Improved SCA-35/ST-35 Performance

[cedricb]

Is this mod the same thing than the infamous LM317 mod (undocumented) for the revision D ?
If so then the PCB should accommodate this EFB without any big surgery? Can somebody confirm please?

I've found this article which describes a cascade of DN2540 and LM317 (figure 13C) which is best suited for tubes... http://waltjung.org/PDFs/Sources_101_P2.pdf

[Shannon Parks]

Dave, very cool article and thank you for documenting this Dyna mod. Seems like a pretty simple mod with an understandable result.

If there ever *is* a Rev E, I probably will allow for grid stoppers at a minimum. On the DIYST35, folks could cut the trace near the socket and use 1206 surface mounts. I think a new biasing scheme (like this) and a LTP would really be cool, too.

Regarding the output numbers on the Z565 I took, I honestly can't remember my test setup. I presume it was single channel. But these max output tests are extremely sensitive to the line voltage, tubes, and a several other factors like test methodology. And I was just a pup at the time. Admittedly, I should probably redo this test.

That said, I did do testing of individual bypassing vs paired bypassing vs all-in-one bypassing and individual biasing *did* test noticeably better. I can revisit this, too, this winter as I have no documentation. So I'm not sure if this is a step back in terms of performance at all with the original circuit. Don't want people to panic, after all.

--Shannon

[dcgillespie]

Cedricb --

Thanks for your continued interest. The EFB modification may appear similar to configurations you found in the article you provided, but is in fact almost the exact opposite. They are similar in that both are using a 3TR device for active control, however, with EFB, the regulator is regulating voltage drop in the face of a varying current, whereas in the article you provide, the effort is to regulate the current itself.

These are two very different concepts. Regulating current is "nice" in that it causes the quiescent bias point to remain at a very fixed point once set, but it is not very good for ultimate performance. This is because in a class AB1 design such as these amplifiers represent, the average current draw increases with power output. Preventing this current draw from increasing by using a current regulator limits power output, and increases distortion as well. With EFB, the cathode voltage is absolutely regulated (from a tube current draw standpoint), meaning the average current draw can increase with power output as it should in an AB1 design, and all the available performance is then possible. The efforts with current regulators "work" only because the regulator is bypassed with a large cap. Without this cap in a current regulator scheme, there would be no output at all, since there would be no change in current flow permitted. In the EFB installation, use of the original Dynaco cathode bypass cap on the output of the regulator is only there for the stability of the regulator itself, rather than any need to couple the audio signal to the load.

Finally, with EFB, the bias is also allowed to be modified by prevailing B+ conditions, so that it will always remain at an optimum point of low distortion operation, rather than drifting away from it as B+ changes with varying demands in output. This feature produces the same low distortion results as that of using individually regulated bias and B+ supplies.

Providing both of these features together with one simple circuit is at the heart of the EFB cathode regulator concept, and I believe is quite unlike any other bias scheme used before. It is easy to install, and the performance gains it provides in the Dynaco amplifiers are very real.

Dave

[EWBrown]

Is this mod the same thing than the infamous LM317 mod (undocumented) for the revision D ?
If so then the PCB should accommodate this EFB without any big surgery? Can somebody confirm please?

The LM317 pads on the Rev D board allow for using one LM317 and itscurrent setting resistor per EL84, it is not compatible with the EFB design. It is undocumented, but pretty easy to figure out. 36 ohms, 1/4 W would get you pretty close to 35 mA cathode current per each EL84. THis allows using any motley assortment of four EL84s / 6BQ5s / 6P14Ps and keeps them all at (near) 35 mA.

If one isn't adverse to using (reasonably) matched quads, then each adjustable cathode biasing "network" can be simply replaced with a 390 ohm, 2W resistor (and its bypass cap). Leave in the 10 ohm resistor if you want to be able to easily measure the cathode curent for each tube. (This also makes the total RK for eachchannel 400 ohms) That eliminates four trim pots, and eight (or twelve) other resistors.

I suppose that if all of the cathode biasing components were just left off the board, and all of the EL84 cathodes (pin 3) were wired together, then an off-board EFB circuit could be easily implemented.

/ed B

[dcgillespie]

Hi Shannon --

No panic intended or wanted on this end either! In my mind, the value of your efforts to this hobby are inestimable. It is simply mind boggling to me that directly because of your efforts, folks can easily build a clone of the original ST-35 with improved physical layout and input stage configuration, from a great set of instructions provided to make it all happen with certainty. It is a testament to you, and Dynaco for the original product they produced. I am quite certain that those who want to install the EFB mod with your boards will find it rather easy to do. As always, where there's a will, there's a way.

I would welcome very much any further documentation you can provide regarding your performance evaluation for both any type of resistive cathode bias concept, and that of EFB operation if you chose to try it.

For the Dynaco 6BQ5 amplifiers, the primary issue at hand is that the loading of the Z-565 transformers is simply too low for any type of traditional resistive cathode bias operation. The Z-565 is an 8K primary transformer, and is an ideal load for these tubes with traditional cathode bias when operated in pentode mode. But these amplifiers use the 575's screen taps which effectively raise the required primary impedance for low distortion operation when traditional resistive cathode bias is used. Since the impedance offered has not been raised, it is therefore too low for proper resistive bias operation. As a result, when resistive bias is used anyway, the excessive current draw developed as a result of the "lowered" impedance causes too much bias deviation across the cathode bias resistor with the application of power. This produces the large shift in operating point, and the raised levels of distortion as a result. Dynaco attempted to resolve this somewhat by connecting all the cathodes together, so that one channel might help hold back on the bias shift in the other channel, as I explained in the article.

The net result of this is that really any type of stand alone resistive bias system -- applied either on a per channel or per tube basis -- will suffer similar results as that which happens when both channels are driven in a stock SCA/ST-35 amplifier. The shift in operating point is excessive, which results in the raised distortion levels. Also, since any type of resistive bias system limits peak current capability, power output is reduced as well. While I did not know who provided the data published on the Triode website when I wrote the article, I do believe it still illustrates valid single channel performance when separate tube or channel cathode bias resistors are used. That it mirrors my data for the stock SCA-35 when both channels are driven is very telling. But again, I would most welcome your confirmation of the results I provided. For the record, my test conditions included:

1. 120 VAC maintained under all operating conditions, as specified for the SCA-35.
2. Tests made from the full 16 ohm winding of the Z-565 transformers, into a 16.0 ohm, 100 watt resistive load for each channel.
3. Maximum power output was measured at the onset of visual clipping in all cases.
4. Distortion readings were taken at a power level 1 db down from 17.5 watts RMS in accordance with the original specifications. Note that this power level is very close to the maximum power your distortion readings were taken at in the data posted at Triode.
5. Equipment included an HP 339A THD distortion analyzer for test signal, power output voltage readings (true RMS), and THD distortion readings, and a Heath IM 5248 distortion analyzer for IM distortion readings.

Understand too that a 2-3 volt drop in B+ is nowhere near the same as a 2-3 volt drop across a cathode bias resistor. The presence of the resistor itself limits peak current draw as mentioned earlier, and also causes the tube to limit it's peak current handling ability as well. Since the dynamic resistance of the EFB cathode regulator is effectively zero, it neither limits the peak current capabilities either due to itself, or through the tube due to a changing operating point. Therefore, power output increases with it's use, and distortion falls.

Finally, as explained, by tapping the bias for the 3TR from the B+ supply for the output tubes, it further causes the bias (once set) to always maintain the same relative relationship with the B+ level, regardless of how or why it might vary -- be it due to changes in AC line voltage, current draw, or otherwise. It is this feature that basically allows the same low distortion figures to be maintained when both channels are driven, or just one is. The use of fixed bias produces a significant drop in distortion to begin with, due to the improved operating conditions it provides for the load offered. But the floating feature of the EFB cathode regulator allows those reduced distortion figures to be maintained under all conditions of use, whether there is a significant drop in B+, or not. It really is a 1-2 punch for improved performance in this case.

Thanks for your comments. EFB is pretty unique in the benefits it offers to these Dynaco amplifiers. I hope you can try it out and report back. I would look forward to it!

Dave

[jgf]

Just a note to say thanks for this info! I tried this mod,
and its excellent. It really cleans up the "compression"
effect from heavy bass notes. I've got st35 repro boards
in a sca35 case, power supply is two JJ cans for 100uF
on the first two stages, then perfboard where the phono
board was, which now holds the EFB parts and 47uF
for the driver stage.

I had already added grid stoppers
and 68pf caps to ground at the EL84 inputs, as per
Morgan Jones' book. This cleaned up high end ringing,
I think the stock circuit, being designed with carbon comps
drops more high end, but with metal films I was getting
ringing and even oscillation with some tubes.

Still playing with the global feedback resistor, now at
47k from original 27k, may try something in between next.

[dcgillespie]

jgf -- Thanks for trying the EFB mod! No doubt the extra power and lower distortion the EFB mod produces provides the improvement in the compression effects you previously noted. I also noted that the overload characteristic is greatly improved as well.

If you have not already done so, I hope that you also included the 100 ohm screen stability resistors, in addition to the grid #1 stopper resistors you installed from your previous work. Since the output stages now operates under conditions of fixed bias, this, along with the power supply upgrades you have performed makes these resistors mandatory to eliminate any tendency towards output tube arcing, that would otherwise exist with high Gm tubes operating under the new conditions provided.

You should also enjoy much longer life from your output tubes now since the actual plate dissipation is now easily under 9 watts per tube when they are biased at the proper level of 27 ma each as instructed.

Thanks again, and good luck with your amp!

Dave

[jgf]

Dave,

You caught me, I haven't put the 100ohms on the screens
yet, I'll do that as soon as I can. I do get quite a bit more
headroom across the frequency range, as you say.
Great stuff! Thanks again,

John

[paart]

Dave,
You have obviously put considerable effort and research into this issue and devised a very unique and simple solution! It's also one that will easily fit into the diminutive chassis that Dynaco "featured" with these models! I'm starting the process of doing a complete rebuild on one of my ST-35s, and look forward to incorporating your circuit.

I have a question concerning regarding the output stage cathode bypass. You stated above, that the bypass capacitor is unnecessary with your EFB design, (as is the case with other fixed bias designs), except for regulator stability. My thinking is that the fixed bias output stage would actually be "better off" under certain operating conditions, with this capacitor completely removed. In checking the recommended operating criteria for the LM337, an output capacitor is needed for regulator stability, as you noted. The recommended value is a 1mfd tantalum or 10mfd AL electrolytic. (????) I'm curious if you have any thoughts on reducing the value from the stock 100mfd. I wonder if something in the 1 to 10mfd range would be appropriate here, allowing, perhaps, the use of a film capacitor.

Thanks!

---Art---

[dcgillespie]

Hi Art --

Thanks for the kind words, and for the interest in the EFB modification.

Regarding your question about the regulator output cap, you are absolutely correct: One is required, but at nowhere near the value the original cathode bypass cap represented in the Dynaco design. I would think that a 10 mfd electrolytic would be more than sufficient to do the job, although I have not specifically tried that value, nor any caps of different make up.

At the time I wrote the article, I was trying to keep the modification as simple as possible to encourage trial, and also to have it in keeping with Hafler and Laurent's basic design philosophy as well. Therefore, the old bypass cap was pressed into service to handle any regulator issues that might arise.

Like you, I also have an ST-35 I need to rebuild. I will be installing the EFB cathode regulator circuit, and had already planned on using a much smaller cap value in that position as you have suggested. For now though, I have been working on an EFB supply regulator that I am initially developing in an Eico HF-89, but will end up being appropriate for any traditional fixed bias amplifier such as the Dynaco MK IIIs or ST-70s. I hope to start writing that article soon.

In the mean time, any experience you develop and can pass along about the value of this cap or the EFB modification in general would be greatly appreciated!

Thanks again for your interest.

Dave

[DeathRex]

I'm making a Eico-89/Eiclone now. I'd love to see your article.

[dcgillespie]

I should be able to start writing the article soon.

The Enhance Fixed Bias modification for the HF-89 will be a supply regulator version that installs in the bias supply itself -- as opposed to the cathode regulator version presented for the Dynaco article. However, like the Dynaco modification, it will be a simple, straight forward installation that should be easy to implement.

The Dynaco EFB mod accomplished two things: It converted the output stages to fixed bias operation, which improved power output and lowered distortion. The Enhanced element of the design then allowed the amplifier to maintain it's low distortion performance regardless of the demand on the power supply. Since the HF-89 is already a fixed bias amplifier, the supply regulator version for it will simply add the Enhanced element to it. Since the 89s amplifiers place a serious demand on it's power supply, the EFB modification makes a significant contribution to performance improvements with it as well.

I will keep you posted on a publish date. Thanks for the interest!

Dave

[paart]

Dave,
I have a variety of caps that I can try on the regulator output. It won't be right away, as I'm just getting to the "ordering parts" stage, but hopefully within a month. I can try some basic measurements and listening tests. But I think assessing absolute regulator stability could be pretty difficult, particularly as yours is tied to B+ variations! I suspect that it would take a separate dedicated test circuit to do this correctly. I may be able to attempt this, at some point.

One additional benefit of your circuit that you didn't mention, and might not be immediately apparent, is that it adds a worthwhile "bleeder" resistor to the original Dyna power supply! (Unadvertised feature)

---ART---

[dcgillespie]

Art --
With regards to absolute regulator stability, while developing the EFB concept for the Dynaco mod, I was certainly mindful of potential regulator stability issues. But they never became a concern at any time, even when I intentionally over biased the tubes (too little quiescent current draw) to check the regulator's stability under conditions of severe crossover distortion during the 1 db down 20 Hz power output tests. I felt that this would be a worst case scenario to induce any potential regulator related stability concerns, which of course never happened. Frankly, if I were to do anything to the modification relating to regulator stability, it would be to add a small film cap (no more than say 1 mfd) across the existing 100 mfd electrolytic to reinforce HF decoupling, but again, no issues were ever noted with the mod as presented.

To be honest, the bigger stability concern I had was to make sure the mod did not significantly disturb the LF stability of the amplifier. As originally designed, the cathode bypass cap is in the audio signal path, and therefore becomes a factor in the LF stability equation of the overall design. We're talking about stability issues that can arise in the 1-2 Hz region with NFB amplifiers. As it turns out, the conversion to EFB operation only strengthened the low frequency stability of the amplifier, as determined by examining the amplifier's before and after behavior under pulsed conditions at various output levels. And, since it is basically no longer in the audio path (the impedance of the regulator itself is extremely low at audio frequencies, typically ~ .01 ohm or so), the value of this cap has a negligible impact sonically.

Finally, while the EFB mod also provides a bleeder action as you mentioned when the unit is turned off, another side benefit I did not mention is that it also eliminates any concern of operating the unit with one or more output tubes removed, which the manual for the original design specifically warns against. While there would be little if any reason to intentionally want to do this, the condition can happen none the less if an output tube ceases to conduct current for any reason. With the original design, this would result in under biasing the remaining tubes, stressing them greatly beyond the extreme limits they were already biased at, and shorten their lives even more. With the EFB modification, if an output tube fails to conduct for any reason, the lost current draw would cause a slight rise in the B+ which the EFB circuit would react to, readjusting the bias on the remaining tubes so that they would basically be none the wiser to the new situation. This "safety net" is also provided when each tube is biased with it's own individual cathode resistor, but of course, that mode of operation permanently reduces power and increases distortion as documented in the article.

Thanks again for the interest!

Dave

[Marshman]

Perhaps I am overlooking this, but what are the wattage ratings on the resistors and what is the voltage rating on the 4.7 cap?

Thanks,

mm