Home |
Search |
Today's Posts |
#1
|
|||
|
|||
27W PP TRIODE AMP May 2005.
I just posted a schematic for a PP class AB1 PP triode amp
at alt.binaries.schematics.electronic. It has several features to explain ideas I use all the time. There are two different CCS, each one using npn or pnp transistors, and an LTP driver using trioded EL84 with a CT choke and additional RL to keep the impedance of the DC supply elements very high between 50Hz to above 20 kHz . The CCS used for an active load for the 12AU7 will reduce thd by at least 8 db over a plain R load. The CCS used for the CCS from common k of the EL84 LTP to the negative supply will balance the LTP automatically. All the possible LF and HF phase tweaking gain tailoring networks are shown for stabilising the amp without compromising the HF performance and without causing any stage to saturate between 16 Hz and 65 kHz at full power providing the OPT is a half decent. It is possible to use up to 12 output tubes in parallel PP with exactly the same input and driver tubes. Possibly a pair of 6H30 could be used in lieu of the EL84 in the LTP, perhaps at 10 mA per tube, not 12mA, as i suggest, and with R18 and R19 increased somewhat. Patrick Turner. |
#2
|
|||
|
|||
|
#3
|
|||
|
|||
Hi Patrick,
I suppose the center tapped choke has an inductance of 100H-CT-100H? Just for the records and since I am looking for a very wide swing PP frontend, what peak-peak swing could be reasonably expected from this LTP? Tnx2u, Tom -- The opposite of a correct statement is a false statement. But the opposite of a profound truth may well be another profound truth. - Niels Bohr |
#4
|
|||
|
|||
Just for the records and since I am looking for a very
wide swing PP frontend, what peak-peak swing could be reasonably expected from this LTP? Looking at the plate curves and assuming an almost horizontal load line due to the choke, this would be in excess of 300 Vpp even when allowing for a -Vg1 safety margin. But since the load line is not horizontal in real world due to the load being driven, I am wondering wether a few more mAs, say, 20 for each EL84, would reduce THD even more? Tom -- The first rule of magick is simple: Don't waste your time waving your hands and hoping, when a rock or club will do. |
#6
|
|||
|
|||
Yves wrote: "Patrick Turner" a écrit dans le message de news: ... I just posted a schematic for a PP class AB1 PP triode amp at alt.binaries.schematics.electronic. It has several features to explain ideas I use all the time. There are two different CCS, each one using npn or pnp transistors, and an LTP driver using trioded EL84 with a CT choke and additional RL to keep the impedance of the DC supply elements very high between 50Hz to above 20 kHz . The CCS used for an active load for the 12AU7 will reduce thd by at least 8 db over a plain R load. The CCS used for the CCS from common k of the EL84 LTP to the negative supply will balance the LTP automatically. All the possible LF and HF phase tweaking gain tailoring networks are shown for stabilising the amp without compromising the HF performance and without causing any stage to saturate between 16 Hz and 65 kHz at full power providing the OPT is a half decent. It is possible to use up to 12 output tubes in parallel PP with exactly the same input and driver tubes. Possibly a pair of 6H30 could be used in lieu of the EL84 in the LTP, perhaps at 10 mA per tube, not 12mA, as i suggest, and with R18 and R19 increased somewhat. Patrick Turner. Hi Pat ! Someone MUST start to comment your job since gurus seem ignoring that thing wich not uses a 300b nor even a 6SN7, but rather relies on tiny noval ****s, not to tell about incongruate SS ;)) So, for those who like DHT, let them use 2A3 for the LTP, and 300B for the outputs. But the input would need to have more gain.... Really clever dicks would have a choke + R from LTP cathodes to 0V in place of a CCS. And maybe they could use paraphase drive to one of the LTP triodes, to make more gain, rather than ground one grid like I have. All sorts of things are possible! So, I pull the choke and turn the key. Great job. The only eye scratching is the huge negative supply where the CCS returns. I undertand that it is somewhat "already available" for biasing the PP, but at much lower current ! Well its easy to find a small tranny with 50v, or use a doubler of 25vrms to get -70v at up to say 100 mA; that's only 7VA. Instead of using small cap values for the -70v supply, I recommend CRC filter with 470 uF, a choke of 2H, and another 470 uF. The CCS supply needn't be completely filtered since the CCS tail is megohms of R, and the common cathode Rin is hundreds of ohms, so PS noise does not get into the LTP. A dozen of volts is more than enough and don't really need to be "well filtered", this could reduce the heat generated by the CCS. On the other hand, with more than 70v accross, perhaps an hi perveance pentode could do the job ? Dunno ! The single bjt is sonically silent, effective, simple, and cheap. The larger the Re for the bjt, the higher the value of actual collector resistance; its far higher than using any damn pentode. In true, this was just to tell something, the only point I wish you enlight is the detail of building this "TURNER SPECIAL" center taped choke. The 'Turner LTP' name tag has been attatched to my way of building an LTP. The choke is an ungapped choke with a CT. Its quite easy to get a sufficient amount of inductance without a gap. The CT chokes I used in my 300 watters has 25mm stack of 25mm tongue GOSS material, µ max = 5,000, taken from some old trannies I stripped down. The bobbin is a preformed plastic type with a divider in the centre as commonly used for small mains trannies wound with primary one side and secondary on the other. I placed an equal number into each space using 0.2 mm dia wire, using random winding carefully built up to fill the bobbin. I got over 200H of L, plenty when Ra is so low. The load the EL84 have due to the 8.2k and L is over hundreds of kohms at 1 kHz, so the main load the EL84 see is the 100k following grid bias R. the bias R could be lower, but then the coupling caps of 0.47 may need increasing to 2 uF. The 8.2k R act to give at least 8K2 load at DC and HF, where the Lp and Cshunt would otherwise cause a 90 degree phase shift. The 8k2 prevent the ultimate phase shifts occuring, and a large amount of NFB can be used. But where NFB isn't to be used, or only 6 dB of NFB, the 8k2 may be be omitted. When that is done the output Z of the LTP is lower, since the two anodes are coupled closely and magnetically. The inductance of the choke varies like all ungapped chokes with applied voltage and F, like a primary of an OPT. So the more turns one uses, the greater the L, and in fact if both lots of turns and a large core is used the choke can be gapped anyway, and this makes the choke have a more stable value of L. The choke I used had about 2,500 turns to each 1/2 bobbin section. So L max was about 650H. At 100Hz the ZL = 400kohms. I also used chokes on my latest 8585 amp which EL84 LTP with a choke using about 8,000 turns of 0.15 dia wire around 10mm stack of 25mm tongue, and that works just fine. The 0.2 mm dia thicker wire I recommend makes the choke more rugged. If a tube saturates the 8k2 resistors develop a maximum short circuit I = 55mA, and the coil won't burn out. 0.2 mm wire is rated for 94 mA at 3 amps per sq.mm. This LTP will have much lower 3H compared to all normal LTP using the same tubes and current. Like all LTP, the Ro of each half is high when measure separately, so if such an arrangement is used to drive an AB2 amp the use of another CF stage is neccessary. But then the need to have fairly high idle current in the LTP gain block is reduced since Cin of a CF stage is very low, so ordinary R loaded LTP is OK. The same choke can then be used to drive an output pair with each 1/2 winding having is own bias voltage to each grid of the OPV, thus eliminating the RC grid coupling. The CF drive with a choke gives good regulation of the bias voltage in fixed bias AB2 amp, and can cope very well with grid current. Since the Ro of the CF is so low, the LF phase shift caused by the choke does not occur until an extremely LF is reached. There must be 50 ways to build a driver..... Another way to use EL84 in an LTP is to run them in pentode, which would make their gain over about 150, and then use NFB to the "idle" input grid from a R divider taken from the appropriate following grid bias R. Instead of increasing gain like one does with a normal paraphase phase inverter, which really is an application of a mild amount of local positive feedback, one can apply local NFB, and reduce the gain to around what one enjoys with the simpler triode connection. The pentodes without NFB would be terrible, since the thd and gain would be quite excessive. Also 6EJ7 in triode would be rather good. Patrick Turner. Cheers, Yves. |
#7
|
|||
|
|||
Tom Schlangen wrote: Hi Patrick, I suppose the center tapped choke has an inductance of 100H-CT-100H? I reckoned 650H a-a max, but 100H a-a is ok. My other post has more about the details. Just for the records and since I am looking for a very wide swing PP frontend, what peak-peak swing could be reasonably expected from this LTP? About 110vrms, or 300V p-p. The trioded EL 84 are operating like they would in a class A power amp with a very high a-a RL. The possible low thd V swing is gi-normous. Patrick Turner. Tnx2u, Tom -- The opposite of a correct statement is a false statement. But the opposite of a profound truth may well be another profound truth. - Niels Bohr |
#8
|
|||
|
|||
Tom Schlangen wrote: Just for the records and since I am looking for a very wide swing PP frontend, what peak-peak swing could be reasonably expected from this LTP? Looking at the plate curves and assuming an almost horizontal load line due to the choke, this would be in excess of 300 Vpp even when allowing for a -Vg1 safety margin. But since the load line is not horizontal in real world due to the load being driven, I am wondering wether a few more mAs, say, 20 for each EL84, would reduce THD even more? The load on the EL84 in this case is about 90k for each tube, so the load line is nearly flat, and there is little to be gained by using more Ia. Patrick Turner. Tom -- The first rule of magick is simple: Don't waste your time waving your hands and hoping, when a rock or club will do. |
#9
|
|||
|
|||
Yves wrote: "Tom Schlangen" a écrit dans le message de news: ... Hi Patrick, I suppose the center tapped choke has an inductance of 100H-CT-100H? Hi Tom, There is a square root missing, but where ? Help Patrick !! Only a square root? 200Ha-a has 50H in each 1/2 section measured separately. But in the circuit it acts as a 200H choke, since current flows at all times in class A. Patrick Turner Just for the records and since I am looking for a very wide swing PP frontend, what peak-peak swing could be reasonably expected from this LTP? Tnx2u, Tom -- The opposite of a correct statement is a false statement. But the opposite of a profound truth may well be another profound truth. - Niels Bohr |
#11
|
|||
|
|||
Yves wrote: "Patrick Turner" a écrit dans le message de news: ... [ . . . ] The bobbin is a preformed plastic type with a divider in the centre as commonly used for small mains trannies wound with primary one side and secondary on the other. I placed an equal number into each space using 0.2 mm dia wire, using random winding carefully built up to fill the bobbin. I got over 200H of L, plenty when Ra is so low. The load the EL84 have due to the 8.2k and L is over hundreds of kohms at 1 kHz, so the main load the EL84 see is the 100k following grid bias R. the bias R could be lower, but then the coupling caps of 0.47 may need increasing to 2 uF. The 8.2k R act to give at least 8K2 load at DC and HF, where the Lp and Cshunt would otherwise cause a 90 degree phase shift. The 8k2 prevent the ultimate phase shifts occuring, and a large amount of NFB can be used. But where NFB isn't to be used, or only 6 dB of NFB, the 8k2 may be be omitted. [ . . . ] The inductance of the choke varies like all ungapped chokes with applied voltage and F, like a primary of an OPT. Duh ? the inductance varies with F ? Yes. The µ of the iron is at a maximum at some low F, around 30 Hz, and then it falls away to almost nothing above 10 kHz. L = ( 1.26 x Np squared x Afe x µ ) / ( 1,000,000,000 x ML ) L in in henrys, 1.26 is a constant, Np is the turns around the core, Afe = stack height x tongue width = sq.mm, µ = iron µ from the data on the iron for the F concerned, 1,000,000,000 is a constant, ML is the magnetic path length around an ungapped core. The impedance of an inductor does not vary upwards in direct proportion to the rise in F. µ falls as F rises, so at 1 kHz, L will be a lot less than at 30 Hz. But the impedance of the inductor, or the reactance, Z, expressed in ohms = L in henrys x 2 x pye x F. = L x 6.28 x F. So if L = 50H at 1 khz, Z = 50 x 6.28 x 1,000 = 314kohms, a very high figure for a load on the EL84 triode. L also varies with applied voltage, so that at 1v a-a, the choke above may have only 20H at 30 Hz, not 650H at say 200 vrms a-a in the LTP. So the inductance appears to be a non linear impedance during each wave cycle, and its Z is lowest at the zero crossing point of the sine wave. The result is that the tubes will see a slightly varying load with voltage, and that means some 3H distortions, usually below 0.03% at the worst at say 5 vrms output, at the choke. However, the Ra of the triodes is so low, and we have the 8k2 "buffer resistors" in series with the choke, so although there may indeed be considerable 3H at the ends of the choke winding present, the low Ra of the triodes of around 2.2k each will divide the 3H down from a high amount to a low amount of say 0.01% worst case at the triode anodes, and at least less than the triode distortion of say 0.01% at 5 vrms output. At 80 vrms output from each anode of this LTP its possible to see less than 0.3 % thd, mostly 3H, since the 2H mostly gets cancelled. despite the problems with iron distortions and its non linear behaviour, it still acts with great benefits in this type of circuit. With a CF dribing a choke, Ra from EL84 inn triode is effectively reduced to perhaps 150 ohms, and this shunts the non linear impedance of the iron, which varies from 10kohms to maybe 500kohms, which does not matter because whatever ZL is, its a far higher, and unable to affect the low Ro of the CF, or indeed the Ra at the anodes from the LTP. So the more turns one uses, the greater the L, and in fact if both lots of turns and a large core is used the choke can be gapped anyway, and this makes the choke have a more stable value of L. The choke I used had about 2,500 turns to each 1/2 bobbin section. So far so good ! Now I have a starting point, Tnx2u So L max was about 650H. At 100Hz the ZL = 400kohms. I also used chokes on my latest 8585 amp which EL84 LTP with a choke using about 8,000 turns of 0.15 dia wire around 10mm stack of 25mm tongue, and that works just fine. Even smaller than I expected ! It took time to wind though. And its easy to break 0.15 wire. I had little room to place the chokes into the 8585. But they work well. [ . . . ] The same choke can then be used to drive an output pair with each 1/2 winding having is own bias voltage to each grid of the OPV, thus eliminating the RC grid coupling. The CF drive with a choke gives good regulation of the bias voltage in fixed bias AB2 amp, and can cope very well with grid current. Since the Ro of the CF is so low, the LF phase shift caused by the choke does not occur until an extremely LF is reached. There must be 50 ways to build a driver..... Plus 1, of course ! I was just wondering to put the CT choke in the grids of the output stage, single ended driven (thru a cap) by a triode feeded by a single resistor or a CCS. In fact, a classic auto tranny PI ! I hope it has less stringent requirements than a real tranny. (Someone asked me how to drive a 6C33 PP ;)) Where little or no global NFB is used, the CT choke becomes really like the primary of an OPT, and both anodes of a PP driver are locked magnetically together. One can wind a 1:1 IST, with P and S windings each with 5,000 turns, although a larger window is needed in the core, but same total Afe of about 600 sq.mm will do. The P and S windings can be divided into 4 quarter windings, with one CT primary taken to B+ and anodes, and the other CT secondary taken to 0V and the grids. The OPV are cathode biased for class A. The ends of the P and S windings have 2 uF caps to shunt the leakage inductance and you have an excellent PP driver for any PP output stage. GE proposed all this back in 1957. If Ra is low enough, and Lp high enough, its possible to make an amp without coupling caps at all and use only magnetic coupling. Ths way the amp can be made to have no proplems with grid current charging up coupling caps, and class AB2 is easy. What I am proposing is for hi-fi amps, and strictly class AB1 only. For grid current you really always need a CF driver stage. Patrick Turner. [ . . . ] Patrick Turner. Cheers, Yves. |
#14
|
|||
|
|||
Thanks for sending the schematic, Patrick.
If you insist on speakers requiring massive amounts of power, like 27W, you will definitely need an amp like this! I thought it was nicely developed, though I would have tried, for no good technical reason, merely because people are prejucdiced, to hold NFB down to 10dB. I'm very keen on constant current loads by MJE 350 myself; note that I call it a load. It sinks current only when applied to the cathode, as in your MJE340 example. I also like the idea of an EL84 as a driver very much. A sweet little tube. I'm not so keen on fixed bias with adjustment, even in DIY hands. One of the great things about autobias is that the amp is essentially service-free until the power tubes give up the ghost. Fixed bias, except by battery for small signal tubes, should be reserved for bass freaks and other street corner boomboxers. I understand that you essentially present this schemo as a teaching exercise to explain stages you use all the time; to that purpose it is a comprehensive success. However, viewing it as an amp someone serious could build I have to conclude that, overall, this is a tremendous amount of engineering that I wouldn't have done if I were you. Since you know how to design and wind your own transformers, and there are already two pieces of iron in the signal chain on this amp, why not avoid all that engineering and design and components and make the entire thing all iron, zero caps, very few resistors? For instance a step up/splitter transformer into the EL84 grids and another beefier IT with secondaries to the KT88 grids and bingo, the circuit is incredibly simplified, more value has been added, a higher price can be demanded. Your split choke is clever but a full IT would be even cleverer. Brian Sowter many years ago gave me an all-iron coupled PP 211 design which was said at the time to be Kondo's personal favourite, not the SE Ongaku which got all the publicity. Studying that circuit, I understood why. Costing the transformers I grasped much more quickly why more people don't do it! Andre Jute Patrick Turner wrote: I just posted a schematic for a PP class AB1 PP triode amp at alt.binaries.schematics.electronic. It has several features to explain ideas I use all the time. There are two different CCS, each one using npn or pnp transistors, and an LTP driver using trioded EL84 with a CT choke and additional RL to keep the impedance of the DC supply elements very high between 50Hz to above 20 kHz . The CCS used for an active load for the 12AU7 will reduce thd by at least 8 db over a plain R load. The CCS used for the CCS from common k of the EL84 LTP to the negative supply will balance the LTP automatically. All the possible LF and HF phase tweaking gain tailoring networks are shown for stabilising the amp without compromising the HF performance and without causing any stage to saturate between 16 Hz and 65 kHz at full power providing the OPT is a half decent. It is possible to use up to 12 output tubes in parallel PP with exactly the same input and driver tubes. Possibly a pair of 6H30 could be used in lieu of the EL84 in the LTP, perhaps at 10 mA per tube, not 12mA, as i suggest, and with R18 and R19 increased somewhat. Patrick Turner. |
#15
|
|||
|
|||
" wrote:
Yo, Patrick: This sounds interesting but it isn't on the ABSE available to me (Syracuse U). Can you please send the schemo to my mailbox: Thanks. Andre Jute Same at this location. After several daze still nothing at ABSE. Not sure why. Please send along your schemo. Thanx, John Stewart |
#16
|
|||
|
|||
" wrote: Thanks for sending the schematic, Patrick. If you insist on speakers requiring massive amounts of power, like 27W, you will definitely need an amp like this! I thought it was nicely developed, though I would have tried, for no good technical reason, merely because people are prejucdiced, to hold NFB down to 10dB. I'm very keen on constant current loads by MJE 350 myself; note that I call it a load. It sinks current only when applied to the cathode, as in your MJE340 example. I also like the idea of an EL84 as a driver very much. A sweet little tube. I have found the dynamics to be truly crisp and real with EL84 in triode. Its almost exactly equal to 5 x half sections of 6SN7/6CG7. 27 watts is a massive amount of power for some, but for others its barely enough. The LTP all works fine without any grid current. Now as i said, one could deploy less than 12 dB of NFB. In one of my other posts i went on to describe Ro without NFB with 2 x KT88/6550. Ra-a is about 2,200 ohms while operating in class A, which is all we need to worry about here because the AB operation where Ro increases the amp has little function, since 90% of users will only use 2 watts average, and won't make much use of the AB working ability. So with a 6k to 6 ohm OPT, ZR = 1,000:1, so without NFB Ro = 2.2 ohms. ( one needs to add winding losses, not much if kept to a nice low 5% ) KT90 used as triode will have lower Ra-a, so maybe Ro = about 1.5 ohms. Using a quad of EL34, or the better 6CA7 with no change to the OPT ZR, Ra-a will be about 1,200 ohms, so the Ro will also be around equal to 2 x KT90 or around 1.5 ohms. In such a case the power output won't increase much, maybe to 35 watts, but you get slightly more class A and less thd, and more current ability. 10 dB NFB would be fine, and as everyone should know, Ro varies as 1 / ( µ x ß ), not 1 / ( gain x ß ), so 10 dB will reduce Ro about 3 times, and thus a DF of nearly 10 is attained even with 2 x KT88/6550. Now of course if we used 300B, we would perhaps use Ea = 420 v max, and about the same PP power is possible, but the Ra-a is like KT90 = 1,600 ohms and considerably lower than KT88/6550. 300B in PP are quite nice. I'm not so keen on fixed bias with adjustment, even in DIY hands. I agree entirely, and cathode bias is a valid option. In my other circuits the bias voltage is fixed, and only the balance is altered, and a two transistor circuit senses out of balance DC of more than 4 mA in the OPT, and a led turns on if it unbalances. There is a balancing pot with a small knob, and a user sees a glowing led, and adjusts the balance until it is extinguished, and thats that, no meter, no measuring, no worry either. If he runs out of travel and an led won't turn off, the tubes or something else is wrong, and active protection takes care of over-currenting in the tube, ie, twice the bias current for more than 4 seconds. So imbalance in bias current is seen across the room. Plenty of warning, and no sudden Mt Vesuvious in the loungeroom, with a lawsuit to follow. The biasing as I have drawn it is optional. Even though fixed bias is the best performing bias method when testing with a sine wave in an AB1 amp, testing with pink noise or busy rock music with clipping just starting to occur on the peaks usually doesn't cause much change in the Ek of either output triode with cathode bias. Of course such biasing isn't everyone's cup of tea. Allen Wright would scoff at cathode bias, and tell us to have commoned cathodes of the output tubes and use a constant current sink ( a power transistor of some kind ) and no bypass caps, rather like the way my PP driver tubes are set up. Thus one gets ONLY class A from such output tubes, and the very maximum of 2H current cancellations, and Allen says it always sounds better. Ppl are welcome to try these options. One of the great things about autobias is that the amp is essentially service-free until the power tubes give up the ghost. Fixed bias, except by battery for small signal tubes, should be reserved for bass freaks and other street corner boomboxers. I agree, but I have sold a few FB amps and with the active balance indicators, and active protection, I have had no customer problems. I understand that you essentially present this schemo as a teaching exercise to explain stages you use all the time; to that purpose it is a comprehensive success. However, viewing it as an amp someone serious could build I have to conclude that, overall, this is a tremendous amount of engineering that I wouldn't have done if I were you. Since you know how to design and wind your own transformers, and there are already two pieces of iron in the signal chain on this amp, why not avoid all that engineering and design and components and make the entire thing all iron, zero caps, very few resistors? For instance a step up/splitter transformer into the EL84 grids and another beefier IT with secondaries to the KT88 grids and bingo, the circuit is incredibly simplified, more value has been added, a higher price can be demanded. Your split choke is clever but a full IT would be even cleverer. I couldn't agree more with what you are saying, and I suggest ppl try things. The step up tranny phase inverter idea is quite nice. But with any global NFB, there could be LF/HF instability since you have time constants caused by LR, rather than much lower and predictable CR time constants. The design as presented uses a CT choke made using 5,000 turns of 0.2 mm dia wire around a core with 25 stack x 25 tongue wasteless, but suitable C-cores would be ideal if one found some. By the GOSS E&I are not that uncommon, and almost any DIYer could make the CT choke. The chokes is isolated from the EL84 anodes, so the coupling to the output tube grids is very pure, direct, without the effects of LL and shunt C and Lp that one would get with an IST, not to mention the trimming one has to do with a zobel to get the response to be flat, not peaked due to resonances at HF, all rather difficult for a DIYer without a CRO, and hardly any knowledge of filter theory. I might at that as drawn, the LTP has anode loads on each EL84 of around 80 kohms at 1 kHz, and its these loads which cause the balance in the circuit while the RL seen by each tube is equal. In my 300 watters, which need about 75 vrms from each EL84 anode, the anode loads are each 6 x 120k in parallel, or 20k, and the choke + 8k2 load together so high it has negligible loading effect, thus the balance is excellent, and thd very low and the grid bias R low enough to control the phenomena of grids slowly going a bit positive in relation to the actual applied grid bias. Therefore, in hindsight, using say 2 uF coupling caps from the EL84 anodes and say 33 k output tube grid bias R would be quite OK and give a pole at 2.4 Hz, instead of the 3.3 Hz using 0.47 uF and 100k. Brian Sowter many years ago gave me an all-iron coupled PP 211 design which was said at the time to be Kondo's personal favourite, not the SE Ongaku which got all the publicity. Studying that circuit, I understood why. Costing the transformers I grasped much more quickly why more people don't do it! PP 211 in class A1 can make around 40 watts without grid I. And a heck of a lot more po with class AB2. The voltage supply to the grids is easy while in A1, or AB1, but as soon as AB2 is considered, the Ro of the driver amp has to be very low, and a pair of CF driving the IST or some such arrangement would become necessary. I have not dallied with 211, or seen what Sowter suggest, but yeah, a good IST can be worth a good OPT. For AB2 drive, an IST with 1:1 ratio makes sense, because there can be two equal windings, each with a CT, or with a secondary with 2 halves, each taken to a bias voltage for each tube, while the active ends of P and S windings can be capacitor shunted using typically 10 uF, which will shunt the leakage inductance between P&S windings, so that at HF the drive does not have the problems of phase shift and response droop caused by the instaneous grid current load applying itself. Usually 1:1 ratio IST have less response problems than 1:2 designs; so we must be able to produce the full output tube Vswing, but I would always insist that this isn't problematic knowing what we do today, and while we keep away from the low fidelity low performance bean counter inspired designs of 60 years ago. I will try to find time to draw up a few more ideas on the subject tomorrow. One fundemental reason why one *needs* to keep the driver stage of a PP amp as linear as possible is that the thd will mostly be 3H of the wave crest flattening type, so any 3H of the driver adds to that of the output stage. One a;lternative to a choke is to use a much higher B+ for the driver stage than I have indicated, and use larger value DC anode supply resistors. To that purpose, if the PSU has a pair of diodes from a CT HT winding so that say 370-0-370vrms is rectified to make +500v at the output of the CLC filter, then it is possible to make +1,000v using a voltage doubler riding on top of the main anode PS and this 1,000v can be CRC filtered down to about +600v, and if Ea of the EL84 = +250v, then the DC RLs can each be 35k if we also creep the Ia of each EL84 down to 10 mA each. The 35k to each anode is sufficiently low value to make sure good balance is possible due to the cathode CCS. The the bias of the output tubes can be as i have shown and RL of each EL84 anode = 26k, or over 10 x Ra, so thd will remain in the very low zone, and we have not had to wind a darn choke. But we still will have needed to create the highish voltage supply, which is nervous nelly time for diyers. I have done exactly this type of extended supply B+ driver supply for one of my amps which has a sixpack of class AB1 6CA7 driven with a pair of 6SN7 in the LTP, one paralleled tube each side of the LTP. ( 6V6 in triode is terrific also ). While an EL84 or maybe 6V6 is as cheap as a 6SN7, or better grade 6CG7, I reckon there is great value with EL84, which will always be with us at least for the rest of my lifetime. If one is determined to make an all iron coupled PP amp, then the IST is a must. but I still think some R damping or termination of the secondary may be needed, and if so, one R of the two can be used to provide the positive feedback to make the PP input stage into parafeed, so making the amp twice as sensitive as where one uses an LTP with single input. Parafeed with a CCS cathode I supply is still best. If one wants to get rid of the usual input tube, then the driver can be made using your cherished 417A, or perhaps a pair of 12AT7, with both halves paralleled. 12AT7 will thus havev µ = 55, Ra = about 10k for 10 mA in the total of two triodes, so a circuit gain of 45 is realizable, so only 2 vrms input is needed to make 90 vrms a-a output. But 12AT7 isn't the most linear driver when making such highish output voltages, and we could wonder that using an input stage like I have shown with medium µ triode LTP is a better approach, since for one thing, the input tube buffers the LTP input from the source. Then the preamp doesn't have to have such high gain, or be necessary at all, just a gain pot will do, unless we have vinyl. But certainly no line stage preamp is needed with a 3 stage power amp. One further way of building an LTP with better linearity using tubes such as 12AT7 is to use an R divider to derive NFB from the opposite tube that one normally takes the positive FB drive to make a paraphase circuit. The CCS must stay, to get balance, and RL of each tube must be equal, but the gain can be reduced from say 45 to 15, and the thd also reduced. Its still bothersome for the DIYer to get right, so I prefer the simplicity of the LTP with drive to one side only and the use of medium µ triodes. The Miller C with the LTP I have shown is the static Cg-a x gain, which I think is about 12pF x 9.5 = 114pF for triode connection, so there is a fair amount of Cin. The 12AU7 Ra of around 5k will make a pole at about 300kHz, and even higher with C7,R8 stepping networks added....so not a worry. Patrick Turner. Andre Jute Patrick Turner wrote: I just posted a schematic for a PP class AB1 PP triode amp at alt.binaries.schematics.electronic. It has several features to explain ideas I use all the time. There are two different CCS, each one using npn or pnp transistors, and an LTP driver using trioded EL84 with a CT choke and additional RL to keep the impedance of the DC supply elements very high between 50Hz to above 20 kHz . The CCS used for an active load for the 12AU7 will reduce thd by at least 8 db over a plain R load. The CCS used for the CCS from common k of the EL84 LTP to the negative supply will balance the LTP automatically. All the possible LF and HF phase tweaking gain tailoring networks are shown for stabilising the amp without compromising the HF performance and without causing any stage to saturate between 16 Hz and 65 kHz at full power providing the OPT is a half decent. It is possible to use up to 12 output tubes in parallel PP with exactly the same input and driver tubes. Possibly a pair of 6H30 could be used in lieu of the EL84 in the LTP, perhaps at 10 mA per tube, not 12mA, as i suggest, and with R18 and R19 increased somewhat. Patrick Turner. |
#17
|
|||
|
|||
I must go cook shortly, so I'll leave all the side issues you raise for
the others to discuss. That design Brian Sowter gave me for PP all-IT 211 was not his design but drawn by Kondo as a companion amp for the Ongaku; I imagine Sowter got it from Peter Qvortrup of Audio Note UK; Peter sent me, many years ago, a stack of truly eye-opening designs that Kondo (or someone on his staff, more likely) collected in Japan. My own experience with 211 is nowhere near as agreeable as with the superficially similar 845. I think the 845 is a fundamentally better-sounding tube, well worth the contortions required in the drivetrain to get it to work. I wondered about replacing the KT88 with EL34 in your design, a possibibility you now volunteer. Whether the American version is truly better is a matter of opinion. My father-in-law worked for Mullard and I still have a few of the best EL34 ever made left. I can also order good quality EL34 for next-day delivery right out of the RS catalogue, which is a definite consideration. Replacing a 12AU7 with a 12AT7 will affect the sound of the finished article, make it more precise and dryer, given of course that the difference is not swamped in NFB (another thing I have against NFB, its homogenizing tendency). Whether you consider that an improvement is a matter of taste. I happen to like the 12AT7, a far superior tube sonically speaking to the wretched waffling 12AX7 that a whole underclass of designers reach for as a conditioned reflex. I also like the 12AU7 for its warmth. Your remark about the EL84 being so many parallel 6SN7 puts it a civilized yet precise tube in perfect perspective. This unassuming little power tube is one of the great class acts of the world. Are you saying, towards the end of your post, that your centre-tapped choked was chosen over an IT merely because it is easier to wind for some notional DIYer? Or are you saying it is fundamentally superior to an IT? Andre Jute Patrick Turner wrote: " wrote: Thanks for sending the schematic, Patrick. If you insist on speakers requiring massive amounts of power, like 27W, you will definitely need an amp like this! I thought it was nicely developed, though I would have tried, for no good technical reason, merely because people are prejucdiced, to hold NFB down to 10dB. I'm very keen on constant current loads by MJE 350 myself; note that I call it a load. It sinks current only when applied to the cathode, as in your MJE340 example. I also like the idea of an EL84 as a driver very much. A sweet little tube. I have found the dynamics to be truly crisp and real with EL84 in triode. Its almost exactly equal to 5 x half sections of 6SN7/6CG7. 27 watts is a massive amount of power for some, but for others its barely enough. The LTP all works fine without any grid current. Now as i said, one could deploy less than 12 dB of NFB. In one of my other posts i went on to describe Ro without NFB with 2 x KT88/6550. Ra-a is about 2,200 ohms while operating in class A, which is all we need to worry about here because the AB operation where Ro increases the amp has little function, since 90% of users will only use 2 watts average, and won't make much use of the AB working ability. So with a 6k to 6 ohm OPT, ZR =3D 1,000:1, so without NFB Ro =3D 2.2 ohms. ( one needs to add winding losses, not much if kept to a nice low 5% ) KT90 used as triode will have lower Ra-a, so maybe Ro =3D about 1.5 ohms. Using a quad of EL34, or the better 6CA7 with no change to the OPT ZR, Ra-a will be about 1,200 ohms, so the Ro will also be around equal to 2 x KT90 or around 1.5 ohms. In such a case the power output won't increase much, maybe to 35 watts, but you get slightly more class A and less thd, and more current ability. 10 dB NFB would be fine, and as everyone should know, Ro varies as 1 / ( =B5 x =DF ), not 1 / ( gain x =DF ), so 10 dB will reduce Ro about 3 times, and thus a DF of nearly 10 is attained even with 2 x KT88/6550. Now of course if we used 300B, we would perhaps use Ea =3D 420 v max, and about the same PP power is possible, but the Ra-a is like KT90 =3D 1,600 ohms and considerably lower than KT88/6550. 300B in PP are quite nice. I'm not so keen on fixed bias with adjustment, even in DIY hands. I agree entirely, and cathode bias is a valid option. In my other circuits the bias voltage is fixed, and only the balance is altered, and a two transistor circuit senses out of balance DC of more than 4 mA in the OPT, and a led turns on if it unbalances. There is a balancing pot with a small knob, and a user sees a glowing led, and adjusts the balance until it is extinguished, and thats that, no meter, no measuring, no worry either. If he runs out of travel and an led won't turn off, the tubes or something else is wrong, and active protection takes care of over-currenting in the tube, ie, twice the bias current for more than 4 seconds. So imbalance in bias current is seen across the room. Plenty of warning, and no sudden Mt Vesuvious in the loungeroom, with a lawsuit to follow. The biasing as I have drawn it is optional. Even though fixed bias is the best performing bias method when testing with a sine wave in an AB1 amp, testing with pink noise or busy rock music with clipping just starting to occur on the peaks usually doesn't cause much change in the Ek of either output triode with cathode bias. Of course such biasing isn't everyone's cup of tea. Allen Wright would scoff at cathode bias, and tell us to have commoned cathodes of the output tubes and use a constant current sink ( a power transistor of some kind ) and no bypass caps, rather like the way my PP driver tubes are set up. Thus one gets ONLY class A from such output tubes, and the very maximum of 2H current cancellations, and Allen says it always sounds better. Ppl are welcome to try these options. One of the great things about autobias is that the amp is essentially service-free until the power tubes give up the ghost. Fixed bias, except by battery for small signal tubes, should be reserved for bass freaks and other street corner boomboxers. I agree, but I have sold a few FB amps and with the active balance indicators, and active protection, I have had no customer problems. I understand that you essentially present this schemo as a teaching exercise to explain stages you use all the time; to that purpose it is a comprehensive success. However, viewing it as an amp someone serious could build I have to conclude that, overall, this is a tremendous amount of engineering that I wouldn't have done if I were you. Since you know how to design and wind your own transformers, and there are already two pieces of iron in the signal chain on this amp, why not avoid all that engineering and design and components and make the entire thing all iron, zero caps, very few resistors? For instance a step up/splitter transformer into the EL84 grids and another beefier IT with secondaries to the KT88 grids and bingo, the circuit is incredibly simplified, more value has been added, a higher price can be demanded. Your split choke is clever but a full IT would be even cleverer. I couldn't agree more with what you are saying, and I suggest ppl try things. The step up tranny phase inverter idea is quite nice. But with any global NFB, there could be LF/HF instability since you have time constants caused by LR, rather than much lower and predictable CR time constants. The design as presented uses a CT choke made using 5,000 turns of 0.2 mm dia wire around a core with 25 stack x 25 tongue wasteless, but suitable C-cores would be ideal if one found some. By the GOSS E&I are not that uncommon, and almost any DIYer could make the CT choke. The chokes is isolated from the EL84 anodes, so the coupling to the output tube grids is very pure, direct, without the effects of LL and shunt C and Lp that one would get with an IST, not to mention the trimming one has to do with a zobel to get the response to be flat, not peaked due to resonances at HF, all rather difficult for a DIYer without a CRO, and hardly any knowledge of filter theory. I might at that as drawn, the LTP has anode loads on each EL84 of around 80 kohms at 1 kHz, and its these loads which cause the balance in the circuit while the RL seen by each tube is equal. In my 300 watters, which need about 75 vrms from each EL84 anode, the anode loads are each 6 x 120k in parallel, or 20k, and the choke + 8k2 load together so high it has negligible loading effect, thus the balance is excellent, and thd very low and the grid bias R low enough to control the phenomena of grids slowly going a bit positive in relation to the actual applied grid bias. Therefore, in hindsight, using say 2 uF coupling caps from the EL84 anodes and say 33 k output tube grid bias R would be quite OK and give a pole at 2.4 Hz, instead of the 3.3 Hz using 0.47 uF and 100k. Brian Sowter many years ago gave me an all-iron coupled PP 211 design which was said at the time to be Kondo's personal favourite, not the SE Ongaku which got all the publicity. Studying that circuit, I understood why. Costing the transformers I grasped much more quickly why more people don't do it! PP 211 in class A1 can make around 40 watts without grid I. And a heck of a lot more po with class AB2. The voltage supply to the grids is easy while in A1, or AB1, but as soon as AB2 is considered, the Ro of the driver amp has to be very low, and a pair of CF driving the IST or some such arrangement would become necessary. I have not dallied with 211, or seen what Sowter suggest, but yeah, a good IST can be worth a good OPT. For AB2 drive, an IST with 1:1 ratio makes sense, because there can be two equal windings, each with a CT, or with a secondary with 2 halves, each taken to a bias voltage for each tube, while the active ends of P and S windings can be capacitor shunted using typically 10 uF, which will shunt the leakage inductance between P&S windings, so that at HF the drive does not have the problems of phase shift and response droop caused by the instaneous grid current load applying itself. Usually 1:1 ratio IST have less response problems than 1:2 designs; so we must be able to produce the full output tube Vswing, but I would always insist that this isn't problematic knowing what we do today, and while we keep away from the low fidelity low performance bean counter inspired designs of 60 years ago. I will try to find time to draw up a few more ideas on the subject tomorrow. One fundemental reason why one *needs* to keep the driver stage of a PP amp as linear as possible is that the thd will mostly be 3H of the wave crest flattening type, so any 3H of the driver adds to that of the output stage. One a;lternative to a choke is to use a much higher B+ for the driver stage than I have indicated, and use larger value DC anode supply resistors. To that purpose, if the PSU has a pair of diodes from a CT HT winding so that say 370-0-370vrms is rectified to make +500v at the output of the CLC filter, then it is possible to make +1,000v using a voltage doubler riding on top of the main anode PS and this 1,000v can be CRC filtered down to about +600v, and if Ea of the EL84 =3D +250v, then the DC RLs can each be 35k if we also creep the Ia of each EL84 down to 10 mA each. The 35k to each anode is sufficiently low value to make sure good balance is possible due to the cathode CCS. The the bias of the output tubes can be as i have shown and RL of each EL84 anode =3D 26k, or over 10 x Ra, so thd will remain in the very low zone, and we have not had to wind a darn choke. But we still will have needed to create the highish voltage supply, which is nervous nelly time for diyers. I have done exactly this type of extended supply B+ driver supply for one of my amps which has a sixpack of class AB1 6CA7 driven with a pair of 6SN7 in the LTP, one paralleled tube each side of the LTP. ( 6V6 in triode is terrific also ). While an EL84 or maybe 6V6 is as cheap as a 6SN7, or better grade 6CG7, I reckon there is great value with EL84, which will always be with us at least for the rest of my lifetime. If one is determined to make an all iron coupled PP amp, then the IST is a must. but I still think some R damping or termination of the secondary may be needed, and if so, one R of the two can be used to provide the positive feedback to make the PP input stage into parafeed, so making the amp twice as sensitive as where one uses an LTP with single input. Parafeed with a CCS cathode I supply is still best. If one wants to get rid of the usual input tube, then the driver can be made using your cherished 417A, or perhaps a pair of 12AT7, with both halves paralleled. 12AT7 will thus havev =B5 =3D 55, Ra =3D about 10k for 10 mA in the total of two triodes, so a circuit gain of 45 is realizable, so only 2 vrms input is needed to make 90 vrms a-a output. But 12AT7 isn't the most linear driver when making such highish output voltages, and we could wonder that using an input stage like I have shown with medium =B5 triode LTP is a better approach, since for one thing, the input tube buffers the LTP input from the source. Then the preamp doesn't have to have such high gain, or be necessary at all, just a gain pot will do, unless we have vinyl. But certainly no line stage preamp is needed with a 3 stage power amp. One further way of building an LTP with better linearity using tubes such as 12AT7 is to use an R divider to derive NFB from the opposite tube that one normally takes the positive FB drive to make a paraphase circuit. The CCS must stay, to get balance, and RL of each tube must be equal, but the gain can be reduced from say 45 to 15, and the thd also reduced. Its still bothersome for the DIYer to get right, so I prefer the simplicity of the LTP with drive to one side only and the use of medium =B5 triodes. The Miller C with the LTP I have shown is the static Cg-a x gain, which I think is about 12pF x 9.5 =3D 114pF for triode connection, so there is a fair amount of Cin. The 12AU7 Ra of around 5k will make a pole at about 300kHz, and even higher with C7,R8 stepping networks added....so not a worry. Patrick Turner. Andre Jute Patrick Turner wrote: I just posted a schematic for a PP class AB1 PP triode amp at alt.binaries.schematics.electronic. It has several features to explain ideas I use all the time. There are two different CCS, each one using npn or pnp transistors, and an LTP driver using trioded EL84 with a CT choke and additional RL to keep the impedance of the DC supply elements very high between 50Hz to above 20 kHz . The CCS used for an active load for the 12AU7 will reduce thd by at least 8 db over a plain R load. The CCS used for the CCS from common k of the EL84 LTP to the negative supply will balance the LTP automatically. All the possible LF and HF phase tweaking gain tailoring networks are shown for stabilising the amp without compromising the HF performance and without causing any stage to saturate between 16 Hz and 65 kHz at full power providing the OPT is a half decent. It is possible to use up to 12 output tubes in parallel PP with exactly the same input and driver tubes. Possibly a pair of 6H30 could be used in lieu of the EL84 in the LTP, perhaps at 10 mA per tube, not 12mA, as i suggest, and with R18 and R19 increased somewhat. Patrick Turner. |
#18
|
|||
|
|||
"Patrick Turner" wrote
I just posted a schematic for a PP class AB1 PP triode amp at alt.binaries.schematics.electronic. It has several features to explain ideas I use all the time. There are two different CCS, each one using npn or pnp transistors, and an LTP driver using trioded EL84 with a CT choke and additional RL to keep the impedance of the DC supply elements very high between 50Hz to above 20 kHz .... I spent some time explaining the faults of this "design" to Patrick, and got foul-mouthed in return, as is usual if anyone questions his "ideas". I have asked for elucidation and got the usual evasive waffle and random babbling...quotations from RDH, I assume. He sold the prototype regardless, as usual, as if it were proven or even properly tested. Simulation confirms it is just what it seems: a pathetic embarrassment to anyone with even half a notion of circuit design. Patrick's unique selling point is his claim to wind decent traditional chokes and transformers. Convincing daft Australians to shell out for an extra 200H to no real purpose is a nothing but a cynical ploy. Tired of arguing, and less eager now to help a grasping fool. Perhaps that's obvious. So, anyone else prepared to defend the use of two EL84, 200H of differential mode choke, and an SS current source, as a general purpose splitter / driver? cheers, Ian |
#19
|
|||
|
|||
"Yves" wrote
Hi Tom, There is a square root missing, but where ? Help Patrick !! Two windings in series on the same core. Inductance is proportional to square of turns, hence 2^2=4 4*50=200 cheers, Ian From the negation of the negation arises new truth...Hegel, more or less. |
#20
|
|||
|
|||
" wrote: I must go cook shortly, so I'll leave all the side issues you raise for the others to discuss. That design Brian Sowter gave me for PP all-IT 211 was not his design but drawn by Kondo as a companion amp for the Ongaku; I imagine Sowter got it from Peter Qvortrup of Audio Note UK; Peter sent me, many years ago, a stack of truly eye-opening designs that Kondo (or someone on his staff, more likely) collected in Japan. My own experience with 211 is nowhere near as agreeable as with the superficially similar 845. I think the 845 is a fundamentally better-sounding tube, well worth the contortions required in the drivetrain to get it to work. Possibly you are right; superficially i see the 845 as a more triodic power triode than an 845; low µ, low Ra, and far easier to wind an OPT for. Maybe the difficulties of the higher RLa-a for PP 211 and the need for good insulation at B+ = 1250v, that ppl get lazt, and wind less than optimal OPTs, but I am guessing a bit about general traits.... I wondered about replacing the KT88 with EL34 in your design, a possibibility you now volunteer. Whether the American version is truly better is a matter of opinion. The US version is really an EL34 imposter, since 6CA7 is a beam tetrode, not a true pentode, like EL34, and its little nephews, EL84, EL86 etc. But I think the 6CA7 is the best poor man's 6550 I have heard music with. My father-in-law worked for Mullard and I still have a few of the best EL34 ever made left. I can also order good quality EL34 for next-day delivery right out of the RS catalogue, which is a definite consideration. Svetlana are supposed to be good. Sovtek are making Mullard copies that look very close to the real thing that was made in many countries; we had tube factories in Oz which churned our millions of tubes of many types, including EL34, and other ELxx family members. Replacing a 12AU7 with a 12AT7 will affect the sound of the finished article, make it more precise and dryer, given of course that the difference is not swamped in NFB (another thing I have against NFB, its homogenizing tendency). Whether you consider that an improvement is a matter of taste. I happen to like the 12AT7, a far superior tube sonically speaking to the wretched waffling 12AX7 that a whole underclass of designers reach for as a conditioned reflex. I also like the 12AU7 for its warmth. I agree entirely; I myself wouldn't ask more than 10vrms from any 12AX7. I'd never use them as LTP halves like in the mullard 520. I did use X7 as a paralleled input triode in an 8585, but and it sounded warm and nice, but when I replaced it with 12AT7 to do the same job in the same place, the sound had more speed and dynamics. 12AU7, like 6CG7, are OK for inputs and LTP. Your remark about the EL84 being so many parallel 6SN7 puts it a civilized yet precise tube in perfect perspective. This unassuming little power tube is one of the great class acts of the world. µ = 20, Ra = 2,200 at 14 mA, and if you had 5 halves of 6SN7 each with µ = 20, and Ra = 10k at 2.8 mA, then the EL84 is like the 5 halves of SN7. I would guess that if one considers the screen to grid1 to cathode dimensions and the relative grid wire pitch, the relationship would be remarkably similar to 6SN7. The later versions of 6SN7 had about 25mm long folded box sections for anodes, and two of these were crammed into a mini 9 pinner botle to make the 6CG7, and all that changed was a screen between triodes, and a derating on the maximum Pd. Same tube otherwise though. Telefunken made 6CG7 with plate structures of folded metal, about 1/2 the size of the larger Mullard versions. The Teles are the best I have heard; those germans knew how to make great tubes. If you can make a 6CG7 which has half the physical size to another brand, yet Ra, µ, and gm are all the same, then one could make a larger plated 6SN7. With slight fiddling with spacings, Ra can be lower, gm higher, and the same useful µ, and it seems that µ = 20 is about an optimal value for linearity. Are you saying, towards the end of your post, that your centre-tapped choked was chosen over an IT merely because it is easier to wind for some notional DIYer? Or are you saying it is fundamentally superior to an IT? It depends how it is used of course. The CT choke + CCS + RC coupling suits only class A1, AB1 amps where no grid current flows to cause a dramatic rise in thd. I cannot see any superiority in an IST over my arrangement. One is linking the driver tube anode signal voltage as intimately as possible to the output grids over the widest bandwidth. One is isolating the effects of phase shift, iron distortions, shunt capacitance if one chooses to use R between choke ends and anodes. The R-L-CT-L-R arrangement to convey the DC to the triode driver tube gives the best possible advantages of both resistive and inductive loading just to get the DC to the tube without incurring the penalty of the far simpler use of simple resistor loads for the same purpose. The choke avoids the current change in such DC carrying resistors, and where I change is heavy in a triode, so to is thd, and if it is an LTP, that means 3H. Today I hauled out my breadboard prototype for this sort of LTP and thoroughly tested it with a variety of loads and coditions of PFB, as used in most paraphase concoctions, and which I don't favour, and NFB, which then means the input voltage has to be too large, so I don't much favour that either. I will be drawing up the schematic I used and values and response results and posting them when i get time. the schematic I used has 4.7k in series at each anode instead of 8.2k. When set up as a normal LTP but driven one side only with cap coupled RL = 25k to each anode, I was able to get 150vrms from each anode at 1 kHz, 3% thd / onset of clipping, Ea = 300, Iaq = 13 mA. The measured Ro at each anode when measured seperately when setting the output at 75vrms, and reducing Rl from 25k to 10k on just one side of the LTP was 1.7k ohms, remarkably low. Clearly, the choke isn't just acting as a choke, but as an auto transformer, and when RL is reduced on one side of the LTP, there is a reaction on the other side, which was a very slight increase in Vo. To make such a measurment, one keeps the input voltage constant. VRL with 25k = 75.2v, VRL with 10k = 68.6v, V change = 6.6v, IRL change = 6.686 mA - 3 mA = 3.86 mA. Source Resistance, Ro, = 6.6 / 3.86 = 1.7kohms One can calculate what Ro would be if the DC carrying RL was 35 instead to each anode, and with no choke. It'd always be a higher measurement compared to the above. I don't care that my LTP idea is neither "fish nor feather" as they might say about a platypus, but hey, even a platypus has surprising addaptability in the inland Oz river systems. So the LTP with a choke has wide bw, low thd, low Ro, and the choke part is easy, not too critical, even if the turns to each 1/2 of the choke are 10% different it won't affect the operation much. What more does one want? Sorry, its not as simple as could be, but then Einstein said something about everything needing to be as simple as possible, but no simpler. The only better thing might be a current mirror. But AAAAHHHHHH, I don't like such things; SS with a high voltage swing is asking for early eventual failure, and SS does have the capacity to lock the + and - v swings together, as is the case with the magnetic element, so that each side of the LTP output is driven by two tubes mutually. T make the equivalent of a CT choke using discrete bjts isn't impossible. It does mean that one would have some interesting circuitry which woukd have to be more active than the two arms of a current mirror. And the choke still allows a voltage swing *above* the supply voltage, something a CCS cannot do. The eternal mysterious beauty of the transformer is its capability to send voltage going +ve and above a B+ supply as a result of a linear release of magnetic energy powered by a change of only voltage at the g1 of a tube. I must draw and scan. Patrick Turner. Andre Jute Patrick Turner wrote: " wrote: Thanks for sending the schematic, Patrick. If you insist on speakers requiring massive amounts of power, like 27W, you will definitely need an amp like this! I thought it was nicely developed, though I would have tried, for no good technical reason, merely because people are prejucdiced, to hold NFB down to 10dB. I'm very keen on constant current loads by MJE 350 myself; note that I call it a load. It sinks current only when applied to the cathode, as in your MJE340 example. I also like the idea of an EL84 as a driver very much. A sweet little tube. I have found the dynamics to be truly crisp and real with EL84 in triode. Its almost exactly equal to 5 x half sections of 6SN7/6CG7. 27 watts is a massive amount of power for some, but for others its barely enough. The LTP all works fine without any grid current. Now as i said, one could deploy less than 12 dB of NFB. In one of my other posts i went on to describe Ro without NFB with 2 x KT88/6550. Ra-a is about 2,200 ohms while operating in class A, which is all we need to worry about here because the AB operation where Ro increases the amp has little function, since 90% of users will only use 2 watts average, and won't make much use of the AB working ability. So with a 6k to 6 ohm OPT, ZR = 1,000:1, so without NFB Ro = 2.2 ohms. ( one needs to add winding losses, not much if kept to a nice low 5% ) KT90 used as triode will have lower Ra-a, so maybe Ro = about 1.5 ohms. Using a quad of EL34, or the better 6CA7 with no change to the OPT ZR, Ra-a will be about 1,200 ohms, so the Ro will also be around equal to 2 x KT90 or around 1.5 ohms. In such a case the power output won't increase much, maybe to 35 watts, but you get slightly more class A and less thd, and more current ability. 10 dB NFB would be fine, and as everyone should know, Ro varies as 1 / ( µ x ß ), not 1 / ( gain x ß ), so 10 dB will reduce Ro about 3 times, and thus a DF of nearly 10 is attained even with 2 x KT88/6550. Now of course if we used 300B, we would perhaps use Ea = 420 v max, and about the same PP power is possible, but the Ra-a is like KT90 = 1,600 ohms and considerably lower than KT88/6550. 300B in PP are quite nice. I'm not so keen on fixed bias with adjustment, even in DIY hands. I agree entirely, and cathode bias is a valid option. In my other circuits the bias voltage is fixed, and only the balance is altered, and a two transistor circuit senses out of balance DC of more than 4 mA in the OPT, and a led turns on if it unbalances. There is a balancing pot with a small knob, and a user sees a glowing led, and adjusts the balance until it is extinguished, and thats that, no meter, no measuring, no worry either. If he runs out of travel and an led won't turn off, the tubes or something else is wrong, and active protection takes care of over-currenting in the tube, ie, twice the bias current for more than 4 seconds. So imbalance in bias current is seen across the room. Plenty of warning, and no sudden Mt Vesuvious in the loungeroom, with a lawsuit to follow. The biasing as I have drawn it is optional. Even though fixed bias is the best performing bias method when testing with a sine wave in an AB1 amp, testing with pink noise or busy rock music with clipping just starting to occur on the peaks usually doesn't cause much change in the Ek of either output triode with cathode bias. Of course such biasing isn't everyone's cup of tea. Allen Wright would scoff at cathode bias, and tell us to have commoned cathodes of the output tubes and use a constant current sink ( a power transistor of some kind ) and no bypass caps, rather like the way my PP driver tubes are set up. Thus one gets ONLY class A from such output tubes, and the very maximum of 2H current cancellations, and Allen says it always sounds better. Ppl are welcome to try these options. One of the great things about autobias is that the amp is essentially service-free until the power tubes give up the ghost. Fixed bias, except by battery for small signal tubes, should be reserved for bass freaks and other street corner boomboxers. I agree, but I have sold a few FB amps and with the active balance indicators, and active protection, I have had no customer problems. I understand that you essentially present this schemo as a teaching exercise to explain stages you use all the time; to that purpose it is a comprehensive success. However, viewing it as an amp someone serious could build I have to conclude that, overall, this is a tremendous amount of engineering that I wouldn't have done if I were you. Since you know how to design and wind your own transformers, and there are already two pieces of iron in the signal chain on this amp, why not avoid all that engineering and design and components and make the entire thing all iron, zero caps, very few resistors? For instance a step up/splitter transformer into the EL84 grids and another beefier IT with secondaries to the KT88 grids and bingo, the circuit is incredibly simplified, more value has been added, a higher price can be demanded. Your split choke is clever but a full IT would be even cleverer. I couldn't agree more with what you are saying, and I suggest ppl try things. The step up tranny phase inverter idea is quite nice. But with any global NFB, there could be LF/HF instability since you have time constants caused by LR, rather than much lower and predictable CR time constants. The design as presented uses a CT choke made using 5,000 turns of 0.2 mm dia wire around a core with 25 stack x 25 tongue wasteless, but suitable C-cores would be ideal if one found some. By the GOSS E&I are not that uncommon, and almost any DIYer could make the CT choke. The chokes is isolated from the EL84 anodes, so the coupling to the output tube grids is very pure, direct, without the effects of LL and shunt C and Lp that one would get with an IST, not to mention the trimming one has to do with a zobel to get the response to be flat, not peaked due to resonances at HF, all rather difficult for a DIYer without a CRO, and hardly any knowledge of filter theory. I might at that as drawn, the LTP has anode loads on each EL84 of around 80 kohms at 1 kHz, and its these loads which cause the balance in the circuit while the RL seen by each tube is equal. In my 300 watters, which need about 75 vrms from each EL84 anode, the anode loads are each 6 x 120k in parallel, or 20k, and the choke + 8k2 load together so high it has negligible loading effect, thus the balance is excellent, and thd very low and the grid bias R low enough to control the phenomena of grids slowly going a bit positive in relation to the actual applied grid bias. Therefore, in hindsight, using say 2 uF coupling caps from the EL84 anodes and say 33 k output tube grid bias R would be quite OK and give a pole at 2.4 Hz, instead of the 3.3 Hz using 0.47 uF and 100k. Brian Sowter many years ago gave me an all-iron coupled PP 211 design which was said at the time to be Kondo's personal favourite, not the SE Ongaku which got all the publicity. Studying that circuit, I understood why. Costing the transformers I grasped much more quickly why more people don't do it! PP 211 in class A1 can make around 40 watts without grid I. And a heck of a lot more po with class AB2. The voltage supply to the grids is easy while in A1, or AB1, but as soon as AB2 is considered, the Ro of the driver amp has to be very low, and a pair of CF driving the IST or some such arrangement would become necessary. I have not dallied with 211, or seen what Sowter suggest, but yeah, a good IST can be worth a good OPT. For AB2 drive, an IST with 1:1 ratio makes sense, because there can be two equal windings, each with a CT, or with a secondary with 2 halves, each taken to a bias voltage for each tube, while the active ends of P and S windings can be capacitor shunted using typically 10 uF, which will shunt the leakage inductance between P&S windings, so that at HF the drive does not have the problems of phase shift and response droop caused by the instaneous grid current load applying itself. Usually 1:1 ratio IST have less response problems than 1:2 designs; so we must be able to produce the full output tube Vswing, but I would always insist that this isn't problematic knowing what we do today, and while we keep away from the low fidelity low performance bean counter inspired designs of 60 years ago. I will try to find time to draw up a few more ideas on the subject tomorrow. One fundemental reason why one *needs* to keep the driver stage of a PP amp as linear as possible is that the thd will mostly be 3H of the wave crest flattening type, so any 3H of the driver adds to that of the output stage. One a;lternative to a choke is to use a much higher B+ for the driver stage than I have indicated, and use larger value DC anode supply resistors. To that purpose, if the PSU has a pair of diodes from a CT HT winding so that say 370-0-370vrms is rectified to make +500v at the output of the CLC filter, then it is possible to make +1,000v using a voltage doubler riding on top of the main anode PS and this 1,000v can be CRC filtered down to about +600v, and if Ea of the EL84 = +250v, then the DC RLs can each be 35k if we also creep the Ia of each EL84 down to 10 mA each. The 35k to each anode is sufficiently low value to make sure good balance is possible due to the cathode CCS. The the bias of the output tubes can be as i have shown and RL of each EL84 anode = 26k, or over 10 x Ra, so thd will remain in the very low zone, and we have not had to wind a darn choke. But we still will have needed to create the highish voltage supply, which is nervous nelly time for diyers. I have done exactly this type of extended supply B+ driver supply for one of my amps which has a sixpack of class AB1 6CA7 driven with a pair of 6SN7 in the LTP, one paralleled tube each side of the LTP. ( 6V6 in triode is terrific also ). While an EL84 or maybe 6V6 is as cheap as a 6SN7, or better grade 6CG7, I reckon there is great value with EL84, which will always be with us at least for the rest of my lifetime. If one is determined to make an all iron coupled PP amp, then the IST is a must. but I still think some R damping or termination of the secondary may be needed, and if so, one R of the two can be used to provide the positive feedback to make the PP input stage into parafeed, so making the amp twice as sensitive as where one uses an LTP with single input. Parafeed with a CCS cathode I supply is still best. If one wants to get rid of the usual input tube, then the driver can be made using your cherished 417A, or perhaps a pair of 12AT7, with both halves paralleled. 12AT7 will thus havev µ = 55, Ra = about 10k for 10 mA in the total of two triodes, so a circuit gain of 45 is realizable, so only 2 vrms input is needed to make 90 vrms a-a output. But 12AT7 isn't the most linear driver when making such highish output voltages, and we could wonder that using an input stage like I have shown with medium µ triode LTP is a better approach, since for one thing, the input tube buffers the LTP input from the source. Then the preamp doesn't have to have such high gain, or be necessary at all, just a gain pot will do, unless we have vinyl. But certainly no line stage preamp is needed with a 3 stage power amp. One further way of building an LTP with better linearity using tubes such as 12AT7 is to use an R divider to derive NFB from the opposite tube that one normally takes the positive FB drive to make a paraphase circuit. The CCS must stay, to get balance, and RL of each tube must be equal, but the gain can be reduced from say 45 to 15, and the thd also reduced. Its still bothersome for the DIYer to get right, so I prefer the simplicity of the LTP with drive to one side only and the use of medium µ triodes. The Miller C with the LTP I have shown is the static Cg-a x gain, which I think is about 12pF x 9.5 = 114pF for triode connection, so there is a fair amount of Cin. The 12AU7 Ra of around 5k will make a pole at about 300kHz, and even higher with C7,R8 stepping networks added....so not a worry. Patrick Turner. Andre Jute Patrick Turner wrote: I just posted a schematic for a PP class AB1 PP triode amp at alt.binaries.schematics.electronic. It has several features to explain ideas I use all the time. There are two different CCS, each one using npn or pnp transistors, and an LTP driver using trioded EL84 with a CT choke and additional RL to keep the impedance of the DC supply elements very high between 50Hz to above 20 kHz . The CCS used for an active load for the 12AU7 will reduce thd by at least 8 db over a plain R load. The CCS used for the CCS from common k of the EL84 LTP to the negative supply will balance the LTP automatically. All the possible LF and HF phase tweaking gain tailoring networks are shown for stabilising the amp without compromising the HF performance and without causing any stage to saturate between 16 Hz and 65 kHz at full power providing the OPT is a half decent. It is possible to use up to 12 output tubes in parallel PP with exactly the same input and driver tubes. Possibly a pair of 6H30 could be used in lieu of the EL84 in the LTP, perhaps at 10 mA per tube, not 12mA, as i suggest, and with R18 and R19 increased somewhat. Patrick Turner. |
#21
|
|||
|
|||
Ian Iveson wrote: "Patrick Turner" wrote I just posted a schematic for a PP class AB1 PP triode amp at alt.binaries.schematics.electronic. It has several features to explain ideas I use all the time. There are two different CCS, each one using npn or pnp transistors, and an LTP driver using trioded EL84 with a CT choke and additional RL to keep the impedance of the DC supply elements very high between 50Hz to above 20 kHz .... I spent some time explaining the faults of this "design" to Patrick, and got foul-mouthed in return, as is usual if anyone questions his "ideas". I have asked for elucidation and got the usual evasive waffle and random babbling...quotations from RDH, I assume. He sold the prototype regardless, as usual, as if it were proven or even properly tested. There you go again Ian, on the train of discreditation. You must be lonesome, because hardly anyone agreed with whatever detailed reasons you had for dismissing the ideas I put forward. Maybe you would like to explain more in detail than I have, but just get all your sums right, ppl who don't are fair game around here. Simulation confirms it is just what it seems: a pathetic embarrassment to anyone with even half a notion of circuit design. But I used NO simulation. You are the one to simulate everything instead of going to your workshop for several hours to measure how it works. Patrick's unique selling point is his claim to wind decent traditional chokes and transformers. Convincing daft Australians to shell out for an extra 200H to no real purpose is a nothing but a cynical ploy. Ah, so you belong also to the School of Mediocrity. Williamson would have been overjoyed to hear you tell him that a minimum of 100H in an OPT for a triode amp was a cynical ploy and plain daft. Lucky you weren't around in 1947, I know who would have been the laughing stock amoungst the real doers of the time. Tired of arguing, and less eager now to help a grasping fool. Perhaps that's obvious. So, anyone else prepared to defend the use of two EL84, 200H of differential mode choke, and an SS current source, as a general purpose splitter / driver? I don't care if nobody does. I am not worried if nobody is prodded to think about their tube craft which was the purpose of the info. There are 50 ways to make a tube amp, and ppl are welcome to try whatever ideas turn them on. You are just as bad today as Oinky when he says we should throw out our tubes and build SS amps. cheers, Ian There is nothing cheerful in what you are saying. You are being just a grumpy old fart, with nothing to contribute to the discussion at all. You need a good night's rest. Patrick Turner. |
#22
|
|||
|
|||
In article , "Ian Iveson"
wrote: "Patrick Turner" wrote I just posted a schematic for a PP class AB1 PP triode amp at alt.binaries.schematics.electronic. It has several features to explain ideas I use all the time. There are two different CCS, each one using npn or pnp transistors, and an LTP driver using trioded EL84 with a CT choke and additional RL to keep the impedance of the DC supply elements very high between 50Hz to above 20 kHz .... I spent some time explaining the faults of this "design" to Patrick, You seem to have explained the faults of the design to Patrick in private communications as I see nothing on this subject posted by you to the entire group. Are you afraid that your comments won't hold up under the intense scrutiny of the entire group? Why don't you post your "explanations" for all to see, rather than slamming Patrick as if we are all in agreement with your "explanations"? You do have an alarming tendency to almost always get your "sums wrong" as Patrick would say. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
#23
|
|||
|
|||
John Byrns wrote: In article , "Ian Iveson" wrote: "Patrick Turner" wrote I just posted a schematic for a PP class AB1 PP triode amp at alt.binaries.schematics.electronic. It has several features to explain ideas I use all the time. There are two different CCS, each one using npn or pnp transistors, and an LTP driver using trioded EL84 with a CT choke and additional RL to keep the impedance of the DC supply elements very high between 50Hz to above 20 kHz .... I spent some time explaining the faults of this "design" to Patrick, You seem to have explained the faults of the design to Patrick in private communications as I see nothing on this subject posted by you to the entire group. Are you afraid that your comments won't hold up under the intense scrutiny of the entire group? Why don't you post your "explanations" for all to see, rather than slamming Patrick as if we are all in agreement with your "explanations"? You do have an alarming tendency to almost always get your "sums wrong" as Patrick would say. For some reason I think Ian is referring to when i first brought the idea of a "Turner LTP" to the group a considerable time ago, some years, I think. But for want of a better name, mine may as well be used, because I can't think of any other tube amp builder publicizing the LR combined load and CCS. Ian wasn't all that happy at the time with the T-LTP, but I have satisfied myself that it works fine, and all the explanations and simulations in the world don't really mean much unless one builds, measures, and uses something, and finds it has some saving graces compared to the alternatives, which can't be perfect either. The experiments I did today using EL84 and the primary of an ancient and truly abominable OPT made for a very poor PA amp using 6BM8, confirmed the benefits I know exist with this concept. The use of crummy inductors in prototypes tend to reveal the problems of a basic design more obviously than starting with a perfectly made L. In this case the early onset of saturation of the core could be observed, and the absense of such distortions in the outputs, thanks to the series R betwen anodes and ends of the choke. The 6BM8 opt had only 30H at 100 Hz, a long way short of 200H which I think a circuit for hi-fi deserves. In this case, 200H is very easily obtained; its far simpler than winding an IST, or any OPT, and random winding can be used instead of layered windings with paper between each layer, which I assure you are real PITA to wind without the proper lathe. Random wound field coil inductors were wound to make speaker magnets before permanent magnets became economically viable. They were subject to tube failures, excessive currents, but many still are working fine after 70 years in a radio, quite remarkable, considering the wire enamel in 1935 wasn't as good as today's polyester-imide. Patrick Turner. Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
#24
|
|||
|
|||
----- Original Message ----- From: Iain M Churches To: Patrick Turner Sent: Wednesday, May 18, 2005 5:39 PM Subject: 27W PP TRIODE AMP May 2005. Iain asked me in a private email about the HF phase tweaking RC circuit shunting V anode to 0V, and I tried unsuccessfully to send the reply following his question ...... Can you give me "starting values" for C7, R8. ? You should determine these values when the amp is wired up with the OPT that is to be used. C15 also has no value, and is dependant on the other HF poles in the tube line up and OPT. R30 and the C = 0.22uF, not numbered, may also not suit all amps or OPTs. But for starting values, use C15 = 470pF, and use R30 with 0.22uF. C7 and R8 may not be needed at all. But R&C should be tried to get a less peaked sine wave response with 0.22 uF across the speaker terminals with no R present. This will give a 5 kHz square wave with the least overshoot and with fewest oscillation waves after the initial overshoot, ie, good "settling time" after transients. Some amps will oscillate violently with 0.22 uF across the output when NFB is connected. So first you have to try the amp without NFB and then add a litle NFB, say 6 dB only, and see what happens with 0.22 uF. What we are aiming for is unconditional stability, and a 15 dB margin of stability with R loads, and 65 kHz of bw with C loads, and with a pure 2.0 uF load, no more than 1 dB of peaking in the response at 20 kHz, no R load used. The maximum undistorted sine wave output voltage with 2uF and no R should not be more than 4 dB lower than the power is at 1 kHz into 8 ohms. No stage should saturate or be driven into grid current during such a test between 10 Hz and 20 kHz. I use a dual 20-400pF variable tuning gang in series with a 20k pot so R&C can be varied to reduce the ring on a square wave with 0.22 uF and still give 65 kHz of bw into pure resistance loads. The Ra of V1 = approx 5k for 12AU7 paralleled, since Ra for 1/2 = 10k with only 2.5 mA of Ia. The Cin to the LTP = approx 100pF, ( stray C + Miller C ), so an RC low pass filter exists between V1 and V2, the driven grid of the LTP. So there is a -3dB HF pole caused by 5k and 100 pF at 318 kHz, where there is a phase lag of 45d. Other phase lags are caused by leakage inductance, L+R, Miller C after the LTP into the output tubes, C+R, And when a 0.22 uF capacitor load is used, it will make the phase lag also, so the total open loop phase shift may sum to -180d without C7 and R8 and C15. For oscillation to occur, the open loop gain must be above 1.0 when the applied FB is -180d with respect to the input voltage which has a reference phase of 0.0 degrees. In other words if there is gain when the NFB has become positive FB it will oscillate. C15 advances the phase of HF signals fed back, as well as increase the amount of FB fed back since c reduces impedance as F rises. Now I have told you what the basic RC structure is between V1 and V2. Suppose 470 pF and 2k are used for C7 and R8. What does that do to the open loop response measured at V1 anode with no FB connected? What will the open loop gain at V1 anode be without C7/R8 at 318 kHz? Assume you have a signal gene with a flat response to 2 MHz, and its Ro = 100 ohms, so the input voltage has a flat response to 2 MHz. I know all the answers to the questions. But why should you not be able to analyse the relevant responses yourself, by applying the most basic ideas of triode models and RLC theory? Now someone said to me tonight, existance is the uncertainty of possibilities. Uncertainty is due to subatomic activity, where lil iddybiddy thinges may of may not be there. A long story it was, mostly about nothing, since all matter is mostly vacuum, and mainly energy........ We can reduce possibilities by reductions of the variables affecting existance. So we can know what responses will be if we choose L or C values and then vary F, taking into account the whole amplifier as a bandpass filter with active elements. The triode can be seen as a voltage generator with an output of µ x Vg, and with an internal R from gene output to where the anode is, and R is the dynamic generator resistance, or anode resistance of the triode, Ra. Unless you try to understand the basics, and write out the whole circuit on a page with all active tubes modelled, R and C and L values all noted, with all calculated poles, and measure and confirm your analysis is about right, you will never properly understand amplifiers. After doing lots of amps you get an intuitive feel for tuning the response and phase shift to give the overall desired outcome after applying the tweaks, while watching square waves in a CRO. Once you have all that about achieved well, the listening tests should confirm that it is difficult to make a poor sounding triode amp if its well constructed. Regards, Patrick Turner. Best regards Iain |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
ECL86/6GW8 triode strapped pentode section plate curves | Vacuum Tubes | |||
Using power triode or power pentode wired as a triode as a split-load phase splitter tube? | Vacuum Tubes | |||
Using power triode/pentode wired as triode as a split load phase splitter tube? | Vacuum Tubes | |||
SRPP questions | Vacuum Tubes | |||
List of NOS mostly tubes | Vacuum Tubes |