Measuring the same signal in different ways, yields different results

[eTobey]

Hi,

i am trying to figure out, how a current flows through a trace. I tried measuring it first with a probe on each end (A and B), and the GND of the probes to the GND (C) of the circuit, and subtracting the measured values with the math function (F2 = C4-C3). The circuit is connected via a powersupply.

The other way of measuring it was done by connecting GND of the probe to "A" and the probe on the other end "B" of the trace.

I thought those signals should nearly look the same, but they dont really.

Anyone can explain this, or give me some suggestions/terms i could google for? (Note, that the timescale is not the same in both pictures.)

[TimFox]

I don't understand:  you are measuring the voltage across a wire (trace)?
Also, "C" in your drawing is grounded.  Is there any connection between that "ground" and the 'scope ground?

[golden_labels]

With us not knowing how exactly the testing rig looks like, we may only guess.

My guess would be: there is no testing rig at all, not one suitable for the signals considered, and probes are just connected to traces. Which means you built yourself some antennas from wire loops. Scope displays whatever those antennas pick up, likely bandwidth limited, affected by transmission line effects and various kinds of noise.

If you wished to repeat AlphaPhoenix’s demos^(1)(2)(3) on signal propagation in wires, note his traces are very long to avoid get around how fast causality propagates, he spent much more time than it’s apparent from the videos on making it all work, has some decent experience, and still is on the tip of what he can do.

Unfortunately my experience in this area ends with knowing what doesn’t work. So beyond telling you, that the setup seems not suitable for the task and you just see garbage, I can’t give much advice. To see, what I meant by “testing rig,” you may see the example from w2aew.⁽⁴⁾ But even best practices will not let you ignore physics: the signal propagates at foot per nanosecond speeds.

⁽¹⁾ “Watch electricity hit a fork in the road at half a billion frames per second”:

⁽²⁾ “I bought 1000 meters of wire to settle a physics debate”:

⁽³⁾ “Why does WATER change the speed of electricity?”:

⁽⁴⁾ “#111: How to make a high performance oscilloscope probe socket”:

[eTobey]

I don't understand:  you are measuring the voltage across a wire (trace)?
Also, "C" in your drawing is grounded.  Is there any connection between that "ground" and the 'scope ground?

Yes, that is one way to measure the current. (With a little bit more work)

That ground in the schematic is the circuit ground. There may be only a connection through: circuit -> supply -> scope.

...

I did the one-probe-measure with a ground spring. That shouldnt pick up much?

And when i measure 2 times some noise (at the same time), and i subtract the values, i should get almost no noise in the result?

[eTobey]

I have now found one issue:
The following picture, shows a difference (F2 - red) from -0.8ms to 1ms. But this is due to the analog frontend (if this terminology is right). I found it out, when i put both probes on one side of the pcb trace to measure, which showed the same signal.



After adjusting the levels of C3 and C4, the trace is almost as one would expect it (0V).



After doing this, i got a trace (white trace with the spike = F2) that i would consiber plausible. For switching on the highside mosfet (detail of trace from above at -0.8ms):



And for switching off the highside mosfet (white trace with the spike = F2) (detail of trace from above at 0ms):

[eTobey]

Pictures of the setup:

[eTobey]

I have made another comparision:

Using a lead (F4) vs. using the gnd spring (F1) makes almost no difference in this case (see attachment for setup):



But seemingly on another scale, but i doubt that. I think it has to do with the frontend issue (https://www.eevblog.com/forum/beginners/probe-compensated-but-level-of-square-signal-gt-20ms-not-looking-good/msg5519191/#msg5519191):

[eTobey]

I think i found the reason for this:
This seems to be a common mode noise issue. I put a ferrit on the cables (wound 2 x) after the power supply. And the same effect (half of it) is achieved by putting the ferrite in to the mains cable.

I dont know where it precisely comes from.
- Is it generated in my circtuit?
- Is it generated from in the powersupply because of the sudden current draw? (the voltage does drop)

[inse]

What exactly are you expecting to happen across this trace?
The only current flowing will be the tiny gate charge during turn on and turn off.
The trace will be in the single or sub mOhm range and the inductance in nH‘s.
So the signal you are searching for is in the sub mV range and will disappear in the noise floor of your scope.
With the vertical setting of 5V/div you get ca. 150mV per bit (assuming an 8 bit scope).
So think yourself what you are able to achieve.
Above that, the node will swing between supply and ground, introducing common mode issues to your measurement.

[eTobey]

What exactly are you expecting to happen across this trace?

Current flow.

The trace will be in the single or sub mOhm range and the inductance in nH‘s.

You sound pretty certain there, how come? I have measurend with a 1A current about 20mV, that is 20mOhm. Calculated i got 16mOhm.

Therefore

So the signal you are searching for is in the sub mV range and will disappear in the noise floor of your scope.

is wrong.

So think yourself what you are able to achieve.

Of course i think. A lot actually.

Above that, the node will swing between supply and ground, introducing common mode issues to your measurement.

Wrong:
It does go below GND look at the inductor and the diode in the lower mosfet and think for yourself.

[inse]

I admit I always try to sound convincing no matter what kind of BS I spread 😁
But for my vindication, you didnt tell you measured the trace resistance before (must be rather thin) and that you own a HD scope.
Nevertheless this job is what differential probes were invented for.
Just do the math of your scopes resolution vs expected signal amplitude.
An additional cross check for the validity of your setup: calculate the area of the gate current during charge and discharge phase and compare them to one another and the MOSFET parameter from the datasheet.

[eTobey]

But for my vindication, you didnt tell you measured the trace resistance before (must be rather thin) and that you own a HD scope.

If i didnt give you numbers, then there is no point in giving me random numbers.
The trace is actually rather thick, its 2OZ copper. 

Nevertheless this job is what differential probes were invented for.

Usable probes for this task would propably cost a couble of grand i guess...

An additional cross check for the validity of your setup: calculate the area of the gate current during charge and discharge phase and compare them to one another and the MOSFET parameter from the datasheet.

Good idea, but i am not certain yet, if that signal is good enough for this.

BUT there is no parameter that can be compared. Mosfets are fairly complex in this regard. Actually i am more interested in (rough) current flow and timing.

[radiolistener]

That ground in the schematic is the circuit ground. There may be only a connection through: circuit -> supply -> scope.

The ground connection must be eliminated to prevent the formation of a large ground loop. Connecting the oscilloscope ground to points A or B under these conditions may damage both the device under test (DUT) and the oscilloscope. Additionally, it poses a risk of severe injury, including serious burns, or even fatal electric shock and death.

For such kind of measurements, make sure that either your oscilloscope or the test circuit is powered by a source with complete galvanic isolation.

For example, you can power your oscilloscope using batteries if the DUT is powered from the mains, or conversely, power the DUT with batteries if the oscilloscope is connected to mains power. 

But under no circumstances should both the oscilloscope and DUT be powered simultaneously from the mains.

[inse]

The gate charge should be published in the datasheet, it gives you a ballpark figure to validate your measurement against.
What peak current did you determine so far?
The most capable gate drivers I came across would drive 4A, but that is only necessary for either huge gate capacitance or hard switching. Usually the gate drivers have been tamed by gate resistors.
Which also makes an ideal aid to measure the gate current.

[CaptDon]

Was the ohmic measurement of the trace done with a 4-wire milliohm meter? your 1 amp measured / calculated observation still seems to be off by what I expect to be a factor of 3. Why is this measurement so important? It could be calculated probably within a 10% error margin based on driver impedance and gate to source capacitance plus a bit of Miller effect. Your setup so far hasn't yielded believable provable information that agrees with calculations by others. I could get close to accurate provable numbers at my former workplace using a Tek scope that cost over $100,000 but sadly, I would have to examine the really ungodly expensive differential probes as I believe they can't handle voltages above 10 volts without damage and the wiring needed to connect the probe to each end of the trace would still be suspect for error that could only be as accurate as doing simple calculations.

[eTobey]

But under no circumstances should both the oscilloscope and DUT be powered simultaneously from the mains.

It depends....

The gate charge should be published in the datasheet, it gives you a ballpark figure to validate your measurement against.
What peak current did you determine so far?
Usually the gate drivers have been tamed by gate resistors.
Which also makes an ideal aid to measure the gate current.

The gate charge highly depends on Vds, which at the moment is not even constant. I am still investigating and measuring. Found values from 0.4A to 3.4A, which have to be validated. Things dont add up yet. I think i also run the chip in unusual conditions (high/low times > 1 ms), that is causing some abnormalities. Also i may have fried that one already, after forgetting another gnd clip. 
I got the 6EDL7141... no gate resistors.

Was the ohmic measurement of the trace done with a 4-wire milliohm meter?

Yes 4 wires were used. And as it is close to the calculated value, there shouldnt be any doubt.

Your setup so far hasn't yielded believable provable information that agrees with calculations by others.

What others calculations exactly?

[inse]

Why do you give us the details only step-by-step?
What peak gate current did your measurement show (white trace in one of the screen shots)?

[eTobey]

Why do you give us the details only step-by-step?
What peak gate current did your measurement show (white trace in one of the screen shots)?

It was never my intention to give you any numbers. I just wanted to know what problem there was, and later where exactly it arose.

Didnt i already gave you some (unverified) numbers?

That microcontroller now gives me problems... cant measure anything. It only works when connected to the PC... 

[eTobey]

Found the problem:
Eliminated any connections / cables, that could interefere.... hifi, Audio and such. Connected somewhere or not. Had a input on a cable "fake" floating. Without the cable it actually started floating, and thus creating controller misbehaviour on my external circuit. 

Now to some more measurments:
Tried 3 measures on "B": One with a shortet probetip, one with a fake ground spring, and the other with a real one (the actual measurement).

First measure for unwanted noise with a shorted probe (The wrong channel is selected, but it has the same scale):


Next measure with a "fake" ground spring, that is actually not connected, but forms a similar loop (note the scale change):


Final measurement: But is it valid? That increasing current (negative dip - up to about 0.5A)  is rather strange. Also i have only setup the gate current to 20mA. That spike is about 2A.

[eTobey]

Had some more time with this. Measured, without actually measuring the trace, with a strip of foil, and other probe tip constructions. Got a few "aha" moments.

Spikes come partially from the inductive coupling of the pcb trace that supplys the fets (measured with foil as "fake trace"). The long dip  (yellow trace from -0.8ms to 0ms) is only slightly influenced by the magnetic field of the supply current of the FETs.

Foil measurment:
You can see the negative deviation from 0V with the foil measure (orange trace) at -800us. This deviation becomes 0, where the current change goes to nearly 0 (-50us towards 0us).


Still there is this 0.5A current, which i think has to do with the dropping of the voltage of point "B", and thus pulling out some current from the driver.

The compensated trace (yellow) is the actual trace measurement minus the measured trace wtih the foil, thus eliminating magnetic interferences.

EDIT:
After a lot of testing, i realized, that its not as easy as it seems. Using an open loop, to compensate for the noise, will not really work, as the actual trace is on another layer, and has also a different shape. Also any piece of conductive material will alter the magnetic field too.

I am still not sure if i can trust the long dip. Measurements will continue in other ways.