Measuring Power Supply Output Impedance Using a VNA


Output impedance is a key quality criteria for a power supply. If a PSU is any good, it is very small. Now, how can something like this be realiably measured over a rather large frequency span ? People from Keysight and other companies have derived a bag of tricks how this could be done with “standard” (*) VNA equipment. I got my hands on two of those, the Keysight E5061B-3L5 VNA that goes from 5Hz to 3GHz, and the Bode100 from Omicron Systems that works from 1Hz to 50MHz.

It can be done – with a bag of tricks. Here are the gory details …

Click Here for Making Power Supply Measurements with a Vector Network Analyzer …

(*) Standard means that you can buy this off the shelf, but it does not mean its cheap or simple in any way. The Bode100 is about 4.5k€, and an E501B-3L5 is about 35k€ (list prices).

Not All LM723s are Born Equal


I made a few tests using LM723s from ST in a self-heated reference circuit. The typical drift curves over 10.5 days looked like this:


Its all covered by the datasheet, saying typical stability is 500ppm/kHour at 25°C (we have ca. 90°C).

The circuitry can be seen here:

Click here to see Silly Circuits – A Heated LM723 Reference …

Still, its a bit disappointing, so I tried a military grade LM723J. OK, now it looks like this:


It drifts less, and it seems to settle down on top of that. Looks promising !

All the gory details can be seen below:

Click here to see Updated LM723 Long Term Stability Results …


Updated Results for the Long-Term Tests of the LM723 Self-Heated Reference

The test setup for the heated references is shown here:


The reference itself sits on top of the 34401A (a metal box with a Pt1000 thermistor hinged on the lid of it). The pile of cables on the right comes out of the breakout box of the 20 channel voltage measurement card (inside the M300). Power supply and shunt resistor are at the bottom of the pile.

Test run two and three have been completed and they show some consistent behaviour. Back to test #1 (ca 4.5 days):


Test #2 (10.5 days) :


and Test #3 (10.5 days):


It seems that the LM723 creeps up about 4mV per 10.5 days.More info here:

Click here to see some LM723 Long Term Stability Results …

Do all LM723s exhibit this behaviour ? Or are chips from other brands or MIL chips better ? Lets see … The logger is running.