PHP: image corrupted if I force the download - php

Question

I've a strange behavior with a simple PHP code. When I try to force the download or print out the image using the correct content-type, the output file is corrupted.
Seems that the webserver (apache) adds two bytes (0x20 and 0x0A) at the begin of the file.
This is the code:
$file = "image.png";
$image = file_get_contents($file);
// Test
file_put_contents("test.png", $image);
// Download
header('Content-Description: File Transfer');
header('Content-Type: application/octet-stream');
header('Content-Disposition: attachment; filename='.basename($file));
echo $image;
I use the same code on other websites hosted on the same server without problems.
The problem is only on download only because test.png works properly. The MD5 checksum of text.png and the original image are equals.
This is the hex code of test.png.
And this is the hex code of the corrupted file after download:
As you can see, there are 2 extra bytes at the begin. If I remove them, the file returns to work properly.
I attach the screen of Wireshark (as you can see is not a browser issue):
How can I fix it?
The server is Ubuntu 16.04 with PHP-5.6 (yes I done the downgrade from 7.0 to 5.6 for compatibility issues with roundcube)
UPDATE 1: I'm trying to find if somewhere in the file there is a space + newline
UPDATE 2:
First of all: thanks.
The code is part of a Wordpress plugin and the download is called using the AJAX system. I wrote a simple plugin test:
<?php
/*
Plugin Name: Test
Plugin URI: http://www.google.com
Description: Test
Author: Anon
Version: 4.0
*/
function downlod_test() {
echo "test";
header('Content-Description: File Transfer');
header('Content-Type: application/octet-stream');
header('Content-Disposition: attachment; filename=prova.html');
die();
}
function iopman_shared_download_doc_ajax() {
downlod_test();
}
add_action('wp_ajax_frontend_download_doc', 'iopman_shared_download_doc_ajax');
//downlod_test();
?>
If I call downlod_test with /wp-admin/admin-ajax.php?action=frontend_download_doc it adds the 2 extra bytes. If I call it directly (by removing the comments), it works.
So the problem now is: how to strip out these bytes that wordpress adds?

Answer 1

$file = "image.png";
header('Content-Description: File Transfer');
header('Content-Type: application/octet-stream');
header('Content-Disposition: attachment; filename=' . basename($file));
header("Content-Encoding: gzip");
header('Expires: 0');
header('Cache-Control: must-revalidate');
header('Pragma: public');
header("Content-Length: " . filesize($file));
header('Content-Transfer-Encoding: binary');
header('Connection: Keep-Alive');
header('Expires: 0');
header('Cache-Control: must-revalidate, post-check=0, pre-check=0');
ob_get_clean();
readfile($file);
exit;

Answer 2

To help you find that unwanted whitespace you can track loaded files with get_included_files(). Additionally, a backtrace could also shred some light on what your script does.
In many cases, it'll come from closing PHP tags at the end of the file. Since they're optional it's recommended to just not use them.
Once you locate the file where that white space is, you only need to load in your favourite text editor and remove them (you might need to enable your editor's Show hidden chars feature).
P.S. I understand that's probably simplified code to illustrate the issue but you may want to give readfile() a try.