The Fisher X-1000 and EFB

Enhance Fixed Bias has been used successfully in a variety of output stage applications to date. In all cases, distortion was dramatically lowered by correcting the operating point in the face of drooping power supply voltages with the application of power, and in some cases, also by correcting for some very much less than optimum operating conditions of the original design. In most cases, quiescent current was also able to be notably reduced (or optionally reduced) to provide for added tube life and cooler operating temperatures, while in some cases, dangerous conditions for the tubes were also resolved. The total benefits achieved in each scenario then obviously depends on the compromises built into the original design.

An analysis of an X-1000's output stages show that they contain some of all of the above:

1. With no real good way to passively regulate screen grid voltages back in the day, many manufactures just powered the screens through a simple dropping resistor -- applying much too high a screen voltage under quiescent conditions, and allowing it to fall in line by way of elevated screen current through the dropping resistor and general supply droop under conditions of elevated power output. The X-1000 is no different in this regard.

2. Such a large drop in voltage to the screens between quiescent and full power conditions produces a major shift in the operating conditions of the tubes, causing a significant rise in distortion as power output increases.

3. Operating the screens at or very near their maximum rated voltage under quiescent conditions always invites potential problems -- and particularly so if no screen stability resistors are installed. Such is the case with the design of the X-1000's output stages, meaning that at the very least, adding screen stability resistors to these units is a must. Is it any wonder that so many of these units are found with damaged cathode resistors, or worse?

In short then, while the OPTs presents a good working load for the output tubes used in the X-1000, the operating conditions the tubes operate under are rather hard on the screen grids, and cause increased distortion at elevated power output. But then further, there is this:

The EL34 output tube used in the X-1000 is rather unique. It is a true pentode (meaning it uses a suppressor grid rather than a beam forming plate), and since it is not a Beam Power Tube, the control and screen grids are not aligned within the structure. The result of this construction is that the screen grids draw a rather large amount of current under the conditions used in the X-1000 at full power output; much more than an equivalent Beam Power Tube would draw.

The up shot of this is that using any kind of a general screen dropping resistor to obtain the screen voltage is rather poor practice with these types of tubes, meaning that any dropping resistor used must be rather small, which only serves to further cook the screens under high power conditions.

As a side note, some pentode tubes work very well with this type of operation -- but the EL34 is not one of them. On the other hand, the Beam Power 6L6 family of tubes is. How many tens of thousands of Fender Amps were (are) produced using this tube, using these very operating conditions, and universally praised for the copious amounts of "clean" power they could (can) produce? While across the pond, in search of a new sound, Mr. Marshall ran EL34s in the same basic way, but his equipment was NEVER known for its clean power capabilities. Now whether Jim took a look in Avery's play book, or Avery took a look at Jim's, the bottom line is, that other than for the excellent OPTs used in the X-1000, its output stages are basically set up the same way they are in a 50 watt EL34 Marshall guitar amp -- which is known to chew up output tubes, and add its own unique coloration to the sound. That's all well fine and good for a production amplifier, but hardly for a REproduction amplifier.

None of this is to condemn the EL34. It just needs to be operated properly to obtain all the benefits of its otherwise excellent linearity under conditions of high power output. In pentode mode, that means operating it with a healthy separation between the plate and screen voltages. Otherwise, the screen grid in this tube can draw enough power so as to suck power away from the plate during periods of maximum power output, producing some rather weird clipping.

Taking all of this into account then and using simulated EFB testing, the same 50 watt RMS power output level can be obtained by reducing the screen grid voltage to about 85 volts under the plate voltage during quiescent conditions, which immediately produced a notable drop in distortion levels. Preliminary tests in stock form had each channel producing just over .5 % THD @ 1 kHz at full power output with each channel driven individually, and well over 1% with both driven together. Using simulated EFB, distortion dropped to just .18% at full power, whether both channels were driven or not.

At 20 kHz, the stock design produces nearly 4% THD at max power output (over 6% with both channels driven), while with EFB, distortion dropped to just .7% in both channels.

The other huge advantage with EFB in this amplifier is how the screen grids are treated. In the original design, the screen grids operate at nearly 140% of rating under full power conditions, while with EFB, the screens approach 85% of their rating -- with the tubes producing the same amount of power output, but with 1/5 the amount of distortion.

Preliminary testing also shows the tubes to require about 38 ma of quiescent current with EFB, which translates to a plate dissipation of about 13.5 watts per tube, for a tube conservatively rated to handle 25 watts of plate dissipation. EFB then creates a low distortion set point with the tubes idling at just under 54% of their plate rating. This operating level represents about 80% of the original operating level -- which was already operating the tube rather conservatively. The difference is that with EFB, not only are the tubes running cooler yet, but also the screens are operated well within their ratings at all times, and distortion has been significantly lowered to boot -- and all while maintaining the same power output production capability. This should all bode very well for extending the basic life of the output tubes -- and protecting them as well from any tendency towards self destruction.

Overall, these achievements are certainly worthwhile enough that the installation of EFB in the X-1000 would represent a significant improvement in performance, and tube life. With the data generated then, the next step is to figure out how to install it into an already well populated chassis. That will be my goal for the next couple of days. For now however, pics include:

1. The X-1000 all connected up for simulated EFB testing.

2. As I said, weird clipping produced by the stock Fisher output stage at 20 kHz as full power output is approached.

3. Sharp eyes will notice that with the THD test set set on its 10% full scale setting, the waveform in pic #2 contains nearly 4% THD.

4. At the same power output level (about 46 watts RMS) with EFB, the wave form is clean right up to the point of overload, with safer operating conditions for the tubes and lower quiescent current levels as well. And........

5. Sharp eyes will again note that on the 3% full scale setting, this waveform now only represents about .7% THD.

Time to ponder, but will report back soon.

Dave

Thanks Dave,
Your attention to the details is quite something to watch.

So, the EL34 screens are regulated at 85 volts. I never played with EL34s, but never would have thought any appreciable plate current would be drawn at 85 volts screen.

The screens are set at 85 volts below the plate voltage under quiescent conditions -- and then the EFB circuit controls them from that point under all conditions of power output.

Dave

audiodon said:

Thanks Dave,
Your attention to the details is quite something to watch.

Click to expand...

+1 wholeheartedly to that, Don!! You hit that one right on the noggin!!
Following Dave's projects are always interesting, and I think it's so cool of him to share his knowledge with us. I'm new at this so I'm studying hard trying to make sense of it all. I LOVE IT!!!

John

Progress Report

The past week has been put to good use not only finalizing the details of the EFB design for the X-1000, but also figuring out what the best way is to install it. As always, nothing is ever as easy as it seems, as both presented challanges.

Room on the underside of the chassis is a premium, while the under deck height really discourages any type of vertical mount board. Anything top side just didn't seem workable, so that meant that the modification had to be contained underneath the chassis, and hardwired without the aid of a circuit board.

To create the necessary room, some of it was gained by removing the old screen dropping resistors between the OPTs, but tie points were also going to be needed, so obviously more than that was required.

The original design of the X-1000 employed a necessarily complex dual mono type of B+ distribution system, used solely for the intent of helping to deal with the heavy amounts of screen current that EL34 tubes ultimately draw with elevated power output, and the higher than appropriate screen voltage employed under quiescent conditions due to the inevitable voltage drop that would occur across the screen resistors when power output was increased. Visually this amounted to a maze of resistors associated with sunk mounted filter cap over near the phono section. With EFB now taking over control of the output tube screen voltage, the complex distribution system for the remainder of the tubes was no longer needed. This presented opportunity.

Removing all of the old distribution system meant reworking finished work in the unit again, but resulted in not only a cleaner look for that system, but a system that is superior to the original: The new system contains more total capacitance than the original system did, and shows significantly less voltage fluctuation between quiescent and full power output conditions. Additionally, general isolation between the power amp and preamp/control sections is also improved, which is always a significant consideration when a common power supply powers both a high power power amplifier section, and matching preamp/control section.

That effort was made all the more interesting by some previous label maker happy individual who labeled the A,B,C&D sections of the sunk mounted can cap wrong. This required its removal to determine which terminal represented which capacitance within the can, and then if that capacitance was correct for the wiring originally connected to it. The bottom line was that the wiring was correct, but the labeling was wrong.

Ultimately then, space has been created between the OPTs, on the T-strip near the sunk mounted can cap, and on an originally used but now unused T-strip behind the volume control. These areas will primarily contain the control grid regulator element of the EFB modification. Additional components will be mounted in precious space over in the power supply supply area, and near the back panel for the EFB screen grid regulator element.

As for the design of EFB itself, the control grid regulator portion requires a bias source that is roughly double that of the original design to operate optimally. In the modified X-1000, some of this needed voltage is obtained simply by the fact that the output stage now operates with less grid bias voltage than the original design did. The rest has been obtained by piggy backing the bias supply onto the DC supply for the small signal tube heaters. Together, these two supplies produce about -95 vdc at the output of the raw bias supply, which is more than enough to let the EFB control grid regulator do its job. The design of the EFB screen grid regulator was rather straight forward, with more work involved in redesigning the new B+ distribution system than in developing the EFB screen grid regulator.

So, as it now stands, design of the complete EFB modification for the X-1000 has been finalized, all of the old B+ distribution system has been removed, the new B+ distribution system has been installed, and space has been made available to properly install all of the EFB circuits, and all without creating that crammed in look. In short, all of the hard work is done, and now the modification itself can simply be installed as planned. This project really is beginning to close in on home rather quickly now.

Pics include:

1. In progress of tearing out the old B+ distribution system. For those interested, look back at some of the older pics and you will see just how convoluted it was in this area.

2. With the old distribution system gone and the new one installed, it left a good amount of needed space between the OPTs,

3. And on the T-strip behind the sunk mounted can. The T-strip will now contain the inverter section of the EFB control grid regulator. Notice too how the new distribution system -- fully installed here -- cleaned up the nest of wiring associated with the old B+ distribution system and sunk mounted can cap.

4. This previously unused T-strip (unused with the new line amp/control section design) will now be placed back into service for the pass section portion of the EFB control grid regulator.

5. In this shot, one of the new B+ distribution system dropping resistors is now located with the main power supply B+ can caps, while at the rear most edge, you can see where the EFB screen grid regulator pass mosfet will be mounted.

There are some other small components to be added of course, but the planning phase of the EFB exercise has paid off very well in allowing for a neat installation, that should not necessarily draw attention to itself.

If all goes well, the next post should show the final finished product, with all modifications fully installed.

Dave

It's always interesting to see how different people address challenges in circuitry and component routing in these "form factor fitting" units designed to be small yet still with big audio goals.
The X-1000 is very Sherwoodesque in trying to cram a lot of stuff into a small space, yet still meet lofty performance goals.
That seems to come at a cost that makes some other (more uncluttered) products more attractive than they once were.
This kind of reminds me of some cars I used to work on where everything was packed tightly and you had to remove a lot of stuff to get to things like the clutch slave cylinder.

Back on topic:
I'm glad you're figuring out how to accomplish getting EFB in there and all I can say is more power to you (and the unit, I'm sure).

Getting Oh So Close!

Today's work addressed some details of the installation, but primarily, completed installation of the Screen Grid Regulator portion of the EFB modification, and tested it for proper operation. Everything passed with flying colors, so next up, will be completing the EFB installation by installing the Control Grid Regulator portion. It will likely be Wednesday before I can do that, so for now, pics include:

1. The majority of the EFB Screen Grid Regulator circuits are contained in the cleared out area between the OPTs, connecting between the T-strips located there. However,

2. Some of it is also carefully tucked away under the new power supply filter section that was added to get a handle on all the 120 Hz PS hum that was present.

3. The center piece of the EFB Screen Grid Regulator -- the power mosfet used to control the voltage to the screens -- is located here, near the back of the unit, and away from the heat of the output tubes. Under a worst case scenario, this device never dissipates over 2 watts of power, even though the screens are drawing about 32 watts of power through it when both channels are developing full power output. It is a 9 Amp, 500 Volt, 30 Watt plastic device, and with such an overrated device being used for the job, its long life should be guaranteed. It is the same device I used in my Fisher 400, my Fisher SA-100 clone, and in Warfcreek's Fisher SA-100 to date with no concerns.

4. An overall view of the complete unit as it now stands.

After finishing the installation on Wednesday, a full battery of tests will be performed to document the performance of the unit with the output stages operating under the full control of EFB, and then some listening tests can begin. This project is approaching the 90 day mark, but has covered an awful lot of ground in that time. While I'm ready to get this one wrapped up, I have a sneaking suspicion that ultimately, I'm going to be sorry to see it go!

Dave

Hey, Dave...since you are currently in an EFB frame of mind, what news on the Universal EFB board?

-D

Hi Derek -- The design has been finalized and vetted through many of these various projects documented here. Once this X-1000 is out the door, fellow AKer BuzzK has been a tremendous help in securing an avenue to rather quickly address the needs of the immediate future, while more permanent long term plans can be finalized.

Dave

Finished!

EFB has been installed in a number of different applications now, always significantly lowering distortion because of EFB action, while also correcting design issues in some cases (Dynaco SCA/ST-35, Late Model Fisher 400), providing fixed bias operation in others (my SA-100 clone), or simply making a given design work better (Fisher SA-100, Eico ST-70). In all cases, EFB operation provided safer operating conditions for the tubes, and allowed for lower quiescent current/plate dissipation levels to be had, extending tube life significantly. The Fisher X-1000 fell into the last category, with EFB allowing the original design to simply work better.

Because the vast majority of Fisher products all use pentode output stages, and because a simple, economical way of regulating the voltages to the output stages did not exist back in the day, it means that most Fisher products will benefit significantly from the installation of EFB. The X-1000 is no different.

In stock form, with each channel of the X-1000 adjusted for the specified quiescent current draw, each channel would deliver right at 50 watts RMS individually, or about 45 watts RMS with both channels driven. But 1 kHz THD was high at .7% for a single driven channel, and ~ 1.5% when both channels were driven (both slightly elevated because of the current inrush limiter installed). Also, because of the heavy current requirements of the output tube screen grids and the resulting way they were powered in the original design, full power operation caused the screen grids to mount a protest and glow in the dark. Granted, it was nothing like the glowing screen grids in the late model 400 design, but any color produced in the screen grids produces possible deformity and resulting change of tube characteristics. And of course, if they ever overheat and warp to the point of touching other elements inside the tube, then really bad stuff happens. Finally, there was also the weird clipping produced at 20 kHz. This is a phenomenon that is unique to EL34 tubes when the screen grids operate at too high a voltage. When that happens, they steal needed current flow away from the plate, which at that frequency and power level is trying to deal with the notable winding capacitance that large OPTs have at that frequency. With the added load of the capacitance, and the stolen current needed to deal with that increased load, the sides of the sine wave collapse in, causing the weird clipping shown in an earlier post of this thread. But upping the screen dropping resistor so that they don't receive so much voltage at full power only causes a larger total voltage drop to these grids when going from quiescent to full power conditions, which then sends distortion to the moon. So, Fisher really had to thread the needle in the X-1000 by overpowering the screens a bit to minimize gross distortion, allowing them to glow a bit in the dark, accepting some higher overall distortion numbers, and also accepting the weird clipping produced at 20 kHz in the process. With the large power reserve, it was all likely looked at as an acceptable compromise to achieve that power level in an integrated amplifier. The upshot of all of this is that while Fisher specified a quiescent current flow of 43 ma per tube in the stock design, the actual low distortion operating point of that design is achieved with a quiescent current flow of about 75 ma per tube! Simply put, in the stock design, this is how much current the tubes needed to pass under quiescent conditions, so that when the plate and screen voltages fall under full power conditions, the tubes will then be properly biased. Looked at from the other direction, when the original design is biased to the lowest distortion operating point under full power conditions, the resulting rise in plate and screen voltages with the return to quiescent conditions then cause the tubes to draw 75 ma each. This level exceeds the ratings of the tubes, causing massive amounts of heat to be dissipated from the unit, and does nothing to address the environment that the screen grids endured. By careful choice of screen dropping resistor and backing down the bias setting to 43 ma per tube, the compromise was had, but distortion rose substantially, as the measured performance shows. Enter EFB.

With EFB, the screen grid voltage can be locked at a precise value below that of the plate voltage -- low enough to keep the screens within their dissipation rating, but high enough to allow the tubes to draw enough current to match the load impedance offered by the OPT -- and then allows that voltage to fall proportionately as the plate voltage falls with increasing power output. For the control grids, the same exact action is mimicked, meaning that the tubes maintain their operating point, regardless of power supply fluctuations. This action is the definition of EFB action, and helps the mighty X-1000 significantly.

Where as full power 1 kHz distortion before was .7% (single) and 1.5% (both), with EFB and the quiescent current set to the low distortion point of 46 ma per tube, 1 kHz THD falls to .21% -- whether both channels are driven or not, it makes no difference. Also, because the screens are always operating at an optimum voltage, the weird clipping is gone, and the screens remain dark in the dead of night. Based on both channels operating, this is over a 7 fold drop in distortion, with only a very slight increase of 3 ma quiescent current per tube over the stock specification. This results in each tube dissipating 17 watts at the plate (+8% over the stock setting), but still only represents 68% of the tube's conservative Design Center rating of 25 watts.

I am setting the bias at 40 ma total per tube. This causes the distortion to rise to only .25% at full power (single or both driven), and reduces dissipation to just 14.5 watts, or <58% of the tube's rating. If you want to mimic the original Fisher specification of .5% distortion, you can achieve that with the output tubes now idling at a mere 28 ma per tube, for a plate dissipation of just 10 watts! Because EFB constantly maintains the output tubes at the operating point they are set at, distortion is now unaffected by whether a current inrush limiter is installed or not. The EFB circuit adjusts the output stage according to prevailing conditions just as it would if a central AC unit kicked on, it makes no difference.

As a durability test, the EFB modified X-1000 was set at full throttle in both channels for one hour uninterrupted. The output tubes were quite content during this time, showing no tendency towards uncivilized behavior -- this no doubt largely due to their screens now operating at an appropriate voltage for the condition. Distortion remained low throughout the period, and power output remained steady at 48.5 watts RMS per channel. The power transformer gave me a dirty look when it was all over, but was no worse for the wear, never exceeding 150 F. I would not consider running this test with the original output stage design, due to the high probability of output tube damage in the process.

So, the final effort has been finished with the X-1000. There are still little things to be done -- some tubes to test, a schematic to mark up, but the work is basically done. With that, I will provide some final installation pics, and then go back to the original X-1000 thread tomorrow for some final thoughts. Pics included:

1. The EFB Control Grid Regulator inverter transistor.

2. The EFB Control Grid pass transistor.

3. A final under the hood shot.

4. EL34's now under the control of EFB, all biased up at 40 ma, rarin' to go.

5. The X-1000 ready to take on whatever you can dish up to it.

More back on the other thread tomorrow.

Dave

This is fantastic news. I love the explanation about stock form vs. modified form, behavior and compromises made because of the prevailing mindset and technology of the day.

OUTSTANDING JOB, DAVE! I actually understood about 60%. Up from about 20% when you started the 400 thread.

Can't wait until you get the Universal boards out so I can do the 800-C's.
I'd rather put in a kit board than try and make it myself. I'm not too good at complex (to me at least) schematics and making something from it point to point. Yet!

Dave,
I was looking for this article. I inadvertently got sucked into the DUT tube testing thread (damn my ADD)
Often at the ends of these threads (well during too). You have finished schematics and before as well as after pics. When you last left this thread (9/13) you wrote more tomorrow. Did we miss out on the conclusions and the final designs?

This was an offshoot of another thread.
See post #122 and after in this thread for the rest of the story.
Hopefully it has what you're looking for.
http://www.audiokarma.org/forums/showthread.php?t=306741

Thanks Don,
For what its worth , I'm sure I read this before. It did seem strange at O dark thirty this morning the thread would have ended so abruptly. Thanks for the direction.

Reviving this older thread. Has anyone else than Dave tried to implement this mod into their X-1000s ? I would love a step-by-step tutorial to do this mod in mine.

I restored my X-1000 at a broadcast museum class a couple years back. I don't ever push it half volume, so, maybe don"t need EFB yet.
Still, would be very interested in some sort of kit package. Eliminating space expanded would be nice for now. Along with and other tweaks, Dave could suggest.

So now it's over 4 yrs, and finally I got around to bypassing the space expander, Low filter, High filter, and tape monitor.
Brave new minimalist approach. Actually, very cost effective. Higher Fidelity for low bucks.
Dumped a ton of grunge. Black background and getting blacker, with selective low microphonic tubes.
A certain 1954 backplate tube in the V4 tone position; cemented the soundstage and nailed tone.
Bias can now be set individually. Instead of two ganged prs, in the outputs.
Sorta in tone whore heaven, right now...except I am waiting on two .10
Caps, to complete the left channel.

hotroady said:

So now it's over 4 yrs, and finally I got around to bypassing the space expander, Low filter, High filter, and tape monitor.
Brave new minimalist approach. Actually, very cost effective. Higher Fidelity for low bucks.
Dumped a ton of grunge. Black background and getting blacker, with selective low microphonic tubes.
A certain 1954 backplate tube in the V4 tone position; cemented the soundstage and nailed tone.
Bias can now be set individually. Instead of two ganged prs, in the outputs.
Sorta in tone whore heaven, right now...except I am waiting on two .10
Caps, to complete the left channel.

Click to expand...

I would like to do both those mods. Do you have a link that explains exactly what to do.