Amp circuits for DIY audio

tcpip's picture
Submitted by tcpip on 16 June, 2013 - 13:13

I have collected below a set of power amplifier circuits which are DIY-friendly and of interest to me. They are interesting because:

  • they are tested -- people have built them
  • there are lots of information available about them, because many DIYers have built them or very similar ones
  • they do not use obsolete or hard-to-find parts; in fact they do not depend too much on the exact component, thus allowing you to substitute parts intelligently
  • they are easy to understand for a student of electronics; they are not esoteric or bleeding-edge circuits in any sense
  • they do not require tricky construction, not even SMD soldering

Most of them have an added advantage:

  • They have fantastic measured performance in terms of low distortion, excellent SNR, and good stability

Most of these circuits are from Randy Slone's books. One or two are from Doug Self's books. You can add others.

Naming convention for many of the circuits

Since most of the circuits listed here have been published in books, I am using some abbreviations to refer to them.

Chip-amp circuits

There are some very popular integrated circuits which contain a complete power amp or a large part of a power amp. In this century, the quality of full-amp-on-chips has improved to the point where some sound quite good and behave well. And the most-of-an-amp-on-a-chip family, a recent category, is delivering outstanding performance in terms of distortion, noise, power headroom, and reliability.

Complete power amp on a chip

The most well known of them among us DIYers are the following:

  • The LM4780: a single chip carrying two complete power amp channels, each rated at up to 60W
  • The LM3886: a single chip with one complete power amp channel, rated at about 60W. A close cousin of the slightly lower power rated LM3875.
  • The LM1875: a lower power version of the LM3886, delivering 20W

These are practically indestructible and can handle abuse of almost all kinds other than an over-voltage on the supply rails. Their distortion ratings, while excellent at lower powers, have a problem competing against the top-end discrete power amplifiers at high power.

Kits/ PCB: There are many sources of kits all over the world, of which one or two are:

Almost complete amps on a chip

These are recent products, and contain everything other than the output transistors (the OPS devices) on a single chip. One can build excellent high-power amplifiers by adding the OPS devices and a small handful of other components.

  • The LM4702 includes the input, VAS and driver stages of a stereo power amplifier. Add OPS devices and you get a very high performance stereo power amp.
  • The LME49830 is a single chip which can drive a MOSFET OPS. It includes everything other than the OPS and a few passive components. The resulting power amp can have rail voltages up to +/- 100V and can deliver THD+N figures of 0.0006% with careful PCB layout and construction. Other closely related cousins include the LME49810 to drive BJT devices.

These almost-complete-amp chips are able to deliver excellent power amplifiers rivalling any fully discrete class B power amps. Some of them can be biased high to even operate in Class A mode.

These circuits are not listed here, because they are available from their manufacturer's application notes.

Kits/ PCB are available for these too, but are more complex than the single-chip amps, because the OPS transistors and associated parts must be added on separately. Here are some kits:

  • LM4702 amp kit from Connexelectronic of China. One LM4702 driving two pairs of BJT transistors, giving 100W/channel with very low distortion
  • LME49810 amp kit from Connexelectronic again, with one LME49810 driving three pairs of BJT OPS devices and giving 300W with very low distortion
  • LME498x0 kits from Panson Audio of Hong Kong
  • Lots of kits from various sellers on Ebay

Discrete: single pair of OPS devices

The first group of circuits are the lower power, smaller ones, which have just one OPS transistor per rail (two transistors per amp channel). These amps will typically never generate more than about 100W per channel. They are usually simpler and smaller, and in simplicity and smallness, Rod Elliott's P3A takes the prize.

Rod Elliott's P3A: CF BJT OPS, one pair, 60-80W

This is a super-simple amp which gives excellent, very clean sound and is easy to build.

  • Power: medium power: 60-80W
  • Input stage: one differential pair with CCS
  • VAS: single transistor, single-pole compensation, bootstrapping via a "T network"
  • OPS: Complementary feedback (CF), one pair of BJT devices
  • Distortion: Very low: 0.04% total THD at power outputs of 1W to 80W
  • Kit/ PCB: See http://sound.westhost.com/project3a.htm

APADH 4.19: EF BJT OPS, one pair, 80W

This is Doug Self's Blameless design.

  • Power: medium power: about 80W
  • Input stage: one differential pair with CCS and current mirror
  • VAS: Darlington with CCS, single-pole compensation and over-current protection
  • OPS: emitter follower (EF) one pair of BJT devices
  • Distortion: Extremely low:
    • 0.0005% of total THD at 1KHz into 8 Ohms
    • 0.003% of total THD at 10KHz into 8 Ohms
  • Kits/ PCB: From The Signal Transfer Company, who are the only authorised vendors selling Doug Self's PCBs and kits
  • Derivative design: from Scopeboy here. Has two pairs of OPS devices, extensive protection circuitry, and other enhancements. Full schematics cialisfrance24.com and PCB designs included.

HPAACM 11.4: CF BJT OPS, one pair, 80W

This is what many believe is Randy Slone's derivative of Doug Self's "Blameless" design.

  • Power: medium power: about 80W
  • Input stage: one differential pair with CCS and current mirror
  • VAS: Darlington with CCS, two-pole compensation and over-current protection
  • OPS: CF, with single-slope SOAR over-current protection, one pair of BJT devices
  • Distortion: Extremely low:
    • 0.0000x% of second and third harmonic distortion at 1KHz
    • 0.0014% of second harmonic, 0.00052% third harmonic, at 50KHz
    • overall THD: 0.0009%

    This is probably the best measured distortion performance among commonly known Class B amplifiers, which is why Douglas Self named it the "Blameless".

APS 6.6: EF BJT OPS, one pair, 80W

I believe this comes closest to an EF OPS derivative of Randy Slone's derivative of Doug Self's "Blameless" design. Take the "Figure 11.4" above, replace the CF OPS with EF, and this is the result.

  • Power: medium power: about 80W
  • Input stage: one differential pair with CCS and current mirror
  • VAS: Darlington with CCS, two-pole compensation and over-current protection
  • OPS: EF with single-slope SOAR over-current protection, one pair of BJT devices
  • Distortion: Extremely low: 0.004% overall THD at 1KHz at rated power, 0.01% at 20KHz at 0.5W power (worst case conditions). SNR should exceed 100dB if well constructed.

APS 6.20: CF L-MOSFET OPS, one pair, 80W

According to Randy Slone, the L-MOSFET CF OPS circuit given in "Figure 11.6" (see below) is an attempt to design an L-MOSFET version of the "Blameless" ultra-low-distortion design. However, this one, the "Figure 6.20", probably comes closer to a "MOSFET Blameless" because it has just one OPS pair.

  • Power: medium power: about 90W
  • Input stage: one differential pair with CCS and current mirror
  • VAS: Darlington with CCS, two-pole compensation and over-current protection
  • OPS: CF with single-slope SOAR over-current protection, one pair of L-MOSFET devices
  • Distortion: Extremely low: 0.003% overall THD at 1KHz at rated power, 0.01% at 20KHz at low power (worst case conditions). SNR should exceed 100dB if well constructed.

Discrete: larger amps

The amps here have at least two pairs of OPS transistors (i.e. at least four OPS transistors per channel). These will all generate more than 120W into an 8-Ohm load, and will also handle high output into bigger loads (i.e. 4 Ohms or so) without increase in distortion.

HPAACM 11.6: CF L-MOSFET OPS, two pairs, 120W

This is Randy Slone's L-MOSFET derivative of his "11.4" design.

  • Power: high power: about 120W into 8 Ohms, 200+W into 4 Ohms
  • Input stage: one differential pair with CCS and current mirror
  • VAS: Darlington with CCS, two-pole compensation and over-current protection
  • OPS: complementary feedback (CF) with single-slope SOAR over-current protection, two pairs of L-MOSFET devices
  • Distortion: Extremely low: 0.006% overall THD

HPAACM 11.12: EF BJT OPS, two pairs, 140W

Randy Slone's emitter follower (EF) OPS based high-performance design.

  • Power: high power: 137W into 8 Ohms
  • Input stage: mirror image topology, two differential pairs with CCS and current mirrors
  • VAS: mirror image topology or push-pull VAS, each half with Darlington pair, cascode, two-pole compensation and over-current protection
  • OPS: emitter follower (EF) with single-slope SOAR over-current protection, two pairs of BJT devices
  • Distortion: Extremely low: 0.0029% overall THD at 1KHz, 0.009% at 20KHz

APS 6.15: EF BJT OPS, two pairs, 150W

Another emitter follower (EF) OPS based high-performance design from Slone. He called it his "favourite BJT amp".

  • Power: high power: 150W into 8 Ohms
  • Input stage: one differential pair with cascode stage, CCS and current mirrors
  • VAS: Darlington pair, CCS, two-pole compensation and over-current protection
  • OPS: EF with multi-slope SOAR over-current protection, buffer stage between VAS output and driver transistors (also called "triples" sometimes), two pairs of BJT devices
  • Distortion: Extremely low: 0.004% overall THD at 1KHz and rated power; 0.008% at 20KHz at low power

APADH 6.31: CF BJT OPS, two pairs, 100W

Doug Self's "load invariant amplifier" design, aiming to retain low distortion into large (i.e. low impedance) loads.

  • Power: high power: 100W into 8 Ohms
  • Input stage: one differential pair with paralleled transistor pairs in the LTP, CCS and current mirror, innovative low-noise feedback network
  • VAS: Darlington pair, CCS, single-pole compensation and over-current protection
  • OPS: CF with single-slope SOAR over-current protection, two pairs of BJT devices
  • Distortion: Extremely low: 0.0008% overall THD at 1KHz and 25W into 8Ohms; 0.004% at 20KHz and rated power into 3 Ohm, 4 Ohm, and 8 Ohm loads; SNR better than 96dB
  • Kits/ PCB: from The Signal Transfer Company

APS 6.21, the "OptiMOS": EF L-MOSFET OPS, two pairs, 140W

This is Randy Slone's "favourite amplifier". It incorporates soft clipping circuitry, which makes the super-sharp MOSFETs behave like valve amps when they clip.

  • Power: high power: about 140W into 8 Ohms, 200+W into 8 Ohms with higher rail voltages
  • Input stage: mirror image topology, with each half comprising a differential pair with cascode, CCS, and current mirror.
  • VAS: mirror image topology, with each half having a signal amplification transistor, a cascode, two-pole compensation and over-current protection
  • OPS: EF with multi-slope SOAR over-current protection, soft clipping, two pairs of L-MOSFET devices
  • Distortion: Extremely low: 0.005% overall THD at rated power, and worst-case THD of 0.04% at 20KHz at low power. SNR may exceed 115dB if well constructed.
  • Schematic: OptiMOS v4
  • Kits/ PCB: From Lede Audio of Australia

DAPA 3.14: EF BJT OPS, four pairs, 200W

High performance BJT design from Bob Cordell.

  • Power: high power: 200W into 8 Ohms, 360W into 4 Ohms, 560W into 2 Ohms
  • Input stage: one differential pair with CCS and current mirror
  • VAS: Darlington VAS with cascode, CCS, and overcurrent protection
  • OPS: EF triple with one set of 2SC3503/2SA1381 driving one set of MJE15032/15033, which drives four pairs of MJL21194/93 BJT OPS devices in (surprisingly) TO220 packages. No overload protection or current limiting. Bias current of each OPS device is 100mA.
  • Distortion: Extremely low:
    • 0.0002% at 1KHz, 0.004% at 20KHz, driving 200W into 8 Ohms
    • 0.0002% at 1KHz, 0.005% at 20KHz, driving 360W into 4 Ohms
    • 0.0004% at 1KHz, 0.0084% at 20KHz, driving 560W into 2 Ohms

DAPA 11.17: EF V-MOSFET OPS, two pairs, 125W

High performance MOSFET design from Bob Cordell.

  • Power: high power: about 125W into 8 Ohms, 200W into 4 Ohms
  • Input stage: cascoded JFET pair loaded with a differential current mirror
  • VAS: cascoded differential VAS with a Darlington cascoded current mirror and transitional Miller compensation which includes the input stage
  • OPS: EF triple with one set of 2SC3801/2SA1407 driving one set of MJE15032/15033, which drive two pairs of 2SK1530/2SJ201 V-MOSFETs. No overload protection or current limiting. Each OPS pair biased at 200mA quiescent current.
  • Distortion: Extremely low: 0.002% overall THD at 20KHz at 125W into 8 Ohms, and it does not increase at lower power. THD of 0.004% at 20KHz at 200W into 4 Ohms.

Class A

Here we deal in two modern Class A circuits.

HPAACM 11.13: Class A, EF BJT OPS, two pairs, 40W

Randy Slone's only Class A design. Emitter follower (EF) OPS.

  • Power: high power for Class A: 40W into 8 Ohms. Quiescent power dissipation 200W.
  • Input stage: mirror image topology, two differential pairs with CCS and current mirrors
  • VAS: mirror image topology or push-pull VAS, each half with Darlington pair, cascode, two-pole compensation and over-current protection
  • OPS: emitter follower (EF) with single-slope SOAR over-current protection, two pairs of BJT devices
  • Distortion: Extremely low, as is typical of high-performance Class A topologies:
    • At 1KHz: 0.00001% second harmonic, 0.000018% third harmonic
    • At 20Khz: 0.00017% second harmonic, 0.00013% third harmonic
    • Overall THD into 8 Ohms: 0.0018%
    • Overall THD into 4 Ohms: 0.0059%
    • DC offset at output: 11uV (and there is no trimpot to adjust)

APADH 10.19: Class A, CF BJT OPS, one pair, 20-30W

Douglas Self's amplifier which can operate in Class A or Class B, and its Class A operation shifts to Class AB if driven with large input signals.

  • Power: medium power for Class A: 20-30W into 8 Ohms
  • Input stage: differential pair with CCS and current mirror
  • VAS: Darlington VAS with over-current protection
  • OPS: CF, with single-slope SOAR over-current protection, one pair of BJT devices, complex bias generator switchable between Class A and Class B
  • Distortion: Extremely low:
    • Class-A. Less than 0.0003% at 1kHz, 20W/8 Ohms. (20kHz BW) Less than 0.0015% at 10kHz, 25W/8 Ohms. (80kHz BW)
    • Class-B. Less than 0.0006% at 1kHz, 25W/8 Ohms. Less than 0.004% at 10kHz, 25W/8 Ohms. (80kHz BW)
  • Kits/ PCB: from The Signal Transfer Company.

Building them

Anyone who has read Randy Slone's and Doug Self's books and has spent some time with a PCB design program like Eagle or Diptrace can design PCBs for these amps and construct them. All the issues in power amplifier construction so clearly described in Randy Slone's books will apply. I will not repeat them here, and I cannot add anything to those descriptions. Help and guidance is always available from DIYaudio.

Using them

The smaller amps

The smaller amps, with single OPS device pair per channel, are excellent when compared to commercial products costing $400 or more. The entry level integrated amps from Denon, Cambridge Audio (their Topaz range), Yamaha, etc. have a preamp with a remote control, but their power amp sections are almost always inferior to even the weakest performing amp in this set. The "Blameless" based designs and derivatives are absolutely top class in performance, and will compete with dedicated expensive medium power commercial power amplifiers. See this passage from the Cambridge Audio 640A integrated amplifier manual:


The design of any purist audio amplifier is mainly centred on two main areas, the Power Supply and the driver stage's ability to drive the output stage effectively. We at Cambridge Audio have researched the best possible ways to achieve the highest performance in these areas, at a sensible price.

The Azur 540A/640A topology uses the same tried and tested output devices that Cambridge Audio have used in previous award winning amplifier models, but many hours of research have gone into the study and development of the preceding stages. This driver circuitry is essentially a matched differential input pair, loaded by a current mirror and driven from a transient compensated current source driving a high beta cascoded voltage gain stage. The thermally compensated output stage is setup to inherently give optimum Class AB conditions (for greatly reduced cross over distortion caused by dynamic heating of the output dies). In addition the topology includes a further improvement to the driver stage consisting of a pure class-A follower to isolate the voltage amplifier transistors from the difficult loading of the output transistors. This increased current drive to the output stage combined with a novel transient feed forward circuit doubles the slew rate to 40V/uS

The successor of the 640A is the 651A, which seems to be selling for $800 or so. The 640A used to sell for $500-600. Now let's compare its feature list with the medium power circuits on this page:

  • matched differential input pair: check
  • ... loaded by a current mirror and driven from a transient compensated current source: check
  • ... driving a high-beta cascoded voltage gain stage: check
  • Thermally compensated output stage: check
  • pure Class A follower to isolate the voltage amplifier transistors from the output transistors: check
  • power output: 65W into 8 Ohms, 100W into 4 Ohms: check
  • THD: 0.005% at 1KHz at rated power, 0.07% at 20KHz at rated power: check
  • Frequency response: 4Hz to 80KHz: check
  • SNR: 92dB: check
  • slew rate of 40V/uS: check (in many designs)

I selected the CA 640A because (i) CA is an honest manufacturer which talks about their products without bullshit, and (ii) the 640A is a highly regarded amp in its price bracket. More glamorous brands than Cambridge Audio sell amplifiers of similar quality and features for twice the price.

I would like to use these single-pair OPS circuits in active speakers as embedded power amps. In that kind of application, I can even remove the overload protection circuitry, and the slightly lower power of these designs will not be a constraint -- they will be five times as powerful as what one may need in an active setup with line level crossovers. For instance, I cannot think of a better fit than a pair of Seigfried Linkwitz' top quality LX521 open baffle speakers driven by a set of eight or ten of Randy Slone's HPAACM 11.4, APS 6.6 or APS 6.20 power amps. This combination will be among the best music reproduction systems at any price. Linkwitz is easily satisfied when it comes to amps -- he is happy with the LM3886 chip amps. I feel his speakers deserve better.

Removing the overload protection circuitry from a MOSFET amp allows us to remove the source resistors in the OPS too, and this reduces crossover distortion in these topologies.

It is amazing how small a power amp PCB can be, for these circuits. One has to actually make a PCB and hold it one's hand to appreciate this. This reduces the work for building such amps, and makes the project more attractive for the constructor.

I would also use these circuits to build power amps for a smaller listening room, with, let us say, very high quality two-way standmount speakers. This speaker topology dictates that high power deep bass reproduction will not be possible, thus reducing power requirements. The rest of the musical spectrum can be handled with the highest quality with these power amps. Very rarely does a person need larger power amps in a domestic setup, specially in an Indian home.

The larger amps

The larger amps include the multiple-pair OPS circuits like the OptiMOS or Doug Self's Load Invariant Amp, and also the larger ones driven by the LME498xx family of driver chips. They are all probably of comparable quality.

These amps will be perfect for slightly larger rooms (by Bombay standards), e.g. larger than 20x15 feet, with floorstander passive speakers with louder bass requirements. Larger floorstanders, specially three-way designs, waste more power in the crossovers and may need more powerful amps to drive them well. They will also be perfect for sound reinforcement applications in commercial establishments, since the LME498xx-based designs can exceed 500W per channel.

Class A

Some people will not touch Class B amps -- they will of course only look at the Class A circuits. I have no problem with them, but I am happy with Class B. Where would I use Class A amps?

Answer: as the treble amp for a multi-way active speaker system with line-level crossover. The reasons:

  • Class B amps have higher distortion in the high frequency regions. Whether I can hear the difference between a Class A and a top-class Class B amp in the 2KHz - 20KHz region is debatable. But I like to believe that I can, or at least will want to try.
  • Class A amps have lower power output, because of power dissipation challenges. But an amp which only handles the higher frequencies will never need to generate more than a Watt or two. Therefore, the power limit is not a constraint here.

Someday, I will build such a system.

Attachment: 

Image icon OptiMOS schematic
Zircon - This is a contributing Drupal Theme
Design by WeebPal.