In my last Picoscope 6424E RoadTest blog I was testing a power supply which uses a hand wound custom transformer.

shabaz asked about the transformer, and specifically how it might behave at lower frequencies (25kHz a rather than the 150kHz I tested at.)


It's not easy to test the transformer in situ in the PSU, so I would need to make another one to test it and that meant that I might as well design a new one just for testing.

The 6424E has some tricks up its sleeve which helped a lot with the testing (which I wasn't actually expecting when I started !).


One of the problems is that I need a power amplifier capable of testing over a frequency range of at least 25kHz to 150kHz.

(To test the transformer we need to drive it at working power levels.)

Standard audio amplifiers are often OK at 25kHz but not many work at 150kHz.

A couple of years ago I did some work on an amplifier that was intended to mange 10W at up to a few MHz. I made a prototype, which was promising but I never got round to refining it or building the active offset null controller it was intended to have.

With a bit of care it should be able to drive 6W or so through a 1:1 transformer with a 5 ohm load.

I wound the transformer so that it should be OK at 25kHz and should still have low copper loss at 150kHz.

Prinmary and secondary are both 16 turns of 4 strands of twisted 27SWG (0.4mm) varnished copper wire. The core is an EPCOS (TDK) epoxy insulated N87 type ferrite. (Farnell 2673405), selected for cheapness, availability and high inductance per turn. This latter feature is important if you mean to wind by hand!

The competed transformer measured like this:

inner winding resistance:     19.3 mOhm

outer winding resistance:     21.4 mOhm

inter winding capacitance:    50pF

A very blurred picture of the primary winding - subsequent pictures were taken using the camera stand !

Primary bound up with polyester tape read for the secondary.

Putting the polyester tape on is worse than winding the wire.

The completed transformer in the test rig with its 5 ohm load (two 10 ohm TO220 resistors in parallel).

The complete test rig.

The Probe Positioning System was brought into play again to support the amplifier board. The little vice in the centre left is holding a small fan. This was introduced when I realized that even driving the load at 3W had the amplifier up to 90C in the hottest spot. The fan got it down to 55 or less, even when driving 6W into the load. You can just about see the current measuring shuntin the bottom right corner. This is a Kelvin connected 0.39R current sense resistor on a scrap of strip-board.

Test set up, the power amplifier connects to J1.


The power amplifier doesn't have its planned active offset controller, so the DC offset is dreadful (300mV or more). C1 isolates the transformer from the DC bias which would otherwise result in a static current of several amps. R1 senses the current in the primary. Channel D is used to measure the primary current, channel C the primary voltage and channel A the secondary voltage. For the purpose of the test we assume that the 5 ohm load resistor is both accurate and a pure resistance.


Only one PicoScope screenshot is needed to tell almost the whole story:

Picoscope setup used for the transformer loss measurements

On the Scope1 view there are two traces, A, in blue which is the secondary voltage measured across the load and the black trace, with a scale on the right in W and peaking to 12W. There are two measurements at the bottom of scope 1, channel A AC RMS and A*A/5 DC Average. The 6424E can do maths on the channel data to make a "Maths channel" and then do measurements on the maths channel. (BTW, to get average power you average the power channel it's already been squared by A*A so doesn't need it don again !)


We do a similar trick in the Scope3 view. only this time we use (C-D)*D/0.39 - C is the voltage across R1 and the primary, D is the voltage across R1, (C-D) is the primary voltage and D/0/39 is the current. Multiplying (C-D) by D/0/39 gives us the power into the primary.

Pop up window for editing maths channels


On the screenshot the input power is 5.929 W and the output power is 5.898W so the total loss is 31mW.

We can read the peak primary current from the D trace and its 0.6V/0.39/1.414 A RMS = 1.088A, so the resistivity loss in the primary will be about 23mW.

The secondary loss will be similar but slightly higher (25mW) - so we can show that the loss must be greater than 48mW but we didn't see that.

Rats !

The reason is almost certainly due to tolerance errors: The load resistors and current sense resistors are only nominally 1% parts and we are using x10 scope probes with their own errors. The power loss we measure is about 0.5% but the measurement uncertainty of the test rig is more like +/- 5% - so all we can say is that the power loss is small.


I measured the transformer at two power levels and over a range of frequencies.




Then, just for fun I measured it at 1MHz:


At 1MHz the leakage inductance is coming into play and there is significant phase shift  between the primary voltage and current. This means that my power amp can't drive the full 6W into the rig.

The loss has gone up, the input power is 4.975W and the output power only 4.674W, so the transformer loss is up to about 0.3W.


Overall this has been a very useful experiment.


We can see that the transformer optimum frequency is at about 40kHz, although there is little change between 25kHz and 150kHz.

The transformer is fine with 6W, and probably good for a lot more.

The Probe Positioning system has proved its worth again, happily supporting a board that was literally too hot to touch. We didn't use the probe holders.

The 6424E's ability to do maths channels, and measurement processing on maths channels, made things a great deal easier than they would otherwise have been.


The amplifier should get an honorable mention too - it's been sitting in a box for 2 years and at last it has done some useful work