Home |
Search |
Today's Posts |
#1
|
|||
|
|||
CPI, RDH4, pg 523
Something else needs to be said about bootstrapping the pentode RL
of the schematic in the mifddle of the page 523. The CPI has R4, 20k for the anode load, and then 40k from cathode to 0V, R5. There is another R2, 40k from B+ taken to the junction of R1/C2. Thus R2 is effectively is in parallel with R5, and coupled via C2. Now I reckon this arrangement of R loads means that at large signal voltage levels, the output from the CPI cathode will cut off sharply above a threshold because of the ratio of R5 to R2, which is 1:1, where ideally to make the cut off distortion occur at a high voltage threshold, the cap coupled R should be a much higher value than the R carrying DC to/from a tube. The penomena is no different to where one has a cathode follower, and the cap coupled load is reduced, so that if RL dc = cap coupled RL, then if the Ia is cut right off, the maximum downward voltage travel is 1/2 the quiescent voltage between cathode and 0V. A CPI has to perhaps make a quite a few volts to drive an output stage, and unless there is a nice large voltage between the cathode of the CPI and 0V, there will be limiting of the voltage travel due to the voltage divider effect of the two RLs being equal. It would therefore not be such a good idea to use this circuit to drive large octal outputs, since the drive is usually over 25 vrms max to each g1. With a normal R loaded pentode stage driving a CPI, the gain will be around 6 times less than the bootstrapped gain pentode, so to boost the gain without the ill-effects of early cut off I have described, a CCS could be used with the pentode, in the form of a triode as one would use in a mu follower config, where the R between top tube is around 10 times a normal cathode biasing R, so about 22k. If the top tube has U = 100, like a paralleled 12AX7, the load seen by the bottom tube becomes about 100 x 22k // whatever biasing R for the top tube is, but at least RL for the pentode would be about 1M, and if U of the pentode is 3,000, Ra = 1M, say for a 6AU6, at Ia = 1.5 mA then gain = 1,500, and that's plenty. Direct coupling of the input pentode is possible to the CPI grid. I'd still prefer to use cascaded low u triodes for the input drive amp, since the linearity is maybe 10 times better. But the Williamson with all 6SN7 with KT66 triode outputs, ( or better, KT88/6550/300B outputs ) and using 20 dB of GFB needs about 2vrms input for full power. PPL may not agree with 20 dB of GFB, and ok, its your choice, and I would maybe use only 12 dB for a triode output stage, and that would reduce Vin needed to about 0.8vrms. Also a 12AT7 be used for the input triode and CPI, since the gain will be more than double the 6SN7 case, and Vin could be reduced to about 0.35vrms, for about 20 watts of output power, and no preamp is needed, just a 50k gain pot. Certainly with KT88 outputs, and with PO max = 28 watts AB1 triode, there is plenty of sensitivity, and no need for a preamp. The 12AT7 has moderately low Ra of around 20k at Ia = 4 mA, and u = 55, so the gain will be around 35 for the input triode, and the CPI will be extremely low distortion since gain reduction is from around 32 to about 2, and the signal level is low, since its a williamson balanced voltage drive amp that does most of the work to make the voltage drive to the output tubes. I have heard music via a pair of pure class A 300B in PP, which give around 18 watts, and its very nice, and will give the SET boys some real worries, but alas, PP has been out of favour for some years at least. About 28 watts class AB1 is possible, and considerably more in class AB2, but i've never ever actually made a class AB2 amp; I have always thought it better to use 4 output tubes to get the extra power lazily, rather than work the guts out of just two tubes. Therefore the low thd sweetzone of the amp is widened, over what one has with just two outputs. Plus the OPT is usually easier to wind, or a 5k to 4,8,16 OPT is easier to source, and we *want* a high value RL if we want a high % of class A, and natural linearity. Patrick Turner. |
#2
|
|||
|
|||
Some random reflections was CPI, RDH4, pg 523
Patrick Turner's dunny copy of the RDH fell open at the pinup on p523. Those of you who still have the text parts of the RDH uselessly attached to the pinups may read my accompanying thread on dealing firmly with the RDH. But Patrick, as ever incapable of keeping to the main idea, raises some tangential points that need a reply. I've split off a subthread for these side issues. First Patrick says, directly as a conclusion to another subject he discussed: I'd still prefer to use cascaded low u triodes for the input drive amp, since the linearity is maybe 10 times better. Sure and all, in days of yore before I had a stash of 417A and, more important, learned from a lot of good, helpful Joenetters and RATs how to use them, I didn't even think before reaching for a pair of 6SN7, the most linear driver among the easily accessible and affordable tubes. It helps that the 6SN7 is beefy too, because, as I tried in years gone by to explain to RATs, the implication of Miller is that you must use enough current on the driver to suit the power tube's requirement. Slew rate calculations for a reasonable HF bandwidth offer a shortcut both to the thinking behind this and the math. (I can post something on calculating this if anyone is interested.) My fave 417A is beefier still, usually used at over 20mA on the plate, which is good all the way up to 845 and Svet 572-xx. Let's just pause here to say that a mickey mouse B9 tube looks like a junior piece of plastic for kids building a little school project out of their pocket money, an octal driver looks a bit more like real glassware, and men drive their power tube with another power tube. (Real men, as seen over in the Kilovolt Tube thread for Big Hairies, have long since had their brains fried by overheated ozone in the region of multikilovolt transmitters. Reckless *******s! Take a sperm count, fellers. G) Then Patrick goes off on the tangent that interests me: But the Williamson with all 6SN7 with KT66 triode outputs, ( or better, KT88/6550/300B outputs ) and using 20 dB of GFB needs about 2vrms input for full power. I'm not a big fan of the Williamson, because in a historical perspective it set the entire audio industry off on a wild goose chase of NFB fixes to self-induced diseases. But 2Vrms input was prophetic and is really convenient these days: straight outta your CD player. I like integrated stereo amps. SNIP Also a 12AT7 be used for the input triode and CPI, since the gain will be more than double the 6SN7 case, and Vin could be reduced to about 0.35vrms, for about 20 watts of output power, and no preamp is needed, just a 50k gain pot. As it happens, I also like the 12AT7 for its precision, but the whole point is to use the 12AT7 for its linearity rather than to hog it out. Now, for those who don't know them, the 6SN7 and 12AT7 are crisp, precise tubes. It used to be received wisdom that you didn't use two of the same tubes in sequence because that would magnify their disabilities. When I was still young enough to believe it smart to smash ivory, I reckoned that if one 6SN7 is precise, two would magnify the precision. Others think it results in an acid sound that rips off your ears. (In sighted tests. In blind tests the same "experts" choose the 2x 6SN7 drivers first over anything else except, sometimes, the 417A--which depends on high current rather than true inherent linearity--and a 6SL7/ 6SN7 combo which is definitely warmer even if less precise. See Steve Bench's netsite for some good silent driver arrangements with these tubes.) SNIPPED, PATRICK'S ARGUMENT TO LEAVE JUST THE DATA BECAUSE I'M ON ANOTHER TRACK: The 12AT7 has moderately low Ra of around 20k at Ia = 4 mA, and u = 55, MORE SNIP, DITTO Now Patrick goes off on another tangent: I have heard music via a pair of pure class A 300B in PP, which give around 18 watts, and its very nice, and will give the SET boys some real worries, but alas, PP has been out of favour for some years at least. PSE 300B give 16W without the crossover distortion of PP. PP 300B are not really out of favour. For instance, Lynn Olson, another ultrafidelista opinion-former, built a PP 300B amp. The circuit is on his netsite. Price the iron he requires to make it work as well as a simple SE amp, then we can talk again about whether the additional advantages of his approach are worth the cost, the effort, the investment in a higher level of expertise, and so on. I admire Lynn's amp as a work of art, but I spent my own time and money instead on developing speakers that don't require such a monster. One of my favourite amps, sitting on the floor at my right on hand of this thread, just waiting to be plugged in when the current disk finishes playing, is the Miyabe SEPP 300B (it is sold as a kit by Triode Supply Japan, costing 1600 USD when I reviewed it for Glass Audio a few years ago; see GA for a full discussion). SEPP is an interesting and useful compromise. About 28 watts class AB1 is possible, Patrick Turner. 24W Class A1 out of triode-linked EL34PP will sound so much better. It really does seem to me that any speakers which require more than 24W must justify their existence. Since a 24W Class A1 EL34 PP amp is relatively cheap and easy to design and build, any bigger amp must therefore justify its existence. Anything that hogs out more power, or requires much NFB to work, seems to me like giving yourself the clap so you can prove how clever you are by "curing" it with a bandaid. Andre Jute Here is Patrick's original in full: Patrick Turner wrote: Something else needs to be said about bootstrapping the pentode RL of the schematic in the mifddle of the page 523. The CPI has R4, 20k for the anode load, and then 40k from cathode to 0V, R5. There is another R2, 40k from B+ taken to the junction of R1/C2. Thus R2 is effectively is in parallel with R5, and coupled via C2. Now I reckon this arrangement of R loads means that at large signal voltage levels, the output from the CPI cathode will cut off sharply above a threshold because of the ratio of R5 to R2, which is 1:1, where ideally to make the cut off distortion occur at a high voltage threshold, the cap coupled R should be a much higher value than the R carrying DC to/from a tube. The penomena is no different to where one has a cathode follower, and the cap coupled load is reduced, so that if RL dc = cap coupled RL, then if the Ia is cut right off, the maximum downward voltage travel is 1/2 the quiescent voltage between cathode and 0V. A CPI has to perhaps make a quite a few volts to drive an output stage, and unless there is a nice large voltage between the cathode of the CPI and 0V, there will be limiting of the voltage travel due to the voltage divider effect of the two RLs being equal. It would therefore not be such a good idea to use this circuit to drive large octal outputs, since the drive is usually over 25 vrms max to each g1. With a normal R loaded pentode stage driving a CPI, the gain will be around 6 times less than the bootstrapped gain pentode, so to boost the gain without the ill-effects of early cut off I have described, a CCS could be used with the pentode, in the form of a triode as one would use in a mu follower config, where the R between top tube is around 10 times a normal cathode biasing R, so about 22k. If the top tube has U = 100, like a paralleled 12AX7, the load seen by the bottom tube becomes about 100 x 22k // whatever biasing R for the top tube is, but at least RL for the pentode would be about 1M, and if U of the pentode is 3,000, Ra = 1M, say for a 6AU6, at Ia = 1.5 mA then gain = 1,500, and that's plenty. Direct coupling of the input pentode is possible to the CPI grid. I'd still prefer to use cascaded low u triodes for the input drive amp, since the linearity is maybe 10 times better. But the Williamson with all 6SN7 with KT66 triode outputs, ( or better, KT88/6550/300B outputs ) and using 20 dB of GFB needs about 2vrms input for full power. PPL may not agree with 20 dB of GFB, and ok, its your choice, and I would maybe use only 12 dB for a triode output stage, and that would reduce Vin needed to about 0.8vrms. Also a 12AT7 be used for the input triode and CPI, since the gain will be more than double the 6SN7 case, and Vin could be reduced to about 0.35vrms, for about 20 watts of output power, and no preamp is needed, just a 50k gain pot. Certainly with KT88 outputs, and with PO max = 28 watts AB1 triode, there is plenty of sensitivity, and no need for a preamp. The 12AT7 has moderately low Ra of around 20k at Ia = 4 mA, and u = 55, so the gain will be around 35 for the input triode, and the CPI will be extremely low distortion since gain reduction is from around 32 to about 2, and the signal level is low, since its a williamson balanced voltage drive amp that does most of the work to make the voltage drive to the output tubes. I have heard music via a pair of pure class A 300B in PP, which give around 18 watts, and its very nice, and will give the SET boys some real worries, but alas, PP has been out of favour for some years at least. About 28 watts class AB1 is possible, and considerably more in class AB2, but i've never ever actually made a class AB2 amp; I have always thought it better to use 4 output tubes to get the extra power lazily, rather than work the guts out of just two tubes. Therefore the low thd sweetzone of the amp is widened, over what one has with just two outputs. Plus the OPT is usually easier to wind, or a 5k to 4,8,16 OPT is easier to source, and we *want* a high value RL if we want a high % of class A, and natural linearity. Patrick Turner. |
#3
|
|||
|
|||
François Yves Le Gal said:
A properly designed push-pull amplifier doesn't exhibit xover distortion. In his book "Modern High End Tube Amplifiers", Menno vdVeen describes an effect in PP output transformers that resembles crossover, namely the point of operation where the Weiss particles are forced to turn because of the current swing from positive to negative. I know, it's probably poorly translated from Dutch, but I suspect you know what I mean. Whether or not this effect is largely theoretical or clearly audible, I don't know. My class push pull KT88 in triode amps sound wonderful despite that. (And yes, Andre, I like them over EL34s) BTW, PSE is highly non linear - even when using closely matched tubes. Agreed, but despite the apparent non-linearity, SE or PSE amps surely have a huge crowd of fans. There must be a reason for it ;-) (though not my cuppa tea, you will note) -- Sander de Waal " SOA of a KT88? Sufficient. " |
#4
|
|||
|
|||
|
#5
|
|||
|
|||
"François Yves Le Gal" wrote: On 17 Mar 2005 10:31:08 -0800, wrote: PSE 300B give 16W without the crossover distortion of PP. A properly designed push-pull amplifier doesn't exhibit xover distortion. BTW, PSE is highly non linear - even when using closely matched tubes. Only pure class A amps have no crossover Dn. But any class AB amp will have some. AJ was speaking about a 24 watt PP amp using a pair of EL34 in triode, and its a fairly class AB situation, so there will be Xover Dn between few first class A watts and the transition to class AB watts. PSE becomes more linear as one increases the number of tubes in parallel, and if triodes, and more linear if the RL is increased in its relative value to Ra. So a pair of EL34 in triode connected to the same RL as used for 1 tube will be far more linear than just using one tube, for the same 1 watt of output power. Why don't you get back to where you belong, Mr. Jute? I see no reason for him to move from where he is. Just a man discussing tubes; I think he is where he belongs. Patrick Turner. |
#6
|
|||
|
|||
On Fri, 18 Mar 2005 18:44:04 +0100, François Yves Le Gal wrote:
snip Mr. Jute has a very long and very disturbed history on the Internet. I believe that he belongs to a psychiatric institution. http://groups.google.com/groups?q=%2... 1&sa=N&tab=wg I extracted one of the posts from that same link: http://groups.google.com/groups?q=%2....co m&rnum=11 Which side belonged in an institution? - No, don't continue this. Its gone, dead & buried. We don't want to revive old wars, there are enough current ones to be going on with. ;-) -- Mick (no M$ software on here... :-) ) Web: http://www.nascom.info Web: http://projectedsound.tk |
#7
|
|||
|
|||
On Thu, 17 Mar 2005 20:24:54 +0100, François Yves Le Gal
wrote: It's IMO very theoretical and at the molecular level, i.e. two or three orders of magnitude below perceptible levels. The flat spot on the B-H curve is easily observable; try it for yourself. Agreed, but despite the apparent non-linearity, SE or PSE amps surely have a huge crowd of fans. There must be a reason for it ;-) The main reason? Most people don't understand how an amp works and follow the audiophool of the month. Doubtless a contributing factor, but not the whole story. Chris Hornbeck |
#8
|
|||
|
|||
On Sat, 19 Mar 2005 01:23:45 +0100, François Yves Le Gal
wrote: On Sat, 19 Mar 2005 00:05:16 GMT, Chris Hornbeck wrote: The flat spot on the B-H curve is easily observable; try it for yourself. It is, but it's not caused by any form of Weiss-related phenomenon. http://www.beigebag.com/case_xfrmer_4.htm This is an interesting reference, but is unrelated to the issue (based on a quick look; I'm off to the opry). The B-H curve also has a small flat spot near zero crossing. This is the one and only reason anyone would ever bother with the single-ended amplifier grief. But for some, who have tried it and found it significant, it is surprisingly important. Surprised the **** outta me, anyway. Who knew? Off to "The Tender Land", Chris Hornbeck |
#9
|
|||
|
|||
Yo, Mick:
I never heard of Le Gal before today but I too looked up the URL he gives. Beyond the first message, which can be checked as factual, it consists, as far as I can see, entirely of unproven accusations against me, typical Magnequest Scum/Pinkerton ad hominem flamebait; notice how I just ignored it.. Perhaps this pompous little French fascist Francois Yves Le Gal believes I should submit myself to the Code Napoleon and just accept guilt by accusation. (Or do you think Le Gal's English is so poor he doesn't understand what goes on in that thread?) Then, unbelievably, he excuses his stupidity by resurrecting a hoary old Magnequest Scum chestnut, that anyone with a kind word to say for me is a sock puppet. Convenient, as it relieves the usual run of morons we get on RAT of the obligation of invdividual thought. Incidentally, several members of the Magnequest Scum actually spoke to Mr Beresford in their efforts to intimidate him into withdrawing his damning report... But Le Gal isn't about to let the facts get in his the way of his malice. You don't need to be Freud to discover Le Gal's motive for his slimy character assassinations. I can reveal that Mr Le Gal's motive is that he once built an amplifier with Magnequest transformers; he has confessed to it. That's enough to warp anyone's judgement for life. I wouldn't mind the quarterwits that hang around RAT so much if they would just be consistent and a spot less slimy. But this hypocritical evolue (1) Le Gal complains about the Magnequest flame wars being hauled up again on the same day as he hurls a googleful of discredited Magnescum character assasinations against my head. Half these morons are in a thread (Dogs) whining about flame wars--while carrying on a flame war on me, some of them like the blustering butcher's boy Stouffer right in that thread. Andre Jute. PS Soon Le Gal will discover Mick is my sock puppet. Who wants to bet money on it? (1) jumped-up barracks-room lawyer from the provinces On Fri, 18 Mar 2005 18:44:04 +0100, Fran=E7ois Yves Le Gal wrote: snip Mr. Jute has a very long and very disturbed history on the Internet. I believe that he belongs to a psychiatric institution. http://groups.google.com/groups?q=3D...Dfr&lr=3D&c = 2coff=3D1&sa=3DN&tab=3Dwg I extracted one of the posts from that same link: http://groups.google.com/groups?q=3D...3D10&hl=3Df = r&lr=3D&c2coff=3D1&selm=3D6vcac4%24dh2%241%40nnrp1 .dejanews.com&rnum=3D11 Which side belonged in an institution? - No, don't continue this. Its gone, dead & buried. We don't want to revive old wars, there are enough current ones to be going on with. ;-) -- Mick (no M$ software on here... :-) ) Web: http://www.nascom.info Web: http://projectedsound.tk |
#10
|
|||
|
|||
Patrick Turner wrote:
Crossover distortion in class A PP triode amps? In the present atmosphere on RAT I can only conclude that you and Phil met behind my back in the basement of the John Curtin Hotel and conspired for a lark not to tell me it was a Class A amp we were taling about. Never mind, I forgive you all your sins. Andre Jute PS Not to make a big argument with you but-- There are tiny but very real issues here which lead back to the superiority of SE amps, but I imagine we'd better get around to subtle detail work when the present silicon heckler has become bored and buggered off. Patrick Turner wrote: wrote: Some random reflections was CPI, RDH4, pg 523 Patrick Turner's dunny copy of the RDH fell open at the pinup on p523. Those of you who still have the text parts of the RDH uselessly attached to the pinups may read my accompanying thread on dealing firmly with the RDH. But Patrick, as ever incapable of keeping to the main idea, raises some tangential points that need a reply. I've split off a subthread for these side issues. The main idea was to tangentially shoot off with discussions slightly away from what is already in RDH4 on this particualr subject, and just talk around the idea of a bootstrapped pentode and CPI. Doncha think everyone would be bored more than they might be already if I didn't ramble OT a bit now and then? One could consider what isn't written in a book, not only what is. First Patrick says, directly as a conclusion to another subject he discussed: I'd still prefer to use cascaded low u triodes for the input drive amp, since the linearity is maybe 10 times better. Sure and all, in days of yore before I had a stash of 417A and, more important, learned from a lot of good, helpful Joenetters and RATs how to use them, I didn't even think before reaching for a pair of 6SN7, the most linear driver among the easily accessible and affordable tubes. Well, the 6SN7 *is* usually nice and linear, to be sure, but its also sounds well. But only most of the time, not all the time, when compared to different brands of the same tube. 6CG7 were also made by cramming halves of 6SN7 into a 9 pin mini bottle, then versions of 6CG7 came out with smaller plates, grids, but with the same electronic u, Ra, and gm. IMHO, Telfunken sound the best overall, but Mullard are nicest with voices, and the EH types I tried were horrid. The linearity with all when I last *listened* to a both in a line stage was all below 0.01% thd, yet inexplicable sound differences were heard by the 4 of us present. Linearity is nice thing if its free, and and it is with a 6SN7, but that's not all there is to it, because its the sound that counts. Finding ultimate listening pleasure is like trying to nail a nice smell to the ground so it won't move. It helps that the 6SN7 is beefy too, because, as I tried in years gone by to explain to RATs, the implication of Miller is that you must use enough current on the driver to suit the power tube's requirement. Slew rate calculations for a reasonable HF bandwidth offer a shortcut both to the thinking behind this and the math. (I can post something on calculating this if anyone is interested.) My fave 417A is beefier still, usually used at over 20mA on the plate, which is good all the way up to 845 and Svet 572-xx. There are plenty of amps with 12AX7 drivers for octal outputs, Leak 60 monos, Mullard 520, etc. Measured BW isn't too bad. Methinks having more current ability and lower Ra in driver tubes makes more a more detailed sound, despite the fact the miller effect is having virtually no effect at low levels, because the effect is also voltage dependant, so the more voltage swing you have, the more a wave at HF is affected by a given capacitance. So replacing a 6CG7/6SN7 with say an EL84 in triode which is equivalent almost exactly to 5 halves of a 6SN7 will give a perceivable improvement in musical dynamics. I used not think that this does not happen, but too many people for whom I have built amps tell me that they like the higher current drivers. 6V6 as a triode is also a nice driver tube. Let's just pause here to say that a mickey mouse B9 tube looks like a junior piece of plastic for kids building a little school project out of their pocket money, an octal driver looks a bit more like real glassware, and men drive their power tube with another power tube. (Real men, as seen over in the Kilovolt Tube thread for Big Hairies, have long since had their brains fried by overheated ozone in the region of multikilovolt transmitters. Reckless *******s! Take a sperm count, fellers. G) What sayest thou? that there isn't anything unmanly about a little nine pinner? I reckon nine pinners work well, but it ain't anything much to do with size. ( but anyway, if you have a small one, use it twice a night to compensate... and if she still divorces you, just tell the court.... ..."well, I did try, your honour..." ) Then Patrick goes off on the tangent that interests me: But the Williamson with all 6SN7 with KT66 triode outputs, ( or better, KT88/6550/300B outputs ) and using 20 dB of GFB needs about 2vrms input for full power. I'm not a big fan of the Williamson, because in a historical perspective it set the entire audio industry off on a wild goose chase of NFB fixes to self-induced diseases. But 2Vrms input was prophetic and is really convenient these days: straight outta your CD player. I like integrated stereo amps. Williamson's amp was just that tad too complex for me. With a CD player output level, preamps are simply not needed with most old tube amps. And the Williamson was one of the most insensitive, but also with a great snr. If one uses 6SN7 thoughout with a mullard 520, instead of EF86 and 12AX7, the open loop input voltage is far higher than the approximtae 20 mV for full power with Mullard. The SET + LTP is actually about 1/2 as sensitive than the Williamson circuit with SET, CPI, and voltage amp, using the same triodes. So if 20 dB of NFB is used, you need about a volt, rather more convenient, and NFB can be reduced from the excessive 26 dB urged by Mullard. 18 dB is all I ever use, but 12AU7 allows for good sensitivity. Leak made all his amps sensitive to 0.1vrms for full power, something I would never do, for noise reasons, and the fact that sources have more voltage these days. But 0.1v sensitivity means that a single EF86 is all one needs for the MM vinyl amp, and low outputs from radios was OK. I just spent all evening at a friend's place comparing SE amps, and the use of sensitivity where 0.6 vrms for 25 watts into 89 db speakers was just fine, and we never used more than the 12 o'clock position on the gain control. SNIP Also a 12AT7 be used for the input triode and CPI, since the gain will be more than double the 6SN7 case, and Vin could be reduced to about 0.35vrms, for about 20 watts of output power, and no preamp is needed, just a 50k gain pot. As it happens, I also like the 12AT7 for its precision, but the whole point is to use the 12AT7 for its linearity rather than to hog it out. Well, all my measurings of any 12AT7 show, that like the 6AQ8, and other high gM triodes mainly for RF use, that the linearity isn't anywhere as good as 6SN7, or 6CG7. BUT, where the 12AT7 etc, is used as an input tube to an LTP with low u triodes, the usual maximum signal is less than 10 vrms. In this region, if RL of the SET input 12AT7 is a CCS, the thd can be kept below 1%. But it matters not if the RL is a generous value of ohms in a resistor and max thd is higher because the range of operation where the tube is used is far less than 10vrms, and the slight 2H is negligible. Everytime I have replaced a 12AX7 which is more linear than 12AT7 and at the input of a power amp, I have got a more precise sound, especially with massed strings and brass. Linearity isn't everything.... 12AY7 is also a nice tube for intermediate gain between a 12AX7 and 6SN7.... Now, for those who don't know them, the 6SN7 and 12AT7 are crisp, precise tubes. It used to be received wisdom that you didn't use two of the same tubes in sequence because that would magnify their disabilities. When I was still young enough to believe it smart to smash ivory, I reckoned that if one 6SN7 is precise, two would magnify the precision. Others think it results in an acid sound that rips off your ears. Maybe more variables are involved than you indicate, and some not all that explicable, which I don't mind because not everything in life has to be able to be explained. (In sighted tests. In blind tests the same "experts" choose the 2x 6SN7 drivers first over anything else except, sometimes, the 417A--which depends on high current rather than true inherent linearity--and a 6SL7/ 6SN7 combo which is definitely warmer even if less precise. See Steve Bench's netsite for some good silent driver arrangements with these tubes.) SNIPPED, PATRICK'S ARGUMENT TO LEAVE JUST THE DATA BECAUSE I'M ON ANOTHER TRACK: The 12AT7 has moderately low Ra of around 20k at Ia = 4 mA, and u = 55, MORE SNIP, DITTO Now Patrick goes off on another tangent: I have heard music via a pair of pure class A 300B in PP, which give around 18 watts, and its very nice, and will give the SET boys some real worries, but alas, PP has been out of favour for some years at least. The PP 300B amp I last heard was a "Vincent" amp from a cheap supplier in HK.... Not too much wrong with the sound I thought... PSE 300B give 16W without the crossover distortion of PP. PP 300B are not really out of favour. Crossover distortion in class A PP triode amps? Gee, I always thought there was none in pure class A PP. But there is a smidgin of 3H; declining to zero as power is reduced, wheras amps with Xover Dn display a hump in the amount of 3H at about a watt to two, or more, because that's where the crossover point between A and AB is located, and there is this dynamic change in gain of the amp during the cycle. Triode amps are the most forgiving of all with AB, since few other odds are made because of the gradual turn off transfer function of the triodes. Pentodes and beam tubes are sharp cut off, and you get far worse odd orders at the Xover region, and because the multigrids have gain more determined by gm x RL, and gain varies more abruptly in the cycle, since the Ra is high. For instance, Lynn Olson, another ultrafidelista opinion-former, built a PP 300B amp. The circuit is on his netsite. Price the iron he requires to make it work as well as a simple SE amp, then we can talk again about whether the additional advantages of his approach are worth the cost, the effort, the investment in a higher level of expertise, and so on. I admire Lynn's amp as a work of art, but I spent my own time and money instead on developing speakers that don't require such a monster. My clients have not the tenacity and passion to develop or buy sensitive speakers to allow low power SET amps to be used, Alas, those who did delve into horns produced only boxes good for firewood. They go right out and buy mainly 89 dB/W/M speakers. One sat my room with a sound level meter, and wanted to see if my SE35 amps produce the LF cone wobbles in my speakers (of that sensitivity,) whilst maintaining an average 95 dB SPL with some appalling rock and roll tunes. I found 36 watts from an SE amp was still able to do the business, and the amps were purchased. I am compelled to use PSE, or one large power tube at least. The bother is to make the sound musical, not bland, and not suppressed sounding. Real music lives, and sounds alive, and its what to expect from a system. I will have to look at Mr Olson's site.... One of my favourite amps, sitting on the floor at my right on hand of this thread, just waiting to be plugged in when the current disk finishes playing, is the Miyabe SEPP 300B (it is sold as a kit by Triode Supply Japan, costing 1600 USD when I reviewed it for Glass Audio a few years ago; see GA for a full discussion). SEPP is an interesting and useful compromise. Well, as a result of tonight's earlier comparisons between using amps with a single 13E1 in SEUL, and amps with 4 x 6CA7 with CFB, ( like an SE version of Quad II, but with 4 KT66 ), I have to say the 13E1 was the more musically involving amp to listen to. I doubt more than a watt was used all evening, we played a pile of vinly and CD. The vinyl came from a Michel Gyrodec with latest upgrades, a very decent MC and MC amp, the line stage is an SET with 6H30 in mu follower mode, Wima caps, decent R, DACT attenutaors etc, nothing anyone could say was crummy, and even Nordost speaker cabling. Speakers are Vienna Acoustic Motzarts, bases full of sand, with a sub-woofer powered to subtle but vital levels with a nice SS amp and control box. ( I may have finally got a good sub recipe.) Dame Joan Sutherland singing in 1962 Galetea with large orchestra was absolutely breathtaking. Who could ever say valves and vinyl cannot be accurate, refreshing, noiseless, and plain stunning? Even with ancient old records... Bill Evans, Miles Davis, wow..... I came away thinking that perhaps it'd be better to swap the 12BH7 I have driving the 4 x 6CA7 with a lone 6BQ5 in triode, and not be so concerned about choosing a triode driver whose increase in thd from wo to go is about the same as the SECFB output stage, so that with 5 ohms, the 2H cancellation is so complete across 3/4 of the whole power range that thd is less than many PP amps. One only gets a lot of 2H cancellation across a range of load values; between 3 and 6 ohms is where cancellation is most desirable. But necessary? I am left wondering. The SE35 with 6CA7 has 12 dB less thd than the SEUL at 1 watt and with 5 ohms. At 1 watt, the SEUL 13E1 has about 0.05%, and its virtually all 2H. The SE35 thd has other harmonics, 3H, 4H and 5H, but these are at the 0.01% level, and obvious on the CRO because the 2H is missing. Both lots of amps use about the same amount of applied NFB. Some NFB is required, because Rout from UL, or CFB amps is always too high without NFB. I don't really think thd measurements mean much if the levels are below 0.1%. So I am left with co-relating what I hear to what engineering I shall use based on a concept of mustering the better sonics associated with a driver with higher idle current. I have a hunch better music can be had, with a leetle further tinkering..... About 28 watts class AB1 is possible, Patrick Turner. 24W Class A1 out of triode-linked EL34PP will sound so much better. It really does seem to me that any speakers which require more than 24W must justify their existence. Since a 24W Class A1 EL34 PP amp is relatively cheap and easy to design and build, any bigger amp must therefore justify its existence. But why not just use a pair of KT90? Pd rating is twice the EL34's, so PO could be a lot more. I heard real music from Joe Rasmussen's "Forced Symetry" PP amps in 1996, at a hi-fi store demo here in the ACT. These were *very* modified Dynaco MkIII, with KT90 in triode, 600v on the plates, and which gave 40 watts, AB1, to power Duntech speakers. 20 people were present on the saturday to hear Allen Wright give everyone a lecture about FS, and how it worked. I wasn't so sure FS was such a big deal, and I had more than a few words to say in the ASON newsletter in the following months. But, fact is, the KT90 triodes were doin it ok, and I doubt a pair of EL34 could have. But the venue was a large room with 20 ppl, so fairly high levels were needed, nobody saw any cone shuffle, or heard any clipping. Anything that hogs out more power, or requires much NFB to work, seems to me like giving yourself the clap so you can prove how clever you are by "curing" it with a bandaid. What's much NFB? To me it is as much as Mullard would have deemed correct in 1960. Good for sales ppl to brag about. Good for nothing else. And a large amount can only be used with a lot of phase tweaking circuitry in place, and with a really good OPT. Alas, ppl try using lots of NFB without either..... 24 watts from a pair of EL34 in A1 PP triode is a lot, and its very AB, ie, not a lot of class A. 12 watts in triode from a pair is more like one would get in pure class A, but when a quad of EL34 are used, then you can have a more reasonable 30 watts class AB, but with a lot more A%. THEN you can still have a high RL a-a, thus usually have a high OPT impedance ratio. Thus the Ra-a is transformed down at the sec to a much lower figure than with just two EL34 with a low Z ratio OPT. So if Ra-a is 1,300 ohms with a quad of tubes, and OPT is 7k to 5 ohms, then Rout = 0.928 ohms, and if Rw = 0.25 ohms, total Rout without any loop FB is around 1.2 ohms. thd at 25 watts = 1%, so at 1 watt, its = 0.2% or less, mainly 3H, and I leave it to better minds than mine to decide if that is seriously awful enough to scurry away to the aid of say 9 dB of global NFB to reduce the thd to below 0.1%. 1 watt gives perhaps 90 dB SPL, and is well above what my average listening levels are, so I could get away with no global NFB, and just let the natural NFB inside the tubes do its work because we have the tubes operating in mainly voltage mode, not current mode. The Allen Wright / Joe Rasmussen camp reckoned in 1996 that one only needs enough global FB applied around a triode amp to reduce Ro to give a sufficiently good damping factor and not a stitch more, for reasons to do with delays, and that delays of the micro-detail signal can't be fixed with NFB. And in fact, the more NFB applied, the worse things get, they opined. "Micro-detail" isn't explained in RDH4, or any other texts I am aware of, but its a modern way of describing the sonics of musical differences heard between amps, speakers etc, which are not fully explainable simply in terms of thd, imd, or phase shifts. Anyway, I figured that for low NFB, or none at all, a low amount of phase shift and literal delay was a plus in any amp *before* any global NFB was applied, so a really beaut OPT is needed if you don't use any global NFB. The NFB user camp says that a poor open loop response with a poor OPT is OK because NFB will correct the delay, ie, the phase shift, and push the sagging response back up level. Plenty of accountants in the pro-lots-of NFB camp. I have always thought a wide bandwidth OPT is desirable whether you use NFB or not. Then when some NFB is used, there is little "delay", or micro-detail errors to fix, and the sound isn't smeared. In two demos at ASON meetings in 1997, Joe and I were able to demo amps which used a somewhat different attitude to NFB. In the first demo, his amps used about 6 dB around 4 6550 PP triodes, mine used 16 dB around 2 6550 in UL, and there was no majority conclusion reached by the gathered audiophiles that global NFB was an evil beast it had been made out to be. In the second demo, my SEUL amps humbled Joe somewhat, and I overheard a dude during the coffee break say to a mate, " this shows NFB don't **** up the SE sound, eh " I recieved a lot of conflicting comments during the days I spent with the fellas, but I do remember thay stayed late to hear the music, and I saw a sea of smiles. It does not mean much, but I remain flexible about NFB.... Each unto his own. Patrick Turner. Andre Jute Here is Patrick's original in full: Patrick Turner wrote: Something else needs to be said about bootstrapping the pentode RL of the schematic in the mifddle of the page 523. The CPI has R4, 20k for the anode load, and then 40k from cathode to 0V, R5. There is another R2, 40k from B+ taken to the junction of R1/C2. Thus R2 is effectively is in parallel with R5, and coupled via C2. Now I reckon this arrangement of R loads means that at large signal voltage levels, the output from the CPI cathode will cut off sharply above a threshold because of the ratio of R5 to R2, which is 1:1, where ideally to make the cut off distortion occur at a high voltage threshold, the cap coupled R should be a much higher value than the R carrying DC to/from a tube. The penomena is no different to where one has a cathode follower, and the cap coupled load is reduced, so that if RL dc = cap coupled RL, then if the Ia is cut right off, the maximum downward voltage travel is 1/2 the quiescent voltage between cathode and 0V. A CPI has to perhaps make a quite a few volts to drive an output stage, and unless there is a nice large voltage between the cathode of the CPI and 0V, there will be limiting of the voltage travel due to the voltage divider effect of the two RLs being equal. It would therefore not be such a good idea to use this circuit to drive large octal outputs, since the drive is usually over 25 vrms max to each g1. With a normal R loaded pentode stage driving a CPI, the gain will be around 6 times less than the bootstrapped gain pentode, so to boost the gain without the ill-effects of early cut off I have described, a CCS could be used with the pentode, in the form of a triode as one would use in a mu follower config, where the R between top tube is around 10 times a normal cathode biasing R, so about 22k. If the top tube has U = 100, like a paralleled 12AX7, the load seen by the bottom tube becomes about 100 x 22k // whatever biasing R for the top tube is, but at least RL for the pentode would be about 1M, and if U of the pentode is 3,000, Ra = 1M, say for a 6AU6, at Ia = 1.5 mA then gain = 1,500, and that's plenty. Direct coupling of the input pentode is possible to the CPI grid. I'd still prefer to use cascaded low u triodes for the input drive amp, since the linearity is maybe 10 times better. But the Williamson with all 6SN7 with KT66 triode outputs, ( or better, KT88/6550/300B outputs ) and using 20 dB of GFB needs about 2vrms input for full power. PPL may not agree with 20 dB of GFB, and ok, its your choice, and I would maybe use only 12 dB for a triode output stage, and that would reduce Vin needed to about 0.8vrms. Also a 12AT7 be used for the input triode and CPI, since the gain will be more than double the 6SN7 case, and Vin could be reduced to about 0.35vrms, for about 20 watts of output power, and no preamp is needed, just a 50k gain pot. Certainly with KT88 outputs, and with PO max = 28 watts AB1 triode, there is plenty of sensitivity, and no need for a preamp. The 12AT7 has moderately low Ra of around 20k at Ia = 4 mA, and u = 55, so the gain will be around 35 for the input triode, and the CPI will be extremely low distortion since gain reduction is from around 32 to about 2, and the signal level is low, since its a williamson balanced voltage drive amp that does most of the work to make the voltage drive to the output tubes. I have heard music via a pair of pure class A 300B in PP, which give around 18 watts, and its very nice, and will give the SET boys some real worries, but alas, PP has been out of favour for some years at least. About 28 watts class AB1 is possible, and considerably more in class AB2, but i've never ever actually made a class AB2 amp; I have always thought it better to use 4 output tubes to get the extra power lazily, rather than work the guts out of just two tubes. Therefore the low thd sweetzone of the amp is widened, over what one has with just two outputs. Plus the OPT is usually easier to wind, or a 5k to 4,8,16 OPT is easier to source, and we *want* a high value RL if we want a high % of class A, and natural linearity. Patrick Turner. |
#11
|
|||
|
|||
Chris Hornbeck wrote: On Sat, 19 Mar 2005 01:23:45 +0100, Fran=E7ois Yves Le Gal wrote: On Sat, 19 Mar 2005 00:05:16 GMT, Chris Hornbeck wrote: The flat spot on the B-H curve is easily observable; try it for yourself. It is, but it's not caused by any form of Weiss-related phenomenon. http://www.beigebag.com/case_xfrmer_4.htm This is an interesting reference, but is unrelated to the issue (based on a quick look; I'm off to the opry). The B-H curve also has a small flat spot near zero crossing. This is the one and only reason anyone would ever bother with the single-ended amplifier grief. When I first hit on my harmonic distribution theory about what makes some amps sound right while others which measure better (in total) chilled me, I was so amazed when tests showed really tiny percentage differences that I had the test repeated while I read Olson again to discover where I went wrong. I didn't go wrong. I just didn't know as much as I thought I did, even though I was trained as a psychologist and in advertising had a larger market research budget than anyone else and spent very much of it on placebo testing (what the poseurs who haunt audio newsgroups call ABX testing). It wasn't rocket science to conclude that there are subliminal reactions whose result we can see, if rather faintly, but for which we cannot describe the cause with any precision. Though I would be delighted if it turns out to be something as easily described as the width of a flat spot on a B-H curve, I rather doubt any effect in audio is really that singular or even straightforward. But for some, who have tried it and found it significant, it is surprisingly important. Surprised the **** outta me, anyway. Who knew? You're following through on a thread I started, Chris. I knew as early as 1992 when I returned to tubes. I suspect a handful of American hobbyists knew. The Japanese knew by osmosis and I wouldn't be surprised to discover that Jean Hiraga described the effect. There's something surprising here at the psycho-acoustic interface. As I have noted before, in blind tests professional musicians prefer Class A1 PP triodes to even 300B single ended often enough to be statistically indicative. Professioanl musicians are, as far as I can tell, no more sensitive to phase than the general body of experienced audiophiles. However, people of any persausion who are sensitive to phase are also likely to choose SE. Connection? Anomaly? Who knows. By the time I realized there was something near enough statisticically significant in it to justify further tests the listening groups were dispersed and I just lack the time (and it must be said the motivation to commit expense and time for results that will be sneered and jeered at by slime like this French jerk Le Gal for the 'scientific' reason that he doesn't like me). Off to "The Tender Land", Do report back. Just checked my record catalogue and I don't have a copy so I can't even listen along with you on record. Chris Hornbeck Andre Jute Listening to Bob van Asperen playing CPE Bach's Prussian and Wurtemburg Sonatas on the harpsichord. Using my Miyabe (Triode Supply, Japan VPD3000 in Oz for some weird reason) SEPP 300B and ESL63. A super clean sound, but when my cat gets off my lap I will plug in the 57s which are truly ideally suited to the harpsichord. |
#12
|
|||
|
|||
|
#13
|
|||
|
|||
On Sat, 19 Mar 2005 02:16:37 +0100, François Yves Le Gal
wrote: "Why is there a gap or 'hole' in the middle of a typical B-H curve? What causes this? I don't know. Personally, I've found that anyone who claims to understand magnetism *or* electricity is trying to blow smoke up your ass. Maxwell's silver hammer. The zero-crossing flat spot is easily observable by anyone interested, and has the same effect as crossover distortion in underbiased gain stages. Don't take my word for it; test to your own satisfaction. You may also have a look at Magnetics Designer http://www.intusoft.com/mag.htm A beautifully made web page, and probably a worthwhile program, but not applicable to this discussion. We're not talking about a large signal problem. In audio, one of the most dangerous assumptions is of monotonicity. We assume that as signal levels get smaller, all bad things in our designs will also get smaller. There are, unfortunately, several exceptions to this rule, and iron cores are one of those exceptions. Magnetic tape would be too, if it weren't so noisy. "Crossover distortion" is the best known example. And all of these errors in monotonicity are cured with bias. Chris Hornbeck |
#14
|
|||
|
|||
|
#15
|
|||
|
|||
On Sat, 19 Mar 2005 06:16:45 GMT, Chris Hornbeck
wrote: On Sat, 19 Mar 2005 02:16:37 +0100, François Yves Le Gal wrote: "Why is there a gap or 'hole' in the middle of a typical B-H curve? What causes this? I don't know. Personally, I've found that anyone who claims to understand magnetism *or* electricity is trying to blow smoke up your ass. Maxwell's silver hammer. The zero-crossing flat spot is easily observable by anyone interested, and has the same effect as crossover distortion in underbiased gain stages. Don't take my word for it; test to your own satisfaction. Even better, don't use an output transformer! -- Stewart Pinkerton | Music is Art - Audio is Engineering |
#16
|
|||
|
|||
Chris Hornbeck wrote: On Sat, 19 Mar 2005 02:16:37 +0100, François Yves Le Gal wrote: "Why is there a gap or 'hole' in the middle of a typical B-H curve? What causes this? I don't know. Personally, I've found that anyone who claims to understand magnetism *or* electricity is trying to blow smoke up your ass. Maxwell's silver hammer. The zero-crossing flat spot is easily observable by anyone interested, and has the same effect as crossover distortion in underbiased gain stages. Don't take my word for it; test to your own satisfaction. You may also have a look at Magnetics Designer http://www.intusoft.com/mag.htm A beautifully made web page, and probably a worthwhile program, but not applicable to this discussion. We're not talking about a large signal problem. In audio, one of the most dangerous assumptions is of monotonicity. We assume that as signal levels get smaller, all bad things in our designs will also get smaller. There are, unfortunately, several exceptions to this rule, and iron cores are one of those exceptions. Magnetic tape would be too, if it weren't so noisy. "Crossover distortion" is the best known example. And all of these errors in monotonicity are cured with bias. Chris Hornbeck With the experiments I have done using iron cored inductors in PP drive amp circuits, such as one may have in an IST, and where the source resistance is high, the iron caused distortion is as bad as crossover distortion in the output stage of a poorly biased PP pair of output tubes. For this reason, I cannot ever recommend pentode use to drive a PP IST, because of the mainly 3H produced at precisely the level at which we want low thd, ie, aty low levels. With extremely low signals, the inductive reactance of the coil, ZL, is very low, so gain is low, and then as voltage is increased, ZL also increases, so you get a sine wave that tends to have flats at the zero crossing, and this is in effect the fundemental plus a 3H signal that tends to peak up the fundemental, and this 3H is of opposite phase to the sort of 3H which causes flattening of the peaks of the fundemental With iron wounds, the flattening occurs sure enough when voltage signals become very high, since the ZL during a cycle is non linear. The practical answer to minimising problems caused by iron distortion is to use a low source resistance. The actual iron cored impedance during a cycle could be modelled using diodes and R components, and their finite value can be established, and its this model network which is strapped across the load we may want to drive, which is also a parallel resistance or impedance with the source resistance. So if we choose to use a pair of 6AU6 with an IST, we will get some really poor response curves and distortion quantities unless we are very prudent with NFB, or passive resistors strapped across the windings. AFAIK, it is very difficult to model a diodes + resistances nework to cancel out the iron wound item's change in impedance at different voltage levels, since a large F range is involved. The 6AU6 would have Ra-a of around 2 megohms. If Lp is say 100H, the LF pole is at 3.2 kHz, and if Cshunt of the L is 200pF to each end of the primary, then the HF pole is at 800Hz. This all indicates that you will never see the HF or LF poles where one would expect them. Looking at the wave forms at almost any F will reveal attrocious levels of odd order distortion, because the non linear proerties of the iron will dominate the load seen by the pentodes, and gain = about gm x RL. Using 50k across the 1/2 primaries would make a reasonable load for a reasonable amount of gain from the 6AU6, and then for 100H, the LF pole would be about 160 Hz, and HF pole at 16 kHz. Its still very poor, and thd would be a varying amount at various output voltages and F, and be quite attrocious and intolerable, and the iron non linear "mystery impedance" is still a large % of the total load seen by the pentodes, and perhaps it varies with voltage and F between 10k and 200k+. Its not unsual to calculate a saturation voltage of say 100 vrms for 50 Hz for the IST primary, when iron Dn will rise to say 10% or more and be all odd order, and still be a few % at 10vrms. The use of GOSS will improbve matters enormously, but still Dn will be way above what the tubes are making at 10vrms. Consider triodes. Say we use triodes, with Ra-a = 4k, ( could be a pair of EL84 in triode, Ia = 20 mA, common cathodes, in an LTP driver.) RL shunting Lp can be omitted, or perhaps only consist of an R + C connected across the secondary, or 1/2 secondaries in a CT sec. R loading is only required at HF, hence the careful RC load. Then with 100H, the LF pole = 6.37 Hz, ok, and HF pole is at 400 kHz, also ok. But there is still the rotten non linear iron impedance, but we will find that the Dn at low voltages may not exceed 0.1 % Its not uncommon to be able to plot the Dn to rise from zero vo to 0.1% at say 10vrms, then it falls to 0.02% at 50v, then it rises again until its a lot at saturation, which is at a voltage we will never be using, and so does not matter. Readers should now refer to the schematic at http://www.turneraudio.com.au/websch...ma550w335h.gif In this we see a PP LTP driver circuit to drive a dozen KT88s, although it also works well for 2 x KT88, or any other output stage needing a low thd drive voltage with a high voltage ceiling if we have a high enough B+. There is a an L with CT used to feed the EL84 in triode of the LTP. Now these triodes are set up with some local current current FB in the form of 470 ohm cathode resistors, which in this case are partially bypassed with 100 ohms + 0.0082 uF to compensate for the phase lag and gain drop in the circuit, but without causing any stability problems. The 470 ohms do increase the Ra-a effectively to 24 kohms. There is 9.4 kohms in series with the ends of the L and the anodes, so at extremes of F the Lp ans Csh does not shunt the tube output, so the gain variation is controlled, along with phase shift. In this case the Lp consists of thousands of turns around a GOSS core of 32 stack of 25 tongue, and ungapped, and U max = 5,000 at 50 Hz. So the maximum L allowing for frequency is well above 100H max and the L mainly acts to increase the impedance of the DC carrying elements to the triodes, so they enjoy a very high load impedance for all of the audio band, and so tube Dn is minimised so that at 200vrms a-a, Dn is less than 1%. Iron caused Dn is also very low, and I found the use of the series R did allow some distortion to appear at the coil ends, but less at the triode anodes, since the 9,4 k R and Ra of the triodes form a divider, lessening the capricious iron effcts to ruin our music. The iron Dn never rises above 0.1% at any voltage, and since I have GNFB, the low level Dn for the amp was less than 0.2% at around 20 watts. Now since I addopted this little used technique of iron cored coils *plus* resistances in series, I have now omitted the 470 ohm Rk from such triode circuits, and found that EL84 in triode as set up with a low stress Ia of less than 20 mA each will last forever, and they don't drift at DC and the current FB isn't needed, and thus with commoned cathodes the compensation networks across the k-k and Rk are all not needed. The 9.4 k in series with each anode and LP ends can be reduced to 4.7k. Thus Ra-a is only 4k, not 24k, and the iron caused Dn seen at each anode is reduced about 4 times to negligible levels, and you have a fine driver stage free of the phase shifting, gain zapping effects of the parasitic Lp and Csh, so GFB can be used around the whole circuit if desired, or not desired, and still the driver Dn will always be low, and a *lot* lower than if you have dc carrying resistors of say 10k only between anodes and the B+. It would be possible to use an air gapped core, but then for this sort of item to work, the turns have to be increased, and core, and the LF pole is higher, since Lp is a lower value once you gap. But Lp is amore constant value, and it changes less during a cycle, so there is less Dn. The alternative to setting up a driver with such an efficient load configurationn is to use a pair of high voltage transistors to make a pair of CCS. But in the way I ahve done it the CT choke acts partially like a transformer, an the outputs are locked a bit together magnetically. For where GNF is very low, it is possible to omit the series R between LP and the EL84 anodes, and just rely on the high inductance with low shunt C to give enough bandwidth because Ra-a is so low. In this case the two driver tubes are combined in action very well, as they operate together on the load, and unlike an LTP with purely R loads, the LTP with a centre tapped auto transformer will have a much better ability to work into a low load, so any variations of load are better handled. The right sort of inductor will minimise Dn to negligible levels. But often we see THD profiles of an amp start off at below noise at 0.0 watts, and then inexplicably leap to 0.02% at 3 watts, then slowly rise slightly faster than a straight line with output voltage to say 0.25% at clipping. Its the typical profile of a class AB amp with GNFB. The sudden increase in 3H at the start of the amps power curve is the iron plus tubes. 0.02% at 3 watts is quite OK. People I know have tried this driver technique for SE amps, where the choke is a gapped one carrying DC, but able to provide a high dynamic impedance to the triode driver, and with a series R, to "isolate the parasitics" and yet avoid loading the driver tube down with a low resistance, and paying the penalty of 12 dB more driver thd. They have told me they liked the sound. But in an SE amp, perfect, or close to perfect linearity in a driver tube where the output tube produces say 4% of 2H at full power, there is little cancellation of 2H from the driver triode. I Have found it isn't necessary to persue perfect even order distortion cancellations between driver stages and outputs, because SE amps are never used near the levels where Dn has climbed to their max, and much varied load valuesof speakers prevent the gains of the tubes and the thd quantities from staying constant , so cancellation can only occur substantially across a limited range of load values. Put more simply, *distortion voltage* cancellation in SE amps is far more difficult to achieve than *distortion current* cancellation in PP amps. The iron caused distortion in an SE OPT is very low at first, because iron is biased magnetically in one direction, so changes in field strengths and L impedance for small signal changes are generaly very low, and linear in each direction, so the iron caused Dn in an SE amp at normal low levels used is usually quite utterly negligible. Some would say the SE amp thus performs sonically better than the PP amp. BUt alas I see not a huge amount of evidence to support this idea universally. AND, one of the very best sounding amps I have ever had the luck of the gods to construct has plain old non oriented iron cores, maximum U = 3,500, and its measurements indicate there isn't much of a problem with the iron at all, although there is a fair bit of iron Dn at 14 Hz and at 50 watts, when saturation has begun. One would expect more observable iron Dn at higher F but the Dn seen is that which one would only ever expect from the tubes, ie, 1% at 50 watts into 8 ohms with a quad of 6550 with 12.5% of CFB, and no GNFB. The thd declines almost linearly with output voltage and without the tell tale hump at low levels. RDH4 gives a formula for calculation of the iron Dn using a graph for GOSS of the 1955 era. AFAIK, the latest GOSS I use generates far less Dn than anything from 1955. Unfortunately, I have had not the time to learn the university physics and math associated with electro magnetics, so I am no expert. But how many experts have any relationship with audio transformer behaviour of the items they wind? I just wind things, and carefully measure them, and if they work under any expected conditions, they are accepted. If not, its back to the drawing board and another read of my books. Some experts are just drips turned off. Patrick Turner. The lelsslwthn with Is cre n In |
#17
|
|||
|
|||
On Sat, 19 Mar 2005 12:20:11 +0100, François Yves Le Gal
wrote: And all of these errors in monotonicity are cured with bias. Or dither. I've had to think (Please stop! Anything but that!) a lot about this, and can almost agree with this viewpoint. You obviously mean something more than the semantic similarity to D/A conversion and have a larger conceptual framework. Taken to an extreme, it raises a question: could transformer iron's zero-crossing discontinuities be avoided with actual (ultrasonic?) dither? IOW, is a flat spot problem solvable by randomizing all small signals sufficiently? This is what's done in D/A conversions, and in any digital word-length shortening, and could be considered to be what's done in magnetic tape recording. Thanks for a very interesting question, Chris Hornbeck |
#18
|
|||
|
|||
On Sat, 19 Mar 2005 07:32:21 +0000 (UTC), Stewart Pinkerton
wrote: Even better, don't use an output transformer! Or, use high impedance speakers. (QUAD... ! ) Your point's well taken that low-perveance electronics couples poorly with low-impedance loads. I'd guess that you and I would disagree conceptually only on the relative importance of various issues. Intrinsic linearity, monotonicity, headroom (not to be scoffed at), classical issues of nonlinear distortion products, group delay errors, etc., harmonic multiplication, resistance to out-of-band-signal overloads of many kinds, bla bla bla. We all draw our individual weightings of lots of factors into even something as relatively simple as an audio amplifier. There's a good reason why religion and politics aren't discussed at the dinner table. Too bloody complicated! Thanks, Chris Hornbeck |
#19
|
|||
|
|||
Chris Hornbeck wrote: On Sat, 19 Mar 2005 12:20:11 +0100, François Yves Le Gal wrote: And all of these errors in monotonicity are cured with bias. Or dither. I've had to think (Please stop! Anything but that!) a lot about this, and can almost agree with this viewpoint. You obviously mean something more than the semantic similarity to D/A conversion and have a larger conceptual framework. Taken to an extreme, it raises a question: could transformer iron's zero-crossing discontinuities be avoided with actual (ultrasonic?) dither? Anyone would think that the zero crossing non linearity is a serious problem. It is not a problem in OPTs that have good material and are well designed, as I pointed out in recent posts with clear examples. So why oh why would anyone want to place a 50 kHz bias oscillator in a tube amps? It would be impossible, since stability problems are bad enough without adding a supersonic oscillator. IOW, is a flat spot problem solvable by randomizing all small signals sufficiently? This is what's done in D/A conversions, and in any digital word-length shortening, and could be considered to be what's done in magnetic tape recording. Thanks for a very interesting question, Chris Hornbeck HF bias on an OPT is a solution looking for a problem. Its been thought of before, and the idea discarded. But don't let us stop you building a prototype. If you are going to have a bias HF current, why not go the whole hog, and use pulse width modulation, and/or make a digital amplifier. Patrick Turner. |
#20
|
|||
|
|||
Patrick Turner writes:
: : Chris Hornbeck wrote: : : On Sat, 19 Mar 2005 12:20:11 +0100, François Yves Le Gal : wrote: : : And all of these errors in monotonicity are : cured with bias. : : Or dither. : : I've had to think (Please stop! Anything but that!) a lot : about this, and can almost agree with this viewpoint. You : obviously mean something more than the semantic similarity : to D/A conversion and have a larger conceptual framework. : : Taken to an extreme, it raises a question: could transformer : iron's zero-crossing discontinuities be avoided with actual : (ultrasonic?) dither? : : Anyone would think that the zero crossing non linearity is a serious : problem. : It is not a problem in OPTs that have good material and are well : designed, as I pointed : out in recent posts with clear examples. : : So why oh why would anyone want to place a 50 kHz bias oscillator : in a tube amps? : : It would be impossible, since stability problems are bad enough without : adding a supersonic oscillator. : : : : IOW, is a flat spot problem solvable by randomizing all small : signals sufficiently? This is what's done in D/A conversions, : and in any digital word-length shortening, and could be : considered to be what's done in magnetic tape recording. : : Thanks for a very interesting question, : : Chris Hornbeck : : HF bias on an OPT is a solution looking for a problem. : : Its been thought of before, and the idea discarded. Has it? By whom, where, when? : But don't let us stop you building a prototype. : : If you are going to have a bias HF current, why not go the whole hog, : and : use pulse width modulation, and/or make a digital amplifier. Like http://www.davidberning.com/ ? |
#21
|
|||
|
|||
RdM wrote: snip a bit, : HF bias on an OPT is a solution looking for a problem. : : Its been thought of before, and the idea discarded. Has it? By whom, where, when? I don't know who, but just about every possible way to connect tubes up with all possible signals has been explored, and had there been money to be made using HF bias in a tube power amp, you'd expect it to have been done. So I can safely assume it has been tried by somebody, sometime, but obviously the benefits are zero since nothing came of whoever tried it. The idea may never have got off drawing board it seems. : But don't let us stop you building a prototype. : : If you are going to have a bias HF current, why not go the whole hog, : and : use pulse width modulation, and/or make a digital amplifier. Like http://www.davidberning.com/ ? That's the place I had in mind. Some how I find myself thinking that using bottles to do what is better done with power transistors or mosfets is a waste of glassware. I could be wrong, but bottles are great for analog amps. Digital amps will become a more common reality perhaps, and buyers will have more of a choice to make. So what? Maybe somebody invents a HF biased power amp successfully and maybe its doable, but I see only huge problems. Somebody must have thought about it before, and concluded similarly. But for the home diyer not concerned about making a buck, its an avenue of latter day audio research that may bear fruit. I hope its taste is sweet. Patrick Turner. |
#22
|
|||
|
|||
"Patrick Turner" wrote in message ... RdM wrote: snip a bit, : HF bias on an OPT is a solution looking for a problem. : : Its been thought of before, and the idea discarded. Has it? By whom, where, when? I don't know who, but just about every possible way to connect tubes up with all possible signals has been explored, and had there been money to be made using HF bias in a tube power amp, you'd expect it to have been done. So I can safely assume it has been tried by somebody, sometime, but obviously the benefits are zero since nothing came of whoever tried it. The idea may never have got off drawing board it seems. : But don't let us stop you building a prototype. : : If you are going to have a bias HF current, why not go the whole hog, : and : use pulse width modulation, and/or make a digital amplifier. Like http://www.davidberning.com/ ? That's the place I had in mind. Some how I find myself thinking that using bottles to do what is better done with power transistors or mosfets is a waste of glassware. I could be wrong, but bottles are great for analog amps. You are wrong about Berning's amps, Patrick. David Berning builds extremely linear all triode tube amps using sweep tubes connected as triodes, but with G1 connected to the cathode and G2 being driven. With some sweep tubes this can give dramatically better linearity than the conventional method of connecting G2 to the plate and driving G1. He then further improves linearity by using a solid state impedance converter (similar to a SMPS) which avoids most of the nonlinearities associated with conventional output xfmrs and presents a higher impedance load to the triode output tubes than is practical with conventional output xfmrs. His amps aren't very efficient but very few conventional amps can compete with their sonics. FT Digital amps will become a more common reality perhaps, and buyers will have more of a choice to make. So what? Maybe somebody invents a HF biased power amp successfully and maybe its doable, but I see only huge problems. Somebody must have thought about it before, and concluded similarly. But for the home diyer not concerned about making a buck, its an avenue of latter day audio research that may bear fruit. I hope its taste is sweet. Patrick Turner. |
#23
|
|||
|
|||
On Mon, 21 Mar 2005 19:59:20 +1100, Patrick Turner
wrote: Anyone would think that the zero crossing non linearity is a serious problem. War, oppression and starvation are serious problems. Homemade audio amplifiers are diversions. It is not a problem in OPTs that have good material and are well designed, as I pointed out in recent posts with clear examples. I'll just have to remain unconvinced. My personal experience has been otherwise, but that's why it takes all kinds to fill the freeways. So why oh why would anyone want to place a 50 kHz bias oscillator in a tube amps? An oscillator would not perform the function of dithering, which requires true randomness, like broadband (or colored broadband) noise. It would be impossible, since stability problems are bad enough without adding a supersonic oscillator. Stability is independent of (small-ish) signal presence. If not, you've got bigger problems to solve first anyway. Nobody has yet commented on whether dither is even an appropriate solution. I think probably not, but I've been wrong *lots*. And it would be particularly interesting to be wrong about this. Thanks, Chris Hornbeck |
#24
|
|||
|
|||
FT wrote: "Patrick Turner" wrote in message ... RdM wrote: snip a bit, : HF bias on an OPT is a solution looking for a problem. : : Its been thought of before, and the idea discarded. Has it? By whom, where, when? I don't know who, but just about every possible way to connect tubes up with all possible signals has been explored, and had there been money to be made using HF bias in a tube power amp, you'd expect it to have been done. So I can safely assume it has been tried by somebody, sometime, but obviously the benefits are zero since nothing came of whoever tried it. The idea may never have got off drawing board it seems. : But don't let us stop you building a prototype. : : If you are going to have a bias HF current, why not go the whole hog, : and : use pulse width modulation, and/or make a digital amplifier. Like http://www.davidberning.com/ ? That's the place I had in mind. Some how I find myself thinking that using bottles to do what is better done with power transistors or mosfets is a waste of glassware. I could be wrong, but bottles are great for analog amps. You are wrong about Berning's amps, Patrick. David Berning builds extremely linear all triode tube amps using sweep tubes connected as triodes, but with G1 connected to the cathode and G2 being driven. With some sweep tubes this can give dramatically better linearity than the conventional method of connecting G2 to the plate and driving G1. I am not an expert on Berning amps. I though he was doing a kind of digital amp, or an amp that worked like a SMPS. With all grids tied together, sure, you get a different sort of triode compared to g1 drive or g2 drive. Everytime I have tried G2 drive, I wasn't convinced it was worth the effort. I have not tried g2+g1 together. Have you any plate data curves for such an arrangement? He then further improves linearity by using a solid state impedance converter (similar to a SMPS) You have lost me..... which avoids most of the nonlinearities associated with conventional output xfmrs and presents a higher impedance load to the triode output tubes than is practical with conventional output xfmrs. His amps aren't very efficient but very few conventional amps can compete with their sonics. I have never heard or seen a real one. Maybe you are right. Patrick Turner. FT Digital amps will become a more common reality perhaps, and buyers will have more of a choice to make. So what? Maybe somebody invents a HF biased power amp successfully and maybe its doable, but I see only huge problems. Somebody must have thought about it before, and concluded similarly. But for the home diyer not concerned about making a buck, its an avenue of latter day audio research that may bear fruit. I hope its taste is sweet. Patrick Turner. |
#25
|
|||
|
|||
Chris Hornbeck wrote: On Mon, 21 Mar 2005 19:59:20 +1100, Patrick Turner wrote: Anyone would think that the zero crossing non linearity is a serious problem. War, oppression and starvation are serious problems. Homemade audio amplifiers are diversions. It is not a problem in OPTs that have good material and are well designed, as I pointed out in recent posts with clear examples. I'll just have to remain unconvinced. My personal experience has been otherwise, but that's why it takes all kinds to fill the freeways. I just don't see a great deal of evidence that at the zero crossing point at any signal level that there is a huge amount of distortion that warrants much intervention. So why oh why would anyone want to place a 50 kHz bias oscillator in a tube amps? An oscillator would not perform the function of dithering, which requires true randomness, like broadband (or colored broadband) noise. It would be impossible, since stability problems are bad enough without adding a supersonic oscillator. Stability is independent of (small-ish) signal presence. If not, you've got bigger problems to solve first anyway. Nobody has yet commented on whether dither is even an appropriate solution. I think probably not, but I've been wrong *lots*. And it would be particularly interesting to be wrong about this. I have an open mind, and it seems you wanna tickle the iron molecules arouns a bit, kinda like distracting them from misbehaving at each zero crossing point. I don't know how you'd do that, without applying a signal across the primary. The amp wouldn't like it, unless it has high Rout at the frequencies of the noise. But hey, maybe its doable if Fnoise is all above 50 khz, and you only want a small "dither" signal. A pink noise source above 50 khz might do, and could be applied to the secondary. Just a small voltage, only 0.5 v max, not enough to upset speakers.... It could injected from the high impedance of a collector circuit of a couple of power transistors. Then rig an on-off switch, to test if it makes any difference. Patrick Turner. Thanks, Chris Hornbeck |
#26
|
|||
|
|||
On Wed, 23 Mar 2005 00:59:01 +1100, Patrick Turner
wrote: I have an open mind, and it seems you wanna tickle the iron molecules arouns a bit, kinda like distracting them from misbehaving at each zero crossing point. I don't know how you'd do that, without applying a signal across the primary. The amp wouldn't like it, unless it has high Rout at the frequencies of the noise. But hey, maybe its doable if Fnoise is all above 50 khz, and you only want a small "dither" signal. A pink noise source above 50 khz might do, and could be applied to the secondary. Just a small voltage, only 0.5 v max, not enough to upset speakers.... It could injected from the high impedance of a collector circuit of a couple of power transistors. Then rig an on-off switch, to test if it makes any difference. Perzactly my thought. Anyway, I've written to an acquaintance who's a math monster, and hopefully he can answer the question of whether the whole idea is as conceptually bogus as it seems. Thanks, Chris Hornbeck |
#27
|
|||
|
|||
On Wed, 23 Mar 2005 00:48:03 +1100, Patrick Turner
wrote: He then further improves linearity by using a solid state impedance converter (similar to a SMPS) You have lost me..... You're not missing anything. Chris Hornbeck |
#28
|
|||
|
|||
"Patrick Turner" wrote in message ... FT wrote: "Patrick Turner" wrote in message ... RdM wrote: snip a bit, : HF bias on an OPT is a solution looking for a problem. : : Its been thought of before, and the idea discarded. Has it? By whom, where, when? I don't know who, but just about every possible way to connect tubes up with all possible signals has been explored, and had there been money to be made using HF bias in a tube power amp, you'd expect it to have been done. So I can safely assume it has been tried by somebody, sometime, but obviously the benefits are zero since nothing came of whoever tried it. The idea may never have got off drawing board it seems. : But don't let us stop you building a prototype. : : If you are going to have a bias HF current, why not go the whole hog, : and : use pulse width modulation, and/or make a digital amplifier. Like http://www.davidberning.com/ ? That's the place I had in mind. Some how I find myself thinking that using bottles to do what is better done with power transistors or mosfets is a waste of glassware. I could be wrong, but bottles are great for analog amps. You are wrong about Berning's amps, Patrick. David Berning builds extremely linear all triode tube amps using sweep tubes connected as triodes, but with G1 connected to the cathode and G2 being driven. With some sweep tubes this can give dramatically better linearity than the conventional method of connecting G2 to the plate and driving G1. I am not an expert on Berning amps. I though he was doing a kind of digital amp, or an amp that worked like a SMPS. With all grids tied together, sure, you get a different sort of triode compared to g1 drive or g2 drive. Everytime I have tried G2 drive, I wasn't convinced it was worth the effort. I have not tried g2+g1 together. Have you any plate data curves for such an arrangement? Hi Patrick, Berning doesn't tie the grids together, he ties G1 to the cathode (along with G3). G2 becomes the driven grid. Audio magazine had a cover story many years ago on a 100W/ch Berning amp that used this scheme with a conventional OT. They included curves for the output tubes in conventional triode mode (G1 driven, G2 tied to the plate) and Berning's triode connection. The curves were dramatically more linear with Berning's hook up, especially near the cutoff end of the curves. This allowed Berning to run his amps very near class B (3mA ilding current per tube, IIRC) without the distortion this would cause in a conventional triode PP amp. The downside is that 5 times more drive signal is required in this mode, thus my comment that his amps are not exactly efficient. I have a 30W/ch Berning amp (with a conventional OT), and the cathode followers that drive the output tubes have 800V across them (+500V on the plates, -300V to the cathode resistors) to get the voltage swing required to drive the output stage to 30 watts. I have a superficial understanding of the switch mode impedance converter he's using in place of an OT in his newer amps, but better minds than mine have become hopelessly confused trying to analyze the circuit ;-) FT He then further improves linearity by using a solid state impedance converter (similar to a SMPS) You have lost me..... which avoids most of the nonlinearities associated with conventional output xfmrs and presents a higher impedance load to the triode output tubes than is practical with conventional output xfmrs. His amps aren't very efficient but very few conventional amps can compete with their sonics. I have never heard or seen a real one. Maybe you are right. Patrick Turner. FT Digital amps will become a more common reality perhaps, and buyers will have more of a choice to make. So what? Maybe somebody invents a HF biased power amp successfully and maybe its doable, but I see only huge problems. Somebody must have thought about it before, and concluded similarly. But for the home diyer not concerned about making a buck, its an avenue of latter day audio research that may bear fruit. I hope its taste is sweet. Patrick Turner. |
#29
|
|||
|
|||
FT wrote: "Patrick Turner" wrote in message ... FT wrote: "Patrick Turner" wrote in message ... RdM wrote: snip a bit, : HF bias on an OPT is a solution looking for a problem. : : Its been thought of before, and the idea discarded. Has it? By whom, where, when? I don't know who, but just about every possible way to connect tubes up with all possible signals has been explored, and had there been money to be made using HF bias in a tube power amp, you'd expect it to have been done. So I can safely assume it has been tried by somebody, sometime, but obviously the benefits are zero since nothing came of whoever tried it. The idea may never have got off drawing board it seems. : But don't let us stop you building a prototype. : : If you are going to have a bias HF current, why not go the whole hog, : and : use pulse width modulation, and/or make a digital amplifier. Like http://www.davidberning.com/ ? That's the place I had in mind. Some how I find myself thinking that using bottles to do what is better done with power transistors or mosfets is a waste of glassware. I could be wrong, but bottles are great for analog amps. You are wrong about Berning's amps, Patrick. David Berning builds extremely linear all triode tube amps using sweep tubes connected as triodes, but with G1 connected to the cathode and G2 being driven. With some sweep tubes this can give dramatically better linearity than the conventional method of connecting G2 to the plate and driving G1. I am not an expert on Berning amps. I though he was doing a kind of digital amp, or an amp that worked like a SMPS. With all grids tied together, sure, you get a different sort of triode compared to g1 drive or g2 drive. Everytime I have tried G2 drive, I wasn't convinced it was worth the effort. I have not tried g2+g1 together. Have you any plate data curves for such an arrangement? Hi Patrick, Berning doesn't tie the grids together, he ties G1 to the cathode (along with G3). G2 becomes the driven grid. Well that's plain old G2 drive. Its very linear sometimes, and gives you a high Ra triode with low U. Audio magazine had a cover story many years ago on a 100W/ch Berning amp that used this scheme with a conventional OT. They included curves for the output tubes in conventional triode mode (G1 driven, G2 tied to the plate) and Berning's triode connection. The curves were dramatically more linear with Berning's hook up, especially near the cutoff end of the curves. This allowed Berning to run his amps very near class B (3mA ilding current per tube, IIRC) without the distortion this would cause in a conventional triode PP amp. The downside is that 5 times more drive signal is required in this mode, thus my comment that his amps are not exactly efficient. I have a 30W/ch Berning amp (with a conventional OT), and the cathode followers that drive the output tubes have 800V across them (+500V on the plates, -300V to the cathode resistors) to get the voltage swing required to drive the output stage to 30 watts. Could I assume the use of a conventional triode driver circuit is used to get the voltage drive? Basically one needs a drive voltage equal or greater than the anode swing voltage. Wouldn't CF operation of the output tubes be more effective? I have a superficial understanding of the switch mode impedance converter he's using in place of an OT in his newer amps, but better minds than mine have become hopelessly confused trying to analyze the circuit ;-) I have never focused long enough. Patrick Turner. FT He then further improves linearity by using a solid state impedance converter (similar to a SMPS) You have lost me..... which avoids most of the nonlinearities associated with conventional output xfmrs and presents a higher impedance load to the triode output tubes than is practical with conventional output xfmrs. His amps aren't very efficient but very few conventional amps can compete with their sonics. I have never heard or seen a real one. Maybe you are right. Patrick Turner. FT Digital amps will become a more common reality perhaps, and buyers will have more of a choice to make. So what? Maybe somebody invents a HF biased power amp successfully and maybe its doable, but I see only huge problems. Somebody must have thought about it before, and concluded similarly. But for the home diyer not concerned about making a buck, its an avenue of latter day audio research that may bear fruit. I hope its taste is sweet. Patrick Turner. |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
RDH4 | Vacuum Tubes | |||
RDH4 mirror is back on-line | Vacuum Tubes | |||
Gregg's RDH4 mirror? | Vacuum Tubes | |||
RDH4: I give up.... | Vacuum Tubes | |||
RDH4 CD's--how many different ones are there? | Vacuum Tubes |