From: Winfield Hill
Subject: Re: Help with audio o/p transformer feedback to inv OP Amp
Date: 14 Jan 2003 09:30:41 -0800
Organization: Rowland Institute
X-Newsreader: Direct Read News 4.20
> Turns ratio is 1:1 using modem type transformer.
> Oscillation is very high freq, looking at scope. I tried a 10pF cap
> across the op amp feedback resistor (thanks Win), and it did stop it,
> but when using a greater ratio of transformer feedback verses op amp
> feedback, the stage still oscillates. I tried a larger cap, but this
> introduced an unwanted phase shift at freqs higher than 1Khz.
> Im thinking I might be able to introduce the transformer feedback
> into the + input instead of mixing with the (-) input.
> I didnt think such a simple addition would be so difficult.
Jeff, what are you trying to accomplish? What are you hoping to
do by applying feedback around your modem-type transformer?
One point, people often forget just how good common transformers
really are. Specifically, if you servo-control the voltage at the
input to the transformer (e.g. with an opamp), the output will be
an exact copy at all frequencies below some upper limit (assuming
reasonable loads), undistorted and with full low-frequency response
(assuming a light load, see * below). At low frequencies the opamp/
driver/servo is required to provide high currents to magnetize the
core, and if measured, this current will show lots of distortion.
But, assuming a very low-Z drive impedance, this current has only
a slight effect on the output signal, via the copper-wire winding-
resistance drop. The point is, there are few problems here for
feedback to solve, so feedback will little improve the output.
OK, what about the high-frequency limit I mentioned? This occurs
at frequencies above that where the transformer's leakage inductance
(L_ell, measured with the 2nd winding shorted) causes an L-R rolloff
with the load resistance, and/or an L-C dropoff with the secondary
winding capacitance plus the load capacitance, if any. The former
L-R pole creates a -6dB/octave rolloff, a response dropoff that in
theory can be corrected by negative feedback. So in this specific
special case feedback is useful to extend the HF response.
By contrast, the L-C point has two poles and a -12dB/octave rolloff.
Usually it's not practical to solve or improve a sharp two-pole
rolloff with feedback, because of struggles with loop instabilities,
such as you are seeing. The bottom line is that if your transformer's
high-frequency response limit is mainly due to an L-C rolloff, you'll
get little or no benefit from feedback. [A better way to deal with
leakage-inductance is to minimize it through the use of some special
winding techniques, sub-1ppm accuracy results are easily achieved.]
The overall bottom line: assuming low-Z drive circuits, feedback is
often not necessary, nor very helpful for transformer performance.
* One exception to this argument is the case of a transformer driving
a power-hungry load, where simple resistive copper-wiring losses can
be significant. Here a separate feedback winding, or even an entire
separate (unloaded) feedback transformer can improve the accuracy of
the output. That's the case we were dealing with in our book, figure
4.90. But here 60Hz power is being delivered to a motor, and we know
60Hz is far below either of the transformer's high-frequency cutoffs.
Well, lunch break is over, so that's all for now on this big subject.