Impedance analyser nonlinearity

[ProbeT]

Noticed differences in resistors plotted with different driving voltage (all other things being equal).  The 5V range is out by a fair way as frequency increases. But as the driving voltage is supposed to be measured at channel one, compensating for the actual driving voltage, then why is this happening?  Tried it with different types of resistors and found similar dependence with the 5V range and variation increased with frequency. Any suggestions?

[JColvin]

Hi @ProbeT,

I have moved your question to a more appropriate section of the Forum where the engineer best suited to help you will be able to see and respond to your question.

Thanks,
JColvin

[attila]

Hi @ProbeT

I repeated your experiment with AD2, using a breadboard and twisted Scope (1+|1-,2+|2-) and AWG (W1|GND) flying wires, using 1K reference resistor and 1K load, with default load first setup.
The results at 1V and 5V, compensated and uncompensated:

[ProbeT]

I tried this again with a similar result to my first post. I got a 1% increase in amplitude for the 5V range at 20kHz; breakpoint around 3kHz. My unit is uncompensated and I am making measurements using the BNC probe adaptor board and using a bunch of BNC to 4mm adaptors.

The parameters determined at 20kHz were

1V  23.65uH & 1.001kO

2V 22.54uH & 1.001kO

5V  66.21uH & 1.011kO

giving a difference of ~1% of the resistor value.  I changed connection lead lengths around to shorten everything, but the result was about the same.

I don't think there is a need to twist wires etc. and measure directly from the AD2 terminals without the BNC adapter board to get a comparative result. I guess it beggars the question Is the unit defective?

 

 

 

 

 

[attila]

Hi @ProbeT

The impedance analysis has pretty complex math behind with many inputs (frequency, scope phase, magnitude, cable resistance, capacitance, reference value...)
Small differences in setup or configuration can influence the result and magnitude/phase level affect the result precision.

Try compensating.
Configure the options (offset, amplitude, start/stop frequency, samples..), perform and enable the open and short compensations for the selected reference resistor.
The scope inputs on AD2 have two gains 5V and 50V pk2pk. The amplitude setting in IA changes gain at about 2.27V.
At 2V and 5V the scope input characteristics are a bit different so separate compensation is recommended.

I used (twisted) wires because I didn't know you are using BNC.

[ProbeT]

I have done as you suggested and performed both measurements compensated at 1V and 5V excitation. I have attached a picture of my setup for the measurement and also the result. The result has improved but is still well below the precision of the 1V range. If I look at it from a difference between 1kHz and 20kHz then for 5V the difference in resistance is 4.6 Ohm or 0.46%, whilst for 1V it's  0.2 Ohm or 0.02%. Your results are 5V 0.4% and 1V 0.2%. Which I guess suggests that my unit is working about as well as yours when compensated, from a change in value perspective. In my case though there is a absolute difference of 4 ohm between the 1kHz values at 1V and 5V compared to your 1 ohm difference. I don't have a precise value for my reference resistor which is a 1% metal film type with a nominal value of 1k. I also note that my graph range is 20 Ohms and yours is 150 Ohms.

When the measurement was extended to 200kHz, then the differences were 5V ~0.8% and 1V ~0.2%.

I added another 1k resistor in series to make the load 2kOhm to reduce the voltage drop on the measured resistor (in case it was voltage coefficient) but the pattern was repeated. On a dB scale these sorts of differences don't amount to much but for precision work where the absolute value is important, this would represent a limitation. 

I guess the basic question to answer still is why do the 0.5, 1V and 2V ranges track each other across the frequency range but the 5V range does not?

 

[attila]

Hi @ProbeT

At 5V the scope inputs of AD are set to low gain (50V pk2pk) and at 2V to high gain (5V pk2pk).
https://reference.digilentinc.com/reference/instrumentation/analog-discovery-2/reference-manual#scope_input_divider_and_gain_selection

We have to investigate this further, but this 1%-0.25% error might be the limitation of such measurement setup...