The PIMETA v1 Headphone Amplifier was a reimagining of the popular META42, the redesign forced by the discontinuation of the wonderful EL2001 and 2002 buffers after Intersil bought out Elantec. I chose to go with the Burr-Brown BUF634, meanwhile making a lot of other design improvements.
The impetus behind the PIMETA v2 redesign was similar: BUF634 availability turned out to be unsteady and unpredictable. There would be plenty for months on end, then dry spells of weeks to months, with no way to tell how long each spell would last. This was irritating enough by itself, but was compounded by my sense of loss for the Elantecs, which I thought sounded better.
I chose the BUF634 by a wish for a low supply current option, matching the Elantecs. I thought that was important at the time, but it’s slowly become clear to me that most people choose to throw this advantage out the window, in favor of better sound quality. (Obvious to me now in retrospect, of course.) Given that low supply current isn’t a critical design consideration for most people, and given the need to choose a different buffer, I began re-examining the field.
The chip that came to the forefront was the LMH6321, primarily because it’s a near drop-in replacement for the BUF634 in the PIMETA v1 circuit, making it easy to do the early tests. I found that I really liked the sound. They drink a lot more current than BUF634s in low-bandwidth mode, but if you dial up the bandwidth or stack the buffers so they draw the same current, they still don’t sound as good as the LMH6321s, in my opinion.
Having made my buffer choice, I began looking at all the other things that had changed since the v1 redesign, both in the world, and in my outlook as a designer. I found lots of room for improvement.
Having resigned myself to using a surface-mount only chip, I decided to allow much more surface-mount components in this amp than in v1, largely due to experience with other audio circuit projects. I’ve come to believe that there is just a lot of FUD out there about SMT, which really has no bearing when you’re talking about relatively large parts like those in the PIMETA v2.
This focus on SMT in turn forced the 4-layer design. With a pure thru-hole design, you have more room between pins and more options for getting traces from one side of the board to the other, using these pins as oversized vias. The much greater density of the PIMETA v2 design wouldn’t have been possible in 2 layers.
Another decision from the META42 and PIMETA v1 days that I revisited in v2 is the need for four rail caps. We did this in the META42 simply because we had the room. When I did the PIMETA v1 redesign, I brought that along without really reexamining it, despite the design goal of making a smaller PPA type circuit. In the PPA, the power reservoir caps span the supply rails, rather than going from power to ground, as in the META42. Running the caps to ground allows you to get away with a weaker grounding scheme, because they lower the ground impedance, but it means you need twice as many caps to get the same effective capacitance. In PIMETA v2, we have a full-fledged split ground channel with a dedicated ground plane. There’s simply no need to swamp the active components with big caps. That in turn let me cut the number of caps in half while still retaining the same effective rail capacitance.
That then opened the door to the scratchpad. I haven’t seen this lead to a lot of FrankenPIMETAs, but one can hope. I hope other DIY amp designs adopt this perfboard section idea.
Another decision I revisted was the class A biasing scheme. The low parts count and high performance of a JFET based CCS makes it attractive, but it has one big weakness: you have to hand-match JFETs for IDSS. No one likes doing this, so I ended up offering matched JFET sets in my shop. In all the years I did this, I saw almost no one match their own JFETs, even after I raised my prices to cover the growing irritation I had in the chore. You may think you hate matching JFETs, but guess how much I like it after hand-binning thousands of the damned things. :) The change to the LMH6321 opened up enough room for a few more parts, allowing the change to an adjustable bipolar type CCS.
Several other good things came out of the redesign: elimination of the C1 jumpers, better bypassing, etc.
The idea to add all the Molex KK connectors is due to linuxworks. He made one of the last PIMETA v1 amps, a highly modular thing with lots of upgrades, with none of these add-ons hard-wired to the board. Even the volume control was off-board. (The latest version of his creation is here.) We’d given a lot of thought to putting pads on common 0.1" centers in past designs, even adding Molex KK pads in several of them, but it made sense to take this further, making room for the somewhat bulky but popular and functional Molex KKs everywhere possible.
Many people offered useful commentary during the initial redesign. Notable among these are bdhornback, cobaltmute, joneeboi, kristleifur, linuxworks, MisterX, PRR, and voodoochile. All of these people were offered prototype boards, and most did build one and offered further helpful commentary. The v2 wouldn’t be as good without their help. Thanks, guys.
Version 2.01, 2009.08.17:
First production run.
Version 2.00, 2009.03.27:
The initial prototypes. Only 7 of them went out into the wide world to be tested.
|Updated Sun Jan 18 2015 04:24 MST||Go back to PIMETA v2 Headphone Amplifier||Go to my home page|