I was searching for an answer why squire wave is so screwed and I was advised by the site’s visitors that it is due to HF restriction of my channel. My HF response does look too low to me – the 23kHz; with all my ultra low capacitance magnetic I was expected way higher.

Something is not right. So, I decided to look at the stages independently.

First I looked at the output stage. I shunted the grid of 2A3 with 250K to ground and put 2uF cap between grid and generator.  Unfortunately my generator can’t swing more than 10V and I was not able to drive the tube into full power. With 10V on grid the 2A3 just started to roll off at 275kHz, but picked a lot of distortions. The clean and not distorted sine wave I got somewhere up to 140kHz, that is meaningless as it was not at full power. So, I drove squire wave there the thing at 10kHz. This time it was tutorial different picture. Obviously the output stage was fast enough and was not the problem of my initial squire wave.

Then it might be the transformer. I still do not know why that transformer in 2:1 configuration gives 30% of voltage lost. Who know, might be something else wring with it? I got rid the grid resistor and coupling cap and put the coupling transformer in, driving transformer’s primary from generator

It looks like the transformer did restrict HF up to the point as I had in the post above. So, the driver stage was one that left. I put 15K anode resistor on the driver stage, drove it from 400V and took output via a coupling cap. Here I had my 23K restriction and screwed squire wave. Hm, it looks like my line-level filter plays a bad game with me introducing too high impedance on the driver’s grid that along with tube miller capacitance of the driver tube shape low-path filter. The original Milq bias had 8.2K grid resistor, now the idea evolved to 32.1K for HF channels.

http://www.goodsoundclub.com/PDF/6-Chennal_Melquiades_DSET_Amplifier.pdf

Well, I need to think about it.  For MF I need the 20K resistor to write the high-pass filter. I can get rid the input filter all together, keeping just 10K-12K impedance on grid (that would give me around 50kHz response) and then I can make the same high-pass filter with coupling cap.  I am a bit afraid this solution as I had problem with sound in this configuration before, when the inner-stage caps acting as filter did not sound right for HF

http://www.goodsoundclub.com/GetPost.aspx?PostID=1234

also, I do not like the idea of coupling cap as with my 33 time gain in driver stage it will deliver too high voltage to the 2A3 grid and I would not wont to be involved into the coupling cap recharging and A2 operations. I might drop voltage on the driver’s grid that will again introduce too high impedance on the grid…

Frankly here is where the 6E5P gain works against me. To use the high gain, current capable driver did make sense for two-stages around 6C33C that has 60-80V on grid but with those flimsy DHT with 40V on grid I would prefer to have 15-20 time gain in driver, something like a half of 6SN7 or alike.

I kind of do not know where to go from here and I did not even start to listen the thing. I am trying to interpret the result I was getting up to now. The original Melquiades idea was perfect but then the value of the grid resistor begins to grow to 10K and then 12.1K and then to 30K that included the high-pass filter. Unknowledgeably to me the HF response of the channel was lowering from above half- hundred KHz to 23Khz. Parallel to it I was loading the MF stage harder and harder from 1.2K on 6C33C (that is VERY idea for that tube) to something that is equivalent  to 450R – a very heavy load.

I might understand why I went this way – I use Vitavox S2 drive that is very fast, very contrasty and that might go into some harshness if it not handled properly. Everyone who use S2 have problem with it but I do not. My S2 is very smooth and superbly eloquent. I was under presumption that I accomplished it by loading the stage that drive S2 but now it look like I also used less fast amp to drive the S2. It might be even beneficial but I am a bit distracted by the fact that when I switched my Sun Audio 2A3 prototype amp from cathode-biased 6SN7 to grid-biased 6E5P I did observed some transformation in transients reproduction. The cathode-biased 6SN7 has no HF limit but the grid-biased 6E5P with 12.1K resistor is effectively a low-pass filter at ~35K. Add to it the roll off of the transformer and we have second order low-pass. Well, I need to reconcile all of it and figure out how much HF I would need in MF channel.  The whole objective of the experiment of conversion the Milq to DHT MF was to inject live more into MF. My presumption was that my single stage tube has no power and no gain to be loaded idle-enough to have good transients. However, it might be ALSO the problem is not with the single stage itself but with the fact the MY CURRENT IMPLEMENTATION of single stage has too aggressive low-pass making the amp not able to throw good transient characteristic.

Hm, this opens a wide thinking about the other opportunities…

Unquestionably in the original Milq the cathode-bias or fix bias did not sound even close to grid-bias. In context of full range the grid-bias literally destroyed all other options… with THIS tube. Would it be because my driver tube itself is very fast (275MHz) and my grid-bias calmed the tube down a bit?

Well, there are many thoughts in my head at this point and I need to think what to do with my MF channel next. I even thought to DC coupe the driver and DHT, why not? I have 400V supple to drive the DHT. If I burn the 150V on DHT cathode and drive the driver at 200V on plate then I can direct-couple them. It will be no drift for PS as my whole amps is driven from AC-stabilized regenerator (good for PP2000). All DC instability might be handles and I never seen my 6E5P to die catastrophically, so nothing will burn out the output stage. It will not however address the problem with the driver stage’s low-pass… To use slower tube with less gain? To go cathode bias on 6E5P but then I need to target impedance to write my line level filter? To go with a cap filtration instead of RL filtration? To go full-range super fast amp and use speaker level filter? I do not like any of the options so far.

That all is very interesting and unpleasant so far. I did not think that the DHT projects would open so many skeletons in the Melquiades closet.

Well, I kind of understand that my objectives for my “perfect” MF channel I have some irreconcilable differences.  Let see:

1)      No capacitors in signal path

2)      Bandwidth restriction, allowing employ DSET type output transformer with minimal inductance.

Sounds simple but it implies use of front filter that automatically implies impedance.  Is it possible to write a filter with of cathode or anode voltage? How about use of substation filter at line-level? The most important - do benefits that I get from use the DSET type output transformer over-weight the problem I get using the line-level filtration? The speaker-level 3uF filter on Vitavox driver works under high current and had less damns to any cap at line level of between stages. That however convert the amp into a full range – not the direction I would like to go.

Options, options, options… I still have no topology that would grab me conceptually and… I am not convinced that my current configuration is something that would have subjective sonic problems on MF.

Still using 6E6P for driver?  Can you just scrub some gain with lower plate values without making such a steep signal path filter?  You'll still be forced to tailor your bandwidth, of course.

Would it help to split the roll-off curve between the stages, maybe step down?

FWIW, I'll bet most of us wind up living with a +/- traditional RC coupler, with a cap in the signal path.

And maybe, given this particular situation, a speaker-level cap and "FR" output to drive it is not really such a problem (except if you need to order another tranny...), since you hardly have to worry about FR, after all, if the cut-off slope is gradual enough.

According to Ohm's Law, that "lost" voltage is going somewhere...

Best regards,
Paul S

Ok, I made some further experiments driving this time my S2 driver from Milq DH DSET. The hell of the sensitively - it was too much.  So, I went for 4:1 inner-stage ratio. The result was very successful and the amp looks like was able to get more voltage in. Now it clips at higher power at bottom half (I do not remember it is voltage or current and I did not get yet which stage or what transformer clips first it will be fun to learn in future).  Take a look what I have now:

The good part is that 4.3V at input is it how much my DAC outputs at -2dB digital. With this volume window of my room will fly away along with frames – I never play that loud. Even at this level I have a driver and the power stage run at half of the grid bias- very good.  The bad part is that I pump a hell of a lot gain in the driver stage and then burn it in innerstage transformers. Well, it is not just do not make sense but it also have relation to my other thoughts…

My other thoughts were that I was listening today my MF channel driver from Milq DH DSET. I actually was listing the Sound.

I like Sound more then I had it with a single stage but it is also not the sound that I would like to have. I have a LOT of to say about the sound of Milq DH DSET with 2A3 and YO186 and about the DH specific but I will do it later what I will be taking about Sound.  So far I am not in Sound mode but I rather looking for first an intellectual stimulating solution of my topological problem.

The subjective initial assessment of sound suggests that although the sound with DH is better but it still the same type of sound – it is gracious, polite and fabulously smooth. With the DH it has some interesting texture and other thing that I will comment about later but it is still a bit on “save side”.

As you might read in my post above I discovered recently that my elaborate Milq-biasing, and filtering crate too much impedance on the grid and consequentially too closer low-pass filer. The ugly square wave response of the driver stage proves it.  So, that all make me to wonder….

Let see, the high impedance on the grid create the filer as it work with Miller capacitance. The Miller capacitance is the plate-to-grid capacitance multiplied by gain plus one.  The 6E6P has 15pf and 33 times gain makes is near 500pF. The 500pF in tube and 32.1K in grid is what I have my 23kHz response and the ugly square wave. Now, I DO NOT claim that the making my MF amp faster in trims of upper frequency response is something that I need and is something that would me to feel better about sound. But it might be the case as the better square usually relates to better transients.

OK, too high capacitance too high impedance and too much gain… the answer is self-evident – I might to look for another driver stage.  The 6E5P/6E6P is voltage tube but I do not need a lot of voltage to drive my DHT. If I look for a new driver then here are some requirements:

1)      Good sound – it shell not be explained further.

2)      Bias or 2-3V that will be left from 4-5V input voltage left after filter.

3)      ability to work with fixed bias

4)      200V on plate

5)      Low Miller capacitance as it still use the filter and prefabs the Milq bias

So, I was looking around and thinking about the the Russian 6N6P. The 6N6P is Russian version of 5687. Bothe tubes  are wonderful sounding but the 6N6P reportedly (not my report) is better as it has more parallel plated vs. the 5687 that has it curved. The 6N6P has 2V bias that would make Milq’s grid bias to be twice lower. I would like to have it 3V to have some spare voltage but you will never know how much it will be unless you try. The 6N6P is double triode and each half has 4.4pF of input capacitance. That is very great as it 3 times lover then my current 6E6P. The 6N6P has gain of 20 that would make the total Miller capacitance of 88pF. The 88pF vs. 500pF – only that will give me way more extend response. Oh, did I tell you that I have around 50 6E6P in my storage?

http://www.goodsoundclub.com/PDF/6N6P.pdf

I do not have a decision yet. I might convert one channel of my Sun Audio prototype amp. But at this point I am opone for options…

I have to admit that although the Macondo “Water Drop Tweeter” sounds very good but I do not have that “shocking sound” that I got from the “Water Drop” during my initial introduction to it. I kind of always felt it and I thought that I get use to it but now I think differently. What is the problem is that in my current version of Milq my HF channel suffers from the same problem of low-passing die to the high grid impedance as I described above.

After all, the grid has the same 500pF of the 6E5P and 32.1K impedance… I can not test the “Water Drop” with other channel as it is direct to plate bound and I was thinking about the two stages and inner stage filtration. I hate to lose the super cleanness of a single stage for HF… and then it came to me: the 6E5P is a TETRODE. The Weekeipida says:

“The second grid, called "screen grid" or sometimes "shield grid", provides a screening effect, isolating the control grid from the anode. This helps to suppress unwanted oscillation, and to reduce an undesirable effect in triodes called the "Miller effect", where the gain of the tube causes a feedback effect which increases the apparent capacitance of the tube's grid, limiting the tube's high-frequency gain. In normal operation the screen grid is connected to a positive voltage, and bypassed to the cathode with a capacitor. This shields the grid from the anode, reducing Miller capacitance between those two electrodes to a very low level and improving the tube's gain at high frequencies. When the tetrode was introduced, a typical triode had a input capacitance of about 5 pF, but the screen grid reduced this capacitance to about 0.01 pF.”

Well, why do not flip my single 6E5P into tetrode?   The datasheet on 6E5P said that in tetrode mode it has 0.006pF. The gain will be higher. Let say 200, so it will be still under the whopping 1pF that will give me response over 100kHz. To do so I need juts to flip one wire on my 6E5P and to listen the result. If it works then it will be so elegant solution for tweeter!!!

Why not try it?  It "should" work to extend response, if nothing else.  Just use the least in the way of caps you can get by with, to avoid "breathing".  And it's not like you need the gain, if, as I recall, that tweeter is a 109 dB target.

Of course, there is the coil(s), as well, to factor in, since the load is direct...

Best regards,
Paul S

Probably not the best place to post this, but in response to the related observation in the referenced post:

How can we impliment the PS that seems always willing to blow itself up in service of the Music, if called upon to do so?

Perhaps when the raw power source is itself made to be uniformly acceptable, it is time to streamline equipment PS to the least that is functionally necessary, so it forms less of a "buffer" to the primal source?  Of course, this still presumes proper isolation, to avoid wholesale signal/PS interaction.

The only basis I have for this notion is that it seems like the better stuff resists the effects of bad electricity, the less "raw" potential it seems to have in terms of Sound character when electricity is good.

Paul S

Romy the Cat wrote:

I have to admit that although the Macondo “Water Drop Tweeter” sounds very good but I do not have that “shocking sound” that I got from the “Water Drop” during my initial introduction to it. I kind of always felt it and I thought that I get use to it but now I think differently. What is the problem is that in my current version of Milq my HF channel suffers from the same problem of low-passing die to the high grid impedance as I described above.

Paul S wrote:

But who says an A2-capable amp should actually be driven into A2 in use?  I thought you planned to operate the amp well within its limits, to get the DHT benefits without subjecting the amp to the sorts of stress that always spoils these little things?

Paul S wrote:

But you aren't using the tube rectifier, are you?!?

Paul S wrote:

OK, I think I get it, except you will have a sand bridge along with your crap-load of Farads guarding ground, and your PS out feeds all six "channels", including the DC MF without dividers?

Romy the Cat wrote:

The biggest question I have now is to kill my 6-8dB in crossover on the grid of the driver stage (my VERT much leading idea), to separate driver anode load and  to drop a cap between  the loading and dropping resistors to ground, or to use another lower gain tube

I will certainly go for a voltage divider in driver grid combining the filter and voltage dropper. I measured today everything and I figured out the I need to drop 6.5dB to be even, it means that if I drop 4dB then I will sufficiently remove the unnecessary stressing voltages from grids and will have some ~1dB reserve to play with gain on the secondary of my OPT, in case the tubes have different gain or something like this… I decided to do tame fate with different driver. The sound of SET in mostly comes from the sound of the driver, so I would like to keep the all chem. Of the Melquiades to be in a way the same and do converts the DSET idea into the ordinary multi-amping.

Romy the Cat wrote:

Ok, here is what I end up with. This version is something that I am planning to give some listening, the fist listening as I this filtered down the intellectuals reasons do not listen it. The amp is fully operational, bit made as a test amp inside the 6Ch-Milq. To finish it would require juts to use better wires and a few better parts…  .  Anyhow, let see how this sucker will sound.

I actually was listening the amp today all day. Was listening my favorite mono recordings, driving my right channel from the old 6-ch Milq with single stage MF and driving the left channel with new 6-ch Milq with 2-stages DC-compiled DH. It was very interesting and very generally I like what I heard. The 2-stages MF amp sounded exactly how I would like to – a bit more contrasty, a bit more dynamic and a bit more expressive. In other word the 2-stages MF amp sound very much how the 1-stages MF amp would sound if I was able to load the 1-stages MF amp twice or trice idler. It kind of surprising how in a way similar the 1-stage and 2-stage sound, despite how much different they are. The 1-stage still is a bit more attractive by its softness and elegance but the 2-stage run on garbage parts, in prototype version… so the 2-stage has a lot of room for improvement. I think I need to finish is as is as it looks like a right direction to go… I have my spare “brisk” and I need to evaluate how well everything else.

Yep, it looks, like very close to “it” design-wise. 

http://www.studiomaudio.info/loftinwhite.html

In the way this amp is one of the manifestations of Loftin-White DC-coupled but it has some own things. The driver stage is the Melquiades Tube to keep all 6 channels to have the same “feel”.  For this MF only version the 6E6P huge gain is totally not necessary, so it was reduced by using the un-bypassed cathode bias. The removing of fix bas got rid the DC dependency of grid from ground and made easy to remove the buildup of impedance from the driver grid, effectively extending up the amp bandwidth. Also, with the bandwidth helped the redaction of the driver gain that droped the total Miller capacitance… I like the driver stage now: it has no explicit capacitance and with RL input filter it is as “clean” as it theoretically possible to be.

The output stage is very ordinary running the power tube very spearing but the measures were taken do not bypass the cathode cap in a normal way. The solution is very elegant and was advised to me by a friend of mine, who generally helped with the DC coupling and who due to respectable reasons prefers to be anonymous.

The PS is interesting. I source power form my main 400V PS that I jacked up to 114V. The supply is my regular: Schotky full-wave, LCRC with enormous last cap (I think I have 15K by 450V or something like this). The huge capacitance assured kills ripples but most important it decuples PS from the amp. Since the huge cap forms very low filter and the AC from the tube get shorted right here in the cap down to a fraction of hertz. The folks who use 8uF or 20uF capacitance in SET are playing different games. The AC from this tube goes back passing the last cap and at lower frequencies it getting shorted via chokes and rectifiers. Thos people spend days and years learning how rectifiers, capacitors, wires, secondary of transformers affect sound. It all does affect but only because all that unnecessary shit in the AC back path. I hate this and I juts cancel all of it by using the large last cap. As the result my amp very little affected by the “sound” of rectifiers, capacitors, wires and the rest irrelevant things.

The large capacitance does not affect bass negatively, quite in contrary, but it does make HF a bit “harder”. So, what I do? After the last large cap I have a decoupling resistor and a high quality film cap to ground. The impedance of the decoupling resistor is much higher than the filtering impedance of the film cap and way below the operational frequency of the given channel. Here is where the advance of the DSET topology again show itself off: I have perfect DC with a fraction or mV ripples, the DC that acts like a huge buttery but at the same time, my HF channel run own AC to ground via a high quality film caps and do not see that large capacitance behind this film cap. This application was advised to me by Dima, I see a lot of rational behind it and it become my way to do the things for DSET applications.

In the latest revision David Slagle helped me with input filter that roll off 2.9dB and introduces 3.2kHz first order roll off. It is regular voltage decider combined with RL filter. David built the 80% nickel filtering choke or right value with high resonance frequency (18kHz). The choke and the Obbligato Film cap are in the mails and as they arrive I will be able to finish the amp in a final version, to close it up and start to listen it integrated with the rest channels… I will post tonight the draft 18, the latest one.