Since I know about a few ongoing attempts out there by different people to build Melquiades I would like to share with some of my observations regarding corrections, simplification and possible modifications of Melquiades amplifier.

Obviously you have read the Melquiades release page:

http://www.goodsoundclub.com/Melquiades.aspx

Please take very seriously my insistence for high sensitively speakers. It was not the paranoia or preferences of a horn guy. There are deep underlying reasons (that I live aside) why I insist to use high sensitively speakers.

Below are a set of randomly organized thoughts about Melquiades that the people who have considered devote/invest into Melquiades might find useful. All below comments will be regarding a regular full-range Milq, not about the Milq DSET.

1) Melquiades, Milq, Melq – across this site they are the same names of the Melquiades SET

2) The Milq’s Driver state tube is easily available on eBay for slightly more then one dollar. It is ridiculously wonderful as the Morons pay a lot of money for 6H30, 437A, “kinky” 6922, and many other drivers but they completely disregard the dirt-cheap 6E5P that in fact sound order of magnitude more interesting then any other indirect heated drivers.

3) There is a mistake in Milq circuit. The B+ of the driver stage has almost insufficient amount of critical inductance. It was because I drew the circuit having the Super Milq in my head where the PS was feeding 3 stages and consequently drew more current. I recently opened my full range Mils and confirmed that I have not 15H coke in there but 30H. So, ether you increase the current you draw across the PS lowering the R23 bleeder or increase the value of the L1 choke (Hammond has 30H and 40H chokes that do very nice). My presumption is that any PS with input choke should be tested in continuity, so it is easy to see what is going on….

4) It is not necessary to run input choke filtration for the Milq’s bias supplies. If you wish to do for CRC filter before the gas tubes then it is fine (use input cap of a small value thought). You will also need much lover voltage on the secondary of the bias supply transformer. With input choke the bias need to be 220V as the input choke will eat up ~10% and the gas tubes need 190V to light up. With input cap you will have an increase voltage from bridge and I would go for 140V-150V on secondly if you use CRC…

5) You might consider do not make a dedicated PS for positive bias but to take it via a resistor right from C2 capacitor. I did it and I did not see any worsening in sound.

6) Do, not screw with Milq’s bias principle for both states. The input stage should be in Milq exactly as it is and the output stage should be fixed biased. Keep the R6 of very good quality, and keep it very short and right at the grid. Bring the return to ground of C5 right to the V2’s ground.

7) The C12 capacitor is not relay necessary if your C9 capacitor physically located close to VR2 attenuator.

8) If you never used 6C33C before then print many posts at this site about this tube and put them in your restroom. Sometimes when you spend long time in there please read about the tube. Do not forget to heat the tube for a long time (40-60) minutes before REALLY stressing it.

9) Amorphous core output transformer is absolutely mandatory for this amplifier. There are a number of companies that do it. Find it, pay premiums and get the amorphous transformers.

10) In the circuit I marked the out stage operation at 225V and 180mA. However, these markings are only for 200mA Lundahl transformer. Lundahl can not gap them for more current as they have fixes core size and any increase of currant will reduce inductance of primary with consequential loosing of bass. Even at 225V and 180mA with 200mA on LL1627, loaded to 8:1 (depends of your speakers) the Milq does very fine. Still it is for my speakers.  If you are not using the off the shelf Lundahl but go for a custom transformers then I would suggest to stay with 190-200V and run 250mA across the tube, with a transformer gaped at 300mA. Sure, it will greatly depending from your speakers - do not forget the Milq has no feedback and therefore it is very sensitive to load. Do not use the transformer taps under any circumstances. Anyhow, you can go as low as 170V and as high as 250V, driving from 180mA to 330ma across the tube, loading it from 5:1 to 12:1. Increase current lower plate impedance and adds transients… You will need to play with current and loading in order to please your speaker… BTW, do not forget that the bass section of your loudspeakers and the HF section of your loudspeakers might have different optimum loading demands…. but I promised don’t talk about DSET in this thread…. 

So, my recommendation for the folks with 300mA transformer would be for average circumstances of above 12R loudspeaker to go for 200V/250mA (on full tube).  It largely would be OK for people who use normal, mid colorfulness Alnico compression drivers. With very flowery drivers you might slightly drop current and with very dull drivers or with ceramic drivers you might drive more current through the output stage drop voltage. It is always a good idea to have an extra gap in transformer that would allow playing with current. The 6C33C will slightly fluctuate current with signal, even with low PS impedance, so to have extra DC capacity in your transformer is good idea to keep away form saturation. Let your transformer maker to advise you about the balance between the transformer core size and the diameter of the turns… Let them also advise you how much excessive gap would be desirable. From one perspective with amorphous core it is hard to get bass out of transformer from another perspective the amorphous core tolerate proximity to saturation with more damage to sound then regular non-amorphous core…

11) Do not save money on the output transformer. It is perfectly OK if the output transformer will cost you as much or more even a few times then the rest of your Melquiades. I assure you that Milq will do sonically wonderful thighs. You assure yours that you will get for Milq a deserving transformer.

12) Put two 10R, 5W resistors (in each leg) between the rectification bridges and secondary of the power transformers.

13) Read my article about the gas regulator tubes:  http://www.goodsoundclub.com/GetPost.aspx?PostID=2792

14) The 6C33C tube sockets are very bad, unless you go for expensive Johnson tube sockets. Still it is not too difficult to replace the cheap 6C33C tube sockets once in 2-3 years. You will extend the live of the 6C33C tube sockets if you clean the 6C33C pins of your every NEW tube. The pins are covered with some freakish candle-like material. Use fine sand paper to brash the pins up and to remove the material.

15) If you do not cool down the Milq internals then allow some space for 6C33C to breathe. Make 3”-4” around the tune as the parts-free zone and drill some ventilation holes around the 6C33C - it runs hot. Twist the filament wires to the 6C33C socket and then solder it. I have seen as in not cooled down tight enclosure the heat of 6C33C melted down the solder on the filament pins. Use the 6C33C pins a high temperature solder. BTW, do not use heavy silver solder anywhere in signal pass. The very light-silver solder (like Wonder solder) is tolerable.

16) If you have a very good quality VR2 that can handle some power then use it and get rid of the R19, R18 chain. I used the 10-turns Burns that are 2W I believe and the R19, R18 are juts to keep the attenuator cooler. Keep the VR2 as good quality as you can get and put R20 as close to the 6C33C’s grid as possible (do not burn if by the tube though)

17) There are a very few wires in Milq – it is very simple amp and most of the parts might be connected by their natural leads. Sure no silver wires must be use anywhere but almond all equally good cupper wires that you will chose to use some wires are more important in Milq then other. You will have a wire between negative pin of C9 and R18. Pay attention to that wire. It layout, the proximity to ground and etc is auditable…

18) R2, R12 and R23 should sit right in the tube socket.

19) The Milq even without loading resistor has enough impedance from the bias resistors but I still like to current-load my cables using 20K resistor as R3.

20) Use good quality tube socket for 6E5P. it is very fast tube and it has some minor tendencies for microphonics when is it getting worm (not as horrible as some other tubes through). The Connex 70032, the 9-pin tube socket does very well and it is not too expensive. Expect approximately on year from a single 6E5P with approximately the same from 6C33C. it is $12 per year – it will be all maintenances that you need to do on Milq.

It is all that I have atop of my mind for now. If something else come to me  then will dump it here….

My project to add 3 one-stages channels to Super Melquiades look like progressing (it means I will have the new 15,5:1output transformers very soon) and in  context of upcoming changes and, after long consideration and  some tests, I desired to  combine the addition of the one-stages with some other modifications. Here is the list of the changes the Super Melquiades will endure (besides adding 3 more channels that itself is a major invasion):

1) Elimination the current filament 6.3V/12.6 transformer and replacing it with   new 12.6V/15A toroidal transformer. All 9 Super Milq’s 6.3V tubes will be paired and driven by 12.6. (Do you know who makes standard serial 6.3-0-6.3 toroidal for 12-15A in order avoid going custom?)

2) Dropping the plate current of the out stage from 220V to 190-200V. The new plate transformer will be introduced: 220V/3A. Since, the HF channels will not be driven by 6C33C anymore I do see a rational to drop voltage and increase current on the 6C33C’s plate.

3) A new fan will be added to cool down the cluster of the very narrow positioned six 6E5P/6E6P tubes.

4) The peripheral wholes around the 4 large B+ capacitors will be sealed, converting the internals of the Super Milq into air-taught with objectives to navigate the air flow from of the Super Milq’s main fan more prissily.

Romy the Cat wrote:

3) There is a mistake in Milq circuit. The B+ of the driver stage has almost insufficient amount of critical inductance. It was because I drew the circuit having the Super Milq in my head where the PS was feeding 3 stages and consequently drew more current. I recently opened my full range Mils and confirmed that I have not 15H coke in there but 30H. So, ether you increase the current you draw across the PS lowering the R23 bleeder or increase the value of the L1 choke (Hammond has 30H and 40H chokes that do very nice). My presumption is that any PS with input choke should be tested in continuity, so it is easy to see what is going on…. 4) It is not necessary to run input choke filtration for the Milq’s bias supplies. If you wish to do for CRC filter before the gas tubes then it is fine (use input cap of a small value thought). You will also need much lover voltage on the secondary of the bias supply transformer. With input choke the bias need to be 220V as the input choke will eat up ~10% and the gas tubes need 190V to light up. With input cap you will have an increase voltage from bridge and I would go for 140V-150V on secondly if you use CRC… 12) Put two 10R, 5W resistors (in each leg) between the rectification bridges and secondary of the power transformers. Romy the Cat

3) For the 30h do you change r23 to like 42k?  the arrow from r23 to l1 seems to imply that r23 depends on the choke. does this effect r2?

4) what is your opinion of small value input cap?  the schematic 4.4 has 100uf with choke input, is this small

12) what does this accomplish, what does this change in the rest of the circuit. If you used high current fred instead of sic schottkey does this still apply

thank you 

Right. I do not mean to suggest that my own solution for arriving at an acceptable global result using the Milq full-range was not a short-cut with regrettable new weaknesses (or just missed or postponed opportunities). But I am curious (wary) about what you mean by the “one-sided” results of using a DHT for the upper midrange (if I misread you correctly). I also wonder whether it is even conceivable that such an amp might use valve rectification and still keep up with the Milq. Suppose further that it was asked to drive not only a compression driver but also a 6” paper cone from 400Hz to 1000Hz? I want to doubt all that, but I haven't thought about it too much. What I do know is that I am quite allergic to the sound of solid-state bridge + 6C33C + amorphous iron in the upper midrange. However, with oversized, steel-core OPT's, lowish end of permissible voltage (not starved!) on the 6e5p, foil caps on C4 and C5, and a system-wide ban on silver molecules (ok atoms), you do start to get a peculiar, stoic elegance out of this tube, and you do not lose too much. Though I am keen on trying a DHT to play along with the Milq, I suppose it's easier said than done. There are a lot of amps out there using celebrated or obscure triodes but which have precious little in the way of tonal sophistication. Less than my Milq, anyway.

As for the M2, I heard it only once in a very controlled environment (one of these demonstration rooms in NYC where you are made to feel that you ought to have paid admission just to behold this or that sonic magnificence, and then kicked out after one movement of something like “Hilary Hahn” played back on a $75,000 CD player). If I'm honest, I was amazed by the sound. But I did not really do much about it at the time. I was using ANJ M77 and Neiro (parallel 2A3) then and had not yet discovered real problems with that amplification. Also, the Kondo, even if it has a particularly strong sonic signature, benefits from an organic-sounding tonal shaping, whereas the Lamm sounded more synthetically processed (and I did not know, or care that much, if it was some Disney-esque production quality of the recording or something else).

dl wrote:

But I am curious (wary) about what you mean by the “one-sided” results of using a DHT for the upper midrange (if I misread you correctly).

dl wrote:

I also wonder whether it is even conceivable that such an amp might use valve rectification and still keep up with the Milq.

dl wrote:

Suppose further that it was asked to drive not only a compression driver but also a 6” paper cone from 400Hz to 1000Hz? I want to doubt all that, but I haven't thought about it too much. What I do know is that I am quite allergic to the sound of solid-state bridge + 6C33C + amorphous iron in the upper midrange. However, with oversized, steel-core OPT's, lowish end of permissible voltage (not starved!) on the 6e5p, foil caps on C4 and C5, and a system-wide ban on silver molecules (ok atoms), you do start to get a peculiar, stoic elegance out of this tube, and you do not lose too much. Though I am keen on trying a DHT to play along with the Milq, I suppose it's easier said than done. There are a lot of amps out there using celebrated or obscure triodes but which have precious little in the way of tonal sophistication. Less than my Milq, anyway.

dl wrote:

As for the M2, I heard it only once in a very controlled environment (one of these demonstration rooms in NYC where you are made to feel that you ought to have paid admission just to behold this or that sonic magnificence, and then kicked out after one movement of something like “Hilary Hahn” played back on a $75,000 CD player). If I'm honest, I was amazed by the sound. But I did not really do much about it at the time. I was using ANJ M77 and Neiro (parallel 2A3) then and had not yet discovered real problems with that amplification. Also, the Kondo, even if it has a particularly strong sonic signature, benefits from an organic-sounding tonal shaping, whereas the Lamm sounded more synthetically processed (and I did not know, or care that much, if it was some Disney-esque production quality of the recording or something else).

Thanks, Romy, for your generous reply. I won't try to address everything about everything, but let me bore you with a mini-history of what is going on with my playback so you will understand where I'm coming from and where I'm trying to go. And how I wound up with the Milq in the first place.

Back in the days when I used the ANJ, I was using a pair of the small Hoerning speakers. These were fine, but you got the sense that it was really a 4-way speaker with one of the channels – mid-bass – removed. Or you might say it is a 2-way speaker with a built-in subwoofer. Anyway, I noticed even with the huge Hoerning “Alkibiades” towers, that as you listened you could “see” the missing...maybe pair of 10” low-mid/upper-bass drivers per side or maybe 12” driver. But Hoerning did not go in that direction – in fact his new big speakers use 16 8” bass drivers per side but downsize the midrange driver from the 8” Lowther DX3 to the 6.5” DX65, which is not nearly as good sounding as the 8” one, first of all, and also makes the one stand-out problem with the old version audibly worse. Anyway, I decided more or less at that moment that I ought to try to build a speaker system myself.

In the meantime, I had heard some horns for the first time, just a pair of old YL compression drivers in metal horns which I stuck on top of my Hoernings, and was seduced. I was not really ready to start a construction project, so I just bought a pair of 2-way horn speakers from Aspara Acoustics – a 12” Precision Devices driver and a sort of radial horn mounted vertically on top. It was not long before I realized that now what I had was actually a 5-way speaker with 3 channels removed. There was no real “bass,” and in fact no tweeter at all, but more importantly there was no midrange anymore. I decided that “audio” was getting way too complicated and was detracting from my enjoyment of music, and I impulsively sold my entire system. I later wound up with a single-driver speaker from Audio Tekne (with a Diatone 610 in a leaky box) and drove it with a push-pull 300B amp. I moved my new system to a small room in my apartment was happy for a long while!

Then I moved to a place with only one large room, and the Diatones just sounded diminutive and “lost” in there. I figured, maybe now is the time to fix those old audio problems. My logic was, I did love that 6” Diatone driver, so why not try a 12” alnico driver and put a horn on top, then look for a midrange driver to put in-between. Last, I would worry about deep bass.

This plan did not work, at least not at first, because the Altec 414 I started with is a crappy and bloated sounding driver, overrated in every way. I did discover the JBL 2440 during that time, which I still use. To replace the Altec, I first tried the 12” alnico driver from Supravox (285-2000). This driver is also a headache for 2 reasons. It has such a strong magnet (2.1T) that it is impossible to get any mid-bass out of it. Second, it has a nasty sharpness around 1-2kHz, which, I am told, is due to the fact that the phase plug has a neodymium magnet inside of it (??!!). I used a series resistor to lift the Q, but I could not put up with the “veiled” (from the resistor) or the piercing (from the phase-plug) for very long. I went instead for the 285GMF, which has a more reasonable magnet (1.4T).

As the transparency of my system began to increase, I discovered that my push-pull 300B amp was actually awful sounding (even though I had no complaints when it drove the 92dB Diatone). I tried several 300B SET amps, i.e., middle-market ones, and these had no bass control whatsoever. I was always high-passing the 285GMF, even with air-tight enclosures. Then I thought: hmm, why don't you get someone to build you a Melquiades amp!

Fast-forward to the present, and I still have that same problem I started with. Namely, there is no “mid” (what I am calling 400Hz-1000Hz) in my setup. If I use the GMF in a high-passed bass-reflex arrangement, the upper-bass really comes to life, but the mid becomes thin and washed-out. Mostly to avoid port spillage issues, I keep it sealed, and though there are no obvious problems, the sound is fairly flat, boring and literal. But as you lift the volume, you realize that the midbass is doing a lot of things that you wouldn't normally expect. It's very tight, but also “broad” and expressive.

The reason I ask about using a DHT “add-on” amp for my non-existent midrange driver (and which would also drive the 2440) is that I worry that I will never find a 6” or 8” driver to play along with the 285GMF. This is partly because of level matching, but also because as you increase the inductance on this “woofer” it sounds more and more veiled, slow and boring. So if I am going to introduce another amp, I would ask it to play from 400Hz up.

But the idea of using different combinations of sections from the LL1693 is interesting, and perhaps I should buy a pile of drivers and start experimenting. Just browsing ebay, you notice lots of things, like this alnico 8” driver with Fs 100Hz and 97dB sensitivity, pulled from a garbage dump:

Who knows. I would like to use such a driver in a Voigt pipe (or I guess sealed since the Q is something like 1.3), then go back to high-passed bass-reflex on the mid-bass driver.

As for my allergy to the amorphous-core OPT's in the high range, it's just a sense of raw, as if all the mechanical elements of the tone were chipped away like flecks of rust, then sprayed into the room a fraction of an instant before the timbre. I managed to counteract this obtrusive sense of distortion, on digital, with a warm and puffy sounding NOS DAC using microphone transformers at the output. FM tended to sound fine. But I was getting headaches with vinyl, even with lush sounding cartridges, and I got really tired of that DAC. With steel-core, the sound is much smoother. At least coming out of the 2440.

-Darren