In the late 1980s I tested as many push-pull transformers as I could get ahold of. The test bed was a modified Heathkit W-6 amplifier with push-pull 6550s. The goal was to get a relative comparison of the quality and characteristics of what was available at the time. Copies of these test results have been floating around for years, and are now reproduced here. I was unable to dig up the original Lotus 1-2-3 files, so these are scans of the last version I made, from December 29, 1990.

Transformer Tests page 1 (PDF: 52Kbytes)

Transformer Tests page 2 (PDF: 48Kbytes)

Transformer Tests page 3 (PDF: 33Kbytes)

Transformer Tests page 4 (PDF: 396Kbytes) (includes sample waveforms)

There have been some misunderstanding on these measurements. The following is the text of a letter I sent to an inquirer in 1991 regarding the tests. It pretty clearly states the methodology and goals for these tests.

I am sorry I haven't gotten back to you sooner, but I have been busy, between finishing various projects and traveling to Asia for three weeks. Enclosed is the latest update of the transformer test listing. I have knocked-down the transformer test set-up, since the chassis it was built on was needed for other purposes, so I won't be measuring more transformers soon. I am still interested in characterizing output transformers, but will likely build a somewhat different test set-up, since I would like to test the transformers in a circuit with feedback.

The characterizations summarize some, but not all, key parameters about output transformers. An important one is the basic high frequency response, along with the square-wave waveform characteristic - this indicates the attention paid to reducing parasitic capacitance and leakage inductance in the winding. Also important is the minimum frequency that full power is possible - this indicates the amount of primary inductance and the quantity and quality of the core material. The Z p-p and Ultra Linear ratio verify the design of the transformer. Note that a discrepancy between the primary impedance between a 4 ohm load connected to the 4-ohm tap, and an 8 ohm load connected to the 8 ohm tap is often due to the transformer actually being designed for a 3.2 ohm tap. There is also some error due to inaccuracies in making the voltage measurements, but these were reduced in the measurements marked with a *.

When measuring an unknown transformer, particularly when it has been removed from its amplifier, I often had to try to deduce what the intended impedances of the secondary windings were, by either using an ohmmeter, or looking at the relative output voltages. In some cases, I may have guessed wrong; this may be why the Fisher 50A transformer has such strange specs. Running a transformer into the wrong load will alter its high-frequency roll-off and square wave response.

I have been criticized for driving all the transformers with the same pair of 6550s, since each transformer should be driven from it's correct driving impedance. Thus a transformer with a 10K primary that is driven by the lower impedance of the 6550s will be overdamped. This complaint is correct, but I was looking at testing the transformers in a uniform environment. The relative differences between transformers of a given impedance are still valid. The primary errors caused by an impedance mismatch are the frequency response and amount of squarewave ringing. If the transformer is used in a circuit with any feedback, these parameters are changed anyway. An area of future research is to see how well transformers behave under feedback. One of my preliminary findings is that transformers that have a smooth (not kinky) squarewave response, even if the -3 dB frequency is low, (such as the Marantz 9 and Acro TO-600) often behave better under feedback than fast transformers with kinks in the squarewave response (such as the Brook 12A or the UTC LS-55).

Related to the issue of squarewave response, my use of the "A, B, C..." ratings to describe the squarewave response has been nearly universally misused and misunderstood. When I first started these tests, I took scope photos of each transformer's response. As the quantities of transformers tested increased, and people asked that the results be put in a tabular form, I dropped the use of individual scope photos, since I found that most transformers fell into generic classes of waveforms. The enclosed sheet showing example waveforms was deemed sufficient. I started testing the better transformers first, so assigned them A, B, etc. Later I got to the worse ones, which tended to get the E and F ratings. These letters are not a grade, but are arbitrary assignments. People have misinterpreted these letters as a pure quality rating, placing "A" transformers at a much higher position than "D" transformers. In fact, Acrosound transformers typically get the "D" assignment (an underdamped ringing, with no kinks), yet do well in feedback amplifiers. As mentioned before, the squarewave response and -3 dB high frequency cut-off are only relative indicators of quality that are easily altered by how the transformer is used in a circuit.

Areas for future research include: measuring core distortion and losses, primary winding coupling (for class B applications), tests with feedback, and last, but not least, listening tests. I have been doing more work on trying to understand what exactly influences the sound of amplifiers, but am often surprised that amps with transformers that don't test well (such as the Brook 12A and many 6BQ5-based amps) sound so good. Often the transformer is not the dominant influence on the amp's sound, but it would still be worth finding out what influence they have. An amp that I am currently building will have provision for mounting four different 60-watt transformers (one at a time); it will be interesting to swap them and hear the differences.

I hope this helps. I am still interested in gathering people's experiences with transformers, in order to learn more about what makes tube amps sound good.
John Atwood

Go to One Electron Home Page