Dear readers; He who has read my test reports knows that I have a special liking to perform both standard listening tests on the device to be tested, and also with other test methods form the most objective opinion possible.

This is the reason I always have supplemented the listening test with a before/after-listening, when this has been possible, i.e. when I have not listened to a device placed first in the sound reproduction chain.


For a power amplifier this test may be done by loading the amplifier with an artificial load (a silent loudspeaker) corresponding as closely as possible to the loudspeaker that would otherwise have been loading the amplifier. The device uses a simple voltage divider to equalize the signal after the amplifier to the same as the input signal. By switching between the input and the output it is easy to hear the tonal nature of those amplifiers that color the music.
This method to test amplifiers is very effective. Of the vast number of amplifiers tested by me over the years, it has been easy enough to hear colorations from 9 out of 10 amplifiers, and then, the majority have been amplifiers of well repute. This has led me to present viewpoints where not only the magnitude but also the character of the coloration is taken into account.
I have too often found listening tests without references to be misleading, not only in the few cases where the tests are not done only to make the advertisers happy, which seems to happen in all hifi publications except for this one!


The before/after-test I use gives a firm link with reality since the comparison with the input signal is an absolute reference for how the output signal should sound.
However, from time to time criticism has been given about my artificial load - which corresponds to my loudspeakers, pi60 - has been a too easy load. Well, I am not impossible even if I have suggested that this load has given very significant results.
Therefore I have accepted the criticism and designed a new stereophonic artificial load that in a better way will cover the main part of the loudspeakers on the market, even those constituting a little more difficult load. I have not given way to the temptation to design the load according to the most extreme impedance monsters since such loads could knock out amplifiers that would be quite excellent to drive 95% of the loudspeakers on the market.


In short; the load corresponds to a difficult 8 ohm speaker, where neither the brain of the designer, nor the speaker he has designed, have been carried out with signs of a short circuit at any frequency. This means in plain English that the impedance does not apply to the specification of an 8 ohm speaker according to the DIN-norm, because such speakers don't grow on trees.
The following describes the impedance of the load more in detail:

* The statical minimum impedance is 3.5 Ohm at 5 kHz.

* The minimum impedance in the bass range is 5 Ohm at very low frequencies and around 200 Hz.

At 38 Hz you will find the top value 42 Ohm, corresponding to a closed box with a moving mass of 28 grams and a force factor of 10 N/A and finally a compliance of 1600 N/m. This compliance simulation also has been made excessively nonlinear in the artificial load. This will force the amplifier to deliver a strongly distorted current, just like with a normal loudspeaker, in order to deliver a correct voltage. This is a very hard task for many amplifiers!

* Furthermore, the artificial load has been given low impedances in conjunction with large phase angles. The impedance curve has a dip at 2000 Hz with the minimum value of 4.5 Ohm, just like many earlier English loudspeakers. The worst spot is just below this dip, where the impedance is 5 Ohm at the same time as the phase angle reaches 70 degrees!

The measured as well as the computer calculated curves are shown below:

Measured impedance


Calculated impedance
Phase angle, top and impedance, bottom.


The electrical circuit

(The unlinear component is made from an E coil former where all E's are turned in the same direction, and the straight pieces are all on one side, placed at a distance from the E-s determined by some strips of mylar tape. The coil is wound for 200mH when placed on a standard E-coil former, and reduced with this tape gap to 57mH. At 50W power input it becomes 30mH. Without the gap it would be too unlinear.)


Another unique feature with this test method is that you may choose what level the music shall be played at when it passes the test object. You do not need to listen at the signal level that stems from the test object, but the level to the loudspeakers is free to choose. This is especially important for analysing loudspeakers with heavy cross over distortion (that may be driven at tiny levels, regardless of a normal listening level) and also to evaluate amplifiers that compress the sound at high levels (where we can play very high signal levels through the test object and save the ears with more reasonable listening levels).


Another thing worth mentioning is that the switching function before/after must be of a non audible quality, or the entire idea with the test method has failed. The quality of this switch is very easy to check, by replacing the power amp, artificial load and voltage divider with a short circuit. Then no difference may be heard when switching the before/after switch.

Layout of the test connection.

© Text and illustrations, Ingvar Öhman.
This article was written originally in Swedish and published in Musik & Ljudteknik no. 1 - 1991.
Translated by Per Arne Almeflo, Sonic Design, with permission from the author.