Above is the HF's high-pass filter for the "Water Drop"

I had never seen anything like that. I like their looks. Do they "sound" as expected? I'm fascinated for your efforts with your system. I think I don't have the endurance to go throughout all that, provided I had the time, space, resources and knowledge, which of course I don't have. Good luck Romy, I hope you get where you want to arrive.

drdna wrote:

The question in my mind Romy is: why?  I will be interested to hear your comment on what your goal was to learn about your system in introducing these and what the results are.  After the extent to which you have described the sound of your system, it will be very instructive to hear your comments. I have toyed with the idea of using these for RIAA circuits in a phono preamp, since the required pF values are appropriate and they should offer minimal distortion.

Pay, attention the Melquiades nowhere uses the “bad-type” (location) capacitors. Wherever the circuit is called upon such a cap the amp uses RL filters or the caps with no dielectric (air, vacuum)

op.9 wrote:

Could you help me understand your biasing scheme with this version? I thought I understood the original Milq circuit - but the only channel that looks like that to me is the D channel. I don't get A and B at all! what is the 50 ohm resistor doing? is it just a grid stopper? I assume you can adjust bias and 0v at input in each stage - but don't the two red legs (on channel B for example) work against each other?

http://www.GoodSoundClub.com/TreeItem.aspx?PostID=5533

…and it would be totally different subject.

Let look at the Super Milq’s input just from purely electrical perspective. The basic premise is to get rid of a DC blocking capacitor at input. To do it we pit a series dose ohms resistor and then apply a contra-collage on another side to keep the DC balance at input at zero volts. Then we need to put at each channel own filter.

Here is the initial filtering idea.

I intentionally kept all drafts on the site to show progress and how one thing flew into another. Eventfully the series capacitors were replaced with RL filters and it make too many series resistors in signal pass, so the objective was to combine the filtering and biasing duty of the resistors together.

So look with what we end up:

Channel A: Has RC (substitute for L) filter at input and the 30K filter resistor is used as yellow biasing resistor. Look at the Draft 3. R14 and R27 are the same resistor and if so then we do not need R17 (red), the purpose of Red resistors is apply positive contra voltage on the left side of bias resistors but in the final draft on the left side of 30K resistor there is a common B+ voltage share be all channels. Surely, since we have much less positive voltage “sipping” through the new bias resistor (30K is more then 12.1K) then we need to apply much less negative voltage to the grid, or to increase the value of greed resistor. The 50R resistor is grid stopper; it might or might not be there – does not matter for THIS circuit. (The need of grid stopper for fast tubes is another subject all together)

Channel B: The identical story with the Channel A only instead of 30K series resistor we have a voltage divider that from a prospective of filter/bias is the same fixed series resistor. Pay attention that of I can the cable on the left side of the divider, disconnection the Channel B then I would need to change value of the Red 402K resistor at the bottom of the image as the total balance of positive (coming from 402K resistor) and negative (returning back from each channel) will be changed.

Channel C: this is practically standard Milq configuration with L-pad at input – there is not filter in here. You would ask me why I did not use the same approach as I used in Channels A and B – getting rid of Yellow and Red and letting voltage to pass via the attenuator to common positive. The reason is because my L-Pas is laden type, what has constant input impedance and variable output impedance. So, the circuit balance the 0V dc on left and right from the L-pad, making the channel DC do not care about the L-Pad. If in my L-pad I would flip not 2 resistors but 4 and the output impedance was constant then I would employ the same approach as I did in the channels above.

Channel D: This is quite complex. We can’t used the approach as above as the DCR of filtering coil is too low (250R) and the negative bias voltage shorts itself to ground via the coil. So, we need here a fill scale biasing Yellow resistor to keep the tube happy. What happened at left side of the Yellow resistor? Some voltage comes from positive bias, via 36.6K resistor and across 250R of coil to ground. So, why do we care, we juts  adjust the voltage of common positive bias and it will be it.

Channel E: The RL filter, the situation is identical to the Channel D. It might be a concern that I have we have too high series resistance in a grid chain. In practice is dos not sound as HF-challenged. That impedance with Miller capacitance of the 6E5P makes ~ over 100K low-pass filter … not so damaging for a fast, oscillation-incline tube.

Channel F: It has input cap that does not transmit DC, so we should not vary about anything and juts run one green resistor with negative bias supply. However, some voltage drift via coil to ground, passing through the L-Pad with variable output impedance… Eh, Not good! So we need to put a Yellow “bias stopper” into the game and to get 0VDC on the left side of the Yellow resistor. Then positive will not flow to coil because it will be blocked by the cap and negative will not flow anywhere because it balanced by red resistor. This way flipping the L-Pad affects nether bias nor impedance. Not that the L-Pad has constant input impedance against which the CL filter is written (it is Bessel). It was very important and a change if the attenuator impedance would lead to change of crossover point and subsequentual throwing the tweeter off the time aliment.

I was slowing mingling the conversion of the power supplied of my second Super Milq from 3 channels into 6 channels and I ask myself: is anything in that amp is made in the way that would allow me to close it up and never review it. Sure if I move in very large space then my Macondo would need more power but it will be another story as it  might be a different Macondo then…

Anyhow, in the 6-channel Super Melquiades I found there is one moment that I feel might be done differently. Currently I have the single-stage amps powered anode supplies that power the Injection channel’s 6C33C. Since the Injection channel is a lot of dB down the plate voltage is very stable. Also the last cap in that supply is 15.000uF, so an extra 100mA that the 3 single-stage amps draw are not a big deal. Still, I feel it is a half ass solution I decided to change it while I am revising the amp. I do not know if the dedicated supply would make any sonic difference, I have no problem with sound of the single-stage amps as now, but it to have the single-stage amps power from own supply would be intellectually more pleasing. The most important it would keep the 6-channel Melquiades logically compile and eliminate the itch to thing about the dedicated supply in future.

So, I will be building-in (thankfully I have space and additional wire in the connector between the PS and amp) a new power supply for single-stage amps. It will be a full scale: 220V/250mA separate power transformer, Schottky full-wave bridge, LCRC filtration. I also contemplate, since the single-stage amps do not care any low frequency to make the power supplies only with film capacitors. I think it might be interesting in HF section of DSET. Does anyone know any low voltage (250V-300Volts), 100uF-300uF film capacitors? Do not propose Solens – they sound like crap. I have some 100uF Ansers but they are 650V – too big…

Wojtek wrote:

has supposedly very nice Obbligato line of film HV caps .

Thanks, the high nominals the have only 630V and they are larger than I would like to have. Does anyone know how those Obbligato sound in PS?

hagtech wrote:

So what is with the ferrite beads?  Do you find they are necessary?  Seems to me the resistors right on the socket would be good enough.  Especially with the grounding of the center pins.  I run beads all the time in my digital signals, but avoid as much as possible for analog.

As you can see I did not leave ferrite beads out and I did use 47R grid stopper at all channels. I even put 2 ferrite beads on MF channel (you can see it on picture). I have to admit that I did use Milq without ferrite beads and without the grid stopper (still keeping 50R between the screen and grid that serve the same anti-oscillation purpose). I did not detect any oscillation-related problems… but only with good tubes. I did not play with “too life” tube in the situation when I did not use stoppers. Also, I did not detect or research that the ferrite beads affect sound in negative way. I do not think they should…

The good 6E5P/6E6P are oscillation-free would it be with ferrite beads or not. They might have some minor oscillation for a few minutes after the start while they are cold and then when they reach their cruise temperature the oscillation should be gone. What is interesting that the 6E5P/6E6P’s oscillation is much softer and less annoying (even dangers) then the oscillation of other UHF 100-300MHz tubes. The WE417A, the WE437A, the 6842 when they got aroused were very atrocious. The 6E5P/6E6P oscillations very mildly and gently – it is like a difference between the Plinian and Hawaiian types of volcano eruption. The 6E5P/6E6P when it age might develop some minor HF noise – it is an easy sign to spot and identify a worn up tube. BTW, I did not detect any oscillation pattern differences between 6E6P-DR – the military and tighter tolerance and the regular 6E5P.

So, the story is that I keep the ferrite beads in there. I do not know if it is necessary. My behavior with the ferrite beads is more like the behavior of that nun who said “Good protect anyone who protect themselves”, while she put a condom on her candle….

Can I stay without the ferrite beads and grid stopper? Probably I can. The grid “bias” resistor is very high that forms with Miller capacitance of the tube a near 100kHz filter. The bias voltage the come directly to the grid is from gas tube. The gas tube voltage is not regular voltage but very “slow” voltage and it might do the additional comforting of the grid. So I might go away without ferrite beads and the grid stopper... but the question is why would I need it?  There are some “protective” benefits from them but are they the course of any problems?  Did you detect that a ferrite beads affect sound in any negative way? If so, then might it be the size and the type of the ferrite beads?

Yes, thanks Charles and thanks for everyone who sent the direct amps on this subject.

The ClarityCaps, similar to the Canadian Solens, British Ansars Supersound, the French MKP, and a few others proposed, are all Metallized Polypropylene caps. I am not a big fan of them generally. There are also an Oil-Impregnated Metallized Polypropylene caps like Siemens/EPCOS MKV, V-Cap, Vishay, ASC, GE, Sprague and a few others –  the motor’s start up caps - I am not a huge fun of them ether. I did not try all of the in PS but I did play a lot with Metallized Polypropylene on speakers and in line-level coupling application and I did not like the results.  So, I have large 100uF Ansars that I will use for now but I will keep my eye open for more “intellectually interesting cap”.

Also, I reach a pathetic point of my live when I decided to build a new PS for my amps and… and … and  … and I  accidently have all parts that I need right in my home – very troublesome symptom. So, the PS will be a 300mA torpid with 220V secondary, Schottky full-wave bridge, inputs choke 15H/150mA and 40mF and then RC with 100uF.  I should be arriving at 200V, 110mA of 3x6E5P and most likely I will bleed extra 20-30mA through the choke – I do not know exactly now – I will measure the choke performance when I build it.

hagtech wrote:

I have not experienced any ill effects from ferrite beads myself.  That is why I asked.  Some report sonic degradation issues (like CH at Ayre).  Some say they can get magnetized.  Listening to the masses I am unable to put together an accurate picture of reality.  Yet.  Detecting a pattern of concensus from within the noise is not happening.

Hm, interesting. I remember a long time ago I did tried to detect the sonic influence of ferrite beads on my 7788-7721 phonostage. At that time I was listing the 7788 with and with the ferrite beads and I did not detect any sonic impact. I did it only once and never thought about it again. I never tried in with Melquiades. I do not know who CH is but I do like the Ayre as a company, so I presume that whoever CH is he is not an idiot. Still, if CH is correct and the ferrite bead gets magnetized with time (it is not difficult to measure with a gausemeter) then why should it affect sound? I really do not know the answer. It would be a good idea to remove the long-worked ferrite beads and to listen if the sound was changed. Unfortunately in my case if the ferrite bead was working for a long time then to physically access to it in the Milq would require an archeological expedition…

Trying to take full advent age of DSE and contusing the thoughts that I developed in the flowing thread:

http://www.romythecat.com/Forums/ShowPost.aspx?postID=4328

http://www.romythecat.com/Forums/ShowPost.aspx?postID=4743

I would like to try making this PS for single-stage amps with absolutely minimal capacitance. It will be very interning experiment something that was planning for a very long time.  Let presume that no switching pulses Silicon Carbide Schottky Diodes are free from any problems and will produce right sound (that would make me do not try the damping diodes) then what we have left – the filters. The first LC chain is not filtrating chain but rather filter entering chain (set right Q + current stabilization). It will very-very slightly reduce the ripples but very slightly. The ripple killing is done with the second RC chain where the last cap is usually is very large in my supplies.

Now we have a DSET where the single-stage amps have just 2.8W and care just MF and HF. This opens a very interning opportunity. I have been saying for a while, perhaps incorrectly, those ripples have no affect to sound. Sure they affect noise but if the noise is not auditable then ripples are not important. So, it will be fun: 120Hz ripples against the HF single-stage 2.8W, 35 times gain amps, 109dB sensitivity. I am going to tell you honestly I will not even planning to measure ripples and juts will be listening the noise from the speakers (I will not use my compression drivers but a full range speakers to hear the LF wobbling to assure the absents of Doppler problem in my compression drivers)

So, I ma planning to find the absolute minimal values of the last capacitance necessary for my application. I hope that I will be able to go away with LCRC were the fist cap will be a couple dozen uF and the last cap will be a few dozen uF. I will be reporting the results.

The issue I see is that of intermodulation distortion on the biasing of the output tube.  It will effectively change HF operating point per supply ripple.  Easy to understand if one imagines a gross ripple of 50V.  Same thing happens with 0.5V ripple, just to a lesser extent.

In my latest ongoing amp project I re-discovered an easy way to deal with some ripple.  The stage runs large bypass cap from B+ to cathode to keep output loop small.  Only thing left is grid.  I drive the grid through an interstage tranny, and set the common mode of the secondary to ground via 50k resistor.  This is for dc bias.  The trick now is to run a 1uF or so cap from B+ to the common mode of the secondary (in my case a center tap).  Thus, any ripple on B+ is also exactly seen on cathode and grid.  The tube has no idea anything is going on, as it only sees a constant dc bias at all terminals.  I got this idea from a WE92A schematic.  It works incredibly well.  My noise floor (on proto) dropped maybe 20dB or more.

jh

hagtech wrote:

The issue I see is that of intermodulation distortion on the biasing of the output tube.  It will effectively change HF operating point per supply ripple.  Easy to understand if one imagines a gross ripple of 50V.  Same thing happens with 0.5V ripple, just to a lesser extent.

Jim, I do not know – it is not so clear to me. First of all the subject of the intermediation is kind of complex.  The contemporary ceramic dives in soft runner suspension have intermediations (via Doppler) order of magnitude higher then compression driver and it doses not prevent them to accepted by morons as “good drivers”. I was thinking abut it before and I do admit that in the case of high ripples the modulations might take place from DC ripples  on plate to AC plate but to which degree I do not know. From one point it is very easy to measure the change of intermediations with change of ripples but from another point I do not know what that measurement would mean. Also I do not know where the subjective problem of intermediations would compete with subjective benefit of lowering capacitance. My objective is to find purely subjectively an equidistant between the minimal capacitance (with presumption that smaller cap would allow to use better cap) and the negative subjective sensation that high ripples might be resulted).  With full a range Milq the minimal capacitance of the last 6C33C cap was 750uF but it was restricted purely by noise. In biasing of 6E5P the minimal capacitance after gas tube is 10uF and it is purely sonic value. I was experimenting in there with everything, starting from no-capacitance and ending with 100uF and in there it is MUCH more critical as I is right on grid). For whatever reason the 10uF was a dead spot where everything come together… So, I presume and hope that the 6E5P plate, in my application, might have own max tolerable ripples and min capacitance. I am not certain in it but I do have a hunch that it might be much smalls capacitance then I ordinarily put in…

hagtech wrote:

In my latest ongoing amp project I re-discovered an easy way to deal with some ripple.  The stage runs large bypass cap from B+ to cathode to keep output loop small.  Only thing left is grid.  I drive the grid through an interstage tranny, and set the common mode of the secondary to ground via 50k resistor.  This is for dc bias.  The trick now is to run a 1uF or so cap from B+ to the common mode of the secondary (in my case a center tap).  Thus, any ripple on B+ is also exactly seen on cathode and grid.  The tube has no idea anything is going on, as it only sees a constant dc bias at all terminals.  I got this idea from a WE92A schematic.  It works incredibly well.  My noise floor (on proto) dropped maybe 20dB or more.

Yep, the idea to have the enter tube to “float on ripples” is elegant but from my point of view it is not the way to fight with problems the ripples might course. You might make the tube impervious to ripples but look how much you pay to do so: pollution of cathode, pollution of grid… in the case of DH you would need to pollute the filaments… Is it worth it, Jim? BTW, some of aspects of your “float on ripples” idea, viewed from a dissimilar angle, is something that Dima and I recently “re-invented” for a VERY different and for almost “revolutionary” application. I am sure the readers of my site will be reading about it in a coupled mouth if the implementation demonstrates itself sonically positive.

the subject of the intermediation is kind of complex