What is the advantage of using a differential probe for power calculation?

[EE4life]

Hi all,

I found online that most professional websites suggest using differential probes to calculate power, along with a current probe, for a DUT. What advantage does the diff probe give us for power calculation?

Thanks

 

[Windadct]

Generally for isolation - most scopes and instrument have the shield of the leads grounded, so connect to a Mains connected source, or to two different locations in a circuit and you just bought a new scope. The Differential function is needed because once you remove the ground reference you need to redefine what you are measuring - the difference between two points in a circuit. There are some ways to cheat in come cases, v divider and maybe an opto isolator - but nothing works as universally as a Differential Probe - they are expensive esp when designed for power.

 

[jim hardy]

Surely things are different nowadays?

In my day oscilloscopes had their case tied to Earth , via the third prong on the power cord.
So when somebody tried to measure the voltage across a circuit element that was not at Earth potential, like in a power circuit,
the instant he hooked the 'scope probe's little "ground" clip to the DUT he got a big flash, a blown fuse, a wrecked 'scope probe and maybe a ruined 'scope too. People learn fast not to do that. It happens because the 'scope has internal connections from probe common to chassis and to earth. Most power circuits won't like being connected to Earth where they're not expecting it.

It is important to have basic knowledge of what is inside your test equipment.Portable battery powered 'scopes can relieve the Earth situation, but check in the manual and verify with an ohm-meter to see whether their two channels are separated from one another.

 

[EE4life]

Thank you for those notes. We know that the way a circuit is setup (shielding, ground paths etc), the phase measurement suffers quite a bit, but the amplitude is not affected near as much. I am having trouble measuring phase (to measure power) using a passive voltage probe.

In a simple situation, where there is no other components between ground and the DUT, do we benefit from using a differential probe or is the passive probe equivalent (for phase)? I know theoretically they are equivalent, but given that the path to ground may produce nonlinear delay errors, perhaps a differential probe is best for the voltage.

 

[jim hardy]

do we benefit from using a differential probe or is the passive probe equivalent (for phase)?

Are we speaking of power line frequency, or much higher?

Reason i ask is i can't imagine 'scope delay being even detectable at power line frequency.

So i probably mis-understood the thrust of your question.
On power mains stuff i use 'scope's "Invert" and "Add" functions to get voltage.. which amounts to differential measurement.

Are you measuring radio frequency?

 

[EE4life]

I am not measuring radio frequency, rather I am using 1kHz to 100kHz range. I measured the phase lag from the current probe and voltage probe at different frequencies and for a noninductive resistor and other tests for very low loss capacitors. I get phase lag in each case, and it is not consistent, ie account for a delay via linear regression does not fix the problem. Also, the phase lag is different for different resistor values. By decreasing resistance, phase lag goes up, especially lower than ~200 ohmns. I am not sure how to verify if and when my phase are correct. Any help is appreciated. I am measuring a resonance type circuit (LCR) and an antiresonance type.

 

[jim hardy]

In a simple situation, where there is no other components between ground and the DUT, do we benefit from using a differential probe or is the passive probe equivalent (for phase)? I know theoretically they are equivalent,

i too wouldn't expect any troubles from ordinary 'scope voltage probes at the frequency you mention.

Without an idea of your test setup it's kind of a guessing game at this point...
Sometimes it's as simple as setting 'scope sweep to 'chop' not 'alternate' and trigger to whichever channel has voltage probe.

What's the frequency response of your current probe? That's where i'd look first. What type is it? Just a current transformer, or active electronic with hall sensor and high frequency compensation?
Can you place sweep in X-Y to make a lissajous pattern, voltage across noninductive resistor vs current probe output?? That should show whether current probe has phase shift.Might be you'll have to use a 1 ohm or so noninductive resistor to sense current.

 

[EE4life]

I have a Tektronix current probe TCP305A w/ TCPA300 amplifier and a standard 10x voltage probe. After speaking with a friend about proper grounding practices, I was able to keep my phase error between -0.5 and 0.5 degrees between 1kHz and 100kHz. On an impedance analyzer(extremely accurate), my resistor has phase of 0.2degrees in this bandwidth.

My goal is to get within 0.1 degrees phase difference from what I measure on the impedance analyzer, with little jitter. From my analysis so far, I have some thoughts
1) Impedance analyzers have extremely stable input signals. Function generators do too, but it is not good enough.
2) The ground loops (ground reference will be slightly different in the circuit,at the oscilloscope, and the function generator due to cables), therefore phase cannot be measured so accurately because the probes will always disturb circuit
3) I have seen that for slightly different grounding (changing the order/position of the probe ground clips, and the input ground clips) the phase can be altered as much as 0.3 degrees or more. At this rate, it is no practical to believe the oscilloscope, because small changes can really affect the measurement of phase. This is not really a problem for a resistor, however, small phase error (eg -90.4 phase vs -89.4 phase) for a capacitor can be the different between positive and negative power!

I think I need to try the differential probe because grounding seems to really affect the phase measurement, although theoretically I can design the experiment such to avoid the issue.

 

[jim hardy]

Great current probe, i used its predecessor in early 1970's.

Its 19nsec delay is 0.19 parts per thousand at 100khz, 0.06 degree if my arithmetic is right.

I think I need to try the differential probe...

If you have one handy,, of course.
Will your 'scope invert and add ?

Sounds like a very interesting experiment.

I have seen that for slightly different grounding (changing the order/position of the probe ground clips, and the input ground clips) the phase can be altered as much as 0.3 degrees or more. At this rate, it is no practical to believe the oscilloscope, because small changes can really affect the measurement of phase.

http://www.edn.com/design/analog/43...al-chips---Part-1--Principles-of-current-flow

however, small phase error (eg -90.4 phase vs -89.4 phase) for a capacitor can be the different between positive and negative power!

Is that what you're measuring, tanδ ?

old jim

 

[EE4life]

I tried a differential voltage probe, it did not fix the problem. I know that my problem is not delay because the delay times vary with frequency. The phase error has a strong dependence on both the impedance of the DUT and the frequency. Also, the error is way beyond reasonable delay differences.

Take a look at some data for frequency sweep using different precision non-inductive type resistors. The y-axis is phase and x-axis is frequency. The setup is red wire, then,voltage probe, then resistor, then current probe, then ground (ground for the signal, voltage probe, and current probe). I get parabolic shape type trends when connecting the current probe before the resistor.

The curves below show a trend. If I plot the slopes of the linear portions, I get a decent looking prediction of the slope with a power regression for the slope versus impedance.

I have not seen anyone measuring phase with 0.2 degrees accuracy. I am not measuring tan delta, but my DUT reaches phase of near 90 deg and minus 90 degrees with a large range of impedance (100kΩ to 10Ω) depending on the working frequency.

Is it possible to measure the phase with 0.2 degree absolute error using an oscilloscope?

Right now, my best guess to tackle this phase error issue is to do extensive calibration by measuring the phase error data for a Pot resistor. By varying the pot resistance and doing frequency sweeps, I can get the phase error for a given frequency and impedance. Then I can adjust the phase according to my calibration tests.

I feel that this is too much calibration without a scientific method behind it, but I am out of methods.

,

 

[jim hardy]

Usually it's something simple...

The setup is red wire, then,voltage probe, then resistor, then current probe, then ground (ground for the signal, voltage probe, and current probe).

If i understand that sentence, your voltage probe will feel the insertion loss of the current probe. Its miniscule Z adds to R.
Current Probe MUST have inductance or it couldn't make any flux to measure.
Try moving current probe ABOVE (ahead of?) voltage probe, so its impedance X_L is part of the source not the load..HP current probes of early 1970's used hall effect for low frequency and transformer action for HF , so the probe had minute insertion loss.
I think they used a current feedback winding to keep flux low, so the probe's impedance likely involves some f(t).

I noticed HP does mention insertion loss for your current probe but I don't remember the number.
Anyhow, move Z_currentprobe out of your measurement loop and see if anything changes. That'll only take a minute.

I could be way out in twilight zone, often am.. no harm intended

old jim

to summarize
Inserting the current probe adds a one turn inductor to your circuit, regardless where in the circuit you put it.
So put it someplace your voltage probe won't see it.

 

[jim hardy]

ps this is interesting. Please keep us posted ?

 

[EE4life]

I measured the phase vs. frequency for the current probe before the voltage probe.
https://www.physicsforums.com/file:///C:\Users\Husain\AppData\Local\Temp\msohtmlclip1\01\clip_image001.jpg
The curves going from down to up, go in decreasing resistance values (all noninductive precision resistors). The orange dots is for 10ohmn.

It seems that putting the current probe before the voltage probe limits the error seen for low impedance (less than 1 degree) and putting it after the resistor (previous post) limits the error for large impedance.

Another experimental point that I need to mention is that I used 10 windings for the current probe to get a stable current signal for the high impedance values. I did all the experimental with the 10 windings. I did an experiment using comparing a 10ohm with and without 10 wingdings and the values were within 100 milli degrees, so I believe the windings are not the main source of this error.

I also did the same experiments with a 100x probe; nothing to special happened, but slightly different trends were observed

My conclusions/suggestions/thoughts:
Measuring phase within 200milligrees is sort of a dream for an oscilloscope, and I have a sort of new one, agilent MSO6014a. If you would like to get within 1 degree accuracy, the right probe will get you there. However, if you would like accuracy within 100milligrees, you need extensive calibration data of your setup. This means measuring the frequency response of non-inductive resistors for several different values. Then, you can calculate the delay times (as a function of frequency) for each impedance. For a given frequency, fit a logrithmtic regression (excel does this) to the delay vs. impedance. The regression looks like A*ln(frequency)+B=delay.
I don't think the solution lies completely in proper probing, but we need to use a calibration technique, like the one I suggested.

Some notes/suggestions to my fellow phase measurement with oscilloscope peoples:
Also, to calculate the phase, do not use the difference in the rising edges. Rather do a power spectrum calculation (involves FFT, labview has this VI) with the voltage and current and extract the power at the frequency you are working at. A pure power calculation is plagued with DC offset error from the original signals, especially current (ac coupling does not completely eliminate it). The oscilloscope phase is quite jumpy (+/- 1 degree), so it is useless to make a tan delta calculation. Also, you oscilloscope should have a decent resolution, not the old kind we know and love (and is cheap :).

 

[jim hardy]

i assume you figured out your trigger and alt/chopped sweep.

Also, an analog scope might be better at this than a sampling 'scope.
Analog scope doesn't take time out to think.

old jim

 

[EE4life]

I have made some changes to my setup and I am receiving better results. I will share my observations once I am ready.

 

[EE4life]

I have made some changes to my setup and I am receiving better results. I will share my observations once I am ready.

Okay, here the grand finale.
An oscilloscope is not the best tool the measure phase. But here is the best setup using it
1. Measure current with a non inductive resistor (test the phase of it on an LCR meter!)
2. Measure total voltage and current using the same model voltage probe. This will help to eliminate any delay differences between the probes
3. Use a vernier scope (highly adjustable time base). Make sure to get exactly a whole number of cycles on the screen. Take the data to your computer. Minus CH1 from CH2 to get the voltage over the sample. CH2 is the current, in phase. Use FFT analyzer to get the phase at the frequency that you are driving at. Minus the phase of voltage and current to get the phase between them.

I also found that the level of the impedance you are measuring effects the phase measurment due to interaction with the probe impedance. Amplitude is very accurately measured, however, phase measurements are strongly perturbed by probe impedance interactions.

The fact about FFT part is critical. Among other reseasons, there is always sometype of bias on your waveform, even with AC coupling. It does not matter too much when the phase is low, but with a near 90 phase a small change in phase due to DC offset could mean negative power!
Also getting exactly getting a whole number of cycles on a screen, for it increases the phase resolution dramatically when you are talking about 100milidegrees resolution.

LabVIEW has good FFT analysis methods and I would recommend using the "estimate main frequency component and phase" option, or its called something like that. Please post if there are questions. Sorry for being a little vague, but I hope it provide some good info.