2026-11-16

Fun with e-paper

I have quite a few of the very nice Waveshare 7.5" mono e-paper panels with laminated black bordered glass front. So I have decided to make a new fun project - a battery based sign.

The project involves a small PCB, and a carefully designed 3D printed case. The display actually comes with some adhesive tape even, so it sticks to the case design.

I then include a 13.7Wh LiPo battery. It includes a recharge circuit to charge from USB.

The board includes an RTC chip, which uses 160nA. This has alarm settings and a button trigger as well. It can then wake the main CPU (ESP32-S3).

I have opted not to use the ESP32 deep sleep mode for this - by shutting down the whole processor the idle power consumption really is very low. The concept is simple - it wakes up periodically, say every 5 mins, checks for new data, and if no new data it goes back to sleep.


It is complicated to get this check and sleep process down to a minimum time, but I have managed to get it to a typical 850ms.

  • Fast boot loader, not checking the image integrity (this does get checked in the event of a crash, and will switch to other image if corrupt).
  • Not running a PSRAM memory check.
  • Not logging boot loader to serial.
  • Caching the channel and BSSID for WiFi connect.
  • Caching the IP, mask, gateway and DNS for IP connect.
  • Fetching a page via http not https (https adds 2s).

The caching uses the RAM and EEPROM in the RTC, and sadly not quite enough space to cache IPv6 - though I may be able to squeeze in for simple SLAAC and zero router address perhaps. One to consider.

So yes, some compromises here, but 850ms is really short.

The system does a more comprehensive start - scanning WiFi, getting IP again, checking s/w update, every day. This takes a few seconds.

To try and get some idea of the battery life I have set one of these up using a tiny 0.9Wh LiPo.

The ADC is not ideal, and estimating battery capacity is always fun. It has been reading capacities around 85% and actually going up and down a bit.

Over the last week it has gone from as high as 87% to 83% for my estimates.

So a crude estimate is that if it has used less than 5% in a week, the 0.9Wh could last 20 weeks? This would mean the 13.7Wh could last 5 years!

Yes, this is a wild estimate - it depends a lot on usage, and a LiPo will have some internal discharge as well. What I think I can confidently say is my original goal of 1 year battery life should easily be achievable. Ideally I want to do a full charge to flat test on the 0.9Wh to get a clearer idea.

2026-10-29

Some simple Eufymake tips 3D print over print

My Eufymake UV ink printer has issues, and so does dealing with Eufymake as a company, as I have blogged, but today, let's look at it working, and doing what I ask.

I have done a few "over prints" now, where I am printing on something that exists - printing on PCBs is excellent. But this time I have a 3D print that I want to accurately over print.

So a few tips.

  1. Dimensional accuracy of the 3D print. My resin printer can be very slightly out on larger prints - depends on the resin used. The models from JLC are unbelievably stable. So measure, with callipers, to check. Do not rely on the Eufymake camera.
  2. Make the artwork to the correct dimensions. My example here is nice as it is a 16x12 grid of 7.5mm square blocks. Almost impossible to get wrong.
  3. Add bleed to the artwork, a few mm over all edges on the most appropriate colour. For this model I copied and pasted half a block left, right, up, and down behind the image.
  4. Make an outline of the image, only like 1mm or something.
  5. Load the outline, centred exactly, and print on paper on the print bed.
  6. Place model on the outline to align exactly.
  7. Load the image, centred exactly.
  8. Know the centre of the image - in my case there is a clear and exact centre.

At this point, when you snapshot, the centre of the image should be right. It is worth double checking it matches the image. Do not trust anything away from the centre - trust the printed outline. Checking the image centre helps double check you have placed on the outline.

If doing many prints the bleed ends up on the paper and gets thicker. In my case with a solid base that means print after print the placement of the image gets easier as there is a thick print line around the edge and the model slots in to place.

Indeed, I printed an outline inside the image and placed on that - an alternative would be an outline outside the image, and print a thick layer on paper with the Eufymake so as to create a positioning slot for your 3D image to slot in to place.

The results. Some happy teenagers who can tap to pay on their Netherite Ingots (whatever they are?!).

I think this has proved a good exercise. The case printing I have been working on could have a thick outer outline printed on paper to perfectly align the case before printing. So yes, good testing even on a silly thing like this.


2026-10-28

PCB mechanical fit testing


One of my latest designs needed more mechanical fit checks that usual, it was to work with specific hardware and I wanted to check the case design with the PCB and battery and so on all fitted before ordering the PCB. So I 3D printed the PCB.


This is not really a new idea, but first I have done it. We actually did recently for some FireBrick work - 3D printing the case, but also 3D printing the PCB. This was FDM, which was fine for checking the mechanics.

But the result in resin is stunning beyond belief. I actually ended up going through the KiCAD models and updating some of the 3D model STEP files even to be even more accurate. Some points of mention.

  • The USB-C socket has the inner part, but you can plug in a USB-C cable properly and not break anything. This allows the exact fit of cable in socket and confirming is it not fouling anything in the case design.
  • The battery connector even has the pins - that is crazy. And again, the plug of a real battery could plug in and not even break the pins! This was great to confirm exactly the right connector and polarity and confirm battery lead length fits as expected and is reachable from where the battery will be located.
  • The ribbon connector has the gap for the ribbon from an e-paper, and allows me to verify the length fits and again is something one can reach in the case to install.
  • The sticky out bits fit in the case design which has slots for them (button, USB, WiFi antenna, and lugs).

I am amazed at the precision. But actually there was one area that was technically too accurate!

KiCAD output the board, and will even output the solder pads, and output the components. But in real life the components would be in solder paste. The actual output had maybe tens of microns gap between component leads and the PCB itself. This meant a print slice layer had nothing.

The result was most components were fine, amazingly, even though all of the SMD had this gap. The contact area after the gap was enough, even on even 0201 resistors. But the leads of SOT parts were curved to a point as they are in reality. In reality solder would wick up those legs.

The result was interesting. The components printed perfectly but very loosely connected to the PCB!

I ended up with a lot of missed components when taken out of the printer (and had to clean the release film) and even more missing after a wash.

What I need (and have a feature request in) is KiCAD to have a solder paste export to fix the component legs to the pads as they would be in real life.

For now I export PCB and components separately and load in to FreeCAD and move one 50um down. The result is a solid print with stuck down components!

2026-10-23

Is TOTSCO fit for purpose

I have explained many of the issues with TOTSCO - the company that does "One Touch Switching" for UK broadband and telephone services.

My main gripes have been over the unbelievably poor specifications and stupidly inadequate processes for accrediting CPs to connect to the One Touch Switching system. This is partly why I made a testing system (NOTSCO) free for CPs to use, and have actively helped many CPs as a "buddy CP" for what testing there is. I am trying to make this work for everyone.

But it gets worse.

TOTSCO are actively refusing to pass on a simple message to another CP that is failing to meet the OTS specification.

In this case it is such an egregious failure one has to wonder how even the inadequate TOTSCO process got this CP on the system. No I am not naming the CP. The error is as basic as they come - a residentialSwichMatchRequest getting a residentialSwitchMatchConfirmation but without an actual switchOrderReference (or implicationsSent). This is the most basic function of OTS and is broken beyond belief. The CP should be REMOVED from the platform until they can show they can actually comply with the spec, and more importantly TOTSCO should have always been testing CPs do, in fact, comply with the spec. Failing to do so is unbelievably irresponsible.

They cite "security", but they is ZERO security issue with passing on such a message. And no GDPR issue either. Their refusal is baffling.

What is crazy is they do operate a "CP to CP" communications system, but signing up for this is not mandatory. When CPs do sign up TOTSCO WILL pass on messages. This highlights that passing on messages is not a security issue. So why not pass one based on a simple email when a CP is not signed up to "CP to CP" communications. There is no justification for this.

If TOTSCO are absolving themselves of ANY RESPONSIBILITY for policing ANY compliance with the One Touch Switch specification, whose job is it? It is not my job as a CP.

So who do we turn to - perhaps OFCOM. Maybe OFCOM will decide TOTSCO are ignoring the most basic responsibility for addressing CPs not complying with the OTS specification.

Should OFCOM police this somehow?

Note this is not the first instance of TOSTCO being so stupid - we have another CP failing to meet the OTS specification that is likely a GDPR breach. They REFUSE to tell the other CP. I have now contacted the ICO (no reply yet). But why REFUSE to pass on such a basic message (not personal data in the message and no security issue).

This needs fixing - OFCOM you need to fix this.

2026-10-10

BT SFI Should not exist any more

It is a while since I have ranted about "Special Faults Investigation" service that Openreach sell.

[see update at end]

Why did SFI exist?

  • Openreach sold a service called metallic path or shared metallic path which is the wires from exchange to master socket. The standard for it was that of wires capable of making a telephone call, though the standard was improved in time.
  • BT Wholesale (at the time pretending not to be the same legal entity as Openreach) buy that, add equipment at the exchange (DSLAMs), backhaul, more equipment (BRASs, switches, etc), more backhaul, and sold a DATA service, at one end being a hand over (typically at a data centre) which is fibre with ATM or Ethernet, and at the other end the master socket with ADSL encoding.
  • ISPs buy that and add equipment in the data centre (LNS, Switches, DNS, email, etc) and transit and peering links, and sell an INTERNET service to customers - sometimes terminating at the master socket as ADSL, sometimes including a supplied router and Ethernet and/or WiFi

This system worked pretty well, and at each level someone is providing a service to some spec for a fee, and if the service breaks and does not meet that spec, then of course, they have to fix it, at no charge to their customer. Each layer builds on many parts to make a service to their customer.

This is where SFI comes in - there is gap. If the metallic path breaks, Openreach will fix it, however if it meets the spec (based on telephone calls) but is not good enough for ADSL data (e.g. below the min sync level BTW have agreed with the ISP), then Openreach DON'T HAVE TO FIX IT because it meets the metallic path spec. So they SELL an extra service to BTW, a Special faults Investigation, which, for a fee, involves the engineer trying to improve the metallic path (beyond the spec) to be better for data (e.g. ADSL). BTW have to buy this so as to fix the broken data service that they are selling to the ISP.

Why was SFI controversial?

All of the above sounds fine, at first. Bear in mind, BTW sell a data service to ISPs, and if a DSLAM breaks they will get it fixed (at not charge to ISP) and the same should apply if the data over the metallic path breaks (e.g. below agreed minimum sync rate), and again as not to spec that should be at no charge to ISP. But BTW DID CHARGE ISPs for the SFI. What is worse is many ISPs then charged their customers for the SFI!

This led to a lot of disputes, and the pettiness got mental. Some examples are BTW used to say SFI was optional so on all faults said we do not opt for SFI. They then refused to actually fix lines unless and SFI was booked but said charge was for work beyond the master socket, so we said do no work beyond master socket. They then said they could not accept that request, and also said that a visual inspection of the end user router and wiring counted as (chargeable) work beyond the master socket. At one point we instructed customers (who did this) to remove their router and all wiring and physically hide it out of sight before the BT engineer arrived. We had pictures even. BTW still tried to charge. On one occasion the engineer immediately closed the fault as "no customer router plugged in", without testing or fixing the fault! We have done well disputing SFI charges at every stage, usually quoting their own terms to them and again explaining that we (as an ISP) are not buying a metallic path.

Note: there is a logic for a charge for wasting time - reporting something that is not a fault, e.g. an engineers visits and finds the end users router is on fire - a charge on that basis, as a penalty of some sort, does make sense and makes sense to be passed on to whoever screwed up.

Why should SFI not exist now?

Now we have the crux here - we are still getting SFI charges, and still spending time disputing them, but for almost all lines SFI should not exist. To explain, that initial bullet point has changed for most lines now - because of the UK telephone service switch over from old copper pairs to the exchange.

  • Openreach sell a DATA service with handover on a fibre at the exchange, Openreach equipment in the street cabinet, and wires or fibre to the end user premises handed over as VDSL or Ethernet.
  • BTW buy that DATA service, and have equipment at the exchange, backhaul, etc..
  • Etc

What we have here is no longer a gap - no longer is Openreach selling a metallic path and BTW using it to sell a data service. If the data service is not working it is Openreach not providing something to spec, and so Openreach should fix that as a NORMAL FAULT INVESTIGATION. There is no gap for which SFI was invented to cover. Indeed, BTW should dispute every such SFI charge before even thinking of charging us (ISP).

There are still some ADSL, but they are quickly going, so SFI could exist for those lines for now - with the same original controversy of BTW trying to charge ISPs, but this is dwindling. Somehow BT try to charge for SFI for VDSL and FTTP services even now where it logically makes zero sense.

If BTW really are separate from Openreach now, they really have no excuse for not disputing every SFI charge related to FTTC or FTTP on this basis.

Giving BT some credit where it is due

One thing BT never had the nerve to do with all their changes to contract terms was make it so that their customer would have to pay them to fix a service they agreed to provide, but were failing to do so. Such a contract, even on a commercial basis, would be very challenging, and I would hope OFCOM could be persuaded to step in if that happened. The basic concept of I agree to pay you for a service and you agree to provide that service relies on the provider having to provide what they agreed, or they fix at their own cost. They tried a lot - the whole visual inspection is perhaps the most creative, but never that.

I may not be able to say that for all providers, but that is another story.

Update

ISPreview published an article on this, thanks. OR tried to justify SFI - their main justification seems to be work in customer premises (which we don't want - they are far to expensive for such work), or failing to find a fault (which makes sense if there is no fault, but does not when there actually a is fault - they should pay us for failing to find it).

However it is very clear to us that OR are still operating SFI charges as if they were selling a metallic path facility. Just yesterday we disputed an SFI charge and got told "Charge valid - SFI2 checklist determined "no fault present" (to SIN 349)".

To explain, SIN349 (which states it forms no part of any BT contract) is the metallic path spec (improved from basic telephone service). BT Wholesale do not buy a metallic path service from OR any more they buy a data service (VDSL/SOGEA), so testing the line to SIN349 is irrelevant - they need to test that the supplied VDSL/SOGEA service is working.

And, of course, as an ISP, we do not buy metallic path service from BTW, and never have, so even in the old day when testing to SIN349 may have made an SFI charge (from OR to BTW) valid, it did not make it valid as a basis for BTW to charge us, as we buy a data service not a SIN349 spec metallic path service.

2026-09-30

Wallet inspector

What is fun with a nice UV ink jet printer is on what you can print...

One thing is a bank card.

This is now my Monzo card, oops.


Yeh, I can print on a card and the new non-embossed cards are even easier.

Someone suggested I make a flashy web site offering this as a service just to see how many people will send me their bank cards to print. It really is the 2026 version of "wallet inspector"!

2026-09-28

Elegoo Jupiter 2 3D resin printer

I decided to upgrade my Elegoo Saturn Ultra 16K to an Elegoo Jupiter 2. So this is how it is going...

Resin printer basics

For those that do not know, a resin printer has a tray of goo (resin) in the base, under which is an LCD. The bottom of the tray is a "release film". There is a flat plate suspended above, it lowers in to the tray, a layer of resin is cured, and it lifts off the release film, and then down again for next layer. Eventually the final print hangs from the base plate. There is some messy cleaning and extra curing then to make the final part.

The Saturn had some challenges

I am sure I have griped on this before, but the Saturn had some issues...

The cover was a big top, sides, and front that hinged at the back top. To lift it up you would obviously grab the top sides and push up. The problem is that this motion would easily tip the whole machine back. The trick is grab the bottom sides and pull towards you and then up.

Then the print bed is clamped in placed by a small lever that you push up to release. It is small and stiff, and you push up with palm of your hand. And, you guessed it, this motion would tip the whole machine back. The trick is hold the top back of the machine with one hand while you do this so it does not tip.

However, the print bed is pushed forward when you do this - I never had this happen, but there is a risk it comes off the mount and crashes in to the tray of goo and LCD and smashes it all in a very messy way. So the trick is hold top of machine with left hand, push left elbow to front of print bed, and then lift level with right hand, and hope you do not get covered in resin.

Why does tipping the machine matter? Well, because you have a tray full of goo and if you tip the machine it pours out of the tray in to the machine. This is bad in many ways, not least of which is the machine then leaks goo from the seams for weeks. But also, at the back, directly behind the tray, is the z-axis motor and screw thread, with a big hole around it - so the goo goes right down in to the z-axis motor. This basically burgers that motor. I managed to get working fairly well with a lot of alcohol, lubricant, and running bed up and down for hours. Still not 100% but good enough to be printing again. It needs replacing.

Obviously you have to top up the tray to a level to cover what you are printing, and that makes spills even more easy to happen and more messy when they do.

Finally, changing the release film is a LOT of screws/bolts and very fiddly. Thankfully I realised I rarely need to change if, after a failed print, I use the cleaning cycle which exposes a whole layer you can peel off with any remaining print debris. However, on one occasions it got pierced somehow and I then had to change the LCD screen protector as well as it had resin sealed on to it. Resin printers can be messy.

Also, the print area is not very big, but more than big enough for what I normally print.

Apart from that it is good, and prints well.

An interesting feature is the tray and LCD tip down at the front as part of the release cycle, neatly peeling off the release film from the latest printed layer. Very clever.

The Jupiter

This is different and a bigger machine.

It has a bigger print area, nice.

It has side opening doors, solving the tip back issue.

It has a very long level for release of the bed, which can simply be lifted between thumb and forefinger with no real effort, or risk of tipping the machine.

The z-axis seems to have a cowling to protect from spills.

Even the release film change is a module with a frame and film that simply unclips and a new one clips in place. Costs more, but a lot less effort.

It even has tubing and an auto fill of tray from a bottle at the back, meaning it keeps a low level in the tray - way harder to spill and way less messy when taking off print bed. It can suck that up back in to the bottle for you as well.

I mean that was almost every single gripe I had sorted, well done.

The downside

Sadly there are problems.

It sucks!

That is the main issue. Unlike the Jupiter the print tray and screen do not tilt down to peel off the film from the print, instead the print head just pulls up after the printed layer. It often strains and snaps up. This clearly sucks on the release film, a lot.

After a few small prints I now have creases in the release film which mess up the print.

What is worse is Elegoo's response.

In summary (a) tough, and (b) my fault somehow. I am using all the default settings the app picks knowing printer and resin, and only printing a few small things with which the Saturn had no issues - even so, saying "yes it has a large print area but you can't actually use it" is mental. Where are trading standards when you need them?

Fun with e-paper

I have quite a few of the very nice Waveshare 7.5" mono e-paper panels with laminated black bordered glass front. So I have decided to ...