@Carl 

 

If you happen to speak to an "engineer who knows things", maybe you could ask what HIS priority wouldl be for reducing electrica noise.  Where would he start, and what would he leave to last?  Articles on the web seem to give conflicting advice.  Which is more important:

 

• Electrical motors (refrigerators, washing machines, dishwashers, garbage disposals, HVAC blowers, garage door openers), or
• Things with switching power supplies (cell phone chargers, HiFi systems, UPS systems)
• Lighting systems, LED lights, lights on dimmers

 

Would he start with electrical outlets that are:

 

• On the same circuit as Powerline adapters
• Physically close to Powerline adapters
• On the same electrical phase as the Powerline adapters

 

Are electrical reflections a source of problems for Powerline?  i.e. a circuit that instead of going from one outlet to another until it finally reaches "the end", has a junction box with branches that might have several outlets strung off in one direction and several outlets strung off in another direction.

 

I am ordering a bunch of surge protectors that claim to have noise reduction.  I have "nothing to los." and the initial investment will be modest.

@Carl 

 

Another question for the engineers.  Is there any value in using one of the EMI Meter devices, such as the ones by Alpha Labs

If spending $160 would allow me to measure the effect of adding EMI filters here and there, it might be worth it.

On the other hand, if this is just more "mumbo jumbo" aimed at people afraid of radiation affecting their bodies, I'll probably pass.

 

 

@Carl 

 

Great.  This is exactly what I wanted.  Here are the followup questions:

 

• "How does the customer mitigate this interference?"
  My "plan" is to isolate these sources of interference by plugging a combination surge suppressor/EMI RFI filter into the outlet, and then plug the source of interference into that.  Would using an extension cord to put the device "farther from the main circuit" (a) work just as well, or (b) not work at all.  Is an EMI/RFI filter (a) essential, or (b) basically worthless.  Does the PA-9020 "pass through" outlet contain EMI/RFI filtering to protect the circuit?  (The UPS for my modem/router/desktop is plugged into the PA-9020 passthrough outlet.)
   
• "Why does the transmission rate reported by the tpPLC utility vary so much from second to second?"  i.e. from 20mbs to 120mbs.  And, "Why is the transmission rate reported by tpPLC greater from a near adapter to a far one than it is in reverse (from the far adapter to the near one), or "vice versa".  Like 80mbs in one direction and 140mbs in the other.
   
• "How can a set of PA9020's work fine for days and weeks, and then one adapter "goes red" and the tpPLC Uutility reports "?" for the transmission rate to it, but reports an actual value on the rate from it?  All of the motors, refrigerators, microwaves, washing machines, etc. get operated all the time.  The "high interference" sources like cell phone chargers are plugged in all the time.
   

Thank You for following up on this.

This explanation does not address the difference between powerlink link speed and internet throughput. Even when powerline link speed is 1200mbps on AV2 2000mbps rated pa9020p the internet throughput is sub 300mbps? Why. Noise on line can relate to powerline link rate but not throughput. It is a badly written firmware which throttles the TCP transfer rate. So please correct your post or amend it.

@Carl 

Do not agree with you sorry.

I have a 500mbps service on a 1gbps ethernet. My WAN speed is very close to 500mbps. Ookla speed test proves it!!!

On 1gbps ethernet with 1gbps service (hyperoptic ISP) you can receive close to 900mbps.

 

So please avoid causing TPlink customers confusion.

The PA9020P is a badly designed product as far as I can see or very poorly marketed with incorrect performance claims. 

 

@Carl 

 

I think perhaps we have wandered off topic.  Yes, we should all agree that there is usually a difference between the listed "Link Rate" and what actually happens.

When I subscribe to 500mb internet and it is delivered on a gigabit ethernet port, the packets actually arrive at gigabit speed, but only 500mb of packets will actually be sent to me.  (More or less depending on lots of factors.  My "200mbs" speed from Spectrum usually clocks in at around 215 to 230.)

 

Powerline has two "link rates", (1) the gigabit ethernet port, and (2) the reported "Link Rate" between any two adapters.  not only different between any two adapters, but different from one second to another and different in both directions.

 

When I attempt to measure the maximum througput between two computers over a gigabit connection using iPerf3, I consistently get about 949mbs.  But, iPerf3 has practically no overhead (no disk writes, no computations. totally full duplex, etc.  Only the minimal TCP overhead.)  Regular activity cannot possibly reach the same level.

 

The issues, as I see it, are:

 

1. How much of the reported link rate between adapters can be realized?
   How much overhead is there in Powerline transmission compared to TCP/IP?  (I believe there is extensive error checking and redundancy.  Does that take up 5% of the bandwidth, or 50%?)  We are so used to working with "switched ethernet" that we forget how networking with shared media was constrained.  (Remember when we had ethernet "hubs" before switches?  or, when ethernet was run on an enormous coax cable?)
    
2. Is transmission between adapters affected by concurrent transmission on the wire?
   i.e. if between A and B I have 200mbs, and between C and D I have 200mbs, can I put a total of 400mbs (less overhead) on Powerline at the same time?
   Or, is it more likely that as soon as any adapter begins sending, the available capacity is immediately reduced?
    
3. What causes an adapter to "lose sync" (Powerline LED turns RED) and essentially quit working?
   Shouldn't it temporarily slow down, but resume when whatever caused the problem goes away?

 

My next step is to attempt to measure #1 and #2.  i.e. run my iPerf3 measurement across Powerline and see how close they come to the tpPLC reports.

@UnWired 

 

Having just spent time crawling around to plug my laptop into Powerline adapters, I thought my results may be of interest.

 

iPerf3 on this Windows 10 laptop achieves about 950mbs when connected to a desktop through a gigabit ethernet switch.  i.e. When TCP overhead is considered, basically maxing out the gigabit connection.

 

My four PA-9020's are arranged in a line, with the test desktop at one end (#1).  tpPLC reports some measure of speed as follows:

 

 

From #2 to #1 - 1,000 to 1,200mbs, return 1,000 to 1,300mbs

From #3 to #1 - 350 to 400mbs, return 400 to 500mbs

From $4 to #1 - 100 to 180mbs, return 40 to 90mbs (but I have seen as low as 20 in the past)

 

When the laptop is connected to each PA in turn, this is what iPerf3 reports

 

From #2 to #1 - 155 to 170,  return 180-200

From #3 to #1 - 80 to 90,  return 88 to 98

From #4 to #1 - 40-45,  return 4-20 (wow!)

 

My conclusions are:

 

• Whatever is being reported by tpPLC is not a measure of actual capacity.  In every case, The amount of data transmitted is in every case 1/5th to 1/6th of the number provided by tpPLC.  It would be really fascinating to learn how the engineers describe this number.
• Whatever these numbers are, they behave as one would expect.  I have no way to measure the actual distance the electrical wire runs within the walls.  It easily could be 100ft. from #1 to #2, another 100ft from #2 to #3, and another 100ft from #3 to #4.  Total distance could be about 300ft (more or less).  We know Powerline is affected by distance.
• All of these raw transmission rates are adequate for almost anything one would want.  4K video requires 15 (Amazon) to 25mb (Netflix).  I should be able to connect 4K tv's at three of these locations with no issue.
• My PROBLEM is with #4. In the few minutes that I was watching iPerf3, the throughput dropped to 4mbs.  I think we also know that electrical noise affects the receiving end more than the sending end.  So, #4 is able to send to #1, but occasionally has real trouble receiving from #1.

 

Armed with a surge/EMI filter and a box of ferrite cores, I will do battle with everything plugged into the circuit near #4.

 

Sorry to rave on for so long.  If I have made serious errors in design or analysis, please point them out.  My quota of mistakes is "5 small and 2 big per day".