Data inconsistencies AIM_882

[g3txq]

Bob,

I wonder if you have any comments on the inconsistencies reported by Owen Duffy; he apparently is seeing significant errors between the SWR, Rs and Xs values in a CSV file, compared to the Zmag and Theta values in that file:
owenduffy.net/blog/?p=3334#more-3334

I've looked at a few CSV files from my AIM4170 and my VNA2180, and haven't yet seen the same thing; but I'm not running the latest software.

Steve G3TXQ

[Bob]

Hi Steve,

I hadn't seen that post before.
This gives me a chance to do a little tutorial on the AIM data files.

Owen is correct that the *.scn file stores Zmag and Phase (in radians), along with the freq and the swr.
The sequence of data in the scn file is: Freq(MHz)   Zmag  Phase  SWR
When you save an scn file, the csv file is also created.

It's not hard to compare the files side by side and see the numbers are equivalent.
For example:
Rs = Zmag*cos(Phase)    (phase is in radians;   most scientific calculators will accept either degrees or radians for trig functions)
Xs = Zmag* sin(Phase)   ( to convert radians to degrees, multiply by 180/pi,   approximately equal to 57.3)

Note that RG58 is not exactly 50 ohms. The "Transmission Line Details" program uses 52 ohms for the nominal impedance. When a 3 ft cable is terminated with 25 ohms, the SWR will slope downward as the frequency goes from 10 to 50 MHz, similar to what Owen shows in his first screen shot of the AIM scan data.

The bottom line is the scn file and the csv files are consistent.

Of course, the absolute accuracy of the test results will depend on several things like: calibration loads, adapters, interconnecting cables, and jumper clips.

If anyone notices any discrepancies in the data please let me.

--73/Bob

[g3txq]

Bob,

I don't think Owen's observations related to *absolute* accuracy; rather, that the values of Rs, Xs and SWR derived from Zmag and Theta were not self-consistent within the CSV file. To quote him:

"So, I have calculated R, X and VSWR from the polar Z data in the .csv file, and compared them to the same values reported directly in the .csv file."

I guess you would need an example of the .csv file from him to be able to pursue the issue.

Also, did you notice the anomalies in the data on his scan display:

1) At the cursor position, RL is clearly in excess of 30dB; but the data table shows it as 10.89dB

2) The data table shows Rs=54.994 and Xs=-3.304, both of which *are* consistent with Zmag=55.094 and Theta=-3.438; but that impedance is totally inconsistent with the reported SWR of 1.799 and the RL of 10.89dB.

The correct results for that impedance (given the Zref of 55) should be SWR=1.062 and RL=30.45dB

Edit: By the way, neither the cable length nor a Zo of 52 Ohms could explain the variation in the SWR values in his CSV file between 10MHz and 50MHz. Once again, if you recalculate the SWR from basic Zmag and Theta values in the CSV file, it is pretty much flat across that frequency range as you would expect.

Try it in TLD. A 1m length of RG58 - even with Zo set to 52 Ohms - only shows SWR(50) going from 1.98 to 2.15 between 10 MHz and 50MHz with a 25 Ohm load. But Duffy reports the SWR values in his CSV file going from >2.15 to <1.9

Steve G3TXQ

[g3txq]

I just pulled out my AIM4170C and ran it with AIM_882 software - I can confirm what Owen was reporting!

I'm attaching the CSV file for you to take a look at. Every entry for Rs and Xs is inconsistent with Zmag and Theta, apart from the very first frequency (10MHz).

As an example, take a look at the data for 40MHz:
Zmag = 89.985881, Theta = 14.849166

should give us:
Rs = 86.9807, Xs = 23.06116

but the values in the CSV file for 40MHz are:
Rs = 78.078881, Xs = 29.288448

Steve G3TXQ
test 882.csv (4.15 KB)

[Bob]

Hi Steve,
Thanks for the feedback.
I noticed that the smoothing function is causing the discrepancy. Your data and Owen's data both used smoothing. Now when I try it with smoothing, I see the problem too. I'll fix that right away. I have a new version almost ready to release so I'll incorporate the fix in that and upload it in a few days.

--73/Bob

[g3txq]

Bob,

I have made hundreds of measurements using the .csv dump from my VNA2180 running VNA_536. Is there any chance the problem affecting AIM_882 could also have affected those VNA results?

Steve G3TXQ

[Bob]

Duffy greatly exaggerated the effect of this issue. It happens when smoothing is used for the AIM or AIMuhf. The normal operation does not employ smoothing and the effect doesn't appear.
The VNA is different. You can spot check some values to be sure.

--73/Bob

[k4hkx]

I have the original .scn and .scv files on which Owen Duffy based his critique..  I would be happy to provide them if requested  In addition to his comments, I have found display problems, specifically, I noticed a very significant offset to the left from accurate alignment when displaying return loss with the SWR and other traces.  I would be surprised if this problem is related to smoothing.

Paul

[g3txq]

Jan 22, 2015 7:37:26 GMT -5 k4hkx said:

I have the original .scn and .scv files on which Owen Duffy based his critique..  I would be happy to provide them if requested  In addition to his comments, I have found display problems, specifically, I noticed a very significant offset to the left from accurate alignment when displaying return loss with the SWR and other traces.  I would be surprised if this problem is related to smoothing.

Paul

I'm sure Bob will tell us the reason for this, but it wouldn't surprise me if it *was* related to the smoothing. Some test .scn files I generated, which produced "anomalous" .csv files, had that same sort of offset - the correct impedance appeared at a lower frequency in the table. Switching off the smoothing cured the problem.

Does anyone know why Owen had the smoothing switched on for his measurements? It would seem a strange thing to do if you want accuracy - particularly if you are using coarse frequency steps.

Steve G3TXQ

[g3txq]

I've been doing some antenna matching calculations for a fellow UK ham, and was surprised when he reported that he couldn't achieve an SWR lower than 3.2:1 using the component values I gave him.

Upon investigation it turns out his AIM4170 was reporting Rs=43 Xs=17 which I thought was reasonable and which I calculate to be an SWR(50) of 1.45:1 However the AIM4170 reports the SWR as 3.2:1 with those values ??

He's running 882, and I've advised he upgrade to the latest version; but what confidence can he have that the latest version will have fixed that problem?

Details here: www.gatesgarth.com/136.jpg

Steve G3TXQ

[Bob]

Hi Steve,

The problem is caused by your use of smoothing in a situation where it isn't needed. The program version 882 did not check for this type of inconsistency so I've added a check for it in the latest version 885A. Generally, smoothing is useful in the VHF/UHF range when the number of data points is relatively large and there may be noise. The size of the smoothing filter is now limited to a few percent of the number of data points. This reduces the chance for distortion in complex waveforms. If there is any doubt about distortion, try it with and without smoothing.

73/Bob

[g3txq]

Hi Bob,

The ham has installed 885 and that seems to have cured the problem.

What I still don't understand is why - smoothing or not - the results are not self-consistent even though they might be wrong. In other words, I can understand that the impedance of 14+j17 might be in error because of smoothing, but I would have expected all other parameters then to be consistent with that 14+j17 figure.

But I guess it doesn't matter now!

Steve G3TXQ