It’s been a while since I’ve posted a random post, so it’s about time I did another. In the “silence” in-between, life has been extremely busy and continues to be and the observations from my daily commutes continue to pile-up.
First Journal Paper and Presentation
It has taken a lot longer than I had anticipated, sickness and all, but I have reached a milestone of putting out my first journal paper. It is titled Photovoltaic powered ultraviolet and visible light-emitting diodes for sustainable point-of-use disinfection of drinking waters published in Science of the Total Environment, and is a review paper focusing on LED and PV development in the context of providing sustainable point-of-use water disinfection. Unfortunately, access is only by subscription – if you’re affiliated with an educational institution, you may be able to get access via ScienceDirect.
As part of that, I have been given the opportunity to hold a presentation about this paper. All are welcome, interested people should note the venue details as follows:
- Title: Barriers to Deployment: Reviewing Photovoltaics and Light-Emitting Diode Technology for Point-of-Use Water Disinfection
- Date: Wednesday 2nd July 2014
- Location: Boardroom, Water Research Centre, H22 Valentine Annexe, UNSW Kensington, Sydney, Australia
- Time: 2pm to 3pm
When there’s trouble, it all comes at once and when you least expect it. That’s a near-universal truth I’ve come to accept. As it turns out, within the space of one week, I had lost three different computers used for experimentation for various reasons.
All of the machines were about eight or more years old, and it turns out, the culprit is mostly bad capacitors. One lost a power supply to bad caps, and replacing the caps did not bring the supply to life (collateral damage to the supply). The board had bulging caps, but because of its age (Socket A), was not worth a repair and was scrapped.
Another machine somehow had a motherboard failure which resulted in it POSTing correctly, but unable to then transfer control to the bootloader on any media (floppy, IDE hard drive, SATA hard drive, IDE optical, SATA optical, USB stick or PXE). It just hung. Changes of parts didn’t help that one, neither did removing all expansion cards or clearing CMOS.
That motherboard somehow collaterally damaged a SATA hard drive in such a strange way that it was detectable and readable in another machine, but the drive was not usable for booting or for installing Windows on. The Windows installer would complain no drives were detected even though the drive shows up correctly with the right capacity in the BIOS and is accessible if it’s a “secondary” non-boot drive on any other machine. At first, I thought it was an AHCI/IDE mode bug, but nope. No matter what the BIOS setting was, the drive was always absent in the Windows Installer. Shoving another drive in managed to have it show.
Another spare board was unpacked, only to find that it too had bulging caps after resting for an extended period. Chances are, the electrolyte was decomposing, even without the help of the heat from the ripple current. Such is the problem of having lots of “old” spares from vulnerable generations.
Even my two Raspberry Pi’s in my roof that had been going for 40+ days uptime decided to disappear from the network. That turned out to be a contact problem on the power-plug end of things.
These strange failures are not easy to track down, and left me scratching my head for ages. I just didn’t think some of these failures were even possible, but my methodical troubleshooting proved that to be wrong.
A New Rig
I decided it was time to start new and grab something very powerful for some “experimentation” I had in mind. That’s the main reason behind all the “new-stuff” reviews that have been popping up here recently. As a result, I ended up building a very pricey new rig, which should do everything I want it to do while remaining moderately sized so making it possible to transport should the need arise. The specs are:
- Intel i7-4770k
- Asrock Z97E-ITX/ac
- Corsair Vengeance LP 1600 16Gb Kit (already had on hand)
- Intel 730-series 240Gb SSD
- 4x Seagate 4Tb 5900rpm HDD
- Lite-On iHBS312 Blu-Ray Writer (already had on hand)
- Asus Xonar Essence STX Sound Card
- Noctua NH-D14 Heatsink
- Corsair VS650 Power Supply
- Bitfenix Sceptre Pro 200mm Green LED Fan
- Bitfenix Prodigy miniITX Case
- Gigabyte Aivia Osmium Brown Mechanical Keyboard
- Steelseries Sensei Raw Laser Mouse
- 3x Viewsonic VA2249S 21.5″ Full HD IPS LCD
- Microsoft LifeCam Studio Webcam
- Logitech Z623 Speakers
- Audio-Technica M50x Headphones
My wallet took a major hit. In all, it would be the most frustrating build I’ve had to date. The miniITX form factor, while being small, is a royal pain to squeeze things in. However, credit does go to the Bitfenix Prodigy case for allowing me to just squeeze everything in – it’s decently powerful for the size.
I was going to go for a Gigabyte mini ITX board, but they had no stock, so I had to settle for Asrock. Not my preference, but I didn’t want to wait. I didn’t opt for a graphic card, instead, using the Intel HD4600 “integrated” solution, as I won’t be doing anything graphically intensive. Besides, the mini-ITX boards only have one PCIe slot, and I wanted a nice sound card so I filled it with the Asus Xonar Essence STX. It’s a pricey card, but it’s very quiet and I have no complaints about it really.
I also had a few spare bits on hand which I wanted to use up, so in went my spare 2x8Gb Corsair Vengeance LP 1600mhz RAM and my Lite-On BluRay burner. Everything was a tight squeeze – the NH-D14 had to be turned upside down, and one fan removed to fit in. The Lite-On fitted in with less than 2mm to spare. To help airflow, the included front 120mm fan was moved to the top, and a 200mm intake fan was fitted (since they had no 230mm in stock).
This is the part that’s most exciting for a new systems builder. Of course, I’m entirely fearless, and the Noctua was specified mainly for overclocking, but it seems my luck has officially run out with this build.
It is a very difficult machine to work with. Very. My assumption was that Devil’s Canyon release soon meant that Intel had perfected its yields and the chips were more-or-less coming out at higher bin speeds with lower power consumption. Not wanting to wait, I thought I’d score a bargain on the i7-4770k as they’re running them out, and live with the thermal paste TIM issue. After all, I’ve read many success stories of people getting to 4.4-4.7Ghz, so that was my expectations.
Initial “blind” attempts resulted in many ‘101’ BSODs which implied more Vcore required. I had no luck at 4.4Ghz – it boots, but it won’t stay stable under Prime95 Blend. I even shoved the Vcore up to 1.25v, and watched the machine shut down due to >100 degrees C on the core. There was no thermal margin to be exploited at all.
Looking at the most comprehensive Haswell overclocking guide I found this statement about the “stock VID” – the voltage the CPU was certified at the factory for running at stock speed (3.5Ghz):
1v CRAZY (TeamAU code: OPA)
1.1v average (TeamAU code: OK)
1.2v Not Good, go kill someone (TeamAU code: JUNK, pull out the 9V battery!)
I found my VID was 1.2590v. Way beyond the “not-good” zone. This one’s literally a chip that “only just made it”. On the side note, yes, I do feel like harming someone now …
but as the chip did technically pass their binning process, I just got very unlucky.
I thought me giving the core 1.160v was being “generous”, as others were getting up to 4.5’s there, but for mine, it was undervolting. If I were to have any chance of overclocking, I’d have to undervolt as well to restore some thermal margin (i.e. stop it from overheating).
The main reason it was overheating was because of the paste between the die and the IHS. Even with a Noctua on the back, which wasn’t working up any sweat, I was seeing 10 degree core-temp differences and overheating when the power dissipated reached ~115w. I know for a fact the NH-D14 works fine even above 160w – my AMD Phenom II x6 1090T BE trips out the VRMs at ~160w and still stays below 45 degrees C.
It looks like I should have waited for the Devil’s Canyon 4790k and ponied the ~$35 differential as that would have made things a bit faster. The replacement of the thermal paste with a better material would help prevent the chip from overheating as well.
It was through this discovery that I realized that many overclockers are novices and don’t know what stability means. Many of them think 10-30 minutes of Intel Burn Test or Prime95 is enough. Others think that “Eh, the 8k FFT in Prime95 causes overheating, so I’ll just run 1024k FFT which is cooler and if it works, you’re good.”
Neither of these scenarios is true stability. Especially for an advanced user, you really want to be able to ensure the machine doesn’t falter on any work-load or shut down on you. As a result, my benchmark for stability is 48 hours of Prime95 in “blend” mode, at the least.
It seems the best I can do is 39 x 100 (3.9Ghz) on all cores and cache at “fixed” speed. Voltage is dialled back to 1.165v. I did overclock the graphics cores to 1650mhz at 1.25v (no graphic driver restarts or artifacts under heavy load), and I have Vin at 2v, Vrin at 1.15v, System Agent Offset at +0.23v, and Analog I/O and Digital I/O offset at +0.2v. Trying to run at 40 x 100 gives me ‘101’ BSOD or shutdown due to overtemp. Drats.
I did start overclocking the RAM – with Vdimm up to 1.60v (fails under 1.50v), my 1600mhz kit is running at 1800mhz, but with even further relaxed latencies of 11-11-10-24 (meh). Unfortunately, it wouldn’t make it to 1866mhz in any fashion at all – no POST whatsoever.
But the problems don’t end there …
So it passed Prime95 for two days straight, the temperatures peaked at 91 degrees (hot, but not lethal, and I really don’t want any less performance), so what’s next?
What’s next is a rash of BSODs which have perplexed me for ages. For one, I got a BSOD involving the Intel xHCI USB 3.0 driver with no apparent cause. Even at stock speeds, after a fresh re-install of Windows 7, with the latest drivers from Asrock, I get the BSOD just once after the install but never again. Curious.
But then I tried to certify my 4Tb hard drives while Prime95 was running, and then came the torrent of iastor.sys and iastorA.sys BSODs. The drives were actually fine, the cabling was good despite the SMART listing Command Timeout (likely because the BSOD meant the drive had outstanding queued commands it couldn’t return to the system).
In the end, I verified I was running in AHCI mode, increased the PCH voltage to 1.152v and 1.560v respectively and that improved throughput on the SATA ports slightly and allowed for Prime95 with one hard drive to be certified just fine.
But then I opened another can as soon as I tried to certify all four in parallel with Prime95 going. Same iastor.sys BSOD!!! ARRGH! In the end, not finding any newer versions of the Rapid Storage driver, I decided to go and uninstall iRST and go with MSAHCI (i.e. generic) driver. So far, touch wood, it’s still good and the performance doesn’t seem to differ significantly.
Another frustration was the listing that the motherboard and CPU can run three screens. So far, my quest has resulted in no fruit – I have one on a DVI to DVI cable, another on an HDMI to DVI cable and the third on a DP to DVI cable (probably active due to a chunky DVI end). Unfortunately, while all three are detected, only two can be activated at any time. Attempting to activate the third results in “Settings could not be saved”. So, I’ve now attached the third to my j5create USB 3.0 adapter, but it’s pretty woeful (even compared to Displaylink USB 2.0).
As you can see, that easily kept me busy, just getting everything stabilized, plus, now I have a rig to build for my aunt which will take me a while. More reviews of stuff might be forthcoming if time permits.
I’m not one to want to pay more than absolutely necessary for networking gear, so after I discovered the WD MyNet N750’s on clear-out, I managed to amass a total of just two. That was enough to upgrade my network to 5Ghz triple-stream N. As I didn’t want to replace the core router, and the existing 2.4Ghz single stream configuration which was “g, n” compatible, I retained my TP-Link WR740N’s.
Unfortunately, just having the same configuration doesn’t automatically mean your 5Ghz capable gear knows what to do. Unfortunately, I found many of them clinging to 2.4Ghz even though they didn’t have to. As a result, I implemented a crude band-steering method.
On my TP-Link WR740N’s, I blacklisted all MACs belonging to 5Ghz capable gear, so they receive no reply to their associate requests. This means that the Wi-Fi devices think they’re “not heard” by the base station and continue to try another detected AP with the right SSID. Eventually, this leads them to associating on the 5Ghz network, thus clearing off the “precious” and limited 2.4Ghz bandwidth.
What a shame not more gear is 5Ghz capable, but it seems none of my neighbours have 5Ghz, so I get to use two 40Mhz wide carriers, one for upstairs and one for downstairs. No more complaints of “microwave” Wi-Fi problems on most of the iPads.
Station upgrades continue all around the network. At Granville, the former bus terminus doesn’t seem to have seen much change.
Wednesday 11th June
Wednesday 18th June
Tuesday 24th June
Of course, that’s not a reflection of the amount of work being done. Some work is more “important” than others. For example, I saw this sign noting stairwell closures:
This closure was in fact to remove the covering over the stairs – it’s now more exposed to the elements. Another improvement is the nearly-station-wide replacement of the fluorescent tubes with LED-look-alikes:
These ones have the same form factor, with red end-caps and no diffusion. It’s clear that the LEDs have “spots” and it seems the power line frequency also results in “strobing” on the image. This is different to the green-capped, diffused Philips LED tubes used to retrofit Redfern station. Maybe they’re trialling different tubes?
The interior of the unfinished refurbished station toilets, as promised in one of my previous posts:
Platform 16 and 17 are finally getting their hydraulic lift replaced, likely with a cabled traction lift.
Interestingly, signage in trains are getting updated too. The old classic “blue and red” style labels seem to be replaced by a newer set of different colours – grey, green, orange, blue.
New style, with references to Opal card
A mixture of colours, and more information than before
Even the “easy access” seat has changed to an orange label.
More DTRS Sites
It seems that there are more DTRS sites being implemented, having spotted a “lower height” one on the Bankstown line and some work around Villawood. Not too sure, but I’ve got a bit better shot of two – this one is near Macdonaldtown, and appears to be in the yard itself.
This one is at the wye near Sydenham – the DTRS nature is given away by the strange alignment of the panel antennas which are optimized for coverage within the rail corridor only.
Millennium Train Display Problems
I boarded a Millennium train this week with some very interesting display issues. For one, the last display on the train sometimes doesn’t come up with the first line, as it should, and flickers as if it’s being reset numerous times before it starts displaying the stops:
It should look like this, and it does sometimes.
Further along the same train, one of the carriages had stuck displays, claiming we were stopping at Circular Quay when at Auburn and also announcing the doors closing message when the doors are opening. A thoroughly confused train.
Bondi Junction Line Meltdown
I just so happened to be caught up in the Bondi Junction line meltdown on the morning of the 24th June. A signal and points failure at Martin Place took out services along the line from Town Hall to Bondi Junction and caused massive delays. Buses had to take over. I was changing over to the line at Redfern when it happened …
You kinda know something bad is about to happen when a later train is listed as arriving before an earlier train. This happened minutes before the board was blanked and passengers were told to change to a City service from another platform and change over to buses.
One thing I noticed is just how old the fluorescent tubes at Redfern Underground were. They were all the “thicker” sort (i.e. think of the 40w tubes as opposed to the thinner 36w “saver” replacements), and many of them were darkened at the ends and running “half wave” with a noticeable flicker.
Tubes which aren’t turned off last a very long time, with degrading light output. I can’t imagine these tubes are very efficient, and it adds to the “dinginess” of the place.
The line remained partially closed from about 9am to 12pm – quite a long time.
Opal Card Observations
The marketing for Opal continues, this multi-part poster spotted at Central. Pretty cool. The more Opal cards I “hear”, the more frustration and error sounds I hear. It seems that part of the reason is that people use their wallets on the reader with multiple ISO style cards, and the readers are selecting the wrong card to work with, or may have committed a transaction and then selected another card within the same wallet right after. The latency of the LCD update has confused many users too.
I saw a user encounter a 9x series error on first tap with the plain card, remove the card, and replace the card with a “Try again” one card at a time message, then tap it again, assume it failed and walk to another barrier. And it was only then, it went green and opened up. Three taps, one card, for one transaction. The latency might have meant that he removed the card *as the transaction was being processed* and causes user confusion. Regardless, I hope these things get ironed out.
One of the major reasons to push for Opal is that some tickets will be retired from 1st September. Unfortunately, the poster is very poor and doesn’t actually tell you which tickets.
A bit of ferreting online reveals the following –
|The following tickets will NOT be sold from 1 September|
It looks like this is to provide incentive for people to change over to Opal and keep the system “fairer”. Still-valid tickets will continue to be honoured past 1st Semptember, but new tickets of these types will not be sold after the 1st September.
Also noted is the strength of the Opal roll-out on buses. More routes are being added to the list. Transdev/Veolia buses here in Chester Hill have mostly been upgraded with readers, but aren’t ready for use yet.
Notice this bus is missing its driver … but here’s another Transdev bus with readers.
All of these have labels plastered over them, so they’re not ready for use yet, but many Sydney Buses are running in the city with active readers. The interesting thing is that I spotted someone who had tapped on successfully on a bus on a non-Opal route but somehow, failed to tap off successfully as the readers were stuck on Please Wait when the doors opened at the destination. So, uh, I still can’t recommend the use of an Opal card on non-Opal routes.
Australia Post Parcel Locker
Since I was buying new computer gear, I visited the new MSY at Eastlakes, near the university. As I did, I walked by Kingsford Post Office, where I got to take a close look at the Parcel Locker. It’s apparently a common sight in Europe, but here, it’s a new thing.
It’s a fairly big machine, as you’d expect, if it were to have any chance of holding a decent number of parcels. The parcel locker has different compartments with different sizes for different sorts of packages.
In the middle is a touch screen which allows you to interact with the machine to release your parcels based on a code in a message sent to your phone.
Some of the lockers are pretty high, so chances are that I wouldn’t be able to reach them.
Instructions are listed on the side, however, the whole idea of parcel lockers is a bit troublesome due to the limitations of what can and can’t be placed in a locker. They also require the mail to be addressed to the locker, so no attempt to deliver to your premises is made.
If only they could try delivering to your premises first, then if undelivered, divert to a locker and allow for international registered post to be placed within them, then it might actually be useful for me.
While on the way to the new MSY, I also came across this opened up splice junction – oh how I would not want to be working on that fragile mess.
Just for the record, the new MSY is staffed by a nice Asian man, and the queues are short. The area is served nicely by the 418 bus route which links Bondi Junction via UNSW past the shop itself.
My room got a little boring, so I decided to spruce it up with 24 sheets of A4 photo paper, a colour printer and some blue tack. Now I have waterfront views every morning!
What happens to Vivid exhibits after the show finishes? Well, they get powered down and left around, looking all sad. This was spotted in the FBE building at UNSW:
One morning on the way to uni, I saw this optimistic sign being cancelled out:
On Anzac Parade near the Eastern Distributor Tunnel entrance, there is a building with a crossed yagi with a 45-degree polarization. I wonder what it’s for. It looks like a VHF antenna, so amateur satellite 2m band came to mind, but it’s not steerable and it doesn’t look like it’s hooked up for circular polarization either. Only one dipole seems to be the driven element from the shot. Curious.
I also spotted an antenna near the Sydenham wye junction on my few days travelling in via the Bankstown Line. It was across Fraser Park and looked like a wire dipole fairly high up. I couldn’t get a sensible shot, so I stole this from Google StreetView and enhanced the wires. It’s pretty big. I wondered what it was for? It would make a good shortwave observation station.
Thanks to the map by Balint, I tracked it down to an AM narrowcast station run by Mars Media.
Ziphosting has been having database issues in the interim too, to the point that even their corporate site went down. If you experience issues, come back in 5-15 minutes!
Also of note is that WWDC and Google I/O have both gone by with little in the way of hardware, and lots of focus on wearables. I like the fact that this makes sensor networks and platforms easier to work with, but I’m still not convinced that a wearable device really is a necessity. Just waiting for a killer app, I suppose.
PNY and Kingston were both caught with changes to their product after review that affected performance – I can’t believe this is still a problem since OCZ was caught doing exactly the same and had to apologize and accept returns from dissatisfied users.
That’s all the random I’ve got time for this week, more random to come in the future!