6L6GC EL34 Single Ended Tube Amplifier

[Mars]

Hi All

I have been listening to my 6L6GC Single Ended amplifier through my Lowther loaded Jericho horns for a few months. The sound leaves little to be desired for, but I suspect the transformers can be upgraded.

I headed off to Yves's Output Design Assistant for some indication of what needs to be done:

Specifications:

• 15 watt power rating
  
• 20Hz to 20kHz bandwidth
• 3000 ohm Primary Impedance into a 6 Ohm speaker load
• 24 Henry minimum primary inductance
• low leakage inductance
• less than 10% losses in iron and copper
• try to keep flux low (<1Tesla)

I was surprised to see that the transformer is going to be rather large. To keep the losses/leakage inductance/primary impedance within spec, I had to opt for EI114 laminations!

I would have loved to have the flux closer to 0.5 Tesla, but this configuration seems to satisfy all the other specifications. I guess I can experiment with different air gaps when the transformers are rolled.

This seems acceptable (I hope )



This would be a significant improvement of what I currently have in my amp.

[handsome]

why are you matching to 6 ohms Karel? for the lowthers or for general compatibility?

[Mars]

Good question. I have seen 6L6GC being used with different OPT primary loads:

See this design (BTW it sounds very good - I have tried this circuit a few times):


I used the Hammond 1627. I noticed that I like the sound better when I connect my 8ohm speakers to the 4 ohm tap; thus effectively having 5K anode load on the 6L6GC. Since I know most speakers have lower than rated impedance at certain frequencies, I generally design for a lower speaker load than specified. I might have to review this policy.

In the circuit above the primary load is 2K5. B+ is more or less what I am aiming for. I am also doing this design for a Decware HDT tower (http://www.decware.com/newsite/HDT.html) set of speakers.

I had a quick look at a possible design for a 2K5 into 8ohm SE OPT. This is the result:


It does look good: low leakage inductance/low losses/low flux density and it fits on a EI96 core.

[dotVIBE]

well, certainly an interesting design.

[alternativeroute]

I am also doing this design for a Decware HDT tower (http://www.decware.com/newsite/HDT.html) set of speakers.

[OFF-TOPIC]
Ooh, slightly off-topic but tell me more about the HD1's. I ask as I have some badly made HD1 cabs in the garage. My intention was to remake the cabs... Currently I am trying the Decaware modded Fostex FE206e drivers in Martin Kings ML-TL...

Martin Kings Project 5; ML-TL with Fostex FE206e by arclients, on Flickr
[/OFF-TOPIC]

I have attached picture form your previous post...

[handsome]

interesting design indeed. it uses two 'lots' of feedback. both global and via the ultra linear connection. have you tried listening without the global feedback loop? i ask simply because so much feedback is obviously creating a higher damping factor, yet you are using it with lowthers which i have heard, prefer a higher output resistance from the amplifier. would be interesting to hear your opinion of the difference in sound.

[Mars]

Currently I am actually listening to 6L6GC SE with a 6DJ8 totem pole current follower driver. But to get back to the EF86 design.

Actually it does sound better without the global feedback. I have also added a bit of feedback to the cathode of the 6L6GC by using the secondary of the OPT. Local feedback sounds better than global feedback in my experience. This way you get the linearity of triode mode with the extra damping/bass/clarity of UL.

Note that my current OPT are compromised. They have barely 6H primary impedance, but still sounds great through the lowther horns.

[Mars]

I think I have found a way to satisfy all parameters:

By having a dual layer secondary, I can have a 4 ohm and a 8 ohm tap on the OPT. It all is rolled on a EI96 bobbin, which has a decent size (nearly a square core section). The flux density is within the limits of the laminations (<1.1Tesla) and the losses is just over 10%. The primary impedance is quite high, so the bass should be good.



I think I might roll this one

[Mars]

Hi All

It has been awhile since I worked on this project. I think it is partly from being lulled into some type of sonic trance by the prototype, which plays every day.

Luckily someone ordered a copy, and I have been pulling out all the plugs to get the design completed.

The biggest challenge was the OPT's. They are now complete. Here is some photos:

The place where it happened!



Ultra Linear primary



The iron is assembled

[iondb]

Looking good. You guys make this stuff look so easy!!!

[Mars]

[Ampdog]

Karel,

If I might make a suggestion ....

That 4,7µF screen decoupling capacitor for the EF86 is unnecessarily large. The -3dB point is somewhere at 0,1Hz! One could easily use no larger than 100nF, which will avoid having to use an electrolytic with its higher leakage rating or an over-expensive polyester type. I have noticed in the past that particularly RCA is fond of using rediculous screen bypassing capacitors. In one circuit they specify 0,25µF for a screen resistor of 820K-ohm, in the next (for the same topology) the value is 40µF! And that was before times of better materials as used today.

[handsome]

Looks really good karel, the simplicity of SE tubes amps is always a wonder to behold! so how does it sound and does it hum any?

[Mars]

Thanks for the input Ampdog. I am still learning about the screen decoupling cap. Usually the screen decoupling is earthed. Does the smaller values also hold for cases where the screen decoupling cap is tied to the cathode?

To clear up some confusion: this amp uses a 6DJ8 SRPP driver. 254V on the totem pole (127V per triode section) running at about 8mA (biased with 330 ohm resistors).

[Ampdog]

Hi Karel,

Yes, correct screen (and cathode) bypassing is a real 'gedoente'. It is easily done with formulae as per RDH (IV) p. 496 - 501 and elsewhere .... but impossible because one does not have the parameters needed there.

Technical Content:
Perhaps of general interest: One starts with the basic relationship of "R = 2.pi.ƒ.C"  (what is the keybord code for pi?), but when bypassing active elements, there of course needs to be an increase in C because of amplification involved, rather like with the Miller effect. Thus R.g2-C of 1 megohm and 10nF will give a passive -3dB point at 16Hz, but using it on G2 of an EF86 will raise that frequency to over 40Hz because there is gain from G2 to anode.

Here I chicken out and use the engineering way. It is easier to hook the circuit up with the specific components and measure with a signal generator, changing C until one has what is wanted (and tabulate the results for future reference).

Per basics the screen is bypassed to cathode (viz. a 'common cathode' topology). This will be the same as taking it to earth when the cathode is properly bypassed. For those looking at the phase effect (important in l.f. stability in amplifiers with NFB), it is useful to realise that phase shift initially caused returns to zero at very low frequency where capasitive reactance become large. THis can be a handy 'tool' to keep l.f. stability within bounds.

In the classic power amplifier topology where NFB is applied to the first stage cathode the screen bypass is returned to the cathode and not earth, so as to preserve a pentode configuration in the first stage. In some designs this is overlooked, giving a funny kind of 'tapped screen' topology, which I have not done measurements on yet. I would believe that that is counter-indicated, seeing as how the NFB-signal is dumped on the input stage with power stage distortion (higher then the rest) in series with the screen.

In your case with the SRPP, Karel, the situation is of course different.