Examples of Class A push/pull
- Thread starter Donnie2112
- Start date
arts
Addicted Member
Why do you say that?
Retrovert
Addicted Member
Mmmmm, not so much. I very much doubt this amplifier is Class A. Without some proof I conclude that whomever told you that did so to extract money from you.
I'm reminded of the exchange (cited in Gladwell's Blink) between Thomas Hoving (director of the Met) and Arthur Houghton (curator of the Getty) about the obviously faked kouros that the Getty was exhibiting:
Here's why I think claims of Class A operation for the amplifier are unwarranted. I again note the work required to refute such claims:
The depicted amplifier is a Western-Electric (made by Dukane) Amplifier KS-16617-L1 dating from circa 1959. Push-Pull 6V6 driven by a 12AU7 voltage-gain stage and then a 12AU7 current-gain splitter. As would be expected, drives either 4 Ω or 600 Ω, plus some sort of additional test line so the output transformer has an extra secondary. The power output set forth in the manual is 8 Watts with a quoted harmonic distortion of less-than 5% from 100 Hz to 5,000 Hz, the rectifier tube is a 5Y3, and the output transformer is quite small once taking into account the dual secondaries. I have both the manual and schematic.
The intended use was about as far from HiFi as one might venture, being test communications in the AT&T phone system, running between a test desk and the central office and some sort of additional output apparently for monitoring. Given the inherently low bandwidth of such connection, and thus the poor fidelity, of the transmission lines used for transmission, and the lack of any need for fidelity, plus the need to minimize wasted power and heat, it would make absolutely no sense whatsoever to use Class A in push-pull. What would be the benefit?
The output tubes appear to be cathode biased, 300 V on the plate, and share (200 Ω || 25 µF). It appears to be much like any other low-bandwidth push-pull Class AB amplifier made for telephony use in the Bell System. If the amplifier didn't have the over-rated Western-Electric name on it the best use would be as a guitar amplifier. Probably bass guitar since the frequency response is poor. Since it does have the valuable Western-Electric name, the best use would be selling it and buying something with HiFi bandwidth.
I think this again illustrates why claims of push-pull Class A operation must be accompanied by schematics, bias settings, and objective proofs. Such data would save debunking time.
I'm reminded of the exchange (cited in Gladwell's Blink) between Thomas Hoving (director of the Met) and Arthur Houghton (curator of the Getty) about the obviously faked kouros that the Getty was exhibiting:
Hoving turned to Houghton. "Have you paid for this?"
Houghton, Hoving remembers, looked stunned.
"If you have, try to get your money back," Hoving said. "If you haven't, don't."
Houghton, Hoving remembers, looked stunned.
"If you have, try to get your money back," Hoving said. "If you haven't, don't."
Here's why I think claims of Class A operation for the amplifier are unwarranted. I again note the work required to refute such claims:
Brandolini's Law on the Asymmetry of Bullshit: the amount of energy needed to refute bullshit is an order of magnitude bigger than to produce it.
The depicted amplifier is a Western-Electric (made by Dukane) Amplifier KS-16617-L1 dating from circa 1959. Push-Pull 6V6 driven by a 12AU7 voltage-gain stage and then a 12AU7 current-gain splitter. As would be expected, drives either 4 Ω or 600 Ω, plus some sort of additional test line so the output transformer has an extra secondary. The power output set forth in the manual is 8 Watts with a quoted harmonic distortion of less-than 5% from 100 Hz to 5,000 Hz, the rectifier tube is a 5Y3, and the output transformer is quite small once taking into account the dual secondaries. I have both the manual and schematic.
The intended use was about as far from HiFi as one might venture, being test communications in the AT&T phone system, running between a test desk and the central office and some sort of additional output apparently for monitoring. Given the inherently low bandwidth of such connection, and thus the poor fidelity, of the transmission lines used for transmission, and the lack of any need for fidelity, plus the need to minimize wasted power and heat, it would make absolutely no sense whatsoever to use Class A in push-pull. What would be the benefit?
The output tubes appear to be cathode biased, 300 V on the plate, and share (200 Ω || 25 µF). It appears to be much like any other low-bandwidth push-pull Class AB amplifier made for telephony use in the Bell System. If the amplifier didn't have the over-rated Western-Electric name on it the best use would be as a guitar amplifier. Probably bass guitar since the frequency response is poor. Since it does have the valuable Western-Electric name, the best use would be selling it and buying something with HiFi bandwidth.
I think this again illustrates why claims of push-pull Class A operation must be accompanied by schematics, bias settings, and objective proofs. Such data would save debunking time.
arts
Addicted Member
PakProtector
AK member
Damn,I wanted to see him back up his assertion.Now you've ruined it....party pooper
I feel your pain...
cheers,
Douglas
Retrovert
Addicted Member
Damn,I wanted to see him back up his assertion.Now you've ruined it....party pooper
I feel your pain...
Well, shucks. I live in NYC. When one sees roaches (Class A or Class AB) one immediately kills them lest they take up residence.
I didn't ruin anything, though. Proof for these Class A claims is never forthcoming. Par for the course. It's all marketing and bragging and confirmation bias; no facts, no engineering, no truth.
The claimant could, of course, provide the data and prove my analysis was wrong.
I just received four boxes of radio and electronics books from a silent key. Saved them from being left at the curb. I was leafing through some of the old volumes, ranging from the 1920s to the 1950s, and noted entire chapters on Class A, load lines, etc. Beautifully clear prose, wonderful tables and charts, fully derived formulas, meticulously crafted and worked-out examples, and top-notch organization. Actual knowledge. Ahhh, if only people would read it.
The lack of understanding of the basics of amplifier classes is why whenever I hear audiophiles crowing about the superior sound of some ultra-expensive triode amplifier with unicorn fur cable insulation, tube rectifier sound, geriatric PIO capacitors with soakage and ESR high enough for BBQing, high-resistance silver fuses without UL ratings, selection of the right getter shape, and mistaken replacement of high-Gm tubes with high-µ tubes, and no negative feedback because that smothers the electrons, I know that not a single one of them could pass Ethan Winer's transistors versus tubes test.
I've come to view Class A discussions like experiments with the xenomorphs:
"It's a queen. She'll breed. You'll die. Everyone in the company will die."—Ripley, Alien Resurrection
This has been a great thread to debunk so many casual claims of Class A operation -- I'd have to go back through to confirm it, but I think the only examples left standing that are truly Class A designs are those from the Brits, with the Williamson being by far the most common example. But with the Vox AC30 (another Brit design) brought up so commonly (but erroneously) as being a "Class A" design, I thought it would make an excellent example to show visually why in fact it is not. It's true class of operation has already been proven through the discussion of conduction angles, load lines, quiescent current draw, and dissipation and B+ levels, but many techs and hobbiests who are otherwise well capable of troubleshooting and servicing vacuum tube gear from a schematic don't live in the world of engineering where developing output stage operation through the use of these design considerations exists. Without a complete understanding of the definitions or relevant issues involved then, it allows claims made by self proclaimed gurus or less than honest marketing departments to become fuel for internet and folklore misconceptions.
To help put a visual face on the difference between Class A and Class AB operation, I've presented a series of pics that show the difference in Class A and Class AB operation of the same amplifier, using one of the designs of discussion -- the Vox AC30. Well, it's not an actual AC30, but a breadboard version of it's power amplifier section, and to be completely honest, it's really only half of an AC30 -- but the output stage is otherwise set up exactly as depicted from the Vox AC30 schematic, high quiescent current level and all. For practicality reasons, it's constructions has minor deviations from the original design, but not in anyway that would affect this conversation. Through nothing more than the use of a dual trace scope and a simple modification to the output stage cathode circuit, both the composite output waveform developed across an external load resistor, and the waveform developed by each (or either in this case) output tube can be shown at the same time. In this way, examination of the two waveforms shows what happens to the tubes at various power output levels within both classes of operation.
We begin:
BELOW: The test bed. Not exactly an AC30, but for these tests, exactly one half of it. For practical purposes, a choke is not used or needed since the amplifier is powered from a well filtered and regulated Heath SP-17A power supply. The choke is replaced by a 680Ω resistor to replicate the same DC drop as produced by the choke in the AC30 under quiescent conditions. A 6SL7 is used in place of the 12AX7 for the long tail inverter (as I didn't have another 9 pin socket), and coupling caps have been reduced in size (to .047 uF) for space considerations, which can be done since the test signal is a 1 kHz sine wave as supplied from a Heath IG-5218 generator. The original 50Ω cathode resistor has been replaced by a 95Ω resistor (bypassed by 100 uF) and two precision 10Ω individual cathode resistors, to make 100Ω total, or double the original value since only two output tubes are used. The output transformer has an 8000Ω primary (and 16Ω secondary), again, double the original value for use with just two tubes. The output tubes have matched quiescent current within 1%, and power output matched within 3%. B+ supplied to the amplifier as determined from the AC-30 schematic is 320 vdc. The barrier strip at the rear of the board allows for the various power supply and scope connections to be made.
BELOW: The test set up, all ready to go with the power supply, generator, scope, and load resistors connected. On the scope, the top line will display the normal composite output signal as developed and appearing across the output load resistor, while the bottom trace will display the signal generated at the cathode of an output tube, as appearing across it's 10Ω precision cathode resistor. This signal could be taken from the cathode of either tube since the signals there are both the same, although 180˚ out of phase with each other. Technically, the signal there includes that developed by the screen current as well, where as the Class definitions are strictly based on plate current flow only. For the purposes of this illustration however, the presence of the screen current in the cathode signal matters little to the wave forms presented. The digital readout at the top of the scope indicates the B+ supplied to the amplifier.
BELOW: The first test is run with a B+ of 320 vdc, and the cathode bias resistor adjusted to mimic the quiescent current of 50 mA per tube, all as taken from the AC30 schematic. For these particular tubes, the total resistance had to be reduced to 81Ω to produce 100 mA of total cathode quiescent current. The long tail phase inverter pulls only 1 mA total at this voltage, and is therefore insignificant to the total current draw. As can be seen, the power supply is delivering the requisite current flow to the amplifier. Screen grid current was determined to be 4.5-5 mA per tube, so based on a 45 mA plate current and a measured 295 vdc plate to cathode voltage, each plate is dissipating 13.28 watts, or 92.2% of the Design Maximum rated plate dissipation of 14.4 watts for these tubes. No wonder these amps have the reputation they do! There is no room for error on the high side, with the quad of output tubes not only needing to be well matched under at least quiescent levels, but also at a characteristic that is appropriate for the fixed 50Ω bias resistor used in the amp.
BELOW: When the amplifier is driven to an output voltage of 10.5 vac (6.9 watts) as shown on the HP meter on top of the scope, the top wave form shows the expected output waveform, while the bottom waveform shows that the tube being monitored is conducting current at all times at this point.
BELOW: However, as the amplifier is driven beyond the previous power level to the onset of clipping (14.85 vac for 13.8 watts), the output waveform increases and shows the expected fattening of the sign wave (this amplifier uses absolutely no NFB) before clipping actually commences. But above 6.9 watts, the waveform at the cathode (appearing inside and then outside the output waveform in alternating fashion) begins to flatten on the bottom side because it is being driven into cutoff. At the onset of output clipping, the cathode waveform is now well clipped as shown having been driven well into cutoff due to increased signal drive, which is the classic signature of Class AB operation.
BELOW: Driving the amplifier fully into output clipping simply clips the tops and bottoms of the output waveform, and drives the monitored output tube further into cutoff on negative going grid signals. Note that the cathode waveform is now entering clipping on positive peaks as well.
BELOW: Next, the amplifier was set up to operate truly as a Class A amplifier, as Randall Aiken described in the text provided by Retrovert. To do this, the B+ level was first dropped to 263 vdc, and the cathode resistor adjusted to again cause the output stage to draw 100 mA of quiescent current as shown below.
BELOW: To produce an output stage quiescent current of 100 mA at this B+ voltage level, the common portion of the cathode resistance was shunt with a 150Ω resistor to produce a total cathode resistance of 55Ω.
BELOW: The amplifier is again driven to the onset of clipping, now producing an output of 11 vac, or 7.6 watts -- which is exactly double that produced by each channel of the Magnavox 8600 when modified into a DGSE-1. This would be expected since here there are two tubes operating in Class A, versus one in the DGSE-1. As previously discussed, a push-pull Class A amplifier develops double the power of a single tube of the same type when operated under the same conditions in SE mode. Note that even though the output waveform is at the onset of clipping, the cathode waveform shows that the monitored tube is never reaching cutoff -- even under full power output conditions. This is the cornerstone of Class A operation.
BELOW: If amplifier is driven beyond full power as shown in this scenario, the output waveform clips evenly on the top and bottom, with the cathode waveform closely mimicking the output waveform -- which can only happen in a true Class A output stage.
The original parameters used to justify the Vox AC30 as a Class A design were the high quiescent cathode current and plate dissipation levels used. These are surely characteristics of a Class A design, but alone, hardly guarantee that is in fact an accurate designation. Output stage loading and B+ levels are equal elements of consideration as well in making the determination -- along with elevated quiescent current and plate dissipation levels. Class A operation can only be assigned when plate current flows 100% of the time when maximum power output of the stage has been reached. Anything less is not Class A.
Known design characteristics coupled with an accurate understanding of class definition can prove (or disprove) the presence of Class A operation in a given design. Scoping a design as was done here can also be an effective tool -- not only to help understand how tubes perform in the different classes of operation under dynamic conditions, but also to quickly determine output stage class of operation for an unknown design as well.
Dave
To help put a visual face on the difference between Class A and Class AB operation, I've presented a series of pics that show the difference in Class A and Class AB operation of the same amplifier, using one of the designs of discussion -- the Vox AC30. Well, it's not an actual AC30, but a breadboard version of it's power amplifier section, and to be completely honest, it's really only half of an AC30 -- but the output stage is otherwise set up exactly as depicted from the Vox AC30 schematic, high quiescent current level and all. For practicality reasons, it's constructions has minor deviations from the original design, but not in anyway that would affect this conversation. Through nothing more than the use of a dual trace scope and a simple modification to the output stage cathode circuit, both the composite output waveform developed across an external load resistor, and the waveform developed by each (or either in this case) output tube can be shown at the same time. In this way, examination of the two waveforms shows what happens to the tubes at various power output levels within both classes of operation.
We begin:
BELOW: The test bed. Not exactly an AC30, but for these tests, exactly one half of it. For practical purposes, a choke is not used or needed since the amplifier is powered from a well filtered and regulated Heath SP-17A power supply. The choke is replaced by a 680Ω resistor to replicate the same DC drop as produced by the choke in the AC30 under quiescent conditions. A 6SL7 is used in place of the 12AX7 for the long tail inverter (as I didn't have another 9 pin socket), and coupling caps have been reduced in size (to .047 uF) for space considerations, which can be done since the test signal is a 1 kHz sine wave as supplied from a Heath IG-5218 generator. The original 50Ω cathode resistor has been replaced by a 95Ω resistor (bypassed by 100 uF) and two precision 10Ω individual cathode resistors, to make 100Ω total, or double the original value since only two output tubes are used. The output transformer has an 8000Ω primary (and 16Ω secondary), again, double the original value for use with just two tubes. The output tubes have matched quiescent current within 1%, and power output matched within 3%. B+ supplied to the amplifier as determined from the AC-30 schematic is 320 vdc. The barrier strip at the rear of the board allows for the various power supply and scope connections to be made.
BELOW: The test set up, all ready to go with the power supply, generator, scope, and load resistors connected. On the scope, the top line will display the normal composite output signal as developed and appearing across the output load resistor, while the bottom trace will display the signal generated at the cathode of an output tube, as appearing across it's 10Ω precision cathode resistor. This signal could be taken from the cathode of either tube since the signals there are both the same, although 180˚ out of phase with each other. Technically, the signal there includes that developed by the screen current as well, where as the Class definitions are strictly based on plate current flow only. For the purposes of this illustration however, the presence of the screen current in the cathode signal matters little to the wave forms presented. The digital readout at the top of the scope indicates the B+ supplied to the amplifier.
BELOW: The first test is run with a B+ of 320 vdc, and the cathode bias resistor adjusted to mimic the quiescent current of 50 mA per tube, all as taken from the AC30 schematic. For these particular tubes, the total resistance had to be reduced to 81Ω to produce 100 mA of total cathode quiescent current. The long tail phase inverter pulls only 1 mA total at this voltage, and is therefore insignificant to the total current draw. As can be seen, the power supply is delivering the requisite current flow to the amplifier. Screen grid current was determined to be 4.5-5 mA per tube, so based on a 45 mA plate current and a measured 295 vdc plate to cathode voltage, each plate is dissipating 13.28 watts, or 92.2% of the Design Maximum rated plate dissipation of 14.4 watts for these tubes. No wonder these amps have the reputation they do! There is no room for error on the high side, with the quad of output tubes not only needing to be well matched under at least quiescent levels, but also at a characteristic that is appropriate for the fixed 50Ω bias resistor used in the amp.
BELOW: When the amplifier is driven to an output voltage of 10.5 vac (6.9 watts) as shown on the HP meter on top of the scope, the top wave form shows the expected output waveform, while the bottom waveform shows that the tube being monitored is conducting current at all times at this point.
BELOW: However, as the amplifier is driven beyond the previous power level to the onset of clipping (14.85 vac for 13.8 watts), the output waveform increases and shows the expected fattening of the sign wave (this amplifier uses absolutely no NFB) before clipping actually commences. But above 6.9 watts, the waveform at the cathode (appearing inside and then outside the output waveform in alternating fashion) begins to flatten on the bottom side because it is being driven into cutoff. At the onset of output clipping, the cathode waveform is now well clipped as shown having been driven well into cutoff due to increased signal drive, which is the classic signature of Class AB operation.
BELOW: Driving the amplifier fully into output clipping simply clips the tops and bottoms of the output waveform, and drives the monitored output tube further into cutoff on negative going grid signals. Note that the cathode waveform is now entering clipping on positive peaks as well.
BELOW: Next, the amplifier was set up to operate truly as a Class A amplifier, as Randall Aiken described in the text provided by Retrovert. To do this, the B+ level was first dropped to 263 vdc, and the cathode resistor adjusted to again cause the output stage to draw 100 mA of quiescent current as shown below.
BELOW: To produce an output stage quiescent current of 100 mA at this B+ voltage level, the common portion of the cathode resistance was shunt with a 150Ω resistor to produce a total cathode resistance of 55Ω.
BELOW: The amplifier is again driven to the onset of clipping, now producing an output of 11 vac, or 7.6 watts -- which is exactly double that produced by each channel of the Magnavox 8600 when modified into a DGSE-1. This would be expected since here there are two tubes operating in Class A, versus one in the DGSE-1. As previously discussed, a push-pull Class A amplifier develops double the power of a single tube of the same type when operated under the same conditions in SE mode. Note that even though the output waveform is at the onset of clipping, the cathode waveform shows that the monitored tube is never reaching cutoff -- even under full power output conditions. This is the cornerstone of Class A operation.
BELOW: If amplifier is driven beyond full power as shown in this scenario, the output waveform clips evenly on the top and bottom, with the cathode waveform closely mimicking the output waveform -- which can only happen in a true Class A output stage.
The original parameters used to justify the Vox AC30 as a Class A design were the high quiescent cathode current and plate dissipation levels used. These are surely characteristics of a Class A design, but alone, hardly guarantee that is in fact an accurate designation. Output stage loading and B+ levels are equal elements of consideration as well in making the determination -- along with elevated quiescent current and plate dissipation levels. Class A operation can only be assigned when plate current flows 100% of the time when maximum power output of the stage has been reached. Anything less is not Class A.
Known design characteristics coupled with an accurate understanding of class definition can prove (or disprove) the presence of Class A operation in a given design. Scoping a design as was done here can also be an effective tool -- not only to help understand how tubes perform in the different classes of operation under dynamic conditions, but also to quickly determine output stage class of operation for an unknown design as well.
Dave
Last edited:
The original Williamson triode output stage most definitely operated in Class A.
Dave
Dave
Pio1980
AK Member
In P/P?
The Williamson is a push-pull all triode configured amplifier that operates in Class A.
Dave
Dave
Pio1980
AK Member
Hmmm, interesting. Both output tubes conducting midpoint at zero crossing?
Would have thought class A/B.
Last edited:
Both tubes do conduct at the zero crossing as they do in any push-pull class A or class AB amplifier. What makes the Williamson a class A design, is that both tubes are also still conducting 100 % of the time when full power output has been reached as well.
Dave
Dave
Pio1980
AK Member
That was what was in my mind, seems wierd, compared to later competing P/P amps from McIntosh, Mullard, Laurent, ect, and my limited experience with 45 P/P amps in "electric" era radios.
Meanwhile, I'm going back down this rabbit hole;
http://www.oestex.com/tubes/williamson.htm
I think that would largely intend on their intended application -- but in any event, you wouldn't know for sure unless you know the operating parameters of the output stage. By topology, an output stage driven by an interstage transformer could be Class A, Class AB1, Class AB2 or Class B1 or B2.
Dave
Dave
You should read the original Williamson article. ;-) Lower distortion and better transient response, among other reasons. Williamson argues that self-biased AB amplifiers exhibit effective changes in gain in the low frequencies. During transients, as the current rises so does the bias voltage which changes the gain of the output stage. Also lower output impedance and better damping factor.
