M³ Project Announcement

[morsel]

M³ Project Announcement

Morsel and AMB present a new community service project and invite your participation.
M³ is a 3 channel MOSFET output headphone amplifier inspired by the PPA and SDS Labs amps.

There is no ETA on boards.

Pronunciation

M³ is pronounced em cubed. The 3 in M³ is HTML ³ (Unicode 00b3). Type MMM if you are lazy. Rumor has it MMM once stood for Morsel's MOSFET Monster, but in fact it refers to the 3 MOSFET output channels. M³ is a codename subject to possible future change.

Design Goals

AC only, no compromises for battery compatibility
Unimpeachable discrete output stage
Better price/performance ratio than the PPA
KISS philosophy (Keep It Simple, Stupid)



Circuit Description

Vbe multipliers bias IRFZ24N and IRF9Z34N 18A power MOSFET pairs into Class A operation for low distortion by establishing a constant voltage drop across the gates, which determines the quiescent current. 1µF film capacitors across the Vbe multipliers help stabilize the voltage drop and lower impedance at high frequencies. FET cascode current sources bias the opamps into Class A operation and set the drive current high enough to overcome the reactance of the MOSFET gates. 220pF capacitors between gate and source of the N channel MOSFETs compensate for the difference in Ciss between N and P channel MOSFETs to provide symmetrical bandwidth rolloff, or the N channel gate resistors may be increased 1.67x, or compensation may be omitted.

3 amplifier channels (left, right, and ground) use the same output stage and noninverting opamp topology. The ground channel sources and sinks the return current from both transducers which would otherwise have been dumped into signal ground or power supply ground. This shifts responsibility for the high current reactive load of the headphones from signal ground to the supply rails of the output stage, thus removing the primary source of signal ground contamination. The transducers are driven by symmetrical output stages with equal impedance and transfer characteristics on both sides, rather than an output stage on one side and the large capacitor bank of the power supply ground on the other. This results in lower output impedance and greater linearity.

Choices

We have deferred some decisions to give the community a greater opportunity to voice opinions and to give us more time to test various aspects of the developing design.

Board dimensions are really critical as they affect cost, size, case compatibility, board mounted components, component spacing, and sound quality. We are leaning toward dimensions 2-3x bigger than a standard eurocard to allow for onboard heat sinks, a less congested layout than the PPA, possible choice of board mounted components, multiple input and output jacks, and compatibility with PAR-METAL 20 Series cases.

Ventilation requirements exist for the MOSFET heat sinks unless the MOSFETs are arranged along one edge of the board and bolted to the case along with electrically insulating heat sink pads. Ventilation slots result in dust ingress but allow for more flexible board arrangements. Heat sink options include Aavid Thermalloy 531202b00000 2" or 531002b00000 1" free standing screw mount extrusions.



External power supplies are best, as they are less likely to interfere with the amplifier. Internal supplies should be as far from the input stage as possible, and preferably shielded with steel or mu metal. We are considering voltage regulators on the pcb, capacitance multipliers or voltage regulators on the opamp power rails (either 3 separate, or 1 for all 3 opamps). LED power indicator current sources and zeners are pointless, since this is an AC only amp. LED and series resistor, power switch, and series diode for reverse voltage protection would be useful.

Ciss compensation capacitors, increasing N channel gate resistors by 1.67x, or no compensation is a choice best left up to the end user, as is bass boost, assuming board space is not a problem. No crossfeed is included, although we have considered a simple resistive crossfeed to reduce the stereo image on hard panned albums from the 60s and 70s.

We tried MOSFET diamond buffers but they had slightly worse IMD than MOSFETs driven directly by opamps. We considered BJT diamond buffers but decided to stick with studly MOSFETs because of lower output impedance, higher input impedance, negative thermal coefficient, and circuit simplicity. Low drive current requirements eliminate the need for more complex driver circuits such as the diamond buffer. We also think MOSFETs sound better than BJTs.

Measurements

Initial tests using an M-Audio Transit and 33 Ohm dummy loads show THD and IMD comparable to sound card loopback. (THD .0015%, IMD .0025%, perhaps someday I will buy a better sound card.) Our prototype drives studio monitors impressively and should be a good match for any dynamic headphone, even the notorious AKG K1000.

Team M³

Team M³ is Morsel and AMB. Thanks to PPL for his bountiful wisdom and lore.
NeilPeart will be helping us with listening tests.

[Glassman]

great, I like that FET+BJT opamp supply isolation better than just FETs.. as to the amp circuit itself - simple, clean and clear.. I like that.. what about different types of MOSFETs or are these really the best?

[Voodoochile]

Nice target chassis choice! Really, I like that box a lot. Glad to see designed for AC from the ground up ('specially with MOSFETs).

I'm in full support of more board-mounted options, such as nested pads for the DaCT attenuator and Alps, plus on-board fusing as an option that can be jumpered if rear-panel access is preferred.

Glad you are planning to drive the MOSFETs directly from the OPAs... I like that.

When will the boards be ready? (kidding- no reply requested!)

[sygyzy]

Very nice. From the little amount of that I actually understood, this seems like a pretty exciting time for DIY.

[Lil_JV]

Looks great! I look forward to building this amp.

JV.

[Earwax]

Originally Posted by morsel

Design Goals

AC only, no compromises for battery compatibility
Unimpeachable discrete output stage
Better price/performance ratio than the PPA
KISS philosophy (Keep It Simple, Stupid)
[/SIZE]

Much of the rest of the announcement is over my head, but I like this part


Looking forward to building one!

[morsel]

Hi Glassman,

Renesas (formerly Hitachi) makes power MOSFETs for amplifiers such as the 2SK216/2SJ79 .5A and 2SK1058/2SJ162 7A pairs. They are really expensive, hard to get, use a nonstandard pinout, have higher Rds, lower Id, and their Ciss is not much better than the IRF parts.

[kyrie]

Looks very cool! Though I don't understand the technicalities...
It says that the M³ is designed to have better price/performance ratio than the PPA, but would you expect it to sound better than the PPA, ignoring price?

[morsel]

Our design goals may or may not be achieved.

For those people who prefer discrete output to HA5002 buffers, it will probably sound better.

[Sovkiller]

morsel:
Any way we could use another OPamps, like the BBs OPA627/637 or the LT1122, LT1028, etc....???

Any way of getting included any onboard processor like crossfeed, bassboost, etc...even if optional or defeatable???

[morsel]

As with any DIY, you can choose your favorite opamps, add crossfeed, etc.

[Nisbeth]

Whoa, the wait is over!
You've mentioned this project a few times before Morsel but I thought you had given up on it. Luckily not .

I'll be following this thread with great interest


/U.

[Lil_JV]

Is it okay if we just write M3 in the future? It is even easier than MMM.

JV.

[morsel]

M3 is a model of BMW. Why not just copy and paste M³ from a previous post?

[Sovkiller]

Originally Posted by morsel

As with any DIY, you can choose your favorite opamps, add crossfeed, etc.

Yep, but what about if you guys, consider those options as part of the initial circuit, that will make the project a more universal one, and will fit more and more persons, why not including them on the original circuit board, maybe optional, but there. And later same as the PPA, if you want them, just populate this part of the circuit, and if not, just bypass it, and period....But the bassboost can't be added later if you want it, IIRC it has to be there in the circuit I think, or am I wrong?