feature request for complex presentation of impedance measurement

[jschwender]

Dear John,
It would be nice to have to have impedance results presented in complex numbers. I think it should be quite easy to implement, as the impedance data is available in polar coordinates ( magnitude and phase over freq.) and can easily be calculated to complex numbers, (real part and imaginary part over freq.) . Right now i do that by exporting the impedance data, and feed it into gnuplot, which then does the calculation and plotting. But it would be comfortable to have that within REW. In my opinion that would be a real added value.

 

[jschwender]

hm, no interest, or is something unclear?

 

[John Mulcahy]

I'm not clear what you are looking for, do you mean a Nyquist plot or separate plots of the real and imaginary parts of the impedance versus frequency? Maybe an image of the gnuplot output would clarify. What does the plot of the complex data help you do?

 

[jschwender]

ok, the idea is: instead of magnitude + phase in a plot, have a plot with imaginary part, real part and Z. This way you can clearly see how much the real part increases with frequency, in this case due to skin effect. That is something that you can't see in a mag+phase plot. Good for analyzing components. With coils you can also see proximity + skin effect. Good for optimizing coil geometry and wire size.

 

[John Mulcahy]

Is it the real part that is of interest or is there a need also for the imaginary part? Imaginary would be awkward since it can be negative.

 

[sm52]

jschwender
From your graph, it appears that the skin effect appears at frequencies from 10 kHz. Is that so, or does the graph say otherwise?

 

[jschwender]

Actually, all three parts are of interest. this is not obvious in this example, but also the deviation of imaginary part from ideal model can be of interest. Is negative impossible because of log y-scale? If negative values are inconvenient, how about taking the abs() as a stopgap measure? Or just cut off negative values at all. Capacities are much closer to ideal model than coils, therefore it is more likely to investigate on positive img part. If it is all too cumbersome: even a plot with only real part, that is better than nothing and helps already.

 

[jschwender]

jschwender
From your graph, it appears that the skin effect appears at frequencies from 10 kHz. Is that so, or does the graph say otherwise?

This is an actual measurement of a loudspeaker cable. It is reality. In this case it is mainly skin-effect.

 

[John Mulcahy]

I'll add a plot of resistance.

 

[jschwender]

Thank you very much i appreciate it!

 

[sm52]

This is an actual measurement of a loudspeaker cable. It is reality. In this case it is mainly skin-effect.

a plot with imaginary part, real part and Z.

Can you show in the impedance plots of two cables, one thicker and the other less, why is it useful to see the skin effect, how can this be used?

 

[jschwender]

Can you show in the impedance plots of two cables, one thicker and the other less, why is it useful to see the skin effect, how can this be used?

you may read this: https://www.schwender-beyer.de/artikel.php?k=4&a=13
i hope the google translation makes a good job.... measurement data always helped when it came to assessing bold marketing claims, for example from cable manufacturers.

 

[sm52]

I have read the information on your site. My opinion in some way agrees with yours, in some way it does not. I made a measurement of the impedance of two branches (plus and minus) of a 5m cable from Audioquest. Each has two cores, the diameter of one is 1 mm, the second is 0.6 mm. This is a plus. The minus is slightly thicker. Can you say this is a good cable?

 

[jschwender]

If i got it right, this is a 4-wire cable, a 1mm and a 0.6 mm wire are connected on both ends to combine in parallel to a plus wire, and two thicker ones are combined in parallel to a minus wire? What did you measure here, each wire individually from one end to the other?

 

[sm52]

The positive wire consists of two cores, each of which is located in a dielectric tube. The minus one also consists of two veins. All four cores are in a common dielectric tube. At the ends, two positive wires are connected in a pigtail. Also negative. For the plus, one core with a diameter of 1 mm is used, the second 0.6 mm. For the minus one core with a diameter of 1 mm, the second 0.75 mm. I measured between the ends of the positive - one graph, and between the ends of the negative - the second graph.

 

[jschwender]

That is not the use case, because you use both plus and minus wire the same time and you run exactly the same amount of current through it. This leads to a partial compensation of the inductance. As your measurement represents the impedance without such a compensation, the result is an extremely high inductance value. What you need to measure for a line measurement is the impedance between plus and minus on the amplifier end while the other end is short-circuited. It would be great if you could do a measurement on this! Is this Type 2 speaker cable?

 

[sm52]

Is this Type 2 speaker cable?

Yes.
Measure the impedance of an amplifier without speaker wires? Or with these wires, which are short-circuited plus on minus at the ends connected to the speaker terminal?

 

[jschwender]

Cable input (amplifier end) must be connected to the soundcard for measuring. Cable output (speaker end) must be shorted, no speaker connected.

 

[sm52]

Did. Without amplifier. Sound card only.

 

[jschwender]

Thank you. Interesting result. The low frequency Inductance is ~330 nH/m, that is equivalent to a Mogami 3082. The DC resistance is equivalent to a 1.2 mm² cross-section, which makes sense with the diameters you mentioned. Accordingly, the resistance value is higher compared to a Mogami 3082, so here the Mogami is better. The skin-effect increase is lower compared to a simple twin braid, but not as good as a flat cable. A simple RG213U performs better in all electrical values.
Under the (simplified) assumption that your load (speaker) is purely resistive and you use that 5m cable for a 4 Ohm speaker, the pure loss at low frequency is 3.7% or 0.3 dB. At 20kHz the loss is about 6% or 0.55 dB.

You asked about the resistance plot, which makes the skin-effect and proximity-effect directly visible: From the pure impedance curve you see a curve with a flat part on the low frequency end and a slope increase (log axis) towards high frequency. The value at the flat corresponds to the Dc resistance, and the level of the slope corresponds to the inductance. In real world both are not constant, and from the sum you don't see that, as the skin-effect is in a part of the curve where you have a slope, and you can't see if this impedance slope is a little bit steeper or flatter. If you see the distinct resistive part, you have a mark how much the increase of the dc resistance is, and the same time this increase corresponds with a decrease of the inductance. Both are quality features for a cable. The flatter they are and the lower their absolute value, the better the cable is. It would generally be great to have the inductance part also in the same plot, but this is a serious scaling issue, as the absolute numbers are much smaller for inductance. For the Type 2 cable such a plot looks like this:

 

[sm52]

0.3 dB and 0.55 dB are small values. I think it will be difficult to understand by ear. From 3 kHz to 20 kHz, the resistance increases from 0.15 ohms to 0.25 ohms. On a good amplifier, this shouldn't be a problem either. It turns out that the cable is good.
Is it possible to compensate in any way for the difference between the resistance at LF and HF in conventional cables, in which this difference can be 8-10 times?

 

[jschwender]

Right, the pure calculated value of .55 dB is not noticable. But this is only half the story because the assumption under which this calculation is right, is most of the times not given. Speakers have complex impedance, and that can cause very unexpected resonant interaction with allcomponents. This can cause deviations of far more than 0.55 dB, and those can be noticable. In practice the biggest disturbing factor is most of the times the inductance of a cable. Compensating this weakness is not at all easy, i have read a publication from Peter Horowitz (AES member), named: Engineered Remote-Sensing Audio Power Amplifier for High-Fidelity Applications. (Convention e-Brief 364, from 2017) where he shows impressive results with stable operation of up to 100 m cable.

 

[sm52]

It is still not clear to me what gives the visibility of the real part of the impedance on the impedance plot. Let's say the impedance and real part graphs go from LF together and diverge by 400 Hz. Or at 3 kHz. Or at 10 kHz. How to apply it?

 

[jschwender]

Without that curve you can only guess the size of resistivity, with that curve it is quantified. This is what measuring is all about.