OnRyoku 845SE amplifier

Disclaimer:

The voltages used in valve (tube) equipment makes DIY potentially very dangerous. Electric shocks are life threatening.

Do not work with this kind of equipment unless you are fully conversant with the safety procedures associated with high voltages.

Even though you may be conversant, users and other people in the vicinity may not be. These people are your responsibility.

Consider your construction carefully, and be sure to comply with safety regulations active in your country.

Especially important are earthing, exposed parts, creepage distances, and flammability. This list is by no means exhaustive.

Having gained much inspiration from Japanese valve DIYer’s, and my interest in Japanese language and culture, I saw it only fit to give this amplifier a Japanese name: “OnRyoku” meaning: Strength of Voice. I decided on this when the test results indicated an even higher power output than expected.

My system at the time was 50 watts, but I had determined that 25 watts would do, and this seemed feasible with SET. I had been put off Parallel Single Ended by the near impossibility of adequate matching.

Amplifier schematic:

Why 845?

The 2 most popular triodes in the class are the 211 and the 845. These valves are very similar in construction: high voltage DHT’s, using Jumbo bayonet sockets.

The 211 would seem to be the easier to drive. Its high mu value means the peak-to-peak drive voltage is small, but because it’s miller capacitance is high, it must be fed from low impedance. Although interesting, I decided against it, as I determined that the maximum power I was likely to get was 18 watts. It’s a great shame; because there are still affordable NOS 211’s available.

So it was the 845 for me. Although I’m not a valve snob, it did concern me a little that the only 845’s affordable by me were the Chinese replicas.

What I decided, was that I would start with the Chinese 845’s with a view to saving for KR’s or NOS in future.

In fact I’m so pleased with the Chinese 845’s I got from Jac at http://www.jacmusic.com , that as long as I get a year out of each, I won’t bother to change.

Yet another 845SE Amp:

There are lots of 845SE projects out there. I looked at all I could find for ideas, most were similar, and a few were different.

Circuitry wise, there is nothing new or unique in this design, sorry if that disappoints, but the combination is unique.

Whether it is better than any of the others remains to be seen. What I can say is that at the time of publication, I claim (and can substantiate) the highest power output of any web published 845SE project.

But this is about quality, not quantity. So I’ll try to explain how they are inextricably linked:

Quality vs. quantity:

It’s true: Most people, including me, only use 1 or 2 watts most of the time.

Harmonic distortion is largely caused by curvature in the transfer characteristic of the active devices (valves) in use. A straight line would mean no distortion.

The middle part of the curve is the straightest. From this, we can deduce that the smaller the part of the curve we use, the less the distortion will be. Or, the longer we can make the curve, the straighter our used portion will be.

That is part of my reasoning for wanting as much potential output power as possible. It minimizes distortion.

The other reason is just quantity. But I’ll explain. Other parameters affect sound reproduction: Transient response is important, and is enhanced by having power in reserve.

This amplifier is a class A1 amplifier with the lowest distortion I can get. But when needed, rather than limiting at 25 watts, it has another 15 watts in class A2 to reproduce the transient. That’s called having adequate headroom.

Power output vs. distortion:

Preliminary results:

1 KHz sine wave into 8 ohms
Distortion RMS Volts out Power out (V*V/R)
5% 18.0v 40.5w
2% 15.2v 28.9w
1% 9.7v 11.8w
0.5% 8v 8w

How to get 40 watts from an 845.

Sounds unlikely doesn’t it? Here’s how it came to be:

Conditions:

I had wanted at least 25 watts from the setup, and had been led to believe from other people’s projects that nearly 30 watts was available when allowed to enter class A2. I have described this in more detail below: How I came to use class A2: (below).

I did not want to over-run the 845’s. Nor did I want to run into an impedance of less than 10K.

The final operating conditions I chose were:

Fixed bias.

1250V

65mA

10K Primary.

That’s about 80 watts anode dissipation at nominal mains voltage. This allows for a rise of mains voltage without exceeding ratings of the 845 or the transformer current.

AC heaters.

Driver:

Having a low impedance driver turns out to be a major contributing factor. Even for class A1, a low Z driver is useful; it can improve distortion and frequency response.

Richard Sears, in his article: http://members.tripod.com/richard984/ uses a triode connected EL34, at 70mA. The triode connected EL34 has quite a high Ra, so requires a very large driver transformer to have enough inductance at this current. It seems to work well, and has a high initial cost (large Iron) but low running cost (cheap valves). It also has indirectly heated cathode, so no hum problem. I nearly went his way. Thank you to Richard for sharing his circuit with me.

The reason I didn’t go that way, was that I was lucky to find a pair of beautiful blue KR VV32B’s at a reasonable price. These are one of the families of “Super 300B” valves, and are normally quite expensive. I also hoped for the lower distortion of DHT’s.

Many 845 circuits use 300B’s as drivers. Eg Andrea Ciuffoli’s http://www.audiodesignguide.com/my/845se.html But I did not want to have to replace potentially expensive valves on a regular basis. By running these at only 40mA, I determined that they should last many years. That made me feel better!

Other than anode dissipation, there are a number of significant differences between the VV32B and the 300B: One is the minimum grid leak value. The low grid resistor value (100K) precludes the use of a resistively loaded 6SL7. I tried a number of valves in this position, and found the lowest distortion and ideal gain came with the 6SL7, so I really wanted to stick with it. I was aware of the research into CCS’s and Mu stages, so thought I’d try. Although semiconductors can be used for superior CCS operation, I really wanted to keep them away from them in the signal path. In the end I went for a 6SL7 with a pentode load. I decided to use a power pentode (EL84) because it takes very little extra power for heaters, but when run at low current, should outlast a small signal pentode. The heaters are of course elevated. In fact the mu stage gives lower distortion whatever the impedance is fed. I chose the EL84 only because of availability. Any similar pentode will do.

AC or DC heaters?

I wanted to use AC heaters on the 845’s and the VV32B’s. After a lot of experimentation I found I had to use DC on the 845's and VV23’s. With AC, I could not get the output hum down to the magic 5mV that I regard as acceptable. I think that I could have made AC work if the driver had been a 2 volt valve such as a 2A3. I had considered these before the VV32’s turned up.

The final operating conditions I chose were:

Fixed bias.

500V

40mA

5K Primary.

That’s about 20 watts anode dissipation, way below it’s capabilities. Barring accidents, they should last a long time.

DC heaters.

How I came to use class A2:

The problem: “Pure class A” as some are fond of calling it, is properly known as class A1. In this mode, the grid voltage is always negative with respect to the cathode. No grid current flows*. Capacitor coupling is regularly used. The “general rule of thumb” says we can obtain about ¼ of anode dissipation as audio output power. For semiconductor people, this is a bit like “depletion mode” in FET’s.

To obtain more power in single ended topology, there is but one enhancement: Class A2. For semiconductor people, this is a bit like “enhancement mode” in FET’s.

As we drive the valve harder, and increase the peak-to-peak grid voltage, the grid starts to go positive with respect to the cathode. Almost as soon as it does, grid current starts to flow, because of the “virtual diode” formed by the grid and cathode. This impacts on the source in 2 ways:

1) The current drawn from the source suddenly switches from nanoamps to milliamps during the signal cycle.

2) This is polarity dependent; therefore rectification takes place, spoiling the carefully arranged bias arrangements.

These 2 problems make capacitor coupling in A2 circuits impossible. Even so, there are articles published, sometimes from “respected” sources that claim, “it works”. It cannot.

* A little grid current flows all the time with DHT’s, but it is insignificant compared with that which flows when the grid is driven positive.

Negative feedback:

Like most 845SE amps, this design uses no negative feedback, global or local.

Electrolytic capacitors:

The years I have spent fixing other peoples designs has taught me one thing about electrolytic: Don’t use them unless absolutely necessary.

They are the least reliable passive component. They have inferior performance in most applications.

There are no electrolytic in the signal path of this design.

Since, in a single ended design, the power supply is in series with the signal, I include the power supply in the “no ‘lytics” rule.

The search for Iron:

Sowter:

This took me to all kinds of places: It started close to home with Sowter: http://www.sowter.co.uk . I have used Brian Sowter’s transformers before with excellent results. Brian tirelessly answered my requests for information and quotations. He deserved the order, but the costs were higher than I’d hoped, and he did not have the facility to “Pot” (Encapsulate) the larger size transformers I wanted.

Bartolucci:

A visit to Richard Sear’s site http://members.tripod.com/richard984/845_se_triode.htm impressed me. His 845 design was the best I’d seen, and it introduced me to Bartolucci transformers: http://inthenet.sm/bartolucci/ . The size and weight alone shows there is no compromise. The recommendations being good too, made me investigate them. Unfortunately these were the most expensive yet. After reading Richard’s article, and communication by e-mail, I’d decided to go with a large interstage transformer instead of a DC coupled cathode follower. I’d also (wrongly as it happened), been put off that topology by other’s opinions. The expense was mounting.

I investigated some other smaller UK suppliers, but I was uncertain of their SE experience. Only one inspired any hope in me, but that one failed to answer my e-mails.

Golden Middle:

A Yahoo search gave Golden Middle: http://www.goldenmidle.mailru.com/This is a Russian company based in St Petersburg, Russia. They have a fine range of transformers and amplifier parts. They are cheap, though the shipping costs mount up quickly. Some e-mail communication convinced me they were competent, but the thought of a money transfer to a Russian bank account left me nervous. I asked for help and opinions on Rec.audio.tubes, and Audio Asylum http://audioasylum.com A number of members helped me with information. One even knew the director of the company and said they had a high reputation amongst Russian DIYer’s, though not the “boutique” type. This satisfied me that they were a reputable company, and safe to deal with.

The shipping costs really were very high, so I thought of a plan: Why not drive to St Petersburg to collect? Make a weekend of it. I mentioned this to one of my “Russian adviser” who said NO. He explained that my route would take me far from cities (my only Russia experience had been a visit to Moscow), the roads were bad, and assistance was variable to say the least. It was easy for a westerner to get into more trouble than he could extricate himself from. Foiled!

Thanks to Genn on AA for much useful advice.

AE (Automatic Electric Europe):

Another AA adviser was Bas Horneman. Bas is a South African, settled in Holland. Bas has a great tri-language website http://home.zonnet.nl/horneman/and professes to be an “amateur”, yet his interest has stimulated him to gain more knowledge than many. He advised something completely different: He, and some friends had used a Dutch company Automatic Electric Europe: http://www.ae-europe.nl/ to wind audio transformers. AE was basically a power transformer company, but has branched out into audio transformers. The big boss Wil Blaauw is an audio “nut” and was spending his spare time perfecting audio transformers, in collaboration with some other Dutch DIYer’s. Bas told me they were really cheap, but really good. Now I’d only just met Bas, and was sceptical. But after investigation I found that the most respected Dutch DIYer’s were using AE transformers. They could not all be wrong.

After much communication, Wil agreed to wind my transformers. I’d changed my mind several times, and he’d patiently revised the quotes.

I’d asked for the transformers by the end of the month, so that friends who were travelling from Holland to England could hand carry them. The month ran on.

On the last Saturday morning, the phone woke me. It was Wil. “Yoor tranfoormers woont werk” he said. I gradually awakened as Wil explained to me that he’d done the simulations, and what mistakes I’d made. He said he’d had a conversation with Dick van der Merwe http://triodedick.com (another great web-site), who had confirmed his findings. What was I to do? “Here’s what I suggest” said Wil: “Forget your parafeed idea this time, and use the same transformer as I wind for Dick as your output transformer”. “Oh, and by the way, if you change your driver transformer from 1.25:1 to 1:1, it will double your bandwidth – it’s up to you…” At 9AM on a Saturday morning I could think of nothing wittier than “do it please”.

I am so pleased I followed his recommendations. He’ll do anything you ask, but do listen to him… I’d had a background worry about class A2 parafeed, this was a release!

When the transformers arrived, I was quick to wire them in and see what I’d spent my money on. They sounded great straight away. But there was some kind of “fur” on the output waveform. It went away when I disconnected the CD player… It was 44.1KHz from the CD player DAC appearing on the speaker terminals. That’s quite something for a 10K primary transformer to pass!

AE’s potted transformers come in a crinkle finish of silver-grey and black. It looks far superior to any crinkle or Hammerite finish that I could do at home.

Parts: Sources and specification.

Transformers: Automatic Electric Europe

Output transformer:

10K : 8R

70mA

Amorphous C-core, potted

25watts. (This is equivalent in size to a 100watt push pull transformer)

Size:

Driver (Interstage)

50mA

5K

1:1

30- 50H

C-core, potted.

Size:

Yamamoto UX4 sockets (for VV32B):

Audiokit Italy.

The lovely Sabena will help you out. Their mail order service was first class.

They also send nice Xmas cards.

Output valves:

845

Jacmusic.

Jac van de Walle: good man.

Driver valves:

VV32

Valve & Tube Supplies

Rod Burman

Other valves:

“Fatbottle” on AA

PIO capacitors:

Tony Welsh

PIO capacitors,

Octal sockets.

Kwtubes,

Gintaras Sakenas

Moving coil meters

AquaBlue

Benny Glass. Good mail order service.

Engineering materials,

Miscellaneous components:

Farnell

Maplin

Rapid Electronics

RS components

Thanks for listening to my ramblings:

Bas Horneman on AA and DIYaudio

Dick van de Merwe, for advice freely given

Doug: “Morse” on AA

Paul Barker on AA

KeithW on AA, for patiently scanning and e-mailing articles on Grounded Grid operation, which I tried, but finally discarded. – Sorry.

Adam Stouffer, and

Daniel Normolle

for advice on class A2 on R.A.T.

“Fatbottle”

for advice on class A2 on AA:

Abbreviations:

SET: Single Ended Triode

PSE: Parallel Single Ended

DHT: Directly Heated Triode

PP: Push Pull

NOS: New Old Stock

Links:

RAT Rec.audio.tubes

AA Audio asylum. http://www.audioasylum.com/

DIYaudio http://www.diyaudio.com/

Bas Horneman http://home.zonnet.nl/horneman/

Richard Sears http://members.tripod.com/richard984/vacuum_tube_audio.htm

Steve Bench http://members.aol.com/sbench101/

Dick van de Merwe http://www.triodedick.com/

Andrea Cuffoli http://www.audiodesignguide.com/my/845se.html

Jacmusic http://www.jacmusic.com

AE Europe http://www.ae-europe.nl/

Audiokit

Tony Welsh http://www.tonywelsh.com/

Kwtubes                                 http://kwtubes.com/

Valve & Tube Supplies http://www.valves.uk.com/

AquaBlue

Golden Middle http://www.goldenmidle.mailru.com/

Rapid Electronics

RS Components

Farnell

Some other DIY amps         http://www.firebottle.com/ampage/homebrew.cgi?cat=ga

Data

845 845.pdf

VV32B

6SN7 6SN7.pdf

EL84 EL84.pdf

John Daniel AKA dhaen

mailto:jd@dhaen.org.uk

Here are some pix of the OnRyoku in service:

Front view of amp (click for big picture)

Top view of amp (click for big picture)

On it's stand with power supply (click for big picture)

Wide shot (click for big picture)

Post script:

Well the beast has been in use some time now, so I thought it only proper to update these pages:

What’s good:

It looks good, sounds great, and is quiet, except when Goldfrapp causes the doors on both floors to rattle in their frames. No shortage of LF power.

It was right to use DC heaters on all DHT’s, and use AC on the other valves.

The Damper diodes work well as slow warm-up rectifiers. They look pretty too, but needed to be covered for safety.

“Always on” bias supplies: to avoid current surges if rapidly power cycled.

Mu-stages for driving the 300B’s. The 2H distortion arises mostly from the output stage. (No need to add further non-linearity to the curve).

I’m more than happy with the Golden dragon 845’s.

Neutral issues:

The full 40 watts caused the 845’s to run well into their limits. After testing how much power was actually needed to fill my room, I decided to turn down the HT to 1100v and current to 90mA. This seems to give about 25 watts @ 2%. I’m still on the original set of 845’s and hope they’ll live long. I didn’t measure the 5% power since.

The amp was designed for almost continuous use, but it has since been relocated. I can’t justify 400 watts dissipation for 2 watts of sound. It radiates more heat than the room radiator (which of course is primarily a convector). So now I play it for enjoyment, usually when I’m actually in the room.

What’s not so good - Criticisms/ what I would have done differently:

Power supply:

Having the power supply remote from the amp caused a lot of interconnects. Furthermore, it was necessary to decouple every feed in the amp. Also I decided to fuse very line that led to complexity.

The regulated PSU didn’t happen. Not enough space or enthusiasm!

HT (+B) power supplies are common to both channels. There does not seem to be an audio performance trade-off, but there are other considerations. When setting the bias, the channels tend to affect each other.

In future I shall build mono-blocks despite the extra precautions needed against coupling of hum.

Although I don’t like electrolytic capacitors, the decision to exclude them wherever possible led to a very large and heavy power supply. Maybe some better compromise next time, or active components for filtering.

Input stage:

I used a balanced input stage at first, but was less than happy with the CMMR. Rather than go back to the drawing board I converted it to unbalanced. It may well be the slightly saggy power rails that caused it.

In August a 6SN7 failed. It was a cheapie, so I decided to buy a pair of NOS RCA 6SN7GTB's. That's the version without a skirt. Although they sound fine, they are far more microphonic than the originals. How much is is the valves and houw much is the amp design, I haven't yet determined.

Last updated 1 October 2006 (still happy with it:). Full schematics added.

Feel free to use them for non commercial projects. I'm always happy to aswer questions.

Construction:

This is not a construction guide, it's just how I did it......

The earliest stage was a breadboard.

As you can see, I'm a particularly messy worker!

The chassis top material used was 6mm thick 500mm x 300mm aluminium.

I printed the CAD drawings onto 4 *A4 label sheets and stuck them on as a machining guide:

The scale of the printing was about 1% out, but I compensated by careful placement of the sheets.

Cutting this particular alloy was hell. It was so soft that it bunged up any and all cutting tools.

I had hoped to bilind-tap into it from below, but you could see where the "blind" holes were by thye deformation of the top surface.

Instead I decided to use cap-head screws using a countebore:

Boring the large holes was a "hairy" experience. Because the material was 6mm thick, conventional ways (tank cutter, stepped drill etc) were out.

This is what I ended up with:

Even with a sturdy pillar drill (drill press) at low speed, the vibration was frightening.

The aluminium was bolted to a sheet of MDF the same size, so that when the cutter broke though it didn't snatch. I wouldn't do it again!

I used a local engineering shop to grain and anodize the aluminium. Even so, it is still easily scratched.

Here are some shot of the build:

Small valve sockest and electronic mounted on sub-chassis:

Power input decoupling and discharge resistors.mounted on aluminium bar

Crikey! I've just noticed the gin-n-tonic. That could explain a few things......

Jumbo valve sockes have their own sub-chassis with independent position and rotation

adjustment for setting best asthetics.

Wiring: PTFE insulated silver plated wire. Why? Because it solders beautifully and the insulation is heat resistant. No other reason! The high voltage wires have a further sleeve of silicon rubber.

This is the connector chosen for the umbilical between the amp and power supply. There were rather opposing requirements of 1.2.KV and 4A. The current is adequately rated, but I chose to surround the HV pins with a ring of unused positions. They are made by Varelco and Edac. Both makes are mateable. There are many combinarions of pins, gender and polarization. Caveat emptor.

Of course the power supply has a female and the amp a male. There is also an interlock loop that carries the power relay coil supply.

Pwer chassis construction:

The big cap is the 200uF 1.2 KV HT resevoir.

The 4 bias controls are at the front

Under test. Damper diodes as rectifiers for slow start:

Top view:

Rear view, showing heater smoothing caps and umbilical:

HT supplies are individually fused.

Underneath:

At top is the power relay, controlled by a logic PCB (middle).

The heater supplies are fused using automotive fuses. Their rectifiers are mounted on the aluminium strip (bottom left) with the 300B common mode chokes visible. These are because their heater transformer lacks an electrostatic shield.

On the right is the "always on" bias supply.

Both units connected and working in the workshop.

The power supply now has a mesh cover for safety and asthetics.

Last updated 1 October 2006