SunAmp PV spares and maintenance issue

[TerryE]

3 hours ago, Ed Davies said:

I can't figure out from this thread is with @TerryE has had any problems getting said support or hasn't tried yet.

 

It's the latter Ed.  Rather than connecting to some help agent who might give you the bum's rush, I thought that Nick or Jeremy might have the direct current contact.  I haven't spoken to anyone in SunAmp since Jan 2017.

 

3 hours ago, joth said:

I'm curious though that with your focus on avoiding serialised failure modes, you have an HA + mqtt setup you are confident enough for this. That means IP network needs to be operational to get heating on? I've seen too many cascading failures with IP networks to trust myself to run a mission critical one in my own home. 

My own policy is lighting and heating must work without any IP. Do you have a write up of your HA setup somewhere? Always interested to see how others have partitioned things. 

 

The current system is based on an RPi + SSD with two USB serial connected ESP8266s doing the sensor collection and relay control, so it is independent of the IP network.  However, after a couple of years running, I've decided that in terms of timeliness Internet access is more critical to the house occupants than control of the SunAmps.  (It took us days for us to realise that one of them had failed. Ditto for the slab which we heat overnight.  I posted separately about once where I had a failure in the Wills heating and the slab heating was off for a day in November IIRC; the house cooled down by about ½°C during this outage.)

 

The way that I demand any heat (whether SunAmp or Willis) is that the maximum on-time request is 35 mins and the so the HA system reissues the on commands every 30 mins as needed.  This gives a reasonable fail safe: if the HA fails then the ESPs will shut down after 30 mins or so.  Minute-by-minute control at the HA level isn't needed.

 

The biggest source of outages since we've lived in the house has been power cuts.  (We bought AEG ovens for our kitchen, and the small Microwave / grill /oven was a piece of junk that kept tripping the power, so we've since replaced it with a Bosch.)  The RPi was OK, but everything else would die because the kitchen and utility are on the same RCB in our power distro box.

 

My current thinking is to keep all of these control devices on a private control WifiLAN: the HA RPi is gigabit-connected to the house switch, and I'd use static IP allocation for the control LAN on a private SSID with the RPi running hostapd.  I don't need to set up routing tables in the RPi, since by design my control ESPs  only communicate with the HA mothership.  I've got a spare RPi, and ESPs so everything is cold-swap repairable.  The control system will keep running if the rest of the house or BT OpenReach network infrastructure fails.  This will be no less reliable than the current setup.

Edited August 4, 2019 by TerryE

[Jeremy Harris]

@TerryE, I'll PM you with the technical contact I've recently been in touch with.

[TerryE]

Thanks Jeremy.  We've got to catch up anyway.

[SteamyTea]

4 hours ago, AnonymousBosch said:

Why has hot water - any hot water -  been made so damn hard?

Because we are trying to do it with lower thermal losses, different energy inputs, consistent and reliable outputs, at known quantities, and lowest capital expenditure.

Why I like vented system heated by a simple electrical element.

[TerryE]

5 hours ago, AnonymousBosch said:

Bugger.  - PV Direct, then.

 

If you've got the space and don't mind the extra heat losses then a TS is probably the simplest and most robust solution at the moment. 

 

Given that you are only using the water as a thermal capacitor, I can't understand why they have to make them cylindrical, and they don't need to be pressurised.  Why not a cube something like this that you bury in the garden lagged with 150 mm of slab PUR?  Do the sums and it's daily heat loss would be less than 1 kWh -- that's a lot better than most TSs and you could always stick a greenhouse on top of it   With a small rocket stove in the greenhouse.   Back to reality ...

Edited August 4, 2019 by TerryE

[Ed Davies]

2 hours ago, TerryE said:

It took us days for us to realise that one of them had failed.

 

There are some good lessons from aviation around this: redundancy is only useful if it's verified. That is, having two things, each of which can work on its own doesn't help if one can fail and you don't notice.

 

E.g., most low-wing light aeroplanes have two fuel pumps of different types. Both types have a fairly limited life and fail once in a while. One is mechanically driven by the engine, the other is electrical. Before you start the engine you turn the electric pump on and listen to it ticking to verify that it's working. You then turn it off and leave it off except for critical phases of flight (take off and landing) so the rest of the time you're verifying that the mechanical pump is working as a side effect of normal operation of the aircraft. The Chipmunk, on the other hand, has two mechanical pumps. As such it probably actually has fewer fuel pump failures, as the mechanical pumps are more reliable than the electric ones, but without the means for pilots to verify that both are operating (they'll only be checked as part of routine maintenance every 50 or whatever hours) it greatly increases the chance of flying around for a while without any redundancy.

 

Sort of similarly, the Challenger Space Shuttle accident can be seen as a failure to take notice of a loss of redundancy. There were two O-rings in the solid rocket booster field joints with the idea that they'd provide some level of redundancy. When it was observed that the first ring was getting burned through not enough attention was given to the fact that redundancy had been lost. They knew about it but didn't really “notice” it.

 

This is something I'm thinking about a bit for my “Mark II” home-automation system: some system to alert me to off-nominal situations so I can investigate before they get to be real problems. E.g., in @TerryE's case the system could notice that there's more than, say, a 5°C split between the output of the two Sunamps and signal this as something to be looked into.

[SteamyTea]

22 minutes ago, Ed Davies said:

This is something I'm thinking about a bit for my “Mark II” home-automation system: some system to alert me to off-nominal situations so I can investigate before they get to be real problems

Can we start something in 'Boffins Corner' about this.

[Nickfromwales]

13 hours ago, AnonymousBosch said:

Thanks Ed. Simpler than PV Direct hot water heating?

 

PV based systems should store excess, so I’m a little confused by your question. PV doesn’t lend electricity to DHW, it has to be diverted there.

For eg; If PV is generating 4kw of juice and the house is looking for 1kW then you’ll have 3kW going begging ( so it'll fulfil the needs of a typical 3kW immersion heater ) whereas 2kw from the roof would give a 1+1 divide, as the house would not ‘switch off’ to allow the full 2kw to go to DHW. 

 

The house ( your base load ) will be a constant and will be the first thing to suckle from the CU, regardless of where the energy is coming from, so managing control / diversion needs to be fully understood   

 

“Direct PV heating” would be for eg a DC immersion fed only from the DC PV panels, with the AC side of the house on the grid mains, or DC panels into an AC inverter that is not grid tied ( eg to a dedicated device only ). 

 

@AnonymousBosch, you’re going to have to buy something for DHW, so why so worried about the SA? A lot of people have them and they’re working fine. Unvented cylinders aren’t without problems / failures, vented are useless without a pump or a CWS tank up in the clouds, so what gives? 

Maybe the question should he asked directly to Sunamp;

”Will you carry spares?”, so I’ll ask   

Simples. 

 

[Ed Davies]

23 hours ago, AnonymousBosch said:

Simpler than PV Direct hot water heating?

 

Just realised I'm unsure what you mean by “PV Direct” - I assumed direct electric heating with an immersion heater of the domestic hot water (not a thermal store) using mostly diverted PV generation. Did you mean something else? Perhaps, direct DC from the PV panels into an immersion heater without an inverter, completely separate from the mains supply?

[TerryE]

I've taken the board out and in fact it doesn't look like the relay, but in fact it's a case of a bad/dry contact overheating and failing / melting the connector.  The board shows signs of thermal damage around the main relay and the through-hole solder joints have melted out.

 

The actual relay itself is a Schrack RZ03-1A4-D012 which is a 240V/12A rated NO relay with a nominal life of 50,000 switching cycles.

 

What really concerns me about the board is the trace geometry both in terms of cross-section and isolation.  IMO, these are far below what you would want for 12A current.  See the equivalent tracks in the Sonoff TH16 device below which do conform to CE requirements.  Note that the 16A rails have been reflowed with lead to build up their cross-section; all of the little dots on the wide tracks are vias to mirror tracks on the top side of the board.  Air gaps are used to isolate physically the phase rail from neutral and 5V logic.

 

I might be able to repair this SunAmp board, but I am be reluctant to use it.

 



 

PS.  I've struck out my previous point on track width.  I feel that JSH has addressed this in his comments below.

Edited August 10, 2019 by TerryE

[Jeremy Harris]

Not sure any isolation is needed, really, as there are no low voltage bits outside the box that can ever be touched.

 

If I had to guess, I'd say that the contact in the connector is the cause of the overheating, not the tracks on the PCB.  I think they've just been heated up from the overheating connector.  Looking at the connector, I suspect that the cause of the problem is the wire termination to the socket.  This looks to be a screw terminal, and Sunamp failed to comply with the requirement to always fit ferrules to stranded wires until fairly recently (our old UniQ control box didn't have ferrules on any wires, the new one does).

 

I'd repair the board if it were me.  Get a pair of new connectors, remove the plug from the board, clean everything up and solder the new one in place.  Trim back the wires to the socket (to get rid of damaged/corroded bits) fit ferrules to the wires and then fit them to a new socket, making sure all the screws are properly tightened.  Loose terminal screws is the number one cause of overheating - happens all the time.

[TerryE]

You are right in that I don't know the copper layer thickness but it would need to be about 2-3× the standard depth for this power trace geometry.

 

I totally agree about the ferrules issue.  I've had this problem before.  Keep you posted.  Thanks.

Edited August 9, 2019 by TerryE

[Jeremy Harris]

5 minutes ago, TerryE said:

You are right in that I don't know the copper layer thickness but it would need to be about 2-3× the standard depth for this power trace geometry.

 

I totally agree about the ferrules issue.  I've had this problem before.  Keep you posted.  Thanks.

 

 

The copper doesn't look to have been overheating, though, so my guess is that they've probably used something like 4oz board, which would be fine for the thickness of those traces and the current they are carrying.  I routinely use 4oz board on the motor speed controllers I've built, and some of them work at 50 to 60 A or more without any problems.  I've just done some quick and dirty sums, and for normal 2oz board for a trace carrying 12 A continuously, a 25mm long track needs to be at least 4.65mm wide.  For the same with 4oz board the track only needs to be 2.35mm wide.

 

The Sonoff, being Chinese, almost certainly uses board with very thin copper.  1oz seems commonplace for Chinese stuff, as it reduces the cost of manufacture a lot (the board is cheaper and it takes half the time to etch 1oz, rather than 2oz).

[TerryE]

2 minutes ago, JSHarris said:

I've just done some quick and dirty sums, and for normal 2oz board for a trace carrying 12 A continuously, a 25mm long track needs to be at least 4.65mm wide. 

 

What a coincidence: that's what I worked out.

 

I don't know about the Sonoff's.  In the quantities that they use they might chose to use 2oz for the top layer.  You just don't know, but really need to rely on the CE certification lab doing their job.

Edited August 9, 2019 by TerryE

[MJNewton]

2 hours ago, TerryE said:

I don't know about the Sonoff's.  In the quantities that they use they might chose to use 2oz for the top layer.  You just don't know, but really need to rely on the CE certification lab doing their job.

 

It sounds like you might be regarding the CE mark as meaning far more than it does. Most CE certification is done through self-declaration by the manufacturer, and even where this declaration has been 'verified' by a 3rd party you will likely find that the certificate (assuming it is legitimate) caveats that the verification is based on 'submitted documentation' and not 'submitted product' i.e. they haven't even seen the device, never mind tested it. Lab testing is expensive, box-ticking documentation is not. Such certificates are pretty much worthless, beyond their marketing value of course.

 

Given it is a self-declaration scheme the CE mark really should be taken merely as the manufacturer/distributor saying 'this complies with all the relevant legislation, honest' and nothing more. You only have to look at eBay being awash with cheap CE-marked power supplies and chargers which, if you open many (most?) of them up, you would find very poor (and unsafe) design and construction.

 

If you are after a 'safety mark' you should be more looking towards something like the UL certification scheme as that *does* involve independent 3rd party lab testing of actual product.

Edited August 9, 2019 by MJNewton

[Jeremy Harris]

+1 to @MJNewton ^^^^

 

I used to be a Notified Body, approving stuff to the LV and EMC Directives.  The vast majority of stuff from China has never been anywhere near an independent test lab.  CE is a complete joke, and to all intents and purposes meaningless in terms of safety or performance certification.  I've posted several examples here of lethal or non-compliant CE marked stuff, from LED lights (from a UK supplier) that were CE marked, had alloy cases and no earth, through to LED power supplies that radiated so much RF that they rendered all radio devices in the house inoperable. 

 

My Chinese built lathe was absolutely lethal, 2 core mains cable and really appalling insulation; no way was it Class II.  Similarly, a Chinese battery charger I bought gave me a bit of a belt.  When I took it apart the L was connected to the -Ve battery terminal, and again it had a 2 core cable, and was marked as being Class II and CE compliant.  There is no independent checking at all under the barking mad EU scheme, it's so bloody dangerous when compared to proper safety certification, like our old kite mark, that someone really needs to step up to the plate and kick the damned EU into touch...

[TerryE]

49 minutes ago, MJNewton said:

It sounds like you might be regarding the CE mark as meaning far more than it does.

 

You mean very like our construction industry approaches BReg compliance. ? The Chinese do it for consumer items. We only do for the housing stock that 95% of new buyers get.

[dpmiller]

CE means China Export, does it not?

 

[TerryE]

On a different note, I've been reading up on the use of ferrules vs. alternatives such as tinning the twisted multicore (which to be honest is what I used to do).  Apparently tinning is an absolute no-no. Why? Because a properly crimped ferrule presents a rigid cross-section whereas a tinned multiccore is plastic over time: it will deform and slacken off.

 

Time to buy a crimping tool and ferrule kit!

Edited August 9, 2019 by TerryE

[MJNewton]

28 minutes ago, TerryE said:

 

You mean very like our construction industry approaches BReg compliance. ? The Chinese do it for consumer items. We only do for the housing stock that 95% of new buyers get.

 

Sadly, quite an appropriate comparison!

 

It was the safety aspect of smart relays, hidden ones in particular, that made me a bit cautious hence why I went with the Shelly as it was the only one I found that's been through UL certification (that's not to say others wouldn't pass of course).

Edited August 9, 2019 by MJNewton

[ProDave]

For a start, I would NOT have speced a 12A relay to switch that load, I would want something a bit larger to be sure of reliability.

 

Some designers just do not seem to be able to design properly for high power loads.  A friend of mine (the tiler) has a solar PV diverter. I forget which one, other than it is no longer made.  It failed.  It was much the same sort of failure, the load to the immersion was carried through PCB tracks that were woefully inadequate and the fly leads off to the heatsink mounted SSR were also showing signs of getting hot and bothered.  It was not difficult to repair and reinforce the current path.

[Ed Davies]

It's worth noting in passing that Sonoff seem to have been improving steadily over time in this respect. John Ward reviewed one ages ago and was pretty sniffy about the current carrying capacity but was otherwise reasonably complimentary. I bought mine on that basis, thinking a limit of about 5 amps would make sense.

 

Like @TerryE's photographed above mine is better than the one John Ward reviewed - shorter and beefier high current tracks. Mine's version 2.1 (2018-06-15) which is slightly different in layout than @TerryE's. A further improvement is that the earth pad is also tinned which would presumably make some fault situations a bit less fraught - if there's no RCD or the RCD doesn't trip you'd rather the fuse blew than the earth track.

[TerryE]

12 hours ago, TerryE said:

it will deform and slacken off.

 

This reminds me.  A 2-year post move-in check is to go around all of the sockets and switches to check the torque on all of the switch / socket screwed connections to make sure that none have deformed to the point of being a loose contact.  I did this every 5 years or so one my old farmhouse -- mainly to check for dampness signs of overheating because we have 300 year-old stone walls -- but even so I'd occasionally come across a loose terminal thanks to deformation of the single core copper over time -- though the main deformation occurs on the virgin round core which is why I think it a good 2-yeear check.

[TerryE]

OK, the board was rather more damaged than I feared.  Here are the before and after with the new connector fitted.  The two RH pins below had got so hot that the through-hole liners very damaged and partially delaminated from the PCB, as had the RH trace.  This is why I bulked up the track between the connector and the next through-hole component (this was the common a.k.a. AC neutral) The image below shows the refitted unit with is working OK.   @JSHarris Note the crimped ferrule connectors   I've got a DS18B20 taped to the replacement connector to see if I have any thermal issues, but it does seem to be working fine.

 

PS.  The 2nd pin (the 240V phase line to the heater) is getting rather warm.  I think that this is because this track is probably also lifting near the pin.  I will probably need to do the same trick and expose the track to the relay and bulk it up with solder.

What I really need is a replacement board.

 

   

Edited August 13, 2019 by TerryE

[TerryE]

And here is the updated board.  The common line is now beefed up from the input to the output connector, and the phase from the relay output to the output connector.  (The latter looks a little untidy because it is covered in a clear potting polymer.)  This seems to have done the trick and the board is running cooler.  Plugging in the P=I²R numbers, a track resistance of 0.025Ω would generate 5W or so resistive heating, and this is enough to create a hot-spot of 50°C or so on a board without any thermal management of this size -- and this was the rough temperature that my DS11B20 was logging. 

Before     and after