Non-inductive resistor that isn't

[Lewish]

Here's a puzzler for you guys. It has me stumped at the moment.

I have a prototype circuit, real pcb, on which I have a Caddock MP2060 non-inductive resistor of 0.05 ohms. I was trying to use it for current sensing. However, the voltage produced across it doesn't match the current measured with either of two different types of Tektronix current probes. I believe the waveform, as displayed from the Tek probes, based on an analysis of the circuit involved.

Using one of HP's best LCR meters indicates an inductance of 35 nano-henries and a capacitance of 5 pF.

The real current waveform should look like a sawtooth. It does from the Tek probes. The voltage waveform looks like the lower 20% and the higher 20% of the waveform is a sinewave. So, only about 60% of the waveform is correct.

The switching frequency of the current is 20 KHz.

For 2 points, tell me what is going on here.

 

[buzzp]

Have you considered the properties of your probes? Also, what about the properties of the traces leading to the part? Are you measuring the current before or after the part?

 

[Lewish]

Hi buzzp,
Good questions, but yes, we have considered those points. The pcb traces to and from the resistor are almost an inch wide. We are using the Tek current probe on a component lead at the end of one of the traces. The current goes thru the component, thru its lead, thru the sense resistor.
The trace is carrying about 30Amps.

 

[buzzp]

One test I would do is bypass the board and prove this is not a problem. Is there a ground plane on the layer below the traces? This can add capacitance, especially with 1" wide traces. What about the 'loop' the current makes? Where is the return line for all of this current?

I am a little unclear as to the voltage waveform seen. Let me repeat it to make sure: the top 20% and lower 20% looks like a sine wave and the rest follows the sawtooth wave of the current probe? This is a little hard to picture. Is the amplitude the same (using the current and resistance)? There must be some phase shift between the voltage and current as well? One other question, what is the load?

 

[walker1]

My first test would be to check if the probes are correct adjusted.

The bottom and the top of a sawtooth waveform are where you need the best performance,
and few nH and pF in your circuit should not give that kind of destortion at only 20 kHz and 0.05 Ohms.


A long time back I had a very similar problem, where the probes to a scope
were way out of tune.
With that corrected, my circuit turned out to be working fine. (as it had all the time )

 

[Lewish]

OK, one new piece to the puzzle. I just tried a magneto-restrictive device in place of the resistor, and got the correct current waveform on the O'scope. The same scope probe as used across the resistor. Hmmm????

To buzzp, yes the amplitude is close to correct. There is about 0.5us of phase shift, but it is hard to tell if it is the resistor or something else.

All ideas are appreciated.

 

[buzzp]

If I imagine the waveform correctly, there is some added capacitance in the circuit that is not letting the voltage change at the rate of the current. I would still try to bypass the board and see what happens. When you are looking at the voltage waveform are you measuring across the component (I assume so)? What happens if you reference the grnd of the voltage probe to system ground and use two channels and subtract the signals on the scope?
I think there is a combination of capacitances (series and parallel) that cause problems with the measurement. This could be the combination of the trace capacitance, probe, and the sense resistor itself.

 

[boffin2000]

This may be a bit late for your puzzle about the current in the non-inductive resistor but have you considered that the TEK cuurent probe may be reading incorrectly because the waveform you are measuring is not a sinewave or a squarewave/pulse. Any form factor other than a sinewave or a pulse may not give the correct answer ... the current is a function of the area under half of a symmetrical waveform so unless wavefrom is symmetrical and the TEK probe knows the wavefrom correction factor you will always get the wrong answer!

 

[Lewish]

Hi boffin2000, that is a good thought, but since both types of Tek probes are rated to measure current to frequencies in excess of 100MHz, and since the current in question is only 25KHz, I think I can safely rule out waveform error.

We solved the problem by changing from the Caddock resistor to a Amploc brand hall-effect current sensor.

BTW, the spice simulation model shows we should have a sawtooth waveform.

 

[electricuwe]

Unfortunatly I can`t imagine the waveform you are getting from the resistor, but generally there might be two reasons for having problems with that circuit:

Common mode voltage:
You can check this by connecting the tip as well as the ground of the probe to the same lead of the resistor. If you see a voltage in that configuration a common mode voltage is driving a current on probe cable shield. Use a 1:1 probe to reduce the effect of this problem.

Connection of sensing leads:
The sensing leads ahve to be connected directly to the resistor leads (Kelvin connection), some types of resistors provide the Kelvin connection internally, and you have to avoid voltages beeing induced into the sensing leads by proper layout.

Solving the problems that way is very difficult in the current range you are working in. Generally the best way is to:

avoid common mode voltages by usning a sensor with ptential separation

using higher voltages for the current signal

If you are intrested to solve the problem without the cost associated with the hall-effect current sensor there are two additional hints:

A low cost current transformer can be used on pulsating DC current if there is gap where no current flows (Used in lots of switchmode power supplies)

The following book provides lots of information for measuring current in power electronic applications (mainly for laboratory use, but also applicable to a product design):

Lappe,Rudolf; Fischer, Friedrich: Leistungselektronik-Meßtechnik, 2. Aufl.; 1993;Verlag Technik; Berlin, München, ISBN 3-341-01068-8 (in German)

 

[Lewish]

Hi electricuwe, it has been over 30 years since I took technical german in college, so I am not up to translating the book. But thanks for the reference.
In the circuit, I have tried to do most of what you mention. One pin of the Caddock resistor is at ground. That makes it easy to probe the voltage across it with a voltage probe on a scope. It is Kelvin connected to the input of an opamp for current sensing. The non-ground lead is also guarded. There is massive ground plane in the area, so there is no ground movement that I can detect. I do not detect any common mode voltage across the resistor.
At this point, I have concluded that the resistor, being a thin-film power device, is capacitively coupling to the heatsink it is attached to, and that capacitance is distorting the voltage waveform.
A current transformer will not work in this particular case, as the AC is riding on top of DC and I need to know both terms. At 20,000+ pieces a year, the device we chose is quite cost effective.
Any other thoughts are welcome.

 

[electricuwe]

You might be on the right track if your heatsink isn't at ground.

 

[a31ford]

I Aggree with the heatsink not grounded thing, I remember a massive D.C. powersupply, that would go into a ramp wave when turned off, the cause was a nongrounded heatsink on a cap. discharge resistor. (hehe, the ground was there from mfg. but I would assume the last tech to work on it, didn't reinstall it, as it was covered in tape when we got it).