There are many DIY amplifier designs floating around. Some are simple chip based solutions, others are fill discrete circuis. They very greatly in quality but arguably one of the best DIY amplifier designs around is the Aleph-X.
The Aleph-X is a strange monster with an even stranger heritage. This is the spiritual child of Nelson Pass from the Passlabs fame, a manufacturer of high-end audio gear. Passlabs differs from other audio companies in that many designs have been openly published to the DIY community after an amplifier has been EOL’d. In addition some of the key design components of Nelson Pass’ amplifiers are patented with very detailed patent descriptions making them easy to follow. In May of 2002 Grey Rollins from diyaudio took two key patented designs, combined them and published a thread on diyaudio. Thus the Aleph-X was born. The last post on the thread occurred in July 2013 after accumulating 3226 posts discussing the design.
Note: This design is based on two patents by held by Nelson Pass and in active use in two different series of commercial amplifiers, the Aleph series and the X series. Patent law permits the use of this design for DIY purposes only.
With that out of the way, the Aleph-X is the combination of two key patented amplifier designs. The first patent covers the Aleph current source. A conventional class A amplifier has a fixed current source biasing the output stage into Class-A. The principle of the Aleph current source is that the source is variable, providing a fixed bias current at DC and modulating the current source with the AC waveform. In a conventional amplifier when supplying a positive signal to the load the output device stops conducting forcing the current from the current source to the load. On the negative swing the output device starts conducting shunting current away from the load. The Aleph current source on the other hand will modulate current with the opposite phase of the output device. Effectively the amplifier can be thus be biased at half the current of traditional designs and still without dropping out of Class-A operation.
The more interesting patent covers the “X” part of the Aleph-X, what has been affectionately called super-symmetry. This design differs from traditional amplifiers which have positive and negative inputs on a pair of stages with some negative feedback. Instead the design is laid out so that the two identical amplifier stages present negative inputs to the source and have their positive inputs tied together. This has several benefits:
- True use of balanced inputs as in a classic “dual mono” design.
- Negative feedback of each stage is fed back to that stage reducing stage distortion.
- The two stages are coupled together feeding the error signal of one stage, out of phase into the other stage. The result then cancels at the output yielding incredibly low distortion.
I won’t talk about how the amplifier works, that is covered extensively by others and there are a myriad of resources and modifications to the design. Instead this page serves as a documentation of my construction issues (the amp has excessive heat-sinking and power requirements) and specific modifications for high power. A custom addition to the amplifier is an auto-power off circuit and a speaker protection circuit. The amp does after all draw in excess of 450watts per channel.
The schematics for the amp are shown above. In red are the components changed from the original values. The original Aleph-X provided a power of approximately 40 watts per channel and were optimised for 8ohm speakers. I was aiming for something closer to 100watts and able to drive with comfort the ~3.5ohm minimum impedance of my speakers. A quick summary of changes:
- Output devices have been quadrupled to handle the higher power requirements and higher bias current without overheating or overloading the stage.
- Current gain changed from 50% from the Aleph source to 66% for improved current output vs voltage gain to play better with lower impedance speakers. This is a combination of the output resistors and the resistor divider in the aleph source.
- Total DC offset was stabilised by reducing output – gnd resistance, this this is at the expense of efficiency and power consumption.
- Start-up DC offset stabilisation rate was adjusted by increasing RMa and RMb, the speakers don’t see this DC offset normally anyway.
- Gain was increased by adjusting feedback resistors.
- C9 C10 C11 and C12 were added as the resulting amp was unstable (more below).
The powersupply was difficult to master. Firstly the chosen transformers are 1kVA +/-25V units which were custom ordered for this project. Most 1kVA transformers available to me off the shelf had secondary voltages >50V used normally by high gain amplifiers, but due to the balanced / differential design the amp is capable of twice the output voltage of a standard design. This is because both speaker terminals are driven with opposite voltages as opposed to one terminal carrying a voltage and the other being held at ground. The powersupply is filtered by a CRC network. The resistor caused thermal problems due to the excessive current in each transformer leg. P=I²R =64*0.15 = 10watts dissipated by the resistor. The diode bridge dissipates approximately 15watts as well. Both items ended up being bolted to the 5mm thick aluminium base plate of the amp for heatsinking.
Heatsinking presented the next problem. Calculations showed that I needed <0.09°C/W of heatsinking capacity to ensure the output device junctions remain below their 125°C. Initially all devices were bolted to a single large heatsink with silicon pads. While running the heatsink reached 75°C. The IRFP044 features an exposed junction facilitating direct temperature measurement. The junction temperature reached in excess of 110°C during winter. In summer the amp started oscillating and one channel required a rebuild. Back to the drawing board.
The pictures above shows the final design. The following changes got the temperature under control:
- Replaced silicon pads with mica pads and heatsink paste. This raised heatsink temps by 3-5°C but the junction temperatures decreased to 90°C.
- A second 0.18°C/W heatsink was mounted back to back with the first pointing into the case of the amp.
- A slow moving silent fan was added blowing into the case and out through the second heatsink fins assisting cooling the primary heatsink.
After these modifications the junction temps have dropped below 80°C.
Stability of this amplifier is dependent on circuit layout, part selection and construction. This was my major downfall as I had several months of battling stability related issues. Grey’s original design added two components C9 and C10 which decreased the bandwidth of the amp at the current source. Unfortunately in my case they didn’t work. The location of the MOSFETs as well as the choice of IRFP044 with it’s higher gate capacitance than the IRFP244 meant the current source wasn’t the only source of oscillations. I had to add capacitors across the output MOSFETs (C12 and C13) as well to further reduce the open loop gain of the amplifier.
The uncompensated output of the amp can be seen above. Without any input signal both sides of the amp were oscillating at 150kHz. While much of this didn’t make it to the speakers thanks to be cancelled by the differential design of the amp, oscillations rob the amp of power and can damage the output stage.
By playing with C9 through to C12 the oscillations ceased and the bandwidth of the amp was reduced to approximately 180kHz. The results of a square wave input above showed major ringing and overshoot (but no oscillations) so C2 and C4 were tweaked to slow down the amp response and reduce bandwidth. After raising the feedback capacitance to 12pF all overshoot was eliminated and the amp still has a bandwidth in the order of 100kHz.
Each monoblock draws in excess of 450watts sitting idle so it was critical for the amps to power themselves off when not in use. The bottom part of the circuit shown achieves exactly that. This is an independent circuit with its own power supply to monitor the signal inputs to the amp. U2B differentially amplifies the inputs with a gain of 100 and then feeds it into comparator U2C. When the amplified signal exceeds 1.3V, U2C’s output hits the 12V rail and charges C7. When there’s no signal C7 discharges through R18. If the voltage in C7 is above the threshold voltage of Q6, the MOSFET turns on and energises the power supply relay. Q5 cascades off Q6 to the speaker protection circuit to ensure that the speakers are disconnected while the amp powers on or off to avoid any thumps when power cycling.
The other part of the custom circuit shown above is speaker protection. The speaker outputs are fed into differential amplifier U2A which low-pass filters the signal at 1.6Hz since we’re only interested in preventing dangerous DC from getting to the speakers and not the sound. It also references the incoming signal to a new virtual ground at 6V as the rest of the circuit is powered by 0-12V and won’t tolerate negative voltages. This amplified DC is sent to comparators U1 and U3 and is compared with a 2 way voltage divider R1, R6 and R10. This arrangement creates a positive and negative voltage either side of R6 using the formula V = ± 12 × (R6 + 10k) / (R6 + 20k) – 6. These voltages set the threshold at which U1 and U3 will drive a negative output forcing the Darlington and relay to switch off and protect the speaker.
The amp is assembled on a main aluminium base plate which is bolted to the bottom of the heatsinks which have been drilled and tapped directly. Another aluminium panel is bolted to the front and rear of the amp. The stepped shape of the back to back heatsinks provide a ledge on which a perforated stainless mesh sits allowing air flow through the fan and out the heatsinks. The fan itself is mounted on a sheet of core-flute which sits snugly between the heatsinks on an aluminium bracket and directs airflow out the fins. All in all it fits well into any living room and sounds fantastic.