Categories
Uncategorized

The Williamson Amp, part one

The original Williamson hi-fi amplifier schematic as published in “Wireless World” magazine (UK) May 1947.  

The Williamson amplifier is considered one of the earliest hi-fidelity audio amplifier designs.  It is certainly one of the most popular audio circuits ever developed for DIY’rs.  Without fail I seem to turn up at least one home-brewed Williamson every year at the local estates+fleas.  A PS Dot Com reader from the UK sent us the original articles from “Wireless World” as published in 1947.   I have yet to build a pair of these myself, and the idea of starting ‘from the top,’ as it were, with the original design, is appealing.  A few things to note: check out the provision to balance the driver stage, and separate bias level and balance controls for the output stage.  Also: check out R25: the formula for determining the feedback loop resistor.  I wish every schematic included this notation.   NB: the ‘L63’ valve is simply a 6J5 – aka, one half of a 6SN7.  the ‘U52’ rectifier is a 5U4 or equivalent.  ‘KT66’ is a better-performing 6L6; feel free to use 6L6 or 5881 if necessary.

14 replies on “The Williamson Amp, part one”

The Williamson articles from the original magazine-reprinted many times on line and in print by Audio Amateur/OCSL-are well worth studying, but not with the intent of building an amp, except as a research exercise. Between the date of its publication in the immediate post-WW period and the end of “mainstream” tube audio gear manufacture, an enormous amount of progress was made.

The original transformer design is a good example. It was larger, heavier and much more complicated than better commercial designs, such as the American Peerless S-265-Q, and various Partridge and Parmeko units from its native Britain. Even in Britain most people bought their transformers after the end of the postwar “gray years”.

It runs a pair of 6L6 class power tubes into 10 K ohms, and gets perhaps fifteen to twenty watts. A more modern design would get a minimum of 35 and perhaps as much as 50 watts from the same tubes.

But the biggest flaw of the Williamson was its lack of unconditional stability. Because feedback is put across so many stages, including the completely unnecessary differential driver stage, it is not stable into certain reactive loads.

An enormous amount of commentary, positive and negative, has been written on the “Willy”, as it’s fondly called, and it is all worth studying. It’s not that D.T.N. Williamson was not a good designer: on the contrary, it’s precisely because the Williamson amp was the first serious effort at a modern high fidelity amplifier to be documented and put forth for public scrutiny, as a component on its own, and because it was a big advance on existing practice-particularly in Britain, which was (not to be parochial) behind the US in audio design-that it deserves such scrutiny. Amplifiers that are simpler, more stable, and put out lower distortion with more power are very much the norm today, but without the success of “The Willy” as well as its shortcomings they might never have been developed.

One of the problems with even discussing a “Williamson Amplifier” today is that because it was so famous and advanced, the term “Williamson” was applied to a lot of things that Williamson had nothing to do with and in some cases was on record as disapproving of.

In the United States, the Williamson came to the attention of the audio world through a series of articles (reproduced in the “Audio Anthologies” reprinted in turn by Audio Amateur/OCSL-Ed Dell’s operation in Peterborough) by Melvin Sprinkle and David Sarser. Sarser (who is still alive, IIRC, and occasionally posts to the Ampex tape machine mail list) and Sprinkle immediately set to improving it and serendipidiously found that the Peerless (in those days this was the real Peerless under Ercel Harrison, not some imitator with dumpster-salvaged blueprints) S-265-Q when strapped for the specified 10K:1 load offered a pair of taps at 50% that made fine screen taps for ultralinear operation just as the first Hafler and Keroes articles on that mode of operation were in print. Henceforth, it was not uncommon to call any amplifier with screen taps on the primary “Williamson style” when Williamson himself had no such thing.

A good Google search will yield a good deal of commentary on the Williamson that should be read and understood before deciding to build one today for actual listening purposes. However, I would start with the “Audio Anthology” articles, if at all available, before plunging into what can be a confusing mass of data.

“But the biggest flaw of the Williamson was its lack of unconditional stability. Because feedback is put across so many stages, including the completely unnecessary differential driver stage, it is not stable into certain reactive loads.”

Rubbish, Bafflegab. There is only one type of feedback applied in the Williamson amplifier (as can clearly be seen from the schematic) and that is overall feedback from the output transformer’s secondary back to the input stage.
Remember: the Williamson is in effect a simple three-stage design because V1 is directly coupled to V2, making V1 plus V2 a single stage.
Furthermore, Partridge established themselves as THE prime supplier of output transformers for those seeking to build the Williamson (Williamson himself never sold OPT’s to the consumer, he merely laid down the transformer specs his amp called for). And what an OPT the Partridge was! It was never equalled by any competitor, let alone surpassed.

“It runs a pair of 6L6 class power tubes into 10 K ohms, and gets perhaps fifteen to twenty watts. A more modern design would get a minimum of 35 and perhaps as much as 50 watts from the same tubes.”

“More modern”? Well, that would have been counter-productive to the whole idea of the Williamson, wouldn’t it? The “Willie”, after all, is a pure class-a design, that goes for maximum sound quality, not for max output power (and thus distortion). By running the KT66’s well below their max. plate dissipation one gains (beside superior sound) the added benefit of dramatically prolonged tube-life.

I think Heyboer makes a S-265-Q clone that’s pretty good if you want to build one of these.

Oh yes: If you build a Willy with US tubes, note that they are NOT quite the same: for best results you will want to tweak some resistor values. Their operating points differ more than a little. Also note that the booklet has two iterations of the circuit: this is the first. You do not want to build it this way, but follow the second, you don’t want the pot at R12. Also note the discussion on stability.

A “CRO” is an oscilloscope: you’ll need one. Also don’t back excite the transformer from the mains: use a low energy source, like an audio oscillator. Note the inductive kick from a good, sizable output transformer if shorted or opened the wrong way can spike back and kill a solid state source if it isn’t well protected. I blew up an Audio Precision test set at an employer once this way. (I paid a week’s salary to have it fixed, which beat unemployment, but the dirty looks continued.) Either get a classic tube HP 2ooCD (or a GR, WE 19C or Bruel & Kjaer BFO set) or use a cheap solid state unit you can replace the output transistor cheaply and easily on.

I built a pair of these using the 1949 updated circuit with 4 6J5 and KT66 outputs and 5V4 rectifier –finished them in 91 . You can jabber jaw about the circuit ….. just listen to it … it sounds like music. I used a common Hammond transformer from AES…3200 primary . –NO FEEDBACK , just shunted the input and raised the wattage on the Kt66 bias resistor and these have wowed everybody that has heard them . No stability issues …. play them all day long ,,still have the original tubes since construction . Don’t worry , be brave ….. its all about the music .

Yes, but then it isn’t a Williamson.

The feedback loop is the whole point-that’s why they went to the trouble of making it just so. Have you MEASURED this amp in this config? I bet you have 5% thd midband…That may well be what you want but you could have had it with a whole lot less trouble.

The entire point of the Williamson was to enable a lot of feedback and it was a poor attempt at that. It is not unconditionally stable and will oscillate into certain reactive loads. I learned this the hard way. The basic flaws are the use of the unnecessary driver stage after the splitter and the DC coupled input stage. Also, running a pair of 6L6 class tubes into 10K is not a great idea at all.

Patrick Turner has some useful information on his website if one wants to research it further.

Why is it bad to run 6L6s into 10k? If the power level obtained is okay with the designer and the tubes are run at reasonable voltages and currents it’s fine.

If you’re running the amp in consatnt overload you do need protection diodes or a spark gap though. And the transformer does need to be hypotted at 10x plate voltage for this service.

The distortion that tubes make is not the same type as transistor distortion … just look at the se amp figures. I have heard amps that measure perfect but sounds like cardboard . When you can hear the ring and harmonics of piano strings ,clear vocals , and acurate reproducion , these are remarkable amps . I heard an original years before I built these …. clear and acurate . Pilot used this design in their amps with 6sn 7’s . I use the original kt66 outputs and glass 6j5’s are better than the metal ones . The way Ihave it set up is 1%@1000 HZ . .. 20 WATTS ….Max is 10 % @50 watts .. . . Sounds great driving my AR-9’s with my SAE amp driving the 4-12″ woofers -bi amped. Cymbals sound like brass . I have heard a lot of amps over the years . I have had these in use since 1990 . Super reliable . It works for me . Nothing like 14 tubes heating the room on a cold winter night!

If you look at specs , you are not hearing the music . I have heard amps that test great on the bench and sound like cardboard. Specs give no indication of how an amp reproduces music and sounds accurate. When you hear all the harmonics of a piano , and cymbals sound like cymbals , vocalists sound like they are in the room with you -its awesome . You listen to your specs -I will listen to the music . I have the feed back loop connected and switched … better with it off …

The reason for Wiliamson VLF instability is that it has two RC couplings plus the OPT in the feedback loop (each gives increasing phase lead as the LF goes lower) The DC-coupled first stage actually helps since another RC coupling would be fatal. A very large primary inductance in the OPT delays phase lead effects, too (Williamson specified at least 100H – expensive!) I have refurbished two of these amplifiers and built one from scratch – all successfully stabilized at VLF – one by using a “VLF gain shelf” and the others by careful choice of RC coupling time constants; don’t make them the same.

I wrote a book about the true History of D.T.N. Williamson.
The book has been published on March 2013, pag. 460, SANDIT Libri Editore, ITALY.
I am a technical contributor of italian magazine SUONO and I take a column on vintage equipment from 1999. I also published a LEAK AUDIO HI-FI book, may 2014.
Regards,
P.Paolo Ferrari

Just finished building our “Prototype Blue(tooth) Tube Amp”.
It is largely inspired by the Williamson. In the front end using a pair of 6SN7. The muscle end is rather more like a Dynaco running 6L6GC twins. With a somewhat more modern power supply (way bigger caps) of my own concoction. Gave it a separate independent bias supply to keep the grids negative and adjustable over a wide range. Also this frees up ~50 volts of lost plate to create a puled up bias. Used a 5K to 8 transformer and a 660v centrer taped yielding ~ 470 loaded plate volts. Biased at 40 ma/ L6 it yealds 37 watts no sine of a clip. Slipped the tone control in the feedback loop, Handed him a blue tooth receiver. I sit here listening to my I tunes wondering why I did not do this a decade ago.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.