Singleton Input Stage in Power Amplifiers

"Been there... done that... I can say I am in the middle of my journey discovering the benefits and drawbacks of the Singleton IPS in power amplifiers. It might seem strange to do not try to improve the Long Tail Pair Differential IPS as this one at least promises a lower H2 and can be used to null the output offset due to it's differential action..... But I am now rediscovering the lovely harmonic distribution possible with a selected single NPN input transistor ideally biased and running in class A."



C3 is the input capacitor needed because Q4 base current through loading resistor R5 creates an offset that might wander due to the output impedance of the preamplifier connected to the input.

R6 and C4 form the input filter that dictates the amplifier high frequency response. This input filter also isolates the base of Q4 from electromagnetic interferences that could cause instabilities.

Q4 is the input NPN transistor connected as a common emitter amplifier with the corresponding collector load resistor R3. Q4 is a KSC1845 Audio Low Noise amplifier with very low Cob 1.6pF and high Ft 110MHz.


The VAS is arguably the most important stage on a power amplifier as it is here that the major voltage amplification is produced. It is important that the open loop gain of the VAS be high so it provides a linear amplification. If only one resistive collector load is used to provide the needed voltage gain, we need to increase the resistor value to a point where the current Ic will drop and with it the transistor gm so we need some sort of active load.

In my design I opted for a bootstrapped resistor acting as a ccs for the VAS collector load as I found that it sounds better due to it's linear response.

R13 and R4 values where chosen to produce Q18 Ic at 6mA and the middle point of those resistors is connected to to the output via C6 (High quality bipolar electrolytic). This combination makes R13 to "see" the same voltage changes in both it's legs and creates a very high impedance ccs.

Q18 wired as a common emitter is the Voltage Amplifier Stage (VAS). To reduce VAS distortion, I opted to increase the amount of local negative feedback around the VAS through Cdom. Gain of VAS is gm x beta x Rc (gm beeing the IPS transconductance and beta is the VAS beta). To increase local NFB I increased the VAS beta by adding the extra transistor Q2 wired as an emitter follower inside the Miller loop.

Q2 is a "Beta enhancer" that effectively increases VAS beta and it's function is VAS linearisation by enhancing local NFB through Cdom.

Q18 is a 2SA1381 in TO126 case Audio Voltage Amplifier PNP High voltage Vceo = -300V with 3.1pF Cob and 150MHz ft.

D2 is a BAV21 fast low capacitance / high voltage clamping diode used to prevent VAS saturation so the amplifier can recover cleanly from clipping. It might be argued that this solution reduces high frequency detail and if the user does never stress the amplifier into clipping it might be left disconnected.


All power amplifiers need some sort of compensation that rolls off gain at high frequencies in order to maintain good phase and gain margins.

In this version I use C12 as Cdom (Miller compensation capacitor) to do the high frequency roll off. To minimize global NFB factor without increasing Cdom I use a nested feedback in the form of R24 and a Lead compensation capacitor C27.

Some might be in favour of TPC (Two Pole Compensation) but in this design I found the single miller capacitor to do a better job.


For this amplifier I chose to use a push pull LATERAL MOSFET output stage.

M1 and M2 are ECX10N20R and ECX10P20R EXICON Laterals Designed specifically for linear audio amplifier applications.

Power mosfets present an alternative to bipolar junction transistors (BJT) with significative advantages... Mosfets have high speed, freedom from secondary breakdown and are easy to drive to very high currents.

This design does not include source resistors in the power devices because they are not very effective in promoting bias stability or current sharing among paralleled devices and they are not required as part of the crossover distortion management.

Lateral mosfets need a lower Vgs turn-on voltage than verticals 0.7V for the 2sk1056 vs 4V for the IRFP240 so we loose less power when using the same voltage for the IPS and OPS.

Mosfets are easy to drive. While BJTs need input current to drive the base, Mosfets have a very high input resistance at DC so they only need input current to charge and discharge their internal capacitances. This also implies that Mosfets are free from beta droop at high currents (an issue present with BJTs where the beta can drop to as low as 20 when current approaches 10A) The ft of BJTs also droops at high currents while the ft of Mosfets actually tends to increase at high currents.

Vertical power Mosfets are capable of very high currents due to their lower Rds(on) and usually require protection from short circuits. Lateral power Mosfets have a smaller maximum current capability. This added with their negative temperature coefficient means that Laterals often do not need an active short circuit protection and can survive a short circuit until a fuse blows.

Lateral mosfets have lower input capacitance than verticals and a zero tempco near 150mA so for increasing temperatures it's drain current will keep going down over this current. This particularity makes this devices virtually immune to thermal runaway so I do not need a complex circuit to monitor output current that might leave it's sonic signature.

This Lateral Mosfet characteristic also helps because I can use a purely resistive voltage spreader instead of the usual VBE multiplier. The passive voltage spreader that actually biases the gates of the Mosfets needs a steady current to operate linearly and this is accomplished by a floating current source that feeds the current mirrors providing the current for Mosfet biasing.

Looking at Id / Vgs curves for Mosfets we see that verticals behaviour is exponential in nature while Laterals exhibit a square law form. this is one of the reasons I prefer the sound of Laterals as they have a more elegant THD distribution.

While transconductance is lower for Laterals, it is more stable with increasing drain current so transconductance droop in the crossover region in a push pull class AB power amp is easier to circumvent with Laterals.

Q13 15 16 17 with the biasing resistors form the floating ccs.

Q1 3 9 10 form the mirrors that feed the ccs current (6.5mA) into the mosfet biasing resistor and trimmer R17 and VR2.


The signal coming from the VAS is fed to the Mosfet's gates via a special resistor arrangement R31 R32 that centers the signal into the biasing resistor R17. This innovative solution gave a significative THD reduction. R14 is included to minimize risk of oscillations due to PCB layout.


Feedback in this case acts to reduce amplifier distortion that principally originates in the output stage.

A small portion of the output signal is fed back to the input stage in inverted form. In this amplifier, the NFB is injected in a low impedance node in the IPS (Signal is fed to the emitter of the singleton divided by the ratio R21/R25 establishing a gain of 21.5X or 26.7dB).

We are dealing with a current feedback amplifier with this setup so R21 is a medium power resistor.


Amplifier output voltage offset is done by the injection of a correcting voltage on the emitter of the input singleton.

Q8 and it's surrounding resistors form a VBE multiplier biased by the zener diode D1.

Correcting voltage is fed to the emitter of Q4 by means of an isolating resistor R26.

To avoid temperature induced voltage fluctuations, Q8 and Q4 must be matched and physically connected with thermal compound.

After trimming, offset remains rock stable with this arrangement.


The Singleton IPS along with the beta enhanced VAS with bootstrapped CCS directly feeding the Lateral Mosfet OPS biased with a floating CCS does produce a high definition sounding amplifier with a sweet extended treble while enabling the listener to hear all nuances in the mids and lower mids.

This is the ideal complement for a high definition front end based on analog or digital sources.

Please feel free to comment and I will try to answer any questions that you might find relevant.

Major Influencers for this project:

Bob Cordell | Dr. Douglas Self | Gareth Ingram | Hugh R. Dean | Joachim Gerhard | Frans De Wit | Keantoken | Michael Børresen | AndrewT | Rod Elliot | Nick Salamouras

Text by the designer: