Hi everyone,
I am torn between a number of possibilities for the front end of a PP amp I'm building, each of which appeals to me in a different way. Trouble is, I don't have enough experience to be able to judge which is the best so, rather than make the wrong choice, I thought I'd ask for opinions from others who've "been there, done that".
The back end of the amp will be PP EL34s strapped as triodes. These will be driven by a 6SN7 differential amplifier, which can either have two anti-phase inputs from an earlier splitter, or it can act as a LTP splitter itself (with a 6AU6 pentode as a current sink in the tail).
When it comes to the front end, I see a number of interesting possibilities:
(1) A triode-strapped pentode (I aleady have some 6AU6s that I could use) or 1/2 of a 6SL7, letting the following 6SN7 act as the splitter.
(2) A 6AU6 in pentode mode, again letting the 6SN7 act as the splitter - a bit like the Mullard 5-20 design.
(3) A mu-stage, using 1/2 of a 6SL7 as the gain tube and a 6AU6 in pentode mode as the active load, again letting the 6SN7 act as the splitter.
(4) A 6SL7 as a differential front end, with a 6AU6 pentode as a current sink in the tail and a shared cathode resistor in the tail of the following 6SN7 driver stage.
If possible, I'd rather not use other types of tubes than those mentioned above, because I already have them and I don't want to have to buy any more!
Grateful for any opinions.
Regards,
Ray
I am torn between a number of possibilities for the front end of a PP amp I'm building, each of which appeals to me in a different way. Trouble is, I don't have enough experience to be able to judge which is the best so, rather than make the wrong choice, I thought I'd ask for opinions from others who've "been there, done that".
The back end of the amp will be PP EL34s strapped as triodes. These will be driven by a 6SN7 differential amplifier, which can either have two anti-phase inputs from an earlier splitter, or it can act as a LTP splitter itself (with a 6AU6 pentode as a current sink in the tail).
When it comes to the front end, I see a number of interesting possibilities:
(1) A triode-strapped pentode (I aleady have some 6AU6s that I could use) or 1/2 of a 6SL7, letting the following 6SN7 act as the splitter.
(2) A 6AU6 in pentode mode, again letting the 6SN7 act as the splitter - a bit like the Mullard 5-20 design.
(3) A mu-stage, using 1/2 of a 6SL7 as the gain tube and a 6AU6 in pentode mode as the active load, again letting the 6SN7 act as the splitter.
(4) A 6SL7 as a differential front end, with a 6AU6 pentode as a current sink in the tail and a shared cathode resistor in the tail of the following 6SN7 driver stage.
If possible, I'd rather not use other types of tubes than those mentioned above, because I already have them and I don't want to have to buy any more!
Grateful for any opinions.
Regards,
Ray
Interesting! What pentode are you using for the front end?
I would think that, as you're using 6SN7 drivers (like I plan to do), the high gain of a pentode could be useful. With higher mu triodes for the driver , like the 12AX7 that Mullard specified in the 5-10 and 5-20 designs, the gain of the EF86 pentode front end is considered by many people nowadays to be exessive. A popular modification for the Mullard amplifiers is to strap the EF86 in triode-mode and to remove the cathode bypass capacitor, to bring its gain down to manageable proportions.
I would think that, as you're using 6SN7 drivers (like I plan to do), the high gain of a pentode could be useful. With higher mu triodes for the driver , like the 12AX7 that Mullard specified in the 5-10 and 5-20 designs, the gain of the EF86 pentode front end is considered by many people nowadays to be exessive. A popular modification for the Mullard amplifiers is to strap the EF86 in triode-mode and to remove the cathode bypass capacitor, to bring its gain down to manageable proportions.
After much rumination and simulation, I think I have the answer. I'm forced to use either pentode or triode mode, by the OP tubes I have (EL34s) and the OPTs I have (3.5k p-p with 8-ohm speaker connected to the 8-ohm terminals, or 7k p-p with 8-ohm speakers connected to the 4-ohm terminals). There are no UL taps.
Pentode appeals more than triode, at the moment, because I already have triode-mode working in an all-differential 6SL7-6SN7-EL34 PP amp and, whilst I like it, I'd like to see if I can get pentode-mode working successfully and have more power.
For pentode mode I need high open-loop gain, to allow me to apply a lot of NFB for the sake of good damping. I'd also like to use directly-coupled cf drivers, so that there is no blocking distortion on transients and also to provide a bit of AB2 headroom. I think the Mullard-type front-end design seems the most likely to succeed, i.e. single-ended input stage, direct-coupled to a double triode differential (LTP) splitter. If there's going to be a cf driver following the splitter, then a high gain, high impedance, low current splitter will do the job OK.
What I plan is:
1. EF86 in pentode-mode with bypassed cathode resistor, directly coupled to stage 2
2. 6SL7 LTP splitter, with 6AU6 CCS in the tail, indirectly coupled to stage 3
3. 6SN7 cathode follower drivers, directly coupled to stage 4
4. EL34s in pentode-mode PP, using fixed bias (applied to the grids of the 6SN7, in the usual way).
Local NFB will be applied in a local loop, from the EL34 plates to the 6SL7 plates, and a global loop, from the OPT secondary to the EF86 cathode. A stable, flat response with good roll-off at each end of the audio spectrum will be achieved using:
(a) for HF stability, phase correction with the usual zobel network across the plate load resistor of the EF86, tweaked with a small cap in parallel with the global NFB resistor to tame overshoot on square waves; and
(b) for LF stability, careful selection of the values of capacitors used for the EF86 screen decoupling to cathode, coupling between 6SL7 and 6SN7 and the splitter AC grounding.
The PS will use a hybrid Graetz bridge (SS ultra-fast recovery and TV damper diodes), feeding a CLC filter. B+ for the OP tube plates (OPT primary CT). B+ for the OP tube screens will be regulated using a series MOSFET (IRF820) with zener stack providing the reference voltage for the MOSFET gate. The regulated screens B+ will also feed an L-C filter for the 6SL7 B+, then there will be another MOSFET series regulator to drop the voltage to ~200v for the 6SN7s B+. A further R-C filter will provide B+ at about 190v for the EF86.
A negative supply will be provided, using a parallel voltage regulator (transistor), for the 6SN7 cathodes and fixed bias and, with an additional R-C filter, for the 6AU6 (CCS) cathode.
Having modeled the entire circuit using LTSpice, it seems viable, with 50w max. output and good distortion levels, judging byt the FFT analysis - 1kHz fundamental @+28dB, 3rd harmonic @-32dB, 5th harmonic $-50dB, 7th harmonic @-60dB. Even harmonics are negligible.
Of course, modeling is not the same as the real world: input signals are not pure sine waves; OP transformers have leakage inductance and capacitance, core losses, distortions and imperfect band width; speakers are not purely resistive; etc. However, I'm encouraged, by comments made by Norman Koren and others, to hope that Spice models can give a good indication of what should be possible, if we use good quality parts and follow good layout and construction practices.
Worst thing about LTSpice modeling is the time it takes to simulate both channels of a stereo amp and a common power supply! I tried to find out how I might speed things up by adjusting the run parameters but I find the LTSpice ‘Help’ incomprehensible and the Yahoo! LTSpice forum too difficult to follow - at least, for me.
Pentode appeals more than triode, at the moment, because I already have triode-mode working in an all-differential 6SL7-6SN7-EL34 PP amp and, whilst I like it, I'd like to see if I can get pentode-mode working successfully and have more power.
For pentode mode I need high open-loop gain, to allow me to apply a lot of NFB for the sake of good damping. I'd also like to use directly-coupled cf drivers, so that there is no blocking distortion on transients and also to provide a bit of AB2 headroom. I think the Mullard-type front-end design seems the most likely to succeed, i.e. single-ended input stage, direct-coupled to a double triode differential (LTP) splitter. If there's going to be a cf driver following the splitter, then a high gain, high impedance, low current splitter will do the job OK.
What I plan is:
1. EF86 in pentode-mode with bypassed cathode resistor, directly coupled to stage 2
2. 6SL7 LTP splitter, with 6AU6 CCS in the tail, indirectly coupled to stage 3
3. 6SN7 cathode follower drivers, directly coupled to stage 4
4. EL34s in pentode-mode PP, using fixed bias (applied to the grids of the 6SN7, in the usual way).
Local NFB will be applied in a local loop, from the EL34 plates to the 6SL7 plates, and a global loop, from the OPT secondary to the EF86 cathode. A stable, flat response with good roll-off at each end of the audio spectrum will be achieved using:
(a) for HF stability, phase correction with the usual zobel network across the plate load resistor of the EF86, tweaked with a small cap in parallel with the global NFB resistor to tame overshoot on square waves; and
(b) for LF stability, careful selection of the values of capacitors used for the EF86 screen decoupling to cathode, coupling between 6SL7 and 6SN7 and the splitter AC grounding.
The PS will use a hybrid Graetz bridge (SS ultra-fast recovery and TV damper diodes), feeding a CLC filter. B+ for the OP tube plates (OPT primary CT). B+ for the OP tube screens will be regulated using a series MOSFET (IRF820) with zener stack providing the reference voltage for the MOSFET gate. The regulated screens B+ will also feed an L-C filter for the 6SL7 B+, then there will be another MOSFET series regulator to drop the voltage to ~200v for the 6SN7s B+. A further R-C filter will provide B+ at about 190v for the EF86.
A negative supply will be provided, using a parallel voltage regulator (transistor), for the 6SN7 cathodes and fixed bias and, with an additional R-C filter, for the 6AU6 (CCS) cathode.
Having modeled the entire circuit using LTSpice, it seems viable, with 50w max. output and good distortion levels, judging byt the FFT analysis - 1kHz fundamental @+28dB, 3rd harmonic @-32dB, 5th harmonic $-50dB, 7th harmonic @-60dB. Even harmonics are negligible.
Of course, modeling is not the same as the real world: input signals are not pure sine waves; OP transformers have leakage inductance and capacitance, core losses, distortions and imperfect band width; speakers are not purely resistive; etc. However, I'm encouraged, by comments made by Norman Koren and others, to hope that Spice models can give a good indication of what should be possible, if we use good quality parts and follow good layout and construction practices.
Worst thing about LTSpice modeling is the time it takes to simulate both channels of a stereo amp and a common power supply! I tried to find out how I might speed things up by adjusting the run parameters but I find the LTSpice ‘Help’ incomprehensible and the Yahoo! LTSpice forum too difficult to follow - at least, for me.
ray_moth said:
Works for me. This is pretty much what I designed, except that I used 807s. Also, I found that a 6J5 with unbypassed cathode that doubled for a gNFB summing node provided enough front end gain. This design didn't require more than 4db(v) of gNFB to clean up any remaining pentode nastiness.
I would, however, suggest using cascoded BJTs to form the active tail load. BJTs have much higher gain, thus impedance, and make better CCSs.
And watch those Vhk ratings.
This may not be necessary.
I would use an active regulator for screen voltage here.
ray_moth said:
If you accept to build an LTP PI with pethodes, you may remove a stage (and the associated complexity).
http://www.dissident-audio.com/PP_6L6/Page.html
EL34 was testedsuccesfully on this amp but do not deliver more power than 6L6 because the load is a bit too high.
Yves.
Thanks, Yves, I see what you mean, although I wonder if the gain would be sufficient for the amount of NFB I like to use.
Also, I prefer to use a CCS with an LTP splitter, and that is very difficult with pentodes, I think. I see two problems: 1) The CCS may be in conflict with the screens in determing the current in each tube. 2) The cathode currents include both the plate and screen currents, so this could defeat, to some extent, the balancing of signals that is the oblective of using a CCS.
Also, I prefer to use a CCS with an LTP splitter, and that is very difficult with pentodes, I think. I see two problems: 1) The CCS may be in conflict with the screens in determing the current in each tube. 2) The cathode currents include both the plate and screen currents, so this could defeat, to some extent, the balancing of signals that is the oblective of using a CCS.
Ray,
Are there any 12AT7s on your "pile"? A 'T7/ECC81 differential splitter has just enough gain to drive triode wired EL34s without loop NFB. If you provide a LINEAR 1st gain block that replaces NFB losses, the "El Cheapo" topology, with its short NFB loop, will work for you.
IMO, the 12AT7/ECC81 is a near perfect tube for LTP splitter duty. It has both high mu and high gm. Also, the distortion spectrum, which is skewed towards the 2nd harmonic, mates very well with PP finals.
Are there any 12AT7s on your "pile"? A 'T7/ECC81 differential splitter has just enough gain to drive triode wired EL34s without loop NFB. If you provide a LINEAR 1st gain block that replaces NFB losses, the "El Cheapo" topology, with its short NFB loop, will work for you.
IMO, the 12AT7/ECC81 is a near perfect tube for LTP splitter duty. It has both high mu and high gm. Also, the distortion spectrum, which is skewed towards the 2nd harmonic, mates very well with PP finals.
hey-Hey!!!,
Try a pair of pentodes for the LTP splitter. CCS under the cathodes, and a dropping R to the g2 node. Bypass g2 to the cathodes. You'll be able to keep g2 current out of the picture fairly well if you pick a decent pentode. The 12BY7 or 12HL7 have worked for me well. Plate load of ~8k, and g2 of ~80V. This will offer an output Z comparable to the 6SN7, and give gain like your proposed 3 stage. Return NFB to the grounded grid, and don't forget to put a small stopper on it.
cheers,
Douglas
Or build a pentode from some triodes, or better still triode on bottom and low capacitance MOSFET on top.
Try a pair of pentodes for the LTP splitter. CCS under the cathodes, and a dropping R to the g2 node. Bypass g2 to the cathodes. You'll be able to keep g2 current out of the picture fairly well if you pick a decent pentode. The 12BY7 or 12HL7 have worked for me well. Plate load of ~8k, and g2 of ~80V. This will offer an output Z comparable to the 6SN7, and give gain like your proposed 3 stage. Return NFB to the grounded grid, and don't forget to put a small stopper on it.
cheers,
Douglas
Or build a pentode from some triodes, or better still triode on bottom and low capacitance MOSFET on top.
Yvesm said:
I suppose that a CCS in the LTP’s tail will do good things for distortion when applying feedback into the LTP.
I did some measurements in a setup like that with and without the CCS in the tail.
In my case it was a standard LTP with triodes, but nevertheless..
Distortion pattern without CCS:
An externally hosted image should be here but it was not working when we last tested it.
Distortion pattern with CCS:
An externally hosted image should be here but it was not working when we last tested it.
Jan E Veiset
Hi ray_moth.
As stated, I run this amp at 8dB global feedback.
Indeed, results are very dependent on how firmly the lousdpeakers need to be damped.
CCS or not CCS ?
In fact it's a matter of available negative rail voltage wich determine the "lenght of the tail"
Since I needed some -115V to feed the cathodes of the CF DC coupled driver I was satisfied with a single 27K resistor.
Also, the symetry of an LTP may be defined as the ratio between the Rp of the tube vs the lenght of the tail. Lo Rp triodes need a longer tail than penthodes !
Since I'm rebuilding this amp in Stereo version, I'll check both way and show the results.
About disturbance caused by screen current, consider it is quite constant as long as the anode volage does not fall too low.
Note that anyway this voltage limit is usually lower that you can obtain with a triode.
Yves.
As stated, I run this amp at 8dB global feedback.
Indeed, results are very dependent on how firmly the lousdpeakers need to be damped.
CCS or not CCS ?
In fact it's a matter of available negative rail voltage wich determine the "lenght of the tail"
Since I needed some -115V to feed the cathodes of the CF DC coupled driver I was satisfied with a single 27K resistor.
Also, the symetry of an LTP may be defined as the ratio between the Rp of the tube vs the lenght of the tail. Lo Rp triodes need a longer tail than penthodes !
Since I'm rebuilding this amp in Stereo version, I'll check both way and show the results.
About disturbance caused by screen current, consider it is quite constant as long as the anode volage does not fall too low.
Note that anyway this voltage limit is usually lower that you can obtain with a triode.
Yves.
Ray,
Before you get too far down the track have another look at SY's Red Light District. The thing that he has used to his advantage there is that with output tubes in pentode mode the Miller Capacitance is very small and he can use high values of grid stop resistors to address blocking distortion without loosing high frequency response. This little gimick is often overlooked when looking at his design because we are too busy looking at his LED bias scheme.
I'm not suggesting you do (since it seems clear that you have some other topology in mind) but his design in toto would easily scale up to EL34 Outputs and a 6SL7 input/splitter.
Cheers,
Ian
Before you get too far down the track have another look at SY's Red Light District. The thing that he has used to his advantage there is that with output tubes in pentode mode the Miller Capacitance is very small and he can use high values of grid stop resistors to address blocking distortion without loosing high frequency response. This little gimick is often overlooked when looking at his design because we are too busy looking at his LED bias scheme.
I'm not suggesting you do (since it seems clear that you have some other topology in mind) but his design in toto would easily scale up to EL34 Outputs and a 6SL7 input/splitter.
Cheers,
Ian
Hi, Ian,
Actually, there are a couple of other considerations, apart from blocking, that cause me to avoid using a 6SL7 to drive OP tubes directly. I see an impedance mismatch between the high OP impedance of the 6SL7 stage and the low IP impedance of a power tube grid using fixed bias (even a pentode). To me, that's bad practice. (And I really want to use fixed bias.)
Also, having that EL34 grid circuit effectively in parallel with the plate and plate load of a 6SL7 would drag down the impedance at that point, making the plate-plate NFB from EL34 to 6SL7 less effective.
I see a directly coupled cf (or possibly sf) buffer as the ideal solution here, since it solves both of these problems and does the job, also, when it comes to the grid going a bit positive on occasions. A large stoppper is OK for minimizing the effects of blocking, but it still can't give any AB2 headroom. (I realise that the EL34 is not exactly a good candidate for AB2, unlike the 807 or 6146, but I only want it for transients.)
LOL! Ya got me, I'm a stubborn SOB.
Actually, there are a couple of other considerations, apart from blocking, that cause me to avoid using a 6SL7 to drive OP tubes directly. I see an impedance mismatch between the high OP impedance of the 6SL7 stage and the low IP impedance of a power tube grid using fixed bias (even a pentode). To me, that's bad practice. (And I really want to use fixed bias.)
Also, having that EL34 grid circuit effectively in parallel with the plate and plate load of a 6SL7 would drag down the impedance at that point, making the plate-plate NFB from EL34 to 6SL7 less effective.
I see a directly coupled cf (or possibly sf) buffer as the ideal solution here, since it solves both of these problems and does the job, also, when it comes to the grid going a bit positive on occasions. A large stoppper is OK for minimizing the effects of blocking, but it still can't give any AB2 headroom. (I realise that the EL34 is not exactly a good candidate for AB2, unlike the 807 or 6146, but I only want it for transients.)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.