Welcome to ned Productions (non-commercial personal website, for commercial company see ned Productions Limited). Please choose an item you are interested in on the left hand side, or continue down for Niall’s virtual diary.
Niall’s virtual diary:
Started all the way back in 1998 when there was no word “blog” yet, hence “virtual diary”.
Original content has undergone multiple conversions Microsoft FrontPage => Microsoft Expression Web, legacy HTML tag soup => XHTML, XHTML => Markdown, and with a ‘various codepages’ => UTF-8 conversion for good measure. Some content, especially the older stuff, may not have entirely survived intact.
- A biography of me is here if you want to get a quick overview of who I am
- An archive of prior virtual diary entries are available here
- For a deep, meaningful moment, watch this dialogue (needs a video player), or for something which plays with your perception, check out this picture. Try moving your eyes around - are those circles rotating???
You can find the posts here replicated onto Diaspora, if you prefer to subscribe there instead.
Latest entries: 
I’ll start with the bad news: my mortgage broker who was handling the Permanent TSB and Bank of Ireland mortgage applications came back to me with unhelpful lending conditions. I had submitted the fixed price quote from EcoTech Homes which brings the house to Builder’s Finish for €606k, so to reach bare minimum viable for the purposes of ticking the mortgage lender’s definition of ‘completed’ at which point the mortgage turns into a normal not self build mortgage:
- Add structural engineering design: €10k.
- Add M&E engineering design: €10k.
- Add site preparation groundworks + ground-air heat exchanger: €40k.
- Add internal painting: €15k.
- Add kitchen, wardrobes, other joinery: €12k.
- Add bare minimum sanitaryware to reach mortgage completion tick box: €5k.
Total: €700k (rounded up)
Bank of Ireland and the Permanent TSB lend the following maximums for self build mortgages by first time homeowners, whichever is the lowest:
No more than four times your average income in the past three years (this is the Loan to Income limit set by the Irish government).
No more than the cost of construction (that would be the €700k above).
No more than 80% of the final valuation of the completed property.
It’s that last one which is my problem specifically, because there is absolutely no way that this property would be worth more than maybe €400k after completion, based on the historical property sale prices as recorded by the national property price register for that region. Therefore, the maximum that PTSB and BoI would lend me is €320k, leaving me €380k short.
At this point the two lenders varied slightly: PTSB wanted me to have the full €380k in cash in my bank account before they would lend a penny. BoI were marginally better: they merely want €350k in cash in my bank account before they would lend a penny. This cash sum must be unborrowed money AND I cannot take out any personal loans after they approve the mortgage, else the mortgage agreement becomes voided.
This creates a particularly large hurdle, because one needs working capital in situations where large outflows occur (stage payments to the builder) which are not replaced until large inflows to replace them happen (stage drawdowns from the mortgage provider, as self build mortgage pay in arrears so you need to front the cash for each stage).
I earn pretty well, but €350k in unborrowed cash is well beyond my means. It would take many more years of saving for me to reach that level of cash. Also, I could not spend any further cash on that site getting it progressed, because every euro I spend on it reduces the cash present hurdle those banks demand and thus puts me further away from acquiring a mortgage.
What is especially frustrating here is that those banks don’t disambiguate between a fixed price contract from a builder and going direct labour. The direct labour cost based on the other quotes I received I reckon is about €450k. The banks want to add +15% to that for cost overruns, so that’s €520k or thereabouts. Despite the obvious large gap in terms of risk and time over a fixed price quote from a professional builder, they treat those two quotes identically i.e. the cash hurdle that they demand before lending a penny drops by €86k, which is two years less time I would have to wait. But then I would have to manage the build myself, chase all the trades, and assume all the risk when a subcontractor does something which messes up the air tightness needed (or indeed anything else e.g. the foundations don’t match the superstructure, and they need to be with three millimeters of spec incidentally!).
This is bad news no doubt. There is one other self build mortgage lender in Ireland, which is the Allied Irish Bank. They’re still processing my mortgage application nearly two months after submission. They have the same lending conditions except that they will lend up to 90% of final valuation, so that’s an extra €40k. I am hoping given how long that they are taking and the extra paperwork they keep demanding that they will additionally offer a personal loan, which in Ireland is capped to €75k. That would supply €435k of lending, which would mean I need unborrowed cash in the bank of €265k before they would lend a penny. That is also well beyond my means, but it does lop a few years off how many more years I will have to wait.
In any case, the chances of commencing a build in 2023 now looks very unlikely 
. Key upcoming dates:
April 2024: Purchase contract requires build to be complete by this date, or the contract is breached. To be honest, even if the build started next month, given the delays in the building industry right now, completion by this date is unattainable.
December 2024: Commencement must begin before this date to get a refund of the 7.5% stamp duty paid for the transfer of land, which was €6,750. Also, the government subsidy scheme ‘Help to Buy’ worth €30k currently expires after this date.
The good news
Mains electricity installation is finally unblocked
Eight months after ordering the solar inverter and the lithium battery storage, those finally arrived while I was at Issaquah. I got the inverter mounted onto the wall ready for being wired in:
This is a 10kW Sungrow SH10RT three phase hybrid inverter. It isn’t the Fronius inverter I originally wanted, but those have lead times of three years right now, so if I wanted mains installed on the site any time soon then I had to choose a Chinese-made not European-made inverter. Sungrow historically are the most reliable of the Chinese inverters with a typical failure rate of 0.31% per annum for their last generation of inverters, this is actually slightly better than Fronius’ typical failure rate. They’re also a lot cheaper, and have some unique features e.g. a built-in mains isolator with < 30 ms switchover time i.e. they’re off grid capable by default, whereas the Fronius needs an expensive additional automated mains isolator added to it for similar functionality.
You do give up a bunch of stuff for the reduced cost however in terms of specification and flexibility. In particular, the Fronius can extract power much much better out of dull days where the solar panels aren’t getting much radiation. And the solar panel string configuration options are nothing like as flexible with the Sungrow, in fact, they are downright rigid, but as this will be a brand new install, we can design around that.
They also get quite hot during use, up to 70 C on the outside, unlike the Fronius which has active cooling. I have accounted for this with plenty of space around the Sungrow, including a metal roof which also doubles in keeping the sunshine off the inverter, as if it get too hot it cuts out.
One unique feature particularly useful to me right now is that the Sungrow is specifically designed to be detachable i.e. you can bring it out in the morning and plug it in, and remove it end of day into safe storage. As I don’t live there, this I will make use of on my site so it doesn’t get nicked.
Anyway, the reason I needed this before I could install mains electricity is because the ESB won’t give you a three phase mains connection unless you have onsite a three phase demand of sufficient power. They instead require you to install a small single phase at a hefty connection fee, then charge you a second hefty connection fee to upgrade the connection to three phase. As much as it cost me many months of waiting, buying these now saves me that second hefty connection fee. There is a further potential cost saving: most builders have to hire in generators which they of course bill straight onto you. If I have 20 kW of three phase mains power available onsite, that should eliminate the need to rent generators, and that will shave a further few grand off total build costs.
Other
I got the 2.5 mm2 three phase power cable and Cat 5 ethernet for the security camera trunked into the conduit for the far services box:
The blue back panel is a sheet of perspex I fitted into the metal back panel. This lets you drill in mounting screws without producing metal filings which are problematic for circuit boards i.e. the ESP32 microcontrollers which will eventually live in these service boxes.
I’m still working on the house (mid) services box, should have that done by end of this week.
I got the lighting inside the shipping container done, and an alarm fitted to the site office (note yellow siren up top):
Turns out that the double width LED strip consumes 113w, which you never know until you test it, as item descriptions for Aliexpress stuff tend to understate power consumption and overstate brightness. Excluding the power supply’s inefficiency, that is bang on 20w/metre, with the Aliexpress item listing claiming 16w/metre, so 20% understated which isn’t the worst I’ve seen. With the far end LED tube added that rises to 136w. The brightness in there is a bit overkill, the shipping container computer when I finish it will be able to PWM dim the LED strip down to perhaps as little as 40w and it’ll still be plenty bright enough.
The site office alarm is a Yale thing which works by AM radio waves. Aldi were selling them cheap for €70, about half their normal cost, so given how slow I am going on wiring up ESP32 boards to act as burglar alarms I figured this was worth the stopgap. It certainly works, the instructions it comes with are inaccurate so it kept going off during installation – indeed, the gards turned up at one point, which is nice to know that they actually do so when a house alarm goes off. With the help of the internet I figure it out, and it now is working as it should. It came with two PIR sensors and a door sensor, more than plenty for my little site office. Any attempt to knock anything off their mounts triggers the alarm, it also goes off if the radio signal gets interrupted. Not bad for €70. Only hassle going forth will be replacing batteries I think.
What comes next
By the end of this week I probably have to stop doing my own wiring, as a RECI certified electrician is needed to produce the ESB certificate of compliance necessary to get the ESB to install a mains connection. Wiring in 2.5, 4.0 and 6.0 mm2 leaf cable myself is one thing, wiring anything in which touches the ESB meter box (16 and 25 mm2 cable) is another. I have started chasing a RECI certified electrician, hopefully I’ll get him to turn up in the next few weeks.
Before any groundworks can occur, I need the ground-air heat exchanger pipes which are a special order with at least a six to eight week lead time, so if I want those installed this summer, I need to get a move on. I’ve started the process of ordering them and creating a storage holder for them in the shipping container as the bulk of those pipes won’t get put into the ground until after the house build is complete, so they’ll need to be stored for several years at my current rate of progress.
Other than that, everything else hangs around getting a mortgage decision, as until I know the exact hurdle they need me to reach, I can’t take any decisions. So here’s hoping that the AIB come back to me soon with a more helpful answer than the other two banks!
I’ve turned my attention since then to installing mains electricity into the shipping container, as I need much better lighting in there than my battery powered lantern. At the far end, I’ve raised the old florescent tube light which was in the kitchen of the house where I grew up – when they were renovating it, I claimed it from being thrown out, I’ve sat on it for years, and now it finally will go into use. That should put out about 5,000 lumen concentrated at the far end where it’s darkest. I’m also raising along one side of the container five meteres of double width LED strip angled 45 degrees downwards, and that should put out another 9,000 lumen. That should bathe in a fair bit of light eight metres of the twelve metre long container, and I reckon the door will illumate the first four metres just fine.
I made a bit of progress on the shipping container computer, but I got stuck for parts. I replaced the IRF520N solid state relays with IRF540N relays, as the latter ought to waste much less heat when derated to a 5v control signal. I found I had lost my 3.3v to 5v level shifters, so I had to order new ones. I soldered pins onto them today, so I just need to get everything wired up – well, and tested piece by piece carefully in case I accidentally shorted something.
Next weekend I head to Issaquah near Seattle for my first face to face WG21 committee meeting since the pandemic grounded all travel. I will be there for a week presenting four of my papers. I hope to get the mains lighting installed and fully functional in the shipping container before I leave, and I don’t have any ambitions this week beyond that.
One other big advancement which I think I got done is to get a mortgage application submitted. It took many more months than I could have ever expected to submit a mortgage application. Firstly we were blocked on my accountant to produce company accounts, which they didn’t complete until December. Then oodles of documentation had to be collated, because the bank/broker seems to think more information is always better, even though they clearly get overwhelmed by the complexity of it all. Then everybody seems to take at least a week and a poke from me to respond to say they need even more documentation. Nobody has actually confirmed that any application has been submitted yet, so I don’t actually know if it’s in. But that alone has sucked down a month, and the banks likely will take another six weeks to decide anything. So it’ll likely be mid April before I know whether a build commencement in 2023 is possible, or whether we’ll have to wait until 2024.
I’ve been disappointed at the lack of quotes from builders, out of thirty requests for quote dispatched, I got just one to take it to Builder’s Finish; one to take it to Second Fix; two for the superstructure only. That’s it. Some builders were blunt enough to say their order queue is full, and a passive certified build is added hassle for them which they don’t need to bother with by sticking with NZEB builds only. Last year Ireland broke its annual record for new build commencements, so no doubt that pipeline is full. However planning permission applications fell off a cliff last year, so I’m guessing that pipeline should start emptying out by 2024.
So maybe if the build commencement does get pushed into 2024, it’s no bad thing overall in terms of cost and build times. Still rather frustrating however, one is within spitting distance now. Just not there yet.
Satellite imaging
Oh, I did do one other thing: I bought an up to date satellite image of the surroundings of my site. This might sound like not a big thing, after all hasn’t Google Maps been shipping high resolution ‘satellite’ images for years now? However, firstly, Google Maps uses really old images, its one for my site are at least twelve years old, and probably fifteen. Secondly, Google’s ‘satellite’ images actually come from an airplane because the 5 - 10 cm per pixel resolution images Google has are definitely top secret nation state spy satellite type of resolution, and an airplane is way cheaper than one of those, even if Google were allowed access to one.
In fact, buying aerial images is straightforward in Ireland. If you go to https://store.osi.ie/aerial-imagery-map.html, for €31 you can get a 30 cm per pixel resolution image for my site. Unfortunately, the most recent images available are from 2014.
Up to date satellite images are until just very recently not for the retail customer to purchase. The reason is cost, you can purchase them yes, but they require you to buy the whole sweep, and it’s at least several tens of thousand of dollars. On HackerNews I noticed that a startup called SkyFi finally had the bright idea of parcelling out those images in much smaller chunks for much smaller prices, ones affordable to the likes of me.
SkyFi’s pricing for my location was US$4 per square kilometre, with a minimum order of five square kilometres, so US$20 before VAT. For that, they would provide an image which captures detail over 75 cm taken in the past year. They don’t have any higher resolution images for my location unfortunately, though they do have a multispectral image a few years old with less resolution, and much higher minimum order area.
Here is SkyFi’s 75cm resolution image from 26th March 2022 11:16am cropped to match the size, but definitely not the resolution, of Google’s Maps:
Undoubtedly the SkyFi image is a pale comparison to the Google Maps image. There are some sensor blowout issues from brightness, and the lack of detail is quite disappointing.
I reckon there is approximately between a ten and a sixteen fold difference in resolution, though note that your web browser may actually be showing you a reduced resolution edition of the Google Maps image, as it is quite high resolution. At 75 cm per pixel, my site there is a mere 53 pixels wide and 52 deep, the house might be 25 pixels wide by 10 high. This is probably such low resolution as to not be particularly useful, true, but much more important here is that I can get a recent satellite image of my site for a small fee. That is very new, and very exciting, because it really won’t take long at all for pixels per dollar to dramatically improve. Commerical imagery already comes in 15 cm per pixel which is 5x better, albeit SkyFi doesn’t offer that yet, their current best is the next grade up after 75cm which is 50cm which I doubt is worth the money for my use case, as it’s only one third better. SkyFi say they’ll be selling 10 cm per pixel images by 2024 however, and that’s not far off Google Maps’ quality.
Out of interest, I looked into how much it would cost to get a man with a drone out to take aerial shots. €550 apparently. For not much more, I could buy a mid range drone with 48 megapixel camera and capture my own aerial shots in glorious 5 cm per pixel resolution.
Apart from the thrill of seeing from the sky the fruits of your labours, I suspect aerial images aren’t actually that useful for a site as small as mine. Well, at least compared to a professional survey anyway. I have a measuring tape long enough to traverse the entire site, so if I need to figure out if something is or isn’t where it is supposed to be, I can just roll out my tape and have a quicker and cheaper answer than any other solution. In that sense, these satellite images are pure navel gazing.
Am I going to spend another €25 inc VAT to get the latest satellite image when SkyFi updates it in April? To see the roof I erected, my shipping container, my portable cabin, and my other works? The chances are high. I like navel gazing!
You can see the four conduits coming out of the ground, with 16 mm2 cable sticking out the top. I’ve since fitted conduit to connect those up and back to the central services enclosure, though I haven’t got to fitting the 25 mm2 earth cable connecting those 16 mm2 earth rods yet. Each earth rod is four foot something long, and is at least a half a metre under the ground, so in theory especially given the high water table there should be a good earth connection here.
Given I was hand shovelling soil, I took the opportunity to do a soil test:
I make that:
- Total height of settled material: 56 mm, maybe minus 2mm for glass = 54 mm
- Height of clay layer: 2 mm (3.7%)
- Height of fine silt layer: 23 mm (42.6%)
- Height of course silt layer: 29 mm (53.7%)
Which is a silt loam soil, which ought to be very fertile and excellent for the vegetables I intend to grow. Drainage will be an issue however, during all that rain during November it was extremely clear that the site struggles to clear all that rainwater on its own – indeed, the house being built next door the entire of their back garden flooded, and had at least four inches of water for the entire month. The only reason it didn’t flood into their house was because their back garden was overflowing onto my site, which won’t be possible after the wall goes in. I don’t know what they’re going to do about that, but I suspect that the builder will consider it not their problem and the new occupants will get an unpleasant and expensive shock after they move in. If I manage to catch them visiting I’ll mention it to them.
I suppose the only other thing where I’ve made visible progress is wiring up the intelligent solar charge controller:
Top right is the ESP32 board with PoE ethernet, though in fact it’ll be powered from the USB socket on the dumb charge controller. Underneath it is a 3Ah LiPo battery, so the ESP32 still has power when everything is turned off. Just above it is a small low power 120 x 80 OLED screen. There are four MOSFETs for PWM dimming a five metre dual column RGBWW LED strip for lighting the inside of the container; there is a relay to disconnect the dumb charge controller from the battery; there is a voltage and current sensor so we can measure how much is left in the 12v battery; and finally there is a coin battery powered real time clock, as the one in the ESP32 is well known to be garbage. I’m still missing a few bits – there is a 12v to 24v boost converter, as the LED strip runs off 24v; a tilt sensor still needs to go on to detect if the container is being lifted by a crane; I still await from China a 3.3v to 5v boost converter, as the relay needs 5v, and we need to be able to turn it on using the LiPo battery when everything else is off. The LED strip is able to draw 90 watts at max power, so max current draw should be under 8.5 amps, hence I fitted a 10 amp fuse.
The 12v battery was reclaimed from an old UPS, I reckon it might have 50-60 watt hours capacity in it, which means I can’t have that LED strip on at full brightness for more than thirty minutes. This kinda sucks, however I hope it won’t be too long before I get mains power, and then the 60v DC mains circuit can power the LED strip and not require anything from the battery. The battery then can serve its primary purpose as a theft alarm even if somebody cuts all the wires into the container before stealing it.
House build spend
As I mentioned last post, I am now out of money so my cash burn rate has significantly dropped, though I did take advantage of the November sales (Single’s Day on Aliexpress and Black Friday over here) to pick up some longer term items needed at discounted prices. My last update on this went up to 1st October 2022, and this will be up to 1st January 2023:
- Spent: €164,463
- Committed to be spent soon: €14,341
- Current three month averaged spend rate: €1,377 per month
The four biggest ticket items in the past three months were: (i) Quantity survey (ii) Three tonne digger hire (iii) Wifi and fibre optic router boards for house (iv) Conduit, concrete screws, conduit mountings.
We’re heading off to the US tomorrow to see Megan’s family, so no more work shall occur until we get back. Have a Merry Christmas everyone!
I did get this finished in the past month however:
Strictly speaking, it’s actually the last four months because I bought the roof sheets in July, and it’s taken me well over two months to get the wall painted and to erect the support beams. As much as that was painstakingly slow, the beams have been designed to aid installing wiring and lighting, so they will enable subsequent stuff to go quicker. And, I now have a sheltered place for the solar inverter and all the other wiring to be installed (plus, that roof is absolutely super when a rain shower comes in, you can go hide under it until the rain passes, or get a lot of cutting and sawing type of work done when it’s raining full belt). The roof panels cost a few hundred euro, as did the paint, as did the raising of the wall. I reckon €1,500 or so of materials has gone into just that little bit of wall. Crazy for what it got me, but that’s the cost of construction I guess. And still vastly cheaper than getting somebody in to do the same work.
I’ve got plenty to keep me busy left to go – conduit needs to be installed for ethernet, 16A three phase mains, local earth and 60v DC to run to the far end of the existing wall. RGBWW outdoor strip lighting needs to be installed under the roof, after waterproof flashing between the roof and the wall is installed (I await a few dry days in vain I suspect). I have two watertight large plastic boxes for wifi APs for each end of the wall, and two security cameras to raise each end of the wall. I have a 12v battery powered alarm system to install, with a roof mounted trickle charging solar panel, and some 50 watts of alarm siren to light the place up if somebody tries to burgle it. At some point the solar inverter will turn up, that and the lithium batteries will need installing. I have months of work ahead of me before I can move in as my daytime work office.
I’m currently working on that 12v system for the container. It will consist of one of my Olimex ESP32-POE boards, a 5m RGBWW PWM controlled strip to light the inside of the container, an infrared movement detector, oodles of 12v siren capacity, a 20w solar panel, a solar charge controller, and an old 12v battery reclaimed from an old UPS which after me reconditioning it appears to still be able to hold 50-60 watt hours. The total cost for all these parts was under €100, which puts the expense of that wall into perspective. Still, I need to wire this up and get a firmware onto it.
The solar charge controller is really cheap and dumb. It draws off the 12v supply to run itself obviously, but it doesn’t turn itself off at nighttime, so it therefore drains the battery for zero good reason. I am therefore going to insert a mechanical relay between it and the battery, and have the ESP32 physically disconnect it each night and reconnect it in the morning. To achieve this, the ESP32 obviously enough needs to know when the sun will rise and when it will fall. Complicating this is that one of the many quirks of the ESP32 is its shitty real time clock, it will happily wander by minutes per day if the CPU is running, but worse, if you put it into deep sleep then it wanders by hours, and obviously I intend to deep sleep it when it’s not needed. To solve the real time clock problem, I bought a cheap DS3231 i2c connected battery powered real time clock which won’t drift more than a minute per year, so now the ESP32 always knows the correct time and date. The next problem is calculating when the sun rises and sets for a given latitude and longitude and date.
The correct solution can be found at https://gml.noaa.gov/grad/solcalc/solareqns.PDF, and it involves lots of maths because the earth isn’t quite a sphere, and it doesn’t go around the sun linearly – it actually slows down and then catches up across the year as the other planets (mainly Jupiter) tug on it. The ESP32 only has hardware floating point for adds and multiplies, everything else is hideously slow, so whilst that is a once per day calculation, I was hoping to derive a simplified estimator of sunrise and sunset for a latitude by assuming that the earth is a sphere and its motion around the sun is linear. I think I have derived a correct simplification:
# Turn your latitude in degrees into radians
latitude = (52.129877 / 180) * math.pi
const1 = 12 / math.pi
const2 = -math.tan(latitude)
const3 = 0.40910517666747085283091311613373
const4 = 2 * math.pi / 365
def calc_sunrise_sunset(days):
diff = const1 * math.acos(const2 * math.tan(const3 * math.sin(const4 * (days + 284))))
sunrise = 12 - diff
sunset = 12 + diff
return (sunrise, sunset)
# Normally pass this time.gmtime().tm_yday
sunrise, sunset = calc_sunrise_sunset(81)
# Must be true at the equinox
assert abs(6 - sunrise) < 0.01
assert abs(18 - sunset) < 0.01
This costs a sine, a tangent and an arc cosine plus three additions and four multiplies. The trig functions will be implemented on the ESP32 as a sequence of Chebyshev multiplies and adds (maybe six of each per trig function?) giving maybe 20-25 adds and 20-25 multiplies, so I’d consider the above fairly optimal. It’s not the most accurate estimator as it is too simplified, it can be up to 45 mins out depending on the time of year, but for my needs I think it’ll do just fine.
Despite what I said earlier about not having any more money to spend, I did take advantage of the Chinese Singles Day sales to pick up an Anycubic Kobra Go 3D printer this past month for under €200 inc VAT delivered. This is a self assembled printer capable of printing in PLA, PETG, TPU, ABS and low temperature formulations of Nylon (just about!) to dimensions no larger than 20 x 20 x 24 cm. It’s very much conventional bog standard cheap 3D printer tech, albeit a very well put together package thereof, so you get what you get for the price point i.e. it’s going to be very slow if you want quality, but all the parts and firmware are the same as any other printer in this price bracket, so it’ll be very easy to maintain and/or upgrade.
My experiences with the printer so far have been extremely positive, given its price. It is perfectly calibrated i.e. 1cm in X, Y or Z on the computer is exactly 1cm in the print. I have been feeding it extremely cheap filament – as in, rock bottom cheap – and the results have been very good considering that. Here is the owl print Anycubic supply with all their printers for you to test its assembly:
This is the owl straight out of the machine uncleaned, and the small amount of ringing at the bottom came off very quickly with a knife. Look at the ears – that is a mightly overhang, yet they do not sag despite the printer going at maximum speed with a 0.2 mm layer thickness.
I’ve had the printer for a few weeks now, and this weekend after much earlier testing to ensure this final test wouldn’t be a waste of a lot of filament, I set it printing this over thirty hours:
(Note the roughness of the bottom half, and the smoothness of the top – a classic example of the effects of moisture getting into the filament. It doesn’t matter for our use case, we’ll be enrobing this with several layers of epoxy resin, which will produce a glossy ceramic-like finish)
That’s the full height (24 cm) and most of the full width and depth (20 cm) using a single thickness over-extruded wall i.e. if anything, absolutely anything, were out of balance or wrong this print would have failed. The nozzle is 0.4 mm wide, but I had it print this by depositing 0.6 mm of material (over extrusion). Obviously, this is highly prone to slipping unless the material is placed absolutely perfectly on top of the preceding material down to micrometre accuracy, and as you can see, the printer is capable of that, albeit it had to run at half speed. There was almost no fluff – only a marginal amount along the Y axis due to the bed slinger – no ringing, no holes, no deformations nor errors of any kind. If you’re willing to wait, this printer is supremely capable at half speed, the quality is second to none. What a more expensive model will buy you is speed, the 3D printing world is currently abuzz about the Ankermake M5 which costs 5x-6x more than mine, but also prints maybe 8x faster using proprietary nozzles and other fancy technology. Indeed, after a few years of stagnation, it looks like a real technological leap forwards is currently happening in consumer 3D printing – high end features typical in industrial printers are increasingly appearing in sub-€1000 consumer printers, and it’s not hard to imagine that in a decade from now for under €500 you’ll be able to buy a 3D printer capable of all non-organic materials printable under 350 C which can ‘plug and play’ turn out a fifty litre print within 24 hours with almost zero chance of print failure.
Speaking of print failure, I ended up spending as much as I did on the printer on preventing print failures. I bought a large sized filament dryer, an enclosure and vacumn seal bags with dessicant for the filament reels to try and keep the moisture out of them. Where I live in the south of Ireland typically has relative humidity exceeding 80%, and I’ll be keeping the printer outside in the garage where it is particularly moist (right now, condensation drips constantly from the roof!) as you can’t safely print ABS indoors due to the fumes produced. I’m hoping that between the filament dryer and the enclosure I can get away with printing in the garage, I certainly don’t have the space here inside the rented house.
I mentioned earlier that I very much cheaped out on the filament. Most filament if bought in Europe costs €20-30 per kg, and if bought from China once you add in the postage it is usually around the same for non-commercial quantities. Bulk filament manufacturers such as https://gst3d.eu/ will get you down to €15/kg if you buy > 10 kg at a time, but if you really want to get down to cheap cheap cheap filament then YOYI filament from Amazon kinda occupies the whole market below €15/kg in Europe at least. I picked up my YOYI filament for €10.50/kg delivered, to reach that price I had to buy 4 kg but I was allowed to choose any colours or materials (PLA, PETG, TPU, ABS) I liked. Remarkably, YOYI charge the same for PLA, PETG and ABS, only the TPU is more expensive.
Based on internet reviews, YOYI filament can deliver excellent results but it is finickety and intolerant. I’ve only had a bit of experience with it, however I can confirm other reviews that for the grey PETG, you need >= 230 C nozzle with a >= 80 C bed for the first layer if you want bed adhesion after which you can dial back to 210-220 C nozzle to reduce stringing. It’s kinda annoying having to hang over the printer until the first layer is done, so when Raspberry Pis return into stock I hope to kit out the printer with a Wifi Raspberry Pi based print manager and then I can watch and control it via Wifi connected camera from within the warmth of my home. Still, for a one third reduction in the cost of running – given the amount I expect to print – I’ll take it.
This brings me to why I bought the 3D printer at all – I have found myself increasingly needing custom plastic parts, and having to work around the lack of them is costing me time and money. I intend to teach myself over the next few weeks how to design end to end a 3D print – so far, I have taken other people’s designs from the internet, sliced them for my printer using my own hand and judgement, and so far so good it appears I understand the tradeoffs given that all my prints to date have been successful. But where I need to get myself is the capability to design and print say a durable custom waterproof case for an assembly of ESP32 automation board with various modules and sensors, which is almost certainly going to require PETG or ABS. PETG is the same stuff from which they make disposable drink bottles, so it’s quite durable and capable. However, ABS has the unique ability to be postprocessed using acetone i.e. if you ‘paint’ an ABS print with acetone it ‘melts’ the plastic, which means you can seal the print against everything other than organic solvants, which includes the rain. Unfortunately, printing ABS is both tricky and toxic, so I’m semi looking forward to attempting a print using the cheap cheap cheap YOYI ABS filament spool I bought, maybe next weekend or the weekend after.
Here’s an example of the kind of custom plastic part I need, these were printed a few hours ago in that YOYI grey PETG using a 0.1 mm layer rather than 0.2 mm to improve vertical strength and resolution:
Despite the low post-first-layer print temperature there was a bit of stringing, but it was very easily cut off to yield the above. These brackets aren’t pretty, but they are functional – they clamp any 65 mm wide circuit board, which is most of the Olimex stuff. That means I can finally mount the Olimex stuff onto an acrylic board, and that in turn means no more rat’s nest jumble of wires and modules with so much that can short or get loose or get tangled. Everything going forth will be fixed onto a solid board and immovable, woo hoo! Down the line, as mentioned earlier, I hope to print waterproof custom cases, but for now even these simple printed mounting brackets will be a big time and hassle saver.
Here is the board I dropped about a grand upon buying fifty of them plus convenience breakouts:
You may remember that a year ago I was not keen on the ESP32, as it is markedly inferior to the STM32F4 in terms of quality and quirks, albeit with bucket loads more features. So why change my mind now?
First reason is cost: when you’re spending a grand of money, saving a third of it is a good move. In my Q3 2021 review post, I reckoned a STM32F4 could be had for €12 inc VAT and a Raspberry Pi for €20 inc VAT, but that didn’t include the PoE adapter which probably can’t be had for less than €10 inc VAT. The Olimex ESP32-POE-IND (with industrial grade components) is under €15 inc VAT if you buy fifty of them, which is nearly one third cheaper.
The second reason is convenience: These boards have already wired a bunch of stuff together for me into a single PCB, which saves me having to wire bits together. Its premounted UEXT connector exposes the UART, I2C and SPI pins plus 3.3v power which will save me soldering in i/o pins in many cases (I also bought a bunch of UEXT daisy chain boards, they were very cheap and let me use dupont wiring instead of hand soldering and crimping). They’re correspondingly more compact, which means I can use smaller much cheaper waterproof enclosures, saving even more money.
The third reason is that for all the things the ESP32 is bad at e.g. Analogue to Digital conversion, or keeping accurate time – I can cheaply buy an I2C peripheral which does have quality ADCs, or quality time. And it’s only rarely I’ll need to do that in any case as the majority of those boards will go on PWM control of RGBWW LED strips. The PWM in the ESP32 isn’t as good as the one in the STM32F4 or the Raspberry Pi, but it can still push 15kHz, which is plenty. One of these boards can PWM up to three LED strips (four channels each, therefore twelve I/Os) if I solder on the i/o headers, but something very attractive is that the UEXT header alone is sufficient to drive two LED strips no soldering needed.
Finally, the ESP32 alone has rather the killer app for my use cases: https://esphome.io/. But more on that shortly.
The Olimex ESP32-POE board
Olimex are a Bulgarian manufacturer of IoT and embedded boards, and have been around for a few decades now so they’re well established. I believe their ESP32-POE board is their most popular by some margin, for some reason nobody else makes a PoE powered ESP32 board for under €31 (the LilyGo TTGO) and more usually €65 (the WESP32; the official Espressif dev board). The nearest cheap equivalent is the WT32-ETH01 which combines an ESP32 with Ethernet, but you won’t get one for under €13 inc VAT, and it doesn’t include PoE. So no wonder that the Olimex board is so popular!
For under €15 inc VAT the feature set is pretty good. You get a 100 Mbit claiming Ethernet port which can draw up to four watts of power from PoE. If you need more, there is a micro USB port through which you can supply up to ten watts. There is a lithium battery connection with a 100 mA charge circuit, this will accept a 3.7v LiPo battery. If the battery is fitted and charged, power can be removed and the board will keep running, however only 3.3v TTL is supplied, the 5v power rail disappears. So you need to make sure any relays you use are fully 3.3v based if you want them to work on battery. Finally, you can connect a SD card if you need local storage, though for me ethernet is where data ought to be stored.
The ESP32 running full belt minus wifi at 240 Mhz will consume about 50 mA, peripherals can’t draw more than 250 mA if on PoE, and perhaps less than that if on battery. An idling ESP32 might draw 4 mA, therefore a 3000 mAh battery could run the device for between 30 and 750 hours (one month) assuming board power overhead of 20-50 mA. If you can put the device into deep sleep, that draws only 0.1 mA, which could be up to 30,000 hours (or over three years)!
I’ve been running mine for several months now, and it has been reliable:
As you may be able to read from its little OLED display, it has three sensors, two are NDIR CO2 sensors and one is a IAQ + Humidity + Pressure sensor. Every few seconds the readings are posted by the board via REST to an Influx DB over the network, and it slowly pulses the LED from fully on to off by PWM. This has been running for a long time now on a ESPHome built firmware, it looks to me as reliable as the Devantech board. You may also note the lithium battery to the left, I have a 3000 mAh unit in there which I reckoned to be the best bang for the buck in terms of capacity to price ratio. Here is Grafana rendering the last week of InfluxDB records:
I guess my only real issue with this board is how hot its PoE DC to 5v circuit runs:
Yup, about 80 C. That’s well below the temperature rating of the components – especially the industrial grade ones which mine are – and in fairness this the eighth revision of the board is the least hot of any of the preceding revisions. Still, it kinda feels like wasted power consumption for no good reason other than reducing cost.
Thanks to this being an open source hardware design, I discovered that the DC-DC stepdown chip is the TX4138 whose datasheet can be found at https://datasheet.lcsc.com/lcsc/1811141153_XDS-TX4138_C329267.pdf. It claims an 84% efficiency. Assuming it’s a linear regulator taking the 5v to 3.3v and therefore burns as heat 33% of the current the ESP32 uses, a 100 mA draw by the ESP32 at 3.3v (one third of a watt) would be 133 mA of 5v, or two thirds of a watt. That turns into a minimum of 0.8 watts of PoE power, which is a best case efficiency of 41%.
Olimex seem to think it draws around a watt of PoE power if you don’t attach energy hungry peripherals, maybe 1.5 watts if the wifi is maxed out. I guess that therefore is +50 watts of constant background heating in my home.
BTW, one watt at €0.30 per kWh is €2.63 per year, so if the board costing €31 were ideally power efficient, it would take 9.1 years to pay back the added cost over the board I chose. I very much doubt that that the €31 board is more efficient however, it looks to me to also use linear converters. The WESP32 specifically mentions in its docs that when running off PoE it will get hot, and I couldn’t find any information on the heat output of the official Espressif dev board. In any case, high efficiency power conversion costs money, and the repay time is unlikely to be worth it for such low wattages.
ESPHome
In my Q3 2021 survey I spoke a lot about ecosystem depth and breadth greatly favouring the STM32 over the ESP32, after claiming I had done a lot of research. Unfortunately I somehow missed the elephant in the room which is the amazing and astonishingly capable ESPHome which I assume at the time I didn’t realise just how amazing it is. And, a game changer in fact, as it made me go all in on ESP32.
ESPHome comes across initially as a dumb microcontroller convenience firmware for Home Assistant, and I think because of that I dismissed it from further investigation at the time. That was a mistake, thankfully realised in time. ESPHome is in fact so capable that you don’t need Home Assistant at all, you can create firmwares which work with other ESPHome firmwares to solve a problem without any central server to coordinate things. That is a killer application for my use case. Furthermore, you can bulk orchestrate the upgrading of the firmwares of many devices chosen by category with zero effort over ethernet, which is another killer application for my use case. Finally, while ESPHome has a limited selection of hardware with direct support, apart from a few mildly painful omissions it does have support for a wide enough range of hardware that so long as you only buy peripherals it supports, it ‘just works’ with very little added effort.
Here is the ESPHome configuration file for the board above. You feed this to esphome run and it will compile an ESP32 firmware for you and automatically write it to the board over ethernet.
esphome:
name: esp32-poe
platform: ESP32
board: esp32-poe
on_boot:
- light.turn_on: led1
globals:
- id: brightness
type: float
initial_value: '1.0f'
- id: increment
type: float
initial_value: '0.01'
influxdb:
host: 192.168.x.x
database: HomeSensors
username: x
password: x
interval:
- interval: 100ms
then:
- light.turn_on:
id: led1
transition_length: 0ms
brightness: !lambda |-
if(id(brightness) <= .0f) {
id(brightness) = .0f;
id(increment) = -id(increment);
} else if(id(brightness) >= 1.0f) {
id(brightness) = 1.0f;
id(increment) = -id(increment);
}
id(brightness) += id(increment);
return id(brightness);
ethernet:
type: LAN8720
mdc_pin: GPIO23
mdio_pin: GPIO18
clk_mode: GPIO17_OUT
phy_addr: 0
power_pin: GPIO12
domain: .nedland
manual_ip:
static_ip: 192.168.x.x
gateway: 192.168.x.x
subnet: 255.255.255.0
use_address: 192.168.x.x
uart:
# Must use UART1 as UART0 is used by the logger!
rx_pin:
number: GPIO36
tx_pin:
number: GPIO4
baud_rate: 9600
id: uart1
debug:
i2c:
sda: SDA
scl: SCL
scan: true
frequency: 400kHz
#spi:
# clk_pin: GPIO14
# mosi_pin: MOSI
# miso_pin: MISO
binary_sensor:
- platform: gpio
pin:
number: BUTTON
inverted: true
name: "Button 1"
# LED
output:
- platform: ledc
pin: GPIO33
id: gpio_33
frequency: "14648Hz" # 80 Mhz clock so 4096 steps
light:
- platform: monochromatic
output: gpio_33
id: led1
name: "LED"
# CO2 sensor
# Temperature, pressure, humidity sensor
bme680_bsec:
address: 0x77
sensor:
- platform: adc
pin: GPIO35
name: "Battery voltage"
- platform: adc
pin: GPIO39
name: "External voltage"
- platform: mhz19
co2:
name: "MH-Z19 CO2 Value"
id: co_1
temperature:
name: "MH-Z19 Temperature"
update_interval: 30s
uart_id: uart1
automatic_baseline_calibration: true
- platform: scd30
co2:
name: "SCD30 CO2 Value"
id: co_2
temperature:
name: "SCD30 Temperature"
id: temp_2
humidity:
name: "SCD30 Humidity"
id: hum_2
update_interval: 30s
address: 0x61
automatic_self_calibration: true
# - platform: bme680
# i2c_id: i2c0
# temperature:
# name: "BME680 Temperature"
# oversampling: 16x
# pressure:
# name: "BME680 Pressure"
# humidity:
# name: "BME680 Humidity"
# gas_resistance:
# name: "BME680 Gas Resistance"
# address: 0x77
# update_interval: 5s
- platform: bme680_bsec
temperature:
name: "BME680 Temperature"
id: temp_1
pressure:
name: "BME680 Pressure"
id: press_1
#on_update:
# write scd30.ambient_pressure_compensation
humidity:
name: "BME680 Humidity"
id: hum_1
iaq:
name: "BME680 IAQ"
id: iaq_1
co2_equivalent:
name: "BME680 CO2 Equivalent"
breath_voc_equivalent:
name: "BME680 Breath VOC Equivalent"
# - platform: ltr390
# uv:
# name: "LTR390 UV Index"
# light:
# name: "LTR390 Light"
# address: 0x23
font:
- file: "gfonts://Roboto"
id: roboto
size: 12
display:
- platform: ssd1306_i2c
model: "SSD1306 128x64"
rotation: 180
lambda: |-
it.fill(Color::BLACK);
it.printf(0, 0, id(roboto), Color(255,255,255), TextAlign::TOP_LEFT, "CO2 A: %.0f B: %.0f", id(co_1).state, id(co_2).state);
it.printf(0, 13, id(roboto), Color(255,255,255), TextAlign::TOP_LEFT, "Tmp A: %.2f B: %.2f", id(temp_1).state, id(temp_2).state);
it.printf(0, 26, id(roboto), Color(255,255,255), TextAlign::TOP_LEFT, "Hum A: %.2f B: %.2f", id(hum_1).state, id(hum_2).state);
it.printf(0, 39, id(roboto), Color(255,255,255), TextAlign::TOP_LEFT, "IAQ: %.2f %%", (id(iaq_1).state/5.0f));
it.printf(0, 51, id(roboto), Color(255,255,255), TextAlign::TOP_LEFT, "Press: %.0f", id(press_1).state);
# Enable logging
#logger:
# level: VERBOSE
# level: VERY_VERBOSE
# Enable Home Assistant API
api:
password: "xxxx"
reboot_timeout: 0s
ota:
password: "xxxx"
# Enable webserver
web_server:
port: 80
As you can see, it’s basically YAML describing the board and what is connected to it and where. Rather usefully it autogenerates a live web page on a web server served by the board, just like the one I made for the Devantech board:
I haven’t really exercised the support for writing out complex logic much yet – you can easily do stuff such as ‘if humidity exceeds 80%, switch relay on’ – but I shall be doing so when I set up my 12v battery installation. I have bought a small 20w solar panel and a cheap low end 12v solar charger which isn’t very clever, so I will also be adding one of these PoE boards with a lithium battery and an added high quality time and date peripheral over I2C. It will calculate sunrise and sunset from the current date, and physically disconnect the solar charger from the battery during night times to prevent it sucking power out of the battery. I then will have various things hanging off the 12v battery, specifically 24v LED strip lighting and an alarm system so if anybody tries to rob me, I can light the place up and set off loud sirens, and it’ll be completely independent of mains power so even if they disconnect or cut the power lines going into the container, it won’t help them.
Once I have the security system in place I can start filling the shipping container with expensive stuff, currently I can’t take the expensive stuff out there as I can’t risk it – even though the shipping container has a CEN class 4 lock on it (the highest CEN grade is 6, and those locks cost half a grand or more each), and it would take a cutting torch to get in there, the easiest way to rob it is simply to steal the whole container and I have no protection against that until I get my battery powered alarm system installed.
), but my wonderful sister Aoife did manage to get me some groundworks done by pulling favours:
Due to being so busy, they could only spare me two days and those cost me more than six grand, but it certainly leapt me forwards a bunch. I got services ducting extended from their entry points at the driveway over to the western wall, specifically mains electricity, broadband, mains water and sewerage. 300mm of hardcore, crushed rock and a thin layer of gravel have been laid down. They felt that my wall was likely to fall down, so they pinned it with a new support column, which was made wide enough to plinth the solar inverter whenever it arrives, and they built up structural support for the far end which I had left open because concrete foundations needed putting in. That, in turn, meant I could seal and finish the ESB meter box, and it enabled me to install a 40ft shipping container:
This shipping container unblocks a vast array of other things as I finally have secure storage onsite – it was a critical blockage point, and without it not much else could proceed. As the photos show, it is a ‘new’ (once used) unit manufactured in China and it did a single crossing from China to Europe. It was expensive at seven grand, however used units are currently costing €5.5k, plus this specific unit is a High Cube rather than standard i.e. it is taller, and has 14% more space internally. I reckoned for the difference in it the slightly extra cost was worth it for a non-banged up unit that is guaranteed completely clean on the inside (which as the photos show, it is).
Most people now ask why buy instead of rent, as is normal for those in Ireland building a house? The reasons are: (i) I don’t expect to build for at least a year probably two as I wait out the coming recession, and rent vs purchase swaps value at around nine months (ii) I want to modify the container to add internal lighting, an alarm system, and forced air ventilation (iii) as a brand new container with modifications highly desirable by other Irish home builders, it should recoup a large portion of its cost when I sell it years from now.
This coming Monday the portable cabin should arrive, and once connected up that will finally give me a working toilet on the site, which would be amazing. And somewhere to shelter when it rains which isn’t my poor car! I’m going to wire it up so it’ll run off my petrol generator for the time being, then I can make cups of tea etc whilst sheltering inside from a deluge, which are common in Ireland. I have a small heat recovery ventilation unit coming for it, once I get electricity being generated onsite it should stop the cabin going mouldy as portable cabins tend to do if not frequently used.
Last few months I’ve been having to load everything in and out of my car per journey, plus my small rented house has been filling up with stuff purchased for the house. Being able to get some of that (the non-expensive stuff) out stored onto the site where it’s exclusively used would be most convenient and save considerable time not unpacking and repacking the car each time. Also, I was now able to order the solar inverter, the lithium battery storage, and the solar panels for the house as I finally have somewhere to store them. When they arrive (likely months from now, there are severe shortages of these items) I should be able to mount twelve of the panels onto the roof of the container, and that should generate 15 kWh of electricity per average day in December. That should be enough to space heat the portable cabin and run my work computer, and every month which isn’t December will generate far more electricity. I could just install mains electricity, however why pay its standing charge when I don’t have to?
In the past month or so I’ve found weekends and mornings before work to finish the construction of my wall, and because its mortar is so unstable due to my mistakes I needed to ensure it wouldn’t wash out with the winter rain, so I got a base coat of paint onto the inner side and I applied a cement and water paste to the outer side to seal the wall.
I’ll get two full coats onto the inner side before winter I would hope, weather permitting, and raise a one meter wide roof along ten metres of it which corresponds to the future outhouse. I’ll then be ready for whenever/if the solar inverter turns up to mount it onto the wall (you can see its plinth above).
Despite the delays in delivery of lots of critical bits, I suspect the single biggest constraint going forth will be my free time. I hope to invest as many mornings before work and weekends as I am either capable of or am allowed to. I’ll get there eventually, and once ready it will become my new place of work so I no longer need to work from home.
House build spend
My last update on this went up to 1st August 2022. I can tell you what the spend will be on the 1st October 2022, more or less:
- Spent: €157,769
- Committed to be spent soon: €15,448
- Current three month averaged spend rate: €12,096 per month
The four biggest ticket items in the past three months were: (i) Portable cabin (ii) Shipping container (iii) Groundworks (iv) Architect fees.
That is obviously rather a lot of money and most definitely is not sustainable – I am quite well paid, but burning twelve grand per month above living costs is very far beyond my means. The only good news is that this three month cash burn rate probably has peaked and will not become so high again until the main house build gets its deposit paid for it. And, I suppose when the inflation rate causes your cash to lose over 10% of its value per year, one of the best value preserving things you can do is get rid of your cash by swapping it for things which don’t lose their value as quickly. I certainly have achieved that anyway!
Apart from all the bitty stuff of wiring things up onsite etc, the next big thing is getting a mortgage application submitted. It is blocked on my company accounts and my 2021 tax return being completed. Hopefully that should happen soon, and we can set that ball rolling. Current backlogs mean we shouldn’t find out what our Approval in Principle value will be until maybe late November. Depending on what max borrowing limit they return, we might start searching harder for builders, or we might kick the can down the road for a year before trying another mortgage application and hope the house building cost proposition has become less insane. We shall see!
The reason I wish to speak of this phone is that my Samsung Galaxy S10 got itself into a boot loop on the first day I went on annual holiday in England, and given that I have ‘stuff going on’ right now and that my holiday rental has no wifi, I needed a burner phone capable of acting as a wifi hotspot for my laptop. I went to the local Tesco Superstore and bought the cheapest PAYG phone they had capable of a wifi hotspot, which was the Alcatel 1 2021 edition for £30 if bought with £10 of credit, which gives me 20Gb of data for the next month (a good chunk of which I have already used in a few days).
The Alcatel 1 phone has a terrible phone experience. It runs Android 11 Go, the edition of Android for low end phones, yet using it is like treacle. Assuming its touchscreen registers your tap at all, it can take up to fifteen seconds to react to your tap because it it so underpowered. It is capable of 4G, but lacks the processing power to download more than 250 Kb/sec or so due to its hideously slow storage. It has a TN panel display with almost zero readability off centre, it cannot be read at full brightness in direct sunlight, and it regularly misreads taps on the wrong items because you tapped sooner than it was ready for. Browsing the web on this device is truly horrible. Apps with any weight to them e.g. WhatsApp spend most of their time appearing to have hanged before eventually getting there. Even non-internet phone calls the phone can take so long showing you the accept or reject display that the call may time out.
The frustrating thing here is that with just a few minor tweaks, the phone experience would be acceptable. It has a quad core ARM Cortex A53 at 1.4Ghz, sure this is slow, but I know from my Android TV sticks that the experience doesn’t need to be like this. I would suspect that the storage is particularly shit, and if they fitted slightly less shit storage, the experience would be considerably better.
Equally, probably the £30 price tag would then get exceeded. I note that the next cheapest wifi hotspot capable phone that Tesco do is the IMO Q4 Pro for £60, double the cost of the Alcatel. For that you get on paper similar hardware specs, yet online reviews say it doesn’t suck anything like as much. I suspect that the cost difference went into storage and the higher resolution screen which is also IPS rather than TN, plus a bigger battery and better cameras. It sounds like an actually usable device as a daily driver, not masochistic like the Alcatel 1. Equally, for just £20 more again (£80), you can get yourself a Motorola e20 which has ARM Cortex A72’s and very decent hardware specs for that price point. As in, hardware specs which would have been near flagship apart from the screen only two years ago, though specification isn’t necessarily quality, albeit online reviews are extremely positive for the e20 for its price point.
As much as I am ragging on the Alcatel 1, I would have to say it is a whole lot of phone for the money. It does have a working GPS and Maps navigation, though I note that one’s position randomly wanders by about 20 metres, it does work. It does take pictures both front and back which are better quality than you’d think for a device at this price point: here is the ocean outside our rental plus a picture of me with Julia:
Those are not half bad considering. The latter picture was taken facing into sun, yet it did a surprisingly good job with applying dynamic illumination to the dark parts of the image.
Furthermore, it does make and take phone calls. You can send messages with it. It is a surprisingly good wifi hotspot, I get far better speeds from my laptop than it itself can do, I suspect because the wifi hotspot to 4G connection doesn’t touch its crappy storage.
Given that a decent dedicated wifi hotspot with 4G capability costs a lot more than £30 and doesn’t come with a battery, on that basis the Alcatel 1 is the best battery powered 4G wifi hotspot on the market by a long mile. On that assessment, this is a great bang for the buck device. Equally, I really hope that the next revision of their hardware chooses better storage. It would make one hell of a difference in usability as a phone. For now, this phone is very much a ‘little engine which could’ experience, as in, you’ll get there eventually, but you’ll be waiting a while.
(In case you’re wondering why I am buying the house solar inverter and batteries now which seems rather premature, it is because I need three phase demand in order to get the ESB to install a three phase mains supply. So I need these now if I am to avoid a second ESB mains installation charge, saving me €2k! Besides, the inverter actually lives outside, I’ll be mounting it on the wall I built below)
I have had the power at least to personally progress the bits I can do myself. To that end weekend before last I built my first concrete block wall in more than two decades:
I also got the temporary fencing raised at vast expense (I bought the fencing new rather than rented, it made more sense given how long I expect the build to take):
Whilst it was great to actually get stuff done instead of never getting replies from anybody, I am not the right person to be building concrete block walls! I was so slow at it that too much air got into the mortar mix from the concrete mixer, and that in turn made the mortar powdery rather than solid. It would eventually wash out with the rain, and then the wall would fall over, however I will be painting it well before then. It’ll do. I might also add that I pulled most of my upper body muscles, and their tendons all swelled up from the trauma. It was a great way of feeling old, unfit, and somewhat past it.
House build spend
It’s been three months since I purchased the site, so I can finally issue the first house build spend update! This will be a rolling spend average over the preceding three months, it will give an idea of the ‘velocity’ of the money flying out the door.
Up to 1st August 2022:
- Spent: €121,481
- Committed to be spent soon: €14,942
- Current three month averaged spend rate: €10,494 per month
This excludes the cost of the site, which was €90,000. The four biggest ticket items in the past three months were: (i) Stamp duty for land purchase (ii) Professional fees (architect, planning consultant, solicitor) (iii) Temporary site office deposit (iv) Aliexpress order of electronics for site lighting and three phase AC to DC conversion.
As I mentioned in the last post, I have spent well over €10k on components, a lot of which were sourced directly from China or directly from component manufacturers as bulk orders. It will cost me time to assemble this stuff by hand, and I need to wait many weeks for all of it to arrive, but it will save me a fortune overall.
Computer renderings of the future house
As mentioned last post, I invested several very late nights into mastering the conversion from Twinmotion into Unreal Engine 5. Unreal Engine 5 does a fair attempt at approximating ray traced quality images yet running at somewhat realtime speeds, at least on my anemic graphics card. I took these screenshots literally from the UE5 game I made of the VR model, and they are undoubtedly better than those from UE4, albeit with a few compromises over the (very slow) ray traced images that UE4 can generate. Note that this is derived from the February 2022 model, a number of small changes were made between this and what was submitted for planning permission.
Let’s start with an aerial view showing the entire site as a bird might see it. The lighting is accurate for April 1st at the exact latitude and longitude:
The main changes in what was actually submitted are additional skylights in the master bedroom (bottom right gable), and far more solar panels on the south roof plus more panels on the outhouse roof, as one needs explicit planning permission for so many.
These are the views from the main road looking at the front of the house and outhouse:
And these are from the back:
Walking in the front door through the lobby one enters the main living space, first entering the kitchen and then looking up:
Continuing from the kitchen past the greenhouse door into the living room:
UE4 didn’t do a bad job of rendering shadows, materials and reflections, but UE5 is very considerably better. Note that the ‘fur’ on the couch is actually furry. Note that the fridge’s stainless steel metal accurately mushes up reflections. Note that the kitchen island very nearly has perfectly accurate reflections, apart from a few rendering artifacts. And similarly, the shadows are complex, yet well defined and have an appropriate blur based on distance. All ‘for free’ in UE5.
Moving upstairs from the top of the stairs onto the bridge:
And then from the mezzanine looking down onto the Living Room – look at how well UE5 renders the reflections of the glass walls of the bridge, with the rightmost glass wall only having a partial reflection of the living room along its top because the bridge shadows the light from below:
The only other two rooms worth screenshotting are the Games Room and the Master Bedroom:
The games room image is another good example of UE5’s much improved light rendering. UE4 could only bounce white light without added effort from the designer, UE5 can bounce light based on the colours of the reflecting surfaces. So because the floor is brown, the light bounced onto the white ceiling is brown. If you adjust the season the sun beams land on the pool table and instead you get green illumination of the ceiling. All this also occurs ‘for free’ in UE5.
There are obviously more rooms in the house: kids bedrooms, shower-toilets, home office. But they’re bog standard rooms you’d see anywhere, so I didn’t bother taking screenshots. If you’d like to wander through the house in VR for yourself, I uploaded a copy to Twinmotion Cloud at:
https://twinmotion.unrealengine.com/presentation/xy1MGtTaLoDcB2Th Password: WLj1jHVq
How long that link will work for depends on Epic Games’ continuing generosity towards providing free cloud hosting and rendering for Twinmotion models, but for now it is good.
In a later post I’ll describe the POE-powered ESP32 based industrial microcontroller boards I ended up buying fifty of costing me over a grand, and how I plan to integrate those into Home Assistant. See you next time!
I’ll do a later post with pictures etc same as before, but tonight I’m going to be talking about some of the stuff I’ve bought for the house build. Since I took title to the land, I have spent as of today €9,952 on stuff for my land. Most of it will be deployed in some way before December, so it’ll be put to use soon. However other motivations were that inflation is eroding my savings at a cracking fast rate, so swapping cash now for inflation protected stuff I’ll need anyway is wise; also a lot of the stuff comes from Aliexpress i.e. China, and therefore takes two to three months to get to Ireland. So I order it now not needing it until October. I am not allowed to clear the site until September, but my hope is immediately after that to clear the site and do an initial set of groundworks to allow the placement of a temporary site office and a temporary storage container, along with installation and activation of services. I then order more stuff with long lead times, let them arrive on site over many months, and once the recession happens and prices become predictable only then will I start asking builders for quotes. Which is certainly a year away, could be more.
Next six months in any case will go on preparing construction detail by my team of designers. I’m currently seeking a Mechanical and Electrical Designer, and for the guy I have in mind I spent most of this weekend writing up this spec document, which others may find useful, so I post it here: My Passive House Plus Mechanical and Engineering Design Brief
My first new toy for the future house build: Security Cameras
Obviously a portable cabin and a storage container sitting on an empty site will be attractive to burglars, so I’m going to need security cameras to watch over them when I’m not onsite. The same cameras should watch over my house once completed, and be able to time lapse record its construction which I think will be fascinating to watch.
I didn’t have time to research what camera to buy, so I simply headed over to https://ipcamtalk.com/ and bought whatever they currently think is the best camera. Lazy I know, but I don’t have the time to do better. They think you ought to buy a 2K camera with 1⁄1.2 inch sensor, these have best in class unassisted night vision, however I really wanted a 4K camera for that twice better image detail. They have literally one choice in their recommendations for a 4K camera: the Dahua HFW5849T1-ASE-LED, also known as the Color 4K-X.
The forum has its own reviews of this model at https://ipcamtalk.com/threads/dahua-color-4k-x-in-depth-review.58999/ and https://ipcamtalk.com/threads/worlds-first-review-dahua-ipc-color4k-x-dh-ipc-hfw5849t1-ase-led-full-color-4k-camera.57013/, and yeah they basically think it the best new camera of 2021. As it is now 2022, the previous generation of cameras has seen price discounts and I picked up two of those with 3.6 mm lenses for €530 delivered to Ireland.
I gave them a quick test to make sure that they work. Here is my living space and kitchen in 4K crisp detail, mould patches and all in superb detail. Note how well the clock’s face resolves:
I now turn the lights off, and the camera automatically turns on its front warm white LEDs to add enough light to make a perfectly fine colour image:
It’s a little blown out by light reflection from the child’s playpen, but yes that’s a detailed, colour, image. Let’s turn off those illuminators:
What you can’t appreciate about the above picture is that it was completely black in that room. I could just about make out the window frames. I could literally see nothing else if I closed my laptop. Yet, this camera made that picture above. In case you’re wondering how blurry motion would be e.g. due to too much sensor gain, I waved my hand in front of it and whilst there was a little blur, it really wasn’t much. That camera just sees better than humans do, more like maybe cats do. It can see when you cannot.
Also note that it’s still a colour image! Note the detail in the tree branches outside – if you give any light at all to this camera, it’ll make good use of it.
I have to say that I am impressed. Mobile phones started seeing better than humans a few years ago, but it usually came with hefty motion blur and loss of colour and resolution. We’ve now reached a stage where there is undoubtedly a floor illumination level below which you get pretty much nothing, but as soon as you get slightly over that floor, you get oodles of detail and colour. Technology is still exponentially improving in this area for sure.
These cameras have a whole bunch of other clever stuff, like they can recognise humans and vehicles entering an area within the frame and send a request for a more powerful CPU to begin analysing the picture with AI. They have two way audio, they can play a really loud alarm sound. They aren’t particularly expensive either in my opinion for fully outdoor cameras. I am pleased with my purchase!
My second new toy for the future house build: Thermal Camera
You may remember my previous post when I received a FLIR One as a birthday present, and I found it very interesting to be able to see low infrared light. It had a resolution of 160 x 120, and via unofficial Android apps it was possible to unlock a 15 Hz frame update rate. Unfortunately, it didn’t last long, within a few months it had stopped responding when connected, so I sent it back for a refund, as it was not at all cheap if I remember rightly.
Thermal cameras are extremely useful for diagnostics of certain kinds of problem, and it is wise to record the construction of your house as it goes with a thermal camera, so problems discovered later can be more easily tracked down without expensively ripping everything out searching for a cause. To that end, I needed to buy a proper professional thermal camera, and I was displeased to find that FLIR cameras remain hideously expensive. Sure, you can recoup half the cost on the second hand market when you’re done with them, but that’s still a grand or so you’re blowing there.
I thus began looking for a ‘good enough’ substitute, some of which are US competitors to FLIR, some are Chinese marks. A recent new entrant to the thermal camera market is a Chinese company called HikMicro, which may or may not be related to HikVision (I suspect it’s a wholly owned subsiduary). HikMicro are most famous for their thermal hunting scopes which have been a huge hit especially in the US, because they deliver smooth thermal video, long battery life and high resolution for a small fraction of anything else on the market. No doubt like US made thermal equipment, there is a substantial military subsidy at work there making these cheap, but I’m not complaining as you’ll see shortly.
After much toing and froing, I eventually plumped for the just released – as in, last month - HikMicro Pocket 2 camera. This has a form factor just like a normal digital camera, with a large VGA resolution LCD touchscreen on the back, and it’s similarly point and shoot. It has a low infrared sensor of 256 x 192 pixels with 40 mK sensitivity and 25 Hz frame rate, a 8 MP visible light sensor, 50 degree wide field of view (most are far narrower), 16Gb of internal memory, dual band Wifi and USB C for transferring images, and is IP54 rated. You get all that for €700 inc VAT delivered at the time of writing, which sounds like a lot and it is. However the closest spec FLIR which is the E6-XT costs almost exactly three time more, and in terms of hardware spec is noticeably inferior to the HikMicro.
As is obvious, as a thermal camera the HikMicro is great – lots of resolution in there, often you don’t need the outlines generated from the visible camera to help recognition. You can set the temperature range manually before taking photos to ensure consistency across shots, and it’s tough to fault the hardware. The hardware is great value for money. The device feels fairly premium, a little plasticky, but definitely robust and the design is both well thought through and it looks swish as well. The waterproof membranes are a little obvious, but ensure you know it’s waterproof.
Let’s look at some images which diagnose a fault – here are the power wires leading up to my LED strip coving:
Let’s look at those in thermal:
Clearly, underneath the junction only one of the pairs of wire is carrying current, and it is mildly overloading the wire, raising it a few degrees above ambient. Above the junction, both wires are carrying current and therefore their temperature is ambient. I fitted two sets of wires precisely because I knew I would be pushing towards the current limits of this very cheap wire, and one set must have gotten knocked loose. As a diagnostic tool, this camera ticks every box you need.
Really it’s the software where things fall short, especially compared to the FLIR’s software. Don’t get me wrong, if you only care about taking thermal photos, this thing is great. A particularly cool feature is the device will do Wifi Direct and broadcast its image live, so you can connect in with you device of choice and watch the broadcast, same as you would with a smart TV from Windows or Android. From that you can then take pictures or record video. Nice touch! The HikMicro app slightly extends this facility with remote control, so effectively your phone becomes a remote control for the camera, which means you can stuff your hand with the camera down into really hard to reach into spots to get the view you want, then use your phone to remotely view and control it. This is a really great feature similar to FLIR’s premium Wifi capable models, and unlike the FLIR E6-XT, the HikMicro Pocket 2 is much less unwieldy to get into crevices. Just make sure you attach the wrist strap, lest it slip from your hand!
No, what I mean by the software letting it down is mainly these issues (note I tested the latest firmware at the time of writing, V5.5.25_220511):
- The UI is somewhat clunky and poorly designed, not really making good use of the touchscreen, and whatever software filter they have on the touchscreen makes it feel resistive instead of capacitive in terms of responsiveness (and yes, I did remove the screen protector it ships with, which does help).
- There is no way of recording video on the device itself, which seems extremely odd as it clearly can pipe video with audio to Wifi Direct.
- The hardware has a good microphone and apparently has a good speaker, yet the software makes no use of either. The microphone is piped into the Wifi Direct video just fine.
- The camera clearly captures and saves the whole visible image, 8 MP resolution if you choose that in the options. You can see that separate visible image when browsing the images on the device AND zoom into it by touchscreen swipe. Yet, you cannot access that image via any other means – you cannot export it, copy it elsewhere, access it. I tried the direct connection and the app. No joy.
- For some reason the fixed touch icons to the right of the screen are duplicated in the touchscreen UI, which seems redundant. Their backlight keeps turning off instead of dimming down, so you keep forgetting they are there at all.
- Exported images are always the VGA image shown on the screen in radiometric JPEG format. You get no other choice – you can’t access the original low infrared image, you can’t access the original visible image. You can’t get a 8 MP image with the thermal image stretched over the high detail of the visibile camera. Why not export all original data if they’re being stored on device anyway?
- I very much like that when you plug the device into a PC it appears as proper USB hard drive rather than as a MTP device like modern phones do. However, using a USB-C data cable to a USB-C laptop doesn’t work, nothing is detected. Only routing the phone’s USB-C port via a USB-B socket with a USB-C converter works. This makes me wonder if the device supports USB-C at all, and isn’t really USB3 with a USB-C socket?
As this is a just released device, improved firmware will undoubtedly follow, however I wonder if they’ll bother fixing most of the above before the next hardware refresh cycle occurs, whereupon they will surely want you to buy the newer hardware instead? I note that earlier thermal cameras in their range do have some of the above missing features, so I am assuming that this is a brand new firmware platform for them, and hence it missing features. Their firmware is a 128 Mb binary, so you’d really wonder what the hell they’re shipping in it given the lack of features.
Being a cynic, I suspect the potential of the hardware will remain unrealised, which is a shame, as for my house build record I’d really like to access the 8 MP visible images accompanying the thermal images. And capturing a thermal video with sound is useful on a windy day! Stll, as a straight shooting thermal camera this model ticks the box, and I am pleased with the cost benefit having tested it. I can live without video and original resolution visible images if it saves me €1,400!
Raw data: http://www.nedprod.com/studystuff/SSDsVsHardDrives.xlsx
This time last year I predicted:
I expect this time next year hard drive prices will have dropped back onto their long term trend line, and SSD prices perhaps will be flat. We shall see!
I called it right for hard drives, but I called it wrong for SSD prices which saw a better than trend improvement. Optane remained not just flat, but inflation adjusted flat which means Intel are proactively marking up Optane’s price with the inflation rate. Which explains why Optane remains a niche technology, as the performance per dollar gap between it and SSDs continues to grow .
Inflation in the past year in the US was something like 8%, so it is surprising that US dollars buy more hard drive and SSDs given the lack of inflation in the East where both are made. However, most raw materials and international trade is denominated in US dollars, so maybe inflation in the home market doesn’t necessarily mean loss of purchasing power outside it. After all, nobody is getting paid more, so Americans are getting poorer quicker than the Asian manufacturers whom only have to deal with raw material and energy price increases. In other words, the West is getting poorer quicker than the East because the West is seeing more generalised inflation, and I suppose that translates into a transfer of relative wealth from the West to the East. Thought about in those terms, I guess price drops could make sense from a Western relative viewpoint.
Looking forwards, it seems all but inevitable that price inflation will exceed 10% in the US and maybe even in the EU, though in Europe non-energy non-materials price inflation is being tamed by war’s effects on energy and materials supplies where outright lack of availability at any price means GDP is impacted, and that in turn dampens price inflation because the economy is weakened overall. I suppose where things get interesting is that lack of energy and materials supply also pushes up prices in the East, so I would assume that where computer components get manufactured is going to become rather more like Europe in the next twelve months i.e. everybody getting noticeably poorer because the cost of living substantially increases. That, in turn, probably means less demand for hard drives and SSDs, but also increased costs to raw materials and wages, so I’m going to suggest that both hard drives and SSDs will improve again this time next year as surplus capacity fights cost of manufacture.