What's good about a moving coil transformer? Support Information from Andy Grove

The question here could be broadened to “why use a stepup transformer for low level signals?” but the MCT will illustrate the point. First we have to look at the kind of signal level of a moving coil cartridge, to illustrate with slightly more impact I’ll use the Audio Note IO Ltd, which has a rated output of 30uV, this is pretty low. Now, there are some low noise valves around, but usually they have other problems, the ECC88 is a fine example, in theory due to its high transconductance it should have low noise, but this valve was intended for use at radio frequencies, and at audio there are other “excess noise” components which tend to delete most of the theoretical low noise figure. And, ECC88s are as microphonic as hell. The elements aren’t solidly braced, and the frame grid rigs at all sorts of high frequencies, the Russian 6922 type thing isn’t so microphonic, but can sound like an overcast day in Soviet Russia, and there are all kinds of glassy sounding ECC88/6922 types. In passing I would mention that usually if you use only one it isn’t so bad, use a whole string of them and you get the chrome plated glass sound.

OK, so the 6922 type is OK, but not great, then there are the other super exotic European and US telephone type valves like the 437A, not exactly common, and again it’s a frame grid type. There are similar types from Russia being used too. In my opinion if you’re looking for transistor type noise figures from active devices then use a transistor (or JFET), it’ll probably sound better.

For more normal types such as the ECC83, EF86 and so on the equivalent input noise at the grid is something like 15nV to 25nV per root Hz, well that’s just not low enough, the noise equivalent bandwidth of a phono stage is about 2kHz, so that works out at 600nV to 1uV, or around 30db.

How about transistors? I mention these even though I’m assuming that the readers of this article will be fairly purist. But lets see how they stack up: A really good bipolar device will have around 1nV/root Hz noise, the noise equivalent bandwidth of a phono stage is around 2kHz, so the equivalent noise at the base is about 45nV, which is -56dB, just about acceptable, not great though, and it’s likely that the old problem of excess noise will make the situation worse. JFETs are usually slightly worse in terms of noise, although the very latest devices approach that of the best bipolar transistors, and with a JFET you don’t have to worry so much about base current and biasing of the bipolar. In many cases I’ve seen schematics with a bloody great electrolytic capacitor at the input of the MC stage, groan…

Now, the solution to our problem is to use a step up transformer of around 1:100 ratio, much more than that and stray capacitances (mostly external to the TX) will start to cause problems, 1:100 is an impedance transformation ratio of 1:10,000, so you really need to use low capacitance cables at the output, and a low capacitance phono stage input, but this isn’t such a big issue. One way to look at this is that the transconductance of your input stage has been multiplied 100 times by the transformer. Or, you can look at the 1uV of the regular valve stage, and then divide that by 100 to get 10nV, well suddenly were beating the transistor. 10nV gives a signal to noise ratio of 70dB, which is OK. If we were to put an MCT in front of the transistor stage then the signal to noise ratio would be about 96dB, excellent, you’d be able to hear individual electrons moving about.

A well designed MC transformer will contribute negligible noise itself, the primary winding will have a DC resistance of fractions of an ohm, and similarly the high secondary impedance weighed against the winding resistance means that the winding contributes negligible noise. A secondary consideration is the isolation both at hum frequencies by breaking any ground loop, and at radio interference frequencies by the same, and by the electrostatic screen which any decent MCT will incorporate. RF interference often manifests itself with bipolar transistor stages by an increase in overall noise, and often by the dreaded breakthrough of the local taxi firm’s base to car radios (or if it’s a radio station, why is it always in a foreign language?)

The frequency response can be as good as you want it to be, as usual cost is a factor, and a good response of an MCT can be ruined by poor selection of cable and so on. I once saw a misinformed comment about Barkhausen noise in MC transformers, well Barkhausen noise doesn’t affect materials at such low levels, this only becomes relevant near the knee of the curve as the domains reorientate.

Step-up transformers are widely used in test instruments when extremely low level signals, especially from low impedance sources such as magnetic detectors of various kinds, are to be measured, and are used in the best microphone preamps in recording studios.

Andy Grove - Audio Note design engineer