I've been doing a little research, as one does.
I agree, this project does not follow the true Williamson 5-20 topology. I bought it as such.
Never-the-less, this implementation of an EL34 P-P amplifier is really easy to build and does sound extremely good.
I thought about KT88s but the impression that I get is that they are brighter in sound than the EL34.
I'm using EL34 IIs. The B&W CM8s2s are bright anyway so the KT88s may not suit them and my English ears.
I agree, this project does not follow the true Williamson 5-20 topology. I bought it as such.
Never-the-less, this implementation of an EL34 P-P amplifier is really easy to build and does sound extremely good.
I thought about KT88s but the impression that I get is that they are brighter in sound than the EL34.
I'm using EL34 IIs. The B&W CM8s2s are bright anyway so the KT88s may not suit them and my English ears.
I'm glad you got that up and running to your satisfaction. With the exception of using EL34's, that is a classic Williamson design. (There's no "5-20" there at all, it's just been misnamed by the seller.) The actual 5-20 design is simpler, with fewer frequency poles, which allows for greater stability and the use of more feedback if desired. Generally speaking, the Mullard 5-20 design does transients better, but the Williamson has more tonal weight and sense of reality, IMO. I've always preferred it.
I think it depends on how much feedback you have as to whether there will much, if any audible difference between output valves. My 5-20 has 20dB of feedback which is 10dB less than a bone stock mullard, and that has pretty much made output valves indistinguishable from each other. I've tried EL34B, E34L, 6CA7 and KT77. The latter are both beam tetrodes vs the former pair being true pentodes.
Not quite sure it is the best, but of a family it is quite fine. Type 4E27 finals on a 600V B+. 6H6Pi/FQP1N60 cascode rigged as a balanced input amplifier. E-Linear connected to 20% taps on custom S-265-Q output Iron. These got done with 20, 30 and 40% taps, and the 4E27's screens share the 20% taps currently. Costs a wee bit of power to run them from the 40%.
No global NFB, just the Schade-style, from the plate tap.
Subsequent build using 6CB5A pentodes as finals, and EL84 voltage amps around S-258-Q opt Iron did quite well also, and do did a big brother amp running S-271-S outputs with 26HU5 finals and 25HX5 voltage amps. The HU5 amp did not take well to using the 50% E-Linear taps, so it runs the standard plate-to-grid, Schade with 100k and 8.2k on the 25HX5 plate loads.
Sooner or later I plan to try using power triodes for finals, but for the time being, Schaded pentodes keep me satisfied.
cheers,
Douglas
I looked at the schematic for this amp in the EL-34 Williamson partner thread to this. I like hearing listening assessments to amps as that's the true test of quality. The big difference to the topology of this amp you've built and the original Williamson is that there is no voltage amplification stage after phase splitting. That's a really good thing. The phase splitter goes directly to the final power amplification stage. That's a big improvement over the Williamson.
You are never going to get two separate parallel amplification stages for each inverse phase for push pull to track each other identically. The fact that this amp eliminates one of those post phase splitting voltage amplification stage can only be a good thing. I would be willing to bet that this change from a Williamson would have a good impact on the sound quality even if measurements do not indicate anything. Just stands to reason.
You are never going to get two separate parallel amplification stages for each inverse phase for push pull to track each other identically. The fact that this amp eliminates one of those post phase splitting voltage amplification stage can only be a good thing. I would be willing to bet that this change from a Williamson would have a good impact on the sound quality even if measurements do not indicate anything. Just stands to reason.
Last edited:
Unfortunately not true.
There are no 2 people with the same exact hearing response curve. And most people can't tell 1% thd from 0.1% thd to save their life. And the vast majority of people are not trained listeners. So YOUR listening assessment is the best test for YOU. Nothing more.
Building a Williamson EL34 amplifier
see item 38 on this page for the OP's schematic. It can be seen that this a very basic change from the topology of a standard Williamson. The second voltage amplification stage in a standard Williamson is right after the Concertina (aka Cathodyne). In the OP's amplifier schematic that I just referenced the second voltage amplification stage is moved in front of the Concertina.
The reason that is important is that, everything being equal (which it never is), if you have two halves of a single phase cycle that are amplified differently it is likely to be detectable in your hearing when they are combined in the last stage. If instead you have two amplification stages sequentually and they don't exactly track each other exactly it is going to create a much more benign type of distortion that you probably will not hear. Unless you have super hearing. Again, everything else being equal.
see item 38 on this page for the OP's schematic. It can be seen that this a very basic change from the topology of a standard Williamson. The second voltage amplification stage in a standard Williamson is right after the Concertina (aka Cathodyne). In the OP's amplifier schematic that I just referenced the second voltage amplification stage is moved in front of the Concertina.
The reason that is important is that, everything being equal (which it never is), if you have two halves of a single phase cycle that are amplified differently it is likely to be detectable in your hearing when they are combined in the last stage. If instead you have two amplification stages sequentually and they don't exactly track each other exactly it is going to create a much more benign type of distortion that you probably will not hear. Unless you have super hearing. Again, everything else being equal.
No it doesn't stand to reason, just the opposite.
The AC balance of an additional amplification stage is no different than the AC balance of a splitter stage. They can both be adjusted very easily with a balance pot.
I just looked at the schematic again. Mea Culpa on my behalf. The second voltage gain stage is after the Concertina, just like the Williamson. I think what happened in my interpretation is that I've always thought the topology of the Williamson had the basic flaw I've been talking about. So when I saw this copy I saw what I wanted to see.
I hope someone will eventually try the Williamson with the second voltage gain stage in front of the Concertina. Better yet eliminate the second one entirely and boost the voltage gain of the first. But keep the Concertina. That probably provides the basic magic of the Williamson.
I hope someone will eventually try the Williamson with the second voltage gain stage in front of the Concertina. Better yet eliminate the second one entirely and boost the voltage gain of the first. But keep the Concertina. That probably provides the basic magic of the Williamson.
Yeah, me too. The best sounding amp I came up with was a 2-stage affair with a 5687 LTP (with center-tapped plate choke) DC coupled to PP 2A3s. The problem was that it took 3.5V peak input voltage to get to full power (6W rms into 8 ohms). It required a line stage with gain, but that was a whole other story...
The problem with a 2-stage LTP to push-pull outputs is that it's not easy for a single LTP to drive tubes like PP 6L6GC, 6550, 6B4G or 2A3 that need 40V peak or more signal at their grids to full power with 1Vrms from your sources.
Also, the LTP will need a negative voltage supply to allow a CCS or other high impedance load in the joined cathodes, whereas the cathodyne has as good balance as you can get driving the benign load of a diff-pair driver stage, without the need of a separate power supply.
The problem I see with a cathodyne phase splitter driving the output tubes is that if the output tubes go into cutoff the cathodyne will no longer see equal load impedances, and its balance will go to pieces. It also would require that a lot of gain comes from the input/voltage amp stage, which suggests high distortion. (Perhaps a 12AX7 into a 6N6P cathodyne would work well?)
Perhaps a cascode LTP with source followers before the output tube grids... But is that then a 3-stage topology?
I think the Williamson topology is a good one for output tubes that require a lot of input voltage at their grids, like PP triode-wired GU-50, or PP 300B. I was thinking one could reduce the driver stages' gain by using local NFB from the output tube plates to the diff-driver cathodes, and perhaps plate-grid feedback around the first stage, instead of a global NFB loop. Keep the more complex reactances of the OPT out of the feedback loop. That should increase stability.
Of course, there is the 'simpler is better' school.
--
Last edited:
