I would like to gain an understanding of interpolated error correction on the PNAs, both old (8720D/8753E) and new (N5241A/N5242A).

*_Q1 - What is the interpolation algorithm?_*

Does the interpolation use simple straight-line interpolation between calibration points?

Does the interpolation interpolate error corrected data points by (1) straight-line interpolation between other error-corrected data points, or (2) by straight-line interpolation of calibratioin points and raw data points, and then applicaiton of the interpolated cal to the interpolated data?

Or is the interpolation more complicated?

*_Q2 - Interpolation Accuracy_*

Agilent has been VERY careful to state only that when you set INTERP (IEC) on the accuracy of the error correction is not guaranteed. But... there is NOTHING that tells you how badly it is degraded.

I know from the manuals that (a) IEC is 'most accurate' when there are 67 (real) calibration points per 1 GHz in the calibration you are interpolating from; and that (b) interpolation doesn't work if the phase shift is more than 180 degrees per 5 data points.

Are there any (reasonable) guidelines that can be used to make interpolation a useful feature?

_*Q3 - Changing "stimulus" parameters in HOLD mode and Interpolation*_

The start freq, stop freq, and number of points can be changed when you are in HOLD (no trigger, no new sweeps) and it works if IEC/interpolation is ON. This seems weird since the parameters are supposedly "stimulus" parameters that affect how a sweep is performed, not postprocessing. But it makes sense - when in HOLD, changing those parameters can be done (subsject to the obvious limitatiion that the start/stop freqs must be a subset of the actual sweep) via interpolation. Since interpolation cannot improve the accuracy of your data (you don't collect more data when interpolating) is the accuracy of an interpolated measurement the same as a non-interpolated measurement *if* you omit error correction accuracy (error introduced by cal interpolation)?

It seems to me that the ONLY source of accuracy degradation when interpolating is the fact that the calibration points in a non-interpolated sweep line are at the exact same frequencies as the data points - and not (necessarily) when interpolating...

_*Q4 - Is there any documentation that addresses these issues? Agilent or 3rd party?*_

Thanks in advance!!!

*_Q1 - What is the interpolation algorithm?_*

Does the interpolation use simple straight-line interpolation between calibration points?

Does the interpolation interpolate error corrected data points by (1) straight-line interpolation between other error-corrected data points, or (2) by straight-line interpolation of calibratioin points and raw data points, and then applicaiton of the interpolated cal to the interpolated data?

Or is the interpolation more complicated?

*_Q2 - Interpolation Accuracy_*

Agilent has been VERY careful to state only that when you set INTERP (IEC) on the accuracy of the error correction is not guaranteed. But... there is NOTHING that tells you how badly it is degraded.

I know from the manuals that (a) IEC is 'most accurate' when there are 67 (real) calibration points per 1 GHz in the calibration you are interpolating from; and that (b) interpolation doesn't work if the phase shift is more than 180 degrees per 5 data points.

Are there any (reasonable) guidelines that can be used to make interpolation a useful feature?

_*Q3 - Changing "stimulus" parameters in HOLD mode and Interpolation*_

The start freq, stop freq, and number of points can be changed when you are in HOLD (no trigger, no new sweeps) and it works if IEC/interpolation is ON. This seems weird since the parameters are supposedly "stimulus" parameters that affect how a sweep is performed, not postprocessing. But it makes sense - when in HOLD, changing those parameters can be done (subsject to the obvious limitatiion that the start/stop freqs must be a subset of the actual sweep) via interpolation. Since interpolation cannot improve the accuracy of your data (you don't collect more data when interpolating) is the accuracy of an interpolated measurement the same as a non-interpolated measurement *if* you omit error correction accuracy (error introduced by cal interpolation)?

It seems to me that the ONLY source of accuracy degradation when interpolating is the fact that the calibration points in a non-interpolated sweep line are at the exact same frequencies as the data points - and not (necessarily) when interpolating...

_*Q4 - Is there any documentation that addresses these issues? Agilent or 3rd party?*_

Thanks in advance!!!

> I would like to gain an understanding of interpolated error correction on the PNAs, both old (8720D/8753E) and new (N5241A/N5242A).

>

> *_Q1 - What is the interpolation algorithm?_*

> > Or is the interpolation more complicated?

Quite a bit more complicated. We call it circular interpolation which is essentially an interoplation in vector space. The exact algorithms are proprietary. I give quite a few more details and several examples in Chapter 3 of my book (I know, getting old saying that, but the info really is all there): http://www.wiley.com/WileyCDA/WileyTitle/productCd-1119979552.html

>

> *_Q2 - Interpolation Accuracy_*

> Agilent has been VERY careful to state only that when you set INTERP (IEC) on the accuracy of the error correction is not guaranteed. But... there is NOTHING that tells you how badly it is degraded.

> I know from the manuals that (a) IEC is 'most accurate' when there are 67 (real) calibration points per 1 GHz in the calibration you are interpolating from; and that (b) interpolation doesn't work if the phase shift is more than 180 degrees per 5 data points.

>

> Are there any (reasonable) guidelines that can be used to make interpolation a useful feature?

If you have 1 point every 5 MHz for each meter of test port cable, you will typically meet the specified perfomance of the VNA (so if you have two, 1 meter test port cables, you will need a point every 2.5 MHz).

>

> _*Q3 - Changing "stimulus" parameters in HOLD mode and Interpolation*_

>

> The start freq, stop freq, and number of points can be changed when you are in HOLD (no trigger, no new sweeps) and it works if IEC/interpolation is ON.

The changes are accepted, but the data is held with the old values. Newer analyzers don't even update the stimulus settings until a new sweep is acquired. Interpolation has NO effect on measured data, it only effects the error terms. You are not interpolating the data, you are interpolating the error terms over the new frequency range and number of points, and then new raw data must be acquired (through a sweep) before the display is updated.

>

> It seems to me that the ONLY source of accuracy degradation when interpolating is the fact that the calibration points in a non-interpolated sweep line are at the exact same frequencies as the data points - and not (necessarily) when interpolating...

>

> _*Q4 - Is there any documentation that addresses these issues? Agilent or 3rd party?*_

>

Why yes, as noted in the link above.

> Thanks in advance!!!

You are most welcome, in retrospect.