I love demoing this trickery to people however I usually take it a bit further and put my feet on a power board or hold a mains cable which capacitively couple huge amounts (for 10 MΩ) of signal onto a bench DMM or oscilloscope (sometimes 50+ V and enough to dimly light an LED), this way you can also demo how grounding yourself almost completely removes the signal and that those capacitively coupled 'ghost' voltages (especially those from isolated switching supplies) aren’t something to be afraid of (but might give you a small tingle).
"and that those capacitively coupled 'ghost' voltages (especially those from isolated switching supplies) aren’t something to be afraid of (but might give you a small tingle)." Oh, that's what's happening with my stuff! Thank you for unintentionally helping me! I was always wondering why anything I plug into my multi-outlet power thing has this 60V and sometimes higher voltage on it's grounded parts. It hurts when I touch it with my forearms. Fun and concerning. Thank you very much! Now I only need to know how to get rid of it without grounding me.
@@GLITCH_-.- You should be able to just ground the negative side of your power thing to remove the leakage. Grounding only yourself will actually increase the current through you for a stronger tingle. The signal/tingle is a result of capacitive coupling between mains and the output through either the transformer or Y class capacitor across the transformer so only special medical grade PSUs have ultra-low leakage currents but that comes at the cost of efficiency and price.
@@WizardTim I live in 230V. Can I really connect mains earth with a wire to a I. E. Li-ion Charger that is powered with a 12V brick, without blowing it up? I mean, I should be, but earth is on a different potential than the 50+V floating charger - or more accurately: the negative of the 12V brick. I feel like this could be problematic for some power bricks.
@@GLITCH_-.- This is not technical advice, but you can always take a 60W or so mains *incandescent* lightbulb and connect it between your points of interest - if it lights up, you'd have blown the fuses shorting those points; if it doesn't, there's not enough current passing there for anything spectacular, even if you can measure some voltage. Don't use LED lamps though, the whole point is to try drawing significant current and LED lamps light up from any leakage.
Most people outside of the electronics industry have no idea how big and ubiquitous the electric fields are that surround us all the time. When I was just starting out in electronics design, I was working on a switching power supply, and was measuring the switching frequency which changes with load on a very old (1950 - 1960) scope. I was doing this in the basement of the house where I was living. Depending on the day, I was seeing both the expected switching frequency but also a signal at 1.160 MHz. I chased that 1160 KHz signal for weeks, trying to understand what was going on. Finally 1 day the signal was so big that dominated the measurement to such and extent that it was clear that this signal had some sort of modulation on it. So I ran it through a diode and then into my stereo input, and presto I was listening to a local AM radio station on 1160 KHz. Turns out that the basement dimensions were just the right size to form a resonant cavity, and that AM station would somedays be up to 1.5 Vpp on open leads. Even just moving to an upstairs room would cause the radio signal to be reduced to more normal radio signal level of uV or nV, un-measurable on my ancient scope. Just a little story about how experience is important when designing anything. :-)
Fun fact: The integration time adjustment on many meters maxes out at 10 NPLC, and the 100 NPLC setting is just an average of ten 10 NPLC measurements for example. This is because of the 1/f noise in the input amps and integrator which becomes problematic with longer integration times. I think the Keithley 7510 that Dave shows in the video performs best in the 1-5 NPLC range, Keithley might have shown a graph of noise vs NPLC in the manual somewhere, very interesting stuff!
These are the videos that both make this channel what it is and underrate it at the same time! But I believe that these will be the videos that people will always come back to in time because the content is so valuable... Great job 👌
Thank you Dave! I have a 5.5 digit HP bench multimeter. I bought it used on eBay and the 1 VDC reading had me concerned. I did a google on the subject and read it was due to the high impedance. I let it go at that. Now watching this great video enlightened me so much deeper on the subject. I had not idea the multimeter was integrating. I never knew about the powerline cycles etc. Now I know, thanks to you. I am now going back through my multimeter manuals to understand how to vary the number of PLCs etc.
The Keithley DMM 7510 is superior in many respects than the Keysight 34470A. The Keithley has a 1 million sample per second reading capability and several other advantages! And as far as the cost difference, if you fully option the Keysight 34470A it is $5,369. The Keithley DMM 7510 has no option up charges because it is already fully optioned at $5,420. This is only a $51 dollar difference! When you do your homework, the Keithley DMM 7510 is a hands-down winner! I have both meters in my electronics laboratory, my go to meter is the Keithley DMM 7510.@@EEVblog
Excelent video!! Welcome to the difference between deterministic noise and random noise. Averaging ( low pass filter after sampling depending on implementation ) is mainly designed to kill high freq random noise. A moving average ( different from averaging frames) can cancel out a deterministic noise if the averaging period exactly matches the signal period( to kill a 50hz signal with a Fs of 100KHz you need 2k samples) . However as you said, "Integration" is effectively anti-alliaisng low pass filter before sampling is performed. By increasing the the value of integration you move the role off frequency under 60Hz killing the deterministic noise.
You shouldn't think of it using the term "roll-off frequency" or as anti-aliasing filter. That's a term used for low-pass filters. The integration time actually makes the device completely blind for multiples of the the sampling frequency. That's a notch reject filter (more precisely: a comb filter) instead. When you integrate for 20ms, you will reject all 50Hz stuff, because you always integrate over a whole period (*without weighting*), so it doesn't matter at what phase you start sampling. On the other hand, you will get a lot of signal for 75Hz noise, because sometimes you catch two negative cycles, sometimes you catch two positive cycles and sometimes you catch some position in-between. The 75Hz will alias to 25Hz. That's why the integration time shouldn't be called "anti-aliasing" filter.
@@tw11tube I agree if they are using a pure comb filter (Not integrator-comb filter) ! As you said, Integration followed by resetting and then integration again creates a filter that has notchs at Fs and other integer multiple of Fs. In fact, this is what a Continuous-time delta-sigma ADC (infinit resulotion) dose by feeding back the result of integration every cycle and subtracting it from the input (resetting the integration). In Daves example the deterministic signal falls exactly at sampling frequency and gets completely canceled out by the "notch". However if the signal was a bit off from the fs it would still get attanuated significantly. In my line of work we call this transfer function an anti-aliasing filter since it always have a notch at Fs and always canceles the aliasing that happens "around" Fs (not a comb filter though check the response of a CT delta-sigma ADC or a integrating ADC). However, I totally agree with you, a pure comb filter with no other attanuation is very very bad anti-aliasing filter only blocking certain frequency
When sampling a signal, you can reject local interference by sampling at a frequency equal to the lowest common multiple of the frequencies to reject, or harmonics of the LCM of the frequencies to reject. EG Sample at 300 SPS to reject both 60 and 50 Hz.
There are a lot of odd harmonics in the mains from switch mode supplies etc. Since you are capacitively coupling the signal into the open test leads, it will naturally boost the harmonics compared to the fundamental.
@@LutzSchafer Even linear supplies create harmonics because of the small conduction angle of the rectifiers i.e. current is only drawn near the peaks of the mains waveform. I can remember wondering why the waveform was distorted when I touched the probe of my old Hameg 20Mhz CRT scope 35+ years back, though I do remember it being somewhat sinusoidal rather than triangular.
Internally for many DMM aperture time is not always equal NPLC, so saying its a symptom is not quite right. For example settings NPLC100 on 3458A is not same as aperture to same time, as meter internally takes NPLC10 readings and filter them together.
You keep telling you cannot get the same result with short integration times and averaging the samples. But I cannot see why averaging would not work, as long as the sample rate is high enough to avoid aliasing AND the averaging time window is large enough. Assuming you have some large sample rate, as the length of the averaging window approaches infinity, the cutoff frequency of the "effective lowpass filter" should be approaching zero.
Is the timing for the integration directly coupled to the mains cycles or is it determined by some crystal oscillator? I imagine that this could lead to some kind of discrepancy/error because of oscillator tolerances and drift.
@@EEVblog i think the point was the power frequency could be 49.9 hz and the MM "PLC" could be referenced to 50.0, so 1 PLC wouldnt necessarily get the job done so cleanly. So not phase but absolute f
@@romand4651 Yes exactly. I don't know how noticable the effect would be, but even a slight difference in absolute frequency could be noticable in a 7 1/2 (or more) digit multimeter. It would be interesting to see if there's any circuitry to get a "mains clock" from the PSU of the multimeter or if the internal oscillator specs are good enough for the effect not to be noticable.
I don't understand why post-averaging does nothing. I could write a simple program to do a moving average on those fluctuating voltages shown and it would definitely have an effect. Either the meter's averaging window is very small as to make it useless or I'm missing something.
I have the same lack of understanding. My best guess is that on low NPLC it's not integrating all the time (analogous to recording 60Hz video with 1/100 shutter speed, the shutter isn't open all the time). The problem with that guess is it doesn't make sense why they'd let you run it with the shutter closed part of the time.
@@eDoc2020 Yeah, it must be something major we're missing. But whatever it is, the exact method of how the meter samples or integrates is clearly not how we think it is. However, so far the post-averaging function is counter intuitive, misleading, defective. It should not fail to work as claimed by its naming. Perhaps Dave could do a small follow-up to explain this.
Actually, you can google this when you need this info. But Dave's visual presentation and charismatic explanation is always the best way to learn and remember something new! 👍
Thanks for explaining the reason for this, I have Keithley 2000 and 2015 bench meters. The voltages are higher on the 2015, thought it was some coupling from the audio circuits until I got the 2000.
I think I remember from my school years, that this phenomenon has to do with Nyquist theory of signal sampling, that states that if you want to measure a signal of a given frequency, you need to set your aperture to at least twice the frequency of the phenomenon. This NPLC setting is doing exactly opposite: if you don't ever want to see a signal of a given frequency (to exclude it as a noise without having to filter it later using math ) you just set your aperture low enough.
Does the equipament measure it's AC supply frequency to determine if 1PLC is 1/50 or 1/60 second? What if the instrument power cord is being fed by a generator or inverter with a frequency that is not in sync with the "real" all around mains electrical fields?
@@georgejobin1744 Personally, I really like the EEVBlog BM786. I do NOT have battery issues with mine. I still do not care for 121GW, but it is "okay" and does NOT have battery issues either. Perhaps you have a bad one. Send Dave a message to get a replacement.
Is the PLC just a fixed value or is the up down integration timing actually coming from the power line of the bench plug ... Ie will it track drifts in the power frequencies?
Sampling high impedance singlas is where this can really come in and bit one. I though sometimes wonder, if 230 +/-20 VAC noticeably worse than 110 +/-10 VAC. Considering that most regions using the lower voltages tends to operate at 60Hz instead of 50Hz, then it should have some stronger capacitive coupling into leads and such, but the reduced voltage should negate that effect by a fair bit more. I guess there isn't an all too major difference for one or the other in the end, and if one is really paranoid, then one would just shield one's main power cables. (And yes, magnetic interference is also an issue, but steel conduit should reduce such. Though, building an anechoic chamber to have a lab inside of is likely the highest degree a voltnut can reach...)
It's quite unscientific I'm sure, but I hooked a fm car antenna up to my scope and walk around and look at all the stuff it picks up. Holding it up by the microwave produces some interesting stuff as well as sockets, cfl's and so on.
They probably just use 100ms (or a multiple thereof). On 50Hz that would be 6 PLC and on 60Hz it would be 5 PLC, but in each case it's still an integral NPLC.
Is the BM786 integration time in the 50Hz filter in this multimeter set in the firmware or is this a physical integrating RC plus some kind of diode chain? In light of some of the minor issues with the EMF pickup reported regarding the measurement of reasonably large resistance around 660-670kOm would be interesting to know
Do these high end DMMs actually measure incoming mains crossing point to sync the NPLC, or does it rely on an internal clock? Just thinking you could get quite a lot of phase error on the final cycle with high NPLC counts if it's not synced.
Is that sampling video available somewhere? Because I don't see the difference between averaging lots of samples taken at way above nyquist rate and integration as long as they represent the same amount of wall time. At least if we remove numerical precision issues in the averaging.
Nice. Nice. On a related note. I wonder if shielded mains cabling in walls and lighting, and from sockets to equipement, would help with anything to reduce the noise floor further. I know there are mains cable with shield or screen, or even both. There is for example a Swiss company called Gotham Cables that manufacturers some shmick multi-shielded mains cables (about 8CHF/m), I am not exactly sure who is manufacturing these cables for Gotham, but maybe labkable from Chaina. Another similar company is Lappcable, Yarbo, Elecaudio, and Oyaide (some of which are definitively total snake oil and don't come with any datasheets or certifcations, but still are interesting) . They obviously sell these stuff mostly to audiophilic nuts, but maybe there is a bit of merit in using them. Similar cables can be found for industrial applications, most notably motors, servos and steppers, but they usually have weird voltage ratings, and usually way more conductors to be useful for mains. Usually one will then connect the shield and screen to earth close to the source (i.e. socket or distribution board). Another option is to put mains cables in walls and stuff in metal pipes or channels, and ground them. Just an idea for future experiments. In some lab settings it is possibility to use that, for example when building a new lab, then one could put proper trunking, grounded pipes, shielded cables, and shielded cords and outlets, and lighting (there are companies making EMI shielded LED drivers and fixtures, as well shielded incandescend light sources, often to be used in RF testing or testing chambers).
Is it fair to say that the difference between PLC integration and simply averaging multiple samples later, is that the samples averaged may not exactly coincide with the mains cycle? Kind of like filming a spinning wheel; only when the frame rate exactly coincides with the rotation speed does the wheel appear to stand still, otherwise regardless of frame rate it will appear to spin one way or the other. The wheel can even appear to slow down and everse direction as the frame rate comes in and out of time with the rotation speed.
How do the high end dmm meter overcome the issue related to noise coupled via the mains supply which is coupled via the mains step down transformer. Even if we have filter capactor at the output stage of the power supply will they still not couple noise on the powersupply lines?
Hey Dave! You should take a look at the Miniware MDP XP portable lab power supply. It can run off a power bank too! It looks like a pretty good gadget worth taking a look at. I'm planning on buying myself one. :)
Shielding mains is generally quite difficult skin depth is quite deep at that low of a frequency, would need a thick layer of steel 10s of cm iirc, impractical Edit(steel is actually quite a bit better probably could get by with less tha. A centimeter, copper/aluminium is abysmal though) Oscilloscope probes are functionally shielded, but they'll grab it just as easily
I think this is what my 3478A is doing. I always thought is was going to break at some point. The resistance is pretty far out of spec on lower resistances but it still works. Are these still worth sending off for calibration?
I've got a cheap handhelt autoranging multimeter, and it jumps around even much worse than these high end meters. It jumps around in the tens of volts...
I am not getting how the mains signal is being picked up. Is it just from a 50/60Hz magnetic field coming off of nearby powered-up equipment and transformers? That would explain why twisting the leads greatly decreases the amplitude of the signal on the bench multimeter. But the leads aren't connected to each other. They don't form a loop! How can a magnetic field induce a current without a conductor that loops? I am thinking here of the Maxwell-Faraday equation where EMF is induced by magnetic flux-- but magnetic flux, by definition, is measured through a *closed* loop. Or is it somehow caused by the power supply of the bench meter? Would a battery-powered meter not show this? No, Dave suggests the reason the handheld DMM doesn't show it is because the signal is integrated over a large enough time interval. That suggests it is a real signal that is coming through space and putting charges on the inputs of the instrument, right? What exactly is the physical mechanism that causes the 50/60 Hz signal to end up measured by the bench meter?
Hi Dave, brymen bm786 is currently sold out on amazon. Is there any plans to resume it's supply in nearest future? I really want to order one. Thanks in advance.
Imagine that your accurate, very high impedance meter has a 1.500V battery (yes, I know; cell) with 1000Meg resistor in series, *across* the test leads and hidden from sight. The meter would read 1.500V, but if you put it across (say) a LED operating at a few mA, would read the correct voltage. It's all a matter of SOURCE IMPEDANCE. With open leads the source impedance is about infinity so any stray charges would dominate. Nothing to do with sampling time and aliasing. BTW: I presume you know there is a DC voltage gradient of about 500 V/m in the open air from ground to at least a few hundred metres height.
The bench multimeters are mains powered but pretty much none of them are referenced to ground. And the amount of 50Hz noise our power grids produce theese days coupled with all the harmonics generated by periodic current peaks from crappy slow rectifiers inside cheap switchmode PSUs we have all around us is so big that it does not take much to receive the signal. You don't need a direct conection to ground, just a couple of hanging wires capacitively coupling into different parts of the surroundings is more than enough for 10M. Imagine an atomic power plant feeding the high voltage lines of the unified national grid somewhere near Murmansk and a shipyard somewhere in Vladivostok consuming the power. It is just 4 huge wires carrying a lot of current hanging above the ground and continiously going trough like 1/3 of the globe. The resulting magnetic radiation just more or less goes everywhere. Maybe apart from some really distant places on earth noone really lives in anyway.
I'm putting my money on SMPS tomfoolery, I've observed my phone charger being able to light up an LED dimly when I hold one pin and touch the ground on the USB plug.
@@EEVblog lol I mean common mode noise coupled between the DC output and earth ground of the supply, though on 2nd thought the meter should be able to reject that.
I love demoing this trickery to people however I usually take it a bit further and put my feet on a power board or hold a mains cable which capacitively couple huge amounts (for 10 MΩ) of signal onto a bench DMM or oscilloscope (sometimes 50+ V and enough to dimly light an LED), this way you can also demo how grounding yourself almost completely removes the signal and that those capacitively coupled 'ghost' voltages (especially those from isolated switching supplies) aren’t something to be afraid of (but might give you a small tingle).
Fascinating!
"and that those capacitively coupled 'ghost' voltages (especially those from isolated switching supplies) aren’t something to be afraid of (but might give you a small tingle)."
Oh, that's what's happening with my stuff! Thank you for unintentionally helping me! I was always wondering why anything I plug into my multi-outlet power thing has this 60V and sometimes higher voltage on it's grounded parts. It hurts when I touch it with my forearms. Fun and concerning.
Thank you very much! Now I only need to know how to get rid of it without grounding me.
@@GLITCH_-.- You should be able to just ground the negative side of your power thing to remove the leakage. Grounding only yourself will actually increase the current through you for a stronger tingle. The signal/tingle is a result of capacitive coupling between mains and the output through either the transformer or Y class capacitor across the transformer so only special medical grade PSUs have ultra-low leakage currents but that comes at the cost of efficiency and price.
@@WizardTim I live in 230V. Can I really connect mains earth with a wire to a I. E. Li-ion Charger that is powered with a 12V brick, without blowing it up? I mean, I should be, but earth is on a different potential than the 50+V floating charger - or more accurately: the negative of the 12V brick. I feel like this could be problematic for some power bricks.
@@GLITCH_-.- This is not technical advice, but you can always take a 60W or so mains *incandescent* lightbulb and connect it between your points of interest - if it lights up, you'd have blown the fuses shorting those points; if it doesn't, there's not enough current passing there for anything spectacular, even if you can measure some voltage. Don't use LED lamps though, the whole point is to try drawing significant current and LED lamps light up from any leakage.
Dave finds the free energy test leads and still doesn't believe in it
BUSTED!
Please tell (or better not tell?) ElectroBOOM about it. :D
I blame Solar Roadways...
quantum vacuum alert! 🚑
It's obviously zero point energy!
Most people outside of the electronics industry have no idea how big and ubiquitous the electric fields are that surround us all the time.
When I was just starting out in electronics design, I was working on a switching power supply, and was measuring the switching frequency which changes with load on a very old (1950 - 1960) scope. I was doing this in the basement of the house where I was living. Depending on the day, I was seeing both the expected switching frequency but also a signal at 1.160 MHz. I chased that 1160 KHz signal for weeks, trying to understand what was going on. Finally 1 day the signal was so big that dominated the measurement to such and extent that it was clear that this signal had some sort of modulation on it. So I ran it through a diode and then into my stereo input, and presto I was listening to a local AM radio station on 1160 KHz. Turns out that the basement dimensions were just the right size to form a resonant cavity, and that AM station would somedays be up to 1.5 Vpp on open leads. Even just moving to an upstairs room would cause the radio signal to be reduced to more normal radio signal level of uV or nV, un-measurable on my ancient scope.
Just a little story about how experience is important when designing anything. :-)
Don't feed the electro-smog zealots! :->
These little fragments of new knowledge is what I love most about this channel. :)
Fun fact: The integration time adjustment on many meters maxes out at 10 NPLC, and the 100 NPLC setting is just an average of ten 10 NPLC measurements for example. This is because of the 1/f noise in the input amps and integrator which becomes problematic with longer integration times.
I think the Keithley 7510 that Dave shows in the video performs best in the 1-5 NPLC range, Keithley might have shown a graph of noise vs NPLC in the manual somewhere, very interesting stuff!
Free energy community: "That's free 2Vpp form empty lead! Gonna harvest that"
And use it to mine bitcoinage..
These are the videos that both make this channel what it is and underrate it at the same time! But I believe that these will be the videos that people will always come back to in time because the content is so valuable...
Great job 👌
Thank you Dave! I have a 5.5 digit HP bench multimeter. I bought it used on eBay and the 1 VDC reading had me concerned. I did a google on the subject and read it was due to the high impedance. I let it go at that. Now watching this great video enlightened me so much deeper on the subject. I had not idea the multimeter was integrating. I never knew about the powerline cycles etc. Now I know, thanks to you. I am now going back through my multimeter manuals to understand how to vary the number of PLCs etc.
That Keithley has an amazing refresh speed, looks like at least 50x a normal multimeter.
50x times the price :-D
Heck, all of em look like light speed compared to the DMMs I learned with...
The Keithley DMM 7510 is superior in many respects than the Keysight 34470A. The Keithley has a 1 million sample per second reading capability and several other advantages!
And as far as the cost difference, if you fully option the Keysight 34470A it is $5,369.
The Keithley DMM 7510 has no option up charges because it is already fully optioned at $5,420.
This is only a $51 dollar difference!
When you do your homework, the Keithley DMM 7510 is a hands-down winner! I have both meters in my electronics laboratory, my go to meter is the Keithley DMM 7510.@@EEVblog
Excelent video!! Welcome to the difference between deterministic noise and random noise. Averaging ( low pass filter after sampling depending on implementation ) is mainly designed to kill high freq random noise. A moving average ( different from averaging frames) can cancel out a deterministic noise if the averaging period exactly matches the signal period( to kill a 50hz signal with a Fs of 100KHz you need 2k samples) . However as you said, "Integration" is effectively anti-alliaisng low pass filter before sampling is performed. By increasing the the value of integration you move the role off frequency under 60Hz killing the deterministic noise.
You shouldn't think of it using the term "roll-off frequency" or as anti-aliasing filter. That's a term used for low-pass filters. The integration time actually makes the device completely blind for multiples of the the sampling frequency. That's a notch reject filter (more precisely: a comb filter) instead. When you integrate for 20ms, you will reject all 50Hz stuff, because you always integrate over a whole period (*without weighting*), so it doesn't matter at what phase you start sampling. On the other hand, you will get a lot of signal for 75Hz noise, because sometimes you catch two negative cycles, sometimes you catch two positive cycles and sometimes you catch some position in-between. The 75Hz will alias to 25Hz. That's why the integration time shouldn't be called "anti-aliasing" filter.
@@tw11tube I agree if they are using a pure comb filter (Not integrator-comb filter) ! As you said, Integration followed by resetting and then integration again creates a filter that has notchs at Fs and other integer multiple of Fs. In fact, this is what a Continuous-time delta-sigma ADC (infinit resulotion) dose by feeding back the result of integration every cycle and subtracting it from the input (resetting the integration). In Daves example the deterministic signal falls exactly at sampling frequency and gets completely canceled out by the "notch". However if the signal was a bit off from the fs it would still get attanuated significantly. In my line of work we call this transfer function an anti-aliasing filter since it always have a notch at Fs and always canceles the aliasing that happens "around" Fs (not a comb filter though check the response of a CT delta-sigma ADC or a integrating ADC). However, I totally agree with you, a pure comb filter with no other attanuation is very very bad anti-aliasing filter only blocking certain frequency
3:30 the stock market uses this as their input for deciding the prices of most stocks lol
And those who follow this trend will henceforth be known as Mains Chartists
Thanks for demystifying this.
Committed to memory. As always, thank you Dave
I always thought my meter was at fault since it was a dumpster find. Good to know it's working just fine. Thanks.
Wow Dave your meter looks so much better than the OEM red meter. Looks good in blue.
wow that was interesting! I never expected such a phenomena to occur!
When sampling a signal, you can reject local interference by sampling at a frequency equal to the lowest common multiple of the frequencies to reject, or harmonics of the LCM of the frequencies to reject. EG Sample at 300 SPS to reject both 60 and 50 Hz.
Dave have you ever wondered why picked up PowerLine noise is never even close to sinusoidal and rather triangular? Where is the non-linearity? ???
There are a lot of odd harmonics in the mains from switch mode supplies etc. Since you are capacitively coupling the signal into the open test leads, it will naturally boost the harmonics compared to the fundamental.
@@ferrumignis Ahhh yes you are absolutely right. Thank you. So I assume in the olden days where only linear psu's were used it was much more cleaner.
@@LutzSchafer Even linear supplies create harmonics because of the small conduction angle of the rectifiers i.e. current is only drawn near the peaks of the mains waveform. I can remember wondering why the waveform was distorted when I touched the probe of my old Hameg 20Mhz CRT scope 35+ years back, though I do remember it being somewhat sinusoidal rather than triangular.
Dave must have found those leads in the Over Unity section on ebay.
I miss Bob Pease , what a star he was.
I recently picked up a copy of his _Troubleshooting Analog Circuits,_ it’s a good read.
Internally for many DMM aperture time is not always equal NPLC, so saying its a symptom is not quite right. For example settings NPLC100 on 3458A is not same as aperture to same time, as meter internally takes NPLC10 readings and filter them together.
I'm already fall in love this meters
When a microphone capsule FET input leg is open, I can hear tons of random hiss.
That's not random, it's how you configure the mic to hear the ghosts talking.. 👻
Love the title, RIP the analogue king
You keep telling you cannot get the same result with short integration times and averaging the samples. But I cannot see why averaging would not work, as long as the sample rate is high enough to avoid aliasing AND the averaging time window is large enough. Assuming you have some large sample rate, as the length of the averaging window approaches infinity, the cutoff frequency of the "effective lowpass filter" should be approaching zero.
Yeah ok, I see your point there. I guess in theory that's possible, but try it on any meter and the lowest NPLC setting still gives a crap result.
Is the timing for the integration directly coupled to the mains cycles or is it determined by some crystal oscillator?
I imagine that this could lead to some kind of discrepancy/error because of oscillator tolerances and drift.
Doesn't need to be synced, as you'll average out to zero regardless of where on the cycle you start.
@@EEVblog i think the point was the power frequency could be 49.9 hz and the MM "PLC" could be referenced to 50.0, so 1 PLC wouldnt necessarily get the job done so cleanly. So not phase but absolute f
@@romand4651 Yes exactly.
I don't know how noticable the effect would be, but even a slight difference in absolute frequency could be noticable in a 7 1/2 (or more) digit multimeter.
It would be interesting to see if there's any circuitry to get a "mains clock" from the PSU of the multimeter or if the internal oscillator specs are good enough for the effect not to be noticable.
I typically leave mine set at 10 NPLC, sometimes I use 100NPLC if I need more accuracy.
At least 1 NPLC is the takeaway!
I don't understand why post-averaging does nothing. I could write a simple program to do a moving average on those fluctuating voltages shown and it would definitely have an effect. Either the meter's averaging window is very small as to make it useless or I'm missing something.
I have the same lack of understanding. My best guess is that on low NPLC it's not integrating all the time (analogous to recording 60Hz video with 1/100 shutter speed, the shutter isn't open all the time). The problem with that guess is it doesn't make sense why they'd let you run it with the shutter closed part of the time.
@@eDoc2020 Yeah, it must be something major we're missing. But whatever it is, the exact method of how the meter samples or integrates is clearly not how we think it is. However, so far the post-averaging function is counter intuitive, misleading, defective. It should not fail to work as claimed by its naming. Perhaps Dave could do a small follow-up to explain this.
So, essentially, you’re deliberately abusing the Nyquist Sampling theory - winner, winner, chicken dinner! 🤣
I learn something!
Actually, you can google this when you need this info. But Dave's visual presentation and charismatic explanation is always the best way to learn and remember something new! 👍
Handy hint. Cheers, Dave.
Thanks for explaining the reason for this, I have Keithley 2000 and 2015 bench meters.
The voltages are higher on the 2015, thought it was some coupling from the audio circuits until I got the 2000.
I think I remember from my school years, that this phenomenon has to do with Nyquist theory of signal sampling, that states that if you want to measure a signal of a given frequency, you need to set your aperture to at least twice the frequency of the phenomenon. This NPLC setting is doing exactly opposite: if you don't ever want to see a signal of a given frequency (to exclude it as a noise without having to filter it later using math ) you just set your aperture low enough.
I literally used a 8.5 digit meter the other day, and saw this exact thing! So coincidental!
Does the equipament measure it's AC supply frequency to determine if 1PLC is 1/50 or 1/60 second? What if the instrument power cord is being fed by a generator or inverter with a frequency that is not in sync with the "real" all around mains electrical fields?
In these ones, yes, I think I showed that in one of the teardowns.
Good topic! 😁
The forum knows why this one came up!
Does not appear to be an issue with the Agilent 34401A 6 1/2 digit meter. I wonder why?
Agilent is a better meter my eev blog meter will kill a new battery in the off position in about 10 days there junk
@@georgejobin1744 Personally, I really like the EEVBlog BM786. I do NOT have battery issues with mine. I still do not care for 121GW, but it is "okay" and does NOT have battery issues either. Perhaps you have a bad one. Send Dave a message to get a replacement.
A 7.5 digit meter has a continuity beeper and cap range?!?!? Yikes. No transistor test??? Aww...
Love the title. Reminiscent of Bob Pease, RIP.
At first I had no idea what could cause that, but I'm pretty sure it's an alias after you showed us the trend. Great video as always!
Is the PLC just a fixed value or is the up down integration timing actually coming from the power line of the bench plug ... Ie will it track drifts in the power frequencies?
There is a mains detection tap, I think I showed that in the teardown of one of these.
Sampling high impedance singlas is where this can really come in and bit one.
I though sometimes wonder, if 230 +/-20 VAC noticeably worse than 110 +/-10 VAC.
Considering that most regions using the lower voltages tends to operate at 60Hz instead of 50Hz, then it should have some stronger capacitive coupling into leads and such, but the reduced voltage should negate that effect by a fair bit more.
I guess there isn't an all too major difference for one or the other in the end, and if one is really paranoid, then one would just shield one's main power cables. (And yes, magnetic interference is also an issue, but steel conduit should reduce such. Though, building an anechoic chamber to have a lab inside of is likely the highest degree a voltnut can reach...)
The video title reminds me of Bob Pease series... RIP EE greybeard
It's quite unscientific I'm sure, but I hooked a fm car antenna up to my scope and walk around and look at all the stuff it picks up. Holding it up by the microwave produces some interesting stuff as well as sockets, cfl's and so on.
Thanks Dave
At 1 PLC does it actually sample a crap load of samples and then integrate them or does it simply sample once a PLC?
It's one sample over one power line cycle (i.e. a 20ms sample integration time)
@@EEVblog Gotcha, thanx :)
takes a crapload of samples and averages them out.
or maybe continuously analogue-integrate the input for one power line cycle
@@victortitov1740 IDK See above from EEV.
Never forgetting this one....
So this means that hand-held DMM that don't allow manual setting of 50/60Hz have to be pre-configured for the target region?
They probably just use 100ms (or a multiple thereof). On 50Hz that would be 6 PLC and on 60Hz it would be 5 PLC, but in each case it's still an integral NPLC.
Is the BM786 integration time in the 50Hz filter in this multimeter set in the firmware or is this a physical integrating RC plus some kind of diode chain? In light of some of the minor issues with the EMF pickup reported regarding the measurement of reasonably large resistance around 660-670kOm would be interesting to know
1:50 Are you sure it's your microphone and not a geiger counter?
Do these high end DMMs actually measure incoming mains crossing point to sync the NPLC, or does it rely on an internal clock? Just thinking you could get quite a lot of phase error on the final cycle with high NPLC counts if it's not synced.
The place it starts in the cycle does not matter, juast that the period is precisely matched.
damn you deserve that IEEE recognition.
Is it (PLC) based on triggering or timer?
And would a multimeter for US market be less precise in a country with 50Hz mains?
With the probes tips unconnected the voltage across them could be anything. You need to connect to a known voltage. People usually short them.
You should watch this video again and then also watch some of Dave's other videos.
@@jeffm2787 Ok. I only saw the first 5 mins before commenting.
Is that sampling video available somewhere? Because I don't see the difference between averaging lots of samples taken at way above nyquist rate and integration as long as they represent the same amount of wall time.
At least if we remove numerical precision issues in the averaging.
Nice. Nice. On a related note. I wonder if shielded mains cabling in walls and lighting, and from sockets to equipement, would help with anything to reduce the noise floor further. I know there are mains cable with shield or screen, or even both. There is for example a Swiss company called Gotham Cables that manufacturers some shmick multi-shielded mains cables (about 8CHF/m), I am not exactly sure who is manufacturing these cables for Gotham, but maybe labkable from Chaina. Another similar company is Lappcable, Yarbo, Elecaudio, and Oyaide (some of which are definitively total snake oil and don't come with any datasheets or certifcations, but still are interesting) . They obviously sell these stuff mostly to audiophilic nuts, but maybe there is a bit of merit in using them. Similar cables can be found for industrial applications, most notably motors, servos and steppers, but they usually have weird voltage ratings, and usually way more conductors to be useful for mains. Usually one will then connect the shield and screen to earth close to the source (i.e. socket or distribution board). Another option is to put mains cables in walls and stuff in metal pipes or channels, and ground them. Just an idea for future experiments. In some lab settings it is possibility to use that, for example when building a new lab, then one could put proper trunking, grounded pipes, shielded cables, and shielded cords and outlets, and lighting (there are companies making EMI shielded LED drivers and fixtures, as well shielded incandescend light sources, often to be used in RF testing or testing chambers).
"wiggle, wiggle, wiggle, wiggle, wiggle, yeah: -- LMFAO!!
Greetings from Caracas Venezuela I just subscribed to your channel, thank you for sharing your knowledge God bless you always!
Is it fair to say that the difference between PLC integration and simply averaging multiple samples later, is that the samples averaged may not exactly coincide with the mains cycle? Kind of like filming a spinning wheel; only when the frame rate exactly coincides with the rotation speed does the wheel appear to stand still, otherwise regardless of frame rate it will appear to spin one way or the other. The wheel can even appear to slow down and everse direction as the frame rate comes in and out of time with the rotation speed.
Dave head not audio synced with vid. Only mentioned because I know you care how your new setup is doing.
How do the high end dmm meter overcome the issue related to noise coupled via the mains supply which is coupled via the mains step down transformer. Even if we have filter capactor at the output stage of the power supply will they still not couple noise on the powersupply lines?
Hey Dave! You should take a look at the Miniware MDP XP portable lab power supply. It can run off a power bank too! It looks like a pretty good gadget worth taking a look at. I'm planning on buying myself one. :)
Not mains and crap ...free energy!!
To Kickstarter!
Damn, free energy, ... have to make a circuit to charge my phone 😂
Just get a solar charger.
hey dave, would or could wifi and radio transmissions be the cause? the waveform looks like its repeating the same way a radio wave would
Keep watching, I show the waveform on a scope
I loved this episode - fun and enlightening topic.
Now.....send me one of those 7.5 digit meters so I can do confirmation tests in my own lab.
Then what is the use of sub unit PLC?
Very interesting stuff. I love these videos and find them interesting. Thumbs up :D
Should that mean then that when using a high end bench meter if they are that sensitive, should the test leads be shielded
Shielding mains is generally quite difficult skin depth is quite deep at that low of a frequency, would need a thick layer of steel 10s of cm iirc, impractical
Edit(steel is actually quite a bit better probably could get by with less tha. A centimeter, copper/aluminium is abysmal though)
Oscilloscope probes are functionally shielded, but they'll grab it just as easily
Could you manually calculate mean of sine of logged data?
I think it's some unproper use of mean function in multimeter in case it don't work...
I think this is what my 3478A is doing. I always thought is was going to break at some point. The resistance is pretty far out of spec on lower resistances but it still works. Are these still worth sending off for calibration?
That's still a nice meter, but you might be able to buy a working in spec ones cheaper than a full cal adjustment
I've got a cheap handhelt autoranging multimeter, and it jumps around even much worse than these high end meters. It jumps around in the tens of volts...
Hi Dave I thought the leads act like a dipole antenna....
What would be the purpose of having a non-integer PLC setting if it needs to be used always for accurate measurements?
Faster sample rate when using a low enough impedance source that the 'noise' becomes a non issue.
Try power the bench meter from isolated supply. Can it make any difference?
I am not getting how the mains signal is being picked up.
Is it just from a 50/60Hz magnetic field coming off of nearby powered-up equipment and transformers? That would explain why twisting the leads greatly decreases the amplitude of the signal on the bench multimeter. But the leads aren't connected to each other. They don't form a loop! How can a magnetic field induce a current without a conductor that loops? I am thinking here of the Maxwell-Faraday equation where EMF is induced by magnetic flux-- but magnetic flux, by definition, is measured through a *closed* loop.
Or is it somehow caused by the power supply of the bench meter? Would a battery-powered meter not show this? No, Dave suggests the reason the handheld DMM doesn't show it is because the signal is integrated over a large enough time interval. That suggests it is a real signal that is coming through space and putting charges on the inputs of the instrument, right?
What exactly is the physical mechanism that causes the 50/60 Hz signal to end up measured by the bench meter?
Hi Dave, brymen bm786 is currently sold out on amazon. Is there any plans to resume it's supply in nearest future? I really want to order one. Thanks in advance.
0:00 I bet it's gonna start with "Hi"
0:02 Knew it!
You mean Hoi!
Fascinating stuff...Oh yeah, can you do an impersonation of Ben Heck ?
Sorry, nope, I suck at seppo impressions.
Thanks for posting dave i got the same thing in my uni t ut61e
Who can kindly share the link to Dave's video on dual-slope ADC sampling? Thx!
So can you run it on a car battery?
I have several top bench meters in my own lab too. In my dreams.
Maybe this Is an idiotic question but what would happen if you were measuring mains power with the nplc set to 1? Do you get zero Volts?
I guess NPLC is just for DC, for AC you have Hz filters.
When will you be designing a filter? :)
Hello Dave! Big fan! Can you tell me when the 121GW will be back in stock?
Imagine that your accurate, very high impedance meter has a 1.500V battery (yes, I know; cell) with 1000Meg resistor in series, *across* the test leads and hidden from sight.
The meter would read 1.500V, but if you put it across (say) a LED operating at a few mA, would read the correct voltage. It's all a matter of SOURCE IMPEDANCE. With open leads the source impedance is about infinity so any stray charges would dominate. Nothing to do with sampling time and aliasing.
BTW: I presume you know there is a DC voltage gradient of about 500 V/m in the open air from ground to at least a few hundred metres height.
This is all a lot of fun when you are in a plane....
How does this affect measuring actual signals at 50 or 60Hz?
And wouldn't higher-frequency signals also average out?
Shouldn't you be using an oscilloscope for that?
Set the meter to AC ?
+1 for sorting out us newbies & self taught old grey beards.
Are the high end units mains powered?
If so, perhaps they're referenced to ground, and are picking up radio waves? That's my only logical assumption.
The bench multimeters are mains powered but pretty much none of them are referenced to ground.
And the amount of 50Hz noise our power grids produce theese days coupled with all the harmonics generated by periodic current peaks from crappy slow rectifiers inside cheap switchmode PSUs we have all around us is so big that it does not take much to receive the signal. You don't need a direct conection to ground, just a couple of hanging wires capacitively coupling into different parts of the surroundings is more than enough for 10M.
Imagine an atomic power plant feeding the high voltage lines of the unified national grid somewhere near Murmansk and a shipyard somewhere in Vladivostok consuming the power. It is just 4 huge wires carrying a lot of current hanging above the ground and continiously going trough like 1/3 of the globe. The resulting magnetic radiation just more or less goes everywhere. Maybe apart from some really distant places on earth noone really lives in anyway.
what about an Old School analog meter?
I'm putting my money on SMPS tomfoolery, I've observed my phone charger being able to light up an LED dimly when I hold one pin and touch the ground on the USB plug.
If your SMPS is working at 50Hz I'd say you have other problems...
@@EEVblog lol I mean common mode noise coupled between the DC output and earth ground of the supply, though on 2nd thought the meter should be able to reject that.
If no one else has said this sure Subway has or will either one but anyways...
And it flat lines....
It's dead Jim!
How to bug the new guys test bench...
Using a digital oscilloscope should be another way of demoing intergration
My fluke model 322 clamp meter was damaged, please can you send me the diagram, friend, I can't get it anywhere
se me daño mi pinza amperimetrica fluke modelo 322 por favor me puedes enviar el esquematico amigo no lo consigo en ninguna parte
very interesting!