First of all, I’ve now completed two years of my undergraduate degree in linguistics. I’ve completed the normal Chinese language instruction stream, or at least as far as I’m willing to take it at UniMelb. Now it’s literature based subjects which I’m actually good at and is much much better taught, so I’ll be feeling a lot better about Chinese in the coming year.

As I’ve said before, my direction has solidified towards the field of computational linguistics although I find myself in niche area that’s quite hard to pilot. I’m more interested in using computational linguistics as a tool within linguistics proper, rather than as a sort of vocationally-focused computer science dominated field of computational linguistics. Just after semester I attended another academic conference and I came away feeling much better about the prospects of finding a good path ahead.

Another major aspect I haven’t spoken about a lot on this blog is my revived interest in all things to do with ‘making’ or ‘hacking’. I have for some time felt pretty strongly that the consumer electronics boom has a lot to answer for in the atrophy of skills and the generally lesser inclination of your average man to get down to the shed and do something for himself. This comes from something of a perfect storm of events, mostly documented on this blog. Starting from the introduction to Python I had with the first computer subject I took as breadth. It might seem absurd, but I had forgotten how … capable (willing more than ability?) I am at just knocking things up out of the whole technical chain from web browser through to little things soldered onto a board. 2011 was the year of awakening of Mat as a Maker.

So at the very end of 2011, where am I at and what am I doing? It’s another couple of months until uni starts again so I’m firmly in the time-rich portion of the year where I get into projects. Some of my projects:

• The Inf0cube – A sort of kitchen-based explosion of old and new visualisation. Heavy electronics and coding project, probably the most impressive single thing I’ve ever built.
• Higgins – The Android-powered telepresence robot. Chassis works, Android hookup works, now crafting some genuinely innovative ways to use a dirty cheap and powerful smartphone as an all purpose robot brain.
•  Home Brew – I have brewed beer for years, on and off, but only as a ‘kit and kilo’ operator. Eg, you buy cans of stuff and just chuck it in a drum. Since I’m so poor now, I began brewing again just so I could remember the taste of beer. As my obsessive personality dictates, this wasn’t enough and I’ve progressed to ‘all grain’ brewing. I think this is a keeper, it’s easy, fun, lots of geek-out potential and everyone appreciates the results.
• Melbourne Mini Maker Faire and new Melbourne Hackerspace. I’m helping to organise this first maker faire event in Melbourne. It’s on something of a tight schedule. I’m doing things to do with writing about makers, bit of PR and marketing, and offering unwarranted opinions. The event is in January.

So that’s where I’m at. Now I should follow up by drilling into those subjects…

In the Android desktop shot of my Nexus One, you can see some cool Chinese written down the left side in a lovely glowing font. This is in fact the time written in Chinese, designed to be more aesthetic than practical. In order to do this I used three different components.

1. Minimal Text: A fantastic text widget that does some stuff like text-based clocks out of the box, but it’ll also accept locale variables.

2. SL4C and Python for Android: This is the actual script that generates the string of Chinese characters, writing it out as utf-8 to a file on the SD card.

3. Tasker: This fires every five minutes, runs the script and then reads two lines out of the file on SD card and loads them into variables. It then copies the variables to Locale variables.

The Tasker bit is a bit long winded but you can’t seem to write locale variables from SL4A. That said there’s a lot of possibilities to lash stuff up in Python and write things onto your Android screen in beautiful fonts. Note that the CJK font has been replaced with something much nicer than the Android stock font as I discussed in a previous post.

Incidentally the background is a live webcam of my back yard in the Dandenong ranges. The temperature display is from an external weather station and the light gadget toggles my wireless ‘infolights’, another project.

The Chinese keyboard plug-in for Smartkeyboard Pro IME is almost as good as Google’s Pinyin IME so that’s what I’ve been using and that’s what TouchPal is going up against.

TouchPal’s motion touch, for want of a better phrase, is still deeply wonderful. The idea is that you are able to type capital letters with a quick gesture upwards starting from the letter you selected. If this seems familiar, it’s because Swype does it. TouchPal did it first. Even more usefully, you can flick down to select a symbol. Most useful symbols are on the standard full-qwerty keyboard. Not only that but there’s genuinely useful buttons to go to other pages such as a web tab which enables you to quickly whack on .com or whatever. Since Swype actually seemed to go from amazing to crap in just a few updates, I’ve gone back to the old style peck-and-predict keyboard but TouchPal is at the pinnacle of this sort of thing, particularly with ease of symbol typing.

There’s also a nice ‘edit’ button which enables cursors, copy and pasting (faster than long-press), deleting and so on and a button for voice input, something I don’t use but you always see people moaning about the lack of it on keyboards in the Marketplace comments. There’s a smiley button which is populated with stupid Japanese smileys so is pretty useless I found – it wasn’t apparent to me how to just add some normal smileys but you’d have to think you could?

The prediction is extremely good, not just for English words but word-ahead in English and for full Chinese sentences. All in all, it’s a very functional, very polished, extremely effective bi-lingual keyboard which has usurped Smartkeyboard Pro for me.

That said it’s not without issues. The standard behaviour is to hit space to accept prediction entries. If you want another entry and click on it, it wont put a space after it, you need to do that yourself. This is counter-intuitive after so many keyboards will insert space after you make a prediction choice. There’s some odd stuff going on with updating. I grabbed the current version from the Marketplace and then clicked on the software update inside the app, at which point it grabbed a new version. Why isn’t the one on the Marketplace the newest?

Furthermore the new version looked substantially different keyboard wise than the one I installed from the Market and not particularly better. I also find tapping keys at the bottom of the keyboard sometimes ends up with undesirable results, closing the keyboard, backing out of the app. One can only assume because I’m touching the soft-keys on the Nexus One. I can’t really say, only that it’s a problem I’ve only really had on TouchPal.

You have a choice of where to download add-ons from, the Android market (which as we’ve seen is not the latest version), CooTek direct or something called the Hiapk market. I imagine these choices are necessitated by the need of providing access outside of the Android market. Some add-ons can be downloaded including a Chinese handwriting IME. I gave this a spin and it’s hands down the best one I’ve ever seen, it recognised my hanzi perfectly, has excellent prediction and is all-round pretty fabulous. It’s faster for me to type pinyin being a laowai but I will still find it useful in a dictionary by sketching a word I don’t know. Something I had to use the rather diabolical Android Hanzi Recogniser to do (or rather more likely pull out my iPod and Pleco).

The tutorial is quite focused on Chinese, the user manual is entirely in Chinese. It seems to have lost some of the complex motion touch features that the IME used to have on WinMo but likely to make it more usable, and I think it kind of works. Given it’s free and it’s actually my favourite Android keyboard, it’s pretty good going really. The focus on Chinese is welcome for me but probably a bit of a mistake in terms of marketability. The keyboard is good enough to go head to head with the best on the Marketplace English keyboards but likely people will be put off by the Asian focus. Oh well.

Anyway, welcome back CooTek. I’s nice to see an old friend return and that they’re still in excellent health.

Firstly, these guys love themselves. Rather than just put the damn thing in the market, for the last year we’ve had to mess around with a craptastic seperate installer app, frequently dig out old usernames and passwords for a beta to get reinvited to download the thing again. On top of that at one point the craptastic installer wouldn’t register the keyboard so it unhelpfully just whinged about not being the right device because it thought you had stolen it. Turns out the solution was a weird hack of pressing back, right at the end, rather than OK. Nice…

Then there’s that fucking pop-up. This is what makes me think they’re not even using their own keyboard. If you select a word that’s down the list of choices, because of a custom word you entered, Swype helpfully pops up this message to tell you what’s happened and tell you that you can delete the word if you like. That’s great, but it pops up every fucking time this happens. There’s no control to disable it. What’s more, it’s persisted through at least one update. This is just plain baffling.

Worse of all, however, is the fact that whatever was good about Swype they’ve just gotten rid of. Where once I used to be able to sketch out words and write away with minimal corrections, now every second bloody word needs to be corrected. Sometimes I’m just tapping the keys because I’m in a bloody hurry and can’t be arsed taking my chances on whether Swype will get it right again. I thought this was maybe me, but today I spoke to someone else who had exactly the same experience. Swype has basically gone shit.

All this time, you can’t buy it, the endless beta bullshit, the lame show stopping obvious bugs, and the virtually broken actual swyping, and we’re ready to call it a day. It’s frustrating because they recently added an actual swypable Chinese keyboard. It’s kind of nifty although they’ve not really thought through about how much of a time waster it is to disambiguate two things in a row for single characters. There’s also a handy way to flick to English, but the long press scroll through entire list of languages to get back to Chinese.

It’s actually easier to use the Google pinyin keyboard and switch to English, even without text prediction, than it is to have to switch entire languages. I switched to the standard Android keyboard earlier this morning and already I’m not going “oh for fuck sake!” at my phone at yet another Swypo. Thanks Swype, you were good when you were young but now you’ve let yourself go and you’ve become far too much hard work, I’m ready for the next big thing.

I’ve been happy with it, I could see it drew bugger all electricity and is bright as hell. It sent me down an avenue of exploring LED lighting more fully but after I constructed my first simple 20W LED grow bar, it was apparent that LEDs were a lot better than this device was showing.

While I’m doing a rebuild of the greenhouse, I thought I’d whip the case off to see what sort of LEDs it has. Ugh, no dice. The thing has been made roughly like someone would make a barbeque. Extensive use of silicon and sealant used as a mounting solution including mains electrical blocks…

What we can see is that there are three PSUs. One of which is a 12V unit that powers the three ‘temperature controlled fans’. Well, it was quite apparent there’s no temperature control but I knew that anyway – they howl all the time. Other than that we have two identical mains LED driver boxes which feed through holes in a circular aluminium plate where the LEDs are which I can’t get to. They went through an electrical block though so I could simply measure the voltage and remove a wire and put my meter in series to measure the current.

I got about 55v on each circuit at 400mA. The current figure is unusual because 1W LEDs, which they’re supposed to be, are generally 300mW. Since we have a total of 110v of DC potential and 90 LEDs, it’s apparent they’ve got two parallel circuits in each side. Which means they’re running the LEDs at 200mW. Those aren’t 1W LEDs at all.

In total I got 22.78W and 22.95W each so 45.7W. Almost exactly half of what the thing was sold as. These are so ubiquitous on eBay and the Chinese sellers on Alibaba that I think there’s actually more than one factory pushing them out in Shenzhen. There’s probably only one outfit making the metal housing and the acrylic face but other than that they probably put what they want in there.

China is still where it’s at for LEDs. Recently some actual branded LEDs with good specification from a firm called Prolight have begun being sold by Chinese vendors on eBay for prices better than bulk deals of no-name questionable heritage LEDs on Alibaba. I recently bought 20 x 3W Red LEDs for $27. They call them 3W, everyone does, but they’re more like 1.75W. The convention appears to ignore the much lower forward voltage of different types of LEDs. Regular 1W LEDs draw about 600mW for reds and genuinely 1W for blue and white. 3W white LEDs tend to be about 2.5W in reality.

This is pure power consumption, not light output. I’m not even going to go there on the issue of luminous flux when it comes to buying LEDs from China. Essentially this is a figure you can completely discard because it’s something that they entirely made up, or they don’t understand such as quoting Lumens for 660nm deep red LEDs which is rather nonsensical.

Anyway, these awesome LEDs appearing on eBay recently means I can buy the equivalent in power consumption (to say nothing of the fact I’m sure they’re a lot better than the ones in this ’90W’ UFO), for about $60. I previously paid twice this price for 1W LEDs and I’m exceedingly happy with those but this time I know where they came from and have the manufacturer’s full data. That’s not normally something you see in China either. One incredibly common practice you see on Alibaba is multiple firms selling exactly the same product but modifying the part number and claiming they’re the manufacturer. You can tell just from the style of the table and the rest of the model number that it’s the same part. Just one more thing I fail to see the point of…

That still means I’d need to mount and power them for $48 which isn’t doable for a one off. However if you were building a larger system, you absolutely could do better. Particularly taking into account a better design, higher efficiency LEDs and so on. From this episode and my experiences of chatting away to suppliers in Shenzhen  is the wilful deception which seems to serve no purpose than to turn you away. In this case the manufacturing technique  is astonishingly low tech to the point that a small investment in tools and they could make these things cheaper and better.

They could make something rather like this 90W grow light that would be vastly superior while being only a little more expensive. That would be preferable because they’re competing against things a lot more expensive than this, there’s no need to go down the cheap and nasty route.

That said, there’s a new kid on the block. You could be forgiven for missing it altogether because it’s called Mouseover Dictionary Framework and as the name implies, it’s actually a framework for implementing this kind of extension for any language. The author, known to us only as dictinfo.blogspot.com, has also made a Chinese-English dictionary. All you need do is install both extensions, ten click on the dictionary icon and select Chinese-English. There’s a Chinese-French one too if that tickles your fancy.

Mouseover, as I’ll call it from now on, is very good indeed. Zhongwen has niggling issue such as leaving artifacts of the pop up notification on/in HTML elements when you turn it on. That’s just cosmetic but the way it constantly appears not to pick up certain compound words in a sentence is not cosmetic and is actually really quite annoying.

There are feature differences between the two extensions to bare in mind. Zhongwen has nifty features to post a highlighted word to your study list on ChinesePod or Skritter. It’s also got tone colours, which you may find nice if you happen to use the same tone colours as Zhongwen does (they’re not configurable). I don’t but having badgered cschiller, he put in a setting to turn them off.

Mouseover has heaps of settings, you can make the box exactly how you like. It’s really rather impressive in this regard. There’s also some nifty shortcuts to enable and disable for all tabs/only that page and to pin the look up word so it doesn’t go away. There’s the usual sort of highlight text and translate thing too. Chrome makes it hard to implement extensions that can work on text in a text input field, Mouseover has a workaround for this with a key combination that performs a lookup – again most welcome.

Update: Unfortunately it seems the Mouseover framework has a critical flaw. It appears to disable the ability to copy text from any page when it’s enabled. At first I thought it was just hanzi which would be annoying but it seems to be everything. This is a humdinger and a showstopper so with tail between legs I’m heading back to Zhongwen.

In other news, Dreamhost has gone back to snailing. Kevin Rudd ejected by shady trade unions and big business largely because he did what they wanted (kill the ETS), new PM chick promptly caves to miners demands making a sell-our liar of herself but the Australian public, who have just been shafted to the tune of billions, increasing their taxes and lowering their public services, really like her because she’s a woman and she doesn’t sound all educated like that Rudd guy. Australian politics is depressing.

It’s one week until Uni results are out. My mid-year application is in, there’s little to do but sit tight. I’m working hard getting up to speed on my part time gig and of course studying Chinese every day.

However my usage statistics showed something else, the device was using a LOT more power when I was out and about on campus. This ended up being a stronger correlation than the usage of Google Maps, although it was easy to blame Google Maps because it tended to be used during these points. What I ended up discovering was that it’s the AMOLED display on the Nexus One that is the power demo. The CPU too, if you can find ways to get it to do a lot of work, is a serious drainer. I think that CPU is better clocked at 850Mhz but the press/public latched onto Snapdragon being 1GHz so it became a marketing arms race.

The reason for the huge change in power drain was because of changing the screen brightness up to the highest level to read the screen in the bright Australian sun. I also want to make some comments on quality of display of LCD vs OLED.

Without making writing a lengthy and technical post about how the technologies work precisely, the difference between LCD and OLED is quite profound. With LCD you have a single backlight source, either fluorescent or white LEDs, and the LCD display itself cuts light making it out of the device. As opposed to OLED which has individual light sources for each pixel. Therefore it comes as no surprise to learn that LCDs use about the same power despite the display content, the only thing that changes power is the brightness of the backlight. There’s also an effect whereby LCDs can still be visable in bright light without needing to ramp up the backlight to such a level to provide a strong enough source of light to overcome the ambient light. Effectively light can bounce off the back drop that enters from outside, and not the backlight.

So what does this mean? Well, an OLED display with a largely black background uses less power because it’s not really doing too much. It’s not zero power for various reasons, but around the 30-40% mark of an LCD. It’s typically trotted out that an OLED display showing off typical program material will use less power than an LCD display. This is where things get complex and where basically the various marketing capacities of technology companies and mobile phones have chosen to measure and therefore state things in such a way that position OLED as being the all-conquering hero. It is not.

The problem for a device like a mobile phone is that you frequently will need to view something on a white background (a web site or a UI) in full daylight. Even the best OLED display is not as good as LCD as this and in order to be in the running we’re talking about 100% brightness. Under these conditions it can use 3-4 times, or more, power than an LCD. If this were rare you wouldn’t have a problem. However it’s not rare. A mobile phone goes outside and the program material often is of a bright nature. Some might be puzzled by Apple not choosing OLED on the iPad. Take a look at Apple’s UI. It’s white. An OLED display on an iPad could be catestrophically power hungry and a panel of that size would also be extrodinarily expensive.

LCD is a highly understood technology with a whole bunch of different implementations of various costs and trade-offs, of which it generally can be easier to find one which works best for a mobile device.  Apple went with an LED backlight IPS panel which is a pretty expensive choice but a very good one for image quality, brightness and wide viewing angle. There is a sub argument here on the readability of LCD versus so-called e-Ink, in general light-source displays are really bad for long viewing in low light. I wouldn’t want to read books on any LCD display.

Lastly another thing about quality of display. Anyone who has seen the display on an iPhone and a Nexus One will be left in no doubt the Nexus One has the better display. I find that the iPhone’s display in particular seems to suffer a major loss of contrast as the backlight is dropped, this is a fairly common issue with LCD and it’s also a symptom of older, cheaper, LCD technology. OLED is not immune from a reduction of image quality at lower brightnesses, in particular it has a fairly poor colour fidelity at lower brightnesses and overall like most organic LED materials, degradation is a lot more rapid than one would like.

To counteract that OLED panels are biased blue from the outset, so the faster drop-off of blue will be less noticable. That would be okay but the vendors then often bias the panel back to linear colour response, which effectively nullifies that approach. It’s possible, although I have no knowledge of this existing, that the driver circuitry could simply know hours of life and alter the colour balance to try attain a linear colour response over a greater proportion of screen life.

AMOLED displays, like the Nexus One, are certainly a bazillion times better than the previous high-res LCD displays. I’m talking as opposed to the more common QVGA like displays. I still have a HTC Touch HD which has the same resolution and size screen, more or less, and it’s not even in the same league. It is not clear to me what the ramifications are for LCD for such high resolutions. It may be that this is the reason that AMOLED is particularly suitable here, but I find it more likely that the relative rarity of displays in that resolution and size and desires to cost-cut weren’t giving LCD the run it could have.

The practical advice here for N1 owners is clear. Get a dark background on your N1 and try to avoid high brightness settings as much as you can. Conscious of the competition from high-spec Android phones it will be very interesting to see what approach Apple will take with the new iPhone. I feel certain it wont be OLED but they surely must be feeling some pressure to increase the resolution and display quality over the 3GS.

The question is whether the public love affair with OLED will vanish before OLED manages to tackle these shortcomings, or will it serve as a reminder that the old faithful LCD with some more recent innovations actually has the solution we’ve been looking for.

If you want your inbox to look like your inbox, eg someone I give a damn about has sent me an email, then the deluge of this commercial junk mail is a problem. It’s also a problem if you want your phone to usefully go bing when you get an email, not the latest fucking el-cheapo LCD TV from Krogan.

My past solution to this was to shift email address every few years but I’m kind of attached to this one. So I decided to claim it back. I would systematically unsubscribe from all of these asshats. Since they’re actual companies “best practice” dictates it should just need a click in the footer and it’s done. Mostly that’s true although there’s plenty that make it harder and some which don’t include it at all which is actually Evil. Bit by bit though I did a few a day. I must have done 50 or so now and I don’t think I’m done yet. However my inbox is looking like my actual inbox once again.

All I need to do is get Google’s Buzz the hell out of my inbox (which seems stupid because it’s actually a folder under your inbox also – why not put this shit there?) and maybe create a couple of extra labels for actual useful (and obligatory apparently) emails from people like my bank and mobile operator telling me that bills have arrived. Of course it also requires that I look at the web in a different way and make absolutely sure people understand I haven’t given them permission to send me junk mail. That also includes checking through the tickboxes when registering on forums and that kind of thing…

That said, after the height of the Dinovo (probably one of the bravest consumer tech launches I can recall) what’s happened since? Well, I tell you what, I’m absolutely amazed I can’t just buy something from these guys that would be a decent couch laser mouse suitable for gaming. Previously I used the MX900 from the Dinovo, this had a good wireless range so worked nicely but the resolution was pretty darn bad and it chewed a set of batteries once every couple of days (so I had a recharger sat nearby to swap batteries).

I thought it was really time I moved up to something a bit more accurate and hopefully something I wouldn’t need to stoke with batteries. Their range is kinda baffling but wireless wise there’s really only the aging G9 and I vetoed that because it uses lithium ions, you have to use their charger, and it chews through those rapidly apparently. So I gave the MX1100 a go. It’s a pretty high end wireless laser mouse. Waaay better battery life, for a start it has an on/off switch which helps.

However it’s got a properly mind-fuck issue, it is absolutely utterly completely hopeless at fine mouse control. What the fuck? It’s a high end laser mouse. It took me awhile to work this out, I thought maybe the laser didn’t work on my couch right (in fact, bizarrrely, it doesn’t but the G5 does), or it was wi-fi range (it just has a shit little dongle), but in the end it was this fine control thing. This makes it actually irritating to use in Windows and absolutely worthless in games.

So what did I do? Well, I decided on a long shot to take the G5 home from work and try that on the couch, and just lump the cable so I could work out if a laser mouse would be okay on the couch. Well it is, it’s absolutely bloody fantastic in fact. So I sit here at work with the MX1100 irritating me so I can use the G5 at home…

How does something as fucked as this end up being released by Logitech? I’ve no idea. I guess I should trawl ebay for a cheap G5 to replace this thing at work too…