

August 1, 1995
Transformer distortion measurements, low distortion transformers:
All transformers (except air core ones) use magnetic material that present a varying impedance
with voltage level. This draws current that varies with level and therefore produce an odd
harmonic voltage drop across the source impedances, resulting in the output of the transformer
to have an harmonic content, defined as harmonic distortion.
1) A modem transformer distorts the receive signal.
2) It also distorts the modem's transmit signal.
The distortion of the receive signal is of no consequence. The transmitted signal is so much
larger than the receive signal, that distortion caused by the transmit signal dwarfs any receive
distortion.
In duplex operation, the modem transmits at the same time as a signal is being received. This
transmitted signal must go through the modem's load resistor, then through the isolation
transformer, before going out on line. Duplex operation requires the transmitted signal, across
the modem's load impedance, to be bucked out, allowing the received signal to be read.
Unfortunately, the signal across the load resistor, because it is feeding into the transformer, will
have a distortion component. However, the bucking signal comes from the modem's oscillator
and does not have the transformer induced distortion component that the transmitted signal
across the load resistor has. Accordingly the transmitted signal across the load resistor has only
the non-distortion component bucked out, allowing the distortion component to contaminate the
received signal.
To measure the transformer's distortion, drive the transformer's SECONDARY (load) winding with a low
distortion generator. This generator should have an output impedance equal to the modem's
LOAD resistor. If this is not available, shunt the generator with 50 ohms, and put in series with
the generator the load impedance value less 50 ohms. Load the transformer PRIMARY (line) winding
with 600 ohms.
It is recommended that the oscillator generate 300Hz and that readings be taken at varying
levels. Recommended levels are -9dBm to +9dBm in 3 dB steps.
There are two instruments that are used for distortion measurements, the distortion meter and
the wave analyzer.
The distortion meter measures the output voltage, traps the fundamental, and reads the balance
as the distortion voltage. In reality the balance also includes, noise, generator supplied
distortion, plus that component of the fundamental that is not trapped.
The wave analyzer tunes into the desired distortion frequency (third harmonic, 900Hz), and
measures it's level. Comparing this value to the load voltage translates into percent distortion.
Limitations: Both above methods allow the generator's distortion to effect the transformer
distortion reading. Both methods have as their limitation the effectiveness of their filtering. In
one case as a wave trap and in the other as a wave receive. In most cases the limitation is how
much one can spend for a low distortion generator and the most effective wave analyzer.
There is a way of measuring a -10dBm signal having -100dB distortion. The basic principal is to
buck out most of the fundamental, before filtering it out. That is, if 99% of the output voltage is
bucked out by inserting part of the generator voltage, the distortion components are, in
effect, 100 times (40dB) easier to measure.
Basically we put a potentiometer across the generator and adjust the wiper until the wiper to
ground voltage equals the load voltage. This is done by putting a battery powered voltmeter, or
scope, between the wiper and the output hot terminal, and adjusting the wiper for a null. A better
null is achieved by putting a variable capacitor from the wiper to one side, or the other, of the pot
and again adjusting for a null. Adjust both the wiper and the variable capacitor repeatedly for the
best null. Using small trimming capacitors and resistors easily allows 99.9% cancellation of the
fundamental. If the null meter is a scope, the reduction of the fundamental can be plainly
viewed, together with the ever present third harmonic and noise. It is reasonable to this point for
the fundamental null to be not too large compared to the harmonics plus noise, and substituting a
wave analyzer for the scope can accurately quantify the third harmonic voltage. Additionally,
this method also bucks out the generator's harmonic distortion.
Generally this method is required for meaningful readings of -10dBm 600Hz signals, but not
needed for 300Hz 0dB, and higher signals. This method is cumbersome, but essential when
levels and equipment limitations require it.
Howard Buschman, President

Home <www.bglabs.com>