**Jim Lux** 2022/12/14 20:43
On 12/14/22 7:55 AM, Raphael Wasserman wrote:

> Jim Lux,

> Your explanation is difficult to understand. I draw diagram of Wheatstone bridge exactly how it is designed for this VNA. So, the bridge all legs have 50 ohm resistors as well the resistor from power driver is 50 ohms and resistor of cross arm where the measured receive power is measured by system. First I looked what happens when the ch0

> sees either short or open condition and what voltage will be measured across the cross arm resistor and later did this when to ch0 port is connected a 10 dB fixed pad for 50 ohms termination, say designed with 3 resistors.

> Again when the output of attenuator is not terminated with 50 ohm load and there is either short or open condition the bridge will see different resistances and therefore different voltages across cross resistor arm will be measured. I am not sure whether those values can be considered as calibration values for further

> Measurement of complex impedance load ( s11).

>

The whole point of the calibration process is that you solve a set of

simultaneous equations to get coefficients using 3 known impedances

(open, short, and load happen to be easy).

So whatever properties the bridge (and fixture, which could be an

attenuator) have, you have 3 measurements and 3 unknowns.

The bridge can be hideously bad, and it still works. The constraint is

this:

The uncertainty of the "calibrated" measurement (after applying the cal

factors) is determined by the SNR of the measurement, and the

uncertainty of the cal factors.

The uncertainty of the cal factors is determined by the uncertainty of

your 3 reference impedances and the SNR of those measurements and how

well the 3 standards "span" the range of measurements.

> Raphael

>> On Dec 14, 2022, at 8:59 AM, Jim Lux <jimlux@earthlink.net> wrote:

>>

>> On 12/13/22 7:26 AM, Raphael Wasserman wrote:

>>> How can you calibrate the output of the attenuator ?

>>> When a short or open is applied on output of attenuator the bridge will see some resistance in both events but the VNA will not see either short or open conditions.

>>> I tried both with attenuator 12 dB and lowering the drive current getting two different results while measuring the load impedance.

>>> With attenuator Rl was 73 ohm without attenuator with lowered current drive about Rl 115 ohm.

>>> Regards,

>>> Raphael

>>

>> Do your open, short, load calibration at the output of the attenuator.

>>

>> It doesn't matter what the VNA actually sees, that goes into the calibration.

>>

>> Consider the bridge inside the VNA - it's not perfect, so even if you put a short on the output, the power seen by the receiver is less than the outgoing power, because of losses in the bridge.

>>

>> No different when there's a 10 dB (or any value) pad.

>>

>> Let's say the receiver sees -3 dBm with a short on Ch0 with no pad.

>> That becomes the reference value (amplitude and phase) for a perfect short.

>> Let's say it sees -3dBm (but phase flipped 180) - with an open. That's the reference value for perfect open.

>> Let's say it sees -40 dBm with a perfect load (yeah, the bridge isn't perfect) - that becomes the reference value for a perfect load.

>>

>> Same thing with a pad:

>> Short - sees -23 dBm

>> Open - sees -23 dBm

>> Load - sees -40 dBm (because it's dominated by the bridge leakage)

>>

>> Then those values go into the calibration math.

>>

>> The overall problem with putting a pad on the port is that you lose dynamic range. The noise floor in the example is, say, -60 dBm. So instead of a -3 - (-60) = 57 dB SNR, it's seeing -23 -(-60) = 37 dB SNR. Which has more measurement uncertainty.

>>

>>

>>

>>

>>

>>

>

>

>

>

>