Dead right.
Ther's two ways to utilise neg feedback to an input tube anode:-
a) To impart a triode even order sound to a low distortion amplifier
b) To actually cancel out distortion when the following voltage amplifying stage is a triode - this requires much less feedback to achive a given distortion spec than does linear resistive feedback.
Where an amp comprises a triode VAS followed by a concertina phase splitter and an ultra-linear output stage or cathode feedback output stage (a la Quad 2), the VAS distortion can be the most significant.
.
I though I had covered this before. You can indeed have a Class B stage that does not draw grid current. It's just a matter of having fixed bias equal to approx the cut off value and signal limitted to less than twice the bias. But the power output will then be less than the same tubes will provide in A or AB1. In A or AB both tubes are carrying the load all or most of the time.
Since Class B distorts more, the only reason for using Class B is to get more power output. So a lower impedance anode load is used and the grids are driven past twice the bias voltage.
In BallPencil's amp, the drive is indeed greater than twice the bias and the output load is very low. (No output transformer - direct into 8 ohms)
Last edited:
There another down side to this daft amplifier of BallPencil's.
The feedback to the input triode imparts a diffrent gain to postive going inputs than for negative going inputs - that was intended apparently.
In a conventional amplifier it won't matter much that the output tubes are driven out of balance. The transformer protects the loudspeaker.
In BallPencils's design however, there is no output transformer or capacitor. So there will be DC fed to the speaker proportional to signal level. Not an ideal way to treat a speaker.
And if one of those 6C33C's flashes over, there will be 150 V on the speaker. 2.8 kW in the voice coil until the HT collapses or the voice coil melts. There will be at best a hell of a bang.
The feedback to the input triode imparts a diffrent gain to postive going inputs than for negative going inputs - that was intended apparently.
In a conventional amplifier it won't matter much that the output tubes are driven out of balance. The transformer protects the loudspeaker.
In BallPencils's design however, there is no output transformer or capacitor. So there will be DC fed to the speaker proportional to signal level. Not an ideal way to treat a speaker.
And if one of those 6C33C's flashes over, there will be 150 V on the speaker. 2.8 kW in the voice coil until the HT collapses or the voice coil melts. There will be at best a hell of a bang.
And in case BallPencil needs still more reinforcment of an important principle:
Respected author Doug Self is the author of an article Class Distinction, Electronics World March 1989, pages 190-195. On page 191 he gives definitions of Classes A, AB, B, C, and D.
He gives definitions much the same as the standard textbooks I quoted earlier except that he uses the word nearly where the others used approximately or similar.
He goes on to explain at some length that a small standing current (to reduce cross-over distortion or for other reasons) in a Class B amplifier, resulting both output devices conducting all the time for small signals, does not mean the amplifier should be called Class AB and not Class B.
In a Class AB amplifier, the standing current is substantial, of roughly similar order to what it would be in Class A. That means that the sound of an AB amplifier is much the same as an A, and not like Class B. (discounting feedback)
Respected author Doug Self is the author of an article Class Distinction, Electronics World March 1989, pages 190-195. On page 191 he gives definitions of Classes A, AB, B, C, and D.
He gives definitions much the same as the standard textbooks I quoted earlier except that he uses the word nearly where the others used approximately or similar.
He goes on to explain at some length that a small standing current (to reduce cross-over distortion or for other reasons) in a Class B amplifier, resulting both output devices conducting all the time for small signals, does not mean the amplifier should be called Class AB and not Class B.
In a Class AB amplifier, the standing current is substantial, of roughly similar order to what it would be in Class A. That means that the sound of an AB amplifier is much the same as an A, and not like Class B. (discounting feedback)
Last edited:
Because the terms are vague. Note that the quoted texts don't give specific numbers. So you have plenty of opportunity to count angels and pinheads.
I don't think the input induced DC imbalance will get past the interstage xfmr. of Ballpencil's amp. But it may indeed unbalance the interstage xfmr. flux, causing distortion. (edit: I see a cap, C2, at the input now, should stop that.)
I thought the whole point of class AB was to avoid crossover distortion. Bipolar SS does have an optimum biasing near class B for least crossover distortion. (The Mosfet case is nearly identical to tubes, and typically improves with hotter biasing without limit.) Are you saying that optimum tube biasing is similarly near class B. The tube datasheets have largely left us in the dark on this, without (usually) any gm curves for power tubes to generate wingspread gm crossover curves.
I thought the whole point of class AB was to avoid crossover distortion. Bipolar SS does have an optimum biasing near class B for least crossover distortion. (The Mosfet case is nearly identical to tubes, and typically improves with hotter biasing without limit.) Are you saying that optimum tube biasing is similarly near class B. The tube datasheets have largely left us in the dark on this, without (usually) any gm curves for power tubes to generate wingspread gm crossover curves.
Last edited:
Look at it this way: The feedback loop gain is different for each half cycle. So one output tube is driven harder than the other one.
While the transformer will most certainly not pass on any DC shift in the opamp output, the DC shift is re-constituted in the output stage.
This is actually a lot more complex than many people realise.
Yes, with tubes, there is an optimum bias point, resulting in a small standing current, where cross-over distortion is minimum. Well, on a Total Harmonic Distortion (THD) test that is.
The human ear reacts differently.
If you set the standing current too low, cross-over distorton manifests as notches cut out of the signal at the cross-over points. Information is subtracted. Partly or wholey lost.
If you set the standing current too high, cross-over distortion manifests as gain doubling around the cross-over point. Information is not subtracted, its' just altered.
For equal distortion on a THD test, it is better audibly if you bias for too much current than for too little. Particularly with MOSFETS and tubes, while there is an optimum standing current on a THD test, in practice it will sound better, or at least no worse, if you set the standing current higher than the THD optimum point.
I think nobody thought of doing wingspread plots back in the tube era. I first saw them when engineers were coming to grips with the more nasty cross-over distortion in transistor amps. Much much later, Doug Self took them up and made them well known.
Due to the statistical nature of music, the zero volt line is crossed much more often than some point away from zero. You can measure a Probability Density Function (PDF) for any given music - the probability that at any given instant of time, the signal volatge will be zero, 1V, 2 V or whatever. The PDF is bell shaped.
The ear doesn't just respond to the amount of distortion. The brain cannot think fast enough. What it responds to is as much the frequency at which distortion occurs, as the amount of distortion.
What this means, is that if you bias for a heavy standing current, as in Class AB, the cross-over distortion instances move away from zero and goes well up on the transfer curve, where you don't hear it so well. The distortion is occurring well down on the skirts of the bell curve. That's why AB sounds just about as good as A.
Yet another reason why a THD figure is no guide to the 'objectionableness' of an amplifier, though it is agood tool for an engineer trying to get everything right.
In summary:-
For equal amounts of distortion as measured in a THD test:-
a) Too much current sounds better than too little
b) A lot of current (as in AB) sounds about as good as Class A.
Last edited:
This is exactly what i am talking about.
We don't land on the moon just by "approximates" and "nears".
It's not important to me that i label my amp as Class AB but it is important for me to stand up to my knowledge. I already feel silly doing this debate, there's no award for being proven right anyway.
This is literally me last night https://xkcd.com/386/
So when does it "approximately" start to become class AB?
I have gone all the way to explain to you why this is not the case. I started by showing that i am pointing at the right node.. and i even showed you the actual Vgk graph that you wanted and you now are saying those graphs are "woffle".
The problem with your calculation above is that you are forgetting that the cathode is not at a fixed voltage. It is swinging along with the output signal. At the peak of the 61Vpk, the cathode of the upper tube is elevated by 15volts. So..
Vk = 15v
Vg = -47 + 61 = 14v
Vgk = 15-14 = -1v
How about that? It remains negative!
The actual number for Vgk at the peak grid swing is not -1v. It's around -5v. This is because you come up with the 122vpp swing yourself when it's actually closer to 110-115vpp. But there, i use your numbers and still end up with Vgk negative.
If you take the time to read my woffle graph, you won't even be using that calculation above anymore. Here i put the graph again. Let's hope you actually read it this time.
An externally hosted image should be here but it was not working when we last tested it.
cnpope saw and understood this way back in post #58.
Alastair E, someone who has actually built and listened to Futterman OTL amp without the MOSFET grid driver, confirms that there is no grid current.
So I guess with unequal AC coupled drives to the outputs, the non-linearity of the outputs, and possible clipping on one side, could cause a DC output imbalance. But wait, how does one get unequal AC drives if they both come from the same interstage xfmr? I think it is just passing on the 2nd Harmonic generated by the N Fdbk triode as if it were original signal.
---------------------------------------
Warning! Sort of Off Topic:
On this optimum biasing, I came up with a new class of operation a while back, that operates in class B (180 degree), but avoids any crossover distortion. Crossover is controlled by switching speed, instead of optimum combining of gm curves.
I might call it class BS for class B + Switching. (Hehe, no worry about marketers using the BS operation mode) It uses a P-P cascode setup, with low voltage class A P-P on the bottom, and high voltage class B P-P up top. Gyrators (Low freq. V stabilized CCS) automatically remove half the class A waveforms from each side, with their voltage setpoints set at the turn-on thresholds for the top stage. Simultaneous switching is guaranteed by the CCS below the bottom class A stage + the auto-balancing of the Gyrators. Any shift in the class A currents instantly upsets the equilibrium of the Gyrators, sending perfect class B currents to the top sides near instantly. (the clamps remove the appropriate half of the class A, AC signal, for each side)
http://www.diyaudio.com/forums/tubes-valves/266913-class-output-without-all-heat-cheat.html
---------------------------------------
Warning! Sort of Off Topic:
On this optimum biasing, I came up with a new class of operation a while back, that operates in class B (180 degree), but avoids any crossover distortion. Crossover is controlled by switching speed, instead of optimum combining of gm curves.
I might call it class BS for class B + Switching. (Hehe, no worry about marketers using the BS operation mode) It uses a P-P cascode setup, with low voltage class A P-P on the bottom, and high voltage class B P-P up top. Gyrators (Low freq. V stabilized CCS) automatically remove half the class A waveforms from each side, with their voltage setpoints set at the turn-on thresholds for the top stage. Simultaneous switching is guaranteed by the CCS below the bottom class A stage + the auto-balancing of the Gyrators. Any shift in the class A currents instantly upsets the equilibrium of the Gyrators, sending perfect class B currents to the top sides near instantly. (the clamps remove the appropriate half of the class A, AC signal, for each side)
http://www.diyaudio.com/forums/tubes-valves/266913-class-output-without-all-heat-cheat.html
Last edited:
I told you - the gains are different for each half cycle. The transformer doesn't prevent that.
But wait, how does one get unequal AC drives if they both come from the same interstage xfmr? I think it is just passing on the 2nd Harmonic generated by the N Fdbk triode as if it were original signal. No harm in that.
probably because according to Keit, the upper tube was a cathode follower, while the bottom tube was a common cathode, do you agree with that smoking-amp....?
i am of the opinion that they are both common cathode amps, as JC or nanana once posted, i hope he chimes in on this conversation...
i am of the opinion that they are both common cathode amps, as JC or nanana once posted, i hope he chimes in on this conversation...
Yes, the interstage xfmr should be making both top and bottom behave like common cathode. Equal drive voltages with respect to the cathodes.
One can look at the 2nd Harmonic in the feedback in two different ways. It makes the Fdbk attenuator different for + and - signal, so different net gains. Or one can look at it as just adding a 2nd H component to the input signal. I think they actually are equivalent, since the unequal +/- Fdbk ends up creating a difference signal at the input. The AC coupling maintains DC balance then, and the interstage xfmr maintains equal drives. So one just ends up with a 2nd H component added to the input signal. Same as if it came from the source.
One can look at the 2nd Harmonic in the feedback in two different ways. It makes the Fdbk attenuator different for + and - signal, so different net gains. Or one can look at it as just adding a 2nd H component to the input signal. I think they actually are equivalent, since the unequal +/- Fdbk ends up creating a difference signal at the input. The AC coupling maintains DC balance then, and the interstage xfmr maintains equal drives. So one just ends up with a 2nd H component added to the input signal. Same as if it came from the source.
Last edited:
We aren't talking about the moon. We are talking about tube amplifiers.
If Class B is defined by an exact half cycle of conduction, then no Class B tube amp in the world can be called Class B. An exact conduction angle is not possible. Theerfore no point in defining a Class B. According to you all tube audio amps are either A or AB.
Over the last 20 years or various schemes have been devised to make the cross-over more precise in solid state amps. Most with limitations. But the Class definitions predate all than by decades.
Nor any award for being proved wrong. It's hard for some folk to admit their wrong though isn't it? But if you can, you will then be able to ultilise new knowlege...
I did enjoy your cartoon. As I post this, its' 1:00 PM here.
What matters is how the amplifier behaves, what sort of power supply and signal drive requirement it has; what type of bias (cathode resistor bias, fixed grid voltage bias etc) will work ok.
These issues are not affected by the conduction angle being not exactly 180 degrees.
I meant woffle in the sense that you are carrying on about leaves and bark without seeing the fundamental property of the forrest - it has trees.
You did claim before that I misunderstood your graph. But I pointed out in a subsequent post that your interpretation cannot be right. You have yet to explain it.
Your graph shows it swinging between +14V and -108V. Total swing 122V.
That 122V pk-pk swing, together with the 47V standing bias is all we need. Those two figures mean grid current (which pushes the whole waveform negative).
If you think your simulation shows no grid currrent than your simulation is faulty (as I said before that can be an outcome of how SPICE works), or you have misinterpreted what its telling you.
It doesn't matter how many times you keep posting graphs, a 122V swing with 47 V bias means grid current. Any time you have a grid swing greater than twice the bias, and you have a grid coupling capacitor, you have grid current. Simple 1st year algebra fer chist's sake.
It doesn't matter how many times you keep posting graphs and re-assert your claims, Koonw's sims, taking into account the three measures he took that reduce grid current, indicates grid current occurs.
Last edited:
thanks smoking-amp, if you ask me, i will never build a 6C33 amp if i have to use global negative feedback, the very low anode resistance of the 6C33 makes making of opt's very simple and cheap...
and if i have to use an interstage traffo, will confine negative feedback to the primary of the IT or else incorporate a feedback winding as one of the secondaries...
and if i have to use an interstage traffo, will confine negative feedback to the primary of the IT or else incorporate a feedback winding as one of the secondaries...
You can look at it as passing on the 2nd harmonic if you like. That's an alternative and perfectly valid way of looking at it.
The thing is, the second harmonic will for one tube increase it's output swing, and for the other tube, will reduce its swing. For one tube the positive peaks of fundamental and harmonic will align. For the other tube the second postitive peak of the harmonic will align with the signal negative peak. Only odd order harmonics can affect both tubes equally.
So the swing of one tube is different to the swing of the other tube. Thus the tubes are unbalanced and there is DC at the load connection.
Last edited:
yup, i can't prove it now because you edited your earlier posts...but i idid recall you said it early on...
- Status
- Not open for further replies.