Skip to content

Category: Amplifier

Nocturne MKIII


Three years ago I build the first version of Nocturne – a transformer coupled tube headphones amplifier – equipped with 6C45p triode and then last year a second version with a more performing ( and more gain ) D3a triode strapped, both intended to drive with excellent performance a AKG K550 back closed headphones ( 32 ohms, 114 dB sensitivity ).

A few months ago I got to try a planar magnetic open back type Hifiman HE400i ( 35 ohm, 93 dB ) and  I was truly impressed with his sound, but due their low sensitivity, I thought they needed one more muscular amplifier. There is a double dissimilar triode family born in 1950-60 years for television use with excellent features also in  the audio area like 6FD7, 6DR7 and so on. These tubes include in the same bulb a mid-high mu triode like 12AT7/12AU7 family and a low mu, low plate impedance triode ( 800 ohms like a 2A3 ! ) a perfect match for a two stage transformer coupled headphones class A tube amplifier.


The power supply section has been upgraded with a hybrid bridge rectifier ( 2 x EY80 plus 2 x UF4700 ), a double LC filter and finally a power HV mosfet ( STP6NK90Z zener protected ) regulated stage per channel to ensure the minimal residual noise. Due to very low plate impedance of output triode ( 800 ohms ) its sufficient a 3kohms primary impedance for SE output transformer, 50mA max current. Secondary side must be 32 ohms or less for best matching Hifiman HE400i load. No global feedback was used ( only local in the first stage ). Please note the ground path separation technique. At last all ground converge in a single point SG and then at chassis ground through a separator ( bridge rectifier plus a RC cell: 10 ohms carbon composition resistor and 10 nF ceramic capacitor ).

Hint of components: first capacitor in PSU section must be a paper in oil or MKP type, on the first stage cathode ( pin 8 ) use a carbon composition resistor while on the anode ( pin 6 ) a non inductive 3/5 W Resista or similar. For the coupling capacitor choose the best possible quality for your wallet (for those who can afford it I suggest Audyn MKP True Copper ). The resistor on cathode of power triode may be a Caddok 25 W TO220 case on  adeguate heathsink bypassed by a Elna Silmic II ( 50V min ) capacitor. As mentioned above there are other excellent tube candidates ie 6DR7. Look at the specs they are very similar.

Sound quality:

headphone listening is naturally a very intimate and personal experience but I think Nocturne is really detailed, warm and very good balanced sound amplifier. Voices are very fine lifelike and excellent dynamics attacks. Every single detail from tonal instrument color to a random cough from a performer is brought to the foreground. It’s amazing!!!

PS: My friend Tony was very entusiast to listen this headphone amp but he put a question: with my Sennheiser HD600 ( 300 ohms impedance ) there is a mismatch with Nocturne ?

Mmh…… High impedance headphones ….maybe an OTL tube amp……STAY TUNED  !!!!

***** under construction ****

Attila 6C33cb SE Power Amplifier


The 6C33C-b is an attractive tube. This indirectly heated, high transconductance, high current capability and low internal resistance, inexpensive triode, from Russia, with a large 60W anode dissipation, originally designed as a voltage regulator for the avionics in Russian military aircraft, it is the ideal “high current” triode for audio applications.

The downside? The 6C33C-B requires a relatively high anode current (typically 180-220mA at 180-210V) a high heater current ( 3,6 A with filaments connect in series or 7,2 A with parallel connection). For this reason the tube runs very hot ( 250-300°C ) and a ceramic socket and adequate ventilation are required. My choice in this project is to use this tube in SE ( Single Ended ) Class A configuration without negative feedback. Due to relatively 6C33C-b low plate resistance, with a 600 Ohm primary impedance output transformer we can obtain 15 W rms at 200 V anode voltage and 200mA of plate current ( 40 W anode dissipation is a good safer choice for a 60 W max tube ). To adopt a 600 ohm output transformer with a very low turns ratio and, as a consequence, extended frequency decade factor (as compared even to the most advanced modern vacuum tube amplifiers), as well as dramatically reduced leakage inductance.

The 6C33C-b needs a very large voltage swing at the grid input to produce 15 Wrms at output side. It’s very important the choice of input/driver tube type and configuration for best sound. I spent two months for testing various front-end tubes and topologies:

At last my choice was for the circuit below:


It’s very simple with a very short signal path but with outstanding performance. The Lundahl LL9201 input transformers, offers a galvanic isolation from the upstream stage, a signal gain ( due to 1:2 configuration ) and lastly but not last  it allows to adopt a fixed bias per the 6SN7GT. I love this old double triode due her exceptional linearity and sound. Each triode has a raw amplification factor of 20 at a moderately low plate resistance of about 7,7 KOhms.

I remember this valve in the famous Williamson Amplifier  the first true hi-fi design. I try this tube with both triode in parallel. In this configuration the plate resistance halves ( 3,85 KOhm ), a must for transformer load to obtain the best low frequency response. As interstage transformer I use a custom made 1:1 from AE Europe, 5 Kohm on both sides, 50H (nominal ) of primary inductance  ( ). 6SN7GT( paralleled ) plus AE transformer, driven under a 15mA current ( 7,5 mA per triode ) at 380 V anode voltage, provide an impressive set of performance: up to 50Vrms ( 140Vp-p ) of output voltage with a very low harmonic distorsion, wide flat bandwith. This stage is perfect to drive a tank-tube as 6C33C-b.

The output transformer is the beautiful, heavy ( 5,2Kg ) James JS-6130 HF. It is rated for 30 W power output with a primary DC current max of 250mA. Primary impedance is 600 Ohm and 3,26 H of inductance. Bandwith at -3dB is extended from 14,3 Hz to 148Khz.


If made in a single chassis this amplifier will be very heavy due to many transformers and choke. This raises two choices: two mono amplifiers or separate power supply from the amplifier chassis. I adopt this last layout because I only have one power transformer for input stage and one for 6C33C-b output stage. Power supply and amplifier it joined with a big military type connector. Please note the last big MKP 330uF reservoir capacitor in the 6C33cb power supply is put in the amp chassis very close to output transformer. Also 33uF MKP capacitor is put very close to interstage transformer and 6SN7.

In the power supply, I adopt a vacuum tube rectifier ( 5V4G ) for 6SN7GT input stage. The first capacitor after 5V4G rectifier MUST be paper in oil type. In the schematic you can see a LCLC type filter for best ripple reduction followed by two other cells LC to separate right, left channel power supply line. Due to high current ( and voltage drop ) of 6C33C-b output stage, vacuum tube rectifier are not recommended. Here a solid state bridge MUR860 600V, 8A ultrafast rectifiers is for me the best choice. After the first 33 uF MKP capacitor, the power supply is splitted in two sections for Left and Right channel. Here I use a LCLC filter with two double 3H chokes per channel, and MKP capacitors to obtain a very low residual ripple. As you can see in the photos, the power supply chassis is made in two levels: in the downside is put the 6C33C-b part ( HT plus filament transformer ) while the upper level houses the 6SN7 power supply and timer. Please note in the 6C33C-b power supply the double choke wound in same core to improve inductance up to 12H and ripple rejection  ( note the phase point in the positive and negative leg ).

Bypass MUR 860 diodes with 1,5-2,2 nF-1000V Silver Mica or Ceramic capacitors. In general I like fixed bias for tubes in each stage but here in the 6C33C-b stage is indispensable to avoid matched tubes, high dissipation in the cathode resistors, and a higher anode voltage for compensate the voltage drop in these resistors. The 6C33C-b has huge inconsistency of parameters when it is new, so it is very difficult to measure the parameters. The reason that these tubes show varying characteristics is not because of bad quality manufacturing but is that as the 6C33C-b has very high transconductance the internal spacing between the electrodes is very small and therefore even a tiny difference in the spacing will have a big effect on parameters. It’s useful to make bias supply available from minus 50V to minus 120V. Biasing resistor that goes to the 6C33C-b grid should be the best possible quality and has value less than 100K. For grid stopper resistors use carbon composite type Allen-Bradley or similar and Blackgate or Elna SilmicII capacitors in the negative side of bias. Although the 6C33C-b marked for 60W plate dissipation, I recommended a plate dissipation 40 W for a full tube ( both halves paralled ). To protect the ceramic tube sockets and maximize air exchange, the sockets are mounted floating on stand-offs, and a forced ventilation with two fan at downside of chassis put in axis of power tube help to dissipate heat. These two 12 V fans are powered at 6V DC to reduce speed and noise.

6C33C-b burn in: when you take as brand new tube, heat its filaments up for 2 minutes and start the amp with 20mA on plate. In 10-15 minutes the plate current will rise to 150mA – let it be this way.  Burn the tune for another 3-6 hours with 100mA-150mA and then consider it ready to be use at full power.
This amplifier employs a relay-controlled timer to allow the tubes time to warm up and stabilize. These circuits provide a 3-minute delay to pre-warm the output tubes. The capacitor Ct set the delay time ( for 3 min,  100uF is the correct value ).

Except for the negative bias circuit and the time delay put in a common board, I adopt the point to point (PTP) wiring technique. All circuit connections are made directly between components, using the leads of those components themselves and very little wire. In the PTP amp, the signal path is usually as short as possible—which, when done right, can help minimize interference and noise in the circuit and the circuit flows logically from input to output. The circuit itself, therefore, tends to look very much like the schematic diagram from which it is built.


When we “measure the bias” we actually measure the voltage drop across the cathode resistor of that tube. In the output stage, check and adjust the bias voltage every 10 hours during the next 50 hours break-in with a millivolmeter across a 1ohm 1% resistor ( TP2 ). I recommend keeping bias voltage at 200mV DC ( 200mA in the 6C33C-b ) for best accordance tube life/output power. In the input/driver stage 150mV across 10 ohm resistor is the correct bias value ( 15 mA in the 6SN7GT group ). Both bias measurement resistors are put in short by push button normally closed P1 and P2: these resistors are insert in the circuit only when we test bias.

About listening experience I found this amplifier is a quite spectacular full bandwidth ‘muscle’ triode amps. Dynamic attacks are in the first class amplifiers but also delicate nuances like female voices or violin strings are played as the best 2A3 tubes. If midrange is true direct-heated triode without the excess bloom, bass has the speed, grip, tremendous impact and shocking growl that lives somewhere between a good transistor amp for transients and control and the saturation of a high-power p/p valve machine. Iron and velvet: This amp sings freely and let’s go with a flourish.




Madama Butterfly

Clipboard02Another 2A3 SE amp !!!! oh no !!!!

What in the added value in my implementation?
What I present in this blog is the result of many, many….. hours of listening sessions to various configurations and components about a SE class A 2A3 Amplifier. My target was the best design to reproduce the body, sonic texture and fine gradation characteristic of natural sound. This is what, in my opinion, is the best result in terms of listening pleasure, that I obtained. First of all, I think is useful adopt the monoblock concept for optimize power supply and internal wiring ( layout), furthermore ensures short cable between amp and speakers. The starting point for any DHT-SE tube amp in the couple audio output transformer- DHT power tube. In this critical stage I choose the well reputed excellent Tamura F-2007 SE output transformer: 3Kohms primary impedance ( 100mA ), 4-8-16 ohms in the secondary side and Shuguang Treasure 2A3-Z as power tube.
About the topologies I test various type starting with the classic Loftin-White direct coupled, RC coupled with various tube front end. I try out pentode ( 6SJ7 ) or triode ( 6J5 and various other ) in the first stage. In the RC capacitor coupled schema I test several type of capacitors : paper in oil NOS or new production, polystyrene, polypropylene and so on. At last I choose the IT interstage transformer coupled topology with fixed bias for DHT power tube.
 As input/driver I found that the old pentode EF6 in triode mode strapped goes very well with Lundahl LL 1660 10mA interstage transformer. The high primary inductance of 130 H ensure good low frequency response. To get out the best performance from couple EF6/Lundahl 1660 I thought to supply it with a fine tube CCS ( Constant Current Source)/voltage regulator.
Being very simple in circuit design, every part is important when building quality DHT SE amp. You should know the tonal characteristic of the part you are using. For my ears carbon composite resistors ( Allen Bradley 2W ) and paper in oil ( or polypropylene in oil ) in PS give the best audio natural performance. In power supply, I want to try TV damper vacuum diode as rectifier for their excellent specs about voltage drop, huge peak currents, and slow warm-ups (about 30 seconds). Furthermore TV damper diodes are the quietest from the viewpoint of switching noise. Power supply for HT is a fine CLCLC schema. The fist capacitor ( 4 uF) MUST be paper in oil the other polypropylene in oil or MKP. For wiring in the signal path use thin solid core copper ( max 0.4mm in diameter ) teflon or cotton insulated. Only for filament supply thicker stranded wiring can be used. In the EF 6 cathode R-C group I think the best choice is Black-Gate or Elna Silmic II and a Allen Bradley Carbon composite 2 W resistor ( or Mills wirewound ). About the schema layout It’s good thing also providing separated ground returns for both low-level ( SG ) and high-level ( PG ) as in the schema below. To reduce hum I adopted a LM317 based regulated slow turn on power supply for EF6 filament. The N ch mosfet souce follower in the negative bias circuit keep constant the 2A3 grid voltage even if a little grid current flow.


Clipboard01nnnnI often want to listen my favorite music at nighttime without disturbing the neighbors….In the modern apartment with thin walls it’s impossible…….Yes I think I need a good headphone and a suitable amplifier.
But common low cost op-amps found in most consumer audio equipment just don’t have the grunt force and the finesse to actually let the headphones do what they do, which is get out of the way and deliver great music. Well, this was a first for me, building an headphone amp, so I decided I’d prototype an amp and see how it goes. The goal was as cheap as possible, with the minimum of hassle with the focus on the circuit and sound, not a final product so to speak. Here’s how it went. Vacuum tube circuit, have a simplicity unbeatable from any transistor amplifier, however, there are not many projects that use a single active component ( minimalism philosophy ).
This is a single ended transformer coupled amplifier that employs a single pentode tube D3A as active device. There are other excellent candidates : E55L, EC8020, 3A167M, 437A but all are more expensive. Normally a pentode in triode strapped reveal unlinear curve. Not so with D3A. This tube in triode mode give excellent data: amplification factor=70, trasconductance=46mA/V, plate resistance=1.5k; plate dissipation 4,5W and very linear anode curve. The perfect choice for my cheap transformer load, compact, easy-to-building, SE headphone amp.
Take a look at schema: I chose the fixed bias for D3A. Put a DC millivoltmeter across a 1 ohm resistor ( TP ) and turn VR bias trimmer until you read 16 mV ( 16 mA in the tube ) with 170 V in the anode. I noted that this is the operating point for the “right sound” with astounding results in terms of transparency and musicality ( low THD at measurement ). If you want, you can easily use the VR pot for little change and find your optimal “sweet spot” . In this application I chose Lundahl LL1689/18mA as output transformer in ALT/R ( 18:1 ) configuration ( ). A primary inductance of 90 H ensure good low frequency response. Use alternative configuration ( 18:4, 18:2 ) for best match your headphone impedance. Put a GS ( grid stopper ) carbon composite 470-1K ohms resistor on tube grid pin to prevent nasty hf-influences. In power supply I suggest to use tube rectifier like 5Y3GT followed by LCLC type regulator. For the lowest hum and best L/R separation use a Mosfet capacitance multiplier stage. The 0,5 uF capacitors in the schema must be the highest possible quality ( polystyrene or polypropylene type ). About the power transformer the secondary windings must have: 240+240 V at 100mA for HT, 6,3 V at 1 A for 6C45Pi filament, 5 V at 2 A for 5Y3GT and 5 V at 100mA for negative bias.
Sound quality
This really was a big surprise to me! The amplifier has a most detailed, warm and very good balanced sound. Voices are very fine lifelike and excellent dynamics attacks. You can listen to this amp for days and days without your ears getting tired – music comes through fast and accurate without any buttery soupiness, but still open and airy without a hard edge. So for little money one can listen to expensive tube sound!
As mentioned above there are other excellent tube candidates : E55L, EC8020, 3A167M, 437A and so on.. but all are more expensive. I found a small triode with excellent data: amplification factor=52, trasconductance=45mA/V, plate resistance=1k; plate dissipation 7,8W the 6C45Pi. In this case the audio output transformer must have a 5Kohms primary impedance and a secondary that well match the impedance of your headphone. A good choice can be Sowter 9351 single ended output transformer ( ): it has a 5K ohms primary and 4 separate secondary windings configurable for 40-150-300-600 ohms. Adjust the bias trimmer for 35 mA in the tube with 170 V in the anode. I noted that this is the best operating point for 6C45Pi. If you want, you can easily use the VR pot for little change and find your optimal “sweet spot” . 6C45Pi has 2 grid pin connection ( 2, 8 ). Use a GS ( grid stopper ) carbon composite 470-1K ohms resistor on each pin to prevent nasty hf-influences. The power supply may be the same of D3A version or adopt a more cheaper type as tube rectifier 5Y3GT followed by CLC type configuration regulator. First capacitor MUST BE paper in oil type 4 uF/ 400-600V.

Falstaff: hybrid circlotron power amplifier….

     ‭ ‬
There has been considerable debate over the characteristics of tubes versus transistors…..The goal of this hybrid power amplifier is to take advantage of best performance of both technologies tubes and solid state device.‭ ‬Tube is highly linear without negative feedback,‭ ‬smooth clipping,‭ ‬wider dynamic range,‭ ‬characteristics highly independent of temperature,‭ ‬and so on‭…‬but its disadvantages such high-impedance devices that usually need a matching transformer for low impedance speakers.‭ ‬I prefer to use active devices in places where they operate the most linear.‭ ‬For voltage amplification,‭ ‬I rate tubes highest because they have a lot of headroom‭ (‬if the operating points and bias are correct‭) ‬and a long straight usable working range.‭ ‬In the power and/or current amplification stage,‭ ‬I rate discrete power transistors best.‭ ‬With them there is no need for an output transformer and it’s easier to achieve very low output impedance to match complex speaker loads.‭ ‬For a hybrid amplifier,‭ ‬MOSFETs combine well with tubes because they need no gate current.
‭ ‬To build this amp I was inspired to a classic Thirties-style, tube push-pull‭ ‬interstage transformer coupling amplifier due to his highly symmetric phase splitting that was invariant with signal level. ‭ ‬In the output stage I‭ ‬tested various topologies and at last I chose‭ ‬Mosfet Circlotron configuration with a single device per phase biased in class A and without feedback.‭ ‬It‭’‬s preserved the fully symmetrical architecture of the amplifier as a standard push-pull‭ ‬ amps but eschews the output transformer.‭ ‬The output mosfet,‭ ‬biased in class A,‭ ‬never switch off but have a constant well-defined drain-source current,‭ ‬there are no crossover or switching distortions.‭ ‬OK let‭’‬s look at first stage:‭ ‬this is a full push-pull tube stage‭ ‬. For the lowest distortion,‭ ‬the drivers need to have essentially horizontal load-lines,‭ ‬i.e.‭ ‬a very high impedance load relative to the plate impedance‭ ‬…‭ ‬preferably more than ten times higher.‭ ‬Transformer bandwidth is improved by having a low impedance on the primary.‭ ‬Tubes with low plate impedance that fit this application are the‭ ‬6h30,‭ ‬5687,‭ ‬7044,‭ ‬7119.‭ ‬After some listening test session‭ ‬I chose‭ ‬5687‭ ‬biased at‭ ‬12mA per triode,‭ ‬210‭ ‬V to anode.‭ The input transformer T1 is the excellent Lundahl LL1676 in 1:1 configuration. Multiturn trimmer on tube bias ( TB ) allows to perfect balance the stage. For GS gate stopper resistors use carbon composite type: any value from 220 and 1kohm is OK.
‬In the output stage‭ ‬I choose the Exicon‭ ‬ECX10N20R N channel lateral mosfet,‭ ‬high performance device designed specifically for linear audio amplifier. ‬It‭’‬s biased at‭ ‬1A of current and‭ ‬38‭ ‬V drain-source voltage.‭ ‬Two device per amplifier are needed and they must be well matched for Vgs.‭ ‬This reduce the output offset and distortion.
‭ ‬For the power supply‭ ‬in the stereo amplifier‭ ‬are needed four‭ ‬independent‭ ‬( floating‭) ‬units‭ ‬able to give‭ ‬38‭ ‬V.‭ ‬I use two toroidal power transformer‭ ( ‬one per channel‭) ‬with‭ ‬two‭ ‬independent‭ ‬secondary‭ ‬at‭ ‬30‭ ‬V,‭ ‬112,5‭ ‬VA,‭ ‬Schottky‭ ‬diode as rectifier‭ ‬and‭ ‬excellent‭ ‬Siemens Sikorel 3.300 uF/63V capacitors‭ ( ‬36‭ ‬total capacitors in the stereo amplifier‭ ‬).‭
In the PS tube stage I use an R-Core power transformer with‭ ‬260+260V at‭ ‬100mA‭ ‬ for‭ ‬5687‭ ‬anode voltage‭ ‬,‭ ‬6,3‭ ‬V at‭ ‬2‭ ‬A‭ ‬for‭ ‬5687‭ ‬filament‭ ‬,‭ ‬5‭ ‬V at‭ ‬3‭ ‬A‭ ‬for GZ34‭ ‬rectifier,‭ ‬6,3‭ ‬V at‭ ‬0,1‭ ‬A‭ ‬for bias power mosfet‭ ( MB ),‭ ‬12‭ ‬V‭ ‬at‭ ‬0,1‭ ‬A for‭ ‬negative‭ ‬bias‭ ‬of‭ ‬5687 ( TB ).‭ ‬5687‭ ‬anode power supply is a very fine choke input type with split L-C filter after first‭ ‬47uF capacitor for best L-R separation.‭ ‬All capacitor are MKP type bypassed with‭ ‬0,1‭ ‬uF fast type.
This amplifier run in class A. Put a DC voltmeter across 1 ohm/20W resistor and rotate the multiturn trimmer P1 for 1V reading ( 1A current ). Adjust P2 multiturn trimmer for minimum VDC output terminal blocks.