Three.ie Broadband Woes...

 Yay, just found something else to gripe about, though unlike my usual 'glass-half-empty' summations, this one appears to have landed sunny-side-up.

I'm with Three (Broadband) coming on 12 years now.  It's been a bumpy road regarding the quality of service provided for a lot of that time.  First off, to my knowledge, Three is still the cheapest provider of Broadband in Ireland.  But click on any one of the Google-spawned links that offer to test your Internet speed, only to then funnel you to a hand-picked assortment of ISP's that actually pay their bills, and you can't help but notice that Three.ie is never even mentioned, let alone included.  Not once in the last 12 years I've been with Three have I seen a single Ireland-based ISP comparison include Three.  The reason, up until very recently anyway, was imo, that Three was by far the cheapest of the lot, so rather than try to compete on price, the other lot simply chose to ignore them, in the hope that they might simply go away.

Well they didn't, and though they are still a small player here, relative to the main Telecom services provider, Eir, who has around 70% of the broadband market, there seems to be a shift of sorts happening among ISP's here at the moment.  Personally, I'd put it down to Musk's Starlink shenanigans.

 For the first time ever it would appear that Ireland-based ISP's are viewing one another as competition and are offering much more competitive broadband packages.  Eir for instance, is now providing a free Huawei B818 modem/router with their current mobile broadband offering, while only charging €30 per month.  This €29.99 per month with a 750GB cap exactly matches what Three has been offering for almost 2 years.  Three is after countering this by cutting their price to €15 per month, though only up until the new year.  I don't know when this was first offered, but had I seen it before now, being 'out of contract', I would have jumped at it, still might.  Thing is, competition is finally having an effect on telecom services in Ireland.  I remember back in the 'noughties', Eircom still selling twin-line ISDN connections (128kb/s combined) as their answer to 'true' broadband offerings from other companies that had begun to surface.  Thing was, these cretins were not only offering something that was laughably inferior technologically, they were actually charging the suckers that fell for this, for two phone lines, not one!!!  I was living in Dublin at the time and had proper broadband from a company named UTVIP.  What has always stuck in my craw was that I still had to pay Eircom line rental, on top of the broadband charges.  But I digress, back on topic...

Three broadband reception has always been spotty where I live.  In the almost seven years that I've been here, the service had gone from being just about acceptable to downright abysmal.  For the last few months, anytime before 1-2am in the morning I would regularly have download speeds in the range of 20-50kB/s - the lower range is only 3 times the dial-up speeds I was experiencing in the 90's.  To put this further into perspective, at its worst, watching uninterrupted, Youtube videos at the lowest possible resolution (144p) was impossible.  My newfound knowledge of the ISP industry in Ireland came from research into finding a Three.ie replacement, as things had gotten that bad. Eir, the main contender, really was never an option, as their reputation among anyone even remotely interested in online privacy, is mud - check out this Wiki-article, the "Controversy" paragraphs at the bottom, in particular.

  Well, denied an honourable way of abandoning the good ship Three, I first decided to see how my immediate surroundings fared broadband-wise.  Using a tablet combined with my old battery-powered Huawei modem, I was amazed to discover that just 3km from here, I could Download at over 30Mb/s, and upload at 10-15Mb/s.  So yesterday, I spent a couple of hours seeing was their any way I could improve the broadband reception here, going so far as strapping the same battery-powered Huawei to a make-shift pole - a couple of planks nailed together -  to see if elevation was the answer.  It wasn't.  It was then I made an amazing discovery!!!

Desolate, with optimism at an all-time low, I decided I'd try a change of in-house location for my Huawei B525 modem.  Signal strength had never been an issue with it.  Worst case, I was always receiving at least 2 4G bars on it, though usually it was 3 or 4 bars.  Moving it from the window of one room at the side of the house, to one on the front, didn't improve signal-strength, there was a falling-off if anything. 

** BUT THE CHANGE IN DOWNLOAD SPEEDS WAS IMMENSE ** 

I couldn't believe it, still can't, given that there seems to be little or no difference in the signal-received strength.  the browser-based B525 diagnostics, which independently provide signal strength, is much the same as before, namely it's a 4G/4G+ signal I'm receiving.  After a day of 'playing' with various web-based Speedtests, I see speeds of between 20-60Mbps download, and once I saw it even hit 70Mbps!!!  Upload speeds are another thing entirely.  When it's operating 'correctly', there seems to be a definite speed-cap in place at around 3Mbps.  Right now though, just after 8pm (which would qualify as 'peak-period' time) my Download speed was 53Mbps and Upload speed, an underwhelming 0.67Mbps.

Nevertheless, I'm finally getting 'value for money' broadband-wise - and it only took 12 years!  Yay!

Edit:

A week or two later, there's still great speeds being had.  Generally, it's around 40-45Mbps, though occasionally it can drop down to the low teens, or jump as high as 70+Mbps.  My talk of a 3Mbps upload-cap was just more idle speculation as I've seen wildly varying figures since then, but onto the point of this edit.

I'm still mystified as to why my transfer speeds have shot up, despite the received signal strength being marginally lower.  So to that end, I've been experimenting.  I moved the B525 to another window on the same side of the house, around 10-15 metres from the first and like from the other location, practically with a clear view to the horizon.  Then using my tablet while in the same room, I speed-tested my connection from the new location.  I expected little or no difference between locations, instead the change was profound - I went from upload speeds of 40Mbps to around 1-3Mbps.  Ironically, upload speeds increased  to 4-5Mbps.  Moving the B525 back to the other room, it reverted back to the 40+Mbps download speeds again.  Intriguingly, I found that my adjusting exactly where I sat it on the window ledge - we're talking mere centimetres here - I could dramatically improve my upload-speed.  I went from the 0.5-0.7Mbps speeds noted above to a much more consistent 4-5Mbps.  I've just done a speed-test right now and it returned a Download-speed of 48.6Mbps, with Upload-speeds coming in at 4.6Mbps.  Consider again that mere weeks ago I was looking to ditch my current ISP due to not being able to stream 144p Youtube videos without them stalling repeatedly - all down to the abysmal download-speeds I was getting (<250Kbps during peak periods, or 4 times dial-up speeds!).

But I am still no nearer to understanding why such small differences in router placement makes such enormous differences in transfer speeds, particularly since there is little or no difference in the received signal strength that I'm seeing.  Had someone suggested stuff like this to me a month ago, I'd have told them that they were talking nonsense.

Remote Control Timeswitch - Additional Switch Units...

I had intended to leave this as an addendum to the original post concerning this old project, but the 'new & improved' Blogger has apparently decided that any editing of past posts will see them treated as being new, receiving the edit date, the original date being lost in the sands of time.  This is bloody stupid!!! [Edit: Ok, scratch that, it reverts to the original publish-date, not the edit-date.  So why the unnecessary alert when you republish - that's bloody stupid! :) ]

Moving on, there's not really much to add to the RTC project, information-wise.  I ordered most of the parts needed years ago.  The PCB's I produced using the excellent Proteus EDA package.  Although all of the PCB's that make up the main controller and the one switch unit, were produced the old-fashioned way, using a Ferric-chloride solution, given the ease & cheapness of ordering them from China these days, I instead opted to send some business JLCPCB's way - and an excellent job the made of them too, highly recommended!

I've moaned on enough about the BC108 transistors I sourced for these in another post, but I cannot resist a gripe about another component problem encountered while putting these together.  I had ordered a few hundred i.c's from the 74LS logic family, between 20-30 varieties, so had plenty of the dirt-common 74LS74 dual D-type flip-flops at hand.  What I discovered was that only one 'family' of the 74LS74's I had received would work reliably - for example, any of the chips that began with 'SN' would either not work at all or unreliably at best.  All of the ones that work perfectly have identical numbers, which begs the question, are the others just faulty crap that's being sold cheap?  If so, of the 12+ 74LS74's I have, only about 4 of them are to spec.  Not a happy thought.

Like I've mentioned before, filtering/line-noise seriously limit their reliability.  Since these units are merely receivers, there is no way for the main controller to verify that a switch has been toggled On or Off.  With 3 switches to play with for the first time, I was quickly introduced to a couple of new firmware bugs.  I would bet good money that Mr Browne, the project's creator, developed the s/w with just a single switch at hand.  But they're fairly inconsequential with 'normal' use as there is a much higher probability of having missed switching due to filtering/line-noise than to it being down to firmware bugs.

The electronics side of the construction was fairly straightforward.  The only complication that arose involved the tuneable miniature mains 'Toko' transformer.  Since (unsurprisingly) these are no longer available, I was forced to 'roll my own', which (surprisingly) proved easier than I had expected - finding this spec sheet online helped matters considerably.

While the hardware side of the builds, though 'fiddly', was with care, doable, finding suitable male-plug parts proved a frustrating affair.  In order to safely secure them to their enclosure boxes, their 13amp fuses had to be contained wholly within the plastic shell-half of the 'prong-side' - otherwise, I would have been forced to cut a hole in the box for the fuse to stick inside.  Situated under the PCB as they are, and with just millimetres between PCB and the live-mains prong, safety would have been compromised and I'd probably be creating a fire hazard to boot!  Thing is, mains plugs of the required construction are incredibly difficult to find nowadays.  All modern plugs have the fuse extending right into the top half of the plug body.  I was actually considering 3D-printing plug parts, the design of which (given my rotten Fusion360 skills) would have taken me ages.  Thankfully, I finally managed to locate two - one from an old Atari ST monitor, the other (scruffy-looking) one came from an old clothes iron.

Note that neither of the plugs used have the plastic 'safety sleeve' on the live & neutral prongs, a good indication of how old they are.  Ironically, the plug I used on the first switch unit, built in the early 90's, has!!!

Anyway, the deed is done, one less thing to do before I croak, so worth the effort.

Linux Mint - The Struggle Continues...

I thought I'd attempt to get at least one new blog entry penned for 2020, even though given the farcical 'lock-down' we've been subjected to over the last few months, and the long stretches of idleness concerned, this ought to have been one of my most productive blogging periods.  Evidently not!  About the only thing of note that I seem to have achieved during lock-down, was the completion of the 2 RTC units that I'd been promising myself for decades that I'd build.  But I'll leave that as an addendum for the blog entry in question.  Besides that, I've seemed to mostly spend my time watching movies or sitcoms, to the point where I'm now sick & tired of television.  Anyway, forgive my little preamble, on to the topic for consideration.

I'm still using Linux Mint 17.3 as my main OS, despite the fact that it has been EOL'ed for at least a year now.  I gave Mint 19.1 a go about a year ago and was initially enthusiastic.  This proved to be short-lived however when I realised that compared to 17.3, it was a bit of a piggy as far as memory was concerned.  As my main computer is still my old Dell laptop - and shall remain so until either it or I depart this realm - I am somewhat constrained by the fact that although the machine has two 2gig SODIMMs fitted, the stupid hardware is limited to using only 2.5gig of this.  I was so peeved when I discovered this (about a decade ago, when I upgraded its memory!) and never would have bought the laptop had I been aware of the fact at the time.

Anyway, as a result, I decided instead to hold off until the next Mint release.  This I only got round to trying a week or two ago - or would have, had it even been available!  You see, the Mint Developers have thought it prudent to drop 32-bit support from Mint 20 onwards, at least with its Ubuntu-based releases.  Instead, what they have added is 32/64-bit versions that are Debian-based, named LMDE 4, in case, and I quote, "Ubuntu were ever to disappear" - which sounds a bit ominous for Ubuntu's future prospects...

So one download of LMDE later, I was happy to discover that it installed & ran without issue.  Initially anyway.  What was soon apparent was that the Nouveau graphics drivers didn't work properly, often only scrolling part of the screen when browsing with Firefox.  It also transpired that the 'Driver Manager' app was no longer available, robbing me of the standard means of installing native Nvidia drivers!  This was unacceptable.  I then proceeded to waste days trying to install manually, driver versions downloaded from Nvidia's site, all to no avail - LMDE seems an impenetrable maze to Nvidia's old installers.

And there it lay, until I decided a day or so ago to give Mint 19.1 another go, more as a last resort than anything else.  On the plus-side, after playing with it for a few days now, Nouveau seems to work well, no glitches so far at any rate.  What had me cursing and swearing yet again at the Mint Developers was the realisation that once more, those idiots had not bothered to fix the audio-preview in the Caja file manager.  What they had done, yet again, was to disable audio-previews by default, in the hope that most users wouldn't notice the bug!  This kind of laziness is maddening to me.  A previous post here will recall the fact that this bug has been an unwanted addition of Mint-Mate releases at least as far back as version 17.1.

Given the amount of time & energy I had spent re-building the various buggy packages from Mate's source-code, I was more than reluctant at the idea of a repeat performance, but persevere I did.  And it proved surprisingly easy this time around!  One huge plus I noticed regarded the compiler.  Prior to this, issuing a './configure' would find the script stopping at each & every dependency issue that it encountered.  This resulted in my having, after resolving the highlighted dependency issue, to re-issue another './configure' - and again & again & again, often dozens of times, until all the dependencies had been resolved - which was soul-destroying.  Now what happens, after the first './configure', ALL missing dependencies, not just the first one encountered, are listed as packages "not found", which allows you to resolve all of the dependency issues in one go.

I was also delighted to find that after building the needed packages (caja-common & caja-extensions) and removing the 'broken' caja-common (which automatically removes Caja, as well as other packages that would conflict with the new build's installation), the new packages installed & worked without a hitch!  All that's needed is to add a Home Folder icon to the Panel to allow a one-click opening of Caja - just drag & drop the Home Folder onto the desktop to there, or navigate to Caja via Menu->System Tools, right-click on Caja and select 'Add to panel'.  Once old Caja has been uninstalled and its executable 'killed' using System Monitor, all that's left is to re-install 'mate-control-center' and fire up 'new Caja' via its panel icon.  In short, it works, no more Caja locking up during audio previews.  I'll include the built packages below, more for my own future use than anything else.

caja-common

caja-extensions

Right now I'm writing this on Mint 19.1, listening to some background music and everything seems to be running fine.  The memory problem hasn't gone away by any means - presently, with just Firefox and Streamripstar running, 2.3gig of my 2.5gig of memory is being used, with swap also starting to be impinged upon. 

Will this prove usable, or will I end up back with Mint 17.3? - only time will tell. 

Edit:

After posting above about the memory limitation of my old Dell XPSM1210 laptop, it got me thinking.  On the one hand, I was 99% certain of my facts - that the laptop had a limitation on the maximum amount of memory that the hardware could work with, 2.5Gig, which was well below the the 4gig limit imposed by 32-bit hardware.  On the other hand, this 2.5gig seemed an almost arbitrarily chosen number, which didn't make sense, so I went looking for confirmation of my long-held belief - and couldn't find any!  But I was certain that I had bought two 2gig SODIMMs when I did the memory upgrade, but was disgusted to discover that only 2.5gig was available to the user when it was installed.  I therefore removed one, replacing it with a 512Meg SODIMM instead.  This is my recollection of events anyway - I can't actually find the second SODIMM that I thought I'd bought!  That was more than 10 years ago. [Did a little checking, (re)discovered that Amazon keeps a long purchase history and that I bought just ONE 2Gig SODIMM, and it was 8.5 years ago, not 10+ - so wrong on both counts!].

 

Well, I was gifted an old HP laptop, minus PSU or working battery pack, about 4 years ago, and which I think I've already commented on in another post.  Long story short, after letting it gather dust for 6-9 months, I finally got a cheap Chinese PSU, powered it up and left it on a soak-test for 24 hours - which promptly killed it, or at least the Nvidia graphics part of it.  What it does have though is 4gig of SODIMM memory, that is Dell-compatible, and which for the first time, I've tried with it.

 

And yes, I was flabbergasted to discover on booting the Dell that I now have gone from having just 2.5gig, to having over 3.2gig of memory available to the user!!!  How could it have taken 10 years for the penny to drop!  Though I should be elated (and am a bit), what's bugging me now is the almost 800meg of memory that's 'unavailable' with 4gig installed, whereas with 2.5gig installed, all of that is available to the user.   The BIOS informs me that this 800meg has been allocated to the system, but nothing is allocated when 2.5gig is installed - so no doubt it's just using the top end of the address-space to map hardware to, with the 800meg left blowing in the wind.  Yep, always moaning, my glass is always half-empty.

Edit2:

Haha, seems I'm on a roll of sorts today.  The HP laptop that donated its memory to the Dell seems to have got a reprieve of sorts.  The graphics problems with these HP laptops is well known.  Despite the exorbitant prices HP charges, they are responsible for some really shoddy designs when it comes to cooling.  This problem effects multiple laptop's of theirs from what I've read,  so it's not a case of my being gifted a dud.  Anyway, what happens is, since the ineffectual cooling employed is incapable of cooling the GPU chip while under load, it literally cooks itself to death, with the soldered BGA contacts melting, and going open-circuit.  What the user ends up with at startup is, a blank screen accompanied by an  annoying Beeping.  Despite multiple past attempts at reflowing the Nvidia chip in question, it always reverted back to the beeping/blank screen.  Until tonight!  It's been running for a couple of hours, right now, playing a movie full-screen, without a bother so far!  This is only the second time I've witnessed it behaving 'normally', the first time obviously having been when I first powered it up.

Wow, all for naught, this thing really knows how to toy with me - after a few hours running perfectly, with the GPU sitting at around 60DegC the whole time, (normal for badly-designed laptops - my 14 year-old Dell, running 24/7 from new, never goes below 50DegC, often well exceeding 60DegC under load), the screen just goes blank.  Pulling the plug and powering it back up, it now exhibits a new behaviour - gone is the infernal beeping, the screen is just remaining blank, but with back light illumination.  I tried it again this morning and it powered up normally, but only remaining so for a few minutes before the screen went black again.

Well, I'm done with it, that thing goes into the bin next week, before garbage day.  I've sweated bullets for that piece of HP crap.  It could only have been worse had I spent money on it, never mind having bought it new.  Anyone considering buying 'prestige' brands such as HP should really do a little research into the quality/reliability of the junk that these clowns produce.  Add to that, this POS company uses 'white-lists' to decide what user hardware upgrades are allowable - anything not on the list is instantly black-listed.  Since HP can't foretell the future, the user is forever prevented from adding 'current-hardware' add-ons.  I discovered this when I tried upgrading the pissy 54Mb/s wireless card that came with this laptop, to something more current - all were black-listed, cited as not having been approved by HP.  Well, Good Riddance to HP.

At least I'm left with 4gig of ram and a usable 100gig hard-drive for my pains.