Williamson on EL84. Is it worth doing?

[bae59]

Hi all
I'm revising my current PP on 84 output tubes and one more envelope for two stages being preampf and phase splitter.
I read about many advantages of the Williamson setup having additional driver stage between phase splitter and output tubes. Of couse that doesn't cover all the advantages persistent to the circuit!
So I wonder if any one has done the WILLIAMSON set up with less powerfull output tubes as compared to 6L6 et al tubes in the reference setup .
Specifically, my question is which stages of the original set up are worth building and even more, is it still good for the sound additional interstage between splitter and output with the EL84 used?
I am quite up to date about the modifications already introduced since 1947. So I don't discuss them now considering them being vital but not covering my current problem of upgrading before I start drilling and welding.
So, any considerations or experience will be welcome and thanks in advance.
p.s. In fact, there is no on the Internet the Williamson setup using EL84 at the output, though!

[ray_moth]

The Williamson circuit is unnecessarily complex for OP tubes with such modest drive requirements as the EL84. Williamson found it necessary to interpose an additional voltage amplification stage between the concertina splitter and the OP stage because the limitations of the splitter rendered it unsuitable for driving the triode-connected KT66s. These limitations are not a problem if you use EL84s, so you can dispense with the differential driver.

Furthermore, the Williamson design with NFB over multiple stages, including 2 sets of decoupling caps, is on the borderline of instability. Although it can give excellent results with very low distortion and certainly has its following, IMHO it is asking for touble for a relative beginner to build one, especially when a much simpler solution is easily achievable.

[richwalters]

Originally Posted by ray_moth

Furthermore, the Williamson design with NFB over multiple stages, including 2 sets of decoupling caps, is on the borderline of instability. ..........

Hey come on...you can do better than that... I've used the same double interstage technique for decades in 7-20-50-150-250W UL power amps and get nearly the same stability results: Design for 20dB global feedback or even less, and with careful 1st stage Zobel CR and interstage cap values, I can increase global feedback onset 35dB tethering both HF & LF instability. In clear text, thats another 15dB margin on top of the 20dB. Who can complain about this ?
It's all about knowing the circuit and the math behind it and better still
I get none of the problems with long speaker leads.

In comparison, my older Citation 2 hates long speaker leads. For short fine, but isn't always practical. The worst offenders to test topend stability, the horrid ceramic piezo tweeter.

I always maintain when starting learning with tube amps keep to a design that has stood the test ot time.

richy

[Shoog]

Originally Posted by richwalters

Hey come on...you can do better than that... I've used the same double interstage technique for decades in 7-20-50-150-250W UL power amps and get nearly the same stability results: Design for 20dB global feedback or even less, and with careful 1st stage Zobel CR and interstage cap values, I can increase global feedback onset 35dB tethering both HF & LF instability. In clear text, thats another 15dB margin on top of the 20dB. Who can complain about this ?
It's all about knowing the circuit and the math behind it and better still
I get none of the problems with long speaker leads.

In comparison, my older Citation 2 hates long speaker leads. For short fine, but isn't always practical. The worst offenders to test topend stability, the horrid ceramic piezo tweeter.

I always maintain when starting learning with tube amps keep to a design that has stood the test ot time.

richy

But there are so many good simple EL84 designs out there. The Williamson requires the best quality Iron - which is not cheap and not easy to get hold of these days.

Shoog

[bae59]

So, I see there are no requirements to use interstage driver envelope in the setup as far as the EL84s are used?
Who has objections to that or different considerations-experience from building the Williamson cct?

Alexander

[bae59]

Originally Posted by richwalters

Zobel CR

What is it? I'd like to know theoretical issues for that before I proceed further.

I've found some tips on that in the neighboring threads but still feel smwhat embarrased thinking I may be don't understand that right.
Thanks to all.

Alexander

[ray_moth]

Ian, a Zobel network is a cap in series with a resistor. It can be used in parallel with the plate resistor of the first stage to prevent a high-frequency peak in the response, causing instability in a NFB amplifier. It needs to be sized to suit the OP transformer characteristics and stray capacitances of your particular amp, using a trial-and-error approach. It could typically be in the order of 1k - 10k in series with 47pF - 220pF.

Rich, I don't doubt that you achieve great results with a Williamson design but I still think that it's far from ideal for a beginner and it's unnecessarily complex, anyway, for EL84s.

[bae59]

Now I got it! Thanks a lot!
I'm a bit far from a beginner, so I would agree with you that the Williamson set up with its intermediate separate driver stage can be regarded unnecessary for EL84s.
But still I would be much pleased to see Mr.J.Brockies' conclusions on this matter as I see this "valve gury" has thoroughly redesigned the cct but he would not reply on my private e-mail as by now. As there is still one issue that remains e.g. matching phase-splitter (split-load setup) with the output tubes that still somewhat embarrassing me if I still go in for the Williamson way.
So, I'd rather and instead of the original setup build the LTP phase spritter with current sink on seperate pentode as Mr. suggests. I'm much impressed by the figures for the THD et al of this cct. Possibly, as far as THD is less than 0.1 % I wouldn't utilise the NFB at all. Even more to this point I don't see the way I'm supposed to do that. Can one give a piece of advice as per how to introduce NFB in the cct in point?

If the gain is unsufficient, I would use pre-amp from the same source "angelfire" (am I quite right in quoting so?) on double triode envelope, where NFB is not needed at all as THD is negligibly small. I use PC card so the output is quite high in level. But still I like the idea to build the so much low THD cct. Of course all the stated is governed by the parts that are available in my stock now.
So, thank you for your time and my kind regards to all!

[richwalters]

Originally Posted by ray_moth

Ian, a Zobel network is a cap in series with a resistor. It can be used in parallel with the plate resistor of the first stage to prevent a high-frequency peak in the response, causing instability in a NFB amplifier. It needs to be sized to suit the OP transformer characteristics and stray capacitances of your particular amp, using a trial-and-error approach. It could typically be in the order of 1k - 10k in series with 47pF - 220pF.

Rich, I don't doubt that you achieve great results with a Williamson design but I still think that it's far from ideal for a beginner and it's unnecessarily complex, anyway, for EL84s.

Ray, your'e quite right. It's too easy to get carried away with success. On a skilled note;
I find the 1st stage Zobel values is critical in either triode or pentode stage.Too high C results in higher upper end thd and some listening tests confirm loss of mid upper range definition. My way is inject low amplitude 2Khz square wave through the amp and adjust the Zobel cap on the global nfb loop in keeping a min cap value in the 1st stage.., higher C values on the output global nfb keep upper end thd down. Throw in the stability dice, reduce the global nfb value and find out how much dB reduction til circuit oscillates.
Each of us have our own procedures of doing this and not particulary pleasant for amp builder to grapple which values to trim for min circuit ring and overshoot.. The Zobel in the UL Anode-Screen is another evil with high gm tubes. The interaction between one adjustment and another is enormous.
The Bode stability map cannot take into account, distortion. This must be ascertained in design and often conflicts with stability. Nothing is easy, alas it shouldn't be !
There is a very strong case for not so experienced amp builders to use low gm tubes for starters, 6L6,807 group where there is less chance to self oscillate.

Morgan Jones doesn't really delve deeply into stability issues. One can see why .

richy

[SY]

Originally Posted by ray_moth

Williamson found it necessary to interpose an additional voltage amplification stage between the concertina splitter and the OP stage because the limitations of the splitter rendered it unsuitable for driving the triode-connected KT66s. These limitations are not a problem if you use EL84s, so you can dispense with the differential driver.

Furthermore, the Williamson design with NFB over multiple stages, including 2 sets of decoupling caps, is on the borderline of instability.

While I absolutely agree with your conclusion (EL84 can use a much simpler input stage), I have some disagreement with the particulars.

The instability on the low end wasn't because there were two sets of decoupling caps but because the time constants of the coupling weren't staggered. Put a 10:1 ratio of time constants and the LF stability problems go away.

A concertina is certainly capable of driving KT66 in triode. Now, interposing the driver is a good idea because it eliminates any possible imbalance when transitioning from class A to AB, but that imbalance isn't very severe. With more modern tube choices (e.g., ECC88) as concertina, the imbalance drops from transient-and-minor to absolutely negligible.