Calculating gain of class A and class B region of class AB output stage

Discussion in 'Tube Audio' started by Alan0354, Dec 29, 2017.

  1. Alan0354

    Alan0354 Super Member

    Messages:
    2,735
    Location:
    Bay Area
    I don't follow post #112, how to have grounded cathode with CCS. Does he mean CCS drawing current from the cathode of each power tube and using bypass cap from each cathode to ground to give an AC ground? I don't see any difference from just simple grounding the cathode. Usually people adjust the bias current to adjust the wing spread AFTER optimize the design.

    Using tube to do CCS is too complicated. If I were to go with modern design like CCS, I would use BJT as CCS. BJT is perfect as the output impedance ( like rp in tubes) is way higher than any tubes can touch. ONLY reason to use CCS is to provide a very high impedance current source so the current source approach an idea current source. Using bypass cap just does not make sense to me.

    Unless one actually hook up the two PP power tubes as differential pair ( long tail pair ). I actually thought about that before also. Then it will make a whole world of sense to use CCS tail to make it ideal. But that only work for class A, not class AB.
     
    Last edited: Jan 1, 2018

     

    Please register to disable this ad.

  2. maxhifi

    maxhifi AK Subscriber Subscriber

    Messages:
    2,221
    Location:
    Edmonton Alberta Canada
    Mostly what smoking amp is doing, is food for thought. The non-linearity he's showing with the transconductance plot is also shown on the regular tube curves with how the spacing between the grid voltage lines get grouped together toward the bottom of the pentode curves. He's just showing the same thing in another way, and looking for ways to mitigate this effect.

    A regular design, simply push-pull, kept in class A, and biased so there's lots of linear region above and below the operating point, will go a long way towards mitigating both of these issues. I'm not seeing a big reason to get too fancy here.. just push pull Class A with the operating point chosen to bias 6L6GC tubes right to the limit of dissipation. The more I study this, the more I think the Quad II looks like a good design... it sure checks all the boxes for quality, and none of the boxes for gadgetry.
     
  3. Alan0354

    Alan0354 Super Member

    Messages:
    2,735
    Location:
    Bay Area
    Yes, I understand what he showed in #53 the spacing is very even with PP class A region.
    What is Quad II? I don't know what that is.

    Main think I got out of the thread is tubes do have crossover distortion just like SS output stage, the wing spread is very similar. This is on top of the gain difference between class A and B region.
     
  4. maxhifi

    maxhifi AK Subscriber Subscriber

    Messages:
    2,221
    Location:
    Edmonton Alberta Canada
    QUAD is a British hi-fi company, the model II is that rusty old British amplifier I saw on the bay area craigslist which I got all excited about a couple months ago. One of the most popular tube amplifiers ever made, and for a good reason - QUAD II is one of the classic designs from the tube era. ,Class A push-pull KT-66, unity coupling, pentode driver tubes, etc.
     
  5. Alan0354

    Alan0354 Super Member

    Messages:
    2,735
    Location:
    Bay Area
    I ran across this link in the thread.https://www.tubecad.com/2015/11/blog0332.htm

    This actually confirm the difference in gain between the class A and B region. He has to double the gm just when the amp goes from class A to class B where the opposite valve is turning off.

    Speaking of the article, I seriously question the article. He use cathode diode to double the gm for tubes to coincide when the other tube just turn off, the diode turns on to double the gm. I'll let Max to comment how good one can match the timing and whether the abrupt turn on of the diode will create more distortion than it's worth.

    Then the guy went into gm doubling in SS amp. To the best of my knowledge, I disagree. First, SS output stage is very very different from tubes. Tube push pull is common cathode stage with gain. gm change the gain, thereby create distortion as I describe in this thread. There are some merit to increase the gm when the amp enter into class B region.......if you can match the rate of increase of gm to the graduate turn of the the other valve. SS output stage is EMITTER FOLLOWER in most case as he describe. We don't care about gm, the gain of the stage is somewhere around 0.95 or so and vary just a little bit, nothing like the variation of the gain of the tube PP output stage. More importantly, you look at all the Wing Spread plot in both books by Cordell and Self, THE GAIN BEFORE AND AFTER THE CROSSOVER TRANSITION IS THE SAME. There is no gain change between class A and class B region.

    More importantly, transistor is NOT like tubes, the transition is quite abrupt, you play games like putting the diode to double the gm, there will be a abrupt transition and that will create high order harmonics.

    Also if you look at the how he adjust the current. He need to set the current so the first emitter resistor ( he called out 1ohm which is a really bad choice) drops 1/2 the Vbe of the diode or about 0.3V. Even at 1ohm, it's 300mA through one transistor. That's a lot of current for one transistor. Dropping 300mV across the emitter resistor is just wrong. This is explained in very detail in both books by Cordell and Self called the Oliver's condition where you want ONLY about 26mV drop across the emitter resistor to get the smoothest crossover region.............................

    Yes, These are just fruitful thoughts, BUT that also can be misleading. I'll let Max to comment on the tube portion. I can only say I am surprised the person in TubeCad claims he has been designing SS amp for years before going to tubes. I am really surprised.

    From books and from experience, the best way to minimize crossover distortion in SS amp is using more output pairs, very detail explanation are given why more pairs lower crossover distortion.
     
  6. Alan0354

    Alan0354 Super Member

    Messages:
    2,735
    Location:
    Bay Area
    I just downloaded the schematic, it has LNFB to the cathode of the power tube through the output transformer. You like that? First, the disadvantage is feedback through the OPT just like UL that you don't like. Also, I don't think there is OPT available that have all these extra winding.

    GNFB go to the screen grid of the differential driver tubes, that's different!!!
     
  7. maxhifi

    maxhifi AK Subscriber Subscriber

    Messages:
    2,221
    Location:
    Edmonton Alberta Canada
    In the Quad II it's better implemented than ultralinear, because the screen voltage can be lower. Also, the damping factor is better than ultralinear, about 2 without any global negative feedback because of the cathode winding.

    I'm not saying build Quad Replicas, just that all this discussion recently proves to me their designer (Peter Walker) made all the right choices, and understood tube theory on a high level.

    You're right, it would be a fool's errand to try and make special one of a kind transformers for one amplifier.
     
  8. Alan0354

    Alan0354 Super Member

    Messages:
    2,735
    Location:
    Bay Area
    I still like what we've been talking about better, no transformer in the picture, very low output impedance, simple. That's the reason I spent over 3 weeks getting those equations together.
     
  9. trobbins

    trobbins Active Member

    Messages:
    435
    Location:
    Melb, Australia
    Gm 'doubling' seems such a misnomer to characterise class A to B operational portions.

    Assessing measured valve transfer curves, and overlaying curves to show a combined PP response is certainly enlightening - especially where the measured transfer curves relate directly with operational levels in an amplifier.

    But perhaps more enlightening would be if the effective total stage gm, or whatever other metric is considered appropriate, could be measured to add some reality. A simple PP amp designed for AB operation, with feedback removed, and a resistive loading, could be tested for change of input to output peak sinusoid voltage, as signal input increased from 0V to clipping. Variation in total Gm would exhibit as a non-linearity in the measured transfer curve of the waveform peak voltage.

    Within the limits of 1947 vintage laboratory test equipment, Langford-Smith and Aston measured no non-linearity for the majority of class A portion of operation in the A515 PP amplifier using 807's in triode mode - although this was for average voltage readings on an AVO 7 (not peak). A Williamson amp would probably make a neat simple test bed for any such gm variation testing, as stages before the output stage are sufficiently linear, and PP output stage bias can be easily changed, and even pentode mode could be simply implemented, or UL if a later commercial variant was on the bench.

    https://www.dalmura.com.au/static/A515.pdf
     

     

    Please register to disable this ad.

  10. Alan0354

    Alan0354 Super Member

    Messages:
    2,735
    Location:
    Bay Area
    I think the real life gain changes between the transition from class A to B is not as obvious as I have been talking about as the gm increase as the plate current increases. I am just making a point that the gain change is theoretically there, not necessary you can see a distinct step.

    I am just argue from the other side, from the plate curve, it's obviously that the spacing between the plate curves increase per at higher current level, even before transitioning from A to B where the lower tube is still barely on, most of the pull is from the top tube. The effect of the lower the tube is already very small. So even when at the point of handing over, the change of gain might not be apparent at all.

    That's the reason when I read the article from TubeLab in post #45 that he actually use a diode to double the gm to match the class A to B transition region, I was more alarmed!!!! I can only comment on the part of the article that talked about SS amp that I absolutely disagree with the article, I hope Max can talk a little more about the tube. I just feel it will do more damage to the signal than helping.

    I wish I have a tube amp so I can actually do some measurements. But without that, I can only talk. This is only a part of the exercise in my learning process, Not that I expect to see a step in change of gain. When I have a tube amp, I would run triangular wave where the side is actually a straight line. Any change in gain will show up like a kink on the straight line.

    This is getting very interesting. I hope someone have a tube amp can run the triangular wave to see whether you get a straight line or you can see a kink at the transition between class A and B.
     
    maxhifi likes this.
  11. 6DZ7

    6DZ7 Super Member

    Messages:
    1,870
    As the current falls to 0 at cutoff in the (-) side, the signal is still rising and carrying the power into the tranny to "fill in" on the (+) side, so there is little observable change in the output wave form at the speaker terminal. The signal stays under clipping. The high power of the conducting tube masks the 0 output of the other. Unless the tube going into cutoff throws noise from an inelegant shutdown there should be nothing to see or hear.
     
  12. Alan0354

    Alan0354 Super Member

    Messages:
    2,735
    Location:
    Bay Area
    I tend to agree. As I said, my whole argument in this thread is all theoretical. I think I confirmed that there is a theoretical change in gain.......not necessary a real gain change. I can see the transition is very gradual and not obvious at all.

    That's the reason I am more alarmed when I see article using diode to abruptly increase the gm at some point. Tubes are more graduate in any transition, not the SS devices, they can switch from one to the other in an instant.
     
  13. maxhifi

    maxhifi AK Subscriber Subscriber

    Messages:
    2,221
    Location:
    Edmonton Alberta Canada
    I think that's the classic case of a solution looking for a problem. AB1 is widely and successfully used in high fidelity applications where more than about 20W is needed, and has been since the 1930s.

    I do however also think there's some elegance to the pure class A operation, and I think the effort smoking gun is making to make total gm of an output stage constant by optimization of class A operating point has the potential to make an audible improvement. That's why I said I would like to see his finished product, that path he's travelling looks like an interesting one.

    With both feet on the ground though, I can't see anything really wrong with a more conventional approach, and certainly would suggest to stay the course with a conventional AB1 push-pull design for the first effort.
     
  14. Alan0354

    Alan0354 Super Member

    Messages:
    2,735
    Location:
    Bay Area
    I don't know what Smoking Amp referred to in his design, I don't quite get his last post, feel like I am missing something. One think though, I absolutely would not use tube CCS, want to go CCS type of modern design, absolutely use BJT where they are much better suit for CCS.

    I am open to mix new design to the old circuit, but I draw the line in stacking tubes just to say it's all tubes. I would not do cascode with tubes also, want cascode, use BJT as the common base stage. Just like want CCS? Use BJT.

    I am actually thinking on and off on the SE amp with 3 parallel tubes. I am thinking of a tube with less grid swing requirement so I can use an ultra low distortion opamp like OPA2604 that can swing from 0V down to -40V. Using LNFB like what we talked about to lower the output impedance, use pentodes output. Use parafeed if needed. No more preamp tubes to deal with. With that, all distortion will be produced by the output tubes as the SS front end will have ZERO distortion. ( well, say 0.0008%). So I don't have to worry about what preamp tube to choose, whether it has enough drive or not. Those opamp can drive +/-30mA without blinking an eye, no cathode followers can even touch this. The front end circuit will have BW of over 1MHz, no more bottle neck............
     
  15. trobbins

    trobbins Active Member

    Messages:
    435
    Location:
    Melb, Australia
    Alan, have you done a quick check of miller capacitance, current slew rate and frequency, and what grid leak current and grid voltage can be attained for what grid current level when using that opamp ?

    It will be interesting to see what hills and bumps appear in the transfer curves for each common valve type and typical range of idle biasing, loadline and B+. It's a good time to see this sample data grow, now that curve tracers and post-processing software are much more available. One needs to know the beast, before it can be tamed.
     
  16. Alan0354

    Alan0354 Super Member

    Messages:
    2,735
    Location:
    Bay Area
    I am all thinking out loud, nothing solid yet. Are you referring to the SE amp I am talking about?

    I have use opamp OPA2604. It is a very nice opamp. I use it in my preamp. I compare a lot of opamp to choose this one:http://www.ti.com/lit/ds/symlink/opa2604.pdf

    It can use +/-24V rails and swing to within 3V to the rail. So I am sure I can make it swing from 0 to -40V. The THD is very low and drive current is very high.

    I have not done any stability calculation, miller caps, pole and zero compensations. I am concentrating on the topology part. I think the topology is the most important thing, getting that right, everything else is going to fall in place. I really not worry about the stability, when I build the circuit, then I worry about it. They are really not that hard. I spent almost a month working out the equations to make sure the topology make sense and make sure I see all the pit falls of the design. That's why I spent so much time on this thread that might not be that important. It's important to me so I know exactly what I am getting into.

    I don't know I want to invest $500 for a curve tracer, I want to build my first tube amp, make sure I like it well enough to go any further before I commit more money on equipment. I just use the info from the datasheet. Sometimes, just looking at the result can mask a lot of the fine details. I want to design the whole thing on paper first.
     
  17. stevo137

    stevo137 AK Subscriber Subscriber

    Messages:
    2,713

     

    Please register to disable this ad.

  18. Alan0354

    Alan0354 Super Member

    Messages:
    2,735
    Location:
    Bay Area
    I have been reading chapter 2, section 6 of RDH4 on the Conversion Factor and see how much I can use to predict the gm, rp with plate current Ib. Section 6 mainly try to use the datasheet to convert to the specific operating condition, not directly address what we've been talking about. I make use of the FACT that the d(Ib)/d(plate voltage) is very small for pentode( high rp), I can use the conversion factor F(gm) = sqrt(Fe). Yes the gm increase with plate current.


    BUT

    Remember, in the gain equation as shown:
    Gain class A and B.jpg

    Both class A and B are related to "u" the amplification factor. u= gm X rp. Conversion factor of rp F(rp) = 1/sqrt(Fe). Therefore u is quite constant as the two conversion factors cancel each other. Amplication factor u was said to be quite constant by text books already, not just me coming to this conclusion. You cannot just look at the gm and conclude the the gain rise, you have to look at the whole equation to judge.

    So, the only thing that affect is the change of plate resistance rp of the tube that is in the denominator of the gain equations. The effect might not be as prominent as people want to believe.

    The transition from class A to class B is when the plate current is TWICE the idle current ( by definition of class AB that when one tube reach 2 times the idle current, the other tube turns off). So rp of the upper tube will reduce to 0.7 (1/sqrt(2)) of the rp at idle. You can plug the number into the equations to find the gain.

    Tell me what you think, I would not count on it too much.
     
  19. PakProtector

    PakProtector AK member

    Messages:
    3,654
    Location:
    Ann Arbor, MI
    the u of a pentode is not constant, unless we have wired it as a triode, then as plate resistance and gm vary inversely u stays relatively constant. In pentode, with gain( or u ) dependent on the gm and loading, its voltage gain is *NOT* going to be anything resembling constant. The plate curves, and the better data sheets show this clearly. With pentodes, while the plate resistance is usually high compared to the load, the plate resistance varies far less than with triodes...and if you measure the circuit you will discover that the change in loading from a-a/2 to a-a/4 is a gradual thing, depending on how fast the finals cut off...and all the while gm is increasing...

    Build one, and then measure it...or measure an existing amp. Beware using triode gain equations on pentodes.
    cheers,
    Douglas
     

Share This Page