This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Which energy meter method more reasonable?

Other Parts Discussed in Thread: MSP430FE4272, MSP430F423, MSP430F1481, MSP430FE423


1 phase. voltage, current and Power. 

-The voltage range is 220v AC. 

-The current range is 0 to 20A.  

1. Use MSP430FE42x MCU with shunt resistor.

2.Analog Devices ADE7569 MCU with shunt resistor.

3.Analog Devices ADE7763 IC + simple MCU with shunt resistor.


Thanks Help...

  • Well, I can't say. With the FE42x you'll have the lowest hardware but have to write the complete software. The ADE devices already implement much of what you need, so you can get the ready calculated power and energy reading as well as effective voltage and current by just reading the ADE chips resgisters.

    We use the ADE7758 and (currently) an MSP1232. You can see that the MSP side is quite small (the 1232 has 8K Flash and 256 bytes ram).

    In any case, usage of a shunt resistor is not a good idea. When using a shunt, your processor is floating on AC level. Since you also want/need to measure AC voltage, you'll end up with 320V on the MSP. If you design your hardware properly, this is possible, but if you cannot ensure under all circumstances that your shunt is always inserted between phase and device or (with other hardware design) device and null/ground, I predict your hardware a very short lifespan. And if the device you measure experiences a shortcut, you'll have +-320Vpeak on your shunt, causing the shunt to catch fire and the rest to be fried by the input voltage.

    To get a decent voltage reading from the shunt, you'll need between +-0.5 and +-1.25V for 20Aeff on the shunt. This would be 17.677669mOhm for 0.5V. A 20mOhm shunt would do. BUT, at 20A you'll get a power of 8W of heat on the shunt. That's a bit much power consumption for an energy meter. And that's for just a +-0.5V voltage swing on the A/D inputs. For greater accuracy, you'd need more and the wasted energy would increase (V²). And (as mentioned above) a shortcut of 220V across 20mOhm would lead to whopping 2.5MW(!!!) momentary power building up at the shunt. Of course this value is purely theoretical, the resistance of the wires and such will reduce it to a mere fraction of this value, and eventually the AC fuse will blow, but the remaining power will surely be enough to still fry your equipment instantly and before any fuse can react.

    Almost all commercial energy meters, even those for several thousands of amperes, do have just 5A or 1A shunts and require a current transformer to transform x-to-5A or x-to-1A. This will not only reduce the power building up at the shunt (0.5W or 0.02W in the above calculation), it also decouples the shunt from the AC voltage (so it doesn't matter where in the power line you place it) and due to the limited energy transmission capability of the current transformer, it also protects against voltage and energy peaks on the shunt. (the magnetic core won't transfer more than its maximum energy, even on massive overcurrent on the primary side)

     But keep in mind that a current transformer tries to keep its output current, so if not attached to a shunt, its output voltage might rise into unexpected heights.


  • Thank you very much.

    i understood. i have to use Current Transformer. 

     ADE7763 one phase energy meter IC is calculating all values that is Irms,Vrms and watt. i dont need to write complex software. 

    and what about  ADE7569? it has 8052 MPU. so it has uart and display driver.

     is it need complex software or easy?

  • The ADE7758 we are using is a three-phase version, but I guess it is similar to the 7763. It is fairly simple to read the momentary values as well as the accumulated ones. The main difficulty is the calibration process. In our first versions of the firmware, we just read the values 'as is' and compared them with a reference, then recompiled the firmware to use a compensation factor. When moving the hardware side from an MSP1232 to the new 54xx and its much bigger flash and ram (together with other module series which are using the 1611 right now), we will include a far more complex self-calibration mechanism, which will allow for faster calibration of the devices and remove the need of reflashing the devices with an individual firmware with the proper calibration factors. I hope it will significantly improve the device accuracy over the full range of operation (the current, but very simple method calibrates only for a certain load point, which was sufficient in most cases).

    I don't have any experience with the 7569. My guess is that the 7569 is some sort of hybrid chip with almost the same hardware for the metering process, but the registers with the results available directly to the 8052 core. But that's just a guess, as I haven't looked at the 7569 datasheet. It is also possible that the other ADEs do have a 8052 core too, which is doing the calculations and providing the results through SPI. If so, you might get sample code or even a complete measuring sequence from AD for the 7549. Or you might have to do it all by yourself like you have on this specialized MSP (forgot the name).

    AllI can suggest is reading the datasheets carefully, and If you have time and money, order a demo board an play with it a little bit. And then decide.

  • 20A is rather low for a utility meter application, so I guess this application is something like power consumption monitoring inside a power supply. If it is, then solutions like the ADI chips probably won't suit your needs. Most of these monitoring applications call for very specific measurements, like "a power measurement result every 100ms or less" which the fixed function utility metering solutions, such as the ADE7xxx devices, can't do. The MSP430F423, on the other hand, is flexible, and can do an excellent job in these applications. TI has reference software for server power supply applications, which can get you started.

  • Hi jens-Michael Gross,

                                              we have started a project  "Three phase Energy metering" using ADE7758 with msp430f1481 and we are facing too many problems in calibration,i.e. as per our design it should calculate the Energies up to 60A ,but while  moving loads (above 10A) the accumulated energies were not accurate so that again we have to chage the multiplication factors,so that it is very difficult for us to give different MFs for different loads.

                                           If u don't mind please share the procedures what you have followed to got the accurate calibration.

    Thanking you.

    Kind Regards,


  • Hi Steve.

    (I guess this application is something like power consumption monitoring inside a power supply.)

    Your guess is true. i want to make a device like this. 

    My design will be this:


    i need a development board for MSP430FE423.




  • Well, 20A is normal for any home usage. Those energy meters for measuring the consumption of a TV or a computer or whatever. Typical update time is 1 second or less. So reason you couldn't do this with an ADE. Also, I had no problems updating the reading in out ADE based project (we use an MSP to read the ADE, so writing that still does not cut TIs profit :) ) every 100 ms and even less. If you don't need all readings all the time, this is really no problem. And with a hardware SPI module for the communication and a 16MHz 5438 instead of a 4MHz clocked 1232 with software SPI, it is even less a problem.

    One argument might be the price. Depending on the number of devices expected the additional software effort might pay. So the 423 might be better suited for this project and the ADE for that.

    But there's definitely no "can't do" for the ADEs. (note that I'm not talking about the ones with programmable 8032 core - I have no experience with them and would prefer an MSP - but the monolithic solutions which deliver the results via SPI).

  • Getting many readings per second from the ADI or similar fixed function chips is no problem, even for a slow MCU. The problem is the interval over which those readings are calculated. Currently, the server power supply makers are working towards the next generation Intel spec for servers, and this calls for the actual measurement interval to be 100ms. They want to see the true burstiness of the load. While the MSP430 is not the only device with the flexibility to serve this requirement, I don't know of anyone considering the ADI or similar fixed function metrology devices. They just can't meet the spec.

  • Hi Kshatriya,

    While I'd love to see you using a TI MSP430F471x6 device, instead of and MSP430 + and ADI metrology chip, I think your current problem is probably with your board and not with the ADI device. These kinds of accuracy issue are usually related to layout and grounding.

    You are probably getting some crosstalk between the voltage and current signals. Remember, the voltage signal is always large, but the current signal varies over a wide range. Careful layout is required to ensure that big voltage signal doesn't leak into the current inputs. If you don't ensure that your layout maintains the full advantage of the differential inputs to the ADCs in the metrology device, you can severely compromise accuracy.

    Another factor which often surprises people is the extent to which surface contamination on the PCB can affect results. Even if you keep the mains voltage wiring, and the potential divider resistors for the voltage sensor input to the metrology chip, well away from the sensitive circuitry, surface contamination can cause a surprising amount of crosstalk. Slotting the board in appropriate places can often do wonderful things for this.

  • Sorry, Steve, I cannot confirm this. The ADE7758 we're using can be read with realiable results down to one wave (which is 20ms on 50Hz). You can also set it up to issue an interrupt request after a given number of zero transitions (e.g. 5 for 100ms interval) and then read the buffered peak values in this period. Anything below one wave (1/2 would be possible) is of course unreliable. If you need to sample the waveforms (other than peak value), you definitely should go for a freely programmable system.

    I don't want to say that the (relatively new) MSP isn't suited for its job. On the contrary, a freely programmable devices is always more flexible than any devices with fixed functionality -  if you require this flexibility. But I haven't found anything I couldn't do with the ADE7758 yet in connection with an MSP that reads and processes the readings. I know that many of the other ADEs have much less functionality (e.g. 3 phase but only sum energy reading etc.) and your statements are true for them.

    As for the problems Kshatriya has, you're right, layout, crosstalk and surface contamination might be a problem. Also parasitic capacitors built by a ground plane on the downside of the PCB. The layout should be single-sided. Also, the calibration process should be done on a significant level of load. The mentioned problems by crosstalk usually appear when there are large (voltage) and small (low current) signals are mixed. At the ADE inputs, there is only a voltage swing of +-0.5Vpp for voltage(line) and current (load). Crosstalks shouldn't be a problemhere if the signals do not cross or are routed too close. But the divider resistor from 220V to 0.5V should be placed far away. And of course no SMD resistor at all but a through-hole type.

    About calibration process itself, I already wrote a private mail. For us, a simple one-point calibration somewhere in the middle of the working range did give a sufficient precision (and more wasn't really possible on the original PIC and the later MSP1232 processor).  In future, we'll go for a higher precision and a more complex calibration with the 5438 processor.

  • The ADE7758 is a 3-phase device, so it is not an ideal choice for use in single phase appliances monitoring their own consumption. Also, it would require a support MCU to achieve the required functionality for server power applications. It would be a very costly solution. The ADE7758 is fairly new, and offers a lot more functionality than the other ADI metrology chips I am aware of. I haven't seen a single phase device from ADI that is comparable.

    There is nothing wrong with using SMD resistors for the potential divider in an power/energy meter. The vast majority of utility energy meters made today use SMD resistors in this area. However, you need to allow for at least 4kV pulses hitting that divider during fast pulse transient testing, and do what it takes to prevent arcing. Divide up the total resistance, and use several large package SMD resistors in a series chain. Keep them well separated from each other and the surrounding parts. Do not use oversize PCB pads, as this will reduce the copper to copper separation across the board. Perhaps most importantly, do not try to run any other signal traces under the resistors. That last point may seem obvious, but I've seen it happen.

  • If you really only need one phase, then it is surely overkill. We use it in three phase setups as well as in triple single-phase measurement (same hardware and same firmware, just different external wiring). Indeed it requires an additional MCU (we used a very cheap PIC first and later an MSP1232). Whether it is costly, depends on the usage. If you plan to produce thousands of devices, then of course every cent counts. In our situation, we produce a lot less devices, so every hour put into developing the algorithms (and testing them and then certifying them) will add more to the per unit cost than the additional costs on the hardware side. Also, it is easier to isolate the ADE (only the few SPI lines) than isolating the RS232/RS438/S0/RF circuitry from the processor. We actually thought of using two MSPs, one for the direct measuring, another one, isolated, for the low voltage side communication. But we'll stay with the ADE as it is a known buddy.

    Using SMD resistors is surely possible if you break down the large 'head' resistance of the divider into several units. The German VDE however requires a minimum gap between phase and ground that must not  be broken down into several smaller gaps, even if possible in theory (this is because the given values include additional space for different dirt/contamination classes etc. and it shall prevent chain-reactions along the isolation path). So at least one of the SMD resistors would need to be large enough fulfil this. We opted to use a through-hole resistor instead. Since we already have some other non-SMD parts (and even the SMD electrolytic capacitors) which need to be soldered manually, it does not add to the cost and also helps a bit relaxing the layout :)

    Anyway, my comment was just to remind that typical SMD resistors (we use 0805 normally) cannot bear 220Veff (actually up to 400Vp between two phases), even if their power rating would be high enough. (just 50mW @ 220V on 1MOhm). Not to mention the 4kV testing pulses. To quote you: It might seem obvious, but I've seen it happen. :)


    Low cost ADE7763 one phase energy meter IC is calculating the values  Irms,Vrms and Watt. it needs external MCU to comm.  Around 1ms time is enough to read any register value by SPI. i need to read three value Irms,Vrms and Watt. i can read these all values in 10 ms. This is enough to calculate kWatt/h. 

  • While you can read values rapidly from an ADE7763, the low pass filter architecture used to produce steady values for the voltage, current and power measurements means the readings do not change rapidly, or represent the readings over a well defined interval. This is fine for some applications. Other applications are looking to track pulses of energy demand more crisply. A fully programmable solution, like the MSP430F42x, has the flexibility to meet a much wider range of requirements.