Hi , We are in the business of Electrical Power and Energy meters. One of our products has a feature to grab samples of voltage and current waveforms , and apply FFT to those samples , thereby generating data for us like the % content of odd harmonics in each waveform . As the electrical power gets 'dirtier' due to usage of equipment like AC Power Drives and other non-linear loads , this feature in our product accurately tells the user as to what is the level of third harmonic and fifth and so on , wrt to the fundamental . One new requirement however has me baffled . The user wanted to know , in addition of the magnitude of the 3rd ( or 5th or any ) harmonic as normalized against the fundamental , ALSO the direction of the third harmonic . In other words , was the 3rd harmonic already present in the mains voltage driving the load , or it was the load that generated it ? At the end of the FFT , run over 128 samples ( two arrays of 128 floats- real and img. - real array filled initially with waveform sample values and img. array filled with zeroes ) , we have usable information in first 64 real and img. How from this result can I know the direction of harmonic current flow ? Or is there some other method which I need to use . Please help .
Detecting the Direction of Harmonic Generation.
Started by ●February 9, 2011
Reply by ●February 9, 20112011-02-09
On Feb 9, 10:45�am, "paragtrinity" <parag.joshi@n_o_s_p_a_m.trinityenergy.co.in> wrote:> Hi , > We are in the business of Electrical Power and Energy meters. One of our > products has a feature to grab samples of voltage and current waveforms , > and apply FFT to those samples , thereby generating data for us like the % > content of odd harmonics in each waveform . As the electrical power gets > 'dirtier' due to usage of equipment like AC Power Drives and other > non-linear loads , this feature in our product accurately tells the user as > to what is the level of third harmonic and fifth and so on , wrt to the > fundamental . > One new requirement however has me baffled . The user wanted to know , in > addition of the magnitude of the 3rd ( or 5th or any ) harmonic as > normalized against the fundamental , ALSO the direction of the third > harmonic . In other words , was the 3rd harmonic already present in the > mains voltage driving the load , or it was the load that generated it ? > > At the end of the FFT , run over 128 samples ( two arrays of 128 floats- > real and img. - real array filled initially with waveform sample values and > img. array filled with zeroes ) , we have usable information in first 64 > real and img. �How from this result can I know the direction of harmonic > current flow ? Or is there some other method which I need to use . > > Please help .not sure...but I think... you need to capture both the waveform of the CURRENT flowing past your measurment location and the VOLTAGE at the location. By examination of the phase relationship between the current and voltage you can deduce the direction of power flow (and you can do this for each frequency component) .....i think... Mark
Reply by ●February 9, 20112011-02-09
On Feb 9, 10:45�am, "paragtrinity" <parag.joshi@n_o_s_p_a_m.trinityenergy.co.in> wrote:> Hi , > We are in the business of Electrical Power and Energy meters. One of our > products has a feature to grab samples of voltage and current waveforms , > and apply FFT to those samples , thereby generating data for us like the % > content of odd harmonics in each waveform . As the electrical power gets > 'dirtier' due to usage of equipment like AC Power Drives and other > non-linear loads , this feature in our product accurately tells the user as > to what is the level of third harmonic and fifth and so on , wrt to the > fundamental . > One new requirement however has me baffled . The user wanted to know , in > addition of the magnitude of the 3rd ( or 5th or any ) harmonic as > normalized against the fundamental , ALSO the direction of the third > harmonic . In other words , was the 3rd harmonic already present in the > mains voltage driving the load , or it was the load that generated it ? > > At the end of the FFT , run over 128 samples ( two arrays of 128 floats- > real and img. - real array filled initially with waveform sample values and > img. array filled with zeroes ) , we have usable information in first 64 > real and img. �How from this result can I know the direction of harmonic > current flow ? Or is there some other method which I need to use . > > Please help .not sure...but I think... you need to capture both the waveform of the CURRENT flowing past your measurment location and the VOLTAGE at the location. By examination of the phase relationship between the current and voltage you can deduce the direction of power flow (and you can do this for each frequency component) .....i think... Mark
Reply by ●February 9, 20112011-02-09
On 02/09/2011 09:28 AM, Mark wrote:> On Feb 9, 10:45 am, "paragtrinity" > <parag.joshi@n_o_s_p_a_m.trinityenergy.co.in> wrote: >> Hi , >> We are in the business of Electrical Power and Energy meters. One of our >> products has a feature to grab samples of voltage and current waveforms , >> and apply FFT to those samples , thereby generating data for us like the % >> content of odd harmonics in each waveform . As the electrical power gets >> 'dirtier' due to usage of equipment like AC Power Drives and other >> non-linear loads , this feature in our product accurately tells the user as >> to what is the level of third harmonic and fifth and so on , wrt to the >> fundamental . >> One new requirement however has me baffled . The user wanted to know , in >> addition of the magnitude of the 3rd ( or 5th or any ) harmonic as >> normalized against the fundamental , ALSO the direction of the third >> harmonic . In other words , was the 3rd harmonic already present in the >> mains voltage driving the load , or it was the load that generated it ? >> >> At the end of the FFT , run over 128 samples ( two arrays of 128 floats- >> real and img. - real array filled initially with waveform sample values and >> img. array filled with zeroes ) , we have usable information in first 64 >> real and img. How from this result can I know the direction of harmonic >> current flow ? Or is there some other method which I need to use . >> >> Please help . > > not sure...but I think... > > you need to capture both the waveform of the CURRENT flowing past your > measurement location and the VOLTAGE at the location. By examination > of the phase relationship between the current and voltage you can > deduce the direction of power flow (and you can do this for each > frequency component) .....i think...Theory at least would seem to indicate that'll work. You'd want to double check with some circuit analysis, but my "wetware SPICE" simulation (I'm thinking about it _really hard_) seems to bear that out. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" was written for you. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●February 9, 20112011-02-09
Mark <makolber@yahoo.com> wrote: (snip)>> One new requirement however has me baffled . The user wanted to know , in >> addition of the magnitude of the 3rd ( or 5th or any ) harmonic as >> normalized against the fundamental , ALSO the direction of the third >> harmonic . In other words , was the 3rd harmonic already present in the >> mains voltage driving the load , or it was the load that generated it ?(snip)> not sure...but I think...> you need to capture both the waveform of the CURRENT flowing past your > measurment location and the VOLTAGE at the location. By examination > of the phase relationship between the current and voltage you can > deduce the direction of power flow (and you can do this for each > frequency component) .....i think...That was my first thought, even before reading later posts. Note, for example, that watthour meters can detect the direction of power usage by the relation between current and voltage. As with the others, I didn't go through the whole derivation, but it seems right. Just measuring current or voltage won't tell you the direction.>> At the end of the FFT , run over 128 samples ( two arrays of 128 floats- >> real and img. - real array filled initially with waveform sample values and >> img. array filled with zeroes ) , we have usable information in first 64 >> real and img. �How from this result can I know the direction of harmonic >> current flow ? Or is there some other method which I need to use .To save some work, you might be able to do it with the voltage in the real part, and current in the imaginary part, and then process the result appropriately. Again, I didn't go through the derivation. Either that, or you can use the well known method of converting 128 real values to 64 complex values for the FFT, and then converting the result back as appropriate. Just to be sure, this should be measured over a whole number of cycles. One way is to phase lock the clock to a multiple of the power line. -- glen
Reply by ●February 10, 20112011-02-10
Commercial power-line meters (those that can trip circuit breakers are called relays) measure some subset of line voltage and current, real and reactive power, positive- negative- and zero-sequence current and voltage (all of those RMS), a few seconds of real-time samples of individual line voltage and current for fault analysis and more I don't recall right now. Relays can trip on any of those and also on undervoltage. I never heard of any that included the mains equivalent of an RF directional coupler. Jerry
Reply by ●February 10, 20112011-02-10
On Feb 10, 3:11�am, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:> Just to be sure, this should be measured over a whole number of cycles.Why?> One way is to phase lock the clock to a multiple of the power line.How? Nah, the OP is right in that he needs to measure both voltage and current. I would use the cross spectrum between these measurements and investigate the phase in the vicinity of the peaks. I would, however, do some preliminary checks to see that the bin widths of the proposed DFT are sufficiently narrow - that the peaks of the harmonic spectrum are about the same size as, or larger than, the DFT bin widths. If the bin widths are 'significantly' (yeah...) larger than the peaks, ant method based on the DFT will strubble to come up with usable results. Rune
Reply by ●February 10, 20112011-02-10
On Feb 9, 4:45�pm, "paragtrinity" <parag.joshi@n_o_s_p_a_m.trinityenergy.co.in> wrote:> Hi , > We are in the business of Electrical Power and Energy meters. One of our > products has a feature to grab samples of voltage and current waveforms , > and apply FFT to those samples , thereby generating data for us like the % > content of odd harmonics in each waveform . As the electrical power gets > 'dirtier' due to usage of equipment like AC Power Drives and other > non-linear loads , this feature in our product accurately tells the user as > to what is the level of third harmonic and fifth and so on , wrt to the > fundamental . > One new requirement however has me baffled . The user wanted to know , in > addition of the magnitude of the 3rd ( or 5th or any ) harmonic as > normalized against the fundamental , ALSO the direction of the third > harmonic . In other words , was the 3rd harmonic already present in the > mains voltage driving the load , or it was the load that generated it ? > > At the end of the FFT , run over 128 samples ( two arrays of 128 floats- > real and img. - real array filled initially with waveform sample values and > img. array filled with zeroes ) , we have usable information in first 64 > real and img. �How from this result can I know the direction of harmonic > current flow ? Or is there some other method which I need to use . > > Please help .As others have already mentioned you'll need to capture both voltage and current. As in real life there is no black and white case (e.g. third harmonics is present only and only due to grid voltage) you should know (measure) at least grid output impedance while load input impedance makes things easier. You need to know this for each harmonic you are interested (I know that power line analyzers usually measure only impedance at fundamental (50/60 Hz). From there you can deduce how no-load voltage would look (if you can disconnect the load once in a while - every 15 minutes - you should not need to know the line impedance) thus all extra harmonics come from load. Mitja Nemec
Reply by ●February 10, 20112011-02-10
On 10.2.11 4:11 , glen herrmannsfeldt wrote:> Mark<makolber@yahoo.com> wrote: > (snip) > >>> One new requirement however has me baffled . The user wanted to know , in >>> addition of the magnitude of the 3rd ( or 5th or any ) harmonic as >>> normalized against the fundamental , ALSO the direction of the third >>> harmonic . In other words , was the 3rd harmonic already present in the >>> mains voltage driving the load , or it was the load that generated it ? > > (snip) > >> not sure...but I think... > >> you need to capture both the waveform of the CURRENT flowing past your >> measurment location and the VOLTAGE at the location. By examination >> of the phase relationship between the current and voltage you can >> deduce the direction of power flow (and you can do this for each >> frequency component) .....i think... > > That was my first thought, even before reading later posts. > Note, for example, that watthour meters can detect the direction > of power usage by the relation between current and voltage. > > As with the others, I didn't go through the whole derivation, > but it seems right. Just measuring current or voltage won't > tell you the direction. > >>> At the end of the FFT , run over 128 samples ( two arrays of 128 floats- >>> real and img. - real array filled initially with waveform sample values and >>> img. array filled with zeroes ) , we have usable information in first 64 >>> real and img. How from this result can I know the direction of harmonic >>> current flow ? Or is there some other method which I need to use . > > To save some work, you might be able to do it with the voltage in > the real part, and current in the imaginary part, and then process > the result appropriately. Again, I didn't go through the derivation. > > Either that, or you can use the well known method of converting 128 > real values to 64 complex values for the FFT, and then converting > the result back as appropriate. > > Just to be sure, this should be measured over a whole number of cycles. > One way is to phase lock the clock to a multiple of the power line. > > -- glenAnother possibility is to apply a suitable window function to the sample set before the FFT. For details, see a text on DSP, e.g. the excellent book by Rick Lyons. -- Tauno Voipio tauno voipio (at) iki fi
Reply by ●February 10, 20112011-02-10
Jerry Avins <jya@ieee.org> wrote: (snip)> I never heard of any that included the mains equivalent > of an RF directional coupler.Yes, that is what I was thinking about. Now, why? If the source impedance is too high, and other loads have high harmonic currents, that could be seen by other loads. Also, there used to be saturable core voltage regulating transformers with high harmonic content in the output. More recently, UPS with inverters that don't generate a sine wave, though some try to approximate one. -- glen
