diyAudio (
-   Power Supplies (
-   -   Network Analyser Measurements (

KSTR 19th December 2012 10:42 PM

Network Analyser Measurements
1 Attachment(s)
Hi all,

Having picked up a used spectrum/network analyser just recently in a moment of big luck, I thought I'd share any meaningful measurement I'm doing with it, over time. Not many people will have access to such an instrument, let alone hobbyists, and measurements of power supply components and systems are scarce.

A network analyser is an extremly poweful tool for impedance measurements which is basically most what power supply issues are about. Looking at the effective impedance seen at the point of load, say, a chip amp's power pins, is what counts, effectively.

The spectrum/network analyser was built late '90ies by the reputed german measurement equipment company Wandel & Goltermann.
The model is SNA-2, the second model in their range which adds phase measurement capability to the earlier SNA-1. Frequency range is 100Hz...180MHz which is just perfect for audio use.

Sitting on my bench it looks like this (note that it is comprised of two really heavy 19" die-cast aluminum boxes packed with electronics to last cubic inch) :

KSTR 19th December 2012 11:03 PM

2 Attachment(s)
The little test jig I made for capacitor etc measurement :

A short coaxial cable was cut in the middle and a tiny double-size blank PCB was soldered in between, forming a tiny capacitor (a few pF at most).
One end of the coax was fed by the tracking generator of the analyser while the other end goes directly to the receiver input. System impedance is 75R. When a DUT (Device Under Test) is low impedance, fractions of an ohm typically compared to the system impedance this forms a constant current source and the voltage seen at the output is proportional to impedance of the DUT.

Plotting the transfer function of the jig alone looks like this :
We can see a pretty perfect line at 0dB with a slight roll-off at the highest frequencies (above 40MHz) coming from the tiny capacitance mentionened and other high-freq effect in the measurement chain. Horizontal scale is ranging 6 decades from 100Hz to 100MHz, in 1-2-4-8 increments for the grid lines (as indicated). This setting is kept for the rest of todays measurements.

KSTR 19th December 2012 11:24 PM

2 Attachment(s)
Using a 0.1R resistor as a reference to adjust vertical dB scale to read in dB ref. 1 Ohm :

Resulting impedance plot after adjusting the reference level (which just shifts the plot) :
Now 0dB=1R, -10dB=0.32R, -20dB=0.1R, -30dB=32mR and so forth. Likewise, positive 10dB step give 1/3.32/10/... values.

The 0.1R wirewound is inductive as seen from it's constant slope from 400kHz upwards, the inductance can be calculated from |Z|=2*pi*f*L, hence L=|Z|/(2*pi*f). Reading 1R (0dB) from the plot at 8MHz we get 20nH.

KSTR 19th December 2012 11:39 PM

2 Attachment(s)
In order to figure out the "baseline inductance" I tried shorting the jig with pliers as close as possible to the cables and using a lot of pressure to get a low shorting resistance :

The according plot looks interesting :
There is an inductive rise again above the <10mR of shorting resistance, coming from the residual path around the PCB edges through the pliers, but the inductance doesn't seem proper in the region where it starts to emerge and it doesn't reach the 20dB/decade slope, only a little more than 10dB/decade. My educated guess is that we see the result from eddy current losses and/or the magnetic properties of the steel pliers (need to investigate that further, with a copper short things looks quite different -- perfectly inductive -- from earlier tests I made on another jig).

KSTR 19th December 2012 11:46 PM

3 Attachment(s)
Now finally hooking a capacitor to it, a WIMA MKS2 type with 3.3uF :

Plot looks just fine :

Comparison to WIMA's spec sheet graphs shows excellent correlation :

KSTR 20th December 2012 12:01 AM

2 Attachment(s)
Let's add a bit of lead length to the 3.3uF capacitor :

The result on impedance in comparison to the previous one (overlayed):
Inductance kicks in earlier (notch shifted to the left) and is about 5dB higher -- which is about twice the original value of the cap intrinsic inductance when flush-fitted to the PCB edge.

This demonstrates the capapilities of a network analyser in an impressive way. Add a mm of lead length and you'll see it instantly!

Capacitance reads as before, calculating from the falling slope via |Z|=1/(2*pi*f*C) we get find C=1/(2*pi*f*|Z|). In this case we read 0.32R @ ~180kHz ==> 2.8uF. That's "close enough" to 3.3uF, from the eyeballing employed.

KSTR 20th December 2012 12:26 AM

2 Attachment(s)
Coming to the last post of today's series, a 150nF film cap (unknown brand) is paralled to the 3.3uF (1:20 capacitance ratio) :

Impedance plot tells us that this is a two-sided sword :

While at very high frequencies the inductance is halved (6dB down, and even more at the resonant frequency of the 150nF cap) as expected from the second "RF short circuit" around the PCB edge, we now also have a nasty resonance peak sitting at 4MHz which is 7dB higher (2.2x) than the original curve of the 3.3uF alone. The high quality of the caps is in fact a problem here, there are close to zero ohmic losses present that could absorb most of the resonant peak (at ~0.4R).

Intrinsic inductance of the two caps should be about the same based on their construction, so their notches should be spreaded apart by a factor of sqrt(20)=4.47 because a LC resoncance is at f=1/(2*pi*sqrt(L*C)). This also is confirmed by the plot in close concordance (use the plot for NO lead length on the 3u3 for this).

DUG 20th December 2012 02:51 AM

These are wonderful tools that teach us a lot about components.

Anything with area has capacitance.
Anything with length has inductance.
Ground is only ground for a short distance.

Thank you for posting these pix.

I hope everyone views them.

jan.didden 20th December 2012 07:56 AM

Excellent measurements! It is indeed a very nice piece of kit.
I assume you will at some point go forward to measure actual power supplies and decoupling effects, looking forward to read about that!


powerbob 20th December 2012 09:28 AM

Thank you for your post. It was a good mini tutorial.

All times are GMT. The time now is 09:03 PM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 18.75%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Copyright ©1999-2018 diyAudio