It’s been a while since I’ve made a random post, or any sort of post for that matter. Don’t worry though – I’m still here, and I’m still busy as always. If there was a picture that could describe how I’ve been feeling in the last few weeks, it’d have to be this one:
Overwhelmed to the point of exhaustion. Flooded with things to do (mainly PhD stuff) and not much energy to do it. Part if it has to do with the awful weather we’ve been experiencing lately in Sydney, and the other part has to do with an unexpected situation which means I have a lot less time to do the things I like – blogging included.
Anyway, since it’s been many weeks since my last random posting, this one’s going to be a long one.
It’s been a while since my last posting, but that doesn’t necessarily mean that there hasn’t been anything done. In fact, I’ve got a backlog so long that I’m sure I’ll forget something. In the past few weeks, the site achieved its peak estimated Alexa rank of under 600,000!
Unfortunately, Ziphosting (my web host) has had a terrible week of trouble which resulted in me filing a ticket almost every single day. No configuration changes were made to the site throughout the time, and yet, it was going down left right and centre.
Before, when there were issues, I really didn’t care too much and didn’t bother filing a ticket, but lately, because of the level of frustration, I have been more diligent in filing them. Unfortunately, this web host doesn’t attend to tickets very quickly, so it feels like I’m reporting issues to a brick wall sometimes.
As it turns out, many of the issues come about because of abuse of resources in shared hosting environments. As a result, many sites can easily go down because of some inconsiderate users. However, I still think that more should be done to make sure the quality of service is improved.
For example, Pingdom Tools has my uptime for the past 7 days pegged at 98.46% – below 99% which is pretty bad. The response time isn’t particularly great either – it’s probably the main reason why loading new pages takes a while.
As a result, please accept my apologies when the site goes down – unfortunately, there’s not really much I can do about it at the moment. Unless I choose to forfeit the remaining half-a-year of hosting which I’ve already paid for, and suffer some downtime in the immediate future, I can’t migrate to another host quite just yet. However, I am seriously contemplating jumping ship sometime around January, and haven’t yet made up my mind about who to go to just yet. I’m thinking of either Hostgator, Bluehost, Dreamhost or VentraIP, although given that this is a hobby site of mine, I can’t justify spending too much on hosting.
I have the full intentions to continue developing and hosting all the information I’ve put my time into producing well into the future, so in case of downtime, please return at a later time and hopefully all will be well. Rest assured, as a precaution, I am now taking regular full site backups, but even if I choose to change host, the time needed to upload the (currently) 10Gb of associated files will take days!
Aside from that, I’ve been running Akismet for comment spam filtering, and this month it seems that spam is at an all time high.
I’ve removed the scale, but the number of spam posts is more than twice the amount in prior months, even though the month isn’t over yet. It seems that with increasing popularity and readership, also comes increasing exposure to spam. I hope this won’t eventually become too overwhelming, but it seems very poor form to continue spamming a site when you know that none of your posts are getting through. It’s a waste of bandwidth, time and effort.
I also had an interesting chance (because I didn’t end up posting anything) to run a little experiment on how much publicity can affect the popularity of a website. Having recently almost fully populated my CPU Corner, I decided to share it as a comment on a posting by a relatively major Tech website’s Facebook post.
Rather interestingly, it blew up my view counts quite significantly, but only for two days. After that, we were back to normal – impossible to distinguish from the background. As it seems, publicity in itself only results in a short-term pay-off, and with reference material like my site, the number of repeat visits are relatively limited. In some sense, you can consider this the “impulse response” of a share – it really only has a shelf life of one day, spread by the time zones around the world.
Aside from that, I’m still coming to grips with the internet as it stands. It seems that there are some “meta” sites out there which are algorithmically generated, and act to “gateway” my content to end users, stripping all of my ads, while presenting their own ads and claiming revenue for what is essentially my work. Likewise, there are some people out there who have taken my images, benchmark results, and other data and “republished” them on other sites with barely a link (or sometimes a small one).
While it makes me annoyed to see a lack of attribution and the loss of revenue (which is peanuts anyway), I’ve come to accept that knowledge should be “freed”, and whatever I put online is almost essentially “free-for-all”, however, the more interesting images will remain watermarked into the future to ensure they always carry my attribution.
So far, I’ve provided a direct contact e-mail address on the sidebar, despite the full knowledge that this is poor form and will eventually lead to unsolicited e-mail spam. So far, the number of automatically sent spam is relatively manageable, however, the number of e-mails requesting personal support is beyond reasonable.
I’m a generally nice and friendly person, but unfortunately, because of the recent rise in time-consuming e-mails, I’ll have to state the following:
- I am not your personal helpdesk. I don’t know everything, I cannot possibly know everything, and I cannot possibly help everyone that comes along with a question.
- If you have a question, it is often better to post a comment so that everyone can see it and any responses I may wish to offer. That way, as many people benefit as possible.
- If you have a very specific question relating to your circumstances, unfortunately, 99% of the time, this will not interest me. Do not be under the impression that just because I’ve written something on a similar topic that I will suddenly answer every question about the topic that you might have.
- Within every field, there is jargon and technical aspects which can only be adequately explained using the correct terminology. I won’t be involved in lengthy “lessons” in trying to privately educate you about some of these technical aspects. Please take the time to self-educate by spending some time on Google. There’s a lot you (and I) can learn from the internet.
- That being said, I am happy to receive e-mails from users when they have something that concerns me – maybe it’s a complement, maybe it’s some well-reasoned disagreement, or maybe it’s some opportunity. Whatever it may be, please be polite and don’t let this deter you from sending me a message. I’m a reasonable person.
From now on, if you send me an impolite e-mail, or one about your specific problem that doesn’t interest me, I will not reply. If you are half-reasonable, I might provide you half-a-reply. There are many forums online where many technical minded people are able to better help you.
I honestly don’t care for e-mails that go “you’ve written about this, so you must know this, so please solve my problem”, or “you’re so smart, so this is going to be easy for you”. If you leave a comment, I may or may not reply depending on whether I have any time, or information to contribute. Just because you leave me a message does not make me in any way obligated to help you. I’m sorry, but I need to live too.
I hope that this will save me enough time so I have a better chance of updating the blog with more interesting, different and exciting experiments.
Sydney Trains Station Upgrades
Not unsurprisingly, the train station upgrades at Granville have been going gangbusters!
Wednesday 6th August 2014
Wednesday 13th August 2014
The side shelter is constructed, sheets laid and the concrete roadway has been constructed for the buses to run on. The new carpark has had its bitumen laid, and the tiling has been completed for the pedestrian walkway section, and the temporary fence moved. I haven’t shot any more recent shots, but I’m sure the recent weather hasn’t helped. It won’t be long before the new Granville interchange may be ready.
I also had an unusual reason to visit Chester Hill station and I found the stairs nosings redone, as well as a complete LED lighting retrofit for their fluorescent tubes. The strange thing is that each station seems to be piloting a different solution – just as Redfern had Philips LED solutions, Central has a “frosted” diffuser hiding its LED solution, and Granville has a “red-capped” style tube, the ones at Chester Hill are again different:
Because of some relatively low-hanging lights under awnings, it was clear to me that these were Osram SustiTUBE Basic tubes. The tubes themselves appear to be retrofitted to existing inductive ballasts with only the replacement of tubes and starter with no rewiring strictly necessary. However, as the tubes will be operating with the ballast in series (i.e. a massive inductor), it affects the power factor and does contribute a small amount of ohmic loss. Electronic ballasts must be completely rewired and re-certified, however. Definitely some quality gear.
The next thing to go seems to be the metal halide lamps used to cover broad platform areas. Already, in Granville station on Platforms 1 and 2, their lamps which look like this pair at Chester Hill have been replaced with a new LED type. It seems there is full support for LEDs as they reduce energy usage and maintenance needs.
At Lidcombe, and possibly other stations, it’s clear that the sun has taken its toll on the plastic tactile “tiles”. The plastic becomes weathered, hard, and brittle and slowly crumbles. They are in the process of being pulled up and replaced – hence we see the pair of drilled holes along the yellow line, awaiting the fitment of the tactile bumps.
DTRS Train Radio Network
As an avid radio enthusiast, I have been keeping a keen eye on the rollout of the DTRS network in the suburban rail network. Rather surprisingly, many base stations have been seen to go up recently. First, I noticed this collapsed pole at Ashfield, already fitted with panel antennas, and soon ready to be deployed.
A few days later, passing the same spot, it had been erected.
Not the only DTRS base to go up, there’s actually one near Central station now. Previously the nearest one would have been the one in the Eveleigh Train Presentation Centre.
With all these towers going up, I was really excited to see whether I could hear the signals from them. I know GSM-R is digital and encrypted, but my interest goes down one level below – that of signal levels and the presence/absence of signals. I travelled the rails with my E4000 based RTL-SDR tuner, and while I heard a lot of strong LTE carriers, I heard nothing whatsoever in the allocated GSM-R bands. As a result, I have to conclude that while the bases have been erected, they haven’t yet been switched on, or are not on all of the time.
However, it seems odd given the range of GSM that bases are being built at around 2-5 station intervals. A little more thought led to the realization that the reliability of this network is paramount – to many degrees higher than the normal telecommunications networks – as the operation of the trains rely on the existence of the network. In order to ensure strong signals all around, there has to be more base stations. By having many closely-spaced stations with overlapping coverage, there maybe scope for maintenance and failure of one base station here and there without loss of train service.
Even more important would be to establish a strong signal to noise ratio especially in the presence of interferers such as out-of-band emissions from 4G base stations and handsets owned by commuters – by having a stronger signal, the protection against close competing-services is improved. It may even be good enough that illegal jammers, if operated, won’t have any significant effect (especially if spatial diversity is employed – e.g. antenna at head and tail of train).
All I can think of is “that’s a lot of money”. I wonder what will happen to the old MetroNet gear – all those Yagi antennas, and BTSes.
Opal Card Rollout
The Opal card assault continues. It seems that now, Opal advertising has entered a new dimension by having new Opal-mode-coloured labels on ticket barriers, as well as taking out advertising on television (with lame Opal-card costumed guy) and on platform-facing billboards.
Stands at popular stations seem to be taking manual sign-ups for the cards, and the number of users seem to be on the rise. Unfortunately, still no sign of concession cards.
Further incentive to change to Opal include the constant reminders of the phasing out of tickets, such as Adult Off-Peak, by 1st September. This is now being advertised on ticket vending machines by labels underneath the screen, and more posters on station walls.
On the whole, most people seem to be getting used to using the cards. However, I did see someone at a non-barriered station trying to tap a paper ticket at an Opal pole, so confusion is always possible.
It seems that everything is on track for buses too, as the local Transdev buses have now one or two functioning Opal readers on my routes – good news as this will entice more users to use it. It seems that the Opal readers rely on the Parkeon Wayfarer head units in the buses, which seems to use GPS to determine what stop a given bus is. Within a certain range of the bus stop, you can see the readers toggle from Please Wait … to being ready.
On that note, I’ve noticed that the Parkeon Wayfarer units seem to print the thermal tickets on the AES Prodata green ticket readers slightly differently.
The left ticket is an older ticket with all rows printed by non-Opal capable DataFare 2000 consoles, whereas most of the rows on the right ticket have been printed by Opal capable buses running the Parkeon Wayfarer units. Notice how the numeric field and one letter field are missing. As far as I can tell, the first 3 characters are the route, followed by a number which is probably the run, whereas the next letter is the direction, followed by date, time, bus number and ride number.
I also got to see some Revenue Protection Officers checking tickets this week on a moving train. These officers were equipped with Opal checkers – these almost look like Nokia 3310 phones – compact, with an LCD screen, which they tap on the card and read the result. It’s much quicker than checking paper tickets, and apparently, it works too!
A Dying LED Light Globe?
This week, I found myself in a predicament I didn’t expect. The Philips 13W LED light globe that I liked quite a lot started buzzing and blinking on start up. The first few days, it would zap a few times and then stay steady. Eventually, I thought it was a contact issue with the Edison Screw, so I cleaned and bent the centre contact outward slightly and tightened the lamp in the luminaire.
This failed to solve the problem, and I was really skeptical that the LED lamp was really failing because it was only a few months old. If it were to actually fail, it would be worse than an average CFL!
As it turns out, it did completely go dead one day. I took it down and I examined it, when a bright idea came to my head. I suspected it was a contact issue with the edison screw cap base – so I got out a pair of pliers and applied pressure along the crimp where the metal screw shell joins the plastic body of the lamp.
Lo and behold, I tested it and it worked perfectly.
Lesson: some things can be fixed yourself, and sometimes knowledge gathered elsewhere can be handy! The reason it came to mind is because I had disassembled a few CFLs before, where the contact was formed by leading a bare wire end through a hole, and then crimping the metal shell onto the bare wire. If the crimp didn’t apply good pressure to the bare wire, then a poor contact would result.
I got this friendly reminder in the post about the local electronics design and assembly expo, ElectroneX. I had meant to go to it two years ago, but I missed out – this year it’s at Australian Technology Park, Sydney, on the 10th and 11th September 2014. If you’re an Electrical Engineer or an interest in electronics design and supply, do register for free and come along to visit some of the major vendors.
Raspberry Pi Power Problems
I’ve been a pretty enthusiastic person when it comes to the Raspberry Pi, and I’ve got at least seven boards now. I’ve found various interesting uses for them, many of them involving radio. As a result, I do like deploying a few in my roofspace, where my antennas are, connected by Wi-Fi to my home network. By doing this, I can avoid lossy coax, and have a local home ADS-B/ACARS and RTL-SDR server, as well as an always-on SSH tunnel for VPN access. It’s beyond handy.
For a while, I’ve powered a pair of them in my roof from a four port Kogan 3.4A power supply, which I obtained on a special offer. Unfortunately, I found my Raspberry Pis completely down with no power from the power supply. After some further debugging, it turns out that the power supply wasn’t capable of putting out more than 1A without tripping the protection.
It was a pretty well glued together power supply, so it was hard to take apart, but I wanted to take it apart to see what it was made out of and to salvage the USB A Female ports for my “replacement” power supply.
The power supply had a habit of running very hot, and you can see some of the heat damage near the R6 silkscreen near the ports. The power supply seems decently built with proper fusing, isolation areas, and adhesive for stabilizing large heavy components. The soldering on the ports could do with a bit of help – note how some of the pins aren’t properly soldered on. The PCB is marked with HLT-1219 Revision 02 dated 22 February 2013.
As a result, I can’t recommend serious Raspberry Pi users to opt for this power supply, although, there is a much better option if you should so choose to build one yourself. For example, element14 had a sale on XP Power VCS50US50S with each unit at about AU$25. Very affordable!
By stealing a power cable from a dead appliance, I managed to wire the open frame supply to a power point, and have the output soldered to “salvaged” USB A connectors. With a 5v 8A rating, and 50mV ripple specification, this power supply is more than ample for four Raspberry Pis with tuners and Wi-Fi cards. Ever since I moved up to this, it’s been running very reliably and putting out little heat! The power supply is even short circuit protected which helps with safety. In fact, I’m now running four Raspberry Pis – three Model B’s and a Model A.
Pushing the Pi – Advanced Overclocking
Having been familiar with the raspi-config option for overclocking, a given number of pre-configured options are supplied which lie within the scope of the warranty. Pretty much all of the boards I’ve owned have managed to get to the 1000Mhz preset with no issues – I wanted to see how much further I can go.
In order to do some advanced overclocking, you will have to modify the config.txt configuration manually. The warranty will be voided by certain changes, but I didn’t really care for it because I had already modified the boards anyway.
As it turns out, my Raspberry Pis top out anywhere between 1050 and 1150Mhz. For the one I played with most, the board was stable at 1100Mhz, but not at 1150Mhz. There are a few options which help – I will explain it with excerpts from my config.txt file:
arm_freq=1100 core_freq=500 sdram_freq=600 over_voltage=8 force_turbo=1 current_limit_override=0x5A000020 temp_limit=99
The arm_freq controls the maximum frequency of the CPU itself. The core_freq controls the frequency of the cache and associated parts. The sdram_freq option controls the clock speed of the RAM. The over_voltage option controls the voltage supplied to the CPU.
For the default overclocks, these options are used with an over_voltage up to 6. However, an over_voltage up to 8 is available, however, to use it, you need to set force_turbo as well which disables the “ondemand” governor (thus forcing the CPU to run at full speed at all times). The magic current_limit_override value allows you to overcome the over-current protection of the on-chip switching regulator which helps overcome reboot-on-heavy-load problems. The temp_limit feature allows you to change the temperature at which the CPU is speed limited to thermally protect itself – defaults at 85. By changing it to 99, even on the hottest days, the CPU should not become speed limited.
As you can tell, I really don’t mind stressing it out just for the fun of it. The warranty bit on the CPU is set if:
(force_turbo || current_limit_override || temp_limit>85) && over_voltage>0
A script is provided to do stability testing on the elinux.org page, which you can copy into a .sh file, and chmod +x to make it executable. This script takes hours to execute if you have a large SD card, but can serve to test stability without resorting to Quake. So far, my overclocked Pi runs a bit hotter, but it’s still as stable as a rock with uptimes > 14 days so far. Every drop of performance is definitely welcome! It’s the overclocker inside of me that is satisfied by learning where the limit really lies.
Generally, in my experience, the core and sdram don’t have any headroom for improvement – they’re already pretty close to their limits at the 1000Mhz setting and not all Raspberry Pi’s can make it there. There is also an overvoltage for the sdram which doesn’t really open up any more headroom in my experience. The arm_freq is where you can make some movement – some have managed even 1200Mhz. Increase it incrementally, and run the test script and watch for freezes, kernel panics, or card errors. If it locks up entirely, you can use hold-shift-key during boot to disable overclocking for that boot or just pull the card and edit the config.txt file even on a Windows machine, because it’s on the FAT32 boot partition.
Mobile Application Permissions Creep
It’s a scary thought, but many applications on mobile devices actually have more permissions than strictly necessary, and it seems many users aren’t too wise to it either. For example, a quick search of the most popular flashlight applications turned out a very popular one which requests the following permissions:
It’s a lot of permissions for something which you want to toggle the camera LED state! You might be able to understand Camera/Microphone, but why does a flashlight need to know information about the device and app history? What about the Wi-Fi connection? Call information?
Users are pretty much willingly permitting the application to reach into such spaces and access your data.
I’m particularly surprised how apps which such obvious abuses of permission requests can still remain highly ranked and more likely to be installed by end users. The better applications request only the permissions they really need, and explain why they need the permissions.
Unfortunately, the way the mobile app space works is that the more popular apps, the better rated apps will always rank higher. This leads to a situation where there are schemes which give away applications so that you install them and boost their popularity or even pay you rewards to install certain applications. Such actions inflate their download and popularity counts even if you then choose to uninstall the applications seconds later. This helps the apps be discovered and rank higher.
It’s also the same reason why many apps feature annoying “please rate me” nag dialogs and also feature pestering notifications of “do you like this? why not my other application?”
In other cases, applications have been caught rigging feedback to favour them by not submitting low-rated feedback to the store, and instead, only having them submitted to their own support team.
In some sense this is a repeat of the “search engine optimization” issue – there’s a commercial incentive to rank higher in search results, which leads to less and less ethical techniques being employed, which push out legitimate results to lower rankings. This leads to “rubbish” pages ranking higher, and raking in ad revenue for the authors. In the end, it only reinforces my view that the mobile app space is mostly a wasteland.
So, how can we make the markets more fair? It’s not an easy question to answer. Maybe a current userbase feature would reflect how many people actually keep an app – most apps are known to be rubbish within a day of being installed anyway. But there are always people who are going to rig the system because of the commercial incentive.
Aside from that, Facebook’s insistence to push Facebook Messenger upon all of us has triggered some realization in some users of the large set of permissions requested by Facebook Messenger and the general pointlessness of splitting the functionality out into a separate application. As it is, Facebook is a big offender of being both a flash memory and RAM hog, chewing over 100Mb of memory at times. By having a second application of theirs running (or a third, if you use Instagram), chances are, they’re just taking over your phone’s resources while not offering anything specifically better.
The reason for the separate app still isn’t exactly clear. Maybe it’s to provide investors with artificially generated “good news”. It’s not as if having Facebook Messenger stops Facebook from checking for notifications in your news feed … or saves any resources.
Maybe in the future, mobile operating systems will offer more fine grained permissions access to applications for more than just location access, or data access. In some sense, Blackberry’s Playbook was ahead of the competition in that regard.
This mind never rests, but the time is always limited. With any luck, the site stays up, and my hairs stay black! Until next time!