Powerline/HomePlug and split consumer unit

jonoro

Powerline communications and split consumer unit.
Since Sy Q was installed some 3 months ago I have had endless problems with the mini box freezing and failing with a variety of errors. The mini has been replaced twice and I now have 3 boosters, but still problems. I thought I would try Powerline. I bought 3 Devolo av1200 units, 2 with wifi.
I searched online to see what I could find re Powerline and split consumer units, i.e. with the routing going through MCBs and RCDs. Some reports say no problem whereas others say they do give problems.

I live in a bungalow with the consumer unit in a detached garage some 7 metres away from the bungalow. The split consumer unit, a British General unit, features the standard 2 RCDs each feeding a few MCBs. My bedroom power sockets are all on one ring and the living areas on the other ring. The router is in a bedroom and the Sky mini in another bedroom, the main Sky Q 2Tb v2 box is in the lounge.
Firstly I plugged all there Devolo units into an extension strip, set the units up and I could see in Devolo Cockpit software speeds of around 1,100Mb/s. I then moved the units to where I wanted them and repeated the look with Cockpit. Whilst the bedroom unit, on the same ring as the host, showed around 500Mb/s the one in the lounge was down to 200Mb/s.  This for where the SkY Q main box was (Lounge1.) I later tried the other side of the lounge (Lounge2).  I had nothing else plugged into any of the sockets or adjacent sockets to the Devolo units.
I decided to try bridging the rings, both live an neutral with a capacitor to see what difference that would make. Looking at the AV2 lowest frequency of around 1MHz I selected a 22nF 630V to start with. This would have a reactance at 1MHz of around 7 ohms and around 150K ohms at 50Hz. This would give a current at 230V 50Hz of around 1.5ma. The capacitors were axial metallised polypropylene at around 50 pence each.
In normal use there would be no 50Hz current through the capacitor as both ends would be at the same voltage. However if one MCB tripped then assuming a heavy load on the tripped ring this could give rise to a current imbalance in the RCD but only 1.5ma which should not present a problem to a 30ma RCD.
I also decided to try 100nF capacitors to see what difference that would make.
The results are below;
CaseAdaptionBedroom Mb/sLounge1 Mb/sLounge2 Mb/sCount
Rcd & Mcb1No change5462054154R & 2M222nF neutral bridge5552475242R &2M322nF live&neutral bridge5682987110r & 0M422nF neutral   22nF MCB5622525250R &1M5100nF live&neutral bridge5882936960R & 0M
Case 1 was with no changes.
Case 2 had a 22nF across the two neutral buses
Case 3 was 22nF capacitors across both neutral and live buses.
Case 4 was 22nF across neutral bus and another 22nF connected to the switched side of the other live bus via a spare MCB
Case 5 was as per case 3 but with 100nF capacitors.
The count column lists the number of RCD and MCB traversals that the Powerline signal has to make from one ring to the other taking the worst case of it using live and neutral pairs.

The results show that in my case a 22nF across the neutral buses brought around a 25% speed improvement and with capacitors across both live and neutral buses a 50% improvement. Comparing cases 2 and 4 showed no improvement when an additional capacitor was used onto a spare MCB (Yes it was closed!).
Increasing from 22nF to 100nF made little difference.

I am not suggesting or inviting anyone to install such capacitors, I was only trying to discover some evidence about split consumer units and Powerline.

As for Sky Q – so far rock solid on the mini box after 3 nights fed only power powerline – I turned off all the wifi and have verified there is no Sky wifi being broadcast.  Fingers crossed.

ChuckMountain

Good write up and findings.

Can I doublecheck the speed you are reporting is the speed in the interface rather than a "real" test?

Would be interested in seeing something like iPerf being used as the speeds above will be duplex and without any overheads.  That way you would see the difference between the reported speed and real speed and you might find your improvements are even more noticeable in those tests.

jonoro

Tank you

Al the speeds were reported by the Deolo Cockpit app running on a hard-wired PC.
I will try and do the same test as it is now with both 22nF in place, on an android app version of Cockpit.  This does report up and down seperately.
Neve used iPrf but will have a go later and report back

jonoro

I just had a look at the setup as is.  This has other devices such as PC etc plugged in adjacent to the main PLC as well as recorders and Sky boxes next to the Lounge unit (Lounge 1)
I used Cockipt n the PC and Cockpit on an android tablet.
PC reports bedroom 381Mb/s Lounge1 101Mb/s
Android Cockpit reports
Bedroom 380 down and 320 up.
Lounge1 101 down and 42 up
So it seems that the figures I gave in the table are download only.

jonoro

I have now had the opportunity to carry out speed tests using iPerf ( jPerf)
I carried out two 10 second runs on various configurations.  I did not allow the units any training time but carried them all out approx. 30 seconds after plugging in the main PLC.
For some reason whilst each download run was with 10 measurements, only 2 were conducted for each upload leg - My error in the command I guess?
The bedroom figures show no significant change with adding the capacitors - to be expected as the bedroom and the location of the main PLC are on the same ring.
For the Lounge1 location (which is where I want the network connection) the improvement in speed by adding both capacitors is 51.5% for download (54.28Mb/s to 82.24Mb/s) and 49.6% for upload (27.82Mb/s to 41.62Mb/s).
These figures I calculate as the average of the 20 download figures and for the upload the 4 figures.
I have attached a text file of the raw figures for jPerf.

ChuckMountain

Good set of results.  Shows a good improvement using your technique but I still think it is shocking what they can get away with both advertising a product at 1000Mbps then stating a quick connection speed in the software.  Yet in reality, you are getting 100Mbps or less

jonoro

Yes, but I did go around a few other power sockets in the house  one on the other side of an internal from where the main PLC is showed almost 1,100Mb/s, so not far off their claimed speed.
All down to distance, interference and high frequency shunts across the lines.

ChuckMountain

That's not quite what I meant.

Your own figures show that the cockpit figures are way off what you actually got with a real test.

Even at 1,100Mbps it will be substantially lower via iPerf.

Plus 1,100Mbps is a duplex figure, if we sold gigabit switches on the same basis they would be advertised a 2,000Mbps

jonoro

Oh I see sorry.
Thank you for pointing that out, I now better understand.

tom 2000

I will jump in even though the technicalities of this thread are well over my head.
For years I have served my outbuildings and stables with home plugs. Currently tp-link 600Mbps. They have worked well enough, chopping my BB speed in half but quicker that intermittent 4G. The stables have their own little distribution board with fuses and circuit breaker.
Lately the Hps have just refused to pair. I have tried multiple combinations and paired them in the main house and also within the stable block but I cannot effect a pair between house and stables.
There has been no change to my installation and potentially noisy domestic appliances seem not to matter.
Any ideas? Thanks.