The Outerlimit ,

the technically most advanced amp on this planet to date in tube country

Two Versions exist , the Poor mans Version for 3000€ .

and the Monoblok complete amp for 9900€ .

 

Schematics construction pictures finished product pictures
construction notes the prototype pictures  

Project 845 by Benny & Stephan


Updated report

Finished project : article covering journey and finished glowing state

as you will all notice this project has slowly developed in a state of the art design
and uses top quality parts everywhere , 2 versions are ready
the full monty and the poor man's version , still costing quite a sum for those who want it ,
but it is for more realistic people then we now are .

A first simple discussion of basic philosophy .

1 What are we aiming to get ?
2 How much power we want ?
3 How high end should this be ?
4 Aesthetic design criteria ?
5 Budget ?

Basically those are common questions ,

as usual the answers make the difference because many contradictions can co-exist in one sentence or life or a project .

The answers .

1 A beautiful and original amplifier that everyone should want to have and that doesn't disappoint when you turn it on .

PS A small note here , that was when we started the project but as it progressed we realised we had a winner here and wanted top quality

2 Well enough of that wimpy 2a3 stuff , 845 is THE choice , it's in fashion it looks good , feels good , has very nice curves

and probably sounds good and can deliver say 25 watts/30 watts max which is more then enough ( are you deaf ?) .

We achieved 33 watts rms and this is almost the maximum you can get .
3 On par with the best we got so far , no hum tolerable . it's better
4 Small , cute , nice light show , nice meters , nice metal/wood look to fit in anyone's home , meaning modernistic ,

antique stylish. Well 45x45x25 cm
5 Well , here is a problem , I mean as soon as you stick a budget you have to save somewhere , well the problem is solved elegantly ,

by having the same design but in 2 versions : The Outer Limit ( of your wallet ) no compromises and the poor mans version

( a year's savings )but still beautifully engineered and top quality,still way ahead of most diy designs

Known problems of the 845 and tube amps in general:

Ahahahaaa yes all tubes have problems but how do you cure them ,
To get more than 13 watts you have to drive the grid positive, in other words dynamic music signals often peak over 13 watts and if they cannot do this without the driver being able to supply grid current
or the other way around, to have a dc low impedance source from the grid point of view means nononononon .
(again lots of designs have a problem here because it's quite expensive to solve)
Resonances of grid wire and internal structure of the tube must be observed and made sure they're not on the same frequencies as the rest.
( every tube has resonances but the 845 has a much heavier grid so it's lower )and having 2 in series makes the problem only bigger.
Bass response: most tube amplifiers have a round warm sound, it's not the warm that bothers me it's the round because that often
refers to sluggish bass response and or slow rise times.,Not so with ours. This thanks to stiff power supply, high quality output
transformers and primarily circuit design (it is often blamed on the output transformer, but that's just one part of the equation).
About the output transformer: this turned out to be a real hassle and quite complicated.
All transformers resonate as well so this had to be checked with tube and power transformer.
Drive signal: full power requires a swing of 400 volts including delivery of grid current. How do you get 400 volts of clean power?
(in other words being able to swing at least 500 volts cleanly on a scope)
This was solved by using an 845 as a driver. This tube swings easily 700 to 800 volts so lots of margin.
Tube loads :original thought inspired by an article in Sound Practices and also by reading some other peoples' experiences was implemented.
Current sources and shunt regulators: these are not a problem to build, but they make the amp much heavier and produce lots of heat.
So the design is ready, but I am reluctant to try it .We tried 811 as a current source and no difference could be heard compared to the choke-loaded versions, so we decided to stick with the chokes, as they do not age. (and we already have enough lights on top of the amp).
We also designed a semiconductor current and shunt regulator for 1250 volts DC.
If any of you are interested in it, it's 1 pcb with 2 current sources and 2 shunt regulators on it and it requires heavy cooling: 200 watts of heat!
The mosfets have to be mounted carefully and isolated ( 2500 volts on the top mosfet at full swing), 3 mosfets to divide the voltage.
For now this idea stays in the drawer.
The circuit
In the beginning we used an srpp (using ecc81 or 82 or 83) to drive the driver, it measured perfectly, great bandwidth without feedback: well over 200khz .Then as driver we used a 2a3 driver then an el34 driver and a 300B and finally an 845,that easily swings 750 volts.
So 500 volts swing should be no problem for this boy.
The output stage uses a classic 845 and Lundhal se1620/80mA transformer. However as much as we tried this just didn't sound right .
(only at higher power, not at low power), so with a bit of measuring here and there and no real problems found, we started discussing
output transformer problems, so we tried 2 1620 in series, double the inductance and the load the tube sees, and we thought that's
original, that's cute and it looks really cool on paper.
What happened? The sound was better, but on the scope it was clearly slower rise time of 40µs (full power) and bandwidth limited to 15khz.
Since I am not a fan of the mu-follower or SRPP, we tried the 6SL7 in voltage gain mode = common cathode and this was an immediate improvement (we did parallel the triode halves so as to gain in transconductance), soundwise at least, we didn't bother to re-measure as our ears agreed that it was better.
As this thing progressed it just kept slugging at 13 watts output. A short glance at the transformer datasheet and it states max 13w se at
11,6K, which is really correct, so we discussed it and remembered the "parafeed" solution
What's this?
Well it's a trick that lets you use a PP output transformer so that you keep primary inductance a high 300H as opposed to 40H in the SE version. But you must load the plate with a very serious choke of at least 60H. First we tried 120H, which is ok and then why not, 300H.
Just to compare: the designs I have seen using this technique use far too weak chokes and small PP output transformers . They do this in order to save cost and weight, but also because decent SE transformers are rather rare, whereas cheap high quality PP trans formers are readily available and chokes have always been available.
So we are not inventing anything new here, but are doing it the right way. Top quality chokes cost as much as an SE output trans former,
but without their problems and very high primary inductance. These are coupled to the PP output tranny ,a top model good for 250watts of power output and a very high primary inductance: the LL1620PP .We also tried the higher 600H version the 9202PP
custom-made by special order. It's more expensive than a top quality output transformer and it's heavier, but it sounds better,
so it's justifiable.
Fortunately from my 300B/300B interstage design I still have some 120H/80mA chokes left.
This gave very good results. Then we tried a current source using an 811 tube: no audible difference, then we used 300H low rdc chokes:
slightly better and more reliable (the tube 811 might break down). Another problem arose.To block dc, a capacitor of at least 4µF/2000Volts paper in oil is necessary. Fortunately I have excellent contacts and had them custom made by a high quality German company.
This seriously started to sound right, rectifying the filaments also reduced hum considerably,
(I just don't understand why so many people use ac on their filaments and even claim that it sounds better), also here we innovated.
After using Schottky diodes and chokes to unburden the transformers and diodes, we made a bridge using mosfets (unique patent pending).
Here also we had very low rdc chokes custom-wound as they have to carry 7 amps and this means serious heating of the coils.
We decided to optimize the coupling of the 6SL7 to the driver, ( loftin white coupling )and removed the input signal cap. The miller cap in the 845 is quite tough to drive, but was necessary. Don't underestimate this mod; this supply must supply both 845's with negative grid bias and also the cathode of the input tube so any ripple here will be further amplified.
Now we also have a new mos-regulated bias voltage supply, one per 845, which shields every 845 even better and has an advantage.
That voltage is obtained by turning a knob, instead of choosing the right zener.
Connecting the blue led supplies to the ground killed the last hum.
But the result is quite pleasing now, bandwidth reached 80khz measured on the speaker terminals at1 watt and at full power (33 watts 30khz).
A minor note: there is a toroid manufacturer who claims to reach bandwidths of 1Mhz. It does state in very small print that this is reached at 1 milliwatt and no secondary load.
I still think toroids are a good choice as power supply transformers and as PP transformers,
but mostly for power transformers.
At full power, the low frequencies were further improved by choosing the 300H as an anode load for the power tube and also the driver tube.
Of course we use negative bias for both the driver and the power stage. Maybe cathode bias is a lot safer, but then we would have to raise plate voltage and we are already at 1440 volts. Since we loose about 200volts on the chokes, we should consider lowering the high voltage to 1350 volts dc.
After a few evolutions and improvements, fresh attitudeand a couple of months later a new name has risen: The Outer Limit
Why so? Because we have become unreasonable and excited about having the best amp a man can get, so the hell with economics. Just get the best! We had monsters of paper in oil caps mnf for this project, all the caps are mainly paper in oil. We custom ordered the right chokes :300H with very low Rdc.
We decided on mercury vapor gas rectifier tubes, since the Svetlana 6D22 didn't last. (they both failed within 3 months and within a week of each other).
The filaments have new (patent pending ) rectifiers and choke input filtering followed by decent banks of capacitors ( costs a fortune )
Built 2 power supplies and kept the high voltage floating , so as to eliminate the input capacitor .
Now all that is left to do is build the aluminum CNC machined housing .
A few blue leds and a few round meters with adjustable backlighting and a correct switch for sequence and we're done .( 5 relays and digital timer for soft start up/ power up mode )
Stephane insisted we need this and he designed a beautiful delay for switching all on :
After 2 seconds the filament transformers are switched on through a current limiting device , then 1 second later the current limiter is shorted .
After 30 seconds the bias voltages are started using the same technique
After 56 seconds the high voltages are started using the same technique
The blue leds , the orange lighting meters , the tubes , a wonderful light show and also a beautiful sound show , both are on par. .
This version requires :
2 chokes 300H/80mA
2 chokes 7H/160mA
2 chokes 30H/20mA
5 huge paper in oil capacitor
2 845
output transformer 11K/8 ohm
choke for grid
5pcs power supply toroids
1pcs filament rectifier input choke
2pcs 866 tube
1 pcs 6SL7GT

A few measurements :

Voltage ohm watt harmonic harmonic ohm
plate input I anode : load : power : dist. 2nd : dB : dist. 3rd : dB : Zout

750v 93 95mA 3500 14 6,2% -25 0,9% -42 3,4
750v 93 95mA 6000 12 3,4% -20 0,3% -33 3,2
1000v 130 90mA 3500 23 10,7% -20 2,3% -33 3,5
1000v 130 90mA 6000 21 5,5% -25 0,8% -42 2,6
1000v 140 65mA 6000 20 9,7% -20 2,3% -33 2,9
1000v 140 65mA 9000 18 5,5% -25 0,9% -40 2,26
1250v 170 80mA 11000 25 4,2% -28 0,5% -46 1,9
1250 170 80mA 22000 17 2,1% -34 0,1% -67 1,4
1350 220 80mA 11000 28 3,1% -34 0,35% -55

note that this is not maximum output , but measured with 16 volts positive grid drive

greetings to all ,

Benny & Stephane

we appreciate any comments from you