AI to UTF-8.

The following are Microsoft encodings that are based on ISO-8859 but with the addition of those stupid control characters.

CP1250 is Eastern European (not ISO-8859-2)
CP1251 is Cyrillic (not ISO-8859-5)
CP1252 is Western European (not ISO-8859-1)
CP1253 is Greek (not ISO-8859-7)
CP1254 is Turkish (not ISO-8859-9)
CP1255 is Hebrew (not ISO-8859-8)
CP1256 is Arabic (not ISO-8859-6)
CP1257 is Baltic (not ISO-8859-4)

If you know you're getting input from a Windows machine with those encodings, use one of these as a parameter to iconv.

For text with special characters such as (é) é which appears at 0xE9 in the ISO-8859-1 and at 0x82 in IBM-850. The correct output character set is 'IBM850' as:
('ISO-8859-1', 'IBM850', 'Québec')

On my system, according to tests, and also as reported by other people elsewhere, you can combine TRANSLIT and IGNORE only by appending

//IGNORE//TRANSLIT

strictly in that order, but NOT by appending //TRANSLIT//IGNORE, which would lead to //IGNORE being ignored ( :) ).

Anyway, it's hard to understand how one could devise a system of passing options that does not allow to couple both options in a neat manner, and also to understand why the default behaviour should be the less useful and most dangerous one (throwing away most of your data at the first unexpected character). Software design FAIL :-/

So, as iconv() does not always work correctly, in most cases, much easier to use htmlentities(). 
Example: <?php $content=htmlentities(file_get_contents("incoming.txt"), ENT_QUOTES, "Windows-1252");  file_put_contents("outbound.txt", html_entity_decode($content, ENT_QUOTES , "utf-8")); ?>

If you need to strip as many national characters from UTF-8 as possible and keep the rest of input unchanged (i.e. convert whatever can be converted to ASCII and leave the rest), you can do it like this:

<?php
setlocale(LC_ALL, 'en_US.UTF8');

function clearUTF($s)
{
    $r = '';
    $s1 = iconv('UTF-8', 'ASCII//TRANSLIT', $s);
    for ($i = 0; $i < strlen($s1); $i++)
    {
        $ch1 = $s1[$i];
        $ch2 = mb_substr($s, $i, 1);

        $r .= $ch1=='?'?$ch2:$ch1;
    }
    return $r;
}

echo clearUTF('Šíleně žluťoučký Vašek úpěl olol! This will remain untranslated: ᾡᾧῘઍિ૮');
//outputs Silene zlutoucky Vasek upel olol! This will remain untranslated: ᾡᾧῘઍિ૮
?>

Just remember you HAVE TO set locale to some unicode encoding to make iconv handle //TRANSLIT correctly!

If you want to normalize a filename on Mac OS X, because it is in UTF-8 NFD and you need UTF-8 NFC
(See: http://en.wikipedia.org/wiki/Unicode_equivalence#Combining_and_precomposed_characters)
you may use:
<?php
$filename_nfc = iconv("UTF-8-MAC", "UTF-8", $filename_nfd);
?>

For those who have troubles in displaying UCS-2 data on browser, here's a simple function that convert ucs2 to html unicode entities :

<?php

 function ucs2html($str) {
    $str=trim($str); // if you are reading from file
    $len=strlen($str);
    $html='';
    for($i=0;$i<$len;$i+=2)
        $html.='&#'.hexdec(dechex(ord($str[$i+1])).
                   sprintf("%02s",dechex(ord($str[$i])))).';';
    return($html);
 }
?>

For transcoding values in an Excel generated CSV the following seems to work:

<?php 
$value = iconv('Windows-1252', 'UTF-8//TRANSLIT', $value);
?>