Hi-Q speaker amplifier "Ceres 6L6/PP"

Since I am very satisfied with 12AX7 gain stage supported by MOSFET, I was curious how to reuse this design elsewhere.

Here it is, a push-pull amplifier with some non standard feedback mechanism. I am tired of millions schematics with low quality feedbacks, like UltraLinear or even global feedback (where signal from output of amplifier comes back again to the front..)  Those mechanisms are old (outdated). Signal which was at the output of the amplifier, and comes again to his input, has often some phase shifts.
This can start awful oscillation, distortion profile is not always improved by these old fashioned feedbacks.
If you carefully use semiconductors in tube amp, end result can be much better. MOSFET does two things here; gives very high impedance to input tube, and low drive impedance (both not possible with anode resistors!) Gains like these are impossible with anode resistors, and will not work here! (..and cathode unbypassed ECC803 sounds lovely too) It´s possible to have >85 gain for ECC803 (with MOSFET), and >95 on 12AX7WB.
  Because gain is so high, we can expect the impedance provided by MOSFET to preamp tubes is really huge = ultra low distortion! 

Two stage PP amplifier with this feedback system is possible. Datasheets for 6L6GC PP mentions "70V Peak AF grid-to-grid voltage" for full output.
 Even if you used small gain input tubes, just wire input transformes in 1:2. I recommend tweak the fb ratio, 10-40% must be enough in all cases. Due to high sensitivity of tetrodes, 20% or less can be enough.  The grid choke is optional, improves DC bias stability, and does change the sound too. Since choke impedance rises with frequency; it will make feedback stronger at treble frequencies. In bass range, its impedance is small, pulling Schade feedback ratio closer to minimum %. (therefore damping factor will be smaller too).
If my expectations are right, amplifier with choke bias, should have beefier bass and negligible distortion at treble. Treble and mids is where ears are most sensitive to distortion. Smaller DF  will make bass feel strong. (virtually making your woofers like higher Qts..) 

Input sensitivity is about 0,8Vrms for max power out. (85x input gain+40%FB)
Sice this is voltage feedback taken directly from primary windings; higher impedance transformer will improve distortion profile. (then you can use less feedback ratio %)   EL34 or 12E1 (CV345)  are even better here.

STAX headphone amplifier "Pluto 801A/SE" p2

Let me show few more examples, how to put this system on even higher sound quality level. 

Warning, additional expenses ahead.

I´ve seen a 801/VT62 on ebay sale. I was curious what changes
are needed to currently used setup.
About 220€ on ebay for a pair, sure it is worth to take a time and recalculate whole amplifier. Currently used heater supplies are not strong enough, update is neccessary..
In A1 class, is possible to get 155Vrms STAX output, enough for medium volume listening + some music peaks.   (LL1660 transformers wired as 4:4,5)  For serious loud listening class A2 modification is needed. As you can see, some complications here and there; but this thing is doable !  
At same volume (100dB) is possible to have with 801 3-4,5x less distortion; than with 2A3 today. Which is really small, for feedback-less amplifier!   (2A3  15Hz....50kHz (-3dB) ~2% THD)
At this level, I recommend to think about amorph version of output trasformers. (~250€/pc) More music details, and less transformer losses, is where amorph core excels.
One remark, this version has less volume headroom than 2A3 version. But will sound better for jazz and similar smooth music genres.

Update 17.5.2016 - Working on directly heated triode input driver, to replace the modern 12AX7WB. Historical tubes sounds nicer than noval types, but are not very hum forgiving. It took about week of work to remove hum, it is fixed now. Sounds way richer in bass extension, treble is softer but does not lack details! Lovely music with detailed soundstage, best tweak so far.  (MAZDA HL23DD)

Design opinions about STAX amplifiers

I will explain here why i decided to make my own all tube design. 

And why i did not used other designs. 

First one explained, let´s start with "Blue Hawaii" (~4000$). This schematic is popular, and people like the sound. So it should be a perfect candidate to build ?

STAX headphone amplifier
Big reason why i dislike this, are the strong global feedbacks ("masking distortion" was discussed in previous article). 
Strong global feedbacks are unavoidable in DC coupled design, because you have to keep working points very stable. (or magic smoke may come out..) Second reason is the EL34´s, used here as rough CCS (wired as triodes). Their job is NOT to amplify signal, more or less, they are protecting from high voltage other components. Any change done to the signal by tubes, will be cancelled by global feedback... 
Designer would do better job, if he replaced tubes with high voltage JFET´s or Depletion MOSFET´s. (IXYS)
Not saying the amplifier is bad, but tubes here are useless.

Second example is the STAX SRM T1. Here situation is better (tubes are driven directly-they do the signal amplification), but again DC coupled stages and global feedback is minimizing tube sound character. Situation is similar like first example. 
If designer AC coupled the headphones, used CCS instead anode resistors, and made tube endstages working more independly.. 
STAX headphone amplifier
Designers have problems sourcing hi-voltage small power transistors. Those were made for CRT TV (are good also for Stax amps). Since CRT TV are gone, these transistors are not in production. I understand this problem, but see no reason to use tubes in crippled way.
I am interested on tube equipment about his sound profile, amplifying characteristics. It makes no sense to me, to build a hybrid amplifier, where any changes to sound would be erased by transistor global feedbacks. Just a waste to use tubes there. Thankfully still some hybrid designs exists, where tube character is amplified (and preserved) by transistors.

Fully balanced "Eris 6V6/PP" STAX amplifier

During development of cheap Stax amplifier, got a idea about hi-end symmetrical model. 

Sort of cheaper version of WooAudio WES (5000$)

The components selected to best linearity and low distortion.
Endstage would put up >550VRMS, choke max 390VRMS. That's about blow both ears easily, and destroy the headphones.
Input sensitivity is 110mVRMS ~ @ 100 dB Stax. Although it is too strong, with this could be used theoretically with a single-stage phono preamp (30-50x).
For example to decrease the sensitivity a  different input transformer should be used (LL1922 8:1+1 or  4:1+1)
STAX headphone amplifierDue to lower demands of Stax, and  PP mode; stages run in narrow ranges of working points.
Therefore very small distortion is made, also no feedback used here. (bonus)
Second harmonic is canceled due to  PP mode, the 3rd is about 0.05%, the 4th is ~0,003%.
Inductors with a mid tap, is good way to improve symmetry of PP stages and distortion.
800H is a huge number, about XL = 5megohms @ 1kHz .. (This gives horizontal loadline => very low distortion)
Next upgrade would be amorphous core or even a silver windings model (~2000€/pc)..
Lundahl does not make PP chokes, therefore a 5 mA model used.

STAX headphone amplifier "Pluto 2A3/SE"

After a many positive STAX reviews I decided to try them out. Wish i got this idea much sooner!  

Soundwise nothing compares to micro detailed, wide stereo spaced sound. STAX are very low distortion electrostatic transducers (here even expensive speakers have troubles).    

I saw on internet original STAX amplifiers schematics, both transistor and hybrid ones, was not satisfied with them. I wanted to go single ended, with some good "iron" instead 

Lundahl Transformers were chosen, because of exceptional quality and endless possibilities of their products. Windings are split into many symmetrical sections, way better than just winding taps.
You can wire them in many ways, to get desired impedance / voltage ratios.
What was used:
Amorph core input transformer LL1544A (10Hz - 70kHz +/- 0.5 dB @ 0,775VRMS):
Interstage standard core transformer LL1660/18mA (better option-amorph core, twice expensive):
I spent about two weeks, calculating various working points of 2A3 output stage. Everything had to be considered, from Stax headphone capacitive load behavior, optimal tube distortion profile, and comparing each transformer parameters (parasitic and primary inductance, optimal transformer airgap).
Wanted to have this amplifier much versatile as possible, with minimum compromises. Reason why 2A3 tube, and why not 300B is simple - better distortion profile at low currents!

Calculated a few transformer examples, LL2745 in stepdown mode 5,6:1 could be quite good for my old Beyerdynamic DT990pro 250R. With some relay contacts (many of them needed; reason why i abandoned this), it could be possible to use this transformer for Stax also. (2,8:4)  Similar examples: LL1671-4:4        LL1692A--3,5:4.
Low impedance, stepdown type: LL1689 - this could be ideal for dynamic headphones or orthodynamics! But not enough for STAX. (18:1   20-50R     18:2  60-200R     18:4  250-800R)

Days later I decided to go with LL1660 / 18mA; this model is made with higher impedances/voltages on mind. It suits better Stax usage. Wired "backward" 2,25:4 voltage ratio, measured primary inductance is ~33H, parasitic inductance ~20mH (very good for such hi-impedance transformer). Parasitic capacitance ~120pF.
This is good for range 15Hz....50kHz (-3dB) !
As you can see, no global feedbacks used. Distortion is low, compared to typical speaker amplifiers.  This shows how good quality the components are.

Endstage parameters during calculated 100dB output: (skewed loadline due to STAX capacitance load at HF)


Endstage parameters during calculated 250V RMS max output: (earbleeding volume here)

Powersupply and amplifier schematic:
(Mercury vapor thyratrons makes powersupply harder than vacuum rectifiers, and got them cheap from ebay. Choke loaded filter is required. Measured hum is negligible. 80mVrms ripple @ 400VDC)
STAX headphone amplifier  STAX headphone amplifier

During autumn 2015, Lundahl announced new interstage transformers. Those are mostly for 1:1 usage, and could be nice with righ rp directly heated triodes, like 10Y, 801A and others. I can only imagine delicate sound coming from such setup with thoriated DHT´s, matched to silver wires-amorph core transformers!   

STAX headphone features 
Electrostatic transducers are ultra low distortion, superior even against most expensive speakers known. Very thin diaphragm moves between two stators by electrostatic force.
No magnets here - like on typical dynamic transducers, or heavy voice coils glued to thick membranes (movement inertia, and resonances).  Micron thick diaphragm does not have own resonance point, even if it had some, surrounding air would dampen it. Diaphragm mass is small, compared to surrounding air.
Therefore STAX has no problem reproducing smallest music details, stereo space is very wide. Turning volume to max, will not change sound character or introduce resonances / distortion or boomy sound.
Standard headphones/speakers will sound harsh if played too loud; this is not happening with Stax..  It gives you desire to increase volume, very easy to loose self control (and damage ears).

My biggest surprise was listen to mp3, it was not bad as i thought it will be. Good quality mp3 has some hidden potential, and Stax was able to recover it. However, lossless formats or HD quality tracks are different monsters, amount of detail is huge. I am rediscovering music albums again, which i was very familiar for years.

Compared to Beyerdynamic DT990pro, the STAX SR207Lambda sound more lifelike, upper range bass does not sound boomy (subbass is rock hard).
STAX Lambda plays "outside of your head"; amount of detail retrieved from music is huge. You can feel the recording room size, also fading reverberation-for example each hit of the drum.
Midrange on Beyers sounds attenuated, acoustic guitar lacks fullness of sound. Treble is quieter too, like under blanket, top end treble range sounds a bit louder (harsh?) --it´s not sounding "all equal" like STAX does. Beyer´s midrange dip is noticeable, also the stereo size is small- music plays "in head".
Comparison to STAX SR007mk2. If entry level STAX Lambda sounds so good, what to expect from ex-flagship? Wear comfort is very good, even if listening to it whole day. (headphones feels lighter than cable, which is very wide)
Soundwise, it is unbelievably good. Price difference is big, but it is worth it. Enjoyable sound without peaks or dips, feels even flatter than SR207. Probably bigger circular diaphragms responsible for sound improvements. Bass is great, no resonances, just deep growl. Very different than standard headphones.

Sound source is a USB DAC, XmosU8 + PCM5102A DAC. It is a assembled unit from ebay. With unlocked drivers capable of 32b/384kHz rates.

Prototype status, new chassis is incoming:

Update 11.1.2016
Recalculated ECC803 input stage again, small change to working point to improve harmonic distortion. Reverb of instruments stereo space feels a tiny bit better, feels more improvement in firm bass/drums sound.
And more gain is just a bonus.
After (left), before (right)

Update 14.1.2016
Changed from JJ ECC803 to Sovtek 12AX7WB. This has a boxy anodes, it should be bit more linear (it has ~+10% gain). Feels like more details there. Voltages are different, at anode and cathode.

Update 19.4.2016
Modified circuit with MOSFET, now the 12AX7WB sings! Really good amount of detail can be heard. Circuit has smal output impedance, drives 2A3 very well. Impedance for driver tube is high, several hundreds kiloohms (lower distortion).

(I think, for treble roll off is responsible Miller capacitance of end tube, which is ~90pF . This represents additional load of 130k, which lowers gain at treble freq.)


Modern tube amplifier construction

What is worth to do, and what is overkill these days in tube amplifiers.

No. 1, which i do not like in modern amplifier, is the vacuum rectifier. GZ34, 5Y3, EZ81..you name it. Makes your power trasformer hotter, internal resistance is high (it makes supply "soggy"), and you MUST make power supply design compromises. If you don´t, then reduced life or damage is possible. In some cases, higher requirements of amplifier will prohibit to use vaccum rectifiers.
Notable exception are gas rectifiers, like 816/866A, 3B28, 394A, CK1006... These don´t have that nasty high resistance, voltage drop <20V is constant. No matter what current you will load them. Plasma inside acts as approximate supply current load display, and you will see the amplifier is turned on.

No. 2, use good quality transformers. If you go with cheap units, dissatisfaction is quite possible.
Examples are-less iron laminations used (bass distortion due to not high enough inductance); muffled treble (too high parasitic inductance due to low quality/bad design windings) or stability problems due to high values of parasitic components.
 Of course you can fight against them with feedbacks, but with low quality transformers you´ll need use more..(stability problems, see previous article)

No. 3, "silicon phobia" in tube amplifiers is not necessary, if implemented correctly. Good examples are MOSFET CCS in cathode phase splitters (improving symmetry), or gyrator instead of anode resistors (more horizontal loadline for tubes-less distortion and output impedance).
MOSFET is cheaper than a good expensive choke, a cheaper way to improve amplifier.

No. 4, audiophile components. To me it does not make sense, to buy a 1000$ capacitors; when a much cheaper teflon or polypropylene capacitors will do same or better job! Use common sense, avoid overpriced "snake oil"

No. 5, simulate or at least verify design´s working points, you are using. Maybe it´s not optimized properly - you would get worse distortion. Tube amplifier schematics are simple, with minimum number of components. Optimal working points is what matters most. Setting up a proper working points is much harder, than draw some schematics.

No. 6, look at schematic blocks the complex way. Think also about powersupply impedance (ideally zero) across frequency spectrum it works in. Maybe you are thinking that a huge electrolytic capacitor will save you, but this in not the case. Thankfully I have access to expensive LC Bridge, and I was shocked to see, how electrolytic capacitors quickly loose their capacity with increasing frequency. Any manufacturer and even lowESR types behave more like coils above 10-20kHz. Around 1kHz many lose half of their rated capacity.
Foil/paper/mica capacitors have quite stable value with increasing frequency. Paralleling electrolytic capacitor with foil capacitor is a must have!
One reason why semiconductor amplifiers does sound bad, is because of high impedance of power supplies. (PSRR decreases with increasing frequency, seen that in many datasheets)

No. 7, tube amplifiers with external bias (-) lacking a voltage stabilisation is not very wise, and can be also a hazard. Bias voltage must come up first, and should be stable much as possible. Typical steepness value of commonly used tubes is in range 3000-10000µmhos.
Yes, that means if bias changes by 1V, quiescent current will change by 3-10mA. Value will shift further with increasing power-tube temperature. Used tubes does shift more.