Rabbit hole tube amp

I'm building my first amp from scratch. The need to try different topologies combined with the curiosity of what if I try this lead me to build what will be a 45/2A3 amplifier which will evolve over the months and years to come.

I'm taking circuit inspiration from my beloved Stellaris, the knowledge from Pete Millette with my gut feeling from building Bottlehead gear. The initial result is a parafeed 6J5/6SN7 DHT amplifier in a rather large dual chassis amp.

I've been planning, researching and visualising the amp for the past months, now the dream becomes a reality using hard work and determination. Anyone can do it with the right motivation. I've had quite a few parts for it, bough most components used except for the iron which had to be custom ordered.

I'm still an initiate when it comes to this stuff, I know some decisions are schechy yet they were made out of cost, lack of materials or just poor knowledge on my part, this is part of learning and fun.

Buying all the components I've been left with nothing for the chassis, an area I've seriously underestimated yet was so fun to tinker with and find a super cheap resolution. In the end I've gone with 2020 aluminium profile for the PSU and I've had an old wood isolation plinth I've repurposed for a top plate and some hammerite paint for my bottlehead beePre.

I'm going for an industrial PSU look with a super clean amp, no screws anywhere.

Next up is to finish the components installation and start soldering.










Can't wait to try the Visseaux 6J5 Elrog 300B combo

 

Neat.

What is your operating point on the 300B? Are your 6J5s capable of driving it?

 

^ wanted to say c3g 300B yet a 6J5 might be able to do it, running at 8mA with a 4V source. The 6J5 is the start point for 45 output. Sowter nickel transformer will be wired for high impedance initially 150ohms or so output.

Apart from the heater circuit which will soon be changed to a current source not a single electrolytic cap is in the signal path.

PSU done after two more days:


3 days later testing started


Now enjoying the first parafeed experience, my own little slice of heaven.


Still many kinks to iron out. I've got hum from less than perfect grounding or other factors might be involved, 45s are running at 200V with 1.8K cathode resistors and 6J5 is at 110V and 470Ω unbypassed cathode resistor.


This is still very much in prototype phase, many upgrade are planned but just having this puppy up and running, playing beautiful music is a huge milestone.

The two most surprising things about this build:
- just the size of the chassis required for certain components. I didn't want to go dual chassis but had to and what space looks empty is already primissed for future upgrades.
- the low end of a parafeed amp needs to be heard to be believed.

 

Amp completed, prototype is a resounding success. Hum was only present during the middle range of volume, at min and max it was dead quiet. This is a sign of heater cathode leakage, either super bad wiring or floating heater. And of course I forgot to ground the DC for the input tube, I've been so hung on DHTs I missed the indirect needs for a CT grounding. Also tried a virtual center tap with two 100Ω resistors and did the trick just fine.

Why build another amp?
- the need to tinker and push boundaries
- curiosity of parafeed and the need to compare apples to apples with my other amps especially Stellaris
- busting some more audio myths
- having a large supply of valves that need to be burned during my lifetime
- getting the type 45 tube itch
- solving the puzzle
- hunch commercial even esoteric amps are gimped in one way or another due to size, heat, cost, maintenance or danger(I'm thinking of direct coupling here)
- fun

Anyone can build a super DIY amp with enough determination and passion. A year ago I had no idea what a cathode is and now I've build a whole amp more complex than the DNA one.

I now feel the circuit was the easy part, a hundred hours on the web will teach anyone if willing. Pete Millett is invaluable for knowledge and parts, his filament and b+ boards simplifies things massively, it's almost half the amp and the hardest bit of the schematic.

Step 1: Parts
Many say build the circuit then get the parts, I went a bit backwards and chose the design based on the best bang for buck parts I could get. There wasn't any point in trying a typical SET amp as it will be worse than the Stellaris without question. Bottlehead inspired me to go parafeed and I had the major advantage of Sowter parafeed transformers, quite accessible for someone in the UK and custom transformers on the cheap. Primary windings was invaluable and John is a true gentleman.
Large parts:
- Sowter 8665 mumetal cored headphone output transformer
- 4x100uF ClarityCap TC used for 80$ from head-fi
- 20uF and 30uF ClarityCap TC new caps (35£)
- 0.15uF Vcap TFTF from another amp the woo wa5
- custom power transformer and chokes from primary windings 100£ which I'd say is a bargain
- 2 x filament transformers from spirit, one for another project and got a second one (35£ total)
- input heater transformer from a different project
- TKD volume pot from different project
- filament boards and b+ from http://www.pmillett.com/ 100$
- 4.3uF Mundorf Silver Gold Oil parafeed cap for 70£ from ebay, mega bargain
- 50uF Jensen oil caps DHT cathode resistor bypass from previous bottlehead crack experiments
- gyrator boards from Ale Moglia
- lifetime of input and output tubes from years of hoarding
- 3D printer for any custom support parts

I was quite strapped for cash and once I got everything I thought I could start. This was a mistake, the resistors, nuts, bolts, wiries, paint, tag boards, connectors, knobs, tube sockets, accessories etc. were a lot more than expected. My 1K max build went up by a bit.

Step 2: Chassis
Without cash I had to get creative. The PSU was build from the cheapest china imported 2020 T frame aluminium and accessories. Once everything was tightened it feel like a solid piece of metal. 50£ overall cost with brackets and three way joints. A lot of elbow grease to cut to size.

For the amp itself I had a leftover isolation plinth from ebay from years ago. Had the right tools to drill the tube holes and leftover hammerite paint from other builds. A dremel is highly recommended. To hold everything in place I used plastic

Step 3: Wiring
This took so long it might as well be a step in its own especially for dual chassis. I couldn't afford a pair of proper amphenol connector so went with cheap aviation dual 8 pin connectors, one for B+ and one for heaters.
The rest of the wiring is a mix of random wiries. The B+ wiring has super cheap solid core and mega generic OFC mixed with multi strand cheap in the umbilicals and Jensen and Duelund OCC in the amp. The two ombilical wires took half a day to complete. What's worse I'd underestimated the number I could fit, wanted 8 yet only 6 could fit. Not an issue for this build but futures upgrade will need 8 for the headers once I'll go all DHT.

Step 4: Putting it all together
The fun part, soldering, positioning, screwing, building something out of nothing. Grounding is a a hybrid buss bar star one, one star for everything input, one for everything output going to the last psu cap. It may look like a rats nest but it has some logic behind it:



Total cost: Unsure but I'd say around 1000£ or over my initial 1K$ target. Even so it will give the Stellaris a bloody nose in a straight up fight and is at the moment the second best amp I've heard yet this is the baseline. I'm planning to use sowter input transformers and go 01A for input as the next upgrade.

Sound:
It's almost confusing to hear a solid state psu with tube bloom. The Cree diodes I'd say are better than 95% of rectifiers with only the most expensive and hard to find ones being in some areas better. I'm a PSU fanatic at this point and believe it's the most import part. We're listening to modulated DC and that DC needs to be perfect, electrolytics will degrade the sound no matter what even the audio types such as Jensen or Unicon.

Low end is the best I've heard hands down and the 45s are not bass monsters like AD1, PX4, even the Stellaris cannot do this level of low end punch, fullness and detail, not without mercury rectifiers anyway.
Mid range is super clean, correct, yet being used to the 422A bloom in some ways I miss it, in other ways I believe it's unnecessary coloration. Top end can be very delicate when needed, surprisingly so as I've though one needs tubes for holographic sound yet it still misses the last bit of airiness or rounded edges, if anything soundstage is more impressive with some instruments having impossible angles and depth.
The overall speed, tightness, rock solid presentation with the typical 45 magic is a winner. It can only get better from here.

Even with all the upgrades I have in mind and including tubes this is <2K DIY build and will put most TOTL amps to shame.
I'm thinking of converting my BH Crack into a parafeed single stage C3G amp.

 

Thanks for sharing, looks like a really cool project.
You're making me curious about parafeed, but not going to start any new projects for the time being.

 

Parafeed isn't something new, the highly regarded Bottlehead Mainline has this topology, ECP has been using it for a while now and the T4 needs no introduction. I wanted to build a big boy versatile parafeed.

Now phase 1 of this amp is completed, I wanted to put the stepped attenuator from the start but had some issues and had to replace it with a standard pot for easier debugging. Adding back the custom stepped attenuator with vishay dale resistors the difference isn't quite hard to hear. Instrument depth is slightly increased, this might not seem like a big thing yet it has major impact in presence as with depth comes more definition. I went with a bottlehead style coarse/fine setup.



For input I went with 6J5 for the octal socket which is plentiful and cheap so I can make adaptors for anything I want. Various 6J5s, L63 and even the mighty Visseaux are no match for the 4V AC/HL large globe tubes. L63 is basically a better 6SN7 and even the best of the best sound seriously gimped by comparisons. I wanted to try all my 6SN7s but I know it's futile to expect anything better. ECC32, 13D2, B65, TSRP, Reliatron, Bad boy etc. just can't compare with the stage, accuracy and linearity of these tubes. Warmth or as I hear it treble rolloff with rounded edges is plentiful and if that's what one is looking for or bounded by the amp tube selection then by all means.

The beauty of this amp is the ability to convert it for any tube in the goal for the best of the best search. No switches this time, to convert it for PX4s I'm changing the cathode resistors and adjusting the regulated DC boards for the right values.

This was my phase 1 goal , all 4V tubes with fat bottle PX4 and Mazda AC/HL


The 45, both ST and globe are very fine tubes and endgame yet the PX4 offers a more complex, rich and detailed sound.
Also I'm running the PX4 at just 6W dissipation, 200V and 29mA and I'm barely able to raise the volume, looking forward for a very, very long lifetime.
Finally I can enjoy what may be unique tubes, mesh PX4 AVVT UX4 tubes


Future major planned upgrades:
Phase 2 - Rod Coleman regulator for output tubes and 100 steps attenuator
Phase 3 - DHT input with 26 or 01A with filament bias and nicket input transformer as second stage.
Experimental 1 - Filament bias output tubes, Osram P2 should work and first attempt without cathode cap. LP2 DHT tube input as second stage.
Experimental 2 - PX4 with filament bias, will need beefy CPU coolers to tame the heat, a pair Thermaltake Archons are on the wishlist. For a full craze setup, 2A3 or 300B filament bias with water cooling.

 

If you like buying super expensive tubes, why not try a PX25 driving a PX25?

You could try to have the two tubes naturally cancel out their distortion.

 

I've got nothing intelligent to add to this thread but it's freaking awesome and inspiring. After many months of debating it, this pushed me over the edge to finally order Morgan Jones' book. Open to any other suggestions you have to get smarter. Thanks and keep up the great work.

 

While I do like your thinking PX25 isn't on my list, just because it's more expensive doesn't make it better, if one needs the additional power yes but from all account PX4 seems a better tube having a different family tree than PX25 so it's not the case PX25 is an uprated PX4.

Also it's an output tube and while there have been successful 45, 801A, 46, 71 input tubes this one really doesn't work.
Simple is best. I will be going DHT driving DHT.

Push pull is an idea yet 4 output tubes that need to be matched within 1% with double the PSU requirements, dual input with interstage or phase splitters, just no. I'm powering 100dB headphones, not speakers.

 

2nd this. Nothing intelligent to add but inspiring. Makes me feel like a lazy bum in comparison.

 

I was kind of half joking with you.

Im not suggesting the PX25 because I think it's a better tube. I am suggesting it because I think it could give you an interesting topology. Kind of like what sakuma does with his amps.

 

You speak of gods, I'm a mere mortal by comparisons. I think he used some crazy custom 1:25 input transformers for the big guns.

 

Yeah, thats kind of why I want to try the px25 into a px25 sometime. You don't actually need a step up transformer.

The px25 has a gain of 9.5 and a bias of 35. You can actually just barely make that work with a 2v input and you would certainly be able to make it work with a 4v input with no transformer gain required.

Think of it as a sakuma lite amplifier.

 

Since I'm bound at home without work, what else to do but experiment, listen and tweak.

Rewired the input for a single 6SN7 to try other combos with the 45 outputs. 5-6 6SN7s, 6F8G, ECC40s, ECC32 etc. mostly good stuff yet nothing soul numbing even with the globes. I feed they don't have enough bite and presence for my preference, hugely detailed, fast, impactful even if soft, uber smooth. I do see why it has a cult yet it's not what I'm looking for.

Rolled the imphamous 46s, somehow easier to peel the layers of music, striking amount of correct fine high end detail yet soft impact. 45 fanatics will enjoy this, maybe not as much as the globes but a step up from the standard.

Adjusted the circuit for 2A3, my love hate tube, 37mA, 200V. I do like the neutrality and sound signature but it's always been plagued with technical difficulties in one way or another even for the best of the best.

My only old 2A3, Sylvania dual flat plates ran with a matching 6SN7 BB. Holy mother of God, synergy bingo. Deep, layered, insightful, holographic and soul numbing. Never has it sounded this way in the Stratus or Stellaris, an average tube yet somewhat obtainable tube in this combo it's in the hall of fame.
Even weirder I'm running the 6SN7 at roughly the same bias just with silicone as the load instead of a resistor, in the DNA amps these are a mess, mushy and slow yet here their character is shown in full based on rolling these on other amps such as Liquid Glass, Teton, Peak etc.

If the bad boy is this good then the other tubes I classify as better should be, right? A metal base ECC40 comes close, ECC32, TSRP, L63 are still trailing behind, 6SN7W is strong yet loses the bloom being closer to a SS voicing. Still waiting on parts to run the BL63.


ECC32
Loved the mid range, fat kind of uncontrolled low end, exceptional tone even if warm. These traits were consistent across 5 amps and was always a 4th tube never cracking top 3.
Usually used in 6SN7 amps, I'm now thinking all of them are gimping it hard. It's typical to use a 22K plate resistor, perfect for the 6SN7 yet for the 14K ECC32 it's far too low, 40K+ at a minimum and the spec sheet wants it at 200V+.
Synergy with NOS PX4 is off the scale with truly remarkable tone and presentation and is the combo to look after.
The bad boy has good synergy but falls short here.

PX4 sounds quite different than in the Stellaris and overall the new amp is less sensitive to tube rolling putting its base signature quite hard on the sound.

Total of 42 combinations tried so far, still many more to go.

 

I'm flawed and lack patience in anticipations of results. One should never ever do this yet I've installed all the upgrades I could at once for the full phase 3 implementation.

Multiple sunrise to sunset days of soldering, measuring, printing, debugging, testing, redoing made the end feel as finishing a marathon.
Work hasn't given me this sort of dedication and long term focus in a very long time, hours felt as minutes, real challenges to overcome for a novice DIYer.






Main upgrades are 01A input tube with SiC filament bias, Rod Coleman's excellent filament regulator boards for input and output, Sowter input transformers.

I'm calling this amp the Basilisk, it's a full DHT gyrator loaded single ended amp designed for PX4 and AD1 output tubes and in its current configuration 40ohm output headphone amplifier passion project.

There's still a lot left to do, various tweaks and to build a metal chassis yet for the foreseeable future I'll enjoy the fruits of labour, pure set bliss and finally what sounds like the full potential of the HD650.

 

It's all fine and good to audition gear in isolation yet the true test comes when A-Bing, so I compared the Basilisk against the Stellaris.


Stellaris has:
- more harmonics in the mid range
- finer detail and shimmers
- far wider soundstage
Basilisk has:
- obvious cleaner transients and instrument definition
- just a hair punchier/deeper low end yet testament to the Stellaris to match a solid state psu, same level of control.
- more quantity in top end yet sharp, coarse by comparisons
- compact stage
- less tone purity

The soundstage is what threw me off, it's very well defined yet 30% or so smaller and suspected rectification might be the difference. Three hours of work later, a hybrid Graetz-bridge is in place in an improvised socket.


The conversion was as follows:
- removed the diodes D1 and D2 from the PMillett board
- added an old filament transformer as the heater, 5V 2.5A perfect for a GZ34 not so much for 5U4G.
- wired tube anodes in parallel with the board, and changed the HV transformer tap from 250 to 275 to account for the voltage drop of the tube.
- linked the cathode of the tube to the D1 cathode on the board. Basically this:


The major other benefit is soft start for the high voltage. Even with a thermistor on the primary the instant the transformer is switched on full B+is applied to the tubes. The tube heaters are already warm as they're on a separate switch that's turned on first but still... Now it takes around 10 seconds and B+ gently rises to the exact same voltage as the full SiC diodes before.

Round two, all the voltages/bias are the same, the only difference two SS diodes replaced with one tube.
Sound stage opened up, detail became very fine and delicate, lost a bit of focus yet gained so much subtle bloom.


Rolling some different tubes on the Stellaris and using the Sowter input transformers for both amps at one.
Stellaris has:
- a wider vocal image and slightly de-emphasised smeared background
- still more harmonics but the gap is much smaller
- a low mid/upper bass bump can be good in some cases but mostly has a muddy/veil/heavy effect track dependent.

Basilisk has:
- still more defined edges but the contrast is not as powerful as before.
- the same size soundstage without the mid focus, background is much more present and clear
- airier top end though not by much, not quite as detailed but within spitting distance
- sounds cleaner, no bump
- quintessential great tube sound
- a few rows more forward in presentation typical of PX4
- amazing rock capabilities

They're more similar than different, tubes will have a big effect but for everyday use I prefer the Basilisk voicing, cleaner, airier and just as savage in the low end.
I didn't wan't to use a tube rectifier yet I can't deny the difference and there's no going back to a full SS psu. Damn.

 

Taking a break from the electronics and drafted a chassis. We'll see if it it's possible to actually make it. Colour is burgundy.

 

This has been one hell of a journey, I added the rabbit hole title as a joke yet it turned out to be true. This amp, now retired and disassembled was my initiation into DIY tube amps, crucially it showed me the potential of a circuit and how to fine tune it to my liking resulting in a sound unlike any other amp I've heard. It must have had over twenty different incremental distinct versions finally setting into a modern PX4 amp.

This is how it looked in its final stage, the amp top left, psu bottom right. Top right was a new type of design which evolved into a sort of power amp.




This was a different take in a smaller footprint, now disassembled as well.


A Push Pull EL84 without caps in the signal path. Everything is a compromise in audio and the lack of capacitor as energy storage isn't necerarily a positive. A fun project overall, 10W and built to see how the Susvara will scale, balanced output just because why not.



From all this I gathered a few parts and supplies and I'm constantly tinkering, visualizing, building something.
This is how a full 801A Elrog amp would look if or when I get the time to build it.


A preamp is also in the pipeline with 26, 01A or LP2. The case work is no easy task, I try to not get to a point where I have to backtrack because of a tolerance or measurement mistakes but hey, that's how I learn.


At the moment I'm working on an upgraded PX4 amp, I loved how that amp sounded, different signature than the spud 801A and knew I'd have to build it the way it deservers. A few critical parts are on the way but in the meantime I have suitable replacements to get it going. This will be my statement prototype amp PX4/AD1/2A3 good for about 3W of class A1.



The whole experience has thought me the headphones are just a part of the chain and overall of minor importance. The range or sound quality I've experienced over the years from dynamics especially HD650/HD580/HD800 but different amps and amp/dacs is frankly astonishing and why I don't trust any reviews because the chance I'll ever hear the same sound as the reviewer is basically zero.
What I value is to pay critical attention to the areas that damage sound quality e.g. volume attenuation, signal conversion, gain stages, tube loading, coupling, energy storage, biasing technique etc. etc. etc. with components in the top 5% for their class. The goal is to have music as addictive as possible and a refuge/recharge station when I need it.

 

Why did you disassemble and retire your creations? Just for parts for new ideas or were you not quite satisfied with your results?

Heck of a lot of work. Interesting read. Makes me curious about doing this myself if I had the time.

 

I'm very proud of each and every amp I've built yet the first amp doesn't have the improvements of the 2nd, same goes for 2nd and 3rd, 3rd and 4th etc.
I'm doing a noamp concept for now, a bunch of hand picked components I can spin up to whatever I want.

Plus this is the the fun of DIY, DIY-ing everything, learning in the process and enjoying the fruits of labour.

Earlier I converted the 801A amp from the active load with gyrators to lundahl chokes, I'm using a sort of quick connect plugs and bare frame with hook-up points everywhere, 20 minutes of work and post voltage checks I have a different signature, this one reminds of of the Stratus but grown up and all it took was 150£ in parts.

A while ago I realised I want all the amps and no single amp will do it for me, even the Stellaris. So I'm working towards the goal of having it all a single modular chassis, it's a long journey to that goal and enjoying every step.