As with most things, time causes change, and so it has now with the cathode regulator iteration of EFB™. This approach to applying EFB to an existing amplifier -- or even new projects -- has been successfully carried out many hundreds of times either through the use of available boards for the small Dynaco amplifiers, hardwired into many of the seemingly endless supply of Magnavox (and similar) amplifiers out there, or planned into new-build construction projects of all types. It's simple, highly effective, and very economical to build as well.
Over the course of the last 10+ years since EFB was introduced, I have been -- or have been made aware of a handful of instances where otherwise successful installations of the LM 237/337 regulator used in cathode based EFB applications failed for no apparent reason. No smoke, fireworks, or explosions -- the regulator simply shorted, to the great consternation of the output tubes, because then all bias applied to them was lost. The failure rate is small -- with best estimates well under 1% -- but still high enough for me to keep a close eye on these occurrences over the years, as I think this failure rate is still much higher than simple manufacturing defects or catastrophic tube failures can explain.
Not being privy to all the factors surrounding each failure, it has made it very difficult to pinpoint any commonality as to a possible cause in any kind of timely way. And of course, actually trying to induce failure to speed that process up has -- well, always failed as well. However, recent work has shed some new light on possible causes, of which I am now confident enough in the findings to release the results of that work, and present the resolves developed that can be done to address the issues involved. No doubt some explanations of the unusual circuit action will be helpful:
STRESS MODE #1
Occasionally, I have been asked why the customary reverse-voltage protection diodes were not added around the regulator used in the EFB design to help protect it. Why? Because they don't apply to the way the regulator is used in the EFB application. In other words, it is not possible for there to be a reverse-voltage condition as customarily thought of, which testing during the development work bore out. As a result, no diodes were used.
A further complication is that the regulator -- as used in providing EFB operation -- is being operated in a way that (apparently) the manufacturers of the regulators either never considered, or at least never intended anyway. That is, all 3 terminal (3T) adjustable regulators are designed to regulate by seeking to produce a constant voltage between their OUT and ADJ terminals, as referenced against an internal voltage standard within the regulator. This is accomplished by using an external resistor voltage dividing network connected between the OUT terminal, the ADJ terminal, and ground. Because this network is effectively connected across the output of the regulator then, any change in output voltage causes the voltage through the dividing network to change the voltage appearing between the OUT and ADJ terminals, which then causes the regulator to act accordingly against its internal voltage standard, so as to correct the output voltage back to its original level. In this way, the output voltage becomes highly regulated in a very simple and economic way, to an exact value needed. This is the way virtually all 3T adjustable regulators operate and are classically shown to be installed in all the application sheets, with power supply caps invariably connected to (usually) each terminal in practical circuits, and protection diodes then often added to protect the regulator against the possible reverse-voltage conditions the caps can potentially create in this type of installation at shut down. Except....... that's not how the regulator is being used with EFB.
In the classic (and intended) design applications, 3T adjustable regulators ultimately act to maintain a constant output voltage to their load by varying their internal conductivity between the IN and OUT terminals to achieve that end -- as based on the changing conditions of the load the regulator is powering. In the EFB application however, the regulator is actually seeking to maintain a constant voltage between the IN and OUT terminals, by once again, varying its internal conductivity between these two terminals to achieve that end -- based on the changing conditions of the current passing through the regulator. In other words, the intended application has the regulator achieving a constant voltage at the output of the regulator, while the EFB application has the regulator achieving a constant voltage across the regulator -- an application the designers had (apparently) not envisioned.
To produce EFB operation, the voltage applied to the ADJ terminal is now no longer sourced from a voltage divider connected to the OUT terminal, but from a source of B+ voltage that ultimately causes the regulator to operate in a sort of "Transistor Mode" -- which represents a significant deviation from the intended use. This is an important distinction to recognize, made all the more important by the one fact that did slowly become evident over time: the regulator failures that did occur were all primarily associated with designs using SS B+ power supplies.
Continuing on, the internal circuits of the regulator operate from the voltage drop created between the IN and OUT terminals when current flows through the regulator. Therefore, a minimum voltage drop is always specified. The customary voltage divider connected between the OUT, ADJ, and ground nodes is designed to pass enough minimum current (usually about 5 mA) to properly reference the ADJ terminal at the intended output voltage, as set by the ratio of the divider resistors used. Then, by ensuring that the overall power supply design provides at least the minimum IN/OUT voltage drop across the regulator (so its internal circuits will operate properly), the divider network's current draw will then ensure a constant output voltage from the regulator -- even if no other load is connected to its output. But what happens if that's not always the case? A worst case example for the regulator in the EFB application is when there is a SS power supply employed (providing instant B+ voltage), coupled with output tubes that have not yet warmed up to conduct any current.
With the scenario just described, there is a period of time then when there is now no current flowing through the regulator, because there is no voltage divider on the OUT terminal to create any draw in this application, and the output tubes have not yet warmed up enough to conduct any current, either. With no current flow, there is therefore no voltage drop produced across the regulator during this time for the internal circuits to operate from. A fly in the ointment of this condition occurs because also during this time, there is a significant voltage applied between the IN and ADJ terminals (thanks to the SS power supply) which in tests I conducted, showed the regulator wasn't particularly fond of -- not outright revolting against mind you (i.e., blowing up), but creating a condition of stress none the less. Specifically, the tests showed that during this time, there was nearly a 6X increase in the voltage appearing between the ADJ and OUT terminals, versus the typical 1.25 vdc that is present during normal operation. This causes a significant increase in the current flowing through the ADJ terminal during this time, that can potentially damage the internal voltage reference within the regulator. If that happens, the regulator will move to a full conduction mode, effectively creating a shorted regulator -- whether it actually is, or not.
To validate my thoughts, I contacted George Ronnenkamp of Audio Regenesis, who took the data I generated, and could not even come close to replicating any of it with any available computer models of the LM337 -- all of which reinforced the very unique way in which the regulator is used in the EFB application (i.e., no known modeling data exists for that configuration). But in physically recreating the conditions, he was able to replicate my findings certainly closely enough to validate them -- but also to suggest that each regulator example likely reacts somewhat differently to the specific conditions of this discussion. George then connected his digital storage scope (I only have analog scopes) to the breadboard mock-up of the EFB circuit he was using, which additionally showed that a rather sharp transient was also taking place at the ADJ terminal at start up, which is on top of the abnormal steady state conditions I had already noted. While by strict definition this is not a smoking gun, the research strongly suggests that it is this 1-2 punch of a strong turn on transient spike followed by a period of steady state abnormal voltage conditions at the ADJ terminal -- all before the output tubes warm up -- coupled with each regulator example's long term ability to deal -- or not -- with these unique conditions, that is the root cause of those few regulators that mysteriously gave up the ghost in SS powered designs. To be clear, the vast majority of regulator examples have readily shown themselves as being capable of handling these events in stride -- but a few have not and so, the quest to determine why -- and then bullet proof the design against those events.
Further study however indicated that there's yet one more stress condition that can be presented to the EFB regulator -- present only in slow warming rectifier tube designs -- that needs to be addressed as well. And this one actually does involve reverse polarity voltages.
STRESS MODE #2
In this 2nd scenario, the output tubes become able to conduct current before any B+ voltage is present because often, slow warm rectifier tubes only begin conducting current well after typical output tubes have become capable to. This then potentially sets up an entirely different stress condition. The source of the reverse polarity voltage is leakage within the power transformer, that can cause the output tube heater winding to become elevated above ground level -- if the winding does not directly reference ground via its own (or a faux created) CT. This is in fact exactly the case in so many designs, where the heater winding is not directly grounded, but returned to a positive source provided within the design to help reduce noise from and stress on the small signal tubes used -- a source only available once the rectifier tube becomes engaged. Until it does however, the heater winding can then float to whatever level the leakage within the power transformer so dictates. When the output tubes initially become active, the heater/cathode interface within them can actually rectify this leakage voltage so that a positive DC potential appears at the cathode terminal of the output tubes (relative to ground) that is then presented to the OUT terminal of the regulator until the rectifier tube becomes active. For the OUT terminal itself, this is not a problem. But with no B+ yet available, the ADJ terminal is referencing ground during this time which is a problem, because the ADJ terminal should always reflect a more positive potential than the OUT terminal does with the 237/337 series of regulators. In fact then, a condition of reverse polarity voltage can be presented to the EFB regulator under the unique conditions described. This event too will be checked to further bulletproof the design.
RESOLVED
I wanted to detail the issues involved to give understanding to their uniquety, the lack of any manufacturer's data to go on, and the amount of time its therefore taken to determine a resolve: I've simply been working with crumbs of information. Thankfully however, the actual resolve to both problems was found to be quite simple, and does involve the installation of four protection diodes -- three of which however are not applied in the usual manner.
Resolving the 2nd stress mode was easily enough, adding a single traditional back diode between the OUT and ADJ terminals, with the Cathode connected to the ADJ terminal as is customary for any such diode strapped to a Negative 3T Adjustable regulator. I recommend using 1N4148 small signal diodes as they tend to have somewhat less voltage drop than larger pieces, and their small size makes their installation easy enough to perform. Really however, any small signal diode will do the job, as the current and voltages involved are quite small.
Resolving the 1st stress mode however requires the diodes to be installed rather differently, because here the problem is not one of reverse polarity, but of an over-voltage condition. The normal operating voltage between the ADJ and OUT terminals is ~ 1.20-1.25 vdc, with the exact value being dependent on the manufacturing tolerances of the particular regulator piece used. If two small signal diodes were connected in series and then strapped across the ADJ and OUT terminals -- opposite in polarity to the reverse polarity diode, then some regulators might work, but many might not because the voltage drop across the two diodes is virtually the same as the normal voltage drop produced by the regulator itself. The use of just two diodes in series then could interfere with normal regulator operation.
To guard against that from happening, a string of three diodes is used, all connected in series, connected across the two regulator terminals as described. Now the diodes prevent the over-voltage condition from reaching beyond about 1.80 vdc (total), but are completely disengaged during normal operation, since the normal drop of the three diodes well exceeds the normal operating voltage produced by any example of these regulators between these two terminals.
The use of these four diodes then has completely resolved both issues described, and I heartily recommend their installation to ensure trouble free operation of the circuit regardless of what type of power supply an amplifier uses -- or example of regulator piece chosen. I have provided some pics showing the construction of the diode array, and its installation in my modified 9300 Magnavox amplifier. Also included are the traces that George captured showing how the diodes completely eliminate the issues detailed in the 1st stress mode.
Time will of course tell, but I have every confidence that the diode array will resolve the mystery failures once and for all.
Dave
The diode array constructed, before the heat shrink is applied:
Installed in the Magnavox test amplifier. By pulling the output tubes, I could test the conditions outlined in stress mode #1, and by leaving them installed but pulling the rectifier tube, I could test the conditions outlined in stress mode #2.
In each scope shot, the upper trace is the ADJ terminal, while the lower trace is the OUT terminal display -- but these are not separately generated displays: They are both generated together in real time with the amplitude of each display properly relevant to each other when generated. It can readily been seen then how the 3 diode string completely eliminates the sharp transient at turn-on, greatly minimizes the difference potential between the ADJ and OUT terminals, and creates plenty of voltage drop across the regulator for the internal circuits to operate properly -- during the period where there is no output tube current draw to produce normal regulator operation. When the tubes begin conducting current, the diodes disengage so that they effectively disappear in the circuit.
Again, I want to thank George Ronnenkamp for his time and effort in validating my original findings, his expert analysis, and permission to use the displays he generated and captured for this work.
Over the course of the last 10+ years since EFB was introduced, I have been -- or have been made aware of a handful of instances where otherwise successful installations of the LM 237/337 regulator used in cathode based EFB applications failed for no apparent reason. No smoke, fireworks, or explosions -- the regulator simply shorted, to the great consternation of the output tubes, because then all bias applied to them was lost. The failure rate is small -- with best estimates well under 1% -- but still high enough for me to keep a close eye on these occurrences over the years, as I think this failure rate is still much higher than simple manufacturing defects or catastrophic tube failures can explain.
Not being privy to all the factors surrounding each failure, it has made it very difficult to pinpoint any commonality as to a possible cause in any kind of timely way. And of course, actually trying to induce failure to speed that process up has -- well, always failed as well. However, recent work has shed some new light on possible causes, of which I am now confident enough in the findings to release the results of that work, and present the resolves developed that can be done to address the issues involved. No doubt some explanations of the unusual circuit action will be helpful:
STRESS MODE #1
Occasionally, I have been asked why the customary reverse-voltage protection diodes were not added around the regulator used in the EFB design to help protect it. Why? Because they don't apply to the way the regulator is used in the EFB application. In other words, it is not possible for there to be a reverse-voltage condition as customarily thought of, which testing during the development work bore out. As a result, no diodes were used.
A further complication is that the regulator -- as used in providing EFB operation -- is being operated in a way that (apparently) the manufacturers of the regulators either never considered, or at least never intended anyway. That is, all 3 terminal (3T) adjustable regulators are designed to regulate by seeking to produce a constant voltage between their OUT and ADJ terminals, as referenced against an internal voltage standard within the regulator. This is accomplished by using an external resistor voltage dividing network connected between the OUT terminal, the ADJ terminal, and ground. Because this network is effectively connected across the output of the regulator then, any change in output voltage causes the voltage through the dividing network to change the voltage appearing between the OUT and ADJ terminals, which then causes the regulator to act accordingly against its internal voltage standard, so as to correct the output voltage back to its original level. In this way, the output voltage becomes highly regulated in a very simple and economic way, to an exact value needed. This is the way virtually all 3T adjustable regulators operate and are classically shown to be installed in all the application sheets, with power supply caps invariably connected to (usually) each terminal in practical circuits, and protection diodes then often added to protect the regulator against the possible reverse-voltage conditions the caps can potentially create in this type of installation at shut down. Except....... that's not how the regulator is being used with EFB.
In the classic (and intended) design applications, 3T adjustable regulators ultimately act to maintain a constant output voltage to their load by varying their internal conductivity between the IN and OUT terminals to achieve that end -- as based on the changing conditions of the load the regulator is powering. In the EFB application however, the regulator is actually seeking to maintain a constant voltage between the IN and OUT terminals, by once again, varying its internal conductivity between these two terminals to achieve that end -- based on the changing conditions of the current passing through the regulator. In other words, the intended application has the regulator achieving a constant voltage at the output of the regulator, while the EFB application has the regulator achieving a constant voltage across the regulator -- an application the designers had (apparently) not envisioned.
To produce EFB operation, the voltage applied to the ADJ terminal is now no longer sourced from a voltage divider connected to the OUT terminal, but from a source of B+ voltage that ultimately causes the regulator to operate in a sort of "Transistor Mode" -- which represents a significant deviation from the intended use. This is an important distinction to recognize, made all the more important by the one fact that did slowly become evident over time: the regulator failures that did occur were all primarily associated with designs using SS B+ power supplies.
Continuing on, the internal circuits of the regulator operate from the voltage drop created between the IN and OUT terminals when current flows through the regulator. Therefore, a minimum voltage drop is always specified. The customary voltage divider connected between the OUT, ADJ, and ground nodes is designed to pass enough minimum current (usually about 5 mA) to properly reference the ADJ terminal at the intended output voltage, as set by the ratio of the divider resistors used. Then, by ensuring that the overall power supply design provides at least the minimum IN/OUT voltage drop across the regulator (so its internal circuits will operate properly), the divider network's current draw will then ensure a constant output voltage from the regulator -- even if no other load is connected to its output. But what happens if that's not always the case? A worst case example for the regulator in the EFB application is when there is a SS power supply employed (providing instant B+ voltage), coupled with output tubes that have not yet warmed up to conduct any current.
With the scenario just described, there is a period of time then when there is now no current flowing through the regulator, because there is no voltage divider on the OUT terminal to create any draw in this application, and the output tubes have not yet warmed up enough to conduct any current, either. With no current flow, there is therefore no voltage drop produced across the regulator during this time for the internal circuits to operate from. A fly in the ointment of this condition occurs because also during this time, there is a significant voltage applied between the IN and ADJ terminals (thanks to the SS power supply) which in tests I conducted, showed the regulator wasn't particularly fond of -- not outright revolting against mind you (i.e., blowing up), but creating a condition of stress none the less. Specifically, the tests showed that during this time, there was nearly a 6X increase in the voltage appearing between the ADJ and OUT terminals, versus the typical 1.25 vdc that is present during normal operation. This causes a significant increase in the current flowing through the ADJ terminal during this time, that can potentially damage the internal voltage reference within the regulator. If that happens, the regulator will move to a full conduction mode, effectively creating a shorted regulator -- whether it actually is, or not.
To validate my thoughts, I contacted George Ronnenkamp of Audio Regenesis, who took the data I generated, and could not even come close to replicating any of it with any available computer models of the LM337 -- all of which reinforced the very unique way in which the regulator is used in the EFB application (i.e., no known modeling data exists for that configuration). But in physically recreating the conditions, he was able to replicate my findings certainly closely enough to validate them -- but also to suggest that each regulator example likely reacts somewhat differently to the specific conditions of this discussion. George then connected his digital storage scope (I only have analog scopes) to the breadboard mock-up of the EFB circuit he was using, which additionally showed that a rather sharp transient was also taking place at the ADJ terminal at start up, which is on top of the abnormal steady state conditions I had already noted. While by strict definition this is not a smoking gun, the research strongly suggests that it is this 1-2 punch of a strong turn on transient spike followed by a period of steady state abnormal voltage conditions at the ADJ terminal -- all before the output tubes warm up -- coupled with each regulator example's long term ability to deal -- or not -- with these unique conditions, that is the root cause of those few regulators that mysteriously gave up the ghost in SS powered designs. To be clear, the vast majority of regulator examples have readily shown themselves as being capable of handling these events in stride -- but a few have not and so, the quest to determine why -- and then bullet proof the design against those events.
Further study however indicated that there's yet one more stress condition that can be presented to the EFB regulator -- present only in slow warming rectifier tube designs -- that needs to be addressed as well. And this one actually does involve reverse polarity voltages.
STRESS MODE #2
In this 2nd scenario, the output tubes become able to conduct current before any B+ voltage is present because often, slow warm rectifier tubes only begin conducting current well after typical output tubes have become capable to. This then potentially sets up an entirely different stress condition. The source of the reverse polarity voltage is leakage within the power transformer, that can cause the output tube heater winding to become elevated above ground level -- if the winding does not directly reference ground via its own (or a faux created) CT. This is in fact exactly the case in so many designs, where the heater winding is not directly grounded, but returned to a positive source provided within the design to help reduce noise from and stress on the small signal tubes used -- a source only available once the rectifier tube becomes engaged. Until it does however, the heater winding can then float to whatever level the leakage within the power transformer so dictates. When the output tubes initially become active, the heater/cathode interface within them can actually rectify this leakage voltage so that a positive DC potential appears at the cathode terminal of the output tubes (relative to ground) that is then presented to the OUT terminal of the regulator until the rectifier tube becomes active. For the OUT terminal itself, this is not a problem. But with no B+ yet available, the ADJ terminal is referencing ground during this time which is a problem, because the ADJ terminal should always reflect a more positive potential than the OUT terminal does with the 237/337 series of regulators. In fact then, a condition of reverse polarity voltage can be presented to the EFB regulator under the unique conditions described. This event too will be checked to further bulletproof the design.
RESOLVED
I wanted to detail the issues involved to give understanding to their uniquety, the lack of any manufacturer's data to go on, and the amount of time its therefore taken to determine a resolve: I've simply been working with crumbs of information. Thankfully however, the actual resolve to both problems was found to be quite simple, and does involve the installation of four protection diodes -- three of which however are not applied in the usual manner.
Resolving the 2nd stress mode was easily enough, adding a single traditional back diode between the OUT and ADJ terminals, with the Cathode connected to the ADJ terminal as is customary for any such diode strapped to a Negative 3T Adjustable regulator. I recommend using 1N4148 small signal diodes as they tend to have somewhat less voltage drop than larger pieces, and their small size makes their installation easy enough to perform. Really however, any small signal diode will do the job, as the current and voltages involved are quite small.
Resolving the 1st stress mode however requires the diodes to be installed rather differently, because here the problem is not one of reverse polarity, but of an over-voltage condition. The normal operating voltage between the ADJ and OUT terminals is ~ 1.20-1.25 vdc, with the exact value being dependent on the manufacturing tolerances of the particular regulator piece used. If two small signal diodes were connected in series and then strapped across the ADJ and OUT terminals -- opposite in polarity to the reverse polarity diode, then some regulators might work, but many might not because the voltage drop across the two diodes is virtually the same as the normal voltage drop produced by the regulator itself. The use of just two diodes in series then could interfere with normal regulator operation.
To guard against that from happening, a string of three diodes is used, all connected in series, connected across the two regulator terminals as described. Now the diodes prevent the over-voltage condition from reaching beyond about 1.80 vdc (total), but are completely disengaged during normal operation, since the normal drop of the three diodes well exceeds the normal operating voltage produced by any example of these regulators between these two terminals.
The use of these four diodes then has completely resolved both issues described, and I heartily recommend their installation to ensure trouble free operation of the circuit regardless of what type of power supply an amplifier uses -- or example of regulator piece chosen. I have provided some pics showing the construction of the diode array, and its installation in my modified 9300 Magnavox amplifier. Also included are the traces that George captured showing how the diodes completely eliminate the issues detailed in the 1st stress mode.
Time will of course tell, but I have every confidence that the diode array will resolve the mystery failures once and for all.
Dave
The diode array constructed, before the heat shrink is applied:
Installed in the Magnavox test amplifier. By pulling the output tubes, I could test the conditions outlined in stress mode #1, and by leaving them installed but pulling the rectifier tube, I could test the conditions outlined in stress mode #2.
In each scope shot, the upper trace is the ADJ terminal, while the lower trace is the OUT terminal display -- but these are not separately generated displays: They are both generated together in real time with the amplitude of each display properly relevant to each other when generated. It can readily been seen then how the 3 diode string completely eliminates the sharp transient at turn-on, greatly minimizes the difference potential between the ADJ and OUT terminals, and creates plenty of voltage drop across the regulator for the internal circuits to operate properly -- during the period where there is no output tube current draw to produce normal regulator operation. When the tubes begin conducting current, the diodes disengage so that they effectively disappear in the circuit.
Again, I want to thank George Ronnenkamp for his time and effort in validating my original findings, his expert analysis, and permission to use the displays he generated and captured for this work.
Last edited:
