Author Topic: Hard wired communication link?  (Read 19251 times)

GAR

  • Full Member
  • ***
  • Posts: 131
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #30 on: August 25, 2011, 05:06:55 AM »
110824-2053 EDT

PV-Skip:

I think the question on the phases was answered, here or one of the other threads.

If you simply go to "export", select the averaging time as seconds, save to a file, open the file with Wordpad, then you have a human readable ASII data set. On 1000 series dumps you must use Wordpad because the file has no carriage return characters. On the 5000 series carriage returns are included and most text editors or word processor should provide a usable output.

.

PV-Skip

  • Jr. Member
  • **
  • Posts: 67
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #31 on: August 25, 2011, 06:23:33 PM »
Yeah, I can see the data nicely with excel.

PV-Skip

  • Jr. Member
  • **
  • Posts: 67
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #32 on: August 26, 2011, 09:10:17 PM »
Installed my dedicated, filtered line for the gateway last night.
Some hick-ups, but the corrupted data % is down to about 3 now.

GAR

  • Full Member
  • ***
  • Posts: 131
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #33 on: August 26, 2011, 09:53:03 PM »
110826-1351 EDT

PV-Skip:

What were the hick-ups?

.

PV-Skip

  • Jr. Member
  • **
  • Posts: 67
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #34 on: August 27, 2011, 12:20:44 AM »
110826-1351 EDT

PV-Skip:

What were the hick-ups?

 After installing the filter and changing the setup to "120V (bk, wh)" it messed up my average values.
It doubled the "KWh used this Month" and "Projected KHh usage" and messed also up the use since midnight.
Had to erase the history last night and this morning I realized that the real time usage was a negative number now.

I assumed a failure of the 2 firmwares, which I installed again, but no luck.
Between all of this my usage jumped around and went up to over 5000 Watts with nothing running in the house.
Finally I read through the trouble shooting guide and found the suggestion to turn one of the main current sensors. Did not make sense to me at that time, but I did it anyway.
My ghost usage was cut in half, so I turned the other one too.
Now the red dots are facing the breakers and everything is running fine.

Ted from the Enphase forum explained the sensor problem to me and it made sense.
I could have also swapped the clamps or the sensing wires to the opposite side of the phase, but at that time I did not fully understand the implications of the accidental change of the power wire to the MTUs.

GAR

  • Full Member
  • ***
  • Posts: 131
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #35 on: August 27, 2011, 06:47:56 AM »
110826-2115 EDT

PV-Skip:

Your description of the problems tends to confirm my judgement that there are fundamental design defects in the 5000 series software. I do not know where or why, but various different problems that are discussed on this forum probably result from some fundamental defects in design. From the outside it is also very hard to know what is going on inside a complex "black box".

First, stored data should not be modified without your specific OK. So changing parameters or setup structure should not change data without your ability to save the data first first.

From experiments I have run on both the 1000 and 5000 where I use a single current sensor and observe the output power reading when I reverse the sensor orientation has shown no sign change. Ideally I would expect a sign change, but apparently there is no display of a sign. Thus, we do not really know what is happening internally. What does occur is if two sensors are used on one load wire the power reading is double the power reading of a single sensor when one is orientated opposite the other, and zero or near zero if phased the same. This is what is expected.

Investigating my defective 1000 series MTU I found the following:
Each current sensor has one red and one black lead.
Both sensors are made and phased identically. Thus, if orientated in the same direction on a single wire and both red leads are connected together, then voltages from black to red for each sensor will be nearly identical in both voltage and phase.
That this is correct I verified by measuring the voltage between the two black leads and it was essentially 0. This is expected when you measure the voltage difference between two identical voltages.
With about 12 amperes flowing thru the sensor the voltage from one sensor was about 175 millivolts RMS. With one sensor reversed the voltage difference between blacks was about 350 millivolts. As expected.
Internally on the PC board the red leads are both connected to the ground plane.
Thus, the circuitry on the board from the sensors is subtracting one sensor from the other electronically in order to add the two signals.

I have to stop the story at this point because there are problems with this website and I can not see what I am typing. The display keeps shifting from the cursor position.

.

GAR

  • Full Member
  • ***
  • Posts: 131
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #36 on: August 27, 2011, 06:50:55 AM »
110826-2249 EDT

I will continue tomorrow.

.

GAR

  • Full Member
  • ***
  • Posts: 131
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #37 on: August 28, 2011, 12:43:41 AM »
110827-1634 EDT

PV-Skip:

I have looked at the datasheet for the Cirrus CS5461A and there is a differential amplifier at the input to each analog channel (voltage and current). The two current sensor signals are essentially added by this differential input. Assuming a similar chip in the 5000 this would be the means to sum the two voltage sources (+ 120 and - 120 phases).

The digital output of average power is two's complement with a range of +/-1.0 . Thus, internally TED could derive sign information for power,

The voltage channel in the 1000 feeds the +Vin pin and I suspect the -Vin is connected to the ground plane. The input voltage divider consists of 120k ohms and 100 ohms. Thus, a 1000 to 1 divider, 120 V in is 120 millivolts to the Cirrus chip. The peak voltage for 130 V in is 1.414 * 0.130 or 0.183 millivolts. Saturation is 500 mVp-p.

For the 5000 series I suspect there is a similar voltage divider to the -Vin pin and thus twice the input voltage to the Cirrus chip.

Suppose the MTU is scaled so that for 240 V applied to the voltage input and 1 A resistive load only thru one current sensor, nothing thru the other sensor, then we expect the power reading to be 120 W. Without changing anything else except removing 120 V from the red wire, then the power reading should change to 60 W. With only the black and white voltage wires used it is necessary for the scaling constant to be doubled to get a correct reading.

Based on the description of your problem this implies that the scaling is being done in the wrong place. The history registers should always be in kW or kWh units. Any scaling of values to compensate for the hardware change between 120 and 240 should be done before that data is combined into the history registers.

Quote
I assumed a failure of the 2 firmwares, which I installed again, but no luck.
Between all of this my usage jumped around and went up to over 5000 Watts with nothing running in the house.
Finally I read through the trouble shooting guide and found the suggestion to turn one of the main current sensors. Did not make sense to me at that time, but I did it anyway.
My ghost usage was cut in half, so I turned the other one too.
Now the red dots are facing the breakers and everything is running fine.
This result makes no sense. So long as the power being displayed is derived from one MTU, then for the current sensors to work correctly both must have their dots toward the meter or alternatively toward the main breaker. Should not make any difference based on my experiments. If more than one MTU is used and the results are combined in a Gateway or RDU I have no idea whether some sign information is used in the combination.


GAR

  • Full Member
  • ***
  • Posts: 131
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #38 on: August 30, 2011, 03:55:10 AM »
110829-1929 EDT

PV-Skip:

I think you have indicated that you want to determine if you are being billed correctly.

My current thoughts are as follows:

Experiment with one TED system first. This means one MTU and one Gateway.

Use two in-line filters. These filters connect to two separate breakers, one on each phase, or one two pole breaker. These breakers will have no other load on either circuit beyond 1 MTU and 1 Gateway. This means there will be negligible voltage drop across these two breakers. just a few millivolts if that much. Therefore the 240 V input to the MTU is essentially directly from the main panel busbars. This voltage will be as close to the voltage measured in the power company meter as is feasible for you to obtain.

One in-line filter feeds the red wire of the MTU and the other in-line filter feeds the MTU black wire. Use the MTU black wire circuit to feed data to your Gateway. This arrangement reduces noise to the MTU black wire circuit from the main panel busbar, and simultaneously reduces the electrical loading of the transmitter in the MTU from devices that at not on the MTU black wire circuit. The only loading will be the Gateway.

The second in-line filter from the other phase allows that phase's 60 Hz voltage to get to the MTU red wire, and reduces noise from the main busbar getting to the MTU and possibly to the MTU black wire side. The filter also reduces loading on the MTU transmitter.

This arrangement will most closely approximate the same signals the power company meter sees.

You will operate in the 240 V mode.

Later I will describe a means to check the TED calibration independent of using the power company meter because that is what you want check.


.  
« Last Edit: August 30, 2011, 03:58:02 AM by GAR »

GAR

  • Full Member
  • ***
  • Posts: 131
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #39 on: August 30, 2011, 06:29:50 AM »
PC-Skip:

You need a high power load resistor to put across the 240 supply.

Most all resistors have a substantial temperature coefficient. This means when operating at a high load the resistance will be different than at room temperature with no current flow. Thus, you can not make an accurate resistance measurement at room temperature and use this for the resistance at high load. Even so you still have to measure voltage and this has to be squared to calculate power knowing the resistance. Thus, if a current meter and a voltmeter of about the same accuracy are used to measure power the voltmeter only with a precisely known resistance does not produce any better power accuracy.

One way to measure power at the high load condition is to measure voltage and current and multiply these together. Using somewhat reasonably priced digital multimeters, about $ 400 each, you can get about +/-1% accuracy. Thus, the power accuracy would be more in the range of +/-2%.

If you had very stable voltage, then one meter could be switched between measuring current and voltage. 

The Agilent U1272A might be a useful choice, and it has other uses. Ideally you would want two instruments so measurements could be made simultaneously. These are limited to 10 A. The chip in the TED1000 has very good linearity and thus extrapolation to higher power levels might not be too bad.

On the other hand if you had a standardized and calibrated TED 5000 it could be your simplest and most economically accurate method. In other words use somebody with better test equipment to calibrate the MTU at a price. Thus, turning the calibrated 5000 MTU into a standard. To do this you would need to determine how much variation occurred from opening and closing the current sensors.

.

GAR

  • Full Member
  • ***
  • Posts: 131
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #40 on: September 02, 2011, 08:52:55 PM »
110902-1040 EDT

Back to my original discussion. Since that point I have learned a great deal more about the TED system.

Some definitions first:

PLC is a name (acronym) for power line communication. This uses a high frequency carrier to transmit data over conventional 120 V power wires in a home or building. TED uses 132 kHz, X10 uses 120 kHz, and others may be at different frequencies. Then there are various devices using the power system that may introduce noise into these frequency bands. Also there are loads that attenuate the signal level.

FILTER generally means to remove or separate. From dictionary.com for electronics the definition is:
"6. Electronics,  Physics . a circuit or device that passes certain frequencies and blocks others."

SIGNAL to NOISE ratio (S/N). How strong a desired signal (signal) is relative unwanted signals or random noise (noise). The higher this ratio the greater is the probability of correcting detecting and extracting data from the desired signal. This ratio can be improved by increasing the transmitting power, filtering to reduce excess noise, reducing noise at the source, and better detection means, such as correlation detection.


Starting with X10 type systems. This system concept is based on the ability to have many different transmitting locations within the power distribution system of a building, and to have many different distributed receiving locations. This has to be viewed as a big network of parallel wires onto which a signal is injected somewhere (transmitted), and that can be received anywhere else. This has major problems from two sources --- shunting loads that attenuate the transmitted signal, and noise generators.

I believe the TED 1000 system started out with the X10 concept that would allow the MTU to be located at the main panel, and then the RDU could be plugged into any outlet for viewing data. This flexibility and no wire installation were selling points. From a marketing point of view seems to make sense. But from a technical view it is unreliable.

Problems were discovered. Namely errors or the system did not work. Thus, in each individual case patchwork solutions are suggested. Experimentation is required and somewhat indirect means are used to see if an improvement occurs. In many cases I suspect that operation may be marginal. Why should a receipt of 20 % bad packets be considered acceptable?

One of the solutions is filters. This is good as a starter. But more tha that should be done.

What is being called a filter I would rather call a filter-isolator to be more descriptive. The word filter being used to mean high attenuation of unwanted signals from the input side of the filter to the output side. This would reduce noise or other signals on the output side of the filter for the TED components of interest on the output side from the noise and other signals that originate on the input side of the filter. This I could do nicely with a cascaded sequence of L-C low pass filter stages. Series inductance and shunt capacitance. But this would be bad for the TED system on the output side because of the last capacitor in the filter. This capacitor looks like a short to the TED transmitter at its carrier frequency. A short to the transmitter lowers its output signal, thus reducing the S/N at the receiver.

So a filter alone is not the solution. At the filter output there needs to be a high output impedance at the TED carrier frequency. Now the TED transmitter won't be heavily load and its signal level will be large. To distinguish the required filter device from an ordinary Corcom filter, such as a 5VR1, I would prefer to call the TED filter a "filter-isolator". I might point out that I have probably 20 or more filters like the 5VR1 scatter through my house. These are on all my fluorescent fixtures to reduce radio noise from the lights.

Note: the filters supplied by TED are X10 filters and these are a "filter-isolator, and do a reasonably good job such that if the TED setup is as I describe next it should perform in a rock solid fashion relative to S/N.

The way to solve the power line communication problem is:

1. Erase the concept that any circuit in the house can be used for for this PLC. Means there is going to be a dedicated circuit for each Gateway or RDU and nothing else is on the circuit.

2. An in-line filter-isolator is installed at the main panel. There is no choice, you use the in-line filter-isolator. If the Gateway uses the red input wire, then a second in-line filter-isolator is required. Again, no choice it must be used.

3. This will be classified as an isolated circuit. There must be nothing else connected to this circuit, except 1 Gateway, and up to 4 MTUs that are associated with this Gateway.

4. Following this approach there should be rock solid communication with no communication problems if all the components are good. I expect bad packets to be essentially 0.

An experiment I ran earlier today with a plug-in filter-isolator as the filter and only a 5000 Gateway and 1 MTU produced a good signal. Adding a resistive load, a 1500 W heater at low power, 735 W (6.13 A at 120 V 19.6 ohms), reduced the signal level to 2/3. Without the in-line filter-isolator, in a normal house with large electrical loads, there could be far more attenuation of the transmitted signal than with the low power heater. That the in-line filter-isolator provides a series high impedance at 132 kHz is why its use prevents loading down of the transmitter from shunt loads on the input side of the filter

.

tazer

  • Newbie
  • *
  • Posts: 13
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #41 on: September 03, 2011, 10:40:06 PM »
1. Erase the concept that any circuit in the house can be used for for this PLC. Means there is going to be a dedicated circuit for each Gateway or RDU and nothing else is on the circuit.

2. An in-line filter-isolator is installed at the main panel. There is no choice, you use the in-line filter-isolator. If the Gateway uses the red input wire, then a second in-line filter-isolator is required. Again, no choice it must be used.

3. This will be classified as an isolated circuit. There must be nothing else connected to this circuit, except 1 Gateway, and up to 4 MTUs that are associated with this Gateway.

4. Following this approach there should be rock solid communication with no communication problems if all the components are good. I expect bad packets to be essentially 0.

Yep, that's really the way to go, and it's what I've been doing here for a couple of months now. It also keeps TED and the Insteon devices separate, since they don't play nicely. I would contend that it achieves what you set out for in the original post for this thread. What should be offered is a single integrated device that is both MTU and filter, and offers a dedicated lead for wiring to the branch circuit where the gateway would be installed. This would differentiate a professional TED install from an amateur one. It might require some of the less tech-savvy users to hire an electrician to wire the dedicated circuit, but that's true of many home improvement tasks. Oh, and put WIFI in the Gateway too, so the Gateway can be located close to (but not inside) the panel without the worry of having to run cat 5 to an arbitrary location.

Although I must say, the TED worked remarkably well before I did the dedicated circuit. The same techniques that were used to eliminate noisy devices from the X10 network sufficed to keep the house clean for TED traffic. The only complaint was that it gave some of my older X10 devices some grief, and the Insteon devices were the final straw the broke the camel's back (the Insteons were more reliable but the dimmers exhibited a weird intermittent flicker that was only present when the MTU was installed).

jrwalte

  • Jr. Member
  • **
  • Posts: 60
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #42 on: September 03, 2011, 11:22:09 PM »
The gateway/MTU doesn't have to be alone. With a wall plug-in filter, it can work fine behind an in-line filter with other electronics on the same filtered breaker. I have a 15A plug-in filter for a UPS, 2 PCs, subwoofer, receiver, xbox 360, router, switch, and cable modem, that are on the same in-line filtered breaker as the gateway and I have no issues. Without the plug-in filter, the gateway has issues and without the in-line filter the gateway wouldn't work.

So it isn't necessary to install a dedicated circuit for the gateway if you can filter other devices on the same circuit with plug-in filters.

GAR

  • Full Member
  • ***
  • Posts: 131
  • Karma: +0/-0
Re: Hard wired communication link?
« Reply #43 on: September 04, 2011, 12:12:21 AM »
110903-1518 EDT

tazer:

Yes it does accomplish much of the goal I set forth in my original post.


jrwalte:

Sure the plug-in filter is a fairly reasonable solution to adding other loads on the same circuit, but it does not eliminate voltage drop changes thru the breaker to the isolated circuit from the variable loads on the isolated circuit. This drop may be small most of the time, but there are a lot of breaker panels and breakers that are not as good as Sq-D QO units. However, there was a time when Sq-D made some panels with aluminum busbars and these could be classified as very poor. I am not sure if this was done in the QO series.

With just TED components on the circuit there is virtually no voltage drop from the breaker interface to the bus and the breaker even with relatively high interface resistance. So TED sees bus voltage.

If you want accurate power measurements, then incorrect voltage to the MTU should be eliminated. I do not have a good solution for the voltage drop from the meter and thru the main breaker to the busbars.

What I was suggesting as a standard approach should produce the least problems for the average user without all sorts of experimenting to solve problems.

There seem to be enough problems from the start that any procedures that can be employed that almost certainly eliminate questions should be employed.

Additionally a lot of electricians need work.