Kid -- could you send that to me? I don't absolutely need it, but it would be nice to have for the record on this end.
Dave
Dave
More Fun With Magnavox: The 9300 Series
- Thread starter dcgillespie
- Start date
Done
Actually, a 5V4GA would probably be a better choice. 350v is a bit much. Last "bone stock" 93xx I had here, ran about 330 to 335v on the B+, with 120v from the wall. With a bucking winding and a 5V4, I'd bet you'd be in that range.
Regards,
Gordon.
Magnavox made consoles with different tuner/preamps, with and without a gain stage, and with and without an FM multiplexer. The multiplexer may have had some residual 19 and/or 38 KHz output, which wouldn't play well with the ultrasonic peak of the version Dave measured. So there were multiple versions with different gain and feedback compensation.
ferninando
Lunatic Member
Indeed.
My 9303 is just waiting to be hacked on when Dave is all finished.
My 9303 is just waiting to be hacked on when Dave is all finished.
IMPROVEMENTS TO THE 9300 -- PT 1
Before investigating to see if there was any potential for improvement in the 9300 series, I checked the build of my unit against the schematic and parts chart that Kidmoe sent me. This showed that units with the model number of my unit here were supposed to include C105 and C205, which are 100 pF caps from the plate of the first AF amplifier stage to ground. However, they were nowhere to be found in my unit. Tacking some appropriate value caps in place to check what their contribution might be found the HF response peak noted earlier tamed slightly (but hardly to any significant degree), and stability improved slightly as well. Therefore, rather than serving any specific function, these components seemed to be insurance policies in case wiring layout and component tolerance didn't get the job done. In any event, it does not appear that my unit ever included these components.
Studying the design then, it appeared that a number of opportunities presented themselves. They are as follows:
1. POWER SUPPLY: Two opportunities are available, with one already having been discussed -- that being using the extra heater winding to buck the AC line. With this modification in place, the voltage applied to the heaters is now quite reasonable (6.4 vac), and B+, as applied to the output stage, is about 8 volts high.
Under these conditions, the output tubes draw a collective 130 ma, for 32.5 ma per tube total. Backing out about 3 ma for screen current, that means that each plate is dissipating about 29.5 ma of current, with ~ 291vdc dropped across the plate and cathode elements, for a plate dissipation of ~ 8.6 watts per tube. This is great for tube life, but really puts a damper on the performance of this amplifier: As maximum power is approached, the sides of the sine wave begin to cave and show a classic notch due to the (once again) severe over-bias condition that exists (just like in the 8600 series) -- before actual clipping even commences. This represents a particularly nasty form of distortion that is anything but pleasant.
With the output tube dissipation running just over 70% of the Design Center rating then, this left some room to see if increasing the quiescent current would help alleviate this condition -- and the easiest way to do that, was simply to replace the high drop 5U4 rectifier tube, with a low drop GZ34 type. Besides, it would eliminate the voltage surge at turn on due to its slow heat characteristics. This move increased power output from just over 10 watts RMS (both channels driven), to just over 11 watts -- with the notch distortion still very present before full power is reached. That meant that the next stop would be in the power amplifier section, but for now it was decided to leave the GZ34 in place because of its slow heat capability, and because the B+ it provided to the output tubes only rose to ~ 340 vdc. At this level, plate dissipation rose to 10.2 watts -- still quite safe, with the power transformer benefiting from the reduced rectifier heater current draw.
2. OUTPUT STAGE: It was determined that increasing quiescent current draw could ultimately eliminate the notch, but by the time it did, the tubes were well into exceeding their plate dissipation rating, which also made things hard on the power supply section as well. At issue is the very same issue I dealt with in the small Dynaco amplifiers some years ago:
A. To achieve maximum practical power output, the output tubes are loaded with an OPT providing a 7600 ohm load -- a load that causes the tubes to operate in the B portion of Class AB much of the time.
B. Class B operation produces significant change in current draw with the application of signal.
C. For reasons of economy, cathode bias is used.
D. Cathode bias and Class B operation do not mix well, because of the major shifts in bias voltage this type of operation creates when using cathode bias.
E. As a result -- and NOT because they were just cheap -- the cathodes of all four tubes were strapped together so that bias was provided by a single resistor and bypass capacitor.
The idea of this last point is that with the current demands of both channels rarely being equal at any point in time under dynamic conditions, being strapped together, they would therefore act to help regulate the bias voltage between themselves better when operating in Class B under dynamic conditions. In fact, this approach has merit. As with the small Dynaco amplifiers, if you drive only ONE channel of the 9300, it is capable of developing about 15 watts RMS, with no notch distortion in sight. That's because the other un-driven channel is acting to regulate not only the bias voltage, but the B+ voltage of the power supply as well. But who listens to only one channel? That's kinda hard to do with stereo. But drive both channels together, and power output then falls by one third in each channel to the previous value stated (just over 10 watts per channel), and distortion at the higher power levels just becomes gross -- this because now both channels are drawing current through the common bias resistor, creating the over-bias condition once again as previously discussed. Further, those well meaning techs that think they are improving the amplifier by using separate bias resistors for each channel -- or even each tube(!) -- are really sending performance down the crapper, as then each tube will ALWAYS be over-biased period, regardless of whether one or both channels are driven.
END PART 1
Before investigating to see if there was any potential for improvement in the 9300 series, I checked the build of my unit against the schematic and parts chart that Kidmoe sent me. This showed that units with the model number of my unit here were supposed to include C105 and C205, which are 100 pF caps from the plate of the first AF amplifier stage to ground. However, they were nowhere to be found in my unit. Tacking some appropriate value caps in place to check what their contribution might be found the HF response peak noted earlier tamed slightly (but hardly to any significant degree), and stability improved slightly as well. Therefore, rather than serving any specific function, these components seemed to be insurance policies in case wiring layout and component tolerance didn't get the job done. In any event, it does not appear that my unit ever included these components.
Studying the design then, it appeared that a number of opportunities presented themselves. They are as follows:
1. POWER SUPPLY: Two opportunities are available, with one already having been discussed -- that being using the extra heater winding to buck the AC line. With this modification in place, the voltage applied to the heaters is now quite reasonable (6.4 vac), and B+, as applied to the output stage, is about 8 volts high.
Under these conditions, the output tubes draw a collective 130 ma, for 32.5 ma per tube total. Backing out about 3 ma for screen current, that means that each plate is dissipating about 29.5 ma of current, with ~ 291vdc dropped across the plate and cathode elements, for a plate dissipation of ~ 8.6 watts per tube. This is great for tube life, but really puts a damper on the performance of this amplifier: As maximum power is approached, the sides of the sine wave begin to cave and show a classic notch due to the (once again) severe over-bias condition that exists (just like in the 8600 series) -- before actual clipping even commences. This represents a particularly nasty form of distortion that is anything but pleasant.
With the output tube dissipation running just over 70% of the Design Center rating then, this left some room to see if increasing the quiescent current would help alleviate this condition -- and the easiest way to do that, was simply to replace the high drop 5U4 rectifier tube, with a low drop GZ34 type. Besides, it would eliminate the voltage surge at turn on due to its slow heat characteristics. This move increased power output from just over 10 watts RMS (both channels driven), to just over 11 watts -- with the notch distortion still very present before full power is reached. That meant that the next stop would be in the power amplifier section, but for now it was decided to leave the GZ34 in place because of its slow heat capability, and because the B+ it provided to the output tubes only rose to ~ 340 vdc. At this level, plate dissipation rose to 10.2 watts -- still quite safe, with the power transformer benefiting from the reduced rectifier heater current draw.
2. OUTPUT STAGE: It was determined that increasing quiescent current draw could ultimately eliminate the notch, but by the time it did, the tubes were well into exceeding their plate dissipation rating, which also made things hard on the power supply section as well. At issue is the very same issue I dealt with in the small Dynaco amplifiers some years ago:
A. To achieve maximum practical power output, the output tubes are loaded with an OPT providing a 7600 ohm load -- a load that causes the tubes to operate in the B portion of Class AB much of the time.
B. Class B operation produces significant change in current draw with the application of signal.
C. For reasons of economy, cathode bias is used.
D. Cathode bias and Class B operation do not mix well, because of the major shifts in bias voltage this type of operation creates when using cathode bias.
E. As a result -- and NOT because they were just cheap -- the cathodes of all four tubes were strapped together so that bias was provided by a single resistor and bypass capacitor.
The idea of this last point is that with the current demands of both channels rarely being equal at any point in time under dynamic conditions, being strapped together, they would therefore act to help regulate the bias voltage between themselves better when operating in Class B under dynamic conditions. In fact, this approach has merit. As with the small Dynaco amplifiers, if you drive only ONE channel of the 9300, it is capable of developing about 15 watts RMS, with no notch distortion in sight. That's because the other un-driven channel is acting to regulate not only the bias voltage, but the B+ voltage of the power supply as well. But who listens to only one channel? That's kinda hard to do with stereo. But drive both channels together, and power output then falls by one third in each channel to the previous value stated (just over 10 watts per channel), and distortion at the higher power levels just becomes gross -- this because now both channels are drawing current through the common bias resistor, creating the over-bias condition once again as previously discussed. Further, those well meaning techs that think they are improving the amplifier by using separate bias resistors for each channel -- or even each tube(!) -- are really sending performance down the crapper, as then each tube will ALWAYS be over-biased period, regardless of whether one or both channels are driven.
END PART 1
Attachments
PART 2
It was this very scenario that led to the development of Enhanced Fixed Bias, or EFB(tm) some years ago, and it is just as effective with in the Magnavox chassis as it is with the small Dynaco amplifiers on which it was developed. A quick review of EFB operation might be helpful to those not familiar with it. Using EFB:
A. The tubes now operate under true fixed bias conditions, where the bias voltage applied to the tubes is independent of current flow through the tubes. As a result, bias voltage does not creep with the application of signal (hence, use of the term "fixed"), so an over-bias condition cannot occur.
B. But beyond simply creating fixed bias conditions however, once the quiescent bias is set to its optimum value, EFB then acts to maintain that optimum value, in spite of changing power supply voltage levels due to either changing AC line voltage conditions, or changing power output (i.e, current draw) demands from the amplifiers. This is far superior to traditional fixed bias, where changing plate and screen DC voltage levels cause the optimum bias setting to vary all over the place under changing dynamic conditions. It is also far superior to traditional screen grid regulation, where regulation of the screen is good, but if it is the only element regulated, while all the other element's DC levels are still varying, then its effectiveness is limited. It is the "Enhanced" portion of EFB that seeks to keep the control grid (and screen grid when necessary) in a relative lockstep relationship with the plate supply voltage, regardless of how it might vary due to changing AC line or dynamic conditions. And, it is very simple to implement as compared to providing tightly regulated power supplies for each element -- and it produces the same low distortion level as the latter approach achieves as well. It is really a win-win situation, with no down side.
Applying EFB to the output stages of the 9300 then produces the following results, based on both channels being driven:
A. Power output in both channels is a sustained 15.5 watts RMS, representing over a 150% increase from that produced by the original design when both channels are driven. When only a single channel is driven, power output reaches 17.0 watts RMS -- this due to less drop in power supply voltages in this scenario.
B. Midband (1 kHz) THD easily drops to under 1% within 1 db of 15.5 watts power output (typically about .70%). In the stock design, notch distortion is just commencing at 1 db below the maximum power output of 10.24 watts, so distortion is significant (~ 2.8%). This means that the modified amplifier is producing less distortion on more power, than the original design is even capable of producing in this scenario. Forgetting the increased power for a moment then, distortion at the 1 db down points has been reduced some 75%.
C. Under quiescent conditions, each output tube idles at a phenomenally low total current draw of just 22 ma (44 ma per channel). Each tube now dissipates just 7.1 watts at the plate -- even less than in the original design -- which amounts to operating at 59.2% of the conservative Design Center rating for the tube (12 watts). This translates to a very long tube life indeed.
D. The operating temperature of the amplifier is notably reduced, particularly that of the power transformer. In combination with the buck connection supplying the correct primary voltage, the reduced heater current draw from using a GZ34 rectifier tube, and now the reduced quiescent current draw due to operating the output stages under the control of EFB, the power transformer reached a low 114F after 2+ hours of operation in a 68F environment. This too speaks to component life in a big way.
Finally, installing EFB of course caused all traces of the notch distortion that appeared in the original design to completely disappear as well.
A few pics are provided, but do remember that this is a development mule........
View attachment 648902
1. The entire EFB circuit easily mounts in the upper right section underneath the chassis. Also visible are the two 10 ohm common cathode resistors for each channel near the bottom center of the pic. Because the tab of the LM337 operates at ground level, no insulation kit is required, although a little silicone grease helps ensure a good thermal contact with the chassis. Under a worst case scenario (continuous full power output in both channels), the EFB regulator dissipates approximately 2.5 watts, which certainly works to ensure its own long life as well.
View attachment 648905
2. On the top side, the new bias control occupies the place of the old hum balance control. Down the road, the three terminals which are currently the speaker output connections will serve as the bias test points for each channel. Three terminals are required since the common connection in this application is not at ground level. Therefore, one terminal would be required to represent the meter common connection, while the other two would represent the test point for each channel. That way, the bias can be adjusted without having to turn the amplifier upside down to do it whenever the tubes are replaced.
View attachment 648909
3. The two images are not calibrated the same for vertical gain so as to be able to show the complete waveforms. On the bottom, the stock design is operating in one channel (using a 200 ohm cathode resistor and bypass cap), and has not even reached the full power point where clipping commences (10.24 watts) and yet, notch distortion is clearly already evident. The top waveform is the other channel with the output stage operating under the control of EFB, developing over 16 watts RMS, as the onset of clipping is approached. Note that besides the increased power, the notch distortion is gone, yet the output tubes idle at an even lower level of quiescent current.
With the power supply and output stages addressed, the next installment will deal with the phase inverter, NFB, stability, and the output transformer.
Dave
It was this very scenario that led to the development of Enhanced Fixed Bias, or EFB(tm) some years ago, and it is just as effective with in the Magnavox chassis as it is with the small Dynaco amplifiers on which it was developed. A quick review of EFB operation might be helpful to those not familiar with it. Using EFB:
A. The tubes now operate under true fixed bias conditions, where the bias voltage applied to the tubes is independent of current flow through the tubes. As a result, bias voltage does not creep with the application of signal (hence, use of the term "fixed"), so an over-bias condition cannot occur.
B. But beyond simply creating fixed bias conditions however, once the quiescent bias is set to its optimum value, EFB then acts to maintain that optimum value, in spite of changing power supply voltage levels due to either changing AC line voltage conditions, or changing power output (i.e, current draw) demands from the amplifiers. This is far superior to traditional fixed bias, where changing plate and screen DC voltage levels cause the optimum bias setting to vary all over the place under changing dynamic conditions. It is also far superior to traditional screen grid regulation, where regulation of the screen is good, but if it is the only element regulated, while all the other element's DC levels are still varying, then its effectiveness is limited. It is the "Enhanced" portion of EFB that seeks to keep the control grid (and screen grid when necessary) in a relative lockstep relationship with the plate supply voltage, regardless of how it might vary due to changing AC line or dynamic conditions. And, it is very simple to implement as compared to providing tightly regulated power supplies for each element -- and it produces the same low distortion level as the latter approach achieves as well. It is really a win-win situation, with no down side.
Applying EFB to the output stages of the 9300 then produces the following results, based on both channels being driven:
A. Power output in both channels is a sustained 15.5 watts RMS, representing over a 150% increase from that produced by the original design when both channels are driven. When only a single channel is driven, power output reaches 17.0 watts RMS -- this due to less drop in power supply voltages in this scenario.
B. Midband (1 kHz) THD easily drops to under 1% within 1 db of 15.5 watts power output (typically about .70%). In the stock design, notch distortion is just commencing at 1 db below the maximum power output of 10.24 watts, so distortion is significant (~ 2.8%). This means that the modified amplifier is producing less distortion on more power, than the original design is even capable of producing in this scenario. Forgetting the increased power for a moment then, distortion at the 1 db down points has been reduced some 75%.
C. Under quiescent conditions, each output tube idles at a phenomenally low total current draw of just 22 ma (44 ma per channel). Each tube now dissipates just 7.1 watts at the plate -- even less than in the original design -- which amounts to operating at 59.2% of the conservative Design Center rating for the tube (12 watts). This translates to a very long tube life indeed.
D. The operating temperature of the amplifier is notably reduced, particularly that of the power transformer. In combination with the buck connection supplying the correct primary voltage, the reduced heater current draw from using a GZ34 rectifier tube, and now the reduced quiescent current draw due to operating the output stages under the control of EFB, the power transformer reached a low 114F after 2+ hours of operation in a 68F environment. This too speaks to component life in a big way.
Finally, installing EFB of course caused all traces of the notch distortion that appeared in the original design to completely disappear as well.
A few pics are provided, but do remember that this is a development mule........
View attachment 648902
1. The entire EFB circuit easily mounts in the upper right section underneath the chassis. Also visible are the two 10 ohm common cathode resistors for each channel near the bottom center of the pic. Because the tab of the LM337 operates at ground level, no insulation kit is required, although a little silicone grease helps ensure a good thermal contact with the chassis. Under a worst case scenario (continuous full power output in both channels), the EFB regulator dissipates approximately 2.5 watts, which certainly works to ensure its own long life as well.
View attachment 648905
2. On the top side, the new bias control occupies the place of the old hum balance control. Down the road, the three terminals which are currently the speaker output connections will serve as the bias test points for each channel. Three terminals are required since the common connection in this application is not at ground level. Therefore, one terminal would be required to represent the meter common connection, while the other two would represent the test point for each channel. That way, the bias can be adjusted without having to turn the amplifier upside down to do it whenever the tubes are replaced.
View attachment 648909
3. The two images are not calibrated the same for vertical gain so as to be able to show the complete waveforms. On the bottom, the stock design is operating in one channel (using a 200 ohm cathode resistor and bypass cap), and has not even reached the full power point where clipping commences (10.24 watts) and yet, notch distortion is clearly already evident. The top waveform is the other channel with the output stage operating under the control of EFB, developing over 16 watts RMS, as the onset of clipping is approached. Note that besides the increased power, the notch distortion is gone, yet the output tubes idle at an even lower level of quiescent current.
With the power supply and output stages addressed, the next installment will deal with the phase inverter, NFB, stability, and the output transformer.
Dave
Morrism -- They do not -- although I certainly understand that on the surface, it could look like they do.
The DC restorer is an approach to deal with what is known as "blocking" distortion -- distortion that occurs when RC coupling is used to couple the signal from a driver stage into an output stage operating with fixed bias, and the output stage is driven to the point of clipping. Under these conditions, the value of the otherwise normally steady DC voltage that is present across the coupling capacitors becomes elevated on positive drive cycles presented to the grid. When this happens, the tube is then "late" in turning back on after the succeeding negative drive cycle. Since this is happening on both sides of the push-pull cycle, it creates -- and here ya go -- an under-bias condition resulting in the very same notch distortion present in the original 9300 design. However, there is a difference.
Notch distortion created in this manner is quite different between different types of output tubes. Some tubes (like the 6L6 family of tubes for example) and circuits are designed such that the output tubes commence clipping when control grid drive reaches a zero bias condition. On the other hand, other tubes -- like the 6BQ5 family of tubes -- actually reach clipping slightly before a condition of zero bias is reached, which acts to prevent blocking distortion from ever occurring. It will occur mind you. But only after a condition of full power clipping has been reached, making its occurrence a moot point. Therefore, the effects of blocking distortion vary considerably based on the type of output tube being used. In any event, the DC restorer acts to prevent the level of DC voltage appearing across the output stage coupling caps from changing as zero bias conditions are reached in fixed bias designs. That in turn prevents the blocking distortion from occurring.
In the case at hand however, the notch distortion was occurring BEFORE output stage clipping was reached -- this because of using an inappropriate bias system for the loading conditions presented to the output stage. Operating output tubes in class B conditions causes significant shifting of the operating point when cathode bias is used, which is exactly why the load impedance must be elevated when this type of bias arrangement is used in Class AB output stage design -- this to reduce peak currents to minimize the shifting, and therefore, the distortion. It is also the reason that (among other things) a properly design class AB cathode biased output stage will always produce less power output than a properly designed fixed bias Class AB output stage, using the same tubes and B+ supply voltages.
However, in this case, the plate load impedance was not appropriately raised, but cathode bias was still used. The operating point then shifts significantly, creating the notch distortion shown in the scope shot. Magnavox obviously never counted on the user never using the theoretical full power capabilities that the unit should have had. This will all be discussed in much more detail in the next installment, which (in part) addresses conditions where this design is used to drive 8 ohm loads. Until then, understand that the DC restorer prevents notch distortion due to the use of RC coupling in fixed bias amplifiers. EFB(tm) -- in the form presented here -- is a simple and convenient way of applying fixed bias operation to an output stage that also automatically adjusts itself to the prevailing power supply conditions at hand, providing the numerous benefits discussed. However, in this case, as well as in other (Dynaco) documented efforts -- it is the basic conversion from traditional resistive cathode bias operation to fixed bias operation, that corrects the poor design (i.e. improper loading) of the output stage, that causes the notch distortion to begin with.
I hope this helps!
Dave
The DC restorer is an approach to deal with what is known as "blocking" distortion -- distortion that occurs when RC coupling is used to couple the signal from a driver stage into an output stage operating with fixed bias, and the output stage is driven to the point of clipping. Under these conditions, the value of the otherwise normally steady DC voltage that is present across the coupling capacitors becomes elevated on positive drive cycles presented to the grid. When this happens, the tube is then "late" in turning back on after the succeeding negative drive cycle. Since this is happening on both sides of the push-pull cycle, it creates -- and here ya go -- an under-bias condition resulting in the very same notch distortion present in the original 9300 design. However, there is a difference.
Notch distortion created in this manner is quite different between different types of output tubes. Some tubes (like the 6L6 family of tubes for example) and circuits are designed such that the output tubes commence clipping when control grid drive reaches a zero bias condition. On the other hand, other tubes -- like the 6BQ5 family of tubes -- actually reach clipping slightly before a condition of zero bias is reached, which acts to prevent blocking distortion from ever occurring. It will occur mind you. But only after a condition of full power clipping has been reached, making its occurrence a moot point. Therefore, the effects of blocking distortion vary considerably based on the type of output tube being used. In any event, the DC restorer acts to prevent the level of DC voltage appearing across the output stage coupling caps from changing as zero bias conditions are reached in fixed bias designs. That in turn prevents the blocking distortion from occurring.
In the case at hand however, the notch distortion was occurring BEFORE output stage clipping was reached -- this because of using an inappropriate bias system for the loading conditions presented to the output stage. Operating output tubes in class B conditions causes significant shifting of the operating point when cathode bias is used, which is exactly why the load impedance must be elevated when this type of bias arrangement is used in Class AB output stage design -- this to reduce peak currents to minimize the shifting, and therefore, the distortion. It is also the reason that (among other things) a properly design class AB cathode biased output stage will always produce less power output than a properly designed fixed bias Class AB output stage, using the same tubes and B+ supply voltages.
However, in this case, the plate load impedance was not appropriately raised, but cathode bias was still used. The operating point then shifts significantly, creating the notch distortion shown in the scope shot. Magnavox obviously never counted on the user never using the theoretical full power capabilities that the unit should have had. This will all be discussed in much more detail in the next installment, which (in part) addresses conditions where this design is used to drive 8 ohm loads. Until then, understand that the DC restorer prevents notch distortion due to the use of RC coupling in fixed bias amplifiers. EFB(tm) -- in the form presented here -- is a simple and convenient way of applying fixed bias operation to an output stage that also automatically adjusts itself to the prevailing power supply conditions at hand, providing the numerous benefits discussed. However, in this case, as well as in other (Dynaco) documented efforts -- it is the basic conversion from traditional resistive cathode bias operation to fixed bias operation, that corrects the poor design (i.e. improper loading) of the output stage, that causes the notch distortion to begin with.
I hope this helps!
Dave
soundmotor
super modified
That is an impressive performance difference.
So I have a question about how to obtain operating parameters for an EFB system in a given amplifier. Since we're talking Magnavox 9300, lets stick with that one. Does the EFB circuit itself change any depending on the particulars in order to track bias relative to plate voltage, or are those circuit values already established and they will work on anything? Beyond that, how do you determine what the proper idle current level is?
ferninando
Lunatic Member
so only 1 EFB circuit is used for the 4 output tubes?
is the schematic coming in the future?
is the schematic coming in the future?