Randy Yates <yates@ieee.org> wrote: (snip)> They may have stated or intimated it, but I don't see why. I've done a > pencil and paper analysis of the scenario Rune proposed (re: the > insulation coming off one of the phases and a resulting connection to > ground) and I don't see a problem in that scenario with deriving the > last L-L from the first two L-Ls. (Whether current is flowing through > the ground-to-neutral impedance or not, the voltage of neutral w/r/t > ground falls out when you do A-B and B-C (or whatever).)Ones I can think of, which may or may not actually matter, are differences in the two being measures. If the filters were different, then the two wouldn't subtract exactly. Also, the gain of any amplifiers between the inputs and the ADC have to be adjusted. (Though more usual might be to calibrate the differences in gain and correct after the ADC.) As I understand it, it is usual for DVMs to do the calibration to the digitized value. Otherwise, it seems to me that it should work. As previously noted, the sample-and-hold should be done at the same time. If the S&H isn't done at the same time, then they should alternate such that on the average they are the same. -- glen
Line-to-Line Sensing: Analog Vs Digital?
Started by ●May 8, 2011
Reply by ●May 15, 20112011-05-15
Reply by ●May 15, 20112011-05-15
On 5/15/2011 10:32 AM, Randy Yates wrote:> > If it isn't necessary, then why do it?***I don't know if it's "necessary" or not. I have no assertions to make in that regard. ***I just said "highly recommended" because having information is generally better than not having it. ***Then I said: "Better is the enemy of good enough". Which might be translated into: 1) If it's easy, do it. 2) If it's easy and you're convinced really never needed then don't do it. 3) If it's hard AND if you're convinced not likely needed, then don't do it. 4) If it's hard and you don't know if it might be needed then you have a tradeoff decision to make. That's OK....>>The > time required to do so is too extreme,***If this is the case then #1 and #2 clearly do not apply. ***If this is the case AND you're happy that #3 applies entirely .. fine. ***If this is the case then you are likely dealing with #4. and there is too great of a risk> of throwing my time away since there seems to be a lot of confusion and > misunderstanding.***I'm not confused. I'm ignorant. Since I'm ignorant I vote for a bit more information if affordable. You're better informed on your system so you can make a better judgment.> > If someone is asserting that there is a problem, they ought to be able > to explain (_precisely_) why.***There can be unknown unknowns .... either you hedge against them or prove to yourself the opposite: - there is NOT a problem Or accept the risk: - the likelihood that the problem exists is low enough to risk it.> >> If you do measure N then you can always look at it. And, you can >> compare A-N to A. If the result is boring then that's a good sign >> that everyone "understands" what's going on, eh? But what if it >> isn't? > > Whoa!!! A-N to A? No, no no no!! We'd be comparing [(A-N) - (B-N)] > to (A-B)!.***Sorry. I was referring to "if you measure N" and A,B,C single endedly i.e. referenced to A/D earth ground or some other suitable reference. I now understand now that you're not likely prepared measure N. My comment was truly trivial: "If N (measured) is zero then A-N (calculated) is the same as A (measured)". What comparisons you make are entirely up to you. I agree with Tim that doing the work in software is the best way to go if it's possible - avoiding an excursion into analog things that are harder to fix/change/etc. You still have to get the A/D stuff working properly anyway. Getting back your original post, you said: ""My current client has a requirement to obtain the the line-line voltages of a three-phase power source inside a DSP given the line-neutral inputs. He's thinking of doing it in the analog domain using opamps to do the differencing, and feeding the resulting line-line into the ADC (we're using the 2809, BTW)."" ***I think the ONLY way to get the line-neutral inputs is with a differential amplifier or a transformer each. If that idea is set aside then you'd have single-ended measurements on each wire .. with or without N. ""Further, there are at least two opamp architectures under consideration. One is the standard difference amplifier between each of A-B, A-C, and B-C."" ***OK. ""The other uses just three opamps to determine A-B and B-C, "" ***I think you mean "just two" .. but OK. ""then relies on the fact that A-C can be derived from those two."" ***Differencing schemes can be iffy. i.e. (A-B) is a measurement so OK. (A-C) is a measurement so OK. (A-B) - (A-C) = -B + C or -(B-C) relying on a difference An error analysis of this would perhaps look like this: -[(A + e1 - B -e2) - (A + e3 - C - e4)] = [(e3-e1) + B + e2 - C - e4] So while you may be willing to accept e2 and e4 as things go, you would be introducing the added error [|e3| + |e1|] in the worst case. I'm not saying that this is a show-stopper because I don't have all the data, likely outcomes, etc. It's just an observation that comes about by effectively measuring "A" twice, once in each of two "channels". They may be "close enough" but will never be "the same". ""The other alternative is to just input the line-neutral voltages directly (after appropriate voltage dividers) into the ADC."" ***Where do you get the line-neutral voltages? Assuming that neutral is ground always and single-ended? OK if you believe that assumption holds. Maybe drop the term "-neutral" and just say "the line voltages"? This is about labels I think. The A/D ground reference should be just as good. If you don't care about N then it won't matter will it? But, if you do care about N then what? ""I'd be curious to learn what folks here think the "best" method would be given considerations of 1) noise, and 2) common-mode (the 3-phase center isn't necessarily the same as the ADC / opamp circuit ground)."" ***OK. I've not really touched on this yet. An awful lot goes into the analog circuit design for noise, CMR, etc. So, you might well use diffamps to deal with common-mode noise even when doing a "single ended" measurement. Depends on coupling methods, etc. Are you considering transformers perhaps? ***You acknowledge that the 3-phase center isn't necessarily the same as the ADC ground. Thus the notion of maybe measuring it too. When you say: We'd be comparing [(A-N) - (B-N)] to (A-B)!. The concerns go like this: What is "A-N"? If you don't measure N then it's a label for something that you did measure. If you measure "A" then I'd not label that measurement "A-N" just to avoid confusion - even if you believe it makes sense. Same with "B-N". Randy, I have a fair bit of analog design experience and I must say (as you can likely see) that I'm a bit confused for the details of the analog side. The devil is in the details. And, I'll bet that there are other folks here who are better at it than I. And, that's the focus of your question. I see two possibilities in a general sense: 1) Make single-ended measurements and deliver them to the A/D. You didn't really mention this approach originally. And, it may have merit. 2) Make differential measurements and deliver those single-ended to the A/D. These are the ones you mentioned. As you recognize, how you deal with each one introduces different issues with noise, CMR, etc. I honestly don't know which one is best for your situation but I'm more leery of differencing methods if they can be avoided. And that applies on the digital side as much as on the analog side. Maybe transformer coupling two lines together would be a reasonable approach and providing the output to a diffamp wouild be reasonable for X-Y in each case. A big part of this would also be: What are you going to do with the data? Fred
Reply by ●May 15, 20112011-05-15
On 05/15/2011 05:43 PM, Fred Marshall wrote:> On 5/15/2011 10:32 AM, Randy Yates wrote: > >> >> If it isn't necessary, then why do it? > > ***I don't know if it's "necessary" or not. I have no assertions to make in that regard.******Why are you using asterisks?!!!> [...]>>> If you do measure N then you can always look at it. And, you can >>> compare A-N to A. If the result is boring then that's a good sign >>> that everyone "understands" what's going on, eh? But what if it >>> isn't? >> >> Whoa!!! A-N to A? No, no no no!! We'd be comparing [(A-N) - (B-N)] >> to (A-B)!. > > ***Sorry. I was referring to "if you measure N" and A,B,C single endedly i.e. referenced to A/D earth ground or some other suitable > reference. I now understand now that you're not likely prepared measure N.N isn't available. The 3-phase systems we're dealing with do not include N. You can attempt to derive N, but then that's not really N. And if you rely on a derived N, there are other issues to deal with. So, no, we're not likely to have a derived N available at the ADC.> My comment was truly trivial: "If N (measured) is zero > then A-N (calculated) is the same as A (measured)".Well, yes. A - 0 = A. I'm well beyond knowing that circuit ground is not necessarily equal to neutral - it was stated up-front (re: common-mode).> ""My current client has a requirement to obtain the the line-line > voltages of a three-phase power source inside a DSP given the > line-neutral inputs. He's thinking of doing it in the analog domain > using opamps to do the differencing, and feeding the resulting > line-line into the ADC (we're using the 2809, BTW)."" > > ***I think the ONLY way to get the line-neutral inputs is with a differential amplifier or a transformer each. If that idea is set > aside then you'd have single-ended measurements on each wire .. with or without N.That's like saying the only way to perform an IIR filter is direct-form I.> ""Further, there are at least two opamp architectures under consideration. > > One is the standard difference amplifier between each of A-B, A-C, and B-C."" > > ***OK. > > ""The other uses just three opamps to determine A-B and B-C, "" > > ***I think you mean "just two" .. but OK.Well that *was* confusing. Thanks for pointing this out. I actually meant "three" but I shouldn't have used the word "just" since direct differential measurement would also require "just" three opamps. What I should have said is "The other uses just three sense strings and three opamps to determine A-B and B-C". The real issue here is sense resistors. Due to the high voltage (277 Vrms), we have to utilize several surface mount resistors in each sense string, and space the sense strings apart. If you perform a direct differential conversion of all three L-Ls, you'd need six such sense strings. I'll not go into the architecture here since it's hard to draw and it may be somewhat proprietary.> ""then relies on the > fact that A-C can be derived from those two."" > > ***Differencing schemes can be iffy. > i.e. (A-B) is a measurement so OK. > (A-C) is a measurement so OK. > (A-B) - (A-C) = -B + C or -(B-C) relying on a difference > An error analysis of this would perhaps look like this: > > -[(A + e1 - B -e2) - (A + e3 - C - e4)] = [(e3-e1) + B + e2 - C - e4] > > So while you may be willing to accept e2 and e4 as things go, you would be introducing the added error [|e3| + |e1|] in the worst case. > I'm not saying that this is a show-stopper because I don't have all the data, likely outcomes, etc. It's just an observation that > comes about by effectively measuring "A" twice, once in each of two "channels". They may be "close enough" but will never be "the > same".> ""The other alternative is to just input the line-neutral voltages directly > (after appropriate voltage dividers) into the ADC.""One problem with that is that the potential level of N-G (N being 3-phase neutral and G being circuit ground) can vary between -480Vrms and +480Vrms, so the necessity of adjusting gain degrades the digital SNR of our desired signal (A-B, e.g.) by almost 10 dB since that part of the signal isn't matching the full-scale range of the ADC.> ***Where do you get the line-neutral voltages?We don't. That's why we have to do a differential somewhere.> Assuming that neutral is ground always and single-ended?No!> OK if you believe that > assumption holds. Maybe drop the term "-neutral" and just say "the line voltages"? This is about labels I think. The A/D ground > reference should be just as good. If you don't care about N then it won't matter will it? But, if you do care about N then what?We care about differences, not N.> [...]> When you say: We'd be comparing [(A-N) - (B-N)] to (A-B)!. > > The concerns go like this: > > What is "A-N"? If you don't measure N then it's a label for something that you didI assume you meant "didn't" here.> measure. If you measure "A" then I'd not label that measurement > "A-N" just to avoid confusion - even if you believe it makes sense. > Same with "B-N".I agree there may be a semantic issue here. See below.> Randy, I have a fair bit of analog design experience and I must say (as you can likely see) that I'm a bit confused for the details > of the analog side. The devil is in the details. And, I'll bet that there are other folks here who are better at it than I. And, > that's the focus of your question. > > I see two possibilities in a general sense: > > 1) Make single-ended measurements and deliver them to the A/D. You didn't really mention this approach originally.The other alternative is to just input the line-neutral voltages directly (after appropriate voltage dividers) into the ADC. Semantics. When I said "line-neutral" what I really meant was a single-ended voltage measurement, albeit to circuit ground (since that's all we have) and not neutral. The single phase voltages A, B, and C are commonly referred to as "line-to-neutral" voltages. I was writing in an object descriptive manner and not an actual voltage measurement manner. I could have been more clear on that.> 2) Make differential measurements and deliver those single-ended to the A/D. These are the ones you mentioned. > > As you recognize, how you deal with each one introduces different issues with noise, CMR, etc. I honestly don't know which one is > best for your situation but I'm more leery of differencing methods if they can be avoided. And that applies on the digital side as > much as on the analog side.They can't be avoided - there is no N. This is common practice (apparently) in the power industry as it saves copper.> [...]> A big part of this would also be: What are you going to do with the data?Determine magnitude and phase via a PLL which uses a DQ transform as a phase discriminator. -- Randy Yates % "Watching all the days go by... Digital Signal Labs % Who are you and who am I?" mailto://yates@ieee.org % 'Mission (A World Record)', http://www.digitalsignallabs.com % *A New World Record*, ELO
Reply by ●May 16, 20112011-05-16
Randy, You are evidently dealing with a delta-connected three-phase distribution system. (Actually, that only needs two transformers. See "open delta".) It would be unusual (and probably violate code) if the center of tee connected resistors were more than a few volts off ground, assuming there were no line faults. That's why single-ended measurements are often sloppily called line-to-neutral. In the end, it doesn't really matter where on the phasor diagram the neutral lies. (See "two wattmeter method".) Neither need you be concerned with a small difference between large numbers on a fault-free system. The line-to-line voltages are larger than the line-to-neutral. What is important is the simultaneity of the measurements, and you seem to have that covered. The relations you want are mathematically related. The fewer independent measurements you need to make, the more reliable the result will be. Jerry -- Engineering is the art of making what you want from things you can get.
Reply by ●May 16, 20112011-05-16
On 05/15/2011 11:50 PM, Jerry Avins wrote:> Randy, > > You are evidently dealing with a delta-connected three-phase > distribution system. (Actually, that only needs two transformers. See > "open delta".) It would be unusual (and probably violate code) if the > center of tee connected resistors were more than a few volts off > ground, assuming there were no line faults. That's why single-ended > measurements are often sloppily called line-to-neutral. In the end, it > doesn't really matter where on the phasor diagram the neutral lies. > (See "two wattmeter method".) > > Neither need you be concerned with a small difference between large > numbers on a fault-free system. The line-to-line voltages are larger > than the line-to-neutral.Jerry, I am primarly concerned with two things: 1. how choice of architecture affects common-mode (G-N) input range, and 2. how choice of architecture affects noise and G-N cancellation> What is important is the simultaneity of the measurements, and you > seem to have that covered.I may care about it, but it's not what I'm asking about here.> The relations you want are mathematically related. The fewer > independent measurements you need to make, the more reliable the > result will be.That is a refreshingly simple and powerful observation - thank you Jerry! -- Randy Yates % "Watching all the days go by... Digital Signal Labs % Who are you and who am I?" mailto://yates@ieee.org % 'Mission (A World Record)', http://www.digitalsignallabs.com % *A New World Record*, ELO
Reply by ●May 16, 20112011-05-16
On 05/15/2011 11:50 PM, Jerry Avins wrote:> [...] > It would be unusual (and probably violate code) if the > center of tee connected resistors were more than a few volts off > ground, assuming there were no line faults. That's why single-ended > measurements are often sloppily called line-to-neutral.Apparently there is something wrong in your assumptions here. One case that introduces signficant voltage between G and N is when the system is corner-grounded. We'd like to be able to operate in this configuration. -- Randy Yates % "Watching all the days go by... Digital Signal Labs % Who are you and who am I?" mailto://yates@ieee.org % 'Mission (A World Record)', http://www.digitalsignallabs.com % *A New World Record*, ELO
Reply by ●May 16, 20112011-05-16
On 5/15/2011 7:06 PM, Randy Yates wrote: Asterisks? Oh! I do that when I'm interspersing comments as a delimiter. Maybe there's a better way.... Fred
Reply by ●May 16, 20112011-05-16
On 5/15/2011 7:06 PM, Randy Yates wrote: Ah! No neutral and "line-neutral" or "line-to-neutral" is jargon meaning what I'd call "line" voltage I guess. OK. Well then, some of what I said doesn't matter. ...... > I see two possibilities in a general sense: > > 1) Make single-ended measurements and deliver them to the A/D. You didn't really mention this approach originally. Now I understand that this is exactly what you have in mind for A,B and C and there is no N. That seems a good approach. Then do all the differencing in the digital side. 2) Make differential measurements between A and B, B and C and, for redundancy or a check maybe A to C. But I don't disagree with Jerry re keeping it simple. I think this could be done with a transformer for each pair. Float the input winding with A on one end and B and the other. The output winding would represent A-B and go single-ended to the A/D or balanced to a diffamp first for CMR. But I don't know the practical issues with doing this. In principle it would accomplish the same thing as a diffamp. But in practice a lot depends on implementation.... Fred
Reply by ●May 16, 20112011-05-16
On May 16, 12:11�am, Randy Yates <ya...@ieee.org> wrote: ...> Jerry, > > I am primarly concerned with two things: > > � �1. how choice of architecture affects common-mode (G-N) input range, > � �and > > � �2. how choice of architecture affects noise and G-N cancellation... To address common mode, use instrumentation amps instead of op amps. Think of them as DC transformers. Jerry -- Engineering is the art of making what you want from things you can get.
Reply by ●May 16, 20112011-05-16
On May 16, 12:14�am, Randy Yates <ya...@ieee.org> wrote:> On 05/15/2011 11:50 PM, Jerry Avins wrote: > > > [...] > > It would be unusual (and probably violate code) if the > > center of tee connected resistors were more than a few volts off > > ground, assuming there were no line faults. That's why single-ended > > measurements are often sloppily called line-to-neutral. > > Apparently there is something wrong in your assumptions here. One case > that introduces signficant voltage between G and N is when the system > is corner-grounded. We'd like to be able to operate in this > configuration.Apparently there is: my apologies. It may not violate code in your case -- I never understood why it should, but I can guess -- but it is surely unusual. Jerry -- Engineering is the art of making what you want from things you can get.
