By December 15th,
I had to insure that I was still on Santa's white not black list. "I have spent much of the intervening time refining the design. The last bits were put in place about a week ago. I had originally considered simply modifying one of the originals. Then I decided that I needed to beef up the power supply. Since it wouldn't do to provide you with less than production quality, I had new channel and supply boards made. Two units are keeping my shop warm. I'm planning to carry the 240V copy back to Auburn this Thursday for shipment and I'll be writing text about them this weekend." Good things can never be rushed. Still, delivery looked set just in time for the holidaze.

At this point, the forum guessing game included "the input is a pair of complementary SITs each idling at 0.3-0.5A; SITs may be cascoded like the FE of VFET2; the SITs are loaded with power resistors to the power rails like in the F5 and thus drive two banks of complementary multiple Mosfets which maybe idling low or high; the Mosfets are the contributors to brute power and the SITs imbue the subjective delicacy of triodes to this ensemble". Given that similarity with the F5 was on record, on how to improve upon it, someone suggested "cascode input for more power, a folded cascode input, a symmetrical double differential input or something like the Sony Vfet pt2 front end, with Schade feedback on output." There was also a call for "lateral Mosfets on the outputs only because Nelson hasn't used them before." In short, things were all over the place. Some suggestions read like a who's who of cool stuff to dream up some idealized beast. Others with "F5-like JFet input, cascodes modulated, shaded Toshiba Mosfet drivers in place of regular F5 outputs with a source follower output" stopped themselves short by realizing that it would be "quite similar to one half of an XA.8", i.e. something Nelson had already done elsewhere. The mystery lingered. Right about then, Nelson published a short article about the Genesic GA10JT12, a silicon-carbide junction transistor and how to adapt his 1994 Zen™ amplifier circuit to use it. It served as a reminder of his fascination with transistors of all sorts; and how getting hipped to a new part usually ends up with him sampling it to enhance his personal data base. When you love what you do, this type of curiosity is completely hardwired. Less is that as soon as Nelson discovers anything promising, he tends to release tips for DIYers to roll their own and share in his discovery.

Discovery is our cue to now recall what in this amp stable came before; what was still available at time of publication. Their basic specs of power, gain, NFB and output transistor type all set the background and stage for the latest F7.

What this chart does not show is mode of operation. That would be terribly redundant. None of these models were biased in anything other than pure class A.

Should you stop at why output power never exceeded 25 watts into 8Ω, the answer is simple. For FirstWatt, Nelson deliberately limited himself to one shared chassis with the same heat sink area. Ergo, max dissipation for each model was limited from the start. Factor that class A bias usually means constant power draw x 10 or higher than actual output power and these figures fall straight in line. To increase power ratings would mean bigger chassis, bigger heat sinks and bigger power supplies, i.e. Pass Labs .8-type cases and fins.

On December 23rd, Dawid and I both got a load of Santa's good cheer: "Your F7 amplifier left Auburn yesterday. It was accompanied by a 240V power cord but not an owner’s manual. Here is the latest draft of that. I hope you both have happy holidays." For the uncut tech dope, click here. For the fully legit but diluted version, 2007's F5 circuit was a class A p/p circuit implemented with 8 semiconductors and 23 resistors. Nelson dreamt of improving it by simplifying the circuit to four transistors and four resistors, with all Fets in common-source mode for both voltage and current gain whereby two complementary input Jfets would drive two complementary power Mosfets whose output voltage fed back to the Jfet's source pins for a current feedback loop. Not entirely content with his initial results, he eventually decided to add a positive feedback resistor. Quoting him verbatim, "modest amounts of negative feedback are balanced in counterpoint to a smaller amount of positive feedback. This creates an equilibrium where the output impedance approaches zero, improving transient and frequency response. Of course you can achieve a similar effect with tons of negative feedback but I think this is more elegant and sounds better. Also, I put more capacitance in the power supply and found a clever way to further reduce the effect of high-frequency DAC noise and environmental RF." The end result is a 20/30wpc into 8/4Ω stereo amp with a 10KΩ input impedance, gain of 14dB, input sensitivity of 0.57V (2.53V will achieve max output) and a damping factor of 100. HF response at -3dB is 100kHz, unweighted full bandwidth noise is 100µV. Class A peaks at 50 watts into 4Ω and power consumption is 160 watts.

Lateral Mosfets from SEMELab Magnatec.

By December 27th, holidays be damned, Pass forum DIYaudio contributor Ihquam aka Lynn Quam, retired computer scientist of Ollala/Oregon, had mocked up this F7 hack "which appears to do the job. I have left the resistor values unspecified to stimulate the DIYer to play around to find them. I have figured out a set of parameters that appear to meet most of the specifications that I will post at a later date. The primary spec I haven't met is that my H2 is about 10dB too high which will require selecting jfets or lateral fets with different transconductances."

In short, with yet no published photos or any intel on Nelson's used transistors, this gent had already taken the sketchy circuit descriptions and plugged into them the owner's manual specs to reverse engineer a basic F7 by software simulation. Only his 2nd harmonic was still too pronounced to match the actual value. In fact, Ihquam had a first rougher mockup ready on the same day I'd updated my preview with the above PDF. This is just one glowing example of the respectful 'open source' learning culture which Nelson and his friends have created over at the Pass Labs forum of the DIYaudio site. Beyond FirstWatt and its Pass Labs peer, I really don't know of any other hifi brands who prompt quite this level of dialogue and DIY interaction. Or as 2picoDumbs put it commenting on this schematic, "I'd be very impressed/surprised if you could get stable operation of those lateral mosfets without any gate resistance." Thrust, parry, deflection. Referee np looked on mightily bemused, considering when and what crumb to drop next. To dumbs, it was a bone dry "then you can be both" (impressed and surprised). For our purposes, we now exit this particular game to let Dawid Grzyb kick off our sonic reportage with his rookie nookie; but not before we show some inside photos to demonstrate the passage of time and Nelson's process of improvement by simplification.

Dawid's game plane in Warsaw was simple. "My plan is to confront this device with Hegel and NuForce stuff. But what I want to do as well is to take it over to Marek. He has efficient Bastanis Matterhorn speakers. That should be rather interesting. At this point I have no idea what will be the outcome with my own Boenicke W5 and KEF LS50 boxes. Those are rather demanding in terms of power, at least that’s been my consensus so far after a fairly long time spent together. So I'm curious as hell." By now we had our man in the absolutely perfect frame of mind.