EEVblog 1379 - What's all this NPLC Stuff Anyhow?

hi this is your humble multimeter and you're used to it just on dc volts here reading zero volts when you're just got your probe sitting there on the bench it's reading nothing and you might be uh familiar of course if you switch it over to millivolts you know it might pick up a little bit of noise there and if you put your hands on there you know you might get a few like tens or even hundreds of millivolts noise but generally it's around about you know zero volts something like that especially if you put it on the voltage range like this well what happens if you take the same leads and you plug them into like a high end six and a half digit or seven and a half digit multimeter like this well uh beulah bueller that's um one and a half volts what's going on there something's a bit weird why are we getting like a volt and a half well let's go over to the keysight meter up here and let's plug in the exact same leads oh we're still getting a vault i can hear you saying dave i know what's going on here bench meters are famous for having like a high input impedance on like the millivolt range and even up into uh several over one or several of the voltage ranges well if we go into there you'll see that no we're 10 meg ohms input impedance and if we go auto input z yeah it goes a bit higher and stuff like that and it might charge up because you've effectively got uh infinite input impedance but um and if we like manually range it like this look on the one volt range we're getting overload overload overload wha what's going on and let's try an older school uh bench meter in this case a um old phillips uh six and a half digit jobby and if you listen very carefully you might be able to hear something you can hear a relay in there range switching because it's just going crazy it doesn't know whether it's volts millivolts or whatever you can see the m uh flash up there very briefly it's just going berserk and this is actually uh very real stuff look if we go into the uh trend chart over here we can see look at that i mean that's we can auto scale that look at that i mean there's real stuff on there at like almost two volts plus minus two volts p to peak it's enormous why as i said we've got 10 meg ohms input impedance exactly the same as your regular multimeter what's going on why does this show zero and these higher end meters show like a couple of volts and we can even choose to like like slowly data log this as well look at this i'm doing like one sample per second one it's like minus one volt um it's all over the place look but you might see it is actually counting down that's interesting and if we do a trend chart of uh this slow ones once per second data logging hmm there you go i left it for a bit and this is what we're getting it's kinda sorta sinusoidal not really but something's going on look at that oh there's a bit of bitter wiggle wiggle wiggle yeah going on down the bottom there and it'll probably go back up shortly you watch come on come on i'm betting it will you betcha here you go go you little beauty up it goes all the way you can do it well add a little jaggy there you can see this is like real interesting stuff this is real logged data once again with the probe sitting in exactly the same position we had for the other meters it's interesting it's picking up something but of course everyone knows what it's picking up it's just picking up mains and crap right and sure enough if we short the probes it is zero and we can go back to our number display there and zero volts and we take our hand off that and there we go it's going up point three point five point seven once again this is ten mega ohms input impedance exactly the same as our 10 mega ohm input impedance multimeter here so why the difference well i'm glad you asked it has to do with the number of power line cycles or the integration time of the multimeter your typical handheld multimeters are like these these are relatively uh slow you used to you know like a really fast one will get five or even some on the market might be like seven times a second or something like that they're really quite slow but these actually have built in 50 60 hertz sometimes it's selectable sometimes it's not um filters in them so they're actually filtering and out the power line frequencies because in any sort of lab or environment where you're measuring stuff 50 hertz is going to be or 60 hertz for you yanks um is going to be like one of the predominant interference sources in a typical environment office or lab environment so your handheld multimeters are being very nice to you and they actually have that uh integration time set so that it takes samples long enough that it actually effectively filters out 50 hertz or 60 hertz uh interference frequencies but your higher end multimeters like this they may or may not do it by default you've actually got to go into the menu and check out the number of powerline cycles so if we go into dc volts here you'll see i'm at 0.2 plc or powerline cycles right so i'll just stop my data log in here and we'll go back to a continuous display right and there we go we're in volts and number of powerline cycles this actually determines the accuracy of your meter as well so nplc is the acronym for it and you can also do time as well um in milliseconds they're effectively like essentially the same thing the number of powerline cycles means it'll do an integration measurement over one 50 or 60 hertz power line cycle so if mplc is set to one and then you can do that in milliseconds as well i mean you know 50 hertz would be 20 milliseconds of course so you can see if i've actually got that value very low i don't get many uh significant digits there and i also get quite a lot of noise here and if i go to 0.02 power line cycles 0.06 we're still getting um you know like volts of noise right and point two we're still getting quite a lot of noise but watch what happens when i go to one powerline cycle ta-da it's magically vanished because it's doing at least one full uh integration of the 50 or 60 hertz power line cycle so you're reducing the noise and you can see of course that we've got more significant digits now so if we go back of course we still have the same number of significant digits there it hasn't changed but because the integration time is not long enough to do any effectively like averaging so to speak even though it's integration i won't go into the differences but but anyway if we go to there and then if we go to 10 powerline cycles watch ta-da we get an extra digit of resolution here and of course we're getting our zero volts there once again if i touch those leads right yeah i can get you know tens of millivolts basically um equivalent to what we get on our um the handheld multimeter here and i can show you how the uh smoothing or you know every mathematical averaging doesn't do the same thing it's actually to do with the measurement integration not the post measurement um smoothing or something like that so let's go down to say point to uh powerline cycles here and then we'll go into math up here and where are we we've got smoothing filter there we go if we turn the smoothing filter on ah it doesn't really do anything so it's doing and the response also um you know 10 readings 50 readings of smoothing it doesn't help so doing post sample uh averaging and smoothing does not help the situation it's all to do with the how the adc works and these are integrating adc's you might have heard of dual slope integration i've probably done a video on dual slope or multi-slope integration the keysight have their multi-slope integration and there's dual slope and there's single slope and all sorts of things but that's basically how your high-end multimeters well even your handheld multimeters as well like even your low end ones they use like dual slope integration so it's you know if you don't have your uh integration time of your measurement uh set to actually uh take into account and average out in the measurement the 50 or 60 hertz noise pickup then yeah you're going to come a gutsy like this and you're going to measure volts and you can get the meter to do weird auto ranging stuff we saw on that phillips one and the keithley one down here exactly the same uh thing like i've got the smooth that smoothing field is actually on right the smoothing filter doesn't do anything it doesn't help your cause at all and check it out it's just going me auto range in there oh look it's even got like 10 volts one volt right it just doesn't know what to do it's just absolutely nuts and you turn on the smoothing filter and well it's still it's a little bit slower of course but the those high voltages are still there it's not getting rid of them and once again we're still 10 meg ohms input impedance but you'll see that we're 0.1 power line cycles so i'll turn off the filter here and we'll change that to one power line cycle bingo it's gone away because we're doing at least what an integration over one full 50 or 60 hertz power line cycle nice and as i said uh meters will typically have like a setup in there for 50 or 60 hertz and just to show you the actual waveform that we are picking up here what i've got is i've replaced the multimeter leads with just uh some banana plug leads flapping around in the breeze there and i've got a uh 10 to 1 probe directly coax connected across there so we've effectively got a 5 mm input impedance now total but we're going to be you know that's still quite high enough to pick up the noise and stuff so if we go in here and we have a look at our trend chart you can see that we're getting like plus minus a volt there does that correlate with the oscilloscope yep it does check it out there you go plus minus a volt there so yeah no worries and i was getting before but i'm not now unfortunately i was getting like large um spikes on there so something was switching in here i don't know what it's gone now of course it is as soon as i hit record white coat syndrome and of course if i touch those leads there you can see yeah it just changes if you twist them it's going to change if you you know it depends where you've got this will change from lab to lab whether or not you're holding them it'll change from one part of your lab to another it'll like just vary all the time because you've got such a large input impedance you can see that change that i just played around with there on the trend chart there and we can probably do that again let me get the leads and i'll actually twist them okay so what i've gone and done now is actually uh twisted the leads like that and you can see that that has significantly reduced the pickup there but of course it all has to do with the number of power line cycles so when you're playing around with your multimeter especially these bench ones um that can do a really fast integration times and stuff like that you need to know about your number of powerline cycles and how not only how it can influence uh the display resolution um but also can influence your noise pick up there it is just magically vanished and if you want the most accurate readings like you're going to put it on like a hundred power line uh cycles and to give you an example of this i'm actually feeding in five volts dc superimposed with a one volt peak-to-peak 50 hertz uh sine wave and you can see that it's bang on five volts because we've got the number of power line cycles equal to one and if we go to point two you know there we go it's jumping around like a jackrabbit 0.02 0.06 there you go it's jumping around like crazy but if we go to the number of powerline cycles at least equal to one it magically vanishes and we can see that perfectly on the uh trend chart here you'll notice it's precisely uh plus minus 0.5 of a volt there one volt peak-to-peak that's exactly what we're uh seeing and if we actually extracted that data and looked in hopefully we can see a sine wave but if we change our number of power line cycles instantly go up to one bingo it's stopped we're actually getting a flat line there now and we can go back to point two power line cycles and you can see at point two it's getting a bit we probably won't if we actually looked at the data and zoomed in we probably wouldn't see a perfect sine wave there but the lower that we go the more solid you see that's going to get right so what i've done is i've pulled this data from the multimeter put it into a spreadsheet here and we can graph it and i'm changing the modes in the power line cycles and you can see like four distinct modes here now this flat one over here this is five volts we're of course feeding in five volts plus minus half a volt 50 hertz signal on there and we're completely flat line in here and you may have guessed this is one power line cycle so that's 20 milliseconds cycle time or aperture time as it's called or sampling time they're all basically the same thing it's just different uh terminology some manufacturers might use a different term but basically an aperture time there of 20 milliseconds so that allows us to get at 50 hertz signal is to be different for 60 but at 50 hertz we get one complete main cycle so it averages out and that's why we get a flat line and it's not averaging out mathematically later as we just saw it's actually doing it in the integration or sampling time of the analog to digital converter the if it takes this much time to sample it in that time the 50 hertz noise has gone exactly one complete cycle and it's just averaged itself out and we get five volts magic but at this point here i then switch to 0.2 power line cycles or 4 and now which is 4 milliseconds aperture or our sample time and as you can see we start to see i you know up to like we start to see the peak there that one volt peak to peak signal there but you might have noticed this it's kind of like modulated you might have seen this before in your oscilloscope this looks like classic alison this is all to do with your nyquist stuff right where you need at least twice the sample right otherwise you get aliasing so we're clearly getting sampling artifacts here of our 50 hertz signal and it it's not good enough because we're only sampling at with an aperture time of 4 milliseconds or 0.2 powerline cycles now at this point here i switched over to 0.06 powerline cycles or 1.2 milliseconds aperture time and as you can see we really start to get a pretty decent signal it's still not absolutely perfect because our sample rate's not very high and at this point over here i switched to 0.02 powerline cycles or 0.4 milliseconds aperture time as you can see we get pretty much a perfect sine wave there and you can see how it's effectively changed what looks like changing frequencies there at at each point because we're taking more samples each time we set each time we change that number of power line cycles is changing our sample rate effectively so there you go we get sampling artifacts just like you would on an oscilloscope or a data logger or anything a bench multimeter is no different it's just a sampling system that's it it's not rockin science just how these things work and of course it all has to do with the input impedance that 10 meg ohms is actually quite high and it picks it all up and if you put go whack a 1k resistor in parallel with it it's going to knock it on the head and when if you go and measure uh like a low impedance voltage source like a battery because it's got like milliohms output well a source impedance and you measure that that's why we can get just the probes there we can get like a volt of noise yet when we measure our battery like that we will get 1.30254 and we'll only get the noise will only be a couple of least significant digits like that so that's all to do with the impedance of your measurement source in this case the impedance of our measurement source is 10 meg and it's just pink and we've got these big antenna uh leads on here picking up the 50 hertz which is like plus minus a volt so there you go very interesting stuff number of powerline cycles that has to do with the integration time which is different to any sort of smoothing or averaging mode which the meter might do after that because you're doing that after the measurement and not before so it's all to do with the measurement time of the analog to digital converter and of course it's going to slow down your measurement the more number of power line cycles you have but you get increased accuracy and rejection of 50 60 hertz noise so i hope you found that interesting if you did please give it a big thumbs up as always discuss it down below and check out my alternative platforms like odyssey i think i'm close to 60 000 subscribers on odyssey now we don't win a chicken dinner catch you next time you

Channel: EEVblog

Views: 51,656

Rating: 4.9489303 out of 5

Keywords: eevblog, video

Id: yKb7xIsPjVM

Length: 18min 25sec (1105 seconds)

Published: Sun Mar 07 2021