Archived Forum

From what you say, it sounds as if your diverter could be correctly balancing the generated power on the phase that it's on, and the meter is recording imported power on the other phases. What is your overall set-up now? Are you generating into all three phases, or just one? How are your other loads split across the three phases?

Hi Robert,

The set up at the moment is only temporary and is somewhat unbalanced.

The diverter is set to monitor the energy of all 3 phases together in one combined 'energy bucket'. The solar is only on 2 of the phases (L1,L3)  and the other phase (L2) has very little load on it, about 35w nominal but then the kettle etc is on that one.

Today wasn't the best of days to turn it (3 phase) on with rain and overcast skies. When I tried we were generating about 2.2kw of solar split between phases L1 and L3 (single phase inverters). We have a base load of about 600w so there should have been 1.6Kw available. I found thought, with the diverter off and a 500w heater on L2 the red Led was still pulsing.

To me, it looks as if the meter is not registering minus x and minus y on phases L1 and L3 but zero and zero. 

That's quite possible. Have a read of "Operation of electricity meters when energy flowing in reverse"

This section seems to say that your meter might be behaving incorrectly. If you can establish that for certain, you need a meter change.

"3.3. Multi-phase meter not designed to net exports from imports
Import multi-phase meters should operate as an integral device so the meter should record the sum of the
imports less any exports. The result should be the same as if all of the load were taken through a single
phase meter."

The example that follows looks remarkable similar to your circumstances.

Thanks Robert.

Thankfully I remembered that the meter installer had phoned me for directions so I had his number and he was a really nice guy. I left him a message and he's phoned back and is going to enquire into it.  I don't quite understand why meter operators install used meters but this one is 4 years old.

The suffix J of my meter is shown as ok but mine is a suffix A

Well it does get interesting.

The meter installers say I have to go back to my supplier and ask them to get the meter installers to install a different meter, A Vector Sum meter which has to be a specific request. Why??? 

On phoning the supplier I eventually managed to get them to agree that my meter is  not reading correctly- but I had to be very very insistent, quoting from the link that Robert put up and saying I would e mail it and a public ofgem letter on the matter to them.

I can't understand why old meters are being installed that don't even meet the meter operators own guidance.

"I can't understand why old meters are being installed that don't even meet the meter operators own guidance."

I can. They probably hadn't got a clue how that particular meter had been programmed. You should have excellent grounds to contest your bill.

Didn't want to go down that line so I switched the supply off and have continued on just on battery and solar power. We had already done 2 weeks like that while the local network was being upgraded. The meters only gone up by 2 units!!

Tomorrow I'm going to try the other new meter on the other 3 phase supply and I'm praying that it's not going to be the same!

I think its just as you say and the meter installers are passing the buck so they get paid again for changing it, even though I don't believe they should have installed it in the first place. 

Guess what, the other 3 phase meter installed through another supplier and a different meter installer company is the same. 

They both went to the same (half-day?) training course?

Well I am really struggling here, no reply from either of the supply companies yet!

As a get round I have gone back to just using everything on one of the  phases but I can't get my diverter MK2i to work without the meter kWh led flashing (the kvarh led doesn't).

If the diverter is off and I am exporting more than the base load the led doesn't flash. If I am importing it flashes away.

If the diverter is on (in normal mode) the kWh led flashes at different rates, normally about once per 8 seconds (that's not nearly as much as if it were just importing the diverter load without the pv).

As a test I have a commercial diverter, which is designed to have a small safety export of 50w, and the meter kWh led still flashes but not quite as often.

Also I do have a similar, but not the same, meter to the suppliers one and with that and using the MK2i the kWh led doesn't flash. (But the kvarh will, but that could be because i am using the inverter to supply all the power and its not as stable as the grid !!)

I'm lost!!!!

It really does look like your meter is "faulty" in the sense that it is not fit for purpose, according to the guidelines. Those specifically state that it should exactly emulate a 3-phase Ferraris meter with 3 induction disks on the same shaft and with a backstop. In other words, the meter should register the nett import and ignore export. if the total import on 2 phases exactly balances the export on the third, the LED should be on steady, indicating that it is in "creep" mode and there is zero nett energy flow. (Test: If you switch everything off for a few minutes, does the meter LED light? It should do.)

It's hard to know what you should do. Reverting to a single phase installation (inverters, loads and Mk2i) should allow you to run a balanced system with zero nett import given suitable conditions, but that's not what you want long-term.

(Circulating vars are not a problem, though they can heat up cables they don't - or shouldn't - cost you anything unless your supply agreement says otherwise.)

Yes, the led lights up constantly if the power is turned off.

Besides the meter actually not being fit for purpose I have actually come to the conclusion that it is faulty as well or it doesn't seem to behave as other meters do with diverters!!

Just using one of the phases and Robins diverter system the kWh led has continued to flash on that meter (5219A) at about once per 8 seconds or in that region (it does vary).

I have now swapped over everything to the same phase of my other supply meter (Ace 3000) and that kWh led does very occasionally flash but nowhere near like the other meter. (This could possibly be due to the 1/2 second time delay between SSR's turning off as the sun goes behind a cloud).

(I have 2 supplies and swapping is easy as I have a generator transfer control panel so the second supply can simply be on the generator input).

Edit, I haven't paid for a 3 phase supply to find i'm unable to use it so the meters WILL be changed!!

As I am now trying everything running on one of the phases I wanted to get to the bottom of why the 3 phase meter was showing the pulsing red kWh led when using the MK2i diverter.

I have modified my sketch, with guidance from Robin, so that I can set it to have a bias to export a little to the grid. I set this high at 450w. At the same time I installed 2 new single phase meters on the phase, one wired as normal and one in reverse. ie. One will show import and one export.

This morning once the pv was generating nicely  I turned onto grid supply.

Results after about 3 hours;

Import meter has stayed at 0.75

Export meter has increased from 0.74 to 2.01 

3 phase meter has increased by 1 unit. (It only displays to full units)

I think that is very conclusive evidence that my meter is faulty (or has a zero energy bucket?) as well as not being suitable. It should be noted that if the diverter is off and I am exporting a lot of solar the led doesn't flash!!

EDIT;

Since writing the above I noticed the red kWh led staying lit on the 3 phase meter showing no import or export, as if it was turned off. A slow moving rain cloud was above and the house base load was just slightly less than the solar pv so the diverter wasn't diverting (remember 450w bias towards export). The export meter was going up very slightly, with a flash of the led occasionally. Does this meter have a strange way of working like no other meter anyone has come across??

Since writing the above I noticed the red kWh led staying lit on the 3 phase meter showing no import or export, as if it was turned off. 

On a single phase meter, the LED being "on" can have two different meanings:

1.  After a period of minimal energy flow, the meter slips into an anti-creep mode and the LED comes on constantly.  This is the mode in which the meter powers up.

2. When a burst-mode diverter is active, the LED light being on means that some energy has been exported through the meter beyond the point where it can be reclaimed.  This is not a problem, it's just a consequence of the way that the two systems are interacting.  There's an explanation here.

In the case of my Mk2 PV Router, the energy bucket needs to fill to at least the half-way point, and real energy is being exported through the meter while this occurs.  Shortly before the router starts to divert its chunk of power, the meter's LED can often be seen to come on - but not always.

While surplus power (SP) is available, if you turn on a kettle or similar to consume all the SP, the meter will pulse slowly to show a low rate of consumption.  Then just before the next pulse is due, turn the kettle off.  The level in the Mk2's bucket will immediately start to rise, and power should be sent to the dump-load before the meter's LED comes on. 

In "normal" mode, there's a 50:50 chance as to whether the meter's LED will be "on" or "off" when the Router is actively diverting SP.  When an anti-flicker algorithm is being used, it's more likely that the meter's LED will be "on" than "off", but it's not a problem as such.

The main thing to watch out for is the meter's LED changing state while power diversion is occuring.  Every off->on transition is a chargeable event, and there should be none of these.  If there are, then a tighter anti-flicker regime needs to be used.

The above findings have all been with single-phase systems.  Sorry, I can't help with your 3-phase difficulties.

I think 3 phase meters are supposed to be the same. All I have seen come on with the LED constantly lit and also are constantly on when there is no export or import. I think the led on different meters can behave differently when exporting. My previous single phase meter lit up constant green and my new ACE 3000 3 phase meter also does that (though that also flashed red at the same time if I import on one phase). The 2 check meters I have still flash the red led but the meter doesnt increase .

This 3 phase meter is like you describe but for some unexplained reason when I have the diverter diverting the led flashes. The two check meters show that I am not importing and that I am exporting the bias I have set up. So why is the 3 phase meters meter going up? It doesn't go up when I export without the diverter on, some meters are programmed to add export to the import. The diverter isn't set in antiflicker mode but the check meters show I am not importing anyway. 

EDIT;

Results after about another 5 hours;

Import meter has increased from 0.75 from 0.95

Export meter has increased from 2.01 to 3.44

3 phase meter has increased by another 1 unit. (It only displays to full units)

I think I have worked out what is happening.

I think my meter is set to ignore any export at all so when the diverter is working it is importing very slightly.

Say at any split second in time I have 2kw of excess pv above my base load the diverter will turn say a 3kw load on for say 1 mains cycle. In that 1 cycle I am therefore importing 1Kw. In the next mains cycle, when the diverted load is off a normal meter will see 2kw of export and offset that within its energy bucket of say 3600J but mine won't as its set to ignore any export at all. So every on cycle it will see 1kw of import. So for me to see no import on the meter my divert load wattage would have to exactly match the spare pv wattage and then the meter wouldn't advance.

Has anyone seen this on a single phase meter? 

Nick,

You appear to be saying that your new meter has zero allowance for any reciprocating flow of energy.  All types of meter must surely allow some free-flow of energy in that manner, otherwise reactive loads wouldn't be recorded properly. 

In the early days, I recall posting exactly the same argument as you have just done above, but that is thankfully not how meters actually seem to work.  Within the meter's "sweet zone", the mains acts as a "trampoline" whereby exported energy can be freely reclaimed when required by your dumpload.

In terms of not being charged, the safest way for you to run your router is for the load to be flicking on and off as fast as possible.  When a decent amount of surplus power is available, your neon/bulb/whatever should be flicking on & off too fast to count.  If it isn't, then the chunk-size in which you are dumping your surplus energy is surely too large for your meter.

Whether you're using my Mk2 code as your basis, or Martin's PLL equivalent, the power diversion mechanism for both systems is essentially the same; maybe you just need to speed it up.  If you were to post the code that you're using, someone may be able to check it over for you.

Robin,

That is maybe common thought but I seem to remember having a similar discussion ages ago about the size of the energy bucket in meters. Common thought was that they were all 3600J but mine wasn't it was 1200J and since then its also been shown that some are 3000J. Just because something doesn't fit in with what is known about certain meters it doesn't mean they all behave the same way.

My load is flicking on and off really fast. I have been using this diverter for over 18 months with no problems at all. The diverter now runs much faster than it used to as I used to run it in anti flicker mode but now I'm using normal mode.

Can you explain why the led is also illuminating with a commercial unit I have while that is dumping? (And I'm not using both at the same time!!)

Why does the 3 phase meter show over 2kWh of import compared with 0.2kWh on the check meter I put in, with the export check meter showing 2.7kWh of export all over the same period and all 3 connected in series?

Reactive loads show up in different quadrants than reverse power flows, so I think you could design a meter that behaves the way Tinbum suspects his does, without it getting confused by reactive loads.

The energy measurement chips I'm most familiar with break the signal down into reactive and real components in a very early stage, and then treat the two identically, but independently.    Real energy gets accumulated in one signed register and reactive energy gets accumulated in a different signed register.   The cpu can then read those registers at its leisure. Even at full-scale deflections, they overflow about every 74 hours.

Should the real energy change directions between reads, then the next read will only reveal the net result because the energy register is doing signed accumulation.   Ditto with the reactive energy register should the signal move from capacitive to inductive, you only see the net result.

So a typical setup is to read the real energy register at a fixed time interval.  I happen to read mine every 10 seconds.  That works fine for me because my whole-house power flow changes direction very rarely (typically 3 or 4 times a day), and in any case I mostly clamp in locations where power always flows in one direction only.

But in order to allow revenue meter manufacturers the ability to count imports and exports separately, the device has a change-of-direction interrupt (one for real energy and another for reactive energy).  So a typical setup in that scenario is to read the register every x seconds, or whenever the change-of-direction interrupt fires.

Now the resolution of those energy registers obviously depends on how you've set your inputs dividers up, and how you've programmed the calibration registers.   I've set mine so that they accumulate actual WattSeconds (or Joules).  Because I read them every 10 seconds I get "net watt seconds used in the last 10 seconds" every 10 seconds.... which is the equivalent of "average power used over the last 10 seconds" in units of 1/10th of a watt (which incidentally is why I can easily see 5W loads coming on and off... they show up as a count of 50).

If I had a requirement to accurately count imports and exports in separate columns, the only way I know of doing that with this device is to enable the interrupt I mentioned above, and read the energy register whenever it fires (as well as every 10 seconds when it doesn't fire).  If I did that, the only grace time you'd get (or bucket size as you call it) would be the latency in that interrupt.  I'm pretty sure most of this stuff is sync'd to zero-crossings, so that latency might be as large as half a line cycle.

Very interesting. I hope I am wrong but it's not looking like it to me.

Perhaps the behavior is a consequence of the fact that the meter is set to ignore export on other phases when calculating how it advances

Hopefully, 'when' my meters are changed to appropriate 'vector sum' meters I won't get this.  If I do then either  the diverted loads have to somehow exactly match the spare pv or I have to have a bias export on the diverter set at the size of the largest diverter load eg 3kw. Neither of which I would want to do.

I should have said WattSeconds in the case of the real energy register, and VARSeconds in the case of the reactive energy register.

I tried googling to see if I could determine what chipset your meter uses, but I suspect such things are kept pretty secret.  I did come across this though:

http://www.maximintegrated.com/en/app-notes/index.mvp/id/4664

Not particularly related to this thread, but I thought interesting enough for inclusion.  It does show just how smart these energy chips are getting.  This particular one does an I vector sum of all three phases and neutral, and phones home when it notices an imbalance (well, raises an IRQ at least... leaving the meter manufacturer to decide how to phone home).

Robin,

The sketch I am using is the Mk2i Multiload sketch that you kindly did for me at the very beginning. As it is the summer I have have gone back to the original sketch in normal mode. I don't need the chargers because the DC pv panels alone are more than able to recharge the batteries. The time delays I don't need because I wasn't able to use the ASHP on my battery supply and I don't need the chargers.

The only alterations to the original sketch you e mailed are;

I use SSR's so my pinout is high not low to turn a load on.

I have put a 1 second delay between different SSR's turning on or off. The sketch was too sensitive and the next higher SSR was coming on then the next and then the next and then it would do the same turning them off all very rapidly.

Oh and I have just added the export bias, but that was to check what I was finding.

The sketch I am using is the Mk2i Multiload sketch that you kindly did for me at the very beginning.

That sketch was only for single-phase, so I don't see how you're expecting it to work in a 3-phase environment. 

Because everything is on one phase.

'As I am now trying everything running on one of the phases I wanted to get to the bottom of why the 3 phase meter was showing the pulsing red .......'

I have a different one for when I tried the 3 phases initially.

For kicks I enabled the real energy change-of-direction interrupt on my energy chip to see how responsive it was.  I rigged things up so that I had 160W of "exports" going, and then turned on a 2300W "import".  Both loads were simple resistive loads so I could easily see what was going on.    The yellow trace is the /IRQ trigger from the device to tell me real energy has changed direction.

So I was happily exporting 160W,  which turned into a 2140W net import when I switched the 2300W load on.  I got lucky and happened to turn it on at a zero-crossing.   So that upward zero-crossing turned immediately into a downward zero-crossing as the "export" turned into an "import".  37.6 msecs later it alerted me that real energy had changed direction.  With a smaller dump load it took longer,  so it does seem to be energy based rather than time based.

I guess the "bucket" that everyone speaks of is the amount of energy represented by the LSB of the meter's internal energy register.    In my case each count represents a WattSecond (aka Joule).  At 2140W, that would tick over roughly every  1/2 msec, so it's kinda' curious that it took as long as it did to alert me.  Perhaps it has some sort of minimum overhead time.

BTW, I'm only rewarded with that LSB resolution because my "meter" is designed for maximum inputs of 20A x 270V.  I can't remember how much head-room I left myself when I configured that resolution, but probably not much.   Still, if I was using this device to build a 100A X 270V meter, I could easily achieve a resolution of just 5 WattSeconds (instead of my current 1), which would presumably then become your bucket size.

I've no idea if that's what Tinbum's new meter is doing, but technically I can't see any reason why it wouldn't be possible for them to do that.

I'm afraid I don't have anything to test my meter further than I have already and besides it's the meter operators meter!! I think I do actually have a new meter thats the same for measuring export for my own use for when I have everything set up permanently with my sunny backups. But would that meter be configured the same?

My meters are now reading the following;

Import 1.32 (was initially 0.75)

Export 21.63 (was initially 0.74)

Three phase 2362 (was initially 2352)

I guess the "bucket" that everyone speaks of is the amount of energy represented by the LSB of the meter's internal energy register.  

Yes, I think of the "energy bucket"  as being the range within which energy can freely flow in either direction without charge or loss to the user.  This range is 3600 Joules (1 WattHour) for many types of meter, but it can be as low as 1000 J  for other types.  A non-reversing disc meter has considerably more allowance, around half a revolution which can be 2.5 Wh.  An anti-flicker setting can be seriously slow in that case.

Internally, a digital supply meter will maintain a detailed record of the instantaneous energy state, as does my own software.  When using integer maths, the numerical capacity of the energy bucket in my Mk2i code is up in the millions, so there's no shortage of resolution.

The device I use has an internal 48-bit accumulator that I can't get to.  When that accumulator hits its threshold it causes 1 to be added to (or subtracted from) the 32-bit register I can get to.  At full-scale deflections, that happens at 8kHz.   So if we assume a residential power meter has to be able to measure 32kW (per phase) maximum, they could rig up their V and I dividers to make 32kW result in maximum deflections.  So when 32kW was passing through the meter, the accessible register would tick over 8000 times per second.   If 4 watts was passing through the meter it would tick over once per second.    I think such a meter would have a "bucket" of just 4 J.

Now that accessible register is itself a signed register, the accumulator could put 100 units in, and take 100 units out and there'd be no trace of anything.   That's where the change-of-direction interrupt comes in.   If the f/w writer just reads the register every x seconds you've got plenty of room to play with, but if they also read it whenever the change-of-direction interrupt fires, it restricts your opportunities a lot.

Tomorrow is the big day for one of the meters to be exchanged and 2 weeks time for the other.

I found the following information on one of the meters that may be of interest for those that are interested in how they work and are configured. I haven't read it all, you'll see why when you see the files!!

http://www.share-pdf.com/d0b57148c93945dda54f178fdfe0b06f/Chapter%202_Su...

http://www.share-pdf.com/542d5214ab2742188a790b28aa7834a6/Chapter%203_Ha...

http://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja...

http://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja...

http://www.share-pdf.com/f841c43fba914cfc92ae11201c1d9b4a/Chapter%206_PM...

http://www.share-pdf.com/01cab30fbb40432896845d549851131f/Chapter%207_In...

Very detailed, thanks for sharing.  That's a great quadrant diagram in the first reference.  Not quite enough detail to reveal the "bucket size", at least not that I could find.   Let us know how the reconfigured meter works out.

Well a new meter is in, exactly the same meter as the old one ( the 5219) but a different letter at the end. It has been configured as a 'vector sum meter' by the meter operators who do it themselves.

It seems to be working ok in regard to being able to import on one phase, while exporting more on the others and the meter not increasing, but the MK2i still doesn't seem to want to work with it unfortunately.

If you calculate the energy contribution on each phase separately, and add them all together in the Router's energy bucket, I would expect the existing power-diversion logic to work with your new 3-phase meter.

Being able to measure the flow of energy on each phase independently is key.  Rather than trying to change my Mk2i Router code straight away, it might be easier to start with a monitoring-only sketch for 3-phase.  Once you can reliably detect the flow of energy on each phase independently, the rest of the code should be relatively straightforward.

If I understand you correctly that is what I'm doing. I've not had much time with this meter yet to look into much detail but judging how the last one was with everything on just one single phase I really do suspect it is how the meter is configured.

I really do hope I have made a mistake somewhere though.

I haven't been able to find anywhere on the forum how MartinR does his diversion on his 3 phase meter. Physically his meter looks very similar to mine but the meter installer yesterday was explaining how the single letter at the end of the meter type made such a difference. He also said that vector sum is a specific request, as if I recall rightly, its not to a national standard or something along those lines. He wasn't your typical installer and was very knowledgeable on meters including the configuring of them.

Regarding my first problem of the meter ignoring export I belief there are many people with meters configured like that just don't know and are being overcharged especially if they don't have perfectly balanced pv and import across the 3 phases.

Are you still able to temporarily move all your consumers and PV to a single phase and see what happen in that case, or did you lose that flexibility when the new meter went in?

Yes I can do that no problem so will try it.

Well after extensive testing I am absolutely convinced that my meter either has no 'energy bucket' at all or has one so small that the Mk2i, in its current form, will result in import being measured from the grid.

I have done lots of different setups, eventually resulting in me setting the diverter up just using one of the phases on my other 3 phase meter. (I can only use one phase on this as it is still to be replaced because it won't calculate my usage correctly with export on 2 phases and import on the other.)

Using this setup the led on the meter lit up constant green and stayed like that showing no import and the meter did not advance.

I transferred everything over to the other meter (again only using one phase) with a 'generator change over panel'  and let it settle. The red led flashed at a rate of about once every 6 seconds showing import. I then turned one of  the diverter loads off so I was exporting and waited for the red led to light constantly showing export. Once that happened I manually turned the diverted load back on, the red led went out immediately and started flashing again and continued to do so.

I have led's on the diverter itself to show if it is exporting or importing, (ie more pv than loads or vice versa) and other monitoring displays and none was showing  I was importing from the grid.

This was all using the diverter in normal mode with a very small export bias.

With this and the other tests (meters in series, see earlier) I am absolutely convinced this meter does not behave like the meters we have seen so far.

So you mentioned above that your PV is usually driving 2 of your phases?  Is that 2 inverters, one on each of the 2 phases, or  does your inverter have a 2-phase output?   How are things wired during your temporary single phase experiments?

Inverters are single phase. At the moment I have it wired as 3 phase, with 3 ct's except where both supplies from the meters come into the generator change over box two of the phases have been disconnected from the contactor inputs and a link put in to join those inputs to the live phase that still connected.

So during the single phase experiments two of  the phases coming out of the new meter are not connected to anything?  That sure sounds single-phase to me.  I was wondering if there was a chance there was some loads you'd forgotten about on those phases, but clearly not if they're just dangling in space.

Yes that's correct.

It's conceivable that any digital meter is only a firmware upgrade away from having a very small bucket.  I wonder if that reflects a new policy, or whether a keen Elster firmware engineer just decided it would be neat to enable the change-of-direction interrupt on the energy chip in your particular meter model's latest firmware update.   As far as I know, that bucket size never appears in any metering specs.  I'm not sure there's much you can do since arguably your new meter is behaving more correctly than those that do allow a bit of undetected bi-directional slop.

I'm with dBC here. The documentation for Robin's Mk2 controller makes it quite clear that the meter has a finite band within which energy can flow backwards and forwards without registering, and that "typically" we have found this to be 1 Wh. However, when you stop to think about the problem, it's quite clear that the bottom limit is actually very small indeed, the only requirement is for the accurate measurement of reactive loads, which of course take energy from and return it to the supply in equal measure (if the load is purely reactive) within one cycle. Therefore, it's conceivable that if your meter is working that way, no burst-mode controller will ever work as intended for you. As I see it, you have two options if you require to exactly balance your generation. You can either employ phase control and fit large (and expensive) filters to comply with the limits on harmonics injected back into the supply, or you can develop a high frequency PWM switch that can use a much cheaper and smaller filter.

A third option is if you forego the requirement to exactly balance your load against your generation. If you can arrange to have many loads of varying capacities, you can arrange to switch them so that you select the best combination to balance your generation at any particular moment. Clearly, in general terms this will never achieve an exact match, but it should be possible to get 'close enough' - unless of course your aspirations are very exacting.

Robert, interesting.

I had thought of the many smaller loads but  not your other ideas. I'm thinking it may be possible to have a set up that has one load that is switched with phase control, this being the largest load with suitable filters for that load . Once this load reached a state of being constantly on, or at a certain diverted wattage (set at the next loads size) then that next load would be switched on but as a 'constant on' the largest load would then take up the slack once again etc etc. The filtering would then be for that one load only.

The other option I had thought of was to have an export bias set up within the code such that the size of the bias would match the size of each load. I have added this into the code so I can try it, the down side is the 'wasted' export. I would also try turning this bias down to see if there was energy energy bucket at all within the meter.

It looks as if your scheme would work. The fly in the ointment is that the harmonic current limits are specified in absolute terms, and high current filters are naturally more expensive. And it requires a high performance, high cost filter.

In an ideal situation, the largest load is half your generating capacity, the next is half that and so on, so that switching in a binary sequence gives you all possibilities, and in that case putting phase control on one that is the same size as the smallest switched one, or on the last but one smallest, will give the full range of control. In practical terms, this will not be possible, so you'll need to get as close as possible. Therefore I'm suggesting that the load that you will phase control should probably not be the absolute largest, but the smallest that will give you continuous control under all permutations of the switched loads. (I hope you see what I'm getting at. The reason is so that you keep the filter cost to a minimum, you need to keep the harmonic currents to a minimum - absolute, not percentage - so that implies choosing as small a load as practical for that purpose.)

I am absolutely convinced that my meter either has no 'energy bucket' at all or has one so small that the Mk2i, in its current form, will result in import being measured from the grid.

Can you remind us please which version of my Mk2i code you are using.  The interim version that I supplied you with ages ago was prior to Mk2i rev5 so this would not seem a suitable vehicle with which to pronounce judgement on your meter (or my code).

Before you can reasonably conclude that you new supply meter has no sweet zone (or one that's too small for a burst-mode diverter to operate within), I think you need to arrange for the following test:

1) Choose a suitable sketch of mine, such as Mk2i rev 5d, and make only the absolute minimum of changes.  I see these as being just the IO pin allocations (to match your hardware) and the loadStates enum (to provide active-high control rather than active-low).  This code needs to be run in its "normal" mode rather than Anti-flicker.  For the purpose of this test, running in AF mode would be just asking for trouble.

2) Arrange for everything in your premises to be running on just one phase; the other two must have nothing connected to them at all (as you've confirmed above)

3) There must only be one CT in use.  The burden for that CT must be sufficiently low so as to ensure that the ADC will never run out of range.  For this purpose, the emon Tx V3 uses 22R and the V2 version used 18R.   Any value below about 40 ohms should be fine.  For the purpose of this test, you can't continue to use 150R as per my original design.  If you do, it will surely invalidate your test, and thereby defeat the objective.

4) Provide one or more decent sized resistive loads as your dump-loads.  If you want a Mk2 sketch that caters for multiple wired dump-loads, there's one on the Downloads page of my website called  Mk2_multiLoad_CAT5_2.ino.  That sketch is ready configured to drive five hard-wired loads in a priority-based sequence.  It is similar to Mk2i rev5d but without the RF overhead.  The display code in that sketch can be safely ignored.  The display logic will run happily in the background without causing any problems.

5) While surplus power (SP) is available from your PV, precisely one of your dump-loads should be cycling on and off at a rate that will usually be too fast to count.  As the level of SP changes, an additional load may need to be utilised or discontinued, and during these inter-load transitions there could be some brief activity at the meter.  But while one of your loads is actively balancing your household's energy state, I would not expect to see any activity at the meter's LED.  The LED may be either on or off, but it should not be changing state.

If you can arrange for all the above conditions to be met, and your meter is still pulsing every six seconds, then you are right.  You have just found the first known case of a supply meter that won't tolerate a burst-mode diverter.  But if you can't arrange for a proper test as above, then I don't see how you can reasonably blame either your meter or my software for not working together in the conventional manner..

My single-phase Mk2-based diverters are now operating reliably in many countries, on four continents, and no-one else has ever reported such a problem. 

Robin, we have been through all this in PM's and e mails.

I am not prepared to go through it all again. You chose to not listen to me about the size of energy buckets in the past and now you are doing exactly the same over this. I have explained to you exactly what I have done and either you are choosing not to listen or you simply do not understand.

I stand by what I said.

Nick, I have no wish to go over any old ground again. 

I'm simply trying to establish whether your latest supply meter has the ability to be used effectively by a burst-mode diverter, such as my Mk2i PV Router, or not.  You have claimed that your latest meter can't be used in this way. and I have spelt out how you can prove that assertion.  You should now either proceed with the test, as I have suggested, or withdraw your claim.

Having recently set up a business around my Mk2 Diverter concepts, I am every bit as keen as you are to establish whether your new meter operates just like all the others, or whether it doesn't. 

*  *  *

On 02/07/014, you said of your new meter:

It seems to be working ok in regard to being able to import on one phase, while exporting more on the others and the meter not increasing, but the MK2i still doesn't seem to want to work with it unfortunately.

Seems to me that this debate is more to do with how surplus power can be used most effectively rather than anything to do with 3-phase.  Being able to automatically activate a load that requires continuous power (such as a battery charger) makes a lot of sense, but additional logic would be required to control this.  I'm hoping to start a new thread on this subject soon.

Robin,

I stand by what I say, I have done numerous tests, go and look back through everything.

Nick,

Everything has changed with the installation of your new meter, so I think it would be a good idea if you could run a simple test on just one phase to establish whether your meter really does have no energy bucket.

But first, you need to make a few changes: I see Robin says that you have a 150 Ω burden resistor. If you need to measure a current in excess of 23.5 A rms, the ADC input will clip and the current measurement will be invalid. Robin chose this value because the normal maximum current when the PV can be balanced is less than this, if you have more than about 4.5 kWp installed capacity, there is a problem there.

Secondly, in order to validate your claim that the Mk2i isn't suitable, you need to take Robin's latest version of the Mk2i and try it unaltered (except of course for IO assignments, etc) as he's asked.

If you can see your way to do that, not only will you settle the point, but you'll have done some valuable research on energy meters that builds on the results that MrSharkey published a while ago that were the trigger for the burst-mode energy diverter and that provided much of the source material for the "Energy Meters" topic in Building Blocks.

Robert,

I do not need to measure over 23.5 rms. I have a display that shows what is wattage is on each phase and it doesn't go over that or get anywhere near it.

Re the Mk2I sketch, Robin supplied me one. I have run it  and Robin knows I have and knows it did not work.

Despite that having the two meters in series has also shown it without doubt as has the fact that it runs on the other 3 phase meter perfectly even using my own sketch.

Two points:
(1) What is the integration time of your display? I meant 23.5 A rms integrated over 1 cycle (i.e. the maximum rms value of any one cycle).  If you are capable of generating more than that, or you have a dump load that can consume more than that, then even if the nett long-term current flow is less, you have a linearity problem that an ammeter with a long integration time will obscure. 
(2) Are you using unaltered, Robins's latest published sketch?
(3) Have you just written (immediately above: "... it runs on the other 3 phase meter perfectly...") that it is Robin's sketch that runs perfectly on another 3-phase meter that you have? That's how I read it.

Nick, I think were losing sight of the goal here.  To check whether your meter is compatible with a burst-mode router, we want the simplest setup that is possible.  The basic Mk2 sketch Mk2_PV_Router_mini_3.ino should be fine for this purpose.  It runs at the maximum speed with no RF or AF to complicate matters.

For this particular purpose, I think your existing burden resistor should be OK too.  Providing that your surplus power lies within the range 0 - 3 kW (if that is the rating of your resistive load), then the load should rapidly cycle on and off and your meter should ignore it. 

For most of us, surplus power is unlikely to ever exceed 3 kW.  In your case, you can presumably turn on fixed appliances as necessary to temporarily reduce the surplus so that it lies within the above range.

It is possible that the power-distribution arrangements in later sketches such as Mk2i rev5 are not compatible with certain types of meter, in which case I am keen to know about this.  If that were the case, the logic could no doubt be returned to its original state.  Mk2i rev5 was a massive update from rev4 - it included 3 major changes which would have been much better done separately.  Hindsight is a wonderful thing.

I look forward to the outcome of your tests.  It will surely be best for everyone if your meter is found to be OK. 

Robert,

1 I'm using an emontx, so no not over 1 cycle but firstly I don't have a load over 3 kw and secondly I have always tested when the whole system has been in a steady state when there has been constant sun, and the one load has been taking up the spare pv. I should also point out that even though I have a lot of pv the actual output is limited by the inverters so results in much steadier output. Eg 13kw Pv  limited to 7kw

2 Yes, the sketch was unaltered, except for pins and sent to me by Robin, I belief based on his latest sketch.

3 No my sketch, I haven't tried Robins on the other meter.

Robin,

I'll try when I get back but I know what the answer will be.

My sketch is NOT  based on the rev 5.

I've said from the very beginning that I had hoped I was wrong as I'd like to continue using a sketch that has run faultlessly for me, for I think nearly 2 years now and despite the modifications I have in it. 

Can someone please explain how, with two meters running in series, one can measure an import of 24.59 kWh and the other 95 kWh?

Nick, whenever you are able to try a simple "Mk2" sketch such as _mini_3, I very much hope that your new meter will work as intended.  Then that same control mechanism can be imported into the sketch that you've been using for the last couple of years alongside your previous meter.

The upgrade that I provided for your code recently was mainly to incorporate the multi-load control mechanism from Mk2i rev5, but I would not regard the resulting sketch as being suitable for doing any performance checks on your new meter.  The benefit of this code to you is that it incorporates the various different loads which you wish to run.  Some of these loads can't cope with being cycled on and off, so additional logic has been added by yourself over the last couple of years to prevent this.

The thread that I started yesterday is directly aimed at this situation: how to incorporate "continuous" loads as part of a mixed dump-load strategy.  Knowing that your meter is OK with a simple resistive load cycling at high speed would be a  great step-forward.   This is precisely what a mixed-load control scheme needs to be based around.

Certain types of load will need to remain on continuously, but ideally they will be "shielded" by other less-particular loads which can be cycled rapidly to maintain the balance.  If all goes to plan, the more particular type of loads will be charged at the optimum times at minimal cost, with the background resistive loads mopping up whatever surplus power is left.

Can someone please explain how, with two meters running in series, one can measure an import of 24.59 kWh and the other 95 kWh?

What I think is happening here is that the reciprocating energy flow is being recorded differently by each meter.  If their sweet-zones are different, then one meter will be more likely to register chargeable events than the other.  With your loads sometimes being held on for longer than would normally be the case, charges can be expected, but they could well be applied differently by the two meters.

For the simple situation where the energy reciprocations are only small, within a range of maybe 200 Joules, I would expect neither meter to register any charge while your Mk2 system is active.  3 kW is 60 Joules per mains cycle, and the control mechanism takes around 30 mS to act.  So, in the absence of any additional control logic such as AF, the energy state is pegged to the mid-point of the energy bucket very firmly.

"Can someone please explain how, with two meters running in series, one can measure an import of 24.59 kWh and the other 95 kWh?"

I concur with Robin. They are set up differently!

The hard part is deciding how each is set up. If you look at the guidelines in the link in the 4th post in this thread, it says in effect that a 3-phase meter must exactly emulate a Ferraris 3-disk meter with a ratchet mechanism. In addition to Robin's, another explanation is that the one that reads 95 kWh is doing what the guidelines require, with a very small or no export allowed (a small or no 'sweet zone') and the one that reads 24.59 kWh is not doing what the guidelines require - this one has a very large or infinite 'sweet zone', meaning that you imported 95 kWh and exported 70.41 kWh, leaving a nett import of 24.59 kWh.  Those are the two possible extremes. Either or both meters might be not at their respective extreme of 'sweet zone' range, meaning that both import and export values are wrong - but you have no means of knowing which and by how much.

I think there's only one way to resolve this: you need to repeat exactly the tests that MrSharkey did originally. If you can't find his posts, an approximate précis of what he did is: he had a spare meter on the bench (OK, I know you can't do this with your new one) and using a known small load calculated the energy knowing power and time; and ran the energy backwards and forwards through the meter observing the LED and registers as he did so.

Having just returned back off holiday I have just run the test you both wanted me to run.

Using the sketch that Robin linked to above (Mk2_PV_Router_mini_3.ino) with no modifications at all other than changing the state for the triac pin as I use an SSR which needs the pin high rather than low to fire.

Using one single phase from the 3 phase meter.

With nothing at all on the other phases, not even connected.

With nothing connected to the phase being used other than one inverter and the Diverter.

Using a kettle as a load, doesn't show wattage but says 10A.

With the inverter showing approx 1400W being produced.

As soon as the kettle was turned on the red led on the meter went off and started flashing at a rate of about one flash every nine seconds.

As I have already said, the Mk2 diverter does not work with this meter.

What is the calibration of the LED on your new meter? The value should be printed on the front.

1000 imp per kWh so showing about 400w if I calculated it correctly.

Interestingly I have also run Robins phase angle version and that doesn't seem to work with it either.

http://openenergymonitor.org/emon/sites/default/files/Mk2_PV_phaseAngle....

Edit, just realized in that sketch there are 2 pins for triacs and I was using pin 9, I can't seem to get the triac to fire at all with pin 10 though- still trying.

I agree your figure of 400 W.  Your nett import should be 10 A kettle load (presumably at 240 V approx, implying 2400 W at full power) less 1400 W PV generation = 1 kW, therefore it seems as if your diverter is working with your meter, even though it is allowing a nett import. I think Robin and I have always acknowledged that while 1 Wh (3600 J) is a common size, it was not cast in stone and it was you who pointed out that you had encountered values of 3000 J and 1200 J, and Robin wrote that had come across a value of 1000 J.

I think the next step is to adjust the size of the energy bucket (line 59: capacityOfEnergyBucket = 3600) to suit your new meter, and repeat the test. If that gives you a balance (with no export bias), then you have a known good working base which you can incorporate into your multi-load sketch.

Robert,

Done that already and still didn't work..

Yes I discovered my single phase meter had 1200J and I've been working my sketch with that for the last 2 years.

I'm thinking that if you look at say;

http://openenergymonitor.org/emon/sites/default/files/resistive2.jpg

what I need to do, in laymens terms, is to chop all the peaks off on the yellow line at a power level that is equal to the surplus pv as I think above that is what the meter is measuring.

Your much more knowledgeable than me on this, would that tie up with the 400W which, after all, is just an average?

"would that tie up with 400W?" Er, no. And you can't do that anyway. Your triac, when doing both phase and burst mode control, chops the waveform time-wise, not amplitude-wise. (A thought: did you change the opto-driver IC when you tried Robin's phase-angle sketch?)  What you're trying to do, in terms of that diagram, is to reduce the area of the yellow part, not by lopping the peaks off but by chopping cycles out. If you were using phase control, you'd be lopping the side off each half-cycle to reduce the area (see Robin's write-up for how that works).

It's very hard to explain the operation of a meter - I assume you're familiar with the BB page about energy meters. The easiest to understand is the Ferraris meter. If your meter had 3 disks on the same shaft, then the speed of the disk would be proportional to the total energy, and if one phase was generating, that would slow the disk down or reverse it. Now add in a ratchet and switch the load on and off and you can see that the disk will jog backwards and forwards, but when it hits the ratchet going backwards, it stops and ignores all further exported power. The faster you switch the load on and off, the less distance it will jog, and the aim is to jog without hitting a tooth on the ratchet. It's the distance between teeth on the ratchet that we need to establish. That's what Robin calls the "energy bucket", and the meters page call the "energy packet".

If my assumptions about your kettle load are correct, then your meter does have an energy bucket, our problem is in establishing the size of it. What sizes of energy bucket have you tried?

Yes I understand all that and how its time based but that's not what I mean. I think my meter is measuring the power above a line that would be set at the available PV power.

I'm sure there will be a way to calculate that power, what is the value then if its not 400W?

Oh and yes, of course  I changed the SSR but i cant get it to fire with the code. using a Crydom  D2440-10

I still disagree about the energy bucket.

Let's deal with the simple things first: Sorry, I forgot you were using commercial SSRs. The D2440-10 does look to be the correct version, and the input also looks as if it ought to be driveable by your AVR, so I can't see a simple reason why the phase control sketch does not work.

Sorry (again), but I don't understand what you mean by "I think my meter is measuring the power above a line that would be set at the available PV power." The meter has absolutely no knowledge of the PV power available, all it does is measure the nett energy passing by. It does however know the phase angle and the direction of power flow. It most likely does have a bottom limit below which it ignores all power, that however will be quite small - a watt or two.

I've been thinking again about what I said about  'above a line etc' and I think I've got myself slightly confused there. I've been thinking of the PV more as a constant power rather than sinusoidal.

The SSR will work with pin 9 in the phase control sketch ok but not on pin 10. I may have to swap the pins round to make sure pin 10 isn't faulty on the UNO. Its pretty well the same and even the same manufacturer as all my other SSR's that I've been using for ages except that they are zero crossing.

I'm going to try again tomorrow with different energy buckets and make more detailed observations. So far I've just looked to see if the red led was still flashing and in all cases it has. I've normally had it set at 1000 but I have tried various values from the 3600 to even as low as 100. I didn't think the capacity actually made any difference though for this sketch as once the triac is firing isn't it just a value that if the bucket is above half of it the triac fires and if it's below it turns off. I know though for my own sketch with its antiflicker it is vital.

I think you're right about the energy bucket size and that sketch. As you say, the load is turned on above half-bucket and off below it - the bucket size determines how long it takes when it is coming back into balance from either direction, i.e. both when there has been excess generation and when there has been insufficient generation. The bucket must be large enough so that no energy is unaccounted for while the system ought to be in balance.

If you have no other indication, it could be useful to put a lamp load in parallel with your kettle just to check that the SSR is doing what it should be.

I have one of those neon socket testers plugged into it.

Think I have found why the phase sketch won't work.I have the energy bucket at 1000J but in the sketch the energyInBucket has to be between 1300J and 2300J to fire the triac.

Well at long last I actually have some good news.

Using the phase angle sketch; 

http://openenergymonitor.org/emon/sites/default/files/Mk2_PV_phaseAngle.ino_.zip

does work with the meter. I've tried different values of PV, by setting the inverter output, ranging from 100w to 1700w (not very sunny here when doing it) and different dump loads from 500w to about 3600w. The meters red led will stay lit for a good length of time and then occasionally it will flash it every so often.  Setting safetymarginWatts at about 100w helps increase this time.

Now to see if I can do my multi load diverting based on using this phase angle. 

edit   strangely at around the 600-700w of spare pv the led is flashing at about once every 30 seconds

As soon as the kettle was turned on the red led on the meter went off and started flashing at a rate of about one flash every nine seconds.

OK, Nick, this sounds an excellent test setup.  When the kettle started, was it on continuously or was it flicking?  With only 1400 Watts of surplus power available, the power to your kettle should have been cycling at approx 10 Hz.  If it was indeed behaving in this way, then I would have to agree with you that your latest meter is not behaving as it should.  But if your kettle was on continuously, rather than flicking, then something must have been wrong with the Mk2's setup.

Think I have found why the phase sketch won't work.I have the energy bucket at 1000J but in the sketch the energyInBucket has to be between 1300J and 2300J to fire the triac.

In most of my Mk2 sketches, the energy bucket is set to 3600 Joules.  This arrangement works fine with all types of meter as long as the switching thresholds are appropriately set.   In some of my later sketches, I have re-sized the energy bucket to match SWEETZONE_IN_JOULES.  This is a much tidier approach than altering the value of JOULES_PER_WATT_HOUR from its scientifically correct value of 3600.

When the kettle started, was it on continuously or was it flicking?

Flickering

Just found a manual of sorts for the meter:

http://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=3&ved=0CC...

Flickering

Interesting.  With my Mk2_PV_Router_mini3  sketch running, import and export will be kept very closely in balance.   Given that your new meter aggregates forward and reverse flow, it does seem to be set up in the proper way for UK use but has minimal tolerance to reciprocating flow. 

Not sure what you can try next.  I trust there won't be a world-wide roll-out of this type of meter any time soon ...

I trust hope there won't be a world-wide roll-out of this type of meter any time soon ...

FTFY.

If that does prove to be the direction in which the metering world is headed, then that is an incentive to develop the high frequency PWM switch concept that I mentioned here a while ago. It should not require the degree of filtering that phase control does.

I do have a filter made by united automation, that I mentioned on this forum some time ago, that someone said would be suitable and it wasn't too expensive.

http://openenergymonitor.org/emon/node/841?page=2 (second posting down)

It will get expensive if each load is going to need one if I can't design a scheme that would just have one load doing the 'flickering'.

What has got me slightly baffled though is that the phase angle sketch also doesn't seem to fully work with this meter either, its much much better but;

Using a load of 2500w and an export bias of 100w there seems to be a range where the meter registers import. I'm not exactly certain of the center of this range but initially it seems to be at about 650w. Its hard to do with a solar inverter due to the sun but I have been able to set the inverter output to get a general idea. The smallest gap between led illuminations has been roughly about 30 seconds ie about 120w so on top of the bias that's about 220w.

I understand the principles of the sketch but not fully what any difference in the energy bucket limits would change and also if

firingDelayInMicros = (asin((-1 * (energyInBucket - 1800) / 500)) + (PI/2)) * (10000/PI); 

would require changing. I can see 1800 is the mid point of the bucket but is the 500 because each limit is 500 above and below this center point? (I've not had time to read up where this equation comes from yet.)

I have just realized that I think I may actually have one of these meters that I bought last year for use with the self consumption on the sunny backup, I can't get at it straight away but will give it a test as well if it is the same.

Oh and tomorrow I have my other meter changed !!!!!!!!!!!! 

That formula was my best attempt to achieve an accurate degree of control over the energy range of interest.  The 1800 figure is the centre of the energy bucket, but this is just a nominal value.  The algorithm operates in a pseudo-proportional manner over a range of 500 Joules on either side.  So when the energy level is below 1300 J, the load is fully off; above 2300, it's fully on.  I posted some video footage of that system working here.

Given that your meter has minimal tolerance for reciprocating energy flows, it will be interesting to see how it gets on with the phase-angle controller.  Albeit on a faster timescale, exactly the same principle is at work.  When running at 50% load, surplus energy  is being exported during the first half of each half-cycle, and then a similar amount is being imported during the second half of each half-cycle.  For a 3 kW load, the 'depth' of this reciprocating flow will be 15 Joules.  

With a bit of luck, your next new meter will behave in a more conventional manner ...

The meter must be really bad then if its still registering when I'm testing with a 2.5kW load ie 12.5J

Unfortunately it's the other meter, that does work with the Mk2i that is being replaced. It doesn't calculate correctly over the 3 phases.

Oh, what a joke, the meter installer has just been but they don't have the meter they have been requested to fit. That's despite it all being agreed weeks ago with them.

"That formula was my best attempt to achieve an accurate degree of control over the energy range of interest." The graph of power vs phase angle is highly non-linear. A large change of phase angle close to 0° and 180° produces very little change in power, whereas around 90° the change in power is great. This of course manifests itself as a change of loop gain. There's little chance of an unstable system resulting, and the feedback loop will take care of the non-linearity anyway, but we thought that an attempt to linearise the response was not a bad thing.

Yes, the phase-angle sketch can use either of two control algorithms, one as above and one somewhat simpler.  Both versions appeared to work OK on the bench.

// simple algorithm (with non-linear power response across the energy range)
//        firingDelayInMicros = 10 * (2300 - energyInBucket);
       
        // complex algorithm which reflects the non-linear nature of phase-angle control. 
        firingDelayInMicros = (asin((-1 * (energyInBucket - 1800) / 500)) + (PI/2)) * (10000/PI);

The p-a MK2 sketch was used to produce our hot water for a few days, so I know that it does work with our standard 1-phase meter.  During that period, it was probably using Robert's complex control algorithm rather than my simpler one. 

I've been using the complex one.

OK, I hope you understand what it's doing.  It looks a bit frightening to me  :)

No! hence the question.

Just found I have 2 of these meters one with suffix A and one B.

Their isn't much inside it at all and the chip is hidden by the LCD, but I'll see if I can get the number from it. dBC seems very knowledgeable on these and may be able to shed some light!!

Photo of the chip in the 5219A Meter approx age 2005

Not one I recognise sorry.

I've had a chance to do a bit more testing on a couple of the meters I have using the following set up;

Arduino Uno running a modified version of the blink led sketch. I have altered it for 4 output pins, with 2 pins high at the same time and 2 low.

Variable delay time

4 zero crossing ssr's.

The idea is the Uno alternates the SSR's between high and low so that they can swap the wires on the 2 inputs for a single phase on the meter.ie simulating export and import in turn.

Testing this on a 5192F meter gave the result I was expecting, with a 500w load and the time set at 7000 milliseconds the red led stayed on, at 7400 milliseconds it went off. (I was expecting a 3600J energy bucket)

Moving to the 5219 meter, similar to my DNO one and doing the same test resulted in the red led going out. Trying different times, even down to 20 milliseconds resulted in the same result. Even at 100 milliseconds the led pulsed every 23 seconds.

I then substituted the 500w load for a 60w one.

At 200 milliseconds time the led was going out after 2mins 12 seconds and then pulsed every 2 min 12 seconds. At 100 milliseconds the result was better with the led going out after 6 minutes and then coming on solid after a further 4 minutes. It then went out again after 31/2 minutes and then came on solid again after another 4 minutes.

I've come to the conclusion that there isn't any leyway with this meter. Perhaps I should take the option of having my own meter installed instead of this one!!

This sounds to be an excellent arrangement for testing your meter on the bench.

At 200 milliseconds time the led was going out after 2mins 12 seconds and then pulsed every 2 min 12 seconds.

With a 50% duty cycle, your 60W load will be drawing 30W in the forward direction.  At this rate of flow, each Watt Hour will take 3600 / 30 = 120 seconds.  If the actual rating of your 60W load (or the supply voltage) is slightly low, then your figure of 2m 12s appears to match this value well.  The obvious conclusion - as you say - is that this particular meter is making minimal allowance for reverse flow; it simply measures the flow of energy in the forward direction.  

When you increased the cycling rate from 2.5Hz to to 5Hz, I suppose it's possible that the anti-creep mode was cutting in and out, depending on how this mechanism actually works. 

When running in "normal" mode at 50% power, I would expect a Mk2 router to be cycling at 12.5 Hz, i.e. 2 cycles on and 2 cycles off.  Because of the interleaving, it's not able to toggle the state of the load every cycle under these conditions.  At either end of the operating range, the load can come on or off for individual cycles, but not when amount of surplus power is near the centre of the range.

Operating your diverter in phase-angle mode would seem the most feasible way of getting around this problem.  Then there would then be an approximate I/O balance during each half-cycle.  Although being rather unkind to the mains, I don't see how the meter could legitimately penalise you for that pattern of energy flow. 

Unfortunately the meter still registers import in phase angle mode as well !!

Is the phase-angle sketch working OK with your other meter(s)?

Robin, I've not tried, will try tomorrow if i have time.

Did a bit more testing with my rig today and actually managed to get the red led to stay on using 5 milliseconds import with 40 milliseconds export and a 100w load. Note the SSR'S are zero crossing and the UNO is not tied to grid frequency. The 40 milliseconds export was just to make sure I always had more export than import.

A setting of 11 milliseconds import the led wouldn't stay on.

Did a bit more testing with my rig today and actually managed to get the red led to stay on using 5 milliseconds import with 40 milliseconds export and a 100w load.

With that sequence, the load is only on for one-ninth of the time, so the average consumption is approx 11 Watts.  That's probably below the meter's anti-creep limit. 

Using a couple of low-power modern light bulbs  (on continuously), you should be establish where the A-C limit is set.  I think it's generally around the 12 - 15 Watts mark.

A zero-crossing SSR is likely to miss many of the 'on' commands if they are asynchronous and of only 5 mS duration.

Unfortunately I don't have another 3 non zero crossing ssr's of the right control voltage.

I'm not looking at this from the point of view of average consumption more what the maximum wattage is at any single point in time that can go through the meter without registering any import (but not as anticreep).

I've just been looking at German meters as they are what is recommended for self consumption usage with my sunny backups. A couple of things I've noted, they seem to do 10,000 imp per kWh and also they have a very neat idea where the meter is just 'plugged' into a back board so makes swapping very very easy. It also doesn't disconnect the supply to the house even when the meter is being swapped. (search for EZH)

Unfortunately I don't have another 3 non zero crossing ssr's of the right control voltage.

The standard BTA41 triac with the MOC3021 trigger will do the same job, as shown here on the bench and here in the garage.   If you apply fixed loads until your surplus PV is less than the dump load's rating, you should only need one of these. 

For my test rig I'd need 4.

Personally I much prefer the SSR's, much simpler, neater and safer. They do come up at good prices if you can wait!!

For my test rig I'd need 4.

Sure, I was just suggesting a simpler arrangement which you could set up today with what you already have to hand.

A bit of good news at last. I had the second supplies meter changed yesterday, some 2 months after reporting it faulty, and I'm pleased to say it calculates correctly over the 3 phases and best of all it has an energy bucket. Now that one is resolved I need to change suppliers as the supply its on is nearly double the cost of the other!

Robin, once I have a bit of time I will see if the phase angle sketch works with this one.

I still need to resolve what to do about the other one, it may be a case of buying one, possibly the same as the one just fitted as above and having someone fit it.

The Elster 1700 meter has now been in and connected for 4 days and it certainly works ok with the Mk2i. It's still reading 00000 kwh's !!!

Nice One.  At that rate of consumption, it probably doesn't matter how much you are being charged per kWh :D

The meter has just clocked over to 00001 so it does work!!

I have just bought a German meter see link below;

http://www.hager.de/produktkatalog-neuheiten/energieverteilung-und-zaehl...

What I find interesting with this is the LED is 10,000 imp/kWh but its not a visible led but infrared which makes it a little harder to test. It does have a nice feature that shows on the LCD screen the actual power in Watts being consumed/exported. One  thing I don't like is you need to use a torch to scroll through the display as it has no button.

I have done some initial testing and this initially looks to be similar in operation to the Landis+Gyr 5219 with no usable energy bucket.

Many webcams and digital cameras can see IR.

Thanks Robert, I have a camera microscope that I have put over the top. Not quite as easy as the eye.