"Veritasium" (YT) - "The Big Misconception About Electricity" ?

[EEVblog]

I just looked at AlphaPhoenix's video again.
He's using 4 probes, and using them in x1 mode which means about 85pF of capacitance between each of those points and ground.
So the 0.2V he's seeing is likely completely dominated by the probe capacitance. If so then the wire plays no part except when the wave hits the end and gets reflected back.
Have yet to play around with the simulation.

Am I wrong?

I thought about this again because I was just talking with Derek last night about a test he's doing and he was asking for any pittfalls.

[bdunham7]

He's using 4 probes, and using them in x1 mode which means about 85pF of capacitance between each of those points and ground.
So the 0.2V he's seeing is likely completely dominated by the probe capacitance. If so then the wire plays no part except when the wave hits the end and gets reflected back.
Am I wrong?

I think this is just another factor, the others being ground lead inductance, geometry and so on, that prevents any meaningful measurement of the leading edge part--like whether the light turns on at 1m/c or 3m/c--but I'm not sure it invalidates the whole test at the microsecond level.  85pF and 1kR means a time constant of 85ns.  So I think he does more or less capture the effect that he was looking for.  If you want to eliminate the types of errors that are in the same class as using the 1X probe, I think a lot of attention would have to be paid to the geometry of the layout, the connections and the fixtures for the test leads. 

[EEVblog]

He's using 4 probes, and using them in x1 mode which means about 85pF of capacitance between each of those points and ground.
So the 0.2V he's seeing is likely completely dominated by the probe capacitance. If so then the wire plays no part except when the wave hits the end and gets reflected back.
Am I wrong?

I think this is just another factor, the others being ground lead inductance, geometry and so on, that prevents any meaningful measurement of the leading edge part--like whether the light turns on at 1m/c or 3m/c--but I'm not sure it invalidates the whole test at the microsecond level.  85pF and 1kR means a time constant of 85ns.  So I think he does more or less capture the effect that he was looking for.  If you want to eliminate the types of errors that are in the same class as using the 1X probe, I think a lot of attention would have to be paid to the geometry of the layout, the connections and the fixtures for the test leads.

Yeah, was just calculating that.
It's about 5xRC for full charge, and it's 42.5pF for two in series, so 212ns just due to the probe charge curve into the load sense probes. He gets a 1.6uS rise.
And then yes the layout parasitics.

[EEVblog]

I think to do this properly you'd have to use a completely isolated scope on the load with the trigger coming from another scope on the switch side via optical fibre. And then you'd have to account for the skew correction in that trigger system as well. Very messy!

[bdunham7]

I think to do this properly you'd have to use a completely isolated scope on the load with the trigger coming from another scope on the switch side via optical fibre. And then you'd have to account for the skew correction in that trigger system as well. Very messy!

Yes, a similar thought occurred to me.  You want the scope right at the load/light--I would first try a resistor soldered across a BNC adapter-- then you want the signal generator at the switch point.  Then you need to make a shielded trigger cable to trigger the scope across the gap.  That would seem to me to be the minimum, and it still isn't going to be down to the single nanosecond. 

Since the original puzzle was turning a light on, an LED with a small solar panel from a pathway light might sort of work, but it probably isn't sensitive enough.  You could use an identical device with the fiber optic cable as the trigger, then whatever delays there were would be...sort of equalish?

Is he going to 'run some wires in the desert' as he said?

[EEVblog]

Is he going to 'run some wires in the desert' as he said?

I won't say at this stage, but he's doing some tests today, it's why he Zoomed me last night. He recorded the zoom for any potential video which may or not happen depending on the results.

[SiliconWizard]

I think to do this properly you'd have to use a completely isolated scope on the load with the trigger coming from another scope on the switch side via optical fibre. And then you'd have to account for the skew correction in that trigger system as well. Very messy!

Yep. As I hinted a while ago, any proper setup for such an experiment will require a lot of care and expensive equipment. Any quick test with common lab equipment is bound to be flawed. Not worth one's time and even less so that of others, if it's to show just patently wrong results.

[adx]

I must admit those coiled up taped up clippie leads made me cringe for a few moments. Then I realised I would do the same but leave them dangling in the breeze at random angles for that test. I'd use a wider line spacing if interested in the 1m/c thing.

A bit of effort put into "de-embedding" the test lead effects could clean up the unknowns more, and comes tantalisingly close to being able to get an answer to the 1(m)/c question even with that 100MHz scope. But reading delay from a scope in that circumstance (at the midpoint of a much slower apparent rise) is a lost cause if the waveshape isn't the same. Yes, my 60MHz TDS210 can easily resolve 600ps of jitter, but it's not the same as measuring a physically different point.

Another option I'm not sure if anyone here has suggested, is a cheap VNA, and do an IFFT on the result (someone has written a script to do it). An amorphous core and some carefully arranged resistors or amplifiers (like used in PDN analysis) could clean up the probing. VNA people do that sort of thing all the time.

Probes are an interesting philosophical situation (to me!) I've been thinking of since this video appeared. They provide a (literally) flexible measure of simultaneity, which even works in a 2D world, rather than requiring a 3rd dimension to observe it. There's no differential delays or assumptions needed. Even the speed of light doesn't come into it. It's only down to how they are folded up in space.

[Sredni]

I think to do this properly you'd have to use a completely isolated scope on the load with the trigger coming from another scope on the switch side via optical fibre. And then you'd have to account for the skew correction in that trigger system as well. Very messy!

You guys are still thinking you can measure that current with an oscilloscope, but that is not certain at all.
The initial current that flows into the load might flow in the load alone and not in the rest of the circuit, including any external probe you might attach to it.
You want to measure the transient when the current is forming in the wire, not the effect of a current flowing in the whole circuit. Do not expect KCL to work in the first few nanoseconds.

[bdunham7]

The initial current that flows into the load might flow in the load alone and not in the rest of the circuit, including any external probe you might attach to it.
You want to measure the transient when the current is forming in the wire, not the effect of a current flowing in the whole circuit. Do not expect KCL to work in the first few nanoseconds.

Leaving aside the issue of why there would be current only in the load and not in the wires connected to it, how long do you think current can flow in a small resistor before the voltage becomes apparent to the attached oscilloscope with a carefully laid out input fixture?

[EEVblog]

I think to do this properly you'd have to use a completely isolated scope on the load with the trigger coming from another scope on the switch side via optical fibre. And then you'd have to account for the skew correction in that trigger system as well. Very messy!

You guys are still thinking you can measure that current with an oscilloscope, but that is not certain at all.
The initial current that flows into the load might flow in the load alone and not in the rest of the circuit

Yes, but now you are down to basic measurement stuff. The 1k ohm is a pretty low impedance, so any induced pickup in the test leads shouldn't be a problem. If you measure the differential voltage (galvanically isolated) across the load then you shouldn't have any major problems. The key is completely isolating from the switch side.

[BrianHG]

I think to do this properly you'd have to use a completely isolated scope on the load with the trigger coming from another scope on the switch side via optical fibre. And then you'd have to account for the skew correction in that trigger system as well. Very messy!

Not too much a problem, 1 single LED laser diode optical source driving 2 identical length optic fibers where on 1 fiber's output, the photodiode triggers the mosfet switch on the wire (battery powered) and on the second fiber's side, another identical photodiode triggers the scope through a matched mosfet with a pull-up/down resistive load.   Fast 5ns photodiodes are only around ~1$ at digikey.  The problem is on the scope, your reference switch-on time will be an equivilant circuit to what happened on the battery powered optical switch on the other side of the cable.  So a fast mosfet with a good ultra low-on impedance, like some of those cheap avalanche T-MOS devices would apply here.  Though, redoing your measurements with the scope probes on the other side can verify the expected matched switch-on time.

With this cheap setup, I would consider the measurement would be good enough for me, but not necessarily a lab grade environment, though, the setup potentially can be improved with expensive optics and special pulse grade laser diodes to be of such a caliber to meet such a scientific measurement down to the picoseconds.

Also, his wiring on the table needs to be tot and straight with the mosfet/battery switch and resistor on the other side as small as possible, parallel and right across from each other with those 2 other resistors in his circuit removed.  Not like what he has in his video with slack and bent circuit.

[Sredni]

What I'm saying is that that transient current can flow in the resistor without giving rise to charge accumulation at its extremes. So you won't be able to 'see' it from the exterior.
The surface charge  induced by the electric field (perturbation) that has propagated from the switch will create localized field lines, that will create a localized flow of current.

When you flip a switch in a 'normal' circuit (battery-switch-resistor), current builds up from the switch due to the electric field lines that form in the conductor when surface charge start to redistribute/recombine. In a way current 'spreads' from the switch to the the rest of the circuit and the build up of charge at the resistor's terminal proceeds gradually from none to complete displacement.
In Derek's case we could have a temporary current in the resistor that will die off (from the EM pulse, so to speak), and then a gradual build up due to the surface charge perturbation travelling on the wires (with the ensuing reflections).

Look at the field in Ben Watson simulation, but do not look at the current plot because as someone else here noted, he computes current by integrating along the whole loop, so that initial step ha contributes from current in the wires where surface charge is propagating along.
(EDIT: actually I just watched it again and he computes the current in selected spots by integrating the magnetic field along a circumference around the cable section, so that current is the average current in that section of wire. So, do look at the current plot as well)

To see this experimentally, I say we need some sort of material that will give off photons when current passes through it (without a threshold) and measure the output optically. My guess is that we will see an initial pulse that will die off and then the steps of multiple reflections.

[EEVblog]

What I'm saying is that that transient current can flow in the resistor without giving rise to charge accumulation at its extremes. So you won't be able to 'see' it from the exterior.
The surface charge  induced by the electric field (perturbation) that has propagated from the switch will create localized field lines, that will create a localized flow of current.

There is just no point thinking about such things IMO. All anyone cares about is the external voltage.

[Sredni]

you are not thinking quadridimensionally, Dave.
the lamp can light up without voltage being seen at the attached scope.
What we are trying to look at here is the buildup before voltage and current behave as we are used to in circuit theory.

[adx]

That's what confused me about what electricity "is" and reason I posted back on page 17. Without an understanding of both the microscopic behavior of 'lectricity, and how that relates to Maxwell's EM, I could be a bit lost. I don't think I am, but experience fills in the gaps, rather than ever having a good understanding at a theoretical level. In other words, I am a bit lost, even though in practice I'm not. I don't "believe" in surface charges, not because I don't think they happen, but because I don't know what it really looks like. Dumbed down diagrams in high school, highfalutin maths at university isn't evidence of fact needed to form what I would loosely call a belief. Anyway, rant off again, but I think this is pretty much universal for engineers, even RF engineers.

This new point is like Mehdi and Walter Lewin's 'disagreement' about voltages in a loop including resistors in a changing magnetic field (I heard of during this thread, haven't watched it all) - in this case current induced in a sense resistor (standing in for the lamp) without dropping voltage (or perhaps vice versa). If such things didn't happen, there'd be no point having more than 0 turns in a transformer, yet there is room in very well-accepted "theory" for arguments to develop, like the Poynting thing not first covered by Derek, but he successfully lit the fuse. I'm kind of in awe at how difficult this simple question has become!

To my mind, making that sense resistor smaller than about 1% of the 1m in question will mean its antenna effects won't have a significant affect on timing, smaller will help a lot with parasitics for measuring timing to even 10% of 1m/c.

If not going the VNA+IFFT or fibre optic transceiver routes, I'd make a moderately high voltage fast step generator, to ease the building of a 1k (or 2k, for 7.2W 120V bulb) load into 50 ohm coax needed for any fast scope or RF gear. Battery powered, self-triggered or a button (string to pull on), with some nod to the way the battery exists in the video's experiment, so that means a fast-closing switch (small GaN fet) even a mercury wetted relay. It doesn't need to be perfectly isolated or balanced, but obviously can't have wires dangling off it or an enormous shielded box.

Sense end I'd probably try a few turns around that amorphous core before giving up and making a simple differential system of centre-tapped 10:1 dividers, ie 500 ohm to 50 ohm with ground in the middle and 2 scope channels, and if that didn't cut the mustard with viewers, then differential amplifier(s) good for a few GHz. Or something equally unpleasant to 'lash up' or buy for one test. Good scopes probably come with 500 ohm probes (there we go; Tektronix P6056). Duplicated for the send end, trigger off either. All checked for behaviour in and out of circuit.

And that's pretty much it - press the button and watch what happens. Only problem is if using a switch (rather than voltage source), the rig will be a very effective antenna, solve with averaging and high trigger level.

[bdunham7]

What we are trying to look at here is the buildup before voltage and current behave as we are used to in circuit theory.

Even if the charges within the light/resistor are just reacting to local EMF induced by an EM wave, they're still moving locally.  For a resistor of size 1cm, over what timeframe could there be internally induced current flow without a corresponding voltage being observable at the terminals?  And is that really what we are looking for in the original question?

[Microdoser]

If veritasium's video was technically wrong, then transformers would not work. If it was correct in the real world, those kooky beamed energy rechargers would work across the room.

Sure, some energy will appear in the wires, and it will appear distance/c seconds from the time the switch is turned on, but useable energy will take the long route.

[adx]

What we are trying to look at here is the buildup before voltage and current behave as we are used to in circuit theory.

Even if the charges within the light/resistor are just reacting to local EMF induced by an EM wave, they're still moving locally.  For a resistor of size 1cm, over what timeframe could there be internally induced current flow without a corresponding voltage being observable at the terminals?  And is that really what we are looking for in the original question?

Maybe. The question is "...how long would it take for the bulb to light up?". Technically that is first light, but it also weakly implies it should stay on ("turn on" would be stronger, but in the context of the question about closing a switch to turn on a light, the expectation is that it stays lit, but it might not, it might pulse multiple times in response to a single clean switch event and the question is vague enough to allow that). The question is over the time taken for that first effect to occur, and more loosely, whether it stays on to any useful degree (to satisfy any expectation of a non-trick question). The context of the question also implies that the time is about the distance travelled, not how long it takes the LEDs or a filament to fire up. The question is thus over how long the electricity takes to get there, perhaps to the part of the lamp that makes light, but the question is (from memory) clear that it is 1m away, in any event the multiguess answer is 1/c not 1m/c, so that answer implies that the "time" might not be 1m.

The spherical wavefront of that first THz emission from the few microns (say 100) horizontally along the switch closure will hit the resistor (or whatever) first at its centre. Femtoseconds later the ends of the resistor will get it (I haven't done the trig, but it's going to be vastly faster than the speed of light in the horizontal direction).

But measuring such a thing, no one is going to be worrying about an effect moving infinitely faster than the speed of light, or a few attoseconds out of some nanoseconds. I'd be satisfied with 10% to accept a multiguess choice.

[EEVblog]

you are not thinking quadridimensionally, Dave.
the lamp can light up without voltage being seen at the attached scope.
What we are trying to look at here is the buildup before voltage and current behave as we are used to in circuit theory.

The "we" is just you!

[Sredni]

you are not thinking quadridimensionally, Dave.
the lamp can light up without voltage being seen at the attached scope.
What we are trying to look at here is the buildup before voltage and current behave as we are used to in circuit theory.

The "we" is just you!

Just look at Ben Watson simulation, here's the screenshot of the currents



He is computing currents in three points of the same loop, and he is getting three different values.
KCL is dead. I would say that current does not behave as we are used to in circuit theory.

But I guess I'll have to add the term "KCLer" to my vocabulary.

[EEVblog]

you are not thinking quadridimensionally, Dave.
the lamp can light up without voltage being seen at the attached scope.
What we are trying to look at here is the buildup before voltage and current behave as we are used to in circuit theory.

The "we" is just you!

Just look at Ben Watson simulation, here's the screenshot of the currents



He is computing currents in three points of the same loop, and he is getting three different values.
KCL is dead. I would say that current does not behave as we are used to in circuit theory.
But I guess I'll have to add the term "KCLer" to my vocabulary.

I don't think you get it, I don't care.
KCL continues to hold for practically every daily use it's put to by every practicing engineer everywhere.
But if you want to faff around the edges of physics, knock yourself out.

[bdunham7]

He is computing currents in three points of the same loop, and he is getting three different values.
KCL is dead. I would say that current does not behave as we are used to in circuit theory.

So what's the problem with that and what's the point?  Did anyone propose using KCL over the whole circuit to predict the short-term behavior at the light in the first few nanoseconds after the switch is turned on? 

[Sredni]

He is computing currents in three points of the same loop, and he is getting three different values.
KCL is dead. I would say that current does not behave as we are used to in circuit theory.

So what's the problem with that and what's the point?  Did anyone propose using KCL over the whole circuit to predict the short-term behavior at the light in the first few nanoseconds after the switch is turned on?

How does your oscilloscope know what happens between its probe points? If KCL is dead, you can have current inside the resistor between the probe points and - say - zero current where the tips are placed (this is an extreme exemplification). How does the internal resistance of the scope develop a voltage that can be 'visualized' if there is no current coming from those nodes? Something might get there through field propagation but again, what is it? Is it a measure of what is going on inside the load resistor or the effect of the field that has propagated through space from the switch? If KCL is dead in the loop under test, it is also reasonable to assume it is dead in the measurement loop.

What is your oscilloscope measuring, really? (And I'm not even considering any loading effect from the input capacitance of the rig)

[bdunham7]

If KCL is dead in the loop under test, it is also reasonable to assume it is dead in the measurement loop.

No, that's not reasonable, it's absurd.  KCL doesn't 'die' here, it just gets a little behind the curve.  It's a transitory thing and that is why I've repeatedly asked you about the timeframe.  What is the permittivity of the conductors in this timeframe?  How large is the measurement loop compared to the test loop?  I would have been more careful than AlphaPhoenix with the test setup, but for the timeframe he was displaying it was fine.  Or at least OK-ish.  Maybe.  KCL doesn't 'work' in the main loop for a microsecond or so due to the propagation speed and self-capacitance and probably other things.  The KCL issues in the measurement loop are going to be in the double-digit picoseconds at most.