Calibration anomoly in Ver. 6.50 and 6.28

[w0qe]

Do a standard AIM4170 calibration (not a custom calibration) with the supplied calibration standards at the end of a piece of coax (my case 14'10" of RG-148 50 ohm, .66VF, teflon coax). Then with the "Limits" window set to start 1 MHz, stop 170 MHz, step 1 MHz, do a scan with the 200 ohm resistor at the end of the coax. All looks well.

Now change the "Limits" window to be a 0.5MHz step size and do a scan. The resulting scan has huge anonolies every 11.5MHz and at every ~11.5MHz thereafter (or every freq. where the coax is a multiple of a quarter wavelength). It appears that any step sizes that are not integer multiples of 1MHz cause problems. By the way ver. 6.28 also exhibits these anomolies but they look different.

Using Ver. 6.50 and fooling around with the custom calibration it "appears" that the step size in the "limits" window must be an integer multiple (1x, 2x, 3x, etc.) of the step size used for calibration or anomolies will appear in the scan when calibrating out to the end of a piece of coax.

I am a new user of the AIM4170 but own an HP8753 and am not new to VNAs or RF measurements.

Larry Benko, W0QE

[w0qe]

There is no RG-148. Sorry, It is RG-142.

[dl4raj]

Larry,

a possible explanation might be that the coax (like any coax)
has a small reactance,e.g. 50 -j1,5 Ohm (reactance varying
over frequency).
Therefore a pure resistive load gives a pure resistance only
every 0,25 lambda,but in between the measured impedance
must contain a small reactance.
Good analyzers or reflection bridges can show this small
reactance.
It woud be interesting to hear what you see on your HP VNA.

73
Clemens
DL4RAJ

[w0qe]

Clemens,

The anomolies I am referring to seem to be some kind of math error where a termination at the end of a piece of coax displays as an incedibly high or low impedance that is not mathematically explainable. This can not be explained by coax that has a characteristic impedance that is slightly reactive.

Do the test for yourself. I can reproduce it easily with many lengths of coax. The problem does not occur if there is no coax and does occur for coax lengths greater than an electrical half wavelength at the stop frequency in the "Limits" table WHEN the step size in the "Limits" table is not an integer multiple of 1 MHz for the standard cal or not an integer multiple of the step frequency in a custom cal.

Larry, W0QE

[n9fyx]

This kind of makes sense, since when you are calibrating at the end of a coax, you are trying to compensate for highly diverging impedances that occur every 1/4 wavelength when doing the open and shorted tests. Then if you do a measurement that requires any interpolation such as when the step size is different, the errors would be very large at those points, even if the interpolation was done right.

[w0qe]

There are no diverging impedances here at all as the problem occurs with a 200 ohm resistor. Let me go thru the procedure again. It is easy to try for yourself.

1.) Attach a piece of coax to the AIM4170 between 6 and 20 feet long.
2.) Do a standard calibration of the analyzer with the calibration standards attached to the end of the coax.
3.) Leave the 200 ohm resistor connected to the coax.
3.) Press the "Limits" button and set the start frequency to 1 MHz, the stop frequency to 170 MHz, and the step to 1 MHz.
4.) Now do a scan. Notice the constant 4:1 SWR (as it should be).
5.) Press the "Limits" button and keep the start frequency at 1 MHz, keep the stop frequency at 170 MHz, and change the step frequency to 1.1 MHz or 0.9 MHz.
6.) Now do a scan. Notice the SWR is no longer constant (as it should be) but has several anomolies.

Where in this experiment do impedances diverge to very low or very high values? This type of calibration is done all the time with VNAs. The problem here is appears to be a software issue but Bob is the person to comment on the cause. 73, Larry, W0QE

[n9fyx]

The first step in the standard calibration is to put an open
BNC "standard" in place and the instrument will compensate
for stray capacitance etc. in the input jacks. If I understood
right, you put a coax on and put the open jack at the end of the coax. Now the instrument sees highly diverging impedances at .25 wavelength freqs and compensates those. Then a
similar thing happens when you do the short test. I agree the 200 ohm shouldn't have highly diverging impedances. Did you do such a calibration with your HP successfully?

[Bob]

n9fyx said:

This kind of makes sense, since when you are calibrating at the end of a coax, you are trying to compensate for highly diverging impedances that occur every 1/4 wavelength when doing the open and shorted tests. Then if you do a measurement that requires any interpolation such as when the step size is different, the errors would be very large at those points, even if the interpolation was done right.

This is what causes the problem alright. The raw cal data at quarterwave intervals goes to zero or infinity. Even a small measurement error causes a large percentage error when the numbers are such extreme values. Then when applying the correction data to a nominal load like 200 ohms, the final answer may be way off at frequencies in the vicinity of the quarterwave points.

There is a test in the standard cal procedure to catch a situation like this. The warning can be disabled by a flag in the config file and it's probably set to zero now by default. Look for a line that says:" issue a warning if a long stub is attached". If you want to enable the reminder message, change the flag value from zero to one.

The custom cal procedure improves on this situation in two ways. Data points are taken at smaller frequency increments and the interpolation is done with a more sophisticated algorithm. I did a test this morning with a piece of RG59/u that is resonant at 28MHz. After custom cal from 10 to 60 MHz with a delta freq=0.1MHz, scans of the 200 ohm resistor with several different values of delta freq resulted in a discrepency of less than 1.5 ohms at the 1/4 and 1/2 wave points.

The cal data check that is done at the end of custom calibration looks at values in between the original cal data points to see if the result will be reasonably close. If the delta freq was too large, there will be a warning that some data points may not be accurate. The cal data can still be used if you're confident it really is okay. This can be checked by doing a scan and if the abnormal readings are outside the range you're interested in, then the data can be used. This might be the case, for example, where you calibrate with a high pass filter and some of the first points at the low end are off, but the rest of the cal is ok in the range that you're really interested in.

-- 73/Bob

[w0qe]

Message for N9FYX. Yes I calibrate my HP VNA at the end of arbitrary lengths of coax at all frequencies from HF thru several GHz. Once I calibrate over a certain frequency range with a certain number of points (201, 401, 801, 1601 are the choices) I can sweep any frequency range within the calibration range but can NOT change the "step size" which is allowed by the AIM4170. Perhaps HP wanted this to be done to minimize the anomolies that occur at the quarter wavelength multipiles. This means that calibrating from 300KHz to 3GHz and then sweeping from 4MHz to 5MHz will produce some gagged looking Smith charts but VNA shows in the display that the results are probably compromized by displaying a warning.

Nevertheless if you insist on changing the step size, the AIM4170 performs much better when a custom cal has been done. I intend to generally not change the step size.

[Bob]

The best scan results will be obtained when the freq steps are equal to the steps used during calibration although this will not be required in many cases.
For the very best results, the scan points can be made to coincide with cal points. This means the delta freq size has to be some multiple of the cal step size and the start freq value has to be such that the points fall on top of the cal points.
For example,
ScanStart = CalStart + N*CalDelta
ScanDelta = M*CalDelta

N and M do not have to be the same.

When custom cal is being used, the start/stop/delta parameters for the cal data can be seen in the status file. Click Help->Status.

Hopefully, the interpolation procedure will make choosing the scan parameters non-critical so you can use any start/stop/delta values and not have to worry about these details.

It might be beneficial to have a mode where the scan points are made to coincide with the cal points automatically. For example, the values for M and N would be selected by the program to optimise the scan results. The only trick is how to make this easy for the user to specify. Where should the "auto" button be located? If anyone wants to offer a suggestion, I'll be glad to get feedback on how to implement the user interface for this feature.

The cal points can be as close together as 1KHZ and up to 1000 points can be used. The number of cal points can be increased, so if this isn't satisfactory, I'll change it. The only disadvantage is that it takes longer to do the cal process. It doesn't take any longer to do the actual scan.

The custom cal procedure is new and subject to change, so let me know if you have any results that indicate something can be improved.

-- 73/Bob

[w0qe]

Bob, As you know,I have made a bunch of measurements with the AIM4170 in the past few days comparing them to my HP VNA and agree with you 100% that the best results occur when the cal. points match the scan points. I think a button in the "Limits" window saying "Force Limits to match cal. data" would perfect. Of course by turning off this button you would be free to use other limits. I also think that displaying the current cal. start, stop, step values might be nice as they may have been forgotten. I am constantly amazed of the accuracy of the AIM4170 when it is used in a manner that produces the best results. Kudos on designing a cool device.