[John L Stewart]

This note describes a simple, low cost power supply improvement. It allows reduction of the power frequency ripple component (60 Hz in North America & 50 Hz in Europe) by up too 20 db. The problem this modification resolves occurs only in center tapped, full wave rectifier circuits. The conditions described are not found in a full bridge configuration. The experimenter needs to be extremely careful. Deadly voltages are present.

The test bed I used for this modification is an amplifier I built in the early 1960's. It uses a Hammond 273BZ power transformer. Nominal HV is 375 volts each side of center tap. The choke is a Hammond 156R, which along with the 35-microfarad capacitor forms the filter section at the output of the capacitor input rectifier. Load used during testing is a 5k, 50 watt resistor.

We would assume that the ripple would be predominantly even order harmonics ( 2nd, 4th, 6th, etc. However, because power transformers are wound serially, one side of the high voltage winding has a higher resistance then the other. This results in a significant ripple at the power frequency.

The LC filter network following the rectifier is only one-quarter (-12 db) as effective at the power frequency as at main ripple frequency. The DC available at the filter output could contain ripple at the power frequency as large or greater then that of the main ripple frequency.

All of this is of little importance to pushpull amplifiers since they have large immunity to power supply ripple (see Note). However, the single ended amplifier & particularly the triode amplifier (SET & SEUL) are adversely affected.

Depending on the loading conditions of a triode power stage, from 2/3 to 4/5 of this ripple voltage will be applied to the load. Unfortunately, most loudspeaker systems have one or two
resonance’s below 100 HZ. One of these could be at or near the power frequency. Presto!! We have objectionable hum in the listening room.

Figure One illustrates an ordinary center tapped full wave power supply. Two sides of the HV winding measured 38 & 40 ohms. Four resistors have been added. The purpose of R1, R2 & R3 is to apply a small AC voltage in series with the DC output voltage at the break point B1. By selecting the polarity the connection provides series aiding for the high resistance side of the HV winding & series bucking for the low resistance part. R4 is a current sampling resistor used during adjustment. The large dots shown on the transformer secondaries are polarity indicators.

Adjustment for optimum results is made by adjusting R2 while monitoring the current charging pulses with a scope at test point C ( TPC ). R2 must be wirewound since the current pulses are large. I used a Clarostat A4320. Adjust for equal size of the current pulses. An example of how they should appear is shown in Figure 2.

To get consistent information from the test scope it is important to use line triggering. The power cord of the test power supply should be polarized. A X10 probe AC coupled at the scope will keep the trace on the screen.

The diagram shows a solid state rectifier. I used an octal replacement unit from New Sensor Corp for these trials. Several vacuum rectifiers were tried with varying results. Their heaters were connected to the 5 volt winding in the normal way. Socket pin connections are shown.

Similar results could be had by using the 6.3 volt winding. In that case the mod could be inserted at break point B2. These same fixes could be applied to a power supply using a choke input filter.

C3 & C4 are ceramic & must be rated for direct connection to the power line. Mallory makes a capacitor approved by Underwriters Labs for this service. The part number is UN103M.

The ripple components at the input & output of the filter section were measured using an HP302A Wave Analyzer. The results in Table One show the reduction in the 60 Hz component. Higher odd order harmonics are also reduced.

Figure 3 is an FFT comparison of the output of the power supply before and after adjustment. The lower trace shows an improvement of about 20 db of the power frequency ripple.

For some there is an even easier fix. If the power supply uses a directly heated rectifier such as a 5R4WGB then it is possible to simply reverse the 5 volt heater transformer leads. No extra parts are needed. The difference will be very obvious when looking at the trace at test point C. The FFT measured about 15 db improvement. This is shown in Figure 4. By the laws of chance 1/2 of these rectifiers will already be properly connected.

Note- Since doing this work in 1996/1997 I’ve had a closer look at some PP amps. If the PS is unfiltered then 120Hz sidebands are clearly visible on the test signals used. This could explain the so-called mudiness, Etc in the rendering of program by some amplifiers. Some reputed to be high class at that. For those spectrums I was using a Pico Technology ADC-216, so 96 db resolution.


More- A simpler fix is possible when using SS rectifiers or tubes such as 5V4G, 5AR4, Etc with an indirectly heated cathode. Just add enough resistance in the lower R HV lead from the PT to make up the differance.

But I've noticed many of those using DHT such as 2A3, 300B, Etc prefer rectifiers such as 5U4G, 5R4G, Etc. Then the voltage drop across the filament becomes a factor.

The original article was published in Glass Audio, Vol 11, #3, May 1999.
A short version appeared in the June 1998 issue of Electronics World.