Wifi booster?

What?

I've been suffering from weak wifi coverage in the workshop from the outset. The problem is that when we designed this whole extension to the house, it made sense to have it built so that it met all the building regs ("code" in the US and Canada), so that it might be worth more than it cost to build. 

I did most of the "second fix" myself, namely the wiring, plumbing, heating, painting, lighting etc, so saved a fair bit of the cost. Including the kitchen / dining room extension (aka "sun lounge"), utility room and toilet, it cost something like £50-60k plus several years of my life. Obvs I got the electrical installation tested and certified by a qualified electrician and the building work was inspected and signed off by the local building control office.

The upshot is that the walls between the workshop and the existing house are cavity wall construction (ie 2 walls, separated by a cavity filled with insulation). While this is great for noise and thermal insulation, it's also wonderfully effective at blocking wifi. Even the roller door is aluminium. I've got a "TP-Link Deco M9 Plus"mesh router system running the whole house but naturally it barely makes it through here. 

I suspect the bottom line is that most of these routers are designed for the US market where most houses are constructed mainly from wood ie relatively permeable to wifi signals.

I even tried a Powerline extender at one point but it barely worked. I think I can blame the wiring for that, as I have an armoured steel cable feeding a dedicated consumer unit in the workshop. This makes the route between the main router and the workshop sockets rather tortuous / torturous (take your pick).

What to do? In a last gasp attempt at using technology, I bought one of these supeydupey Wifi 6 range extender things.

Did that help? Did it shite. The problem remains the same ie maintaining contact with the mothership in the house.

Plan B, then:

I've considered this before but never quite got round to it - fit an external aerial / antenna outside the workshop where it can be pretty much guaranteed line of sight to the main Deco router.

To my limited recollection, these wifi routers typically have an array of 2 or more aerials that connect to the main PCBA by means of small "UFL" (UHF) connectors. If I can unplug one of those and replace it with a length of coax leading to an external, wall mounted twig, surely this should do the business.

Found a handy teardown report on these Deco M9 Plus things and it showed how to open them up. Sure enough, there are 6 such PCB aerials distributed around the circumference of the PCBA. So let's get an aerial and see if it works....

Bought one of these Chinesium twigs:

Managed to thread the coax cable through the roller door frame:

The Deco unit is disembowelled while I test it out. 

Works nicely, so let's tidy it up by mounting the SMA plug on the top cover. 

This holder thingy keeps the wiring tidy and provides a means of mounting the Deco using the mains plug and PSU.

A suitably positioned hole allows the SMA socket to be mounted on the cover.

Here it is in its final embodiment.

And showing a reasonable connection back to the main Deco in the house.

The connection to the Nucbox in the workshop isn't bad now.

We'll call that done. Hopefully my Samsung work laptop will also work - it's the most profoundly deaf of the various wifi devices there.

Stained glass window soldering

Here's our inner front door. It's probably one of the few original doors in the house and has an original stained glass panel. Not surprisingly, the lead joints have fatigued after 93 years of the door being slammed several times a day. The middle of the panel flaps a good half inch when you waggle the door open and shut.

I couldn't stop The Stupid Fat Bloke from removing it and starting to remove the panel. He'd whipped the door off and made a start before I knew what he was up to. Oh well...

Got the panel out without it completely disintegrating:

I previously bought a fine Chinesium soldering iron for the purpose of resoldering the damaged joint. It has a massive boss end tip and is rated at a Chinese 200W. I have to say it didn't fuck about but after 10 mins or so, the temp had risen way above the recommended setting of ~350-380C, so that the lead melted almost instantly and left a gaping hole. Not ideal.

So, out with the Pace iron. It's only rated at 80W but has thermostatic control.

Some of the joints were royally buggered.

There was even one that the original craftsman had missed:

If you look closely you can see the 2 fractures in this joint:

There. Both sides fully resoldered and ready to go back in when the door has been stripped and refinished using Osmo Polyx Tint 3072.

Job done.

Well - did the Big Beautiful Zener work?

I suppose the answer is sort of yes and no.

The circuit works on the bench at a (fixed!) threshold of around 58V. Problem with that is that the OVP threshold on the DMM Tech servo drives is lower than than. When I turn up the deceleration in the CNC controller, I get overvoltage shutdowns, usually on the Y axis which has the greater mass (both the table and the saddle).

As I can't easily adjust the threshold using hard wired zeners, this solution either works or it doesn't. In this case it doesn't, so I'll need to revert to a more complicated solution.

Here's the DC bus voltage captured during a series of rapid moves. The scope is AC coupled so I can resolve the voltage at a decent level. 

The overvoltage transient reaches about 8V above steady state, which is a nominal 48V for the X and Y axes. So the overvoltage threshold must be somewhere above 56V. Naturally the DMM Tech documentation doesn't specify what their threshold is but I'm guessing around 60V from what I've seen.

What now then?

As I have only 15V and 27V zeners available, it's back to the TL431 and a pot.

This looks about right, with the pot chain sensing the bus voltage and the 431 protected against the full voltage by use of series zeners:

However, the clamp voltage is still dependent on the IGBT's gate threshold voltage, which makes a mockery of the 431. I should just stop trying to cut corners and do the job properly.....

The classical "braking module" concept uses a hysteretic controller to switch an IGBT on and off, to maintain a bus voltage within a window. What I am doing here is using an IGBT in a linear mode, to mimic a zener diode. So instead of implementing hysteresis, I'll use closed loop / negative feedback to regulate the bus voltage. This requires a PNP transistor to translate the 431 cathode signal to a gate voltage:

This looks about right, although the actual tuning of the error amp (the 431) may need to be optimised, as I can't easily determine the circuit parameters in my system. I can play with C1 if the thing is unstable. If I make it too big, I may start to see an overshoot.

In order to adjust the threshold, I'll make R8 a 100k pot. Varying its value between 100k and 10k (easy to change in the simulator - and 10k is close enough to zero) results in a range of ~44V to ~63V. That should do the trick. 

Now build the bugger....

Found some stripboards and the required compts in my "collection". Off we go...

There. Not actually that complex and only took 20 mins or so to build.

And bizarrely, it actually worked first time out. I fitted a cardboard bezel and calibrated the knob using a DVM.

Then refitted it into the cabinet:

And bugger me, it actually does the business. I can now turn the feedrate up to over 10,000mm/min (400" per min). That should be good enough for now.

I may actually have to try out the combined plasma / CNC thing soon at this rate.

Workshop Tetris (again) - and new benchtop

Tetris?
Finally got round to getting some machine skates from Temu and AliExpress so that I can move the Tree CNC lathe from the middle of the workshop where The Mad Jocks left it, blocking any kind of movement of machines and large objects.

I've never managed to find any mention of the weight of the Tree in any of the various brochures, manuals, forums etc. I started to model it up in Fusion but that was partly due to being bored off my tits at the time, rather than any serious attempt to estimate its mass. 

The "frame" of the machine is cast iron, of a piece with the bed - rather than being fabricated as a support for the bed as is often the case. That's why it's so effing heavy.

As it stands, this model is reported (by Fusion) to weigh in at just under a tonne. However, it lacks the actual machine bed ways, saddle, cross slide, turret, tailstock, hydraulic pack, spindle motor, spindle, headstock, enclosure etc, so perhaps 2.5 tonnes or so wouldn't be far off the mark for the completed machine, if I ever got around to completing the modelling work. Don't hold your breath.

Finally, I got a message from a fellow Tree owner who reckoned he'd heard "6000lbs" somewhere. In the absence of any other figure, that's the most credible number I can go on. That's about 2.7 tonnes, which feels about right. The Shizuoka is around 3.5 tonnes and gives a similar impression of weight.

These machine skates claim to be for 8t and 15t respectively, although those are Chinese tonnes of course.

View from the outside door (fire escape?) corner of the room:

With that bench removed, I can see a space where the lathe could sit up against the wall out of the way:

Here's the "8 tonne" skate with a sort of pretend turntable on top...

...and 2 of the "15 tone" skates at the other end of the machine:

Progress is slow when you are single handedly trying to move this kind of lump.

Using a wooden lever to inch the thing along without completely destroying the floor:

Nearly there. Now it "just" needs to swing in, flush with the wall.

Ooof. Nearly there.

There we go:

And the bench?

That bench I moved out of the way has been bugging me for ages. Ideally it'd be reasonably flat but this one has never been remotely so to my recollection. 

FFS, the centre of the bench has sagged an inch. There's no way back from that, as they say. Although it seems to have a nice working surface, the substrate is just MDF, so perhaps that's no great surprise. This is destined for the dump.

B&Q to the rescue. I was torn between 19mm "moisture proof" tongue and groove chipboard and "marine" plywood. The former isn't anything of the sort and neither is the latter. But at £65 for a 4'x8' sheet cut to size, the 19mm plywood didn't look too bad a deal. So with a 900mm x 1800mm section cut out, it simply screws to the frame:

Yes, I used a router to round off the corners and a sander to smooth out any rough edges.

Finally, some Osma wax / oil to give it a degree of protection against oil, crud etc. Varnish wouldn't be much improvement, as it would soon chip off. Besides, I have this stuff to hand. It's what I used on the oak worktop that we recently installed in the kitchen and seems to hold up

There. That looks better. If I can find some patience, I may even apply a second coat once this has dried. Admittedly, that seems unlikely but you never know.

That's it for now...

Let's build this "giant zener" thing...

The simulation sort of shows the general idea but wouldn't necessarily work as shown in my application:

• The 431 doesn't like more than about 36V on its cathode. I will need a series zener diode to reduce the max voltage it sees.
• I need to ensure at least 1mA in the 431 to keep it "lit", yet I need also to ensure that 1mA in the potchain doesn't turn the IGBT on.
• I may need some capacitance across the 431 to keep it stable. Possibly not - but it needs to be considered.
• I have only a limited selection of components - despite the sacks on obsolete stuff I've accumulated over the years, I don't have "comprehensive coverage" of all compt types and values.

Let's see what compts I can find after a good rummage in my "stuff":

• BZX85-C27 ie 1.3W zener, 27V +/- 5%
• BZX85-C15 ie 1.3W zener, 15V +/- 5%
• 470uF 35V electrolytic
• 100uF 35V electrolytic
• 4.7uF 400V electrolytic
• 22uF 160V electrolytic
• 100k single turn pot (Chinesium - acquired for the Motorhead controller)

That looks like a good start. Along with the RS431 voltage regulators (a Chinesium clone of the original TL431) and the Big Fuck Off IGBT modules, I should have everything I need. So now, let's do some sensible calculations to make this thing work in my application:

With this scheme, using 2 of the zeners I just found, the 431only needs to cope with 18V. If I use 2 x 27V, I will 

Here's the RS431 datasheet - it's a Chinesium clone of the industry standard TL431, originally from TI:

IRGTI050U06 

You can see from the old datasheet that the gate threshold is pretty soggy but typically lies around 5-8V. BUT - bollocks to this. If I simply put two 27V zeners in series with the gate threshold, I might expect around 60-64V operation of the device. Hell, this would be a lot simpler than buggering about with 431s. 

I can't be arsed to put it into SIMetrix Here's the circuit:

The zeners would need to pass at least 2mA or so before thew B.F.O. IGBT would conduct, so they would be well into their "flat" region. Equally, a 2mA current wouldn't overheat them by a long chalk.

The 15V gate zener would protect the IGBT against overvoltage and its 1k resistor would limit the current if the input exceeded ~69V (ie 27+27+15).