Category Archives: TV

Subtitling tries to make me dumber

I was watching Todd Sampson‘s Redesign My brain S1E1 Make Me Smarter and noticed the subtitling was annoyingly wrong. FMRI was subtitled as MRI. Baseline became based on – and there was more errors. My hearing’s not super-great, but even I could tell that these weren’t right.

Twice I’ve seen subtitling so bad that I’ve been prompted to find out who did it. Last time it was Jacqui Mapoon at CSI.

This time it was Jacqui Mapoon at CSI. Either Jacqui does a lot of work for CSI and sometimes has bad days, or she does a little work and often screws it up. What are the odds that on the two occasions I notice very bad subtitling, the same person’s behind it?  Subtitling is a very specialized field, so there can’t be that many people doing it, but at the same time a lot of TV is subtitled. I know from personal experience that subtitling takes at least 5 minutes per minute of show, and can take more if it’s particularly speech-heavy. There are a few hours of TV a night requiring subtitles, and it’d take one person one day to subtitle one hour of TV, so there’s probably a few dozen people in Australia doing it; live subtitling is a different specialty. Perhaps work processes need to be changed; I know I proof my subtitling after having done it, and spot errors. Perhaps someone other than the original subtitler ought to do the final proofing? Proof-reading error rates would show whose work needed more attention.

Most of the subtitles that I’ve seen are great – precisely timed transcriptions of the spoken dialogue, either exact reproductions or well thought through précis, contracted just enough to be faithful to the words and the intent whilst also fitting on the screen. For some reason American stuff is all caps unless the character is off-screen. Given so many in this industry can get it transparently right, why does one person’s work repeatedly poke me in the eye? Somebody give Jacqui some training, stat!

Replace a missing remote control with an Arduino and a laptop

I recently found myself without a remote for my WDTV Live media player, and limited resources to do anything about it – but I did have an Arduino, a breadboard and the local Jaycar had an IR LED.  Controlling IR devices is common practice with an Arduino. I would even be able to hack in functions that didn’t exist on the manufacturer’s remote – like creating a three minute skip by switching to 16x speed for 12 seconds.

The first port of call was to obtain Ken Shirriff’s Arduino IR remote control protocol library – as opposed to communications protocols, of which there are quite a number; did you know the first cut of WiFi included an infrared version? Without the remote, I wasn’t able to record and playback the IR signals sent to the WDTVLive, as you would with a learning remote. I had to find what to transmit from my custom remote. I little googling and I found the WD TV Live infrared remote control codes, which also helpfully reveals that the protocol is NEC.

I knocked up a quick proof of concept, installed it and watched it not work. Given I can’t see in infrared, I didn’t know if my circuit was working. I hooked a red LED up in parallel, and it didn’t light up; I thought I had cathode and anode swapped around, so flipped the red LED – and it didn’t light up. I pulled the IR LED, and then the red LED worked… I was shorting out the red LED. I couldn’t – with the bits I had lying around – confirm the device was transmitting anything. Rather than put the LEDs in series, I got a cheap camera-phone with video function, and it could see IR just fine. And it turns out the IR LED was transmitting something, but the WD TV Live media player wasn’t listening. Why?

The NEC infrared control protocol transmits 32 bits in one of two formats, one old (as in elderly) format encodes for 256 devices with 256 commands each, and the other encodes for ~64K devices with 256 commands each. The first 16 bits encode the device, and the second 16 bits encode the command. 16 bits for one of 256 commands, you ask? Well, one byte of the second 16 bits is the command, and the other is – for error checking purposes – the one’s complement of that. Further details of the pulse timing and protocol contents are available in various places, but they neglect to mention the extended addressing format. There are many IR control protocols. To use Ken’s IR library you need to know which protocol is used (which the google search revealed), and you can determine the protocol from the timing data found in the LIRC definition of a protocol, in this case the LIRC infrared control protocol for WDTV Live media player remote. The LIRC protocol defintion format is described by WinLIRC, so you can see what the timings are. In this case, the NEC protocol is revealed by the header, one and zero definitions, along with the fact that each code has 16 bits of ‘pre-data’ and 16 bits of data (a 32 bit package). Everything I could see was showing that the two, separately arrived at sets of command codes that were empirically sampled from the real world were compliant with the spec. One of the things the spec taught me was to transmit the NEC code twice, and to wait 70ms between re-transmissions.

I wasted time finding other codes for the remote, in other formats; I checked for byte ordering issues. Nothing worked.

The actual problem was the unsigned long for the command was previously an int; failing to notice this simple error led me to spend a long time trying to figure out why nothing was happening when I transmitted a command. One of the problems with the C language is the guarantees about data sizes aren’t worth much.  My entire life has been spent programming on architectures that have 32 bit data words; C compilers on these machines have all defined an int as 32 bits, but I’ve always been aware that the language spec says that an int is at least as wide as a short, which is at least as wide as a char with actual widths being up to the compiler implementation (although why you’d have different words for things of the same size is beyond me).  The AVR microcontroller in question has an 8 bit word; mathematical operands typically yield an 8 bit result (multiply is an exception) with compilers needing to implement more instructions to yield greater data widths. The defines express the codes as four byte values, which were then wrangled into a two byte int, and then again into unsigned four byte integer when passed to the IR library. Truncated bits in a protocol like this were the cause of inactivity.

Even with this fundamental problem solved, confusion was added by the fact that one of the memory cells in my Arduino is faulty. Once IR control code transmission was working, I noticed that sometimes it didn’t work. I decided to echo the command to the serial port, and the command being transmitted didn’t match that for the key pressed – the second byte was wrong. I added code to work around this memory corruption (not shown in the code below, because this is a pretty unusual). I’ve never come across this kind of problem before, recognising and then solving something like that is pretty old-school.

/*
Pin 3 is hard-wired into the IR library as the emitter
 */
#include <IRremote.h>
//#define DEBUG

IRsend irsend;

#define btn_enter  0x219E10EF
#define btn_right  0x219E906F
#define btn_left   0x219EE01F
#define btn_down   0x219E00FF
#define btn_up     0x219EA05F
#define btn_option 0x219E58A7
#define btn_back   0x219ED827
#define btn_stop   0x219E20DF
#define btn_rew    0x219EF807
#define btn_ff     0x219E7887
#define btn_play   0x219E50AF
#define btn_prev   0x219E40BF
#define btn_next   0x219E807F
#define btn_eject  0x219E08F7
#define btn_search 0x219EF00F
#define btn_home   0x219E609F
#define btn_power  0x219E48B7

// Pin 13 has an LED connected on most Arduino boards.
// give it a name:
const int onboard_led = 13;
const int retransmit=2;
unsigned long play_after=0;

void setup()
{
  pinMode(3, OUTPUT);     
  pinMode(onboard_led, OUTPUT);     
  Serial.begin(9600);
  Serial.println("WDTV Live serial controlled IR remote");
  Serial.println("~ Power    Eject ^ & Search   Rew - + FF");
  Serial.println("  w         Back q e Enter   Play  P");
  Serial.println("a s d (Arrows)     x Stop    Last < > Next");
  Serial.println("3 - FastForward three minutes");
}

void loop() {
  unsigned long cmd=0;
  if (Serial.available()) {
    switch (Serial.read()) {
      case 'E':
      case 'e':
      case ')':
      case '0':
      case 'O':
      case 'o': cmd=btn_enter; break;
      case 'q':
      case 'Q': cmd=btn_back; break;
      case 'P':
      case 'p':
      case ' ': cmd=btn_play; break;
      case 'S':
      case 's': cmd=btn_down; break;
      case 'W':
      case 'w': cmd=btn_up; break;
      case 'A':
      case 'a': cmd=btn_left; break;
      case 'D':
      case 'd': cmd=btn_right; break;
      case '-':
      case '_': cmd=btn_rew; break;
      case '=':
      case '+': cmd=btn_ff; break;
      case ',':
      case '< ': cmd=btn_prev; break;
      case '.':       
      case '>': cmd=btn_next; break;
      case '/':
      case '?': cmd=btn_option; break;
      case '~': cmd=btn_power; break;
      case '!':
      case '1': cmd=btn_home; break;
      case '^':
      case '6': cmd=btn_eject; break;
      case '*':
      case '8': cmd=btn_search; break;
      case 'x':
      case 'X': cmd=btn_stop; break;
      case '3': 
        if (!play_after) play_after=4; break;
    }
  }
  if (play_after > 0) {
    if (cmd) {
      play_after=0;
    }
    else if (play_after > 5) {
      if (play_after < millis()) {
        cmd=btn_play;
        play_after=0;
      }
    }
    else {
      cmd=btn_ff;
      if (--play_after == 0) {
        play_after=millis()+12000;
      }
    }
  }
  if (cmd) {
    digitalWrite(onboard_led, HIGH);   // turn the LED on to indicate activity
    for (int i = 0; i < retransmit; i++) {
      irsend.sendNEC(cmd, 32);
      delay(70);
    }
#ifdef DEBUG
    Serial.println(cmd, HEX);
#endif
    digitalWrite(onboard_led, LOW);    // turn the LED off - we're done transmitting
  }
}

In other links, How-To: IR Remote Control your Computer

"Spaced" DVDs – volume problems

I've been re-watching the DVDs of Simon Pegg/Edgar Wright sitcom Spaced, and noticing that the volume levels go up and down all the time. I didn't remember that being the case from the first time I watched them… which was on a previous DVD player (a Pioneer).

Turns out I'm not the only one to have this problem — those reporting there noted it was an issue on some Toshiba players, but I'm currently using a Sony.

I also have a Panasonic Blu-Ray player I could use, but it's not multi-region, and these discs are Region 2. Damn. And I can't see any easy way of hacking the player to make it multi-region, alas.

That's okay though, because taking a cue from the forum led me to this post about DVD sound problems, and also to a Wikipedia description of Dynamic Range Compression — leading me to think this was causing the issue.

I found this in the DVD player's setup menu. Once I'd switched it from “TV mode” to “standard”, all seemed to be okay again, though in contrast, a newer DVD of Parks & Recreation now seems to have its volume fluctuate, so perhaps I need to switch it back for everything other than Spaced. Odd.

Subtitling now in the Suez

I imagine that Jacqui Mapoon from CSI (Captioning and Subtitling Australia… or International) is someone who helps them out very occasionally, judging by the quality of her work on The Doctor Blake Mysteries: Season 1, Episode 9 “All That Glitters” –  atrocious work.  The gaff that stood out most was the transcription of sewers, but there were so many problems.

Don’t these subtitling services get given the script?  There’s a job listed in the credits as “Post production scripts”, surely they’re able to hand the script over electronically, and it’s just a matter of timing, pagination and confirmation – no transcription, no transcription errors?  I doubt Tim Pye – the writer of the episode – would have got that wrong, nor made the other homophonic errors.

How to fix YourTV.com.au’s annoying Sydney default

I quite like the YourTV.com.au web site. The TV guide it displays is quite usable, and can be customised to show your correct channels.

But why does it keep forgetting your region every few weeks, and reset itself to metro Sydney?

Your TV Sydney default

Very irritating. (Well, if you live outside metro Sydney.)

Using your web browser, you can check the cookies. This article describes how, in various browsers.

That’s where the problem is: it looks like the “TvFixGuide” cookie, which seems to hold details of what region you’re in, is only set for a month.

Your TV cookie

It doesn’t look like either browser allows you to extend the time range of the cookie, or otherwise modify it. I suppose there’s legitimate reasons for that.

It is possible to hack it by deleting the cookie, setting your computer’s clock, say, a year into the future, before going back to the site and setting the option.

Yep, it seems to work:

Your TV cookie modified

Don’t forget to set your clock back afterwards.

USA news clips don’t export well

News clips from the USA are often 4:3 – why?  Is it a technologically backward country, or do they only export their news in a universal format?  I see work-arounds to disguise this fact, like framing the whole clip in a themed border, or widening the clip by tacking onto the pillar-box sides a blurred-out duplicate of the clip that’s been zoomed, stretched or clipped.

And while I’m noting this, why is it that the watermark on these clips is almost universally blurred out and overlayed with the local broadcaster’s watermark – can’t they get the raw, unwatermarked footage from the provider?

The spinning globe

I’ve long been a fan of TV idents, and I used to love seeing the late-80s BBC1 globe animation, when it occasionally popped-up on television here. As well as the Alas Smith and Jones spoof version.

Here are a couple of fascinating articles on how it was generated: by a standalone computer, which animated the 12 second rotation, at the PAL standard of 25 frames per second.

(Update: Found a better video)

Bye bye Teletext

In Australia the 7 Network’s Austext teletext service is to shut down at the end of September, with only Supertext subtitles/Closed captioning continuing to be transmitted. (I wonder if they’ll move them from the current page 801 to the default 100, to make them easier to use?)

Over in the UK, ITV’s Teletext service is to shut down in January 2010.

The BBC’s Ceefax will last until analogue TV is switched off in 2012.

Hardly surprising really. I’m sure demand for text-based news and other information has plummetted since the widespread adoption of the Web. In fact I’m surprised teletext has lasted this long — I struggle to think of anybody I know who uses it.

YouTube goes widescreen

Youtube wide with 4:3 contentYouTube has gone widescreen (note: this link currently breaks if your YouTube preference is for a non-US locale, eg for Aussies you end up here, which currently displays nothing).

The only catch of course is that 4:3 videos now appear letterboxed… or whatever the vertical term for letterboxed is.

My question is: why? Why not just make the player (at least on the YouTube web site) the aspect ratio of the video that it's playing?

In fact at the moment, embedded 16:9 videos still appear letterboxed; 4:3 videos “full screen” … what should happen is that the embedding code should define the player size and so match the video's aspect ratio.

Surely it can't be that hard to avoid those black bands?