Heathkit's W-3AM: A New Beginning:

[dcgillespie]

INTRODUCTION

And so we come to the last of Heath’s early Williamson offerings, the W-3AM. For reference, the previous two entries into this Heath saga can be found here:

http://audiokarma.org/forums/index.php?threads/regilding-the-gilded-lily-heaths-w-2m.767851/

http://audiokarma.org/forums/index.php?threads/from-the-frying-pan-into-the-fire-heaths-w-3m.769029/

As the title of this thread suggests, the W-3AM was in fact a new beginning for Heath’s high fidelity line of amplifiers. After the disastrous W-3M, Heath burned the midnight oil and wisely adopted measures of improved stability previously presented by others, that they had otherwise ignored up until that time. They went even further and introduced an additional new stability measure of their own to help ensure that the problems of the previous model were completely annihilated: Dubbed by Heath’s marketing department in succeeding models as the “Tweeter Saver”, the W-3AM was the first of their models to employ this feature (a Zobel network across the OPT secondary winding) to help ensure HF stability. Of course, the W-3AM sports the same genuine Acrosound UltraLinear transformer and basic circuit design as that of the W-3M. But the new model includes a number of measures designed to contain all the ill manners of the previous offering. They include:

1. Adding an input grid stopper resistor.
2. Adding a step network across the plate load resistor of the AF amplifier stage.
3. Reducing the value of the driver stage input coupling caps by a factor of 5.
4. Adding the aforementioned Zobel network across the 16Ω winding.
5. Reducing the NFB level employed.

All of these elements play a role in taming the untenable low and high frequency stability characteristics of the W-3M. It is from their application then, that the W-3AM was born, and the improvements they make are readily apparent, too. Consider the following:

1. The new design is stable with any level of capacitance added to any properly loaded tap.
2. In an unloaded condition, the new design can handle as much as .22 uF across the 16Ω tap, .15 uF across the 8Ω tap, and an unlimited amount of capacitance across the 4Ω tap.
3. new design demonstrates fast settling under pulsed conditions.
4. The new design is also stable enough at low frequencies when unloaded, although bounce increases, and settling slows significantly under pulsed conditions when unloaded.

This stability level is simply lightyears ahead of its predecessor’s, making for an amplifier that you don’t have to worry is going reach critical mass with the first little transient (musical or otherwise) that comes along. Of course, other things changed as well:

1. Frequency response is no longer ruler flat out to 100 kHz, but now follows the published curve quite closely, with a measured results of -.10 db @ 20kHz, -1 db @ 53kHz, -1.2 db @ 80kHz, -3 db @ 91kHz, and -5.0 db at 98kHz, as measured at the 16Ω tap. This will be discussed in greater detail in a little bit.

2. With the reduced NFB, full power total harmonic distortion takes a hit, but still turns in respectable performance measuring as: .76% THD at 20Hz at 24.01 watts RMS, .46% @ 1kHz @ 25.38 watts, and 1.60% @ 20kHz @ 23.77 watts. Maximum power output levels are unchanged from that of the W-3M.

3. IM Distortion also increases (basically doubling) to 1.25%

4. The reduced NFB significantly increases sensitivity as well, now requiring just under 1 volt (0.96 vac) to produce full power output.

5. The NFB level itself, while stated as being 18 db, actually measures only 16.6 db. When this is compared to the measured 21.5 db of the W-3M, it can be seen that the reduction is notable (cutting it almost in half), making it a significant element of Heath’s effort to improve overall stability. It is a basic principle that as NFB is increased, performance is also increased. But increasing amounts of feedback bring with it the potential for increased instability as well. This is particularly true with global NFB designs like the Williamson that include the OPT within the NFB loop. Therefore, the amount of feedback used plays a significant roll in achieving good stability in a global NFB design. As both versions of the W-3 series clearly show then, when feedback is used in appropriate amounts, it can be a good thing. But when not, the results can be stunningly horrible, which helps to fuel the arguments for or against its use that rage on to this day.

So with the enhancements afforded the new model, how did Heath do in their makeover of the problem child W-3M? Overall, the W-3AM is a huge improvement over its predecessor, making for a very capable amplifier, whereas the previous version was anything but. For those who currently enjoy the performance of their stock W-3AM then, they can do so without the worry or concern for the amplifier itself or the speakers connected to it that the W-3M presented. Improved as the new design is however, there is still opportunity to improve performance further yet for those wanting the best from the W-3 series. Some of the improvement comes from implementing the LF modifications developed for the W-2M. But the rest of it comes from unlocking all the performance capabilities of the most prominent feature of the W-3 series ……..

End Pt. I

 

[dcgillespie]

THE ACROSOUND TO-300 OUTPUT TRANSFORMER

These fabled transformers truly represent a double edge sword, comparable in many ways to that of any thoroughbred: Superior performance in so many ways — but with collateral issues to go along with it. Specifically, the TO-300 has superior performance across the audio spectrum, but with supersonic characteristics that are somewhat challenging.

With the Peerless transformer of the W-2M, producing a stable HF transient response under feedback conditions can be achieved at any offered output impedance with the use of a simple single tap FB network. This is possible because all or the majority of the secondary windings are always used in all output impedance configurations. That means that the supersonic characteristics of the transformer remain constant for all impedance configurations as well. Therefore, as the output impedance is changed, all that is required to maintain consistent amplifier performance is to adjust the NFB resistor and phase advance cap accordingly.

On the other hand, Acrosound transformers have a single separate lead to represent each impedance tap, meaning that at worst case (i.e., 4Ω), only half of the secondary winding is actually used. The rest of the winding is just taking up space in the coil form. While easy connection impedance taps were a relatively new thing at the time, the quality of the Acrosound transformer is such that full rated power output from the amplifier is easily available at any frequency from 20 Hz to 20 kHz — at low distortion, and at any tap — making for a ruler flat high quality power response curve across the full audio spectrum. That was heady stuff for 1951, representing the side of the sword that most others could only hope to dream about at the time. From this perspective then, as well as its simple impedance tap lead configuration, the Acrosound transformer of the W-3 series was seen as an improvement over the Peerless transformer of the previous model: Gone was having to re-strap the secondary output lugs to change the output impedance, and gone was the upper midband dip in power output, too. But then there’s the other side of the sword..…..

Unfortunately, this wonderful audio band performance comes with rather unique supersonic characteristics. Specifically, these transformers have dual points of resonance (causing double peaks in supersonic response), with supersonic phase and frequency response characteristics that are distinctly different at each output tap. This makes the dual points of resonance appear at different frequencies at each of the three taps, which makes the stability performance and square wave presentation different at each tap as well. Detailed response information for each output tap is presented at the end of this material.

As an example of how this manifests itself in the stock W-3AM, consider the frequency at which a -1 db response first appears from a 1 watt power level at each output tap (ref: 1 kHz):

@4Ω = 30 kHz.
@8Ω = 42 kHz.
@16Ω = 53 kHz.

Because of the double response peak in these transformers however, a second -1db response characteristic also appears in the 120 kHz - 130 kHz region at all three taps, following a +2 db to +2.5 db rise appearing in the 110 kHz to 112 kHz region. Based on measurements then, the 1 watt frequency response curve provided in the W-3AM manual follows that presented at the 16Ω tap only, with the response characteristics above that frequency or at the other taps as discussed here, not shown. More detailed response information for the stock W-3AM is given at the end of this material.

In documenting the performance of the W-3AM, Heath used a test oscillator (HP 650A) that was certainly capable of detailing performance in the supersonic region of interest (50 kHz to 150 kHz), but likely chose to keep things simple for the diy home constructor and presented just the 16Ω response curve, particularly since that impedance covered most speakers at that time. And, since most output transformers of that day were not capable of response north of 100 kHz, it was response up to that point receiving all the press attention, so the response specification as presented fit right in. But where is all of this going?

The discussion regarding a transformer’s supersonic performance is important because the application of NFB — necessary to achieve the qualities of highest fidelity — will inherently try to extend an amplifier’s frequency response to infinity. Of course, that’s an impossibility, but the amplifier will still try to do so none the less. In trying to do so however, the amplifier will invariably become unstable at some supersonic frequency due to phase shift that occurs around the resonate frequencies within the output transformer. When that happens, response starts to increase, negative feedback can then become positive feedback, and hello instability. This means it is also very important to control the HF response of an amplifier after NFB is applied as well. If a high quality output transformer is used with an extended flat response well up into the supersonic region, the problems of phase shift will then occur way above audio frequencies. Then when measures are applied to control amplifier response with FB applied, they will limit the response so as to be safely attenuated at the transformer’s highest significant resonant frequency, while at the highest audio frequencies, response will be unaffected. Not only will the amplifier then remain stable with FB applied, but do so with a flat response throughout the audio spectrum. But besides controlling the amount of feedback and circuit response with its application — as Heath did in the W-3AM — there is also potentially one other important variable to consider.

For transformers like the Acrosound that have a varying supersonic response at each output tap, the approach used to retrieve the feedback signal from the secondary winding is also very important. At the time, it was customary to use the full secondary winding as the source for NFB. This is how Hafler and Keroes themselves portrayed it in the circuits published in most versions of their own transformer resource booklet, and of course how Heath did it in the W-3 series as well. These transformers will in fact remain stable so configured with 20 db of NFB when using a purely resistive load on the output. However, things change considerably when the load is made up of the complex elements that a speaker system impedance represents. When that type of load is connected, and a dynamic trigger (music) is added to the mix, then stability becomes rather poor.

In revising the original W-3M train wreck, Heath certainly had to become aware of how this all played out with the TO-300 transformer, making it no doubt the primary reason why the NFB level was cut so significantly in the revised version. It helped to achieve a better balance of performance characteristics for the qualities offered by the Acro transformer, which always makes for improved overall amplifier performance. Ringing was reduced and stability improved, while distortion and frequency response remained within the bounds of acceptable performance. If only because of the reduced NFB level then, the W-3AM is an improved design over its predecessor.

Ultimately however, because of the particular supersonic characteristics these transformers possess, the effective HF transient stability that can be achieved with a single tap FB network is limited, and particularly so when taken from the 16Ω tap. Alternately, it will be found that the use of a more complex multi-tap feedback arrangement helps Acrosound transformers achieve the best performance of which they are capable, by allowing a more accurate composite type feedback signal to be used. It won’t slay all the supersonic dragons, but it will allow for a full 20 db of NFB to be applied around the transformer, while still maintaining excellent stability. Lacking this understanding at the time, Heath’s single tap feedback arrangement meant that the only way gains could be made in HF stability was for other performance areas to be compromised through a reduction in FB. It is against this backdrop then that a more appropriate feedback arrangement will help the W-3AM achieve better results in all performance categories.

Interestingly, for all the changes Heath made to address stability in the W-3AM, the topic doesn’t even get a single mention in the owner’s manual. Whether this was due to not wanting to draw attention to past problems, or because of the reduced performance that the stability corrections caused, the reason is unknown. But boy did that ever change in succeeding models! In the manual for the W-4 series, the problems of potential instability are at least finally acknowledged for the first time, with a hint of past problems even alluded to, while in the W-5M manual, the topic of stability is expounded on even further, and in the W-6 series manual, further yet. But it was the W-3AM that first gave stability the attention it’s due, and marked a turning point in how Heath would specify the performance of their high fidelity equipment for years to come.

Next up, we’ll look at how to turn these thoughts into practical results.

Dave

 

[dcgillespie]

MODIFYING THE W-3AM

The goal in modifying the W-3 series amplifiers is to achieve best possible performance in distortion, low and high frequency stability and transient response, and with a smoother supersonic frequency response than is produced in the stock W-3AM, that will ultimately translate into improved sound quality. To be sure these goals were met, a number of different FB networks were tried and evaluated. Besides the single tap 16Ω network of the original design, four other feedback arrangements were investigated including a single tap 4Ω network, a dual tap 4/16Ω network, a dual tap 8/4Ω network, and a triple tap arrangement. Of these, the dual tap 4/16Ω network showed the most practical promise with the TO-300 transformer. Compared to the original network, using a dual 4/16Ω network allows:

1. The use of a full 20 db of NFB, and the benefits it provides. This level is in keeping with the original Williamson specification, and that emulated in the W-2M (@16Ω) and W-3M amplifiers.
2. An overall improvement in HF stability.
3. The smoothest supersonic frequency response curve at all three output taps with the most gradual change in response throughout the tapered supersonic region.
4. A significant reduction in distortion.
5. A commensurate improvement in HF transient stability (10 kHz square waves).

There is only one area where performance is technically reduced by the use of a dual tap network, and that is with regards to unloaded HF stability on the 4Ω tap. In the original design, stability on this tap is absolute, with no amount of capacitance added to an otherwise unloaded 4Ω output causing it to break into sustained oscillation. With the dual tap arrangement however, the maximum amount of capacitance allowable is .05 uF when otherwise unloaded. Understand however that even with this reduced stability, as long as this tap is loaded with even 10X its rated impedance, then the amplifier will remain absolutely stable, with no amount of capacitance inducing oscillation. Therefore, the practical chance that any speaker system could really cause instability on this tap is beyond remote. Normally loaded, no amount of capacitance will cause either the stock design, or dual tap network to come close to producing sustained oscillation on the 4Ω tap.

In return for this one caveat however, HF transient damping is improved on all three taps, and the 8Ω and 16Ω taps now display absolute stability, whereas they were previously conditionally stable when unloaded before. Also, the supersonic response curve at each tap is smoother, with response never rising above 0db on any tap — and all this while still mimicking the response curve given by Heath for the original design on the 16Ω tap (up to 100 kHz). But the big payoff is that with a dual tap network, a full 20 db of stable NFB can now be used, with all the benefits that provides. Therefore, even though unloaded 4Ω stability is downgraded, overall performance at this tap is still enhanced due to the net increase in benefits achieved. All in all then, the slight technical reduction in 4Ω tap HF stability is considered to be more than a worthwhile trade off.

To improve LF stability, Heath reduced the size of the coupling caps into the driver stage, in addition to the aforementioned reduction in NFB to help accomplish the task. Loaded LF stability was in fact improved with these measures, but at the expense of LF distortion. The goal here then is to improve LF stability further yet to enhance even unloaded conditions, while reducing LF distortion levels to at least W-3M levels or lower.

Approach

Implementing these goals will require:

1. Re-establishing 20 db of feedback with a new dual tap NFB loop.
2. Adjusting the step network appropriately to eliminate the need of the output Zobel
network.
3. Implementing the same LF stability modifications developed for the W-2M (including new
PS decoupling network).
4. Converting miscellaneous areas as discussed for the W-2M:
Input network.
Driver stage bias resistor.

In reviewing these four items, it will be seen then that once modified, the ONLY difference that will remain between a modified W-2M and a modified W-3 series amplifier includes:

A. The output transformers.
B. The negative feedback networks used.
C. The step networks used.
D. In the W-2M, the optional use of the Zobel network developed for it.

This is as it should be. Since the only difference between these two models is the output transformer, then the only circuit differences between them will be those items that effect the proper installation of each particular transformer into an otherwise identical circuit.

 

[dcgillespie]

RESULTS

Two W-3AM amplifiers were modified using new configurations and circuit values developed for items 1&2 above, along with modifications taken directly from the W-2M work to address items 3&4. The results are as follows (averaged between the two units):

Power Output: 20 Watts RMS, from 20 Hz to 20 kHz.

Total Harmonic Distortion: (@ full power output, all greater than 20 watts RMS): @20Hz = 0.31%, @1kHz = .23%, @20kHz = 1.80%.

Intermodulation Distortion: @ 24 watts equivalent power output (maximum power output) = 0.74%.

Frequency Response (ref: 1kHz):

@16Ω Tap: Flat from 10Hz, -.10db@ 20kHz, -1db@ 55kHz, -3db@ 85kHz, -8.7db@ 104kHz, -8.5db@ 113kHz, falling quickly thereafter, with no further dips or sub-peaks noted.

@8Ω Tap: Flat from 10Hz, -.20db@ 20 kHz, -1db@ 42kHz, -3db@ 81kHz, -7.5db@ 98kHz, -5.5db@ 113kHz, falling quickly thereafter, with no further dips or sub-peaks noted.

@4Ω Tap: Flat from 10 Hz, -.50db@ 20kHz, -1db@ 28kHz, -3db@ 57kHz, -7.8db@ 92kHz, -2.0db@ 116kHz, falling quickly thereafter, with no further dips or sub-peaks noted.

NOTES: The .50db drop at 20 kHz on the 4Ω tap is characteristic of TO-300 OPT, and is present in the original design as well. It is discussed further below. Note also the different supersonic frequencies to reach similar response levels at each tap. This is an inherent quality of Acrosound catalog output transformers. Finally, note that the second supersonic peak is almost fully eradicated at the 16Ω tap, limited to a 2 db rise at the 8Ω tap, and limited to < 6db rise at the 4Ω tap, with all sub-peaks on all taps comfortably under 0 db.

Transient Response: No bounce under pulsed conditions either loaded or unloaded, with very rapid settling for any load condition. Square wave pics provided showing HF transient response.

Stability: The modified W-3AM is absolutely stable for any load condition, with any amount of capacitance added on the 8Ω or 16Ω output taps. Maximum capacitance for unloaded 4Ω tap = 0.05 uF.


ALTERNATE SINGLE TAP 4Ω NETWORK

For those who are dedicated to 4Ω speakers and want to improve the frequency response within the audio spectrum on this tap, an alternate single tap 4Ω feedback network is offered that produces flat response to 20 kHz, eliminating the .5db droop at this frequency that is present with either the stock or dual tap network configurations.

High frequency stability on the 4Ω tap with this network is identical to that with the dual tap network configuration, with the maximum capacitance allowed being .05uF in an unloaded condition. Stability on the 8Ω and 16Ω taps remains absolute.

High Frequency transient stability is worse with this network however, with nearly an 8 db rise associated with a 123kHz sub peak on this tap, a 5.4 db rise on the 8Ω tap at 123kHz, and a 3.3 db rise on the 16Ω tap at 127kHz. All other modifications remain the same when using this FB network option.

OPTIONAL OUTPUT TUBE INSTANT B+ CUT AT SHUTDOWN

The Williamson amplifier has numerous LF time constants within its design. Those in the coupling circuits of the modified design have been adjusted to achieve a high level of LF stability under normal operation. At shutdown however, these time constants all collapse at different rates, which can produce a muffled thump in the speaker when turning off the amplifier. In the modified design, this tump is controlled and minimized because of the inherent level of LF stability present. However, any thump still represents a pulsing of the output tubes when turnoff is initiated. As a result, it is recommended to include a 120 vac relay with its coil connected across the AC line, and its NO contacts used to make/break the B+ power to just the OPT CT. At shutdown then, the contacts instantly open to prevent any pulse in the output stage. This is hardly a mandatory modification to make, but a recommended one that absolutely eliminates any concern for the output tubes when shutdown is initiated. This modification will be included in the revised schematics presented.

SCHEMATIC

A schematic of the modified amplifier is presented at the end of this material, detailing the component value and configuration changes required to modify either W-3 series amplifier.

 

[dcgillespie]

EPILOG

Acrosound output transformers had a limited presence commercially as a brand name, but their design quickly ruled how Chicago, Stancor, and others wound their high performance output transformers throughout much of the 1950s. As a result, UL transformers from these companies for use in at least Heath’s W-4 series and Eico’s HF-20, 22, 35, 50, 87, and 89 amplifiers all display Acrosound’s same unique supersonic characteristics. To Eico’s benefit however, all of their models after the original HF-20 were designed with a dual tap FB network, and operate stably with (or with nearly) 20 db of NFB as well. But like Heath, their original HF-20 (which dates to the same time period as the W-3M) only uses a simple single tap 16Ω FB network. Because of this, the HF-20 operates with a limited amount of FB (14 db) so that a workable level of stability could be achieved — just as Heath had to do when revising the W-3M into the W-3AM.

Output transformer design was improving rapidly at the time however, so the quirky supersonic characteristics of the groundbreaking Acro offerings were tamed somewhat (but still present) in later transformers produced by Acro’s competitors for the amplifiers identified above. But it wasn’t until David Hafler (surprise, surprise) parlayed his founding Ultralinear work at Acrosound into his patented para-coupled series of output transformers produced under the Dynaco name that the issues of HF transient stability were really stamped out. More than a mere marketing slogan, his para-coupled transformers display the same supersonic and HF transient response on any output tap, with each tap displaying absolute stability whether loaded or unloaded when operating with 20 db of NFB. What’s more, these transformers accomplish this using a simple single tap FB network from any desired output tap, and of course with an easy change impedance tap lead format. The superb Dynaco A-470 transformer as used in that company’s ST-70 has all of these qualities, which forever cemented the Dynaco name, the man behind it, and that model’s place in audio history.

Before all that happened however, Hafler and Keroes’ first offerings from Acrosound — supersonic quirks and all — made a splash in the audio world that is still felt to this day, with features and performance that were unmatched in their day. Quirks aside, Acrosound output transformers ultimately have the necessary supersonic response, such that when coupled with appropriate FB and control networks, still set performance benchmarks within the audio spectrum that few transformer manufacturers can match today. For the W-3AM then, a little re-engineering to its early stability efforts will make the most of its measured performance capabilities, and its audible presentation as well. The impact is significant: The modifications really tighten up bass response on the low end, while presenting a new clarity in midrange and high frequency detail as well, and all with zero fatigue. As with the W-2M modifications previously presented, for those of you with W-3 series amplifiers, I think you will find the modifications presented here to be well worth your time to investigate.

One final comment is offered. There’s other W-3AM information out there, including a restoration article on the web in which the individual performing the restoration comments that he did not want to modify the design (beyond some very basic dependability changes), so as to be able to hear the unit as the designers intended. I’m sorry, but comments like this show a gross lack of understanding as to the environment that these amplifiers were developed in. If you read the original writings of Williamson, Sarser and Sprinkle, Hafler and Keroes, and so many others who helped pioneer high fidelity audio, you will not find any comments in their work relating to an “intended” sound. At the very most, they may throw a sentence in at the end of an article stating that a given design sounds like the best, or something to that effect. But that’s it. The early pioneers were all about achieving the highest reproduced accuracy based on measured performance, with that being the true understanding then of what the concept of high fidelity sound was all about: measured accuracy. One look at the amount of specifications that the typical Heath manual grew to include versus a small corner box showing the minimal specifications of a modern unit today (if they’re available at all) will testify to this fact. Assuming then that the early pioneers were chasing after a type of or intended sound is an entirely modern effort at redefining (or a gross misunderstanding of) the original concept, where any embrace of standards is all but thrown out the window. The salient point then is that the modifications offered here, are entirely in keeping with what the original designers would have intended in their quest for improving sound quality. If they had had these measures available to them, they surely would have been included. Therefore I hope you will consider that the modifications are not offered to change what the original designers intended, but rather, to enhance it.

Happy listening!

Dave

 

[dcgillespie]

FREQUENCY RESPONSE OF THE STOCK W-3AM BY IMPEDANCE TAP


@16Ω @8Ω @4Ω

1kHz = 0db 1kHz - = 0db 1kHz = 0db

10kHz = -.05 db 10KHz = -.08db 10kHz = -.12db

20kHz = -.10db 20kHz = -.20db 20kHz = -.45db

53kHz = -1.0db 42kHz = -1.0db 30kHz = -1.0db

91kHz = -3.0db 85kHz = -3.0db 53kHz = -3.0db

98kHz = -5.0db 95kHz = -5.0db 88kHz = -7.75db

111kHz = +2.0db 110kHz = +2.0db 1 112kHz = +2.5db

117kHz = 0db 116kHz = 0db 125kHz = 0db


Response falls off rapidly above the highest frequency listed, with no further dips or sub-peaks noted.




FREQUENCY RESPONSE OF THE MODIFIED AMPLIFIER USING THE ALTERNATE 4Ω SINGLE TAP NFB NETWORK


@16Ω @8Ω @4Ω


1kHz = 0db 1kHz = 0db 1kHz = 0db

10kHz = -.08db 10kHz - -.05db 10kHz = +/-0db

20kHz = -.20db 20kHz = -.10db 20kHz = +/-0db

43kHz = -1.0db 49kHz = -1.0db 58kHz = -1.0db

75kHz = -3.0db 77kHz = -3.0db 73kHz = -3.0db

102kHz = -10.5db 98kHz = -9.8db 94kHz = - 8.8db

127kHz - -7.2 db 123kHz = -4.4db 123kHz = -.90db

Response falls off rapidly above the highest frequency listed, with no further dips or sub-peaks noted.



GENERAL TESTING NOTES (through out article series)

Frequency Response measured when properly loaded at a 1 watt power level. All frequencies above 20kHz are rounded to the closest kHz in all data presented.

Equipment used for response testing includes: HP 651B Test Oscillator, HP 5381A Frequency Counter, and HP 400EL AC Voltmeter.

Equipment used for Harmonic Distortion measurements includes: HP 339A

Equipment used for Intermodulation Distortion measurement: Heath IM-5248.

Equipment used for Power Output testing includes: Heath IG-5218 Audio Generator, HP 3466A True RMS DVM.

All load resistors non-inductive.

Scope: Tek 475A

 

[dcgillespie]

Below: A top and bottom view of one of the W-3AM amplifiers used for testing in this series.


Below: Stock, 16Ω with load resistor. A 10 kHz square wave is used in all scope displays.

Below: Stock, 16Ω with .22uF cap only load.

Below: Stock, 4Ω with load resistor.

Below: Modified, 16Ω with load resistor.

Below: Modified, 16Ω unloaded.

Below: Modified, 16Ω .25uF cap only load.

Below: Modified, 8Ω with load resistor.

Below: Modified, 8Ω unloaded.

 

[dcgillespie]

Below: Modified, 4Ω with load resistor.

Below: Modified, 4Ω no load.

Below: Modified, 4Ω using 4Ω single tap FB network with load resistor.

Below: Modified, 4Ω using 4Ω single tap FB network, no load.

Below: Underside of the modified amplifier with dual tap NFB network.

Below: Both W-3AM amplifiers, modified, and ready to go. The one on the left is powered by a power supply cobbled up to handle the job, supplying the same voltages and current capability as the Heath power supply.

Below: Schematic of the modified W-3AM.

Below: Schematic of the modified Heath power supply for use with the modified W-2M and W-3 series.

 

[ab2forever]

Interested to know being as it's 6L6 based, aren't some of the more curious behaviours due to tendency to G2 oscillations or snivets?

 

[warriorpoet777]

There should be a SUPER LIKE button for Dave's work! Wish you had the color by number version of the build for us mortals.

 

[trobbins]

Top stuff Dave - exemplary.

 

[x3workshop]

Always very appreciative of your work and thorough explorations of the solutions, Dave! Thank you!

 

[crispycircuit]

DAVE YOUR HIRED !

 

[dcgillespie]

ab2 -- Not in this case, but a good point of discussion none the less. In all of my work with the 6L6 family of tubes that are fully base terminated, rarely if ever have screen grid parasitic oscillations been a concern in UL service, as the operating Gm of this tube just isn't high enough to support those kinds of problems, unless poor lead dress is involved. Now the 807 is a different animal, where stoppers are typically needed on all the grids to ensure consistently stable operation. The more common problem you will sometimes see is parasitic oscillations that will intermittently develop and ride on high power LF wave forms. These are due to less than ideal coupling between the primary halves within the coil form, and can usually be stamped out with an appropriate amount of small capacitance added across 1/2 of the primary winding. But even this problem is invariably relegated to Acro's competitor transformers, as never once have I observed it with a genuine Acrosound device. Of course, G2 stoppers are required for any kind of PPP configuration, or in production amplifier service for protection of the grid.

Dave

 

[Steve O]

Very interesting findings regarding stock performance. Your meas and sq wave pics pretty much duplicate what I found with my rebuilt stock W3AM so I can conclude my TO300 opt is not damaged as I originally feared (or yours and mine are damaged in the same manner )

Interestingly, the nature of my mods are very similar to what you developed. A major difference is that I chose to keep FB ~ 16-17dB and optimized things for 8 and 16 ohm loads only.

Overall a nice and very instructive write-up.

 

[ab2forever]

ab2 -- .. Now the 807 is a different animal, where stoppers are typically needed on all the grids to ensure consistently stable operation....These are due to less than ideal coupling between the primary halves within the coil form, and can usually be stamped out with an appropriate amount of small capacitance added across 1/2 of the primary winding.

But even this problem is invariably relegated to Acro's competitor transformers, as never once have I observed it with a genuine Acrosound device...

Dave

Ah interesting, when playing around with Partridge transformers years ago it was recommended practice to wire a capacitor and resistor from A-A to stop the ringing occuring at HF, especially visible with square wave.

That was with 6L6 btw, but I can't remember what we did with that AB2 thing we did with 2 x 807....
They do say, the 6L6 only becomes properly linear when it's about to melt.....

 

[dcgillespie]

Thanks Steve --
The dual 4Ω/16Ω network I presented certainly favors the 8Ω and 16Ω taps as well -- which frankly the transformer as a whole seems to do. You almost get the impression that the 4Ω tap was a stepchild, since the only reason I could think of for that tap being used back in the day would be for driving multiple speaker systems. The AR 4Ω speakers didn't come along until much later that decade. In general however, with all of the Acrosound catalog transformers I have worked with, the 4Ω tap is a useful tool in helping to achieve the most HF stability under FB conditions. Even if I had 4Ω speakers, I would personally still used the dual tap network, as I feel it provides the best overall performance from the transformer.

Thanks for weighing in!

Dave

 

[Brice]

Hello Dave,

Any particular reason why you did not add screen resistors in your upgrade?
Thank you,

Brice.

 

[dcgillespie]

In this case, Screen Stability resistors are simply are not needed. The 6L6 family of tubes is not a particularly high Gm design (one of the factors), and they are operated with cathode bias (another factor). There are other factors that help to determine their need, but overall, the two factors alone are plenty enough to discount their use. As for the more traditional reasons, the dissipation rating of the screen grids is not exceeded with the particular loading, biasing method, and screen tap location used, and Acro transformers are wonderfully void of inducing the more garden variety form of parasitic oscillations that would otherwise require their addition, so there was just no advantage to adding them.

Dave

 

[Brice]

Got it,
Thank you.