Home |
Search |
Today's Posts |
#1
|
|||
|
|||
A few general PP power stage questions
Hi RAT's:
Below are a few questions that don't seem to be specifically addressed in the references I have: 1 When using multiple output tubes in parallel on each differential side with cathode autobias, can you theoretically use one coupling cap, and/or grid load resistance, and/or one grid stopper resistance for all of them as a group? Seems likely. If so, would you practically want to, or not want to, for any reason? 2 Has anybody used the Blumlein "garter" cathode bias method? I've seen it mentioned but have not seen anyone expound on it. It seems like a very nice and simple biasing method. Is it not so that once you are in the general neighborhood of the desired idle current that relative fine balance between both diff pairs is what it is all about? 3 Can OPT based cathode feedback be used in cathode fixed biased PP output stages? I have only seen schematics where fixed bias is used and the transformer leads are connected directly to the cathodes. |
#2
|
|||
|
|||
A few general PP power stage questions
"Wessel Dirksen" said:
1 When using multiple output tubes in parallel on each differential side with cathode autobias, can you theoretically use one coupling cap, and/or grid load resistance, and/or one grid stopper resistance for all of them as a group? Seems likely. If so, would you practically want to, or not want to, for any reason? Don't do that. Grid stoppers have to be at the closest proximity of the grid pin. Further, tubes vary in parameters. Connecting all the grids in parallel to a single grid grounding resistor and coupling capacitor will probably work without signal, but not when driven. Why? My gut instincts tell me. I'm just a simple tube hobbyist, not a number cruncher. There are some "professional engineers" here who can probably tell you the theoretical ins and outs of this. 2 Has anybody used the Blumlein "garter" cathode bias method? I've seen it mentioned but have not seen anyone expound on it. It seems like a very nice and simple biasing method. Is it not so that once you are in the general neighborhood of the desired idle current that relative fine balance between both diff pairs is what it is all about? I haven't the faintest, sorry. 3 Can OPT based cathode feedback be used in cathode fixed biased PP output stages? I have only seen schematics where fixed bias is used and the transformer leads are connected directly to the cathodes. It can. See e.g. the Quad II circuit, where a single common cathode resistor + bypass cap are connected to the cold side of the cathode windings to ground. -- "Audio as a serious hobby is going down the tubes." - Howard Ferstler, 25/4/2005 |
#3
|
|||
|
|||
A few general PP power stage questions
Wessel Dirksen wrote: Hi RAT's: Below are a few questions that don't seem to be specifically addressed in the references I have: 1 When using multiple output tubes in parallel on each differential side with cathode autobias, can you theoretically use one coupling cap, and/or grid load resistance, and/or one grid stopper resistance for all of them as a group? Seems likely. If so, would you practically want to, or not want to, for any reason? You need separate grid stoppers for each tube. But nothing wrong with a 2uF cap and a 68k resistor to bias say 4 parallel EL34. But each EL34 should have its own Rk and Ck so bias regulation is good for each tube. In old amps where just one bias Rk and Ck are used the bias currents can vary by 100%. 2 Has anybody used the Blumlein "garter" cathode bias method? I've seen it mentioned but have not seen anyone expound on it. It seems like a very nice and simple biasing method. Is it not so that once you are in the general neighborhood of the desired idle current that relative fine balance between both diff pairs is what it is all about? The sox and trousers method of bias is the one I'd stay with if i were you. Without the sox and trousers adjusted properly, a lotta instability can ensue. 3 Can OPT based cathode feedback be used in cathode fixed biased PP output stages? No problem, I do it all the time. Old Quad II amps benefit from having the CFB winding CT grounded, with 500ohms + 1,000uF as bias networks between the ends of the CFB windings and the cathodes. I have only seen schematics where fixed bias is used and the transformer leads are connected directly to the cathodes. Quad II is cathode bias with CFB, but with a common R and C between CFB CT and 0V. Its a cheap nasty way of doing things. Patrick Turner. |
#4
|
|||
|
|||
A few general PP power stage questions
Patrick Turner schreef: Wessel Dirksen wrote: Hi RAT's: Below are a few questions that don't seem to be specifically addressed in the references I have: 1 When using multiple output tubes in parallel on each differential side with cathode autobias, can you theoretically use one coupling cap, and/or grid load resistance, and/or one grid stopper resistance for all of them as a group? Seems likely. If so, would you practically want to, or not want to, for any reason? You need separate grid stoppers for each tube. But nothing wrong with a 2uF cap and a 68k resistor to bias say 4 parallel EL34. But each EL34 should have its own Rk and Ck so bias regulation is good for each tube. In old amps where just one bias Rk and Ck are used the bias currents can vary by 100%. 2 Has anybody used the Blumlein "garter" cathode bias method? I've seen it mentioned but have not seen anyone expound on it. It seems like a very nice and simple biasing method. Is it not so that once you are in the general neighborhood of the desired idle current that relative fine balance between both diff pairs is what it is all about? The sox and trousers method of bias is the one I'd stay with if i were you. Without the sox and trousers adjusted properly, a lotta instability can ensue. Thanks Patrick and Sander, Patrick, Sox and trousers? That's funny (Sox=Rk, Trousers=Ck ?) Let me see if I understand you correctly. Combining the coupling capacitance and grid leak resistance is OK to multiple output tubes, but grid stoppers and the autobias cathode parts should kept individual per tube? The Blumlein circuit is described here in the middle of the web page. The other circuits described look good but too complex. http://www.tubecad.com/2005/May/blog0046.htm I was hoping to get a more in depth expert opinion on this circuit. It seems that a normal recommended value of grid leak resistor is fed to the cathode circuit of the other differential side where it is inserted between 2 Rk's of normal calculated value in series with each other and straddled by Ck. I guess this way each diff pair would have twice the cathode resistance to ground and as mentioned kill some B+. I really like the "differential auto balance" concept though. Worth it? 3 Can OPT based cathode feedback be used in cathode fixed biased PP output stages? No problem, I do it all the time. Old Quad II amps benefit from having the CFB winding CT grounded, with 500ohms + 1,000uF as bias networks between the ends of the CFB windings and the cathodes. Ok, I think I get it. By "CT" you mean grounding the center tap of the CFB part of the OPT right? If so, then the Blumlein autobias circuit could theoretically be combined with CFB. I have only seen schematics where fixed bias is used and the transformer leads are connected directly to the cathodes. Quad II is cathode bias with CFB, but with a common R and C between CFB CT and 0V. Its a cheap nasty way of doing things. I tend to like cheap and nasty. Is CFB a good form of local FB? It would seem to be quite tidy with primarily the reactive parts of the output tube being in the feedback loop and by stepping up the damping factor only at the output stage where it suffers the worst impedance mismatch. The rest of the stages in the chain I am considering have 100 ratio of Zin/Zout thanks to 6H30's. Patrick Turner. |
#5
|
|||
|
|||
A few general PP power stage questions
Wessel Dirksen wrote: Patrick Turner schreef: Wessel Dirksen wrote: Hi RAT's: Below are a few questions that don't seem to be specifically addressed in the references I have: 1 When using multiple output tubes in parallel on each differential side with cathode autobias, can you theoretically use one coupling cap, and/or grid load resistance, and/or one grid stopper resistance for all of them as a group? Seems likely. If so, would you practically want to, or not want to, for any reason? You need separate grid stoppers for each tube. But nothing wrong with a 2uF cap and a 68k resistor to bias say 4 parallel EL34. But each EL34 should have its own Rk and Ck so bias regulation is good for each tube. In old amps where just one bias Rk and Ck are used the bias currents can vary by 100%. 2 Has anybody used the Blumlein "garter" cathode bias method? I've seen it mentioned but have not seen anyone expound on it. It seems like a very nice and simple biasing method. Is it not so that once you are in the general neighborhood of the desired idle current that relative fine balance between both diff pairs is what it is all about? The sox and trousers method of bias is the one I'd stay with if i were you. Without the sox and trousers adjusted properly, a lotta instability can ensue. Thanks Patrick and Sander, Patrick, Sox and trousers? That's funny (Sox=Rk, Trousers=Ck ?) Let me see if I understand you correctly. Combining the coupling capacitance and grid leak resistance is OK to multiple output tubes, but grid stoppers and the autobias cathode parts should kept individual per tube? That's correct. The Blumlein circuit is described here in the middle of the web page. The other circuits described look good but too complex. http://www.tubecad.com/2005/May/blog0046.htm I was hoping to get a more in depth expert opinion on this circuit. Blumlien shows genius yet again. The garter circuit does improve matters of dc balance quite substantially, and in a worthwhile amount. But what does one do if you have a dozen output tubes, 6 per side of the PP circuit? With two tubes all is ok, but balance is impossible with 12 tubes, so some active method to equalise bias current would ahve to be used. I was just joking about the trousers and sox bit. But before leaving for the office tomorrow morning, try cutting one leg off your trousers at the knee, and wearing a long pink sock on the exposed leg. It'd be a miracle if your boss doesn't threaten you with the sack during the day. Conventions are silly eh? The above website is **very** interesting, and it has some bias circuits using LTPs with bjts to auto balance the bias of a pair of output tubes. I have tried almost exactly what they have there with PNP bjts and found they work extremly well until one applies loop NFB, and then the circuit becomes impossible to stabilise at LF. It seems that a normal recommended value of grid leak resistor is fed to the cathode circuit of the other differential side where it is inserted between 2 Rk's of normal calculated value in series with each other and straddled by Ck. I guess this way each diff pair would have twice the cathode resistance to ground and as mentioned kill some B+. I really like the "differential auto balance" concept though. Worth it? Yes. 3 Can OPT based cathode feedback be used in cathode fixed biased PP output stages? No problem, I do it all the time. Old Quad II amps benefit from having the CFB winding CT grounded, with 500ohms + 1,000uF as bias networks between the ends of the CFB windings and the cathodes. Ok, I think I get it. By "CT" you mean grounding the center tap of the CFB part of the OPT right? Yes. If so, then the Blumlein autobias circuit could theoretically be combined with CFB. Not really because there is a fairly high signal of opposite phase at each cathode, so effectively you would at least be halving the time constant of the CR coupling to the driver stage. I don't know how exactly the stability would be affected. But I like a more hands on approach where an LTP using bjt is used to monitor the Ik of each output tube and if the Ik should drift appart more than 3mA then an led lights up to tell you which tube is conducting more current. There is then a pot where the bias can be slightly adjusted so the two leds near each output tube go out when Ik is within 2mA, and swinging the pot will tune the bias balance even closer. Ths adjust mechanism tells you when a tube is playing up, wearing out, and if you cannot adjust the high Ik tube down with a pot you *know* its stuffed, or there is some other problem. Most times where you buy a pair of unmatched tubes you will find that with plain cathode bias of individual tubes that the bias current sets itself within 3% balance, which is quite OK. The sound starts to go to mud when imbalance drifts to 10%, and the levels are high because the OPT core has its iron µ effectively reduced so LF signals tend to saturatae the core and cause intermodulation with higher F. Quad II amps are renowned for such problems when the tubes age and many samples I have serviced have 40mA in one output tube and 90mA in the other which sometimes has some red patches on the anode, and the sound becomes bad above 1/2 a watt. Just placing individual biasing reduces the imbalance to within 7% even with tubes so far apart and allows many more years of service from the tubes. Peter Walker really made a huge mistake in his original design imho by addopting his use of a single RC for biasing both output tubes. I have only seen schematics where fixed bias is used and the transformer leads are connected directly to the cathodes. Quad II is cathode bias with CFB, but with a common R and C between CFB CT and 0V. Its a cheap nasty way of doing things. I tend to like cheap and nasty. Is CFB a good form of local FB? It is if well implemented. The idea that Quad used is quite valid. It allows the full tetrode power of KT66 to be realised but the tubes act as though Ra was a lot lower, but yet they remain easy to drive. The more CFB you have, the more grid drive you need, so the more linear your drive amp has to be. McIntosh used equal % of P turns in the anode and cathode winding, so you can have 125Vrms at anode and cathode, and if the tube gain = 15, then you need 17Vrms between cathode and grid to cause the 250Vrms between anode and cathode and so you need 125V + 17V applied to each output grid, and hence the complexity of the McIntosh driver stage with its extra stage. The high drive voltage means the thd contribution of the driver becomes greater than that produced in the output stage so you would need global loop NFB to reduce it down. McIntosh used around 20 Db of global and with the 15Db of local NFB in the output stage their amps have around the same thd as many other amps which don't use large amounts of CFB and just rely on modest global NFB. But where McI amps shine is with the Rout of thier amps; the total applied NFB is around 35dB, so Rout is much lower than many other amps; Quad II is about 1 ohm, but McI is about 0.2ohms. McI amps allowed one to obtain 50 watts from a pair of 6L6, a huge amount of power for 1949, and thd and Rout was quite low due to the mainly class B action of the output tubes. Their operation is substantially improved when they are altered for lower B+, and class A operation, when the Rout is halved and thd reduced to vanishingly low levels, so that a watt thd 0.01%. Nevertheless, a good Williamson or class A UL amp will also make around 0.02% at a watt and will use one less stage of gain than the McI, and the OPT is a lot simpler. Hence hardly any DIYers made copies of the McI amp; they voted with their wallets for the simpler Wiliamson, or the Mullard 520. Very few have copied the Quad examples. I believe it to be worth the trouble, and its the only way i make PP amp but I use a higher % of cfb than Quad ever did. It would seem to be quite tidy with primarily the reactive parts of the output tube being in the feedback loop and by stepping up the damping factor only at the output stage where it suffers the worst impedance mismatch. The rest of the stages in the chain I am considering have 100 ratio of Zin/Zout thanks to 6H30's. CFB allows the output tubes to be substantially linearized via a short feedback loop involving only the OPT and the output tubes. The reactive elements of the output tube are negligible since the Miller C is low, and you only have *one* lot of Miller C, not 3 lots, which is the case with a 3 stage amplifier. Therefore the use of moderate CFB, say up to 25% of the total signal voltage between a and k, is a great way of making the output stage behave well and the global FB can be reduced to say between 6 and 12dB which is a low amount. Quad II amps without their global NFB loop and when using the 8 ohm outlet config produce an Rout = 9 ohms. The Rout from pure beam tetrode output stage with KT66 would be about 132 ohms, so that 10% of CFB Quad uses is *very* useful and the effect on the thd is also very beneficial. The global NFB then reduces the Rout to around an ohm for a reasonable damping factor. The idea is that the majority of signal error correction by NFB is carried out in the output stage and not in the driver amp. The Quad method of OPT set up is as effective as the use of the 40% of UL tappings which has about the same effect as the 10% of CFB, except that perhaps the Quad method is more efficient because the feedback is mainly applied in the grid-cathode loop, with also some slight feedback effect in the grid-screen circuit, wheras the UL circuit has all its feedback applied via the anode-screen circuit. Triode connection of tubes with screens is a case where *all* the anode voltage is used as feedback in the screen circuit, ie, the UL tappings are at 100%. The trouble with triodes is that unless they are pushed into class A2 or AB2, the output power is severely limited to half what the tube can deliver, so the UL connection was tried to allow the tube to approach triode fidelity but without the power loss penalty. CFB achieves the same result, only slightly better, ie, Rout and thd are slightly lower than UL circuits where power output is still near maximum for pure tetrode and the output stage sensitivity is about the same; ie, the grid voltage applied to the output stage is about the same for the two ways of connecting the tubes. Quad also liked to be different, to have the best, and they patented their 10% of CFB, so everyone else had to make do with something less, but bean counters loved UL because it was so much easier to make an OPT without a separate tertiary winding devoted to CFB, and probably easier to get the OPT to have low leakage inductance in the right places to allow good stablity; Leak had only 3P sections and two S sections, and got away with it. But of course all the cheap and nasty bean counter driven OPT designs that were the hallmark of a lousy minded consumer audio product industry of 1955 can be easily improved upon, and its for this reason i regard Leaks and Quads as very mediocre amplifiers, things that can and do sound well when properly set up but which all measure quite poorly by today's standards where we have good sound and less winding losses, better stability, and more reliablility, so fewer OPT write offs every time an output tube died in a vesuvious of over current excitement. The Dynaco ST80 was one of the last Dynaco tube amps to be made, and its vastly superior to its predessor, the ST70, and anything made by Leak, but you'd expect that after 20 years. I measured an ST80 they other day, and its OPTs are a lot better than in the ST70, and its quite stable, even without any R load, and using just a 0.47uF or 0.22 uF cap as a load. Dynaco finally got with it, and did the required homework to get their circuit to work properly, after spending the extra $1 for a bit more irn and wire, and $2 for the extra labour. Finally, I have cathode bias in my 300 watt amps for the 12 output tubes. This saves having 24 bias adjustments for 2 channels. In the latest versions i use a patentable and little understood method of shunting excess cathode signal current with a simple two bjt circuit that stops the change of dc cathode voltage that occurs when you apply high signal levels to a class AB amp with cathode bias. So my 300 watt amps work exactly like fixed bias amps with their lower thd ability; the cathode bias voltage changes less than 10% between idle and full power with a sine wave, and while in class A the ac signal bypass circuit remains utterly dormant since it only works when the cathode current change goes over the twice the idle bias current. No need to have active biasing of the grid circuits. Class AB cathode bias amps suffer when driven to the point where musical transients cause the bias voltage at the cathode to begin undulating around, and the dc currents lurch about; but not in my amps. I did post a circuit of the basic approach at abse about 2 years ago, but nobody cared, understood, or tried it out, so I won't be giving any other details now, so youse can all go and invent the solution to easy biasing and dynamic regulation on your own, now that I have spelled out the ideas. Patrick Turner. |
#6
|
|||
|
|||
A few general PP power stage questions
Patrick Turner schreef: Wessel Dirksen wrote: Patrick Turner schreef: Wessel Dirksen wrote: Hi RAT's: Below are a few questions that don't seem to be specifically addressed in the references I have: 1 When using multiple output tubes in parallel on each differential side with cathode autobias, can you theoretically use one coupling cap, and/or grid load resistance, and/or one grid stopper resistance forall of them as a group? Seems likely. If so, would you practically wantto, or not want to, for any reason? You need separate grid stoppers for each tube. But nothing wrong with a 2uF cap and a 68k resistor to bias say 4 parallel EL34. But each EL34 should have its own Rk and Ck so bias regulation is good for each tube. In old amps where just one bias Rk and Ck are used the bias currents can vary by 100%. 2 Has anybody used the Blumlein "garter" cathode bias method? I've seen it mentioned but have not seen anyone expound on it. It seems like a very nice and simple biasing method. Is it not so that once you arein the general neighborhood of the desired idle current that relative fine balance between both diff pairs is what it is all about? The sox and trousers method of bias is the one I'd stay with if i were you. Without the sox and trousers adjusted properly, a lotta instability can ensue. Thanks Patrick and Sander, Patrick, Sox and trousers? That's funny (Sox=Rk, Trousers=Ck ?) Let me see if I understand you correctly. Combining the coupling capacitance and grid leak resistance is OK to multiple output tubes, but grid stoppers and the autobias cathode parts should kept individual per tube? That's correct. The Blumlein circuit is described here in the middle of the web page. The other circuits described look good but too complex. http://www.tubecad.com/2005/May/blog0046.htm I was hoping to get a more in depth expert opinion on this circuit. Blumlien shows genius yet again. The garter circuit does improve matters of dc balance quite substantially, and in a worthwhile amount. But what does one do if you have a dozen output tubes, 6 per side of the PP circuit? With two tubes all is ok, but balance is impossible with 12 tubes, so some active method to equalise bias current would ahve to be used. I was just joking about the trousers and sox bit. But before leaving for the office tomorrow morning, try cutting one leg off your trousers at the knee, and wearing a long pink sock on the exposed leg. It'd be a miracle if your boss doesn't threaten you with the sack during the day. Conventions are silly eh? The above website is **very** interesting, and it has some bias circuits using LTPs with bjts to auto balance the bias of a pair of output tubes. I have tried almost exactly what they have there with PNP bjts and found they work extremly well until one applies loop NFB, and then the circuit becomes impossible to stabilise at LF. It seems that a normal recommended value of grid leak resistor is fed to the cathode circuit of the other differential side where it is inserted between 2 Rk's of normal calculated value in series with each other and straddled by Ck. I guess this way each diff pair would have twice the cathode resistance to ground and as mentioned kill some B+. I really like the "differential auto balance" concept though. Worth it? Yes. 3 Can OPT based cathode feedback be used in cathode fixed biased PP output stages? No problem, I do it all the time. Old Quad II amps benefit from having the CFB winding CT grounded, with 500ohms + 1,000uF as bias networks between the ends of the CFB windings and the cathodes. Ok, I think I get it. By "CT" you mean grounding the center tap of the CFB part of the OPT right? Yes. If so, then the Blumlein autobias circuit could theoretically be combined with CFB. Not really because there is a fairly high signal of opposite phase at each cathode, so effectively you would at least be halving the time constant of the CR coupling to the driver stage. I don't know how exactly the stability would be affected. But I like a more hands on approach where an LTP using bjt is used to monitor the Ik of each output tube and if the Ik should drift appart more than 3mA then an led lights up to tell you which tube is conducting more current. There is then a pot where the bias can be slightly adjusted so the two leds near each output tube go out when Ik is within 2mA, and swinging the pot will tune the bias balance even closer. Ths adjust mechanism tells you when a tube is playing up, wearing out, and if you cannot adjust the high Ik tube down with a pot you *know* its stuffed, or there is some other problem. Most times where you buy a pair of unmatched tubes you will find that with plain cathode bias of individual tubes that the bias current sets itself within 3% balance, which is quite OK. The sound starts to go to mud when imbalance drifts to 10%, and the levels are high because the OPT core has its iron µ effectivelyreduced so LF signals tend to saturatae the core and cause intermodulation with higher F. Quad II amps are renowned for such problems when the tubes age and many samples I have serviced have 40mA in one output tube and 90mA in the other which sometimes has some red patches on the anode, and the sound becomes bad above 1/2 a watt. Just placing individual biasing reduces the imbalance to within 7% even with tubes so far apart and allows many more years of service from the tubes. Peter Walker really made a huge mistake in his original design imho by addopting his use of a single RC for biasing both output tubes. I have only seen schematics where fixed bias is used and the transformer leads are connected directly to the cathodes. Quad II is cathode bias with CFB, but with a common R and C between CFB CT and 0V. Its a cheap nasty way of doing things. I tend to like cheap and nasty. Is CFB a good form of local FB? It is if well implemented. The idea that Quad used is quite valid. It allows the full tetrode power of KT66 to be realised but the tubes act as though Ra was a lot lower, but yet they remain easy to drive. The more CFB you have, the more grid drive you need, so the more linear your drive amp has to be. McIntosh used equal % of P turns in the anode and cathode winding, so you can have 125Vrms at anode and cathode, and if the tube gain = 15, then you need 17Vrms between cathode and grid to cause the 250Vrms between anode and cathode and so you need 125V + 17V applied to each output grid, and hence the complexity of the McIntosh driver stage with its extra stage. The high drive voltage means the thd contribution of the driver becomes greater than that produced in the output stage so you would need global loop NFB to reduce it down. McIntosh used around 20 Db of global and with the 15Db of local NFB in the output stage their amps have around the same thd as many other amps which don't use large amounts of CFB and just rely on modest global NFB. But where McI amps shine is with the Rout of thier amps; the total applied NFB is around 35dB, so Rout is much lower than many other amps; Quad II is about1 ohm, but McI is about 0.2ohms. McI amps allowed one to obtain 50 watts from a pair of 6L6, a huge amountof power for 1949, and thd and Rout was quite low due to the mainly class B actionof the output tubes. Their operation is substantially improved when they are altered for lowerB+, and class A operation, when the Rout is halved and thd reduced to vanishingly low levels, so that a watt thd 0.01%. Nevertheless, a good Williamson or class A UL amp will also make around 0..02% at a watt and will use one less stage of gain than the McI, and the OPT is a lot simpler. Hence hardly any DIYers made copies of the McI amp; they voted with their wallets for the simpler Wiliamson, or the Mullard 520. Very few have copied the Quad examples. I believe it to be worth the trouble, and its the only way i make PP amp but I use a higher % of cfb than Quad ever did. It would seem to be quite tidy with primarily the reactive parts of the output tube being in the feedback loop and by stepping up the damping factor only at the output stage where it suffers the worst impedance mismatch. The rest of the stages in the chain I am considering have 100 ratio of Zin/Zout thanks to 6H30's. CFB allows the output tubes to be substantially linearized via a short feedback loop involving only the OPT and the output tubes. The reactive elements of the output tube are negligible since the Miller C is low, and you only have *one* lot of Miller C, not 3 lots, which is the case with a 3 stage amplifier. Therefore the use of moderate CFB, say up to 25% of the total signal voltage between a and k, is a great way of making the output stage behave well and the global FB can be reduced to say between 6 and 12dB which is a low amount. Quad II amps without their global NFB loop and when using the 8 ohm outlet config produce an Rout = 9 ohms. The Rout from pure beam tetrode output stage with KT66 would be about 132 ohms, so that 10% of CFB Quad uses is *very*useful and the effect on the thd is also very beneficial. The global NFB then reduces the Rout to around an ohm for a reasonable damping factor. The idea is that the majority of signal error correction by NFB is carried out in the output stage and not in the driver amp. The Quad method of OPT set up is as effective as the use of the 40% of UL tappings which has about the same effect as the 10% of CFB, except that perhaps the Quadmethod is more efficient because the feedback is mainly applied in the grid-cathode loop, with also some slight feedback effect in the grid-screen circuit, wheras the UL circuit has allits feedback applied via the anode-screen circuit. Triode connection of tubes with screens is a case where *all* the anode voltage is used as feedback in the screen circuit, ie, the UL tappings are at 100%. The trouble with triodes is that unless they are pushed into class A2 or AB2, the output power is severely limited to half what the tube can deliver, so the UL connection was tried to allow the tube to approach triode fidelity but without the powerloss penalty. CFB achieves the same result, only slightly better, ie, Rout and thd are slightly lower than UL circuits where power output is still near maximum for pure tetrode and the output stage sensitivity is about the same; ie, the grid voltage applied to the output stage is about the same for the two ways of connecting the tubes. Quad also liked to be different, to have the best, and they patented their 10% of CFB, so everyone else had to make do with something less, but bean counters loved UL because it was so much easier to make an OPT without a separate tertiary winding devoted to CFB, and probably easier to get the OPT to have low leakage inductance in the right places to allow good stablity; Leak had only 3P sections and two S sections, and got away with it. But of course all the cheap and nasty bean counter driven OPT designs that were the hallmark of a lousy minded consumer audio product industry of 1955 can be easily improved upon, and its for this reason i regard Leaks and Quads as very mediocre amplifiers, things that can and do sound well when properly set up but which all measure quite poorly by today's standards where we have good sound and less winding losses, better stability, and more reliablility, so fewer OPT write offs every time an output tube died in a vesuvious of over current excitement. The Dynaco ST80 was one of the last Dynaco tube amps to be made, and its vastly superior to its predessor, the ST70, and anything made by Leak, but you'dexpect that after 20 years. I measured an ST80 they other day, and its OPTs are a lot better than in the ST70, and its quite stable, even without any R load, and using just a 0.47uF or0.22 uF cap as a load. Dynaco finally got with it, and did the required homework to get their circuit to work properly, after spending the extra $1 for a bit more irn and wire, and $2for the extra labour. Finally, I have cathode bias in my 300 watt amps for the 12 output tubes. This saves having 24 bias adjustments for 2 channels. In the latest versions i use a patentable and little understood method of shunting excess cathode signal current with a simple two bjt circuit that stops the change of dc cathode voltage that occurs when you apply high signal levels to a class AB amp with cathode bias. So my 300 watt amps work exactly like fixed bias amps with their lower thd ability; the cathode bias voltage changes less than 10% between idle and full power with a sine wave, and while in class A the ac signal bypass circuit remains utterly dormant since it only works when the cathode current change goes over the twice the idle bias current. No need to have active biasing of the grid circuits. Class AB cathode bias amps suffer when driven to the point where musical transients cause the bias voltage at the cathode to begin undulating around, and thedc currents lurch about; but not in my amps. I did post a circuit of the basic approach at abse about 2 years ago, butnobody cared, understood, or tried it out, so I won't be giving any other details now, so youse can all go and invent the solution to easy biasing and dynamic regulation on your own, now that I have spelled out the ideas. Patrick Turner. Patrick, Thanks for the valuable info. This is exactly what I asked for and a great read with some history. I've saved it. You mentioned UL. Perhaps a dumb question since I have never seen this, but why can't UL leads be used as CFB; perhaps with some attenuation? Or is the phase wrong at 43%. Does this imply that a CFB coil has an equal # of turns as the primary? BTW: With this continuing first PP EL34 project of mine, I don't like UL at all compared to triode no matter how the output has been driven. I can't imagine ever going back (at least not in hi-fi) despite the power penalty. Wessel |
#7
|
|||
|
|||
A few general PP power stage questions
Wessel Dirksen wrote: Patrick Turner schreef: Wessel Dirksen wrote: Patrick Turner schreef: Wessel Dirksen wrote: Hi RAT's: Below are a few questions that don't seem to be specifically addressed in the references I have: 1 When using multiple output tubes in parallel on each differential side with cathode autobias, can you theoretically use one coupling cap, and/or grid load resistance, and/or one grid stopper resistance for all of them as a group? Seems likely. If so, would you practically want to, or not want to, for any reason? You need separate grid stoppers for each tube. But nothing wrong with a 2uF cap and a 68k resistor to bias say 4 parallel EL34. But each EL34 should have its own Rk and Ck so bias regulation is good for each tube. In old amps where just one bias Rk and Ck are used the bias currents can vary by 100%. 2 Has anybody used the Blumlein "garter" cathode bias method? I've seen it mentioned but have not seen anyone expound on it. It seems like a very nice and simple biasing method. Is it not so that once you are in the general neighborhood of the desired idle current that relative fine balance between both diff pairs is what it is all about? The sox and trousers method of bias is the one I'd stay with if i were you. Without the sox and trousers adjusted properly, a lotta instability can ensue. Thanks Patrick and Sander, Patrick, Sox and trousers? That's funny (Sox=Rk, Trousers=Ck ?) Let me see if I understand you correctly. Combining the coupling capacitance and grid leak resistance is OK to multiple output tubes, but grid stoppers and the autobias cathode parts should kept individual per tube? That's correct. The Blumlein circuit is described here in the middle of the web page. The other circuits described look good but too complex. http://www.tubecad.com/2005/May/blog0046.htm I was hoping to get a more in depth expert opinion on this circuit. Blumlien shows genius yet again. The garter circuit does improve matters of dc balance quite substantially, and in a worthwhile amount. But what does one do if you have a dozen output tubes, 6 per side of the PP circuit? With two tubes all is ok, but balance is impossible with 12 tubes, so some active method to equalise bias current would ahve to be used. I was just joking about the trousers and sox bit. But before leaving for the office tomorrow morning, try cutting one leg off your trousers at the knee, and wearing a long pink sock on the exposed leg. It'd be a miracle if your boss doesn't threaten you with the sack during the day. Conventions are silly eh? The above website is **very** interesting, and it has some bias circuits using LTPs with bjts to auto balance the bias of a pair of output tubes. I have tried almost exactly what they have there with PNP bjts and found they work extremly well until one applies loop NFB, and then the circuit becomes impossible to stabilise at LF. It seems that a normal recommended value of grid leak resistor is fed to the cathode circuit of the other differential side where it is inserted between 2 Rk's of normal calculated value in series with each other and straddled by Ck. I guess this way each diff pair would have twice the cathode resistance to ground and as mentioned kill some B+. I really like the "differential auto balance" concept though. Worth it? Yes. 3 Can OPT based cathode feedback be used in cathode fixed biased PP output stages? No problem, I do it all the time. Old Quad II amps benefit from having the CFB winding CT grounded, with 500ohms + 1,000uF as bias networks between the ends of the CFB windings and the cathodes. Ok, I think I get it. By "CT" you mean grounding the center tap of the CFB part of the OPT right? Yes. If so, then the Blumlein autobias circuit could theoretically be combined with CFB. Not really because there is a fairly high signal of opposite phase at each cathode, so effectively you would at least be halving the time constant of the CR coupling to the driver stage. I don't know how exactly the stability would be affected. But I like a more hands on approach where an LTP using bjt is used to monitor the Ik of each output tube and if the Ik should drift appart more than 3mA then an led lights up to tell you which tube is conducting more current. There is then a pot where the bias can be slightly adjusted so the two leds near each output tube go out when Ik is within 2mA, and swinging the pot will tune the bias balance even closer. Ths adjust mechanism tells you when a tube is playing up, wearing out, and if you cannot adjust the high Ik tube down with a pot you *know* its stuffed, or there is some other problem. Most times where you buy a pair of unmatched tubes you will find that with plain cathode bias of individual tubes that the bias current sets itself within 3% balance, which is quite OK. The sound starts to go to mud when imbalance drifts to 10%, and the levels are high because the OPT core has its iron µ effectively reduced so LF signals tend to saturatae the core and cause intermodulation with higher F. Quad II amps are renowned for such problems when the tubes age and many samples I have serviced have 40mA in one output tube and 90mA in the other which sometimes has some red patches on the anode, and the sound becomes bad above 1/2 a watt. Just placing individual biasing reduces the imbalance to within 7% even with tubes so far apart and allows many more years of service from the tubes. Peter Walker really made a huge mistake in his original design imho by addopting his use of a single RC for biasing both output tubes. I have only seen schematics where fixed bias is used and the transformer leads are connected directly to the cathodes. Quad II is cathode bias with CFB, but with a common R and C between CFB CT and 0V. Its a cheap nasty way of doing things. I tend to like cheap and nasty. Is CFB a good form of local FB? It is if well implemented. The idea that Quad used is quite valid. It allows the full tetrode power of KT66 to be realised but the tubes act as though Ra was a lot lower, but yet they remain easy to drive. The more CFB you have, the more grid drive you need, so the more linear your drive amp has to be. McIntosh used equal % of P turns in the anode and cathode winding, so you can have 125Vrms at anode and cathode, and if the tube gain = 15, then you need 17Vrms between cathode and grid to cause the 250Vrms between anode and cathode and so you need 125V + 17V applied to each output grid, and hence the complexity of the McIntosh driver stage with its extra stage. The high drive voltage means the thd contribution of the driver becomes greater than that produced in the output stage so you would need global loop NFB to reduce it down. McIntosh used around 20 Db of global and with the 15Db of local NFB in the output stage their amps have around the same thd as many other amps which don't use large amounts of CFB and just rely on modest global NFB. But where McI amps shine is with the Rout of thier amps; the total applied NFB is around 35dB, so Rout is much lower than many other amps; Quad II is about 1 ohm, but McI is about 0.2ohms. McI amps allowed one to obtain 50 watts from a pair of 6L6, a huge amount of power for 1949, and thd and Rout was quite low due to the mainly class B action of the output tubes. Their operation is substantially improved when they are altered for lower B+, and class A operation, when the Rout is halved and thd reduced to vanishingly low levels, so that a watt thd 0.01%. Nevertheless, a good Williamson or class A UL amp will also make around 0.02% at a watt and will use one less stage of gain than the McI, and the OPT is a lot simpler. Hence hardly any DIYers made copies of the McI amp; they voted with their wallets for the simpler Wiliamson, or the Mullard 520. Very few have copied the Quad examples. I believe it to be worth the trouble, and its the only way i make PP amp but I use a higher % of cfb than Quad ever did. It would seem to be quite tidy with primarily the reactive parts of the output tube being in the feedback loop and by stepping up the damping factor only at the output stage where it suffers the worst impedance mismatch. The rest of the stages in the chain I am considering have 100 ratio of Zin/Zout thanks to 6H30's. CFB allows the output tubes to be substantially linearized via a short feedback loop involving only the OPT and the output tubes. The reactive elements of the output tube are negligible since the Miller C is low, and you only have *one* lot of Miller C, not 3 lots, which is the case with a 3 stage amplifier. Therefore the use of moderate CFB, say up to 25% of the total signal voltage between a and k, is a great way of making the output stage behave well and the global FB can be reduced to say between 6 and 12dB which is a low amount. Quad II amps without their global NFB loop and when using the 8 ohm outlet config produce an Rout = 9 ohms. The Rout from pure beam tetrode output stage with KT66 would be about 132 ohms, so that 10% of CFB Quad uses is *very* useful and the effect on the thd is also very beneficial. The global NFB then reduces the Rout to around an ohm for a reasonable damping factor. The idea is that the majority of signal error correction by NFB is carried out in the output stage and not in the driver amp. The Quad method of OPT set up is as effective as the use of the 40% of UL tappings which has about the same effect as the 10% of CFB, except that perhaps the Quad method is more efficient because the feedback is mainly applied in the grid-cathode loop, with also some slight feedback effect in the grid-screen circuit, wheras the UL circuit has all its feedback applied via the anode-screen circuit. Triode connection of tubes with screens is a case where *all* the anode voltage is used as feedback in the screen circuit, ie, the UL tappings are at 100%. The trouble with triodes is that unless they are pushed into class A2 or AB2, the output power is severely limited to half what the tube can deliver, so the UL connection was tried to allow the tube to approach triode fidelity but without the power loss penalty. CFB achieves the same result, only slightly better, ie, Rout and thd are slightly lower than UL circuits where power output is still near maximum for pure tetrode and the output stage sensitivity is about the same; ie, the grid voltage applied to the output stage is about the same for the two ways of connecting the tubes. Quad also liked to be different, to have the best, and they patented their 10% of CFB, so everyone else had to make do with something less, but bean counters loved UL because it was so much easier to make an OPT without a separate tertiary winding devoted to CFB, and probably easier to get the OPT to have low leakage inductance in the right places to allow good stablity; Leak had only 3P sections and two S sections, and got away with it. But of course all the cheap and nasty bean counter driven OPT designs that were the hallmark of a lousy minded consumer audio product industry of 1955 can be easily improved upon, and its for this reason i regard Leaks and Quads as very mediocre amplifiers, things that can and do sound well when properly set up but which all measure quite poorly by today's standards where we have good sound and less winding losses, better stability, and more reliablility, so fewer OPT write offs every time an output tube died in a vesuvious of over current excitement. The Dynaco ST80 was one of the last Dynaco tube amps to be made, and its vastly superior to its predessor, the ST70, and anything made by Leak, but you'd expect that after 20 years. I measured an ST80 they other day, and its OPTs are a lot better than in the ST70, and its quite stable, even without any R load, and using just a 0.47uF or 0.22 uF cap as a load. Dynaco finally got with it, and did the required homework to get their circuit to work properly, after spending the extra $1 for a bit more irn and wire, and $2 for the extra labour. Finally, I have cathode bias in my 300 watt amps for the 12 output tubes. This saves having 24 bias adjustments for 2 channels. In the latest versions i use a patentable and little understood method of shunting excess cathode signal current with a simple two bjt circuit that stops the change of dc cathode voltage that occurs when you apply high signal levels to a class AB amp with cathode bias. So my 300 watt amps work exactly like fixed bias amps with their lower thd ability; the cathode bias voltage changes less than 10% between idle and full power with a sine wave, and while in class A the ac signal bypass circuit remains utterly dormant since it only works when the cathode current change goes over the twice the idle bias current. No need to have active biasing of the grid circuits. Class AB cathode bias amps suffer when driven to the point where musical transients cause the bias voltage at the cathode to begin undulating around, and the dc currents lurch about; but not in my amps. I did post a circuit of the basic approach at abse about 2 years ago, but nobody cared, understood, or tried it out, so I won't be giving any other details now, so youse can all go and invent the solution to easy biasing and dynamic regulation on your own, now that I have spelled out the ideas. Patrick Turner. Patrick, Thanks for the valuable info. This is exactly what I asked for and a great read with some history. I've saved it. You mentioned UL. Perhaps a dumb question since I have never seen this, but why can't UL leads be used as CFB; perhaps with some attenuation? Or is the phase wrong at 43%. Does this imply that a CFB coil has an equal # of turns as the primary? The CFB *is part* of the primary. Only in McIntosh amp does the turns used in the cathode windings equal those in the anode windings. You can use a UL OPT so that the 43% tapps are electrolytic cap coupled to the cathodes which are each grounded through a separate choke with CT. Say each 1/2 primary is part A and part B, each with opposite phased signals. Then UL tap from A goes to the screen in the tube with its anode to B, and UL tap from B goes to screen in the tube with its anode to A. So its cross coupling; the phases are right for CFB but its not as wonderful as it may look because you have a very high amount of CFB, say -70Vrms at the cathode and +168Vrms at the anode, so if you need 16V between g1 and k to get the 238 total a to k signal volts then you need 86vrms of drive to the grid. BTW: With this continuing first PP EL34 project of mine, I don't like UL at all compared to triode no matter how the output has been driven. I can't imagine ever going back (at least not in hi-fi) despite the power penalty. I don't know anyone who has been able to detect a difference in the sound between using triode or UL connected tubes where the amount of FB was such that the Ro of the amps was the same. Its all good. Using 12.5% CFB with a pair of 6550 give an Ro without global FB of about 3 ohms, somewhat lower than 40% UL which would give about 6 ohms using the same turn ratio and a-a loading. Global NFB is then required. Pure triode without global FB will give Ro = about 1.5 to 2 ohms, and that's a bit high, so I like to use 12 dB of global FB, and all this became patently clear when i tried to use EL34 in a pair of Quad II amps in triode. The OPT winding losses in 8 ohm config is 17%, so Rout due to winding resistance is 1.4 ohms eeven if the ra of the tubes was zero, so Rout with triodes fitted to Quad II amps without FB is dissapointing, bearing in mind the impedance ration of 4k : 8 ohms, so the Ra-a of the 2.5 k of ra-a of the tubes reduced to 4.8 ohms, to which we have to add the 1.4 ohms so final Rout without any FB = a massive 6.2 ohms. This is reduced somewhat by the CFB, but triode gain is low, so the amount of CFB with CFB = 10% is only around 3dB, so Rout with CFB and no glbal FB is about 5 ohms, so the CFB merely compensates for the winding resistance of the OPT. With only 6dB of global, the bass was woolly, dynamics poor in comparison the the other amp the guy owned which ran smaller 6GW8 UL with 18 db FB. The little 6GW8 10 watter was definately better than the 13 watt EL34 triode amps. When i replaced the EL34 with Sovtek KT88 in triode and applied 12 dB of fb to allow 20 watts in class AB triode from these Quad amps they began to sing as well as the smaller amp, and the owner then also enjoyed the better headroom with the larger tubes. The load for triodes is critical, and the 4k a-a for triode EL34 is too low; one should have 10k. But with KT88, 4k is OK. Rout is below 1 ohm, just, and thd lower than the original Quad II. The driver amp is all different, using a pair of twin triodes with SET input + LTP rather than the pair of EF86. I have since revised my assumption that triode drivers are always better and the last pair of Quad II retained the two EF86, but were arranged with a long tail taken to -400V with a resistor, so the pair are operated as a true LTP rather than a floating paraphase, and I gote extremely nice sound using the original config of KT66, but with a **much** better power supply plus the separate cathode biasing and bias balance adjustment. You gotta weigh things up without too much prejudice, but good sound can be had from triode amps but the best of it is the result of hard work. Patrick Turner. Wessel |
#8
|
|||
|
|||
A few general PP power stage questions
Wessel Dirksen wrote:
Hi RAT's: Below are a few questions that don't seem to be specifically addressed in the references I have: 1 When using multiple output tubes in parallel on each differential side with cathode autobias, can you theoretically use one coupling cap, and/or grid load resistance, and/or one grid stopper resistance for all of them as a group? Seems likely. If so, would you practically want to, or not want to, for any reason? 2 Has anybody used the Blumlein "garter" cathode bias method? I've seen it mentioned but have not seen anyone expound on it. It seems like a very nice and simple biasing method. Is it not so that once you are in the general neighborhood of the desired idle current that relative fine balance between both diff pairs is what it is all about? 3 Can OPT based cathode feedback be used in cathode fixed biased PP output stages? I have only seen schematics where fixed bias is used and the transformer leads are connected directly to the cathodes. The one to 10 K, 1/2 watt resistors are in the grid circuits to prevent parasitic oscillation at RF, since most power tubes still have lots of gain at several MHZ. The stray C & L in the wiring form resonant circuits at frequencies well into the RF. The R is in the circuit so that Q is lowered & oscillation prevented (we hope). Cheers, John Stewart |
#9
|
|||
|
|||
A few general PP power stage questions
Patrick,
Aha, I would like to try this because my OPT has 33% UL taps (as it turns out) which are not being used. After giving this some thought while waiting for your response, I had partly figured out that decoupling and attenuation of FB would be necessary. But I am not sure what you mean about the whole business of grounding with a choke between the cathodes and CT. 1 Undoubtedly I'm overlooking something and am anxious to learn why, but why is a choked connection to ground needed via the CT? 2 I understand the crossed UL leads and the decoupling bit, but how do the cathodes, chokes, CT, and ground connect together in this setup with UL leads as CFB? Thanks again, Wessel |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
rec.audio.car FAQ (Part 1/5) | Car Audio | |||
A Strawman, Constructed and Destroyed-Williamson's Folly? | Audio Opinions | |||
A Strawman, Constructed and Destroyed-Williamson's Folly? | Vacuum Tubes | |||
KISS 121 by Andre Jute | Vacuum Tubes | |||
rec.audio.car FAQ (Part 1/5) | Car Audio |