It was said in http://www.GoodSoundClub.com/TreeItem.aspx?PostID=3418

Romy the Cat wrote:

The HF channel of the Super Melquiades should be revised. Something is wrong in there and it does not sound as I would like to. It mostly relates to improper amount of “density” of sound at HF. Interesting that the full range Milq had no such a problem. Also, the HF channels in the Super Milq are not high-passable for whatever reasons. The only different between the Super Milq and full range Milq is the output transformer, use of half of the tube of the last cap in B+. I observed the high amount of capacitance in B+ make HF “hard”. I am planning to go from 15.000uf to 1000-2000 and change output transformer. In addition I presume that the way in which I used my current HF OPTS was a bad way to do it. I use 0.9H transformer with 4 primaries and 8 secondaryes and in order to load the tube properly I use one of the primary as secondary… As I was informed it is not kosher way to get the best out of transformer in case of serious demands.

Glad to feel that this problem was resoled by “small blood”. I loaded the 6C33CD more heavily for HF channel and it made all magic. The remedy turned out to be a magical as the HF channel of Super Milq limitedly become as good, actually even better, then the full range Melquiades was. It has that sense of softness but at the same time the sense of extreme HF deficiency. Wonderful, absolutely wonderful. You are saying about making mistakes… I had 1.5 year running the HF channel in less perfect configuration. I was mildly complain about it but had no idea WHY it might be. The interesting thing is that my current HF loading (~5R) is kind of too soft for ½ 6C33C. However, the HF channel drive very “fast” Vitavox S2 and has very “fast” core. All together it works very nicely…

Now I am a happy pussy ready to be critical about my new coming tweeter.

Wow, I mean really, really, really, really, really WOW!

A week ago I converted one of the Super Milqs to 5:1 loading and yesterday I converted another one. I loved what the last week converted amp did. It was very nice but today I put the entire system back together and recalibrated everything with the new HF channels loading. Wow! I knew that it moved me to the correct direction but I honestly did not anticipated that it will be so stunningly good!

The Vitavox S2 driver like went crazy and suddenly begin to throw such amassing effects that my jaw dropped. It always was kind of super fast driver with amassing, unique among all other drivers tone and very high contrasts. However, now, driver with overloaded output stage it suddenly reduced its speed to a civilized rate but instead it bloomed with beyond belief colors and ability to glide within those colors with amusing faintness, softness and neutrality. It is truly stunning?

It’s is frankly speaking a little freaky to almost painful to experience. The driver not juts produces those very reach tones but it almost anticipate what will happen next and kind of repair air around itself before burst with new color. Still here is what even MORE amassing. By heavier loading of the output stage I increase power, harmonics and distortions for this driver. Among with it the driver should become “sallower” and it did in a way. However it at the very same time the driver becomes even more intelligently (!!!) contrasty and eave way more discriminating WITHIN own shades of colors. This I did not even predicted but it is an incredible triumph.

To me, one of the most complex tests during my “target listening”  when I need a playback to do certain “tricks” in a very certain way is very specific requirement that I have to playback while a playback trying to render a very good performance of very ambianic recorded cello.   In real live a single cello playing in 150 feet away among the noise of the full Grand Station still maintains an identity of “performing event”. Out listing alertness acknowledging the cello is able to accommodate out hearing to the performance and tune our awareness from the ambiance noise of the Grand Station to the details of the cellos sound. This “distance tunneling” is wonderfully difficult to replicate by playback, particularly across a wide spectra and preserving the “tone”. The Morons usually love to love those strings recordings with up-close microphones where the strings’ schmaltz-quality hits them right into their faces. The strings in ambiance they do not find worthy… because their playbacks can not handle it. However, the real richness of Amati, Gagliano, Tecchler and Strads, in the hands of a good player, opens itself up ONLY in ambiance environment…

So, what Vitavox S2 driven by over-dumped stage doe sis in a way tune itself to “tunneling”. It links of play the unclosed tones slightly softer (in color sense, not in dynamic sense) but into the far-located tones it injects a new colorful life. That all crates amazing “chromatic dynamics” when the bubbles of imagined  sonic clouds  fry within listening space like birds, flopping with winds and teasing with their chameleonic unpredictability….

I do not know… I’m listening not and am wondering…either I east something very “special” today at lunch or I really hit something devilishly fantastic!

After a few years I have again pair of Lamm ML2 in my room and I decide to take advantage and render my experience with this amp into a format of my today’s understanding. I know ML2 quiet we and used it with a number of loads. Over the time I develop for myself my own way to use ML2’s output stage: the manner that gave me better result subjectively. However at that time I have no idea why I did what I did – it juts sounded better to me subjectively. Today I measured the ML2, discovering what I did in paste did in past and correlating with my knowledge of 6C33C that I develop during the Melquiades project - it was educational, at least to me.

A few words about the ML2’s output stage. Lamm insist using his amp at 175V and 310mA. Running the amp into 8.0R reactive load I got minus 1dB 18.5Hz at 21W with examplerary-perfect symmetric clipping. ML2’s outstage clips with voltage and current at the very same time – very-very good I shish all SET builders would understand how important it is.

Now, I never liked this dealt configuration with ML2 and I always used my ML2 differently. I always used my 16R loaded connected to 4R ML2’s output taps and then I drove the out stage with less current and more voltage. My major problem with default ML2 configuration was that ML2 was just a bit too leisurely and too slow. With perfect harmonic envelop it never went for transient velocity if music was called for it. This sort of velocity mellowness made SOME music slightly more legato then I would like it to be. I found it to be one of the biggest ML2 coloration. Do not forget that I used 4 ML2s in by by-amp configuration so I have a LOT of leverage to play with very much variety of different loads/drivers, particularly at MF.

So, knowing what I know how: what did I? The default ML2 has very bold loading. The outstage converts 17.710V on the 6C33C’s anode to 3.002V that makes 5.9 times transformer ratio, let call it 6:1. Converting the ratio into impedance we have the ML2’s outstage is load to 2900R, let call it 300R - that is very-very hard loading. The 6C33C’s plate is 100R, load into 3000R – this is VERY heavy load that shell produces a lot of power, and a lot of harmonics and consequentially a lot of distortions. Lamm in this configuration uses his 6C33C with little voltage and a lot of current – to drive the anode impudence all the way down. Then ML2 employ feedback to drive output impedances lower, to lift the low biased 12AX7 over the input voltage and the most important to balance out the distortions in out stage.

That all was fine but it is still too “slow: from what I would like to be. So, what I did was driving my 16R speakers from 4R tap that made the ration 8.5:1 and loaded ML2’s output anode at 1140R. The 1140 Ohm is very much too idle load for the 6C33C plate but then I did the right thing – I was raising the 6C33C plate impedance to dial in the right loading. The 6C33C is a tube with a LOT of plate impedance run – it might be from 80R to 180R. Let take 120R as default. Increasing the plate current we lower the plate impedance and this is why I used higher voltage and lover current on my ML2 (220-280mA inst of 310mA). Doing it I raised the anode impedance making my 1140R load less idle. In the end I was getting slightly on “faster side” then default ML2 was able to handle but I got my own control how I might optimize the ML2’s outstage for best sound at my load.

Of cause with Melquiades that has no feedback to cheat and where I have a full control of the transformer ration I am at very deferent level of control loading. In the Melquiades DSET I load each of 12 channels independently with a respect to the very specific sound I would like to get from the given speaker. In DSET I mostly do not care for MF and HF about the output impedance and dumping factor. The commercials full range ML2 has no such a luxury and the amp need to be able the handle some barbaric things: like the current reaction of ported enclosures and so on.  Still, in the condition of targeted application even the ML2 might be “played alone” and to make the amps to sound slightly better and with higher respect of the harmonic and transient needs of the given loudspeakers. In ML3 SET Lamm introduced a variable feedback to play with dumping of load. This is as well unnecessary cheating in my view as the right way to play with loading would be modifying transformer ratio and to play with current/voltage of outstage‘s operational parameters.  Do not want to make too many taps? Use a few mR bifilar coils in the transformer that might be connected positively, negatively of grounded, alerting the ratio with high flexibility. With ML2, where the output stage is regulated it is so simple to change the operation voltage without losing the power supply’s impedance that it is a sham do not use this  option as a tool…