yet another modified 5-20 amplifier

[DF96]

I am part-way through building a modified Mullard 5-20, using Sowter transformers on a Hammond chassis - see my 5-20 build. I use half an ECC81 (12AT7) as input for each channel. Since taking the last picture I have built the HT supply, and reformed the reservoir capacitor (I bought it a few years ago). The next job is to check DC and AC voltages in the input and phase splitter stages.

I will try to keep the web page updated with progress.

If you follow the links you can also find lists of Brimar factory codes and Pinnacle (a UK valve rebrander) P numbers. These don't seem to appear anywhere else on the web, and for some reason that well-known search engine doesn't include them in the index.

[Eli Duttman]

I'm very fond of Mullard style topology, but the 12AX7/ECC83 is NOT a good choice for the LTP. High gm/low R_P types make better phase splitters, as they have good drive capability and provide resistance to slew limiting.

IMO, you still have too much gain. Think in terms of the 2 VRMS O/P of a "standard" CDP, as the source of the driving signal. Employing the 12AT7/ECC81 in a PP amp makes considerable sense, as the net open loop HD spectrum is an ear pleasing "waterfall". The high gm and low R_P of the 'T7 make it an excellent choice for the LTP. A linear triode of modest mu as the voltage amplifier yields all the gain necessary. The 12BH7/CV5042, 6GU7, 6CG7, and 6FQ7 are candidates for the voltage amplifier job. I hope your power trafo has the extra 300 mA. of filament current needed by the better medium mu triodes. IMO, the 12AU7/ECC82 lacks the linearity necessary in a voltage amplifier.

[audiowize]

I strongly agree with you Eli. I messed around with different combinations in my 829B push-pull amplifiers, and I settled on a low gain version of the 12AT7 for the first VA stage and an E80CC (12V 6CG7 more or less) for the LTP. In my next project, I will use a 6AV11 to handle all three stages (three 12AU7 triodes in one bottle), although the 6U10 is also tempting in that it has a higher mu triode along with the other two, but that would involve using an interstage transformer with some step down to get the gain correct.

[DF96]

Thanks for your comments. I decided to stick with the ECC83/12AX7 as used by Mullard because provided that it is not asked to drive much current it should be more linear than an ECC81/12AT7. Unlike beam tetrodes, the EL34 is reputed to be easy to drive. I will redo the slew rate calc - I'm sure I did it some time ago but I don't know where I put it. Does anyone have a good figure for the input capacitance of an EL34 in UL mode?

Because of its non-linearity the ECC81 makes an excellent RF mixer. I can get away with using it at the input because signal levels are lower, when NFB is taken into account. The ECC82/12AU7 is too non-linear for audio, except possibly as a cathode-coupled line stage when the curvature in the two halves roughly cancel each other out. Yet people often seem to use it for cathode followers etc. where an ECC81 would be much better - I guess they get confused between the different requirements for low o/p impedance for different types of valve stage.

The main non-linearity in the original 5-20 is the output stage. This is because the EL34 are biased to somewhere between class A and class B, so their small-signal gain is significantly greater than their large-signal gain. Mullard did this because they used pure cathode bias. I am using a mixture, so hope to reduce this problem.

[audiowize]

Chosing a 12AX7 over a 12AT7 or 12AU7 is definately the choice of someone who looks at datasheets and doesn't sit and listen to each tube. The low current/high gain 12AX7 falls on its face when you can implement a 12AT7 instead (given that you don't need all the gain of the AX7). The AT7 on paper looks every bit as linear as the AX7.

Running 3-5ma of current instead of .5-1, the reduced plate resistance, and a more tamable gain make the AT/AU better choices unless you're building a phono preamp.

Listen to all three, the AU7 is a sweet tube with choke/active loading, the AT (or a lower gain variant) is generally my choice as an input tube, and the AX7 is ideally designed out of everything.

[Miles Prower]

So far, I haven't had the need to try the 12AX7A, though I did use the similar 6SL7. Same deal: high u-factor, large rp (static and dynamic) low current. In that case, I included cathode follower grid drivers between the 6SL7 LTP and the finals. Cathode followers present a nice, friendly load: Hi-Z and with minimal capacitance so that high frequency slewing doesn't become a problem. No way would I try either a 6SL7 or a 12AX7 as a driver for the finals.

[DF96]

If I had beam tetrode finals I would probably do it your way.

Anyway, lets look at the ECC81 splitter. It needs to run at a reasonable current, say 3-5mA as suggested. As it has lower gain than an '83 it will need a higher input voltage. This means that it will hit grid current earlier. Then you either accept the resulting 2nd HD or drive from a low impedance like a 6CG7 input stage. The result is much lower overall gain. This might be OK if your only input is a CD, but I have older (analogue!) devices with lower output. I would have insufficient spare gain to use enough NFB to linearise the output, so better then to run with no NFB than not enough NFB. The result is a completely different design, with a different set of compromises. Maybe I will do it that way next time.

I will listen to it once it is built. To build it I need to design it. To design it I can either look at datasheets, or just follow fashion - I don't fully trust my ears as I suspect that, like most people, I probably prefer a little 2nd HD to warm things up. There, I've admitted it - in audiophile terms I'm a heretic! Maybe it's because I'm a physicist.

Changing the subject slightly, I'm thinking of starting with Mullard CV4024 (654 code) as the input (low noise), Ei (telefunken clone) '83 as the splitter (good for large signals), and RFT EL34 (cheap but good NOS). Any comments?

[audiowize]

If you're worried about gain, why not emply a current source as a plate load on the first stage, that should help you out a bit. I've found an AT7 with a current source to have significantly more gain than a resistor-loaded AX7 in the same position.

[SY]

Originally posted by audiowize
If you're worried about gain, why not emply a current source as a plate load on the first stage, that should help you out a bit. I've found an AT7 with a current source to have significantly more gain than a resistor-loaded AX7 in the same position.

Have you tried this with the 12AT7 degenerated? Grounded cathode (LED bias or bypassed cathode resistor), I've seen the same thing- high gain, excellent linearity. But with an unbypassed cathode resistor, things don't look happy.

[audiowize]

I have done it both with a bypassed cathode resistor, and an LED, and an unbypassed cathode resistor. I tend to prefer the LED (HLMP-6000), although the unbypassed cathode resistor seems to hold up nicely also.