MyRef Fremen Edition
As you have probably noticed, my reviews are free of advertising. Instead of distracting you with annoying ads, I kindly request your donation. If you find the contents of this page to be useful, please consider making a donation by clicking the Donate button below.
MyRef Fremen Edition
- Build Guide
- Summary of Measurements
- Performance Graphs
- Performance with Recommended Power Transformer
- MyRef vs Modulus-86
Overall, the MyRef FE is a quite capable amp. It is a bit noisy and sensitive to RF interference, but it provides ample power and low distortion when powered by a laboratory power supply. Unfortunately its performance does degrade a bit when powered by a transformer. The degradation is unlikely to be audible, but I find it unfortunate that some performance was left on the table. That said, it is likely possible to remedy this with a relatively small PCB layout change. I encourage those who wish to improve upon their MyRef FE to read my article on grounding of the LM3886.
If another layout revision is planned, I would also encourage the layout engineer to provide mechanical support for the vertically mounted components in TO-220 packages and to allow easier access to the mounting screw of the LM3886. I also suggest including an EMI/RFI filter to prevent interference from RF sources such as cellphones.
I do like the simplicity offered by the MyRef FE regarding the choice of power transformer. The recommended transformer will allow for the maximum available power of the MyRef FE and will work will with both 8 Ω and 4 Ω loads.
The My_reference power amplifier was originally designed by Mauro Penasa and released for DIY use. You can find the design documentation for it on DIY Audio. It has garnered a bit of a following, in particular after Mr. Penasa's death.
As shown in the My_reference design documentation, the architecture of the My_reference amplifier is a bit different than your typical audio power amplifier. Its output stage is a Howland current pump based on the LM3886. An LM318 opamp is used for the input stage and the combination of the two form a composite amplifier. Due to the Howland current pump output stage, the MyRef FE is not as efficient as other Class AB amplifiers. For example it only provides 40 W into 8 Ω from a ±35 V supply versus the nearly 60 W provided by a regular LM3886-based amplifier under the same conditions. That said, if efficiency is your main concern a Class AB amplifier would probably not be your first choice.
One popular incarnation of the My_reference amplifier is the My_Ref Fremen Edition or MyRef FE for short. The MyRef FE uses Mr. Penasa's original amplifier circuit and adds speaker protection circuit and power supply. The PCB layout is by Dario Inserra ('ClaveFremen' on DIY Audio) who also organizes the group buys of the circuit. Dario was kind enough to provide me with a PCB for review.
The fully assembled MyRef FE is shown below. The PCB is Rev. 1.6. The most recent revision at the time of writing is Rev. 1.8, but the differences between Rev. 1.6 and 1.8 are minimal. The main difference is that Rev. 1.8 allows for the use of opamps other than the LM318 used in Mr. Penasa's original circuit.
There are several parts lists (BOMs) for the MyRef FE. This allows you to build the amp with different tweaks to the frequency compensation and with/without the use of SMD components. On Dario's recommendation I followed the "Evo A Mod" BOM. The parts cost is about $85 per board.
The PCB layout is pretty decent overall but assembly could be facilitated with a few mechanical changes.
The LM3886 needs to be mounted to a heat sink, but access to the mounting screw is blocked by several components, including the two power supply capacitors, as seen below.
The official remedy is to create a stud using a grub screw and mount the LM3886 with a nut, which is possible but not convenient. Some have also suggested using a socket head cap screw tightened by a hex key cut to fit the allotted 10-15 mm of working room. A better solution could be to use an Aavid MAX08NG mounting clip. This will likely allow for access to the mounting screw for the clip with a ball-end hex key.
The components are packed together very tightly, which I generally like. Unfortunately, the compact layout has the spade connector footprints boxed into some pretty tight spots as seen below.
In particular the ground connection between the two capacitors is inaccessible without tools. Should you ever need to disconnect the ground connection, you will have to pull on the wire, which is generally a bad idea. Thankfully, the connector footprints include a larger hole in the centre, which allows for soldering a wire directly to the board.
The PCB is 2.0 mm in thickness for "vibration control". This is a curious choice as there is no evidence to indicate that controlling vibrations or mechanical resonances has any impact on the circuit performance in a solid state amplifier such as the MyRef FE. It is unfortunate, however, that the concern for vibration tolerance was not carried through to other mechanical aspects of the PCB layout.
As seen in the pictures above, the MyRef FE contains several vertically mounted components in TO-220 packages standing unsupported. I prefer to have the components mounted flat against the PCB and secured with a screw to prevent the component pins from failing due to metal fatigue caused by vibration. Alternatively, the three components with small heat sinks could be secured to the PCB using heat sinks with solder pins. The diodes in the power supply could probably be secured with some well-placed blobs of RTV. The 2.0 mm thick board also prevents the two snap-in type power supply capacitors from snapping in properly as the capacitors are designed for use with the industry standard 1.6 mm board thickness. This makes mounting these capacitors flush with the board a bit tricky.
Those who enjoy tweaking circuits will be delighted that the MyRef FE contains footprints for both surface mounted and through-hole resistors in many places. Many of these resistor footprints have polarity markings for the resistors for even further tweaking. Note, however, that resistors are not polarized. The polarity markings are even more amusing when you consider that only two of the marked resistors conduct DC current. These two resistors are part of the ±14 V regulator for the LM318 opamp. The rest of the polarity-marked resistors only ever see AC current, i.e. conduct current equally in both directions throughout the signal swing. Rotating these resistors will, thus, have no impact on the circuit performance.
Do note that a small handful of components in the MyRef FE are SMD-only. If soldering surface mounted parts is not your cup of tea, you may need some help soldering those parts.
The MyRef FE Build Guide is quite thorough and covers most aspects of the assembly of the PCBs along with some suggested chassis layouts. It also covers various audiophile tweaks of the circuit. Beginners will appreciate that the build instructions include information on how to select solder, how to bend the component leads to fit the board footprints, and how to orient the polarized components correctly.
Note that some revisions of the MyRef FE PCB have errors. Revisions 1.10 and 1.7 have wiring errors on the PCB. That's a bit disappointing but, thankfully, the errors are easily fixed by adding a piece of wire in the appropriate spot. All revisions up to Rev. 1.6 have the polarity on a capacitor marked incorrectly as well. The instructions provide detailed information on how to address these errors. Builders – especially of the older revisions of the boards – should read the instructions closely before assembly.
The instructions could benefit from better and more consistent photography. A mention of electrostatic discharge (ESD) and how to prevent ESD damage to the components during assembly would be a good addition as well. It would also be nice with a proper test procedure for the first power-up. In addition to measuring the DC offset, I suggest also measuring the gain of the amp, which is easily done with a multimeter and an appropriate test tone.
I do appreciate that the PCB and Build Guide include some test points where the DC operating points of the circuit can be verified. This will be handy for fault-finding.
Summary of Measurements
The measurements of the MyRef FE are summarized in the table below. For these measurements, the MyRef FE was powered by a pair of HP 6643A laboratory power supplies providing ±35 V. I chose this power supply to highlight the performance the MyRef FE is capable of when provided a clean power supply.
|Output Power||40 W||8 Ω, < 0.01 % THD+N|
|THD||-111 dBc (0.00028 %)||1 W, 8 Ω, 1 kHz|
|THD||-108 dBc (0.00040 %)||40 W, 8 Ω, 1 kHz|
|THD+N||-106 dBc (0.00049 %)||40 W, 8 Ω, 1 kHz|
|Output Power||70 W||4 Ω, < 0.1 % THD+N|
|THD+N||-103 dBc (0.00067 %)||67 W, 4 Ω, 1 kHz|
|IMD: SMPTE 60 Hz + 7 kHz @ 4:1||-102 dBc||40 W, 8 Ω|
|IMD: DFD 18 kHz + 19 kHz @ 1:1||-109 dBc||40 W, 8 Ω|
|IMD: DFD 917 Hz + 5.5 kHz @ 1:1||-99 dBc||1 W, 8 Ω|
|Multi-Tone IMD Residual||-130 dB Ref.: 40 W||AP 32-tone, 40 W, 8 Ω|
|Input Sensitivity||570 mV RMS||40 W, 8 Ω|
|Bandwidth||2.9 Hz – 84 kHz||1 W, -3 dB|
|Slew Rate||±10.5 V/µs||8 Ω || 1 nF load|
|Total Integrated Noise and Residual Mains Hum||54 µV RMS||20 Hz - 20 kHz, A-weighted|
|Total Integrated Noise and Residual Mains Hum||69 µV RMS||20 Hz - 20 kHz, Unweighted|
|Output DC Offset Voltage||< ±4 mV||Typical performance|
|Dynamic Range (AES17)||107 dB||1 kHz|
|Signal-to-Noise Ratio||108 dB||40 W, 8 Ω|
|SINAD||98.8 dB||5 W, 8 Ω|
|Damping Factor||210||1 kHz, 8 Ω|
|Damping Factor||152||20 kHz, 8 Ω|
|All parameters are measured using a pair of HP 6643A lab power supplies providing ±35 V.|
The THD+N vs output power for 8 Ω and 4 Ω are shown below. The MyRef FE provides 40 W into 8 Ω and 70 W into 4 Ω at the onset of clipping.
The THD+N vs frequency plots are shown below for 40 W into 8 Ω and 70 W into 4 Ω, respectively. Note that the measurement bandwidth in these plots is 60 kHz to allow the capture of at least three harmonics of 20 kHz. The increased bandwidth also captures more noise, which results in higher measured levels of THD+N.
The multi-tone intermodulation distortion (IMD) measurement of the MyRef FE is shown below. The 0 dB reference level corresponds to 40 W into 8 Ω. The multi-tone residuals are 130 dB below the reference level, which is excellent.
The measurement below shows the IMD at 1 W into 8 ohm. The two test frequencies are 917 Hz and 5.5 kHz. Siegfried Linkwitz argues that low IMD in this test is correlated with a positive listening experience. The MyRef FE measures -99 dBc, which is excellent.
Finally, the two traditional IMD measurements (60 Hz + 7 kHz SMPTE and 18+19 kHz) are shown below. These distortions are quite low for the MyRef FE as well.
The residual hum and noise of the MyRef FE is shown below. The measurement shows a little mains hum, even when the MyRef FE is powered by a lab supply. This is normal. The noise level is a bit on the high side at 69 µV RMS (unweighted, 20 Hz - 20 kHz bandwidth). That's high enough that I notice it with my ears within 30 cm of a relatively inefficient (85 dB @ 1 W, 1 m) bookshelf speaker. During this listening test I also noticed that my cellphone (40-50 cm away from the amp) would cause a burr/buzzing noise in the speaker when it synced with the cell tower.
The output impedance and damping factor for the MyRef FE are shown below. It's peculiar with the abrupt transition at 3-4 kHz. This transition shows up both with 8 Ω and 4 Ω load.
For completeness, I measured the amplitude response and gain flatness of the MyRef FE. The results are shown below. The bandwidth of the MyRef FE extends past the audio band, so I measured it using an HP 33120A function generator and Keysight DSOX1102G oscilloscope. The bandwidth of the MyRef FE measures 2.9 Hz – 84 kHz (1 W, 8 ohm, -3 dB ref. 1 kHz).
Performance with Recommended Power Transformer
Composite amplifiers tend to be insensitive to the power supply used, so it surprised me that the THD+N of the MyRef FE degraded a bit when I powered the amplifier with the recommended 2×25 VAC, 500 VA Antek AN-5225 power transformer. The transformer was connected directly to the power supply terminals of the MyRef FE and its wiring was kept tightly bundled using wire ties.
The THD+N vs output power of the MyRef FE when powered by a 2×25 VAC power transformer is shown below. I included the THD+N vs output power measurement taken with the HP 6643A power supplies for reference.
I decided to explore a bit to see if I could find the cause of this degradation.
The noise floor and mains hum of the amp appear to be nearly identical for the two setups as seen below. The mains hum does increase slightly when the MyRef FE is powered by a transformer, but not alarmingly so.
Although the MyRef FE does show a few more mains harmonics when powered by a transformer than when powered by a lab supply, I doubt the level of these mains harmonics is enough to be responsible for the degradation in THD+N.
The two plots below show the harmonic spectrum of the MyRef FE when providing a 1 kHz sine wave at 1 W into 8 Ω for the two power supply conditions.
The 60 Hz and 120 Hz mains components are nearly 10 dB higher for the transformer-powered MyRef FE. In addition many higher order harmonics of the mains frequency have sprouted up. It seems likely that these mains harmonics are responsible for the degradation in THD+N observed in the transformer-powered MyRef FE. They are likely caused by a suboptimal connection between the signal ground and the power ground on the PCB.
MyRef vs Modulus-86
Over the years many have asked how the MyRef compares to the Modulus-86. The plot below shows the THD+N vs output power into an 8 Ω load for the two amps when powered by the HP 6643A lab supply and the Antek AN-5225 toroidal power transformer, respectively.
The Modulus-86 offers lower noise, thereby, lower THD+N. It is also considerably less sensitive to the power supply.
Did you find this content useful? If so, please consider making a donation by clicking the Donate button below.