Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

[SOLVED] Strange problem with high frequency shunt

Status
Not open for further replies.

lw1ecp

Member level 3
Joined
Oct 28, 2010
Messages
62
Helped
17
Reputation
34
Reaction score
17
Trophy points
1,288
Location
Buenos Aires
Activity points
1,810
I need to measure the current on the power transistors of a resonant induction-heater generator. I know the current towards the load is quite sinusoidal, and expect a half-wave at the drains or sources. If I use a 0.1 ohm shunt built with 18 paralleled resistors, 1.8 ohm each, I get an oscillogram close to the ideal. But if I use a homebrew 0.1 ohm four terminal shunt built with a few millimeters of nichrome wire, the wave is GROSSLY distorted!. The anomaly is not related to self inductance. The same happens with a few centimeters of plain iron wire, with non inductive winding.
What is going wrong?. If this is not known to you, I can post some photos. I am not speaking about ringing at the beginning, nor common-mode coupling into the scope.
Or: what are the commercial alternatives?. The peak current is 30A at 50kHz, I may soon need 100A at 200kHz.
Thanks!
 

I would imagine all those paralleled resistors have a lot more inductance than your piece of nichrome or fencing wire. So, is it possible that what you are seeing with the nichrome is real but the paralleled resistors (with several cm of connecting wire) are changing the conditions to be better? I have used nichrome for shunts and loads with a 50KHz maximum power point solar tracker and fencing wire for a 3KW load (but at lower frequencies) with nothing unexpected. But then, these usages are rather different from yours.

Regds,
Dave
 

davidwkerr, thanks for your hints!. I went back to the thing, replaced the shunt with a current transformer, but sorry, the waveform was very like that with the paralleled (film) resistors.
I thought it could be some nonlinear contact between the Ni-Cr wire and the copper bolt and nut, but about the same defect was shown with an iron wire soldered to the copper wires.
I am suspecting of a time-varying skin effect: in the few microseconds after current begins in the half cycle, current flows in a thin skin and gradually fills more and more wire, does this sound known to anybody?.
 

What makes you think few cm's of uncoiled wire does not have inductance?
 

Many thanks for your replies!. Please take a look at

Here you have the three shunts I tested, plus a simulation with a 100nH parasitic inductance (huge, should be more than 10cm total straight wire) . They were shooted in not exactly the same phase condition but you'll get the idea. It's not the ringing what is ruling out both wire-based shunts, it's the very strange distortion highlighted with "?". (Also please disregard the thickness of the waves, the power stage is fed with unfiltered rectified AC).
The simulation was made with a sine generator plus a diode, but the real photos are drain currents of MOSFETs. The ringing is common-mode ingress into the scope.
I'm pretty sure the distortion is not due to stray inductance, but I can't figure any explanation...
 
Last edited:

Now I am viewing the thread in a different computer and the image is missing. Just in case, its url is
https://obrazki.elektroda.pl/3_1288847879.png
BTW, my suspect of a time-varying skin effect is growing... This would affect less the paralleled resistors just because they are FILM units. Maybe I end up with a Rogowsky coil.
 

At first sight, both "strange" waveforms seem to indicate RF oscillations of the power stage. This would be actually another expectable result of placing an inductive shunt into the source connection of a power stage transistor.

The waveform marked as "O.K." is showing a considerable amount of negative mutual inductance. It's far from resembling the simulated effect of self inductance...

So in my opinion, all three cases indicate, that you apparently underestimated the effects of self and mutual inductance for a 60 kHz measurement. If you study the design of commercial low inductance shunts, you'll notice a design quite different from your attempts.

But I agree, that a making a current transformer from a small ferrite toroid is promising better results. A rogowski coil is basically an elegant tool, but I'm not sure if you'll manage to design the integrator amplifier.

P.S.: I'm not sure about the capabilties of your oscilloscope, but you should be basically able to visualize a RF oscillation hidden in the "strange" waveforms.
 
Last edited:

RF oscillations?. Ha ha, please don't make me remember about the nastiest side of our hobby. Yes, that's what I thought in 1st place, but please believe me (*), if you expand the sweep there isn't much more to be seen (except at the transitions, of course). By contrast, if you keep the scope at 2us/div, but delayed sweep, triggered from mains, you will have a slim wave.
"Far from resembling the simulated effect of self inductance": yes, but because conducted interference is getting in my scope, sorry I didn't bother to clean it before shooting.
Not a Rogowski, but a toroidal current transformer, flat (resistively loaded at secondary) gave almost exactly the same wave as that of the film resistors, that's why I hold their innocence.
FvM, you've made a great point mentioning COMMERCIAL low inductance shunts. Please, could you suggest me some brands offering such an item, or ANs?. I searched a few, e. g. Vishay - manufacturer of discrete semiconductors and passive components but couldn't find considerations relevant to my case.
(*) I'm a 53-old electronics engineer, a homebrewing ham, and have written RF_EXPERIENCIAS con circuitos caseros a spanish site fully devoted to radio experimentation, includig the chapter "Damn Oscillations!". No flaunting intentions!!!, just to make you rest assured I'm not new at the topic.
 

You're showing small, blurred oscilloscope photos. Of course, I can't decide from the photo, if the visible broad bands are caused by RF oscillations, or an unwanted (subharmonic) modulation. Or a different kind of interference caused by your way of probing the current. There are many ways to find the answer, but only if you have hands on the instrument. At least oscillations are a plausible explanation compared to the "time varying skin effect" nonsense. If you are familiar to the ham stuff, you surely know how to detect RF oscillations. Viewing the waveform at a slow timebase can easily reveal a modulation/subharmonic effect.

"Far from resembling the simulated effect of self inductance"
I wasn't referring to the interferences. I simply wanted to point to the fact, that the remarkable clear waveform is showing an inversed polarity of the part modelled as self-inductance. Negative mutual inductance is a simple explanation. But I don't know the complete circuit, it may be the real current waveform (and an inappropriate simulation model).

To make a shunt measurement with low self and mutual inductance, you can either use a small flat resistor with suitable planar wiring, or some kind of tubular resistor with coaxial wiring. The latter is done for high current RF shunts. In both cases, a basic magnetostatic analysis can predict the parasitic inductance. I have used the free FastHenry tool from Fast Field Solvers for the analysis of circuit inductances.
 
  • Like
Reactions: lw1ecp

    lw1ecp

    Points: 2
    Helpful Answer Positive Rating
FvM, thank you for your very detailed answer!. And the mention of FastHenry. Unfortunately I don't have the induction heater at hand to investigate and take more useful photos, but I will do something better: build a simple stage with resistive load and pure squarewave driving. AND, most important of all, pure DC supply, not unfiltered fullwave.
FvM, RCinFLA and davidwkerr: when I have the results, I'll request again your valuable contributions!.
 

AND, most important of all, pure DC supply, not unfiltered fullwave.
Unfiltered supply would be a simple (and satisfying in my opinion) explanation for the waveform of course. Nice, that you finally mention this point. In this case, the only remaining question is, why no amplitude modulation can be seen with the 10x shunt measurement. Either the supply is different, or the CRO trigger.
 

FvM: I did mention the 1st photos had been shot with unfiltered rectified mains, but I must recognize the resemblance of the wide traces to an HF oscillation was very tempting. Also, yes, the wave for the paralleled resistors had been taken with a different trigger setting, that's why it was so clean. I beg your pardon if these far-from-optimal pictures made you waste time in the wrong direction.
Here I am posting the promised better photos.
14_1289195699.jpg

https://obrazki.elektroda.pl/14_1289195699.jpg
These are the currents of both transistors. One, measured with current transformer. The other, with the NiCr shunt, grossly distorted.

https://obrazki.elektroda.pl/12_1289197080.jpg
Here current through the shunt is now a nice square pulse. Supply is pure DC. I don't know if the overshoot on the wire shunt is because of self inductance or time-domain skin effect, but definitely the pack of paralleled resistors gives (again) a "textbook" wave.
Self inductance?. Well, there seems to be a time constant of about 0.4us. At first, I thought this could be produced by a quite realistic 40nH in series with the 0.1ohm. I made this simulation including this L:
38_1289196023.png

https://obrazki.elektroda.pl/38_1289196023.png
This gave much shorter transients (I chose 1MHz to get more meaningful plots), and higher amplitude. In fact, the series R in the time constant is not 0.1 but 0.1+2.5 ohm.
So I stick to the skin effect theory. I found two articles relating to this, at least in the frequency-domain:
- Elimination of the skin effect error in heavy-current shunts
www.haefely.com/pdf/scientific/e1-48.pdf
- Survey of instrumentation and measurement, pages 257 & 258
Survey of instrumentation and ... - Google Libros
Again, I don't care about ringing during turn-off. It is caused by stray Ls together with C drain-source. This does not happen at turn-on because of the Rds(on) short circuiting Cds.
The bottom line: whatever the explanation may be, I choose to use the R pack. I am *almost* convinced about the time-domain skin effect, but if you feel the parasitic L more reasonable, please let's keep on talking!. BTW, the very 1st time I saw the strange waveforms with the NiCr shunt, I thought there was some nonlinear contact effect between the NiCr and copper (such as in galena detectors)...
Many thanks!
 
Last edited:

I think, your "theoretical" rectified sine waveform in post #5 has confused the discussion a lot, together with the unfiltered supply and different triggering conditions. Obviously, the circuit operation is very different from the zero current switching case assumed in the theoretical waveform. But it seems like the current transformer is showing the real current waveform.

I doubt, that you can exactly distinguish between self inductance and skin effect components in the "distorted" waveform, because both are most likely present. But there seems to be a resonant effect in addition. I assume, that the waveform driving the shunt is already modified, to know it for sure, you have to measure the same transistor current with both current transformer and shunt.

The skin effect impedance part can be rather precisely modelled for a resistance wire from parameters given in the text books, so it's influence on the measurement can be estimated for a known resistor geometry. But 0.1 ohm can be easily achieved with metal film resistors, that are effectively free of skin effects in the frequency range of interest. Avoidance of self inductance effects is more difficult and may be impossible without a differential sense amplifier. If no DC measurement capability is required, the current transformer variant seems much more convenient in my opinion.
 
  • Like
Reactions: lw1ecp

    lw1ecp

    Points: 2
    Helpful Answer Positive Rating
Sorry I disn't close the thread earlier. It is solved. It is the skin effect that distorts those HF waveforms. The lesson I learnt is: don't use ferromagnetic materials such as nichrome for HF shunts because their high permeability reduces their skin depth much more than other materials, and as we know this depth is frequency -dependent.
Thank to all the friends that asisted me!
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top