Twin- T Filter For Audio Distortion Measurements

It is difficult to beat the twin- T filter for low audio distortion measurements, given its simplicity.

The twin- T filter (fig. 15) belongs to the group of non- minimum phase shift circuits- they have several interesting properties: The input signal is split in two (or more) parts, each following a different path and being subject to different processing, before being recombined at the output. In this particular example, the top limb is capable of providing in excess of 90^o of output voltage phase lead, while the lower limb can deliver more than 90^o of phase lag. Thus, at some frequency, the currents into the output node from each limb are exactly 180^o out of phase and have the same magnitude- they cancel out completely.

The filter should be driven from a low impedence source and terminated in a high impedence: For the values shown, any oscilloscope will do to monitor the output; Also, all frequency generators/ preamplifiers/ amplifiers qualify as far as driving the filter input is concerned (amplifiers should have an 8- Ohm load connected to their outputs.)

A twin- T filter was built using 1% components, and delivered a pleasantly surprising voltage attenuation level of almost 70 dB’s at the fundamental frequency (1kHz.) While the absolute values of components will only affect the frequency of maximum dip, not its magnitude, the impedance ratio of 1:2:2 must be preserved intact. Of course, such performance is not achievable if the filter is expected to deliver its best performance spot- on at a specific frequency: Rather, the input generator must be swept in an interval containing the nominal frequency of maximum rejection until the actual frequency of the dip is established. It may be possible to improve upon that figure, by making the horizontal resistors variable, and using a parallel combination of smaller value capacitors and/ or a trimmer for the horizontal capacitors. However, this has not been tried for several reasons:

It was thought that at such levels of attenuation, non- ideal component behaviour could become dominant.
It is quite acceptable to subtract the fundamental frequency residue from the total filter output in order to measure the distortion products more comfortably (provided the products are not much lower than the residue.)
In the first instance, the filter was employed to assess the distortion of a home- made preamplifier which delivered a distortion figure of some 0.9% at full swing, so the magnitude of the dip was quite adequate.

It is unreasonable to expect such a simple filter to obliterate the fundamental frequency, but allow the not- too- distant harmonics to pass unscathed, so the following table shows the attenuation exhibited by the circuit at multiples of the dip frequency:

Frequency Attenuation (dB’s)
2f 9
3f 5

4f 3

5f 2.3

6f 1.7

Harmonics which are an odd multiple of the base frequency may be largerly innoffensive to the ear, since they do not disturb waveform symmetry. Apart from the obvious advantage of being armed with figures before comparing different generator/ preamplifier/ amplifier comfigurations, it is also possible to determine how much distortion can be tolerated by different people at different frequencies.

More links:

Site Menu (Non- frames version):
Page Index (Frames version)- at end of section.