Hi all,
I just won a quartet of matched JJ-Tesla KT88 on ebay, at a very low price; so I have thus decided to build a PP amplifier.
In the past two weeks I read a lot about tube amplifiers, and I love long tail (I'm a sand guy, and I'm used to use them in semiconductors projects), so this is the way I want to go.
I learned something about negative feedback, and as far I've understood it is not possible (and not suggested) to have a lot of global negative feedback; so why do not have local feedback, instead?
I ended up with two solutions, and I like to ask you if my ideas can work: attached two schematics.
Please do not care about the components values (resistor and caps), I have not yet calculated them, those values are only "casual".
The first schematic (1.pdf) uses local feedback directly from the plate of the differential pair, the second one from the output tubes (both using the R5, R17, C6, C7 networks).
Could they work, or are these stupid ideas? In case, which is the best, in your opinion?
Remember, I'm a newbie in this field, I'm still thinking with "sand" in my mind...
1.pdf
2.pdf
Ciao,
Giovanni
I just won a quartet of matched JJ-Tesla KT88 on ebay, at a very low price; so I have thus decided to build a PP amplifier.
In the past two weeks I read a lot about tube amplifiers, and I love long tail (I'm a sand guy, and I'm used to use them in semiconductors projects), so this is the way I want to go.
I learned something about negative feedback, and as far I've understood it is not possible (and not suggested) to have a lot of global negative feedback; so why do not have local feedback, instead?
I ended up with two solutions, and I like to ask you if my ideas can work: attached two schematics.
Please do not care about the components values (resistor and caps), I have not yet calculated them, those values are only "casual".
The first schematic (1.pdf) uses local feedback directly from the plate of the differential pair, the second one from the output tubes (both using the R5, R17, C6, C7 networks).
Could they work, or are these stupid ideas? In case, which is the best, in your opinion?
Remember, I'm a newbie in this field, I'm still thinking with "sand" in my mind...
1.pdf
2.pdf
Ciao,
Giovanni
Hi Giovanni,
Many good amps use overall negative feedback as well. The amount is far lower than you see in solid state designs due to transformar phase shifts. The feedback is typically returned to the cathode of the voltage amplifier tube, but you could use the grid of a diff pair. I've been thinking of trying that. Local feedback in the form of ultra linear or unity coupling may be used in the output stage to further improve operation.
-Chris
Many good amps use overall negative feedback as well. The amount is far lower than you see in solid state designs due to transformar phase shifts. The feedback is typically returned to the cathode of the voltage amplifier tube, but you could use the grid of a diff pair. I've been thinking of trying that. Local feedback in the form of ultra linear or unity coupling may be used in the output stage to further improve operation.
-Chris
For what it is worth, I like it best applied it from the transformer through a voltage divider into the bottom grid of the LTP since I don't like feedback in the input stage....it sounds less organic to my little tin ears. I usually form the voltage divider by splitting the "tail" resistor, but with THE CCS down there, it might be tougher to do it that way. I have fed it into the both grids, cross feedback from the plates before, but i didn't like it as much that way, but maybe I just used too much FB.
There are other ways to apply local feedback, and in general, if you need to apply feedback even if only to reduce the overal amplifier gain, it is a good idea to apply it locally in order to reduce the required amount of global feedback to get the same effect.
High(er) valuse of global feedback are possible if phase shifts can be made smaller. In particular, direct couping of stages helps on the low end of the spectrum, however, as Chris said, the transformer is the element that is not easily avoided.
One more way to apply feedback locally is plate to plate feedback between output and drivers. However, since we are generally talking triodes in the drivers, the feedback also acts within the driver stage, thus avoiding the extra phase shift of a coupling cap in the feedback network from plate to grid.
Other methods involve partial local degeneration by only partially bypassing cathode resistors. For LTPs, this is commonly done by splitting the tail resistor into two of twice the value, one for each triode cathode. Then the cathodes are joined by a smaller value resistor, which can even be made adjustable. This way it is possible to vary the 'differential' action from full (small resistor = 0) to none (small resistor = infinite), without appreciable change in bias. BTW this is also possible for a tail CCS but requires a slightly inspired combination of CCS and current mirror for the double tail. It is also possible to use a capacitor instead of the small value resistor in order to make the LTP act as a LTP only for AC (for instance in order to enforce DC balance).
High(er) valuse of global feedback are possible if phase shifts can be made smaller. In particular, direct couping of stages helps on the low end of the spectrum, however, as Chris said, the transformer is the element that is not easily avoided.
One more way to apply feedback locally is plate to plate feedback between output and drivers. However, since we are generally talking triodes in the drivers, the feedback also acts within the driver stage, thus avoiding the extra phase shift of a coupling cap in the feedback network from plate to grid.
Other methods involve partial local degeneration by only partially bypassing cathode resistors. For LTPs, this is commonly done by splitting the tail resistor into two of twice the value, one for each triode cathode. Then the cathodes are joined by a smaller value resistor, which can even be made adjustable. This way it is possible to vary the 'differential' action from full (small resistor = 0) to none (small resistor = infinite), without appreciable change in bias. BTW this is also possible for a tail CCS but requires a slightly inspired combination of CCS and current mirror for the double tail. It is also possible to use a capacitor instead of the small value resistor in order to make the LTP act as a LTP only for AC (for instance in order to enforce DC balance).
My two pence:
http://www.dissident-audio.com/PP_ECL86/Schema.gif
Here local feedback is around the power stage.
It also improves balance.
Yves.
http://www.dissident-audio.com/PP_ECL86/Schema.gif
Here local feedback is around the power stage.
It also improves balance.
Yves.
aletheian said:
No, there is no improvement only on the account of the double tail, if you use triangle to star network conversion you will see that it's really the same thing.
The advantage of a split tail becomes apparent when you need to add adjustable differential gain through local degeneration. Conventionally, you would put two extra unbypasssed resistors into each cathode of the pair. The problem here is that adjustment requires both to be adjusted. In the split tail configuration, only a single component is adjusted. In addition, differential gain can be varied down to nearly zero if the cathode to cathode resistor is made very large, without any change to the bias conditions. Differential gain is maximized when the cathode to cathode resistance is zero, at which point you have a standard LTP.
Now, if the tail resistor is replaced by a current sink, making that a split tail configuration requires an unusual dual current sink, but it does indeed improve DC balance. In fact, using BJT composite sink + mirror, insures DC balance to within fractions of a percent. AC balance can also be independently adjusted by including a plate to plate pot that has the wiper connected to +B via a cap.
Amazing what one can learn looking at schematic diagrams of tube oscilloscopes
Yes, but in the case of the original poster, 12AX7 definitely has not enough guts to cope with the additional loading by plate shunt FB AND driving triode strapped 6550. Compute the plate load imposed on 12AX7 in this case of the original poster and see yourself.
A beefier high gain alternatives to 12AX7 would be for example EC91 (single system 7 pin miniature, µ=100). Maybe 12AT7, too.
Tom
I don't undedrstand how dual current sinks would work. AFAIK, the purpose of a current sink in the tail of an LTP is to ensure that the summ of the currents in the two halves of the splitter is constant, thereby giving "perfect" AC balance. Splitting the tail resistor above the current sink helps to improve DC balance, as well as providing negative current feedback. Why go further than that?
On their own, they would not. Reread the post, there is mention of a cathode to cathode resistor in a dual current sink system.
Making the sum of currents constant does not insure AC balance, not unless you count differential output, but the 'balance' does not truly apply. If the two triodes have different characteristics, both DC and AC balance will be compromised in the sense that the signal out of one plate and the other will not be equal amplitude.
Splitting the tail resistor above the current sink will improve DC (and AC) balance but reduce gain due to current feedback. In fact, effectively you are lowering the Gm and making the two triodes 'more equal'. With a current sink, there will not be an appreciable change in DC conditions as the sink still insures that the sum of currents remains constant. Things however get interesting if you need your LTP to maintain near perfect DC balance. See attached picture... it was just meant as an example, not necessairly as something the original poster should use in his design, the idea can be used for other things (like auto biassing of class A PP outputs).
Attachments
If you are just starting out with tube LTP, you might want to ditch the current sink idea AT FIRST, and just build it up with a tail resistor. Then when you get it working, you can sub in the current source later on. The tail resistor develops it's own voltage in conjunction with ohms law, so you don't need a negative voltage supply either.
Actually, in order to save on coupling caps, you may need to use a negative voltage.
The reason for this is that a LTP behaves as a LTP and not merely as a cathoe coupled amp as long as the tail current is fairly constant. In a LTP the two cathodes connect and form a common mode node, where the voltage WRT to ground is the mean of the + and - input + bias voltage. When used as a phase splitter, with one input grounded, the AC component on this node will be roughly half of the input voltage for a reasonable mu. If this is an appreciable part of the total DC voltage on the tail resistor, you cannot assume the tail current is constant. In order to do so, you need to increase the DC voltage drop on the tail resistor to make the AC appear smaller hence the current more constant - this eats away at your B+ and also rises the grid voltage necessary for proper biassing above 0V level, which means you will need coupling caps. Of course, this is a non-issue if they are already there, but the B+ loss can be a problem. This is why a separate negative supply is used, essentially for that voltage to be 'wasted' on the tail resistor in order to better approximate a current source.
The reason for this is that a LTP behaves as a LTP and not merely as a cathoe coupled amp as long as the tail current is fairly constant. In a LTP the two cathodes connect and form a common mode node, where the voltage WRT to ground is the mean of the + and - input + bias voltage. When used as a phase splitter, with one input grounded, the AC component on this node will be roughly half of the input voltage for a reasonable mu. If this is an appreciable part of the total DC voltage on the tail resistor, you cannot assume the tail current is constant. In order to do so, you need to increase the DC voltage drop on the tail resistor to make the AC appear smaller hence the current more constant - this eats away at your B+ and also rises the grid voltage necessary for proper biassing above 0V level, which means you will need coupling caps. Of course, this is a non-issue if they are already there, but the B+ loss can be a problem. This is why a separate negative supply is used, essentially for that voltage to be 'wasted' on the tail resistor in order to better approximate a current source.
ilimzn said:
That's true. I come from the musical instrument amp world, where coupling caps give a nice opportunity to sculpt the midrange tonality by using different compositions to color the sound. I occasionally forget that many hi-fi guys come from the "caps in the signal path are bad" school of thought. heck, I use mylar/paper caps and carbon comp resistors just about everywhere but input stages and power supplies.... want to talk about coloration!
local NFB and LTP you say? seen this one? I clal the circuit E-Linear, as it is voltage( hence the 'E' ), and Ultra-Linear combined. Works quite well actually. Works SE as well, as published by Pete Millett in AudioXpress earlier this year.
regards,
Douglas
The tail load can be as simple as a resistor and voltage, or a small voltge rail, as in fixed bias and CCS of your choice.
regards,
Douglas
The tail load can be as simple as a resistor and voltage, or a small voltge rail, as in fixed bias and CCS of your choice.
Attachments
I don't see how that can be true. With a good CCS in the tail, any change in plate current in the first half of the LTP must be matched by an equal and opposite change in the second half, regardless of how unmatched the LTP triodes may be. If that were not true, then the CCS would not be doing its job. With a CCS tail, AC balance is assured, assuming the plate load resistors of the LTP are accurately matched.
DC balance, however, still depends upon matching of the triodes making up the LTP. A really well-matched double triode (e.g. 6SU7) should achieve acceptable DC balance. If necessary, DC balance can be assisted by using separate cathode resistors above the CCS. As you say, this introduces negative current FB, with consequent loss of gain and elevated plate resistance.
Ray_moth, now that I've re-read you qouting me, I figured out the problem, I must have been more scatterbrained than usual while I was writing that!
Of course, you are right - I was thinking one thing and writing another, namely not AC balance but AC gain per triode in the LTP. This of course, remains different looking at the AC voltage between grid and cathode, but differentially (which is the proper way of looking at that circuit), this does not apply, as the common mode is taken out and the exact gain distribution between the two triodes in the pair becomes largely irrelevant (but of course surfaces as the DC imbalance). Due to the sum of currents being fixed by the CCS, of course, the phase splitting action is always 'perfect' WRT plate current, and if the plate circuit impedances are matched, the AC output is always balanced.
Sorry for the confusion...
Of course, you are right - I was thinking one thing and writing another, namely not AC balance but AC gain per triode in the LTP. This of course, remains different looking at the AC voltage between grid and cathode, but differentially (which is the proper way of looking at that circuit), this does not apply, as the common mode is taken out and the exact gain distribution between the two triodes in the pair becomes largely irrelevant (but of course surfaces as the DC imbalance). Due to the sum of currents being fixed by the CCS, of course, the phase splitting action is always 'perfect' WRT plate current, and if the plate circuit impedances are matched, the AC output is always balanced.
Sorry for the confusion...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.