EEVblog #662- How & Why to use Integration on an Oscilloscope

[HighVoltage]

I am using the Integral function very often to measure discharge energy
Voltage input to Channel 1 [Voltage in kV]
Current input to Channel 2 [Current in mA]
Math function one: CH1  * CH2 [Power Watt]
Math function two:  INT (Power) [Energy in mJ]

Once you set the probe correctly, the Energy is shown correctly
in mJ on the oscilloscope measurement output.

This works well on the modern Agilent scope, but the old
FLUKE / Philips scope PM3394B did this already in real time
in the 1990's

Great Video, Thanks

[miguelvp]

I did a little experiment in Excel so I could graph what I was imagining and it works.
I broke up a triangle into little squares under the curve then added them all up and drew a graph.

That is what is shows per cycle if I do a single integral, but the next cycle  goes on top of the other one.
What it's needed is an offset in the triangle wave so that it shows that at a 90 degree phase for half a cycle and then for the other half cycle starting at 270 degrees it produces the inverse.

But when I scaled the single integral of the triangle wave without offset it showed a secondary wave that surely will affect the process above, not sure why that happens yet.

[dentaku]

I did a little experiment in Excel so I could graph what I was imagining and it works.
I broke up a triangle into little squares under the curve then added them all up and drew a graph.

That is what is shows per cycle if I do a single integral, but the next cycle  goes on top of the other one.
What it's needed is an offset in the triangle wave so that it shows that at a 90 degree phase for half a cycle and then for the other half cycle starting at 270 degrees it produces the inverse.

But when I scaled the single integral of the triangle wave without offset it showed a secondary wave that surely will affect the process above, not sure why that happens yet.

I wish I knew more about he match behind it. I just did another of my crude Adobe Illustrator and Excel experiments with a full cycle and this is what came out of it.

[EEVblog]

Didn't expect the level of math nit-pickiness on my latest video, that was rather foolish of me 

[fvdpol]

Thanks Dave for the (as usual) clear explanation.    Some time ago a found on the Jeelabs blog an interesting post where the power consumption of a board was analysed by type of operation. Great to see correlation between the code/logic executed, specific hardware operated and the power consumption. 

http://jeelabs.org/2011/12/30/anatomy-of-a-room-node-transmission/

[HighVoltage]

Didn't expect the level of math nit-pickiness on my latest video, that was rather foolish of me 

I guess, when you use the word "Integration" as a math term in your video description,
all the mathematicians are coming out for offering suggestions and corrections.
 

[EEVblog]

I use Math functions as well, they are very useful too for quick easy measurements. They also help build reports as well. Either Baffle with Bullshit or hope someone else knows what is being displayed.

I'd sometimes put bullshit units into reports to see if anyone notices, like Furlongs per Fortnight 
https://en.wikipedia.org/wiki/FFF_system
No one ever did! 

[dentaku]

Didn't expect the level of math nit-pickiness on my latest video, that was rather foolish of me 

At least it was civilized math nit-pickiness.
I don't really understand the math but at least this video finally made me understand what integration is.

[EEVblog]

I don't really understand the math but at least this video finally made me understand what integration is.

Well, there was hardly any math in this really, it was meant to be a video showing the concept of integration on a scope and a basic usage case for it. If that helped you, then successful video!

[Circlotron]

One of the first things I did with my scope is measure the milliJoules energy output of a number of different car ignition coils. The integration function is a perfect fit for this. Just clamp the high voltage side of the coil to 1700V max with a string of zeners (typical arcing voltage) and integrate volts x current. One really interesting result was after measuring the output of some boxes from a REALLY big name CDI manufacturer. They claim with regard to one particular box "100mJ spark energy" but that is only what their box pumps into the coil primary. Actual coil secondary output with one of their CDI-specific coils was only 38mJ, and with a more conventional coil it was as low as 25mJ. See pics for good conventional ignition output vs CDI under same test conditions.

[dentaku]

I don't really understand the math but at least this video finally made me understand what integration is.

Well, there was hardly any math in this really, it was meant to be a video showing the concept of integration on a scope and a basic usage case for it. If that helped you, then successful video!

By "the math" I mean the stuff other people have posted in the forums, your video actually showed it visually so that's how it helped.

[DrJoe]

Why is the integral wave less than (below) the wave of the input?  If the integral and the pulse begin together, it would seem to me that the integral function would initially follow the pulse curve, then level out as the pulse ended, but it should always be greater than the input and continually rising since it is, after all, the integral of the input (assuming the input is always positive).

Was the integral's scale much, much greater? I can only guess that the scales were different or the integral function actually began before the impulse. Or I'm misunderstanding something...

Very interesting video, BTW. The colors and sound were fine on my computer.

[robrenz]

The vertical scale of the integral is representing area under the curve. Even though the current pulse height is very high it is also very narrow yielding a small area.

[David Hess]

Does anyone know if the quantization noise distribution veers from the ideal for other types of ADCs?
Also, what kind of ADCs are typically found in oscilloscopes?

They used to use predominantly flash converters but now they use folding interpolating or pipeline converters and of course interleaving is common which will alter the characteristics and *that* may change depending on the number of active channels.

If you can fiddle with the interleaving, then the change in the distribution of the noise may be revealing.

[T3sl4co1l]

Why is the integral wave less than (below) the wave of the input?  If the integral and the pulse begin together, it would seem to me that the integral function would initially follow the pulse curve, then level out as the pulse ended, but it should always be greater than the input and continually rising since it is, after all, the integral of the input (assuming the input is always positive).

Integration is always "PLUS A CONSTANT", which for a proper integral (one which spans a defined width, such as what's shown on screen) is technically zero at the start.  It seems this scope adds an additional offset so it remains balanced visually; which is kind of handy, as most integrals veer off to infinity and starting at mid-screen means you only get to see half of anything.

As long as the horizontal axis is indicated, it's a visual effect only, the math is still correct.  (And anyway, measuring a difference with cursors subtracts the "PLUS A CONSTANT" so even if the offset were wrong, it disappears again in that measurement.)

Was the integral's scale much, much greater? I can only guess that the scales were different or the integral function actually began before the impulse. Or I'm misunderstanding something...

The scale is necessarily different, because the units are different.  They're always adjustable, which is definitely handy!

Tim

[T3sl4co1l]

Oh, another wonderful use of integration:

Current sense probes.

If you do the old "scope probe grounded to clip" trick and wave that over a circuit, you're bound to see impulses and ringing.  The ground loop isn't very particular at that size, so what you can do is, wind a small coil (maybe 10-20 turns of fine wire) and fit it in the end of a 6mm brass tube, which has been slotted up one side.  Add a 50 ohm series termination resistor, and run coax up to the scope (with source termination, it's not necessary to terminate the scope end).  Now, the brass tube shields stray E and B fields from your probe, except the ones that are axial with the tip.  Wrap it with tape, and you can wave it around your circuit and sniff the currents flowing through components, traces, even individual vias!

But you're not actually measuring current.  You're measuring EMF.  Which is the derivative of current.  So, you can integrate it, and now instead of impulses, you have steps -- corresponding exactly to the current steps in your circuit!  Of course, it's not calibrated, but knowing the general proportions and how the signal varies with distance, you can get a feel for how much current is flowing in a circuit, and where.

Tim

[rs20]

If you do the old "scope probe grounded to clip" trick and wave that over a circuit, you're bound to see impulses and ringing.  The ground loop isn't very particular at that size, so what you can do is, wind a small coil (maybe 10-20 turns of fine wire) and fit it in the end of a 6mm brass tube, which has been slotted up one side.  Add a 50 ohm series termination resistor, and run coax up to the scope (with source termination, it's not necessary to terminate the scope end).  Now, the brass tube shields stray E and B fields from your probe, except the ones that are axial with the tip.  Wrap it with tape, and you can wave it around your circuit and sniff the currents flowing through components, traces, even individual vias!

Awesome idea! When you say slotted up one side, do you mean a cut all the way down one side so that the tube has a "C" cross section? Is the idea to prevent the tube from acting as its own shorted-out coil and producing its own cancelling EMF? How long would you normally make the tube?

[T3sl4co1l]

Slotted lengthwise near the end, yes.  It doesn't have to be full length.  The closer the end of the slot is to the coil, the more it shields the coil in that direction, which makes it less sensitive, more directive, and pushes the pattern off center (imagine the tube were mitered, instead of slit, to the same depth).

More info: http://www.bcarsten.com/?page=appnote

Tim

[fpliuzzi]

Here's a nice closeup of the tiny coil in the tip of the EMI sniffer probe referred to by T3sl4co1l. This image was extracted from page 55 of AN70 by Linear Technology (Jim Williams). This closeup detail came in handy when I made my EMI sniffer probe a few years ago.

AN70... http://www.linear.com/docs/4159

[MarkL]

Another great app note by Jim Williams - thanks!

Here's another clever use of integration on a scope, this time to measure output load capacitance:

  http://www.edn.com/design/test-and-measurement/4370468/Oscilloscope-math-functions-aid-circuit-analysis

This was done on a Tek DPO3034 which has generalized, software-driven math functions (like many Tek scopes).  This particular example can't be done on the Agilent MSOX because it needs a second computation after the integration, which isn't a supported combination on the MegaZoom ASIC.

It seems this scope adds an additional offset so it remains balanced visually; which is kind of handy, as most integrals veer off to infinity and starting at mid-screen means you only get to see half of anything.

(Referring to the Agilent MSOX...)  It does.  There's an adjustable constant that pops up on the Math menu when integration is selected.  Dave didn't show that or maybe he didn't have to do anything since the input offset was already close enough to zero.

[KedasProbe]

Calculating Power Value in the selected time range.
http://blog.hameg.com/?p=1799#more-1799

[SNGLinks]

Pity my Agilent DSO-X 2024A doesn't have integration. At least I can't find it in the Math functions.

[rs20]

Pity my Agilent DSO-X 2024A doesn't have integration. At least I can't find it in the Math functions.

That's surprising, but I've had a quick look through the manual for the series and I concur that it doesn't seem to be there. However, the average of a waveform is just its integral divided by its width on the screen. So if you position your signal of interest to be the only thing on the screen, then measure the full-screen-width average (page 176 of the manual) and multiply it by the width of the screen, you'll get exactly the same result.

[SNGLinks]

Thanks for that info. I was surprised about the missing feature. It's all done in software so would be fairly cheap to implement. 

[Digital Corpus]

I'm happy to see another video from you Dave. As I'm working on a measurement project for automotive use, I've needed to quantify some signals coming out of the cars electronics, specifically some old diesel fuel injection ones. In the some of the first commercial direct inject diesel applications, primitive monitoring of fuel injection is required for to maintain engine speed and operation. This was done on one fuel injector and the signal lets the engine control unit know what's going on (clicky if you want to ready more). Well, no one outside of the manufacturers ever really hooked up an oscilloscope to see what this signal looked like, so I decided to.



The cursors are setup at the beginning and end of injection and the voltage is a function of the movement of the needle. In order to figure out its physical location, an integral needs to be taken. I exported the on-screen data plot into a spreadsheet, I ran the integral manually to compensate for the DC offset also so I wouldn't just idle my car forever in a day.



I don't have applicable units, but knowing other parameters of the injector such as fuel pressure, opening pressures, closing pressures, nozzle orifice size, a pretty accurate estimate as to the amount of fuel injected is able to be calculated from this reading. The fuel is also used as a lubricant so there is more fuel that flow through the injector than what is actually injected as well.