Modulus-686240W (8Ω), 360W (4Ω) @ <0.0002% THD
The Modulus-686 is the most powerful member of the Modulus family of amplifiers. Thanks to the Neurochrome Modulus error correction circuit topology, the Modulus-686 provides nearly 400 W into 4 Ω at vanishingly low distortion levels.
Feel free to join the discussion of the Modulus-686 on DIY Audio: Modulus-686 (DIY Audio).
Recommended Power Supply & Heat Sink
The recommended power supply for the Modulus-686 is ±36 V with at least 10 A (RMS) available per channel. Thus, for a traditional unregulated power supply, I recommend a Power-686 with a 2×24 VAC or 2×25 VAC, 350-500 VA power transformer. The Antek AS-4225, Hammond 1182S24, and RS Components P/N: 123-4040 are all good choices. Several additional choices are listed in the Modulus-686 Design Documentation.
The selection of switching power supplies suitable for the Modulus-686 is rather limited. The Mean Well RPS-400-27-C is the best suited candidate. The Mean Well supplies are available from Mouser, Digikey, and others. You will need two of these SMPS units per Modulus-686 board. The resulting amplifier will provide 130 W into 8 Ω when powered by the RPS-400-27-C.
The Modulus-686 is a Class-AB amplifier, thus dissipates a large amount of heat. It will therefore need to be fitted with a sizeable heat sink. For builds intended for music reproduction, my heat sink recommendations are as follows:
- ±36 V supply voltage: 0.25 K/W or better (lower)
- ±28 V supply voltage: 0.39 K/W or better (lower)
These recommendations are based on an ambient temperature of 25 ºC and a maximum heat sink temperature of 60 ºC.
The 4U/400 size ModuShop Dissipante chassis offers heat sinks with a thermal resistance of 0.23 K/W, thus is suitable for a stereo build of the Modulus-686, assuming one Modulus-686 module is mounted on each heat sink.
Based on the 4U/400 ModuShop Dissipante chassis, I have designed a chassis for a stereo Modulus-686 amplifier. You can find the CAD files for the chassis here: Modulus-686 Stereo 4U Chassis. Note that the files are provided as-is, where-is.
Those who find the 4U (165 mm tall) chassis size a tad too big, should consider lowering the supply voltage to ±28 V. This will allow the Modulus-686 to deliver approx. 130 W into 8 Ω and 200 W into 4 Ω and reduces the size of the heat sink and chassis considerably.
The ModuShop 3U/300 chassis size features heat sinks with a thermal resistance of 0.41 K/W would allow for adequate cooling of a ±28 V Modulus-686 build. Note that for ±28 V, a 2×22 VAC, 225-300 VA transformer should be used (per channel). The Antek AS-3222 or Hammond 1182P22 are both good candidates here.
The key features of the Modulus-686 are listed below:
- Fully balanced mono construction.
- 4-layer, gold plated, ROHS compatible circuit board. Professionally assembled in Calgary, Canada.
- Output power: 360 W (4 Ω) – 240 W (8 Ω).
- Ultra-low 0.00025 % THD+N (140 W, 8 Ω, 1 kHz).
- Ultra-low 0.00036 % THD+N (280 W, 4 Ω, 1 kHz).
- Ultra-low noise: 25.0 µV (RMS, A-weighted, 20 Hz - 20 kHz).
- Gain: +26 dB. Changeable by resistor option. Minimum gain: +20 dB.
- Differential/balanced input (can also be used with single-ended/unbalanced sources).
- On-board EMI/RFI input filter and ESD protection.
- Power supply agnostic circuit architecture. The Modulus-686 performs equally well on an unregulated power supply as it does on a well-regulated power supply.
- Board size: 8.25 × 2.30 inches (210 × 59 mm). The finished module measures 8.35 × 2.50 × 1.35 inches (Approx. 212 × 65 × 35 mm).
The Modulus-686 circuit board is available as a fully assembled and tested module as well as a DIY-friendly partially assembled option with all the surface mounted parts pre-populated. The cost of the additional parts needed to complete the SMD pre-populated board option is approximately $56, resulting in a total build cost of just over $300 per mono amplifier channel. Note that this cost does not include the cost of the mounting brackets, backplate, mounting hardware, and wiring harness.
The Modulus-686 circuit boards are professionally assembled in Calgary. My assembly vendor uses a pick-and-place machine for component placement and solders the boards using a 13-zone solder reflow oven. This procedure eliminates the issues with component cracking that you often see from the overseas assembly houses, resulting in a high-quality, high-reliability product. Thus, by relying on the professional assembly, I am able to deliver a better product at a lower price, while also increasing the chances of success for the builders.
The Neurochrome Modulus composite amplifier topology uses a precision amplifier to perform error correction on a less precise power amplifier. The Modulus-686 is a fully balanced composite amplifier, which uses an LME49724 to perform error correction on six LM3886es configured as a bridge-parallel power amplifier. This results in an amplifier which has the precision of the LME49724 and the power of six LM3886es. This error correction is the central point of the Neurochrome Modulus composite architecture. The composite design will correct for many types of error, including distortion and power supply induced errors.
The error correction circuit in the Modulus-686 has its own regulated power supply. Consequently, the power supply for the error correction circuit is clean and free of ripple, even if there is some ripple voltage on the power supply to the board. In addition, the error correction circuit (LME49724 and associated components) has its own power supply rejection (the PSRR of the LME49724 due to its design and architecture). The end result is that the error correction circuit will correct for any distortion and supply-induced errors in the LM3886es. This is done without introducing any errors of its own, while staying within the performance limitations of the LME49724. The end result is a very powerful amplifier with vanishingly low distortion.
As mentioned, the error correction circuit also corrects for power supply induced errors in the power amplifier. This makes the Modulus-686 indifferent to the type of power supply used. When operated at levels below clipping, the Modulus-686 performs as well on a well regulated switching supply as it does on an unregulated power supply. The block diagram of the Modulus-686 is shown below.
As mentioned in the Key Features, the Modulus-686 has a differential (balanced, XLR) input. There are two reasons for this:
- Differential signalling sounds better.
- Differential signalling measures better as it rejects hum.
Using differential signalling moves the ground connection between the various pieces of equipment out of the signal path. This results in a reduction in mains hum of about 90 dB (31,600×), which is nearly as good as you would get from an input transformer (but without the distortion of the transformer). Thus, I recommend using a differential connection to the Modulus-686. Sadly, many consumer and prosumer sources do not feature differential outputs. In those cases, I suggest using a pseudo-differential cable between the single-ended (unbalanced, RCA) source and the differential (balanced, XLR) input on the Modulus-686. These cables can easily be made by the savvy DIYer. They are also available commercially.
The specifications for the Modulus-686 are tabulated below. These were measurements of the first production sample.
|Output Power||240 W||8 Ω, < 0.01 % THD+N|
|THD||< -112 dB||1 W, 8 Ω, 1 kHz|
|THD||< -109 dB||140 W, 8 Ω, 1 kHz|
|THD+N||-112 dB (0.00025 %)||140 W, 8 Ω, 1 kHz|
|Output Power||360 W||4 Ω, < 0.1 % THD+N|
|THD||< -112 dB||280 W, 4 Ω, 1 kHz|
|THD+N||-109 dB (0.00036 %)||280 W, 4 Ω, 1 kHz|
|IMD: SMPTE 60 Hz + 7 kHz @ 4:1||-104 dB||100 W, 8 Ω|
|IMD: DFD 18 kHz + 19 kHz @ 1:1||-113 dB||100 W, 8 Ω|
|IMD: DFD 917 Hz + 5.5 kHz @ 1:1||-104 dB||1 W, 8 Ω|
|Multi-Tone IMD Residual||< -110 dBV||AP 32-tone, 100 W, 8 Ω|
|Gain||26 dB||Resistor programmable. +20 dB min.|
|Input Sensitivity||2.0 V RMS||200 W, 8 Ω|
|Input Impedance||48 kΩ||Differential and single-ended|
|Bandwidth||127 kHz||1 W, -3 dB|
|Slew Rate||16 V/µs||8 Ω || 1 nF load|
|Total Integrated Noise and Residual Mains Hum||25.0 µV RMS||20 Hz - 20 kHz, A-weighted|
|Total Integrated Noise and Residual Mains Hum||31.5 µV RMS||20 Hz - 20 kHz, Unweighted|
|Output DC Offset Voltage||< 2.0 mV||Typical performance|
|Residual Mains Hum||< -126 dBV|
|Dynamic Range (AES17)||> 122 dB||1 kHz|
|Common-Mode Rejection Ratio||91 dB||1 kHz|
|Common-Mode Rejection Ratio||67 dB||20 kHz|
|Damping Factor||330||1 kHz, 8 Ω|
|Damping Factor||120||20 kHz, 8 Ω|
|Dimensions||245 × 65 × 40 mm||W × D × H|
|All parameters are measured at a supply voltage of ±36 V unless otherwise noted.|
|Gain||20 dB||R19 = DNP|
|Input Sensitivity||4.0 V RMS||200 W, 8 Ω|
|Total Integrated Noise and Residual Mains Hum||15.2 µV RMS||20 Hz - 20 kHz, A-weighted|
|Total Integrated Noise and Residual Mains Hum||19.0 µV RMS||20 Hz - 20 kHz, Unweighted|
|Dynamic Range (AES17)||> 127 dB||1 kHz|
|All parameters are measured at a supply voltage of ±36 V unless otherwise noted.|
The performance of the Modulus-686 exceeds that of my Audio Precision APx525 audio analyzer. Thus, the THD+N graphs show mostly the noise of the APx525 source and the noise floor of the Modulus-686. Similarly, the THD+N vs frequency is mostly dominated by the noise of the measurement gear, which is all state of the art. The biggest take-home message here is that the Modulus-686 contributes only a minuscule amount of distortion, intermodulation, and noise to the input signal. It is as close to a straight wire with gain as you can get, and the measurements below confirm this claim. Terms like "transparent" or "wire with gain" are overused, but those really are the best descriptors for the Modulus-686. It is difficult to describe what "The Source Material, Amplified" sounds like. Open and natural, I guess. But those words are overused too... Rather than devolving into marketing babble, I'll let the performance measurements speak for themselves.
The measurements shown below were performed using a pair of Mean Well SE-600-36 switching supplies forming a ±36 V @ 16.6 A power supply. Note that these supplies are equipped with fans, thus are not desirable for use in home HiFi systems. They do work incredibly well as laboratory supplies, though.
The graph below shows the THD+N vs output power for 8 Ω load. The amplifier delivers 225 W at the onset of clipping and 240 W when the THD+N crosses 0.01 %. Note that the sharp jumps (aside from when the amplifier clips) are caused by range switching in the APx525. The THD+N vs output power plots basically show the THD+N floor of the measurement system.
Repeating the measurement with a 4 Ω load reveals:
Soft clipping starts at just over 300 W and 0.1 % THD+N is reached at about 360 W. The THD+N vs frequency plots for 200 W into 8 Ω and 4 Ω, respectively, are shown below. Note that the measurement bandwidth was changed to 60 kHz to capture at least three harmonics of 20 kHz. This also increases the noise bandwidth, hence the THD+N, of the measurement.
The Modulus-686 operates in Class AB, so the plot below may appear a bit out of place as it shows the THD+N vs output power and frequency measurement commonly found in data sheets for Class D amplifiers. I am including it here to showcase that the Modulus-686 performs 10-100× better than most Class D amplifiers.
Siegfried Linkwitz argues that the 1 kHz + 5.5 kHz intermodulation distortion (IMD) measurement is one of the measurements which is more indicative of the perceived sound quality. He bases this argument on the fact that IMD products in this measurement fall in the frequency range where the ear is the most sensitive (see the Fletcher-Munson curves for more detail). I think this argument carries a good amount of weight, so I measured the Modulus-686 accordingly. The measurement is shown below. Note that due to a limitation in the DFD IMD source of the APx525, the frequencies used must be an integer multiple of each other. Thus, I measured at 917 Hz (5500/6) + 5.5 kHz. I performed this measurement at 100 mW, 1.0 W, and as function of signal level. The three results are shown below. Note that the performance of the Modulus-686 is over 20 dB better than the performance of any of the amps shown on Linkwitz's site.
The more conventional IMD measurements are shown below. The two plots show the SMPTE (60 Hz + 7 kHz @ 4:1) IMD and DFD (18 kHz + 19 kHz @ 1:1) IMD, respectively. Poor SMPTE IMD is often indicative of thermal issues or power supply issues in the amp. The 18k+19k IMD is indicative of the loop gain available in the amp near the end of the audible spectrum, which can be telling of an amplifier's sound quality. The Modulus-686 provides world class performance on both of these measurements.
Audio Precision has developed a multi-tone test signal, which contains 32 tones from 15 Hz to 20 kHz, logarithmically spaced in frequency. This test signal sounds a bit like an out-of-tune pipe organ. It is basically the closest I can get to music with a deterministic test signal. Thus, I argue that this multi-tone signal should be used in an IMD test for the best correlation between measurements and perceived sound quality. I run this test at levels just below clipping. Note that even the tallest IMD components are down at the -110 dBV level. This is likely why the Modulus-686 sounds transparent. This measurement shows that it does not add anything (or at least extremely little) to the source signal, even at levels just below clipping where the amplifier is working the hardest. Also note that the amplifier output is completely free of mains-related hum or noise.
The Modulus-686 shows only a minuscule amount of residual mains hum. Note that this measurement was taken with the amplifier board sitting unshielded on a lab bench, thus, actual performance once enclosed in a metal chassis is likely to be better. This is the residual hum and noise of the Modulus-686 when powered by an SMPS:
This is the same measurement with the Modulus-686 powered by a Power-86 and Antek AN-5225 toroidal power transformer.
For completeness, I measured the amplitude response and gain flatness as shown below.
As mentioned in the Key Features, the Modulus-686 features a differential input. The common-mode rejection of this input is shown below.
Finally, the output impedance and resulting damping factor for 8 Ω load are shown below.