Modifying the Fisher 400 with EFB

[dcgillespie]

Installation

It's called a development/proving model for a reason. Before installation, I did a bench check of the board to test both the EFB bias and screen supply regulator functions of the board to make sure they would both react properly with a changing B+ level applied. The test revealed a small glitch that was basically due to the characteristics of (some of) the newer components I used for the actual build, versus the junk bin parts I used to prove the basic model. A short off course excursion was made to deal with that issue, which has now been put to rest. As a result of this little distraction however, sharp eyes will note a couple of additional components on the board, which now works perfectly.

Planing is everything in a project like this, and it paid off well in this case. The board dropped right into place, and mounted up as planned. One (original) screw mounted the L-bracket, and after a small dab of silicon grease was applied to the Mosfet, a small machine screw, lock washer and nut mounted it down through a screw hole that helps to mount the RF tuner assembly in place. This required the bottom cover of the RF tuner to be removed for access to install the hardware, but otherwise, the mechanical installation was uneventful.

Two groups of 3 wires lead from the board down through two of the cooling holes provided around the tuner filter cap. One group contains the B+ in, EFB screen supply regulator output, and ground leads, while the other contains the two AC power leads from the bias winding of the power transformer, and EFB bias supply regulator output leads. Heat shrink tubing is used to separate these lead groups, and protect them where they pass through the chassis cooling holes to prevent any chafing of the wires. The original bias supply components remain as Avery installed them, with the original white bias supply lead to the output tubes simply disconnected, isolated, and neatly tucked away. The EFB bias supply regulator output lead connects to the point where the original white bias lead was removed from.

Just as the original bias supply lead has been islolated, the screen grids also need to be isolated from their original power supply (but remain connected together), so that they can be powered from the EFB board. This requires removal of the of the original 1.2K 7 watt resistor, and replacing it with a 2.2K 2 watt resistor. Because the screens are now isolated from the original screen supply, this new resistor can be moved over into the power supply area, and connected between the main B+ cap terminal, and original screen grid filter cap terminal. This resistor is raised in value since it no longer has any screen grid current passing through it. Therefore, to maintain the original voltage drop produced by the original resistor, it must be raised slightly now.

All of the connections from the EFB board then either connect to existing connections (i.e, the ground, B+, and AC power leads), or replace an existing connection (the screen power and bias supply leads).

The sum of invasive modifications made to accommodate EFB then includes:

1. Installation of 10 ohm resistors at the cathode of each output tube socket, and

2. Disconnecting and isolating the original bias supply wire from its original connecting point in the output stage area, and

3. Isolating the screen grids from the original screen grid supply, and replacing the original screen grid dropping resistor with a new resistor.

Additionally, a terminal strip was added at the top rear of the chassis, with leads running to the cathode terminals of each output tube. With the EFB bias control accessible from the top side of the chassis now (on the EFB board), and the cathode terminals now accessible top side as well, it makes for very easy bias adjustments on the amplifier. The terminal strip and its leads were also accommodated by small cooling holes at the rear of the chassis, so again, no physical alterations were made for the installation of EFB in the 400.

Side Bar: Many folks prefer to install IBAM circuits, which the EFB modification will easily accommodate. Whether such controls are present or not, I always prefer to install a quad of well matched tubes, as this not only helps ensure the best push-pull balance within a given channel, but also ensures the best performance match between channels as well. My 400 does not currently have an IBAM installed, but does employ a tightly matched quad of NOS tubes, making adjustment of the bias very easy now.

With the EFB modification fully installed now, basic tests were run (on full AC power) to see where things stood. Observations include:

1. Power output remains exactly as it was in stock form, with each channel individually producing 25 watts RMS @ 1 kHz by itself, or 22.5 watts RMS each, with both channels driven.

2. Full power 1 Khz THD with both channels driven now produces an average of just .15% THD. This is a very low distortion level, approaching that of some of the very best vacuum tube amplifiers, and is a major improvement over the original design, where both channels were averaging .95% THD with both channels driven.

Side Bar: Much (>50%) of this is due to the EFB screen supply regulator tightly managing the screen voltage now. In the original design, the screen voltage dropped about 55 volts under worst case conditions, destroying the original low distortion operating point of the stage as power was increased. With EFB, the screen voltage still drops slightly, but now, about 1/5 of the original amount -- and most importantly -- it drops in exact relation to any drop in the plate B+ supply, helping to preserve the low distortion operating point. For pentode output stages, effective control of the screen grid is one of the single most important design requirements to maximize performance from the stage.

3. Because the EFB bias and screen supply regulators both provide significant filtering in addition to providing EFB action, residual power supply hum is now extremely low. Measurements were not made regarding this, but it is now very hard to tell if the unit is even turned on through my Cornwalls with the volume at minimum. Subjectively, there is virtually no power supply hum.

4. Original tests indicated that the low distortion point would be achieved with a total quiescent idle current of 22 ma per tube. But those tests were not actual EFB tests, but simulated EFB tests. With the real deal installed, the low distortion point came in at just 21 ma total quiescent current per tube. This is an unbelievably conservative mode of operation for 7591/7868 class tubes. With an actual plate B+ of 425 vdc, and figuring that 20 ma is flowing to the plate (to account for screen current), that equates to a plate dissipation of just 8.5 watts per tube -- which is just under 45% of the rating for the tube. If that doesn't extend tube life, then nothing will!

5. At quiescent conditions, the screen grids are now running at just over 300 vdc. This is nearly a 100 volt drop from what they were running in the stock design, which has taken away all the stress that the stock design placed on them. The EFB screen supply regulator now maintains the new screen/plate voltage relationship regardless of AC power, power output, or B+ level conditions, maintaining the low distortion operating point of the output stage whether each individually, or both channels are driven to any power level up to maximum power output. An additional huge benefit of the new operating conditions is that the screen grids no longer show any color at all, even when grossly over-driven. This too will have a huge effect not only on tube life, but also in maintaining the tubes' original match conditions: Since the screens now run so much cooler, they will not be inclined to distort their shape, which will help preserve either the original match characteristics of a quad or pair, or the that of the individual tube.

6. The new operating conditions in the output stage not only do not require any screen stability resistors to be installed (although they won't hurt if already installed), but also almost certainly will not require the control grid return resistors to be lowered either. Because the tubes now run so cool (in tube degrees that is), the chance of any reverse grid current effects under the new operating conditions is virtually nil. Like the screen stability resistors, lowering the grid return resistor values certainly won't hurt -- although the phase inverter won't like you for doing it. But that will be dealt with in the next installment.

7. After 3 hours of operation, the power transformer was nicely warm to the touch, but hardly hot, and you could easily keep your entire hand on it for as long as you wished. This is out of case of course, but subjectively, it feels notably cooler than with the stock circuit. With a reduction of quiescent current in the output stage of nearly 45 ma, this equates to over 100 ma of current no longer being drawn from the power transformer, considering its voltage doubler configuration, and the charging current required by the doubler caps for the operating current saved.

Pics include:

1. The completed board awaiting installation, glitch free. Green lead group is for the EFB bias supply regulator, red lead group is for the EFB screen supply regulator.

2. Close up detail of the Mosfet mounting. In operation, there is no discernible heat generated by the Mosfet after hours of normal operation, or 1/2 hour of operation at full power output in both channels, ensuring its long life.

3. Detail of the lead groups dropping below the chassis.

4. The completed top side installation. The bias control is easily accessible at the top right of the board.

5. Terminal strip for monitoring output tube current draw.

Next time: Underside shots of the installation, and clean up work: dealing with that pesky phase inverter.

Dave

 

[audiodon]

Interesting.
So, there'll be some rethinking necessary if my takeaway with this installed is correct:
1. No 100 ohm screen stability resistors will be needed, but if there will not need to be removed.
2. Grid resistors of 330k can be left there. Is that what you mean by screen grids?
If you've already put 220k or 200k there, just leave 'em?

 

[dcgillespie]

Don -- you've got the gist of it.

Grids are numbered and named by their historical development in time, and by what they do.

The original grid was simply called "the grid" or "control grid" since that is what it did (controlling current flow). No other designations were needed, since only single grid tubes (triodes) existed when the development of the grid was first discovered.

The second grid to be developed was aptly called grid no. 2, and/or the screen grid, because that is what its function was: The screen grid "screens" the original grid from the plate, by forming an electrostatic shield. This was one exciting discovery, because all of a sudden, higher RF frequencies could effectively be amplified, and the voltage gain was significantly increased over that of the original triode. Tubes with two grids are called Tetrodes.

Tetrodes had their problems though, which were solved by the development of the third grid, called......wait for it.........grid no.3, and/or the suppressor grid, because again, that is what its function is: The suppressor grid suppresses any electrons that are knocked off the plate from being attracted to the screen grid, and drives them back to the plate. This formed the well known pentode tube, and perfected the tube from the standpoint of its being able to handle very large plate currents over a very wide voltage range, with still yet even more gain available.

In audio work today, the control grid is associated with the 330K resistors of discussion, and operate at a negative voltage in fixed bias amplifiers. The screen grids are those which operate at a high B+ voltage, and often need the screen stability resistors added to them that I have written about. The suppressor grids (or beam forming elements in beam power tubes) are often internally connected to the cathode, or sometimes brought out to their own external connection, as with the EL34 (via pin #1).

In my next installment with the 400 EFB modification, I will go into detail about the issues of the phase inverter, and the effects of lowering the control grid return resistor values on the phase inverter.

Dave

 

[BrokenHill]

Brilliant work Dave.

 

[BuzzK]

We are witnessing a "one size fits all" solution that represents, IMHO, the greatest general advance in tube craft technology that I have ever seen. The documentation and attention to detail is exquisite, in Dave's usual "style."

Bravo Dave! Truly great stuff here.

 

[gadget73]

Tetrodes had their problems though, which were solved by the development of the third grid, called......wait for it.........grid no.3, and/or the suppressor grid, because again, that is what its function is: The suppressor grid suppresses any electrons that are knocked off the plate from being attracted to the screen grid, and drives them back to the plate. This formed the well known pentode tube, and perfected the tube from the standpoint of its being able to handle very large plate currents over a very wide voltage range, with still yet even more gain available.

Just a side note, not all pentodes actually have 3 grids in the traditional sense. Many tubes commonly called pentodes (6L6 variants are a good example) actually contain 2 grids and a structure called a beam forming plate. These plates form the electrons into "sheets" and serve the same electrical function as that third grid. These type tubes are known here stateside as beam power tubes. In England, they are sometimes called a Kinkless Tetrode, and the KT66 is a well known example of that.

 

[dcgillespie]

Absolutely right! I made simple mention of the beam forming elements in the next paragraph of that post, but your detailing their distinction and operation punctuates the difference between their operation, and that of an actual very loosely formed Suppressor Grid. Thanks for adding that clarification!

Dave

 

[dcgillespie]

EFB Wrapup

With this post, I plan to wrap up the EFB portion of the project, but will also mention that my overall Fisher 400 project is hardly over. This has been one of those projects that's like a thread you pull on that never quits. The more I work with it, the more I see that can be sincerely improved. None of this is meant to take away from the original design of the 400, or even throw stones at its stock performance. But time has moved forward since this unit was built, and answers now exist to design compromises made at the time, that can be greatly improved upon today. EFB is one of those answers.

In the 400's audio section, the basic power supply, output tubes, loading conditions, and biasing arrangement were first chosen to deliver a target power output level. From there, the rest of the output stage operating parameters can be developed. In this case, it dictated a screen voltage of about 300 volts for the loading conditions chosen. The problem is, you cannot use a simple resistor to drop over 100 volts from the main B+ supply to power them. The reason is because the screen grid current draw rises dramatically with increasing power output. In this case, it rises over 400%, meaning that whatever value of resistance would be correct under quiescent conditions, would leave the screens with hardly any voltage at all as power output increased. The result would be that only a small fraction of the power output capability would be realized, and distortion would be astronomical. So what do you do?

Since cost constraints were not completely out the window at the time this unit was developed at Fisher, you go ahead and use a dropping resistor, but make it much smaller, so that as screen grid current increases with power output, there is still enough screen grid voltage present to produce the full power output capability of the design. But we can't size the resistor to drop the screens to 300 volts at full power, because that would mean that the screen voltage is dropping about 100 volts from quiescent conditions where the tubes are biased at. While such a drop in screen voltage would allow the full power output of the unit to be developed, it would do so with highly elevated distortion, because the diminished screen voltage would seriously alter the operating point biased under quiescent conditions. So what do you do?

You make the screen dropping resistor even smaller. Now the screen voltage doesn't drop too awfully bad, so full power AND reasonable distortion can be attained -- but at a cost to the tubes. With the screens operating at such an elevated voltage relative to what they should operate at in this design, it still requires the quiescent current to be elevated under quiescent conditions to minimize distortion when the screen voltage does drop, has the screens handling much more current than they should (causing them to glow), and hardly produces optimum performance. The result is "acceptable" distortion, and stressed screen grids within the tubes -- but tubes were plentiful and cheap at the time, so there's your compromise.

EFB solves all of these problems by initially dropping the screen voltage down to the appropriate level, and from there only allows it to drop in relative lock with any drop in plate voltage as well. Furthermore, it does the same thing with the negative bias voltage as well, so that the low distortion operating point established under quiescent conditions is maintained throughout the power capability range of the amplifier. For the tubes, it allows a much lower quiescent current to achieve the low distortion operating point (which lowers plate dissipation), and keeps the screen grids cool by operating them at the proper voltage to begin with. Full power is still developed, distortion is held at the lowest possible level, plate dissipation is minimized, and in this case, the screen grids operate safely now. It is truly a win-win-win-win situation.

Fisher was certainly aware of the screen grid issue with their design, and sought to protect them by choking the performance of the phase splitter as I have previously mentioned. In looking at that issue, that produced the never ending proverbial thread I mentioned earlier, so I have decided to start a new thread on dealing with that, and the remaining issues relevant to this project. For now, a few pics and a schematic to close the EFB portion out:

1. The EFB wiring is routed between the heat shield and the tuner assembly, around the tuner assembly, and combined with existing wiring of the unit.

2. The EFB wiring disappears into the unit:

A. The extra red wire on the main B+ terminal feeds the EFB board.
B. The black wire (barely visible) at the 10 o'clock ground lug position on the upper can cap is the ground lead from the EFB board.
C. The red lead going to the screen grid terminal on the right most output tube socket is power for these elements from the EFB board, and
D. The black wire above the screen power lead is the bias supply lead from the EFB board.

Also notice the new 2.2K 2W resistor connected between the two can caps, that replaces the original dropping resistor supplying the screen grid in the original design. It is necessary to continue powering the remaining circuits, but is raised in values since it no longer powers the screens.

3. The original bias lead is neatly tucked away over in the area where the original bias supply resides.

4. The schematic, as installed and drawn on the back of my Fisher service manual.

My unit has some 50 hours on it now with the EFB modification installed, and operates notably cooler. Sonically, the impact is positive, with the same tonal presentation, but notably stronger dynamically. This has been the case with previous EFB installations as well. The biggest impact in this case however, is on the tubes and their life expectancy. Their operating conditions have been radically changed from the stock design, to one that is much more friendly and optimum for them to operate in.

With EFB installed, this paves the way for improvements in the driver stage, which produces significant measured and sonic improvements as well. These will be detailed in the thread coming up.

Happy listening!

Dave

 

[sony6060]

Tidy

In a knee jerk reaction while reading your post, I ripped my pocket off my pants getting at my wallet. Where do I send funds?

 

[larryderouin]

:zoom::yippy::yippy::yippy:

 

[Ths364]

I'm reading this thread, and nowhere can I find a definition for what EFB stands for. Somebody? I presume that the FB means feed back...

 

[larryderouin]

Enhanced FIXED BIAS (Edited)

 

[larryderouin]

Scroll down thru the threads a couple pages and read the threads Dave has put up about it.

 

[dcgillespie]

If I may, it actually stands for Enhanced Fixed Bias. It is a concept that maintains the the low distortion operating point in the output stage, regardless of how the AC line voltage or power output conditions in the amplifier(s) may cause power supply voltages in the amplifier to sag or vary. It automatically adjusts the quiescent grid bias voltage under dynamic conditions as necessary, and also does the same thing for the screen grid voltage for pentode based designs as well. The end result is much lower distortion, even with both channels driven, and output tubes that can idle at much lower quiescent current levels for greatly increased tube life.

I hope this helps!

Dave

 

[Ths364]

Excellent! Thanks guys- I felt like I was missing out on something big, but couldn't read between the lines enough to sort it out. I looked through two threads about this trying to figure it out. I'm going to read the whole thing now!

 

[dcgillespie]

If you want to start at the beginning of EFB, here is the original article where it is introduced:

http://www.tronola.com/A_New_Look_At_An_Old_Friend.pdf

Dave

 

[larryderouin]

Sorry Dave. Somehow I got Feedback and Fixed Bias mixed up in the cobwebs between the ears.

Larry

 

[IMTubeHead]

I'm in as well. I have SO MUCH to learn and I can't think of a better teacher than you, Dave...

 

[boojum]

Hi Dave, I contacted you some time ago in regards to an EFB mod for a Fisher 500B, and it looks like that possibility may be getting closer. I'll follow the thread with great interest.:yes:

 

[Fur Face]

I’m working a Fisher 400 with serial number 49038 and it has the X OPTs. My intention is work my way from one end to the other. But I’m having a problem with Mr Gillespie EFB circuit that I assembled. Any input to help resolve my mistakes is appreciated.

I’m using the schematics that Mr Gillespie designed for the Fisher 400 with the X transformers. The problem I’m having is no + output voltage going to screen grid of the power tubes.

With 50v ac going to the amplifier I’m getting 10v ac at the input of the bridge rectifier of the EFB circuit and after the bridge is -21v dc then going to the first transistor the collector is seeing -21v also. The emitter of the first transistor going to a IBAM is seeing -.369v and the base is seeing -.87v on the same transistor. Across the pot the second transistor is going to ground at the emitter and -.87 at the collector and -.56v at the base pin.

The mosfet is seeing .450 at the gate the drain is seeing the B+ 165 on the other side of the 1.8k 4w resistor and the source is at -.537.

Pictures with the transistor hooked up to a piece of test equipment with two different results. Stand alone and in circuit.