Back in 2010, when we were working at Gramofone GF01/GF02, we were determined to try as many power supply options as possible, and of course to check each particular option and its sonic effects in a controlled way. So we tried both switching and linear supplies, going there both passive and actively regulated routes, and combining conceptions changes with parts changes.
Once we definitely settled into linear campus, some of the most interesting points were about transformers. And here are the most important ones.
1. EI transformers sounded better than toroidals, for their better dynamics, more quiet background, and more natural tone.
2. Lower power transformers sounded better than higher power ones. This finding was totally opposite to what audiophile common sense would predict or recommend by the usual over-sizing recipe – higher power units sounded more thin and more bright, whereas lower power units sounded more smooth and dynamic.
3. EI transformers’ positioning was critical, as one might expect, due to their higher external electromagnetic dissipation; interestingly, Gramofone units have external supply units, physically well separated from active electronics, so supply couldn’t upset it, yet this issue with positioning also appeared when we dealt with mutual positioning of two transformers (we had one isolation unit, and another one “main” unit). We find this setup sounds best when two transformers were placed at 90° one with respect to the other.
Now, all this was extremely interesting, and it hugely counteracted the common audiophile wisdom. Several checks hence took place, in an effort to possibly find technical reasons behind, and frequency response measurement was the most revealing one. I took several samples here, all 1:1 (isolation) units, made of the same material, and put them into the measurement setup loop. Also, all these EI transformers were the same split bobbin construction – primary and secondary were completely separate coils sitting side by side, and not one on top of another, which excludes capacitive coupling between the windings. So, all else being the same, it was obvious that the higher power transformers had a higher bandwidth and, as we want mains transformers to act also as filters, passing by only 50/60 Hz, and stopping all above that as much as possible, this is unwanted performance. If we also consider that the higher power transformers dissipate more around, we are able to understand the previously said sonic disadvantages of higher power transformers.
Time to look at the graph. (Click on it to see the full resolution, please.)
PSU transformers’ frequency response
This same measurement also pointed out that the bandwidth of the toroidal unit, even if it had an electrostatic shield between primary and secondary, was higher than the bandwidth of any of these EI units. In addition, it apparently had some reactive components at the upper end, peaking just before its “natural” roll-off, and apparently suffering from capacitive coupling, which continued its frequency response above 70 kHz.
There is a stream among audio transformer designers that prefer toroidal above EI units for signal applications, due to their tight coupling and hence a higher frequency response – and this approach might have its time here. Still, it would be really interesting to see a toroidal transformer with controlled (and low) frequency response, for power supply purposes.
Now, we could sum up some things about EI vs. toroidal.
EI advantage:
1. Lower bandwidth (i.e. better mains filtering).
Toroidal advantages:
1. Lower external dissipation.
2. Lower mechanical hum – EIs are not always dead quiet, as toroidals usually are.
3. Less in size for the same power rating.
Obviously, we may want all these advantages at once, and there is no reason to claim them impossible to get, but remember: improvements can come only from actual research, and in reality designing work is a set of compromises. Low external field and low interwinding capacitance don’t come together. Or to rephrase: do you want your windings tightly coupled or fully separated?
Practically speaking, the final choice should be based on particular requirements. The transformer’s power does have its limits, so it is not recommended to go beyond its power specifications. Also, you should consider that in our applications transformer could not work right at its specified power rating, as if the current flew constantly. Its output is rectified and it must fill capacitors with 100 Hz (or 120 Hz) spikes, and these spikes must be adequate high current pulses. But once you take this into account (circuit simulators can be of help here), and fulfill these power rating criteria, you better start thinking “less is more”, and stay at that power. Going with a higher power transformer you only increase the transformer’s bandwidth and its external electromagnetic field. In other words: yes, there is the right power for every transformer used in the audio.
So, as said, some of these things probably counteract what you have learned about mains transformers’ sonic properties. But, as one fellow audio enthusiast wrote in the early days of the internet (when the internet was young and pure), “most of the things we could learn about hi-fi were simply wrong anyway”.