Heathkit W4-AM modifications

[magnetmaz]

Hey, all.

I've been in conversation with a member of AK ( and DIYaudio), Dcgillespie, because the two of us have the same primary listening rig: Heathkit W4-AM's into Klipsch Cornwall 1's. He revealed that he had done some modifications to his W4's that seemed quite impressive to me that i'm going to try for myself:

1) A modification that fixed the well-documented instability these amps have ( ringing ) by modification of the feedback loop and a few other small changes. This mod is not unlike the Bricktop mod that most people with Heathkit W4's are familiar with. His comments on it are as follows:


Bricktop's modification is certainly good, although he fundamentally changes the topology of the feedback system. His approach is to apply some local feedback around the output tubes, and reduce the global feedback around the amplifier to improve stability that way. It certainly works, and I am hardly trying to imply it is not a valid approach.

The problem I have with it, is that the OPT is what produces the most distortion in the amplifier at the frequency extremes, and reducing the feedback around that component only serves to reduce the ultimate performance that can be achieved. It also reduces speaker damping as well.

My approach leaves basically the original global feedback factor in place that Williamson specified, but applies it in such a way the the amplifier is very stable, and does not need the hang-on network to load the amplifier, when no actual load is in place. Such networks are notorious for reducing power and performance in the upper frequencies of most typical designs. Note if you will the will the final version specifications of the W4-AM, versus those of the original W-4. Granted, the final version was stable, but at a significant cost to performance. With the feedback and stability networks I devised, it achieved the performance of the original version, with the stability of the later edition.


2) An EL-34 in Triode mode conversion to the W4, allowing the unit to run in Class A. This schematic is forthcoming; I will add it as i get the notes and redraw it. (note that this modification prohibits use of the heathkit preamp that bundled with this amp, because the extra current capabilities of the PT are used on the EL-34's instead of the preamp tubes)


So, here is a link to the feedback loop-modified Heathkit W4-AM schematic. I've included all of the notes for what to modify in the lower right corner of the drawing, and all changes to the original schematic are done in red. Enjoy.

 

[audiodon]

This is a great gift. Thanks.

 

[magnetmaz]

it is pretty excellent. I had nothing to do with it other than recognizing that Dave had some awesome stuff in his head, and needing to get it out to the community.

also, sweet MS Paint skillz.

 

[vinylkid58]

Thanks for posting.:thmbsp:

jeff

 

[DanO55]

Looks like I have a couple of changes to make to my pair. Thanks Guys, esp. Dave

 

[GordonW]

That looks almost just like the same type feedback loop compensation, and the same shelf-filtering, that's done on the Eico HF22/HF35.

From Dave- given his affinity for the HF22/35, that's not terribly surprising. It's a good adaptation of what worked there, to the W4.

In actuality, it might even work BETTER on the W4 than on the HF22, since the W4 has a triode first stage (instead of pentode, as the HF22/35 has). The lower impedance of the triode seems to work better with the RC compensation (and accompanying loading of the front end tube) than a pentode...

This does lead to a thought, though- if we were to appropriately increase the open-loop gain of the amp (by increasing the gain of the front end tube- by changing resistors and such, or by maybe replacing the 6SN7 front end tube with a 6SL7 or such, and adjusting resistor/cap values accordingly)- could we implement BOTH Dave's and Bricktop's approaches at the SAME TIME? In essence, use Dave's global feedback (full magnitude), AND Bricktop's local feedback. This could wind up with an arrangement, with similar final gain (with feedback) to the original design, but with better performance than either type feedback alone...

Regards,
Gordon.

 

[dcgillespie]

Gordon -- Excellent observations/comments as always. It is in fact the same shelf style network that Eico (sometimes) used, but is not used here out of any want to make things more complicated, but due to the particular winding geometry that Chicago and others used in some series of their UL transformers.

A marrying of this approach and bricktop's could indeed produce ultimate performance capabilities, because the big advantage of bricktop's approach is that it lowers the driving impedance to the OPT, allowing maximum performance from it, while the global network (which would include the transformer) would really drive home the aspects of lowering output impedance, distortion, and flattening overall response.

However, there are limits to how far this approach can all be taken -- and not necessarily because of stability issues. In fact, the use of multiple loops actually works to enhance stability. However, there is still an optimum combination within which these two types of networks operate best, such that the idea of if some is good then more the better, is clearly not the case here.

There were a few amplifiers back in the day that boasted the use of nearly 40 db of NFB, achieved by a combination of local and global loops similar the that being discussed here. The problem is, 40 db of NFB implies a reduction in distortion by some 100 times -- except that the designs never achieved anywhere near that kind of performance, most often achieving little better than that of what global NFB alone could produce, all else being equal.

So, while the approach of multiple loops is certainly valid, as always with feedback, it must be carefully applied to achieve maximum results. There are way too many examples where feedback in general was poorly applied (hence this thread regarding the W4), so it's little wonder that NFB got a bad stigma attached to it. In deed, I would much rather have an amplifier with no feedback at all, than one where it was poorly applied. It is so true that feedback makes a good amplifier great, but can do little for an amplifier that is poor to begin with. I've always said it is much like seasoning on food: With none at all, things can taste pretty bland; use too much or the wrong type, and you ruin it. But with the right type in the right amount, it is simply delicious on a level that cannot be achieved otherwise. That pretty much sums up the capabilities of NFB as well.

Dave

 

[DanO55]

Note to Self: Self, hold off making any changes to your pair of W-4s until additional comments are posted on refinements to the original circuitry for increasing overall performance.

Note to Others: The W4 is one of those amps that has always flown under the radar. It is a very fine amplifier that can be purchased for below the average cost of comparable tube equipment. Film at eleven.

 

[GordonW]

I could see something in the middle ground- the normal global feedback (I haven't done the numbers on the network here- but with most Williamsons, it's between 16-20dB), with maybe at most, say 6-10dB of local feedback.

No, I wouldn't be inclined to be as "feedback happy" as Hegeman was on the HK Citation II... but if it were possible to nest 6dB inside the loop, and still have enough overall gain, with the global feedback as-written above... that might be very interesting, as far as (especially LF) performance is concerned...

Regards,
Gordon.

 

[dcgillespie]

Agreed. In my experience, about 6-8 db applied to the local loops does a sufficient job of minimizing the phase shift and extending the response of the stages within the local loop appropriately, which then really enhances the effectiveness of the global loop. Hegeman really was a feedback happy guy, wasn't he?

Dave

 

[GordonW]

Hegeman really was a feedback happy guy, wasn't he?

Dave

Yeah, but the proof is in the pudding, as it is said. The Citation II is one of the only "vintage" amps that truly had close to "ideal" specs, and performed as well as its specs indicated.

Sure, it's a complicated design- but it was pretty much head-and-shoulders above everything else of the time... and it still holds up today, IMHO.

And, the other Hegeman designs, including the Citation V, the Award series and such- have all been pretty stellar performers, too. Many of those also had some interesting features, for the purpose of allowing him to use more feedback than would normally be possible- and the proof of concept is in the great sound they all produce, IMHO...

My take here, is that by careful application of the same principles and some of the practices Hegeman advocated, we can get many of the same benefits, within the Heathkit W4/Williamson design, that he got from his own designs. That's always a good thing, IMHO.

Regards,
Gordon.

 

[dcgillespie]

Couldn't agree more!

Dave

 

[magnetmaz]

Here's the final version of the Heathkit W4-AM schematic. Dave remembered a couple more mods he did to the stock configuration, including adjustment of the coupling caps and a few other resistor changes / additions.

I've included everything on the original schematic, with any changes made to adjust stability at both high and low frequencies done in RED, and some basic additional changes to extend tube life and reduce distortion further in BLUE.

I also have all the instructions from Dave on how to convert your W4's to EL34 triode mode / class A operation, and will be doing these up in an alternate schematic this weekend.

So...final version of the 5881 / super-stock configuration is attached.

 

[Dukiedook]

Why would I want to change from the buttery smoothness of the 6L6 to an EL34?

 

[magnetmaz]

high cholesterol?

 

[GordonW]

Why would I want to change from the buttery smoothness of the 6L6 to an EL34?

I'm kind of doing the opposite.

I'm reworking a pair of Dynaco MkII/MkIII chassis/board assemblies, to work with 6L6GCs. 6600 ohm UL Hammond output transformers, and replacing power transformers to give proper B+ with the 6L6s (about 440v, with solid-state choke-filtered supply)...

So, I guess I kind of agree with the above sentiment. Why use another tube, when you could use a 6L6?

Regards,
Gordon.

 

[magnetmaz]

perhaps you've got a pile of Marshall Plexi's and you want to run a set of them as hifi mono blocks.


Seriously, though... i just looked through Dave's notes, and here at the bottom he gives the final specs. Perhaps this description will provide some illumination.

This modification operates with 19 db of negative feedback, and is a very stable design. It is also a real class A triode amplifier now, remaining in class A operation throughout its entire power output capability range of about 12 watts RMS. Because of the use of a push-pull triode output stage, class A operation, and using the OPT at a slightly reduced power level relative to its design capability, the distortion characteristics of this unit are unbelievably low. At 12 watts output, THD at any frequency from 20 Hz to 20 kHz is less than .50%, while IMD is no more than .15%. The frequency response is flat within +0/-.5 db to 50 kHz. Of course, these numbers were obtained with reasonably matched output tubes and a driver tube with reasonably matched internal sections.

With this kind of frequency response, power bandwidth, power output, distortion, and stability, this amplifier as modified has very formidable performance with speakers of appropriate sensitivity. They are perfect companions to my Cornwalls.

The modified amplifier operates the output tubes at 60 ma each quiescent cathode current. With the new GZ34 rectifier tube, this places about 400 vdc across the plates and cathodes of the output tubes, resulting in a plate dissipation of ~ 22 watts per tube, and a screen dissipation of ~ 2 watts per tube. At these levels, the output tubes will have very long lives in deed.

So modified, this amplifier now embraces all of the high performance qualities originally sought by Williamson in his amplifier, but also has impeccable manners in the process of delivering them to virtually any type of load. In this regard, the original design was left wanting, such that the differences now achieved are immediately obvious. If this type of design is of interest to you, I have no doubt of the satisfaction it will provide.

 

[dcgillespie]

With these modifications -- as with all designs to begin with -- it is all about compromises, and principally, achieving the best mix of them.

The basic topology of the W4 circuit is wonderful, and in its original form, is capable of providing about 18 watts of superb performance. Its only drawback is the poor stability produced by the original design. The modifications offered for that design correct those issues.

For the modification using EL34s, the decision to use that tube should be kept in context, as the goal was not driven by the sound of that tube, but by the sound that a different topology would produce.

The original W4 employs a class AB1 UL 5881 output stage, where as the goals that ended up becoming the EL34 modification were to convert the unit to a class A triode output stage, in keeping with the original Williamson concept.

The original version of Williamson's design employed KT66 output tubes, which produce 10 watts RMS power output in his design. However, at the time of America's introduction to the Williamson, the KT66 was not available. The closest tube by characteristics that could handle the voltage was the 807, so all of the earliest American versions of the amplifier used that tube.

When the 5881 come along, the ability to have the same characteristics as the 807 bottled up in a convenient octal base format with no top cap, and voltage ratings that were acceptable for the Williamson design, meant that all such copies of the amplifier quickly converted over to the 5881. However, compared to the KT66, both of these tubes resulted in a power loss, as they could only realistically provide 6-7 watts RMS of power output in Williamson's design. By the time the KT66 did become available in America, UL had already made its big splash, and the triode output stage was quickly abandoned, whether the KT66 was available, or not.

But returning to Williamson's design, the only practical tube he had available for his design at the time was the KT66. Therefore, even though it is a wonderful tube in its own right, it became his tube of choice principally by default. The design was published, and the rest is history.

One of the drawbacks of his design -- as history shows -- is that it requires a notably greater input signal to achieve full power output. This is typically over 2 volts, compared to the 1+ volt figure that was quickly becoming the norm of the day with faithful conversions of his design to UL operation, and other designs as well. In any event, the 2+ volt driving requirement coupled with only 6-7 watts of usable power here in America, meant that the stage was set (no pun intended) for UL to quickly take off, and take off it did.

In my efforts then to produce a really good version of Williamson's efforts -- using his topology and design goals -- then the EL34 -- unavailable at the time of Williamson's design -- makes for an ideal tube. Consider that:

1. It can easily handle the voltages of the Williamson design.

2. All else being equal, it can produce more power output than the KT66 can in the Williamson design and still run well within rated dissipation levels.

3. Due to its higher Gm rating, it can be used in the design and produce a very normal 1.0 vac drive requirement, while still employing the same amount of NFB specified by Williamson.

Therefore, use of the EL34 became an obvious choice, as it overcame the original American problem of low power output, and even improved on Williamson's original design specification for sensitivity.

The only other matter requiring attention was the famous high and low frequency instability problems associated with the design, which I have also addressed with the EL34 modification.

So the effort of converting the W4 to use EL34s was not so much an effort to convert the W4 simply to use another tube, but because of the excellent and perfectly suited transformer set offered, convert the unit to represent the finest example of the heritage it grew out of, with appropriate fixes for its blemishes attended to along the way. Using the EL34 then was just one of the considerations made to achieve that goal.

I hope that helps!

Dave

 

[Dukiedook]

I just scored a pair of these puppies and will be digging into them soon, from what I can gleam it seems the output transsformer was one of the weak links in the product. Is there another new production output transformer we could plop in place to remedy the weak one currently there? Do you think you would have to do these mods even if you do put in a better output transformer?

 

[dcgillespie]

Dukie -- You would be hard pressed to find modern OPTs that have the performance capabilities of the original output transformers used on these amplifiers. The output transformer was not really a weak link at all, but the efforts to stabilize the circuit they were installed in surely was.

Any vacuum tube amplifier employing global feedback from the output transformer secondary winding requires some means of controlling circuit response so that it will remain stable into various types of loads. While many transformers can use a similar approach to controlling response, no two transformer types will require the same component values or networks to achieve similar stability levels.

Ultimately, with any feedback design it is always best to use the very best OPTs possible, and then apply whatever networks are needed to achieve stable operation with that transformer. Virtually any transformer you might install in the W4 circuit will require some type of network(s) be added to control response, so use a transformer's performance capabilities as your criteria to judge it, and not the relative complexity of the networks required to stabilize it.

Dave