Difficulty level Medium
Load regulation Poor
Line regulation None
Ripple & noise rejection Good – Very good
Stability Good
Thermal drift Excellent
Output impedance Poor

The capacitance multiplier filter power supply is a very interesting topology which can have great results if designed well.  The name implies that it somehow multiplies capacitance – in some aspects it does act as if capacitance is multiplied however it doesn’t truly do this.  The principle of operation is to filter the power supply rail heavily based on a small current draw which provides us with a cleaned “signal”.  This “signal” is then fed into a bipolar transistor or MOSFET and the current is amplified through it to provide the output of the power supply.

The basic starting point of the supply consists of a simple capacitive filter: bridge rectifier followed by a capacitor in parallel (C1, C2).  After this we add a transistor (Q1), resistor (R1) and capacitor (C3) as seen in the figure below.

For the purposes of this article, we will assume the use of the following components:

Transformer: Nuvotem Talema toroidal; 2 Output Toroidal Transformer, 300VA, 2 x 25V ac.  Spec sheet
Bridge Rectifier: Vishay GBPC3506W-E4/51, Bridge Rectifier, 35A 600V.  Spec sheet
Filter Capacitors: Epcos 4700μF 63 V dc Aluminium Electrolytic Capacitor, B41303 Series.  Spec sheet
Transistor: On-Semi MJE15032G.  Spec sheet

Input to the transformer is via an AC power source operating at 50Hz.  Voltage is shown as 325.22V peak to peak which is about 230V RMS.

We will be using only 1 of the secondary taps on the transformer.  Looking at the spec sheet you will see that the primary coil has a DCR of 4.7Ω.  This is represented by R_Prim in the figure below.  For a voltage rating of 25V on the secondary, we have a DCR of 0.1667Ω and output current of 6A per secondary.  This is represented by R_Sec in the figure below.

The bridge rectifier is chosen because it is readily available and capable of high voltage and current.

The capacitors are chosen because they display excellent characteristics with very low ESR of only 39mΩ (shown as ESR and ESR1 in the schematic below).

Cx_Fig1_BJT_Schematic

How does it work?

Capacitors C1 and C2 provide ripple filtering and bulk energy storage as with the standard capacitive filter powr supply.  After these two capacitors we now place an NPN transistor with it’s collector connected to the capacitor output.  The emitter of the NPN transistor becomes the output of the power supply.

Resistor R1 which is shown as 47 ohm in conjunction with capacitor C3 shown as 220uF create a RC low-pass filter which the rail voltage passes through.  This filtered voltage acts as a “signal” which drives the base of the pass transistor Q1.  Transistor Q1 is a current amplifier – it allows a large current to flow from collector to emitter based on the “signal” presented to it’s base.

Output votlage & rail sag

The following graph shows the output voltage when faced with a constant DC load stepped from 0 Amp up to 4 Amp in 1 Amp steps.

Cx_Fig2_BJT_LoadStep

Ripple voltage for a 1 Amp load is around 130mV P-P which is significantly better than any of the other filtered power supply methods we have looked at so far.  As expected the ripple voltage gets worse with increasing load current, though even at a full 4 Amp DC load, the ripple is only 440mV P-P which is a great result!

The output voltage sag between load increments of 1 Amp seem to be near the 2V mark.  This is as a result of the output impedance being a bit higher.

The nice thing about this kind of supply is that the shape of the ripple in the graph above is very nearly sine wave in form.  This indicates significantly less higher order harmonics which is great!

The output voltage under load for a capacitance multiplier is typically a bit lower than that of a C-R-C supply because we have an additional VBe drop to compensate for in transistor Q1.

Output impedance of the circuit into a 1 Amp load

Cx_Fig3_BJT_OutputImp

Output impedance is slightly higher than expected because of transistor Q1 now also being part of the chain.  This low frequency impedance explains the output voltage sag we see with increasing loads.

Power dissipation

The transistor Q1 will dissipate power through it.  This will typically need a heatsink to keep the device within reasonable temperature limits.

The graph below shows the power dissipation through the transistor with a constant DC load stepped from 0 Amp up to 4 Amp in 1 Amp steps.

Cx_Fig8_BJT_PowerDissipation

You can see from the graph above that at 1 Amp load the transistor is dissipating around 1.5 Watt through it.  At a 4 Amp load the dissipation rises to an average of around 22 Watt through it.  A small PCB mounted heatsink may be enough with using this circuit for loads of up to 1 Amp, but anything more will require more significant heatsinking solutions.

Changing the BJT (Bipolar Junction Transistor) to a MOSFET

Next, let’s take a look at what happens if we change the transistor Q1 from an NPN BJT to a N channel MOSFET.  This circuit is identical to the first one with the exception of the transistor change:

Cx_Fig4_FET_Schematic

The MOSFET used here is a IRF540N by Infineon (International Rectifier).  Spec sheet.

Many other MOSFETs can be used as long as the MOSFET maximum Vds and current is suitable for use.  I chose this specific device because it’s easy to find, inexpensive and very important:  It has a nice low RDSon of only 44mΩ.

Let’s look at the output voltage and rail sag figure using the MOSFET.  The following graph shows the output voltage when faced with a constant DC load stepped from 0 Amp up to 4 Amp in 1 Amp steps.

Cx_Fig5_FET_LoadStep

The shape of the output voltage waveform and amount of ripple seems to be very similar to what we see with the NPN Transistor.  The output voltage ripple has reduced by a small margin – at 1 Amp load it is now down to only 100mV P-P with the 4 Amp load ripple coming in at around 368mV P-P.

You will also notice that the output voltage sag between load increments of 1 Amp seem to be near the 1V mark which is a significant improvement over the BJT.  This is as a result of the output impedance when using the MOSFET being lower than that of the BJT circuit.

Output impedance of the entire power supply – MOSFET version
(This includes the transformer and bridge rectifier at a current draw of 1 Amp)

Cx_Fig6_FET_OutputImp

The output impedance is significantly lower than the BJT version.  The low RDSon of the IRF540N has a lot to do with this difference.

Output impedance of the filter only – MOSFET version
(This excludes the transformer and bridge rectifier – power is supplied by a perfect voltage source with a current draw of 1 Amp)

This graph represents the output impedance including capacitor ESR and ESL but does not include trace or connection parasitic factors.

Cx_Fig9_FilterImp

This is a slightly pointless exercise for this filter because the perfect voltage source driving the filter simply bypasses the capacitors in the circuit, thus the graph above really only shows the RDSon of the IRF540N.

Power dissipation

The MOSFET Q1 will dissipate power through it.  This will typically need a heatsink to keep the device within reasonable temperature limits.

The graph below shows the power dissipation through the transistor with a constant DC load stepped from 0 Amp up to 4 Amp in 1 Amp steps.

Cx_Fig7_FET_PowerDissipation

You can see from the graph above that at lower currents the MOSFET dissipates more power than the BJT.  This is normally a result of the MOSFET having a higher Vgs then the BJT equivalent Vbe.  At higher currents this reverses and the MOSFET becomes more efficient.

At 1 Amp load the transistor is dissipating around 3.8 Watt through it.  At a 4 Amp load the dissipation rises to an average of around 18 Watt through it – around 4 Watt less than the BJT.  A medium sized PCB mounted heatsink may be enough with using this circuit for loads of up to 1 Amp, but anything more will require more significant heatsinking solutions.

Load regulation
(How well the PSU is able to maintain a steady output voltage with changes to the load applied to it.)

From the graphs above it can easily be seen that the output voltage sags with increasing load current therefore the load regulation of this type of supply isn’t good.

Line regulation
(How well the PSU is able to maintain a steady output voltage with changes to the input voltage.)

If the AC Mains voltage fluctuates this will directly impact on the output voltage because there is no mechanism to stop it doing so – this type of supply has no line regulation at all.

Filter

The combination of R1 and C3 form a low-pass filter which provides the cleaned up “signal” with which to drive the BJT or MOSFET.  By increasing the values of either of these two components the -3dB point of the filter can be lowered in order to clean out the output ripple and noise even more.

If you are using a BJT, be careful to use a value for R1 which allows sufficient base current to drive the BJT.

Noise

Any noise present or induced into the AC mains line will be damped by the transformer primary because it has resistance, inductance and a little capacitance.  The dominant factor which determine noise transfer in the transformer will be the large inductance of the primary coil.  This inductance causes rising input impedance with frequency, ie it is a low-pass filter.  Any high frequency noise on the line should therefore be reduced rather well.

On the secondary winding things are slightly different.  There are only 2 significant sources of noise: the noise the bridge rectifier introduces (switching noise) and the possibility of induced noise on the line.  When a power supply is housed inside a grounded metal case there shouldn’t be any significant amount of external noise which will affect the secondary winding or subsequent rectified supply line.

The potential exists for an AC power carrying cable which passes near-by the power supply to induce some 50Hz noise into the line.  With a bit of forethought and care in placement this shouldn’t be a problem.  In difficult situations you could always use shielded cables to carry the AC mains inside the case which will significantly reduce any stray magnetic coupling into the supply.

If some noise manages to be transferred onto the output stage of the power supply it is filtered by the R1; C3 low pass filter in the supply.  Because we are able to filter this “signal” driving the pass transistor much more heavily than the supply line itself this power supply has a “very good” rating for ripple and noise reduction.

Stability

Even though we are using an active device in the form of the pass transistor (Q1) in this power supply stability shouldn’t be much of an issue under normal operating conditions – the main reason being that we are not employing any feedback.

Thermal drift

Temperature has no significant impact on the circuit.  If using a BJT transistor for Q1 it will conduct the current through it easier with higher temperatures, though this shouldn’t cause any adverse effects.  In fact this should contribute to lowering the output impedance slightly which is beneficial.

As usual, capacitors suffer most with high temperature which drastically shortens their life.  A rule of thumb is that for every 10ºC rise in temperature the capacitor’s life is halved.

Output impedance

The output impedance of this type of supply is heavily dependent on the components used.  At lower frequencies the impedance is dominated by the bridge rectifier, transformer and pass transistor combination while at higher frequencies the impedance is dominated by the ESR of the capacitors C1 and C2 with the internal resistance of the pass transistor being added to the overall output impedance.

Conclusion

The capacitance multiplier filter is moderately difficult to implement only because it contains a transistor, but the circuit itself is very straight-forward.  The resulting performance in terms of ripple and noise reduction makes this a serious contender if you need clean power for your project.