It’s rarely the case I don’t get a chance to post until the very end of the weekend, but this week has been one of those weeks. Busyness knows no bounds, made worse when you get some surprise events thrown in your way …
Thursday night (14th March) was certainly a night to remember, as we had not one but three hailstorms pass our house, dropping golf-ball sized hail at decent velocities. It may not have broken any windows (luckily), but it did shred almost every single fly-screen and put dents and holes through my mesh C-band dish.
As the dish had survived at least two other hail-storm events without any visible scars, this was not a just another hailstorm – this one was serious causing a fair amount of damage across Sydney. That being said, some areas were lucky, escaping with just rain.
Even rain was not normal, as over the next few days we had heavy, sustained rain at levels causing flooding and breaking monthly averages within the space of a day.
As a result, I spent quite a bit of time replacing fly-screen to make sure the pesky mozzies don’t get in – especially with all the rain we’ve been having. The process is quite simple – but if you’re doing it on your own, the most difficult part is probably maintaining some tension in the screen while you roll the spline in.
My tip? Get some bulldog clips – they can help hold the tension on the screen and don’t “release” when you drop the frame to the ground like clothes pegs might. They also have a fairly low profile, so don’t get in the way of rolling the spline much. I normally use four clips – one on each corner, trim down the screen and then undo the clip on the corner that I start on, roll the spline in, and then when approaching the next corner, remove the clip when you’re about 5cm out. This way you prevent over-tensioning the screen and having “wobbly” alignment.
We had some left-over stock of flyscreen, so I did as much as I could, but that wasn’t even half of the affected windows. I’ll have to restock on the screen and continue the job sometime later …
But that being said, given my C-band dish was already somewhat marginal on gain prior to the hailstorm, I think this may well push the dish into “scrap”, at least for the purposes of feed hunting. At least the panels aren’t tearing out … yet.
Floppy Disk Torture
Last week, I was the recipient of a rusted and face-plate-less Panasonic 3.5″ floppy drive. Having plenty of stock of fresh Sony MPF920s, I decided that I could probably do without this drive, so I thought to myself – how long do floppy disks last?
Maybe I should qualify the question a little more by noting that floppy drives are a “contact” read/write medium where the head rides on the disk itself, wearing away the surface and the head of the drive. Whenever a disk is inserted, the heads are “loaded on” the flexible magnetic disk. As a result, most floppy drives and controllers only enable the motor when reading or writing, so as to minimise unnecessary wear.
So, how long does the disk last? And, on a related note, how long does the drive last? On a check of the spec-sheet, it seems an MTBF of 300,000 hours is standard for this model of floppy drive but others claim a more realistic 30,000 hours. The MTTF ranges from about 15,000 to 30,000 hours. Some spec-sheets also claim that a disk can handle 3-million passes on a single track.
I’m somehow skeptical of the figures. Perhaps they can be achieved in an ideal world, but what about a drive with perhaps slightly dirty/scuffed heads and some “late model” floppy disks which are perhaps slightly lower quality? What happens then? The one thing I didn’t see in the lifetime was duty cycle – the lifetime is certainly not calculated assuming the drive is being used 24/7 – the lifetime is based on power-on-hours. As a result, I suspect they have a duty cycle of perhaps 5%, thus 1,500 hours or so of head-life is my expectation (~2 months).
But if the figure for a disk is correct, 3-million passes would be about 6.94 days of spinning at 300rpm. That’s more than plenty for most uses for floppy disks … but if you had a stuck-on motor … that isn’t hard to achieve on a system that’s regularly powered on. Another thing to consider is that many floppy disks have internal “sleeves” that clean the media and also cause friction with the surface – would this wear down the disk even quicker?
So without much of a real aim except wanton destruction, I set out to test it out and see what happens to a fresh Imation late-model floppy disk that’s kept spinning in an old Panasonic drive. I jumpered the drive select, motor enable and side-1 pins to ground so as to have the drive running, powered it from a Manson HCS-3102 power supply and connected the index and data lines to 1k pull-up resistors and into the Rohde & Schwarz RTM3004. I tried to look at the raw data stream but it is just a series of fixed length pulses with variable spacing – its alignment is notably “jittery” due to the real analog nature of the system, so instead, I decided to go with an FFT view to see the actual signal. On my desktop, I set up a looping wget job to pull screenshots at a fixed interval of 10s from the oscilloscope over LAN.
Perhaps my settings should have had a slightly wider bandwidth and higher sampling rate, but here, we can just see the three bands corresponding to the MFM modulation at 500kbit/s – namely one at 500kHz, one at 1MHz, and one at 1.5MHz. The experiment was started just before 7pm on 11th March.
I don’t know about you, but sleeping with a floppy drive going “shick-shick-shick-shick-shick” is rather therapeutic … that is, until the power supply fan kicks in. Initially, things were all good – the drive was pretty much on-speed and the resulting bands were stable (for an analog mechanical device anyway).
However, about 11 hours later, or about 200,000 revolutions in, trouble started to occur. The drive current consumption started to increase slightly, a sign of increased friction in the system. The drive also started to have spindle speed deviations – it was trying to maintain stability but having some difficulty. At this point, because I wasn’t actually looking at the data stream itself, there may have already been read-errors starting to creep in.
Not wanting to cut the experiment short, I let it go for about four and a half days … but by then, it was clear the drive was struggling. Motor current seemed to increase further until the drive started to squeak and then the motor speed was very much off, rotating at nearly half speed for some time before returning back to full speed just for a short moment and then repeating the cycle.
I’m not sure if this was because the drive motor was overheating or something else was the cause, but the heat could have contributed to case warpage and this could have increased friction on the disk. The squeaking was probably from oxide accumulation on the head, although the FFT amplitude and background noise didn’t seem to change much which was a little surprising given the advanced stage of its ailment.
Leaving it for a little while longer, the motor finally developed enough torque that the “button” that goes into the spindle slot pulled out, disconnecting the motor from the disk for a revolution or two before re-engaging, resulting in the familiar “schlick-clack” sound. The disk was toast and the heads … probably not much better.
So how did the disk fare? Not well. It’s clear that oxide had been cleared off a path corresponding to Track 0 and its neighbouring tracks due to the shape of the head. A few concentric score-lines were also visible due to possible dirt-accumulation on the inner liner.
The wear seemed to be worse on the underside, as the top side did not seem to have the oxide torn up. This may be due to head geometry and shape, but a new interesting pattern to the disk emerged – concentric rings were the surface seems to have been unevenly worn. This may be due to the liner material. A radial pattern was also beginning to take shape – was this to do with the substrate itself or its manufacturing? I’ve not seen this before.
Being a newer, cheaper floppy disk, the cleaning liner only exists on a small wedge, rather than the full inner surface. It has notably become discoloured, especially along the ribs which exert pressure towards the donut, collecting oxide.
So I suppose now I know – keeping floppy disks spinning is a bad idea, with the later disks probably only good for about 11-hours of spinning. Drives are a bit more durable … but head clogs are never a good thing. That being said, the rotational speed may betray the health of the disk surface and head even before any failure occurs – I wonder if timing the index pulses was ever used as a measure of disk health?
But hey … destruction is fun … *throws the lot into the bin*. Maybe next time, I would better do it with some software on the computer so that I would check for data integrity on the disk.
More Audio-Technica Earpad Annoyances
I’ve written in the past about repairing the earpads on my ATH-ANC9, and my recent bad luck with Cowin products, so I’ve reverted to my pair of ANC9s I bought on my holiday to Japan to replace a pair that failed on the way. This time, the earpads on these ones decided to go … but not subtly.
The day before, I put them back in their case and they were fine … today, I take them out and it’s totally blown a seam – more than half-way around. This makes a neat repair a little more difficult. I immediately bought some super-glue to attempt a repair.
Securing the ends, and then securing sections at a time, I was able to get a neat beginning to the repair …
… but then at the top, it was a bit difficult to get the front and back to line up well, so it was a little unsmooth. But it’s still better than a failing ear-cushion! At least it’s as comfortable as before with no risk of falling to bits – but I’m on notice for further tearing, especially on the other ear cushion that’s still okay.
Field-Day Haul – Netgear GS724Tv3 24-port Smart-Switch
The thing I paid the most for at the recent Wyong Field-Day was a Netgear GS724Tv3 24-port smart switch. Readers may have been aware of my use of VLANs in the home through dumb switches and my testing of a low-cost smart-switch with rather hilarious flaws, so I was hoping to get something a little more serious. That being said, buying old networking equipment from a boot-sale isn’t exactly what I’d recommend – you never know (given the few minutes you have to inspect and pay for it) whether if it’s working or not, whether it has any strange configurations pre-loaded, or whether it’s been used in an arduous corporate environment. Regardless, I thought $50 was a fair price to pay to gamble even with the body scratches – after all, the low-cost 5-port unit cost me almost $40.
After I got it home, I cracked it open by undoing the screws and taking the front panel off, followed by the main body cover. Visual inspection proves that aside from a small dead cockroach inside, the circuitry looked okay with rather large heatsinks and thermal-pad connection to the chassis. All capacitors looked to be decent quality Nichicon/Nippon Chemi-Con/Rubycon capacitors with the exception of one L-Tec that was as a DC-input filter from the switching power supply … so was not under much stress. There was no bulged capacitors, so I think it’s probably fine to redeploy.
Powering it up, it takes just short of a minute to come online. Plugging in my configuration netbook with their “SmartControlCenter” software installed, it was easily identified on 192.168.0.239 with “defaulted” settings and firmware that was way out of date.
I was able to login to the web admin with no trouble once I set myself onto the same subnet. Before fully commissioning, I tested every port and found them all to be operational at gigabit rates – then I upgraded the firmware to the latest version. Then I set about seeing what it was like to live with, security wise …
I was somewhat dismayed that Netgear would use HTTP configuration that sent the login information in plain text. That’s not great, but at least the switch does offer HTTPS as an option – but it needs to have a certificate uploaded to it first. Configuration is awfully slow, but at least it didn’t seem trivial to crash. SmartControlCenter uses broadcast to discover the switch as well … which is not ideal.
PORT STATE SERVICE VERSION 80/tcp open http Netgear GS724T http config | http-methods: |_ Supported Methods: HEAD GET OPTIONS |_http-server-header: Web Server |_http-title: NETGEAR GS724T 4242/tcp open vrml-multi-use? 60000/tcp open telnet Broadcom FASTPATH Switching telnetd 161/udp open snmp Broadcom Corporation SNMPv3 server | snmp-info: | enterprise: Broadcom Corporation | engineIDFormat: mac | engineIDData: 20:4e:7f:xx:xx:xx | snmpEngineBoots: 0 |_ snmpEngineTime: 9m09s Device type: switch Running: eCosCentric eCos 2.X, HP embedded, Netgear embedded OS CPE: cpe:/o:ecoscentric:ecos:2.0 OS details: HP ProCurve 1810G, or Netgear GS108v2, GS110TP, GS716T, or GS724TP switch (eCos 2.0) TCP Sequence Prediction: Difficulty=253 (Good luck!) IP ID Sequence Generation: Incremental Service Info: Device: switch; CPE: cpe:/h:netgear:gs724t
I did a thorough nmap scan to find a few strange results – port 4242 seems to be used by their Java device viewer feature which doesn’t work anymore on modern browsers as the Java plugin is depreciated. Port 6000 is used by the Broadcom CLI telnet server … something I wasn’t expecting.
Connecting to it seems to provide a message asking you to wait … but if you wait, absolutely nothing happens. Well, ain’t that a bummer – but it’s because it’s a login prompt in disguise – thanks to this page I was able to work it out. Login with admin and your password … and you get this …
Then, in order to use the CLI, you have to enable it – but it wants a password. Putting in my own password didn’t work – the password is blank – so just hit return.
Then we’re in to the CLI which is apparently partially broken. But now I know it’s powered by the BCM53314. Since I’ve disabled SNMP on the switch, I found it strange it’s still advertised as a service – but it probably isn’t responding to requests, so that’s (perhaps) okay.
Looking around for CVEs, I found this advisory which was worth a bit of a chuckle – the unit was (with older firmware) exposing admin passwords (encoded) through the configuration backup URL which was accessible without logging in, and crashing the switch was as simple as requesting /filesystem/ via HTTP without logging in.
I decided to see what would happen if I tried it – it didn’t break without logging in, so that was nice. But requesting /filesystem/ after logging in still crashes the ability to configure the switch needing a hard power cycle to recover. So half-arsed fixes it is!
Knowing this, I felt confident that I could deploy it in my network – it’s not like my network is filled with attackers and it seems to do the job just fine. The problem is that a large switch like this is not something you just dump on a table-top, especially not on a small table like mine. I don’t have any proper rack infrastructure and I’m not buying any IKEA furniture to hack one up either …
Instead, I thought vertically, by unscrewing and reconfiguring the rack mounting “ears” so as to be rotated at 90 degrees, I could mount it underneath my “student desk” on the rear panel so it takes absolutely no desk space.
The gap in the panels is enough to fit the TDK subwoofer upside down, which is nice. In theremaining back-panel space, I mounted my Asus Tinkerboard which does logging and tunnel duties, the Raspberry Pi B+ which does my RasPBX, the Fingbox with its Fing-shell and my open-frame XP-Power 5V/8A supply that runs the three along with a green cable which leads to an ESP8266 weather display on my desktop. All of this, underneath the table with my tower on the right and (one of) my microserver(s) on the left … racing pedals in the middle. That’s space-efficient!
It’s been a busy weekend of random-stuff … but I’m sure there’s more stuff to come when the time permits. The hail storm was a bit of an unwelcome variation to my routine, as was the flaky Audio-Technica ATH-ANC9 earpads … but at least I managed to torture a floppy disk and get my network up and running on the second-hand smart-switch and have it placed somewhere that doesn’t get in my way.