Home |
Search |
Today's Posts |
#1
|
|||
|
|||
6LU8 SET or SEUL Amp
I did this one about 12 years ago. It compares very well with what you can do with a 2A3, but at a much lower cost. And 6LU8s are still cheap, all built while production knew how to make a reliable tube.
The distortion tests were done with an HP302A Wave Analyzer for both triode & UL modes at One Khz. Another short set at 100 Hz, One Watt compares Hammond 125E Universal with a Hammond 1628SE. The advantage of more iron is quite obvious. ULTRALINEAR WITH FEEDBACK WATTS 1 2 3 4 5 6 HARMONIC DISTORTION % 2ND 0.28 0.44 0.45 0.53 0.74 1.18 3RD 0.16 - - 0.35 0.36 0.96 DAMPING FACTOR = 17 ULTRALINEAR WITHOUT FEEDBACK WATTS 1 2 3 4 5 6 HARMONIC DISTORTION % 2ND 2.71 3.93 5.09 6.65 7.71 7.62 3RD 0.70 0.80 1.10 1.20 1.43 4.05 DAMPING FACTOR = 2.3 TRIODE WITH FEEDBACK WATTS 1 2 3 4 5 HARMONIC DISTORTION % 2ND 0.71 1.04 1.30 1.66 1.98 3RD 0.22 0.25 0.49 0.50 0.79 TRIODE WITHOUT FEEDBACK 2ND 3.95 5.83 7.11 8.93 8.83 3RD - 0.23 0.21 0.20 2.07 Following measured at 100 Hz OUTPUT TRANSFORMER 125E 1628SE A) ULTRALINEAR WITHOUT FEEDBACK HARMONIC DISTORTION % 2ND 15.3 2.9 3RD 3.1 0.4 B) TRIODE WITHOUT FEEDBACK HARMONIC DISTORTION % 2ND 12.2 2.8 3RD 1.5 --- Hope this comes out your end OK. Cheers, John Last edited by John L Stewart : March 16th 11 at 04:21 PM Reason: Problems with tables |
#2
Posted to rec.audio.tubes
|
|||
|
|||
Determination of Maximum Positive Feedback in Bootstrapped Driver
On Mar 17, 12:36*am, John L Stewart John.L.Stewart.
wrote: First of all, calculate the driver gain in grounded cathode mode from published specs- For the 6SN7 family at 250 volts supply * * * * Mu = 20 * * * * Rp = 7.7 K while grid volts is 8 Let A1 be the gain with these conditions, Then A1 = ( mu*Rl ) / ( rp + Rl ) * * * * = ( 20* 27 ) / ( 7.7 + 27 ) * * * * = 15.56 But 27k is the RL only when the 27k is terminated by a B+ which has no ac, or put another way, when there is no signal voltage in the OPT and no Vac bootstrapping tap. Beginners may have failed to understand you. BTW, Ra at 7k7 is a bit optimistic unless Ia is fairly high; Ra varies considerably with Ia. Then find the maximum gain possible for any non-feedback triode stage- Then A2 = mu * * * * = 20 Gain change is A2/A1 * * * * = 20 /15.56 = 1.29 In DB becomes 20 log A2/A1 * * * * = 2.18 db The feed back in the case of the bootstrapped driver is to the triodes anode, so no gain above mu is possible. There is no gain thru the triode being plate driven. OK, this point becomes more obvious the more you think about it, but there's more to consider. One could connect the 27k load to a tap on the OPT where twice the 6SN7 anode signal exists, ie, cross coupled to the opposite side output tube anode, and then you have a case where the 6SN7 would rise above µ, but not much, because the 6SN7 Ra is much less than 27k, and to make a large amount of gain occur one would need to have a feed point for bootstrapping several times the anode signal at the opposite output tue anode. But usually when one feeds more Vac to a bootstrapped RL one has real PFB, not just something which leads the tube towards being loaded by a dummy CCS, ie a load with no current change. When I tried doing this the results were dissapointing, increased Rout and THD NTL, some experimenters have actually built working amplifiers this way using many triodes such as the 6S4 & 12B4. Starting digging, you will find them. Hope this all mkes sense. Cheers to all, John Indeed what you are saying does make sense, but the the effect of bootstrapping where it is a form of mild PFB on THD is yet to be formularized. In RDH4, PFB is mentioned and a schematic is given where a PFB loop between two cascaded small signal voltage triodes is boosted 12dB and the same amount of global NFB is used. The effect of PB when the signal is below say 10Vrms is to increase open loop sensitivty without much increase in THD or reduction of bandwidth so that global NFB may still be applied while retaining stability, at least with resistance loads. RDH4 goes on to say what a marvel PFB can be. Its as if PFB offers "free voltage gain" without using expensive hardware to get it, just resistors are needed. But I have never ever seen any commercial amp with PFB used deliberately to increase OL Gain except of course in samples of the parafeed phase inverter where V1 anode output is divided down to drive V2, to make another phase of drive signal "for free". Anything for free in electronics usually has a price, and THD increase with any form of PFB is one of the costs. Even in Quad-II there is 6dB of PFB so that the two EF86 may generate just enough gain between them to drive the KT66 AND allow about 10dB of global FB. Some other type of input-plus-driver stage with the same tube gain and no PFB would have input sensitivity needing 2.8Vrms input for clipping instead of only 1.4Vrms as Quad-II requires, if the same amount of global FB is maintained. In earier times amp makers wished there amps to be sensitive to 0.2V for clipping; Leak amps with GNFB needed only 0.1Vrms for clipping. Mullard 520 had EF86 in pentode followed by a 12AX7 LTP. But no PFB, and no bootstrapping. Williamson had triode output stage and 6SN7 low µ triode drive stages and the concertina phase inverter with lots of local NFB as a buffer stage between SE input and balanced amp driver. Sensitivity was 2Vrms for clipping. Again, no bootstraping or PFB. As I have mentioned, boostrapping RL of driver anode loads to OPT taps is mainly used to achieve a high voltage swing without the driver going near cut off or grid current. Bootstrapping from an anode winding on an OPT does require a well smoothed B+ rail for the OPT CT. Any hum at the CT also exists ant any tapping points on the anode primary winding and for bootstrapping to work one needs to have any PFB feed of noise to be fed back while remaining common mode and it would be all too easy for some imbalance of noise and the noise becomes differentially applied to OP tube grids resulting in its amplification as any other wanted differential signal. Many old amps like Quad-II had quite high hum at the OPT CT, 17Vrms in the Quad case because there was no beans to pay for a plate choke and additional filter cap, and there must have been a bean counter at Quad's head office. They had a guy there who passionately believed in CMMR and preached its virtues so suppressing any murmers about yet another bit more engineering to make a toy amp heavy. So bootstrapping anything from a conveniently phased OPT tap somewhere can be a problem unless is noise is suppressed by a good PSU. In Quad-II, it would be possible to have say series 82k plus 100k anode supply RL to each EF86 anode. The total of 182k is virtually the same as what is the standard original 180k. Then you could have 22uF electro caps from junction of the two R to the CFB winding which has a CT that is virtually grounded. This would make the load seen by the EF86 to become higher than the 180k, maybe double because pentodes have high Ra not reduced by the internal NFB in a triode, and hence the EF86 gain would nearly double, and GNFB reduced by 1/2 to maintain the same amount of GNFB. But what's the use? I think ppl would find the 0.7Vrms input sensitivity welcome but THD and Rout would not be less. The CFB winding in the Acoustical Connection does not have a common mode hum signal to upset the apple cart. But I doubt much would be really achieved in overall THD reduction or anyother betterments. What works better in Quad-II is to make an LTP with EF80 with 3 times the anode current in EF86 and therefore raise OLG by nearly 2X and construct the input pair as a true LTP with CCS, and with input sig to one side and NFB to the other, and no need to bootstrap or worry about PFB and its THD increasing propensities. 6EJ7 is also a nice driver pentode. Radford used a pair in the LTP for the ST100 PP amp with 2x KT88. 50 ways to leave a lover, slip out the back Jack, find a new plan Stan, and 50 ways to make an amplifier. Patrick Turner. -- John L Stewart |
#3
Posted to rec.audio.tubes
|
|||
|
|||
Determination of Maximum Positive Feedback in Bootstrapped Driver
On Mar 18, 3:05*am, flipper wrote:
On Thu, 17 Mar 2011 05:19:15 -0700 (PDT), Patrick Turner wrote: snip In RDH4, PFB is mentioned and a schematic is given where a PFB loop between two cascaded small signal voltage triodes is boosted 12dB and the same amount of global NFB is used. The effect of PB when the signal is below say 10Vrms is to increase open loop sensitivty without much increase in THD or reduction of bandwidth PFB is PFB and has the attendant increase in distortion and reduced bandwidth regardless of the 'voltage level'. But RDH4 didn't make your simple generalisation or come to the same conclusion about PFB. It is possible to reduce distortion with bootstrapping which is mild PFB. Or PFB may be added to increase gain say by +12dB in a low voltage level input stage/s without increasing THD by +12dB. Then a higher amount of NFB can be applied and the overall THD is much less than without PFB. And, last but not least, Positive Current FB reduces Rout. Sure, PCFB increases THD and reduces bandwith and increases instability, but you don't need much of it to reduce the Rout of some amp from say 2 ohms to say 0.2 ohms, or down to 0.0 ohms, or in fact - 0.5 ohms, or a negative figure which simply means VO increases if a lower load is connected. PCB is DANGEROUS because if RL becomes very low, then the amp tries desperately and vainly to keep VO constant and it can oscillate itself to death easily at some inconvenient frequency, and with some resultant high expense. The reason PCFB and PVFB isn't used is because many dumbcluck amp designers and fuctard production managers will not be able to get a mass made amp to conform to the behaviour of a well researched and tested prototype. R&D and design time costs money, and what CEO can afford to pay anything while he dreams about hs latest Cadilac? so that global NFB may still be applied while retaining stability, at least with resistance loads. RDH4 goes on to say what a marvel PFB can be. Its as if PFB offers "free voltage gain" without using expensive hardware to get it, just resistors are needed. I don't know where the heck you got that idea because RDH4 doesn't say a thing about 'free' or PFB being a 'marvel'. RDH4 was kept brief. It was a 1,600 page book, and brevity needed to be mantained, so much of it comes across dry as desert, and without any distracting notions about "free", or "marvels" . What it does say for the referenced two stage amplifier, and what you left out, is if stage 1's distortion is *much lower* than stage 2 then the increase in stage 1 distortion from applying local PFB is more than offset by GNFB reducing overall distortion because that is dominated by stage 2, which is in the global loop but not the local. You are, in effect, 'redistributing' the distortion from stage 2 to 1. Just in case you are confused about which section of RDH4's many pages about FB to which I may have been referring to, I suggest you turn to page 352 and 353 of RDH4, 4th Ed, 1955. On page 352 there is a 3 stage feedback amp with two 6SN7 and a pair of 6V6 output tubes in beam tetrode mode. The amount of internal loop PVFB is +26dB. The distortion results at the bottom of the page say... "The intermodulation is 40% with no feedback, 8% with negative feedback alone and 1.9% with combined positive and negative feedback under the following conditions--output 8 watts, 4:1 ratio with frequency 60 and 7000 c/s." I rest my case with the jury.... Every 6dB of PFB doubles the distortion but as Cyrano Jones noted when negotiating the price of Tribbles, "twice nothing is still nothing." Now, the first stage isn't quite 'nothing', of course, but if it's very low compared to the second stage then it's, to reference another famous quote, "close enough for government work." Or amplifiers. E.g. if you double a very small number and then halve the big one the net result is lower overall. But it ain't 'free' as stage 1's bandwidth is also reduced, complicating stability. You'd be better off with a high gain wide bandwidth section, like say 2 stages (assuming you don't introduce new problems) instead of the one with PFB, but that costs more and just "how low can you go" before it isn't worth dancing that hand jive? I don't like any PFB anywhere myself. But it is inherent with bootstrapping which is usally a very small amount of PFB, certainly not 26dB. Ditto with paraphase input stages as in Quad-II amps. In amps which John Stewart was referring to which did have bootstrapped driver RLs, the PVFB is small but it does a lot. In Quad-II, the paraphase inverter does indeed double drive amp THD. But overall is mainly due to the OP stage where near clipping its THD might be 10 times the driver stage, so driver THD may be halved or doubled without greatly affecting the overall measurements. By the paraphase connection does the mavellous thing of increasing amplifier sensivity to a sensible figure and thus prevent the Quad-22 preamp having to produce such a high drive voltage which might be higher han the power amps. Everything needs some careful weighing of many factoids before any simple conclusion may be made. I would guess the schematic shown in RDH4, page 353 might be a dog of a thing to get unconditional stability as easily as the text suggests. There's no report there of square testing at low levels with various cap loads...Much is damn well missing! So, what's missing must be found out about and learnt about by real world engineers like myself, not too much dementia yet, because I don't assume and I still ask questions. Patrick Turner. |
#4
Posted to rec.audio.tubes
|
|||
|
|||
Determination of Maximum Positive Feedback in Bootstrapped Driver
On Mar 18, 9:55*pm, flipper wrote:
On Thu, 17 Mar 2011 22:27:28 -0700 (PDT), Patrick Turner wrote: On Mar 18, 3:05*am, flipper wrote: On Thu, 17 Mar 2011 05:19:15 -0700 (PDT), Patrick Turner wrote: snip In RDH4, PFB is mentioned and a schematic is given where a PFB loop between two cascaded small signal voltage triodes is boosted 12dB and the same amount of global NFB is used. The effect of PB when the signal is below say 10Vrms is to increase open loop sensitivty without much increase in THD or reduction of bandwidth PFB is PFB and has the attendant increase in distortion and reduced bandwidth regardless of the 'voltage level'. But RDH4 didn't make your simple generalisation or come to the same conclusion about PFB. Yes they did and it isn't 'my conclusion'. It's a fact you'll find in every electronics textbook on the subject. And I quote RDH4 page 355: "Each 6dB increase in gain due to positive feedback will double the distortion in this stage ; for example 24dB increase in gain will increase the distortion in this stage by 16 times." Page 355 is concerned about different operating condition to those mentioned on Page 353 which I quoted. I remain correct about what I said, and I selected the example omn page 353 because its content is relevant to John Stewart's OP about a power amp including gain stages and power tubes. I cannot find time to answer your further concerns about what I said, maybe your right, maybe your'e wrong. I know people who have reverse phase connected UL taps on OPTs to increase OP tube gain to more than double OP tube gain. The THD was reduced and Rout reduced and bandwidth remained OK and stability fine because they knew how to do it all, unlike most who'd end up with an oscillator. The man who told me about it was Neville Thiele, of the famous and very well educated Thiele and Small gang who worked out all those incomprehensible equations for speaker enclosure design. Theile was 75 yo when I phoned him up one evening in about 1995 for a casual chat about something else which puzzled me. He'd also written some fine articles in Electronics Australia, one being about the PP amp in a deluxe model of Kreisler TV sets which were made in Oz. The tube amp had CFB OPT and a pair of 6BM8, but with no PFB except in the paraphase inverter. Unless careful Nyquist parameters are considered, PVFB is a nightmare. But Neville knew all about Nyquist. For what he did at home for fun, he could get away with using PFB which I would not bother with - too much trouble, and lowering THD with PVFB or PCFB may not improve the sound much. In McIntosh MC275 amps, the 12BH7 has bootstrapped load resistors and there is an amount of PVFB which is substantial, but not huge, but which allows the production of high drive voltages at lower THD than could otherwise be easily achieved. I again rest my case with the jury.... Delete the rest of whatever you said..... Patrick Turner. |
#5
|
|||
|
|||
[quote=Patrick Turner;928239]On Mar 18, 9:55*pm, flipper wrote:[color=blue][i]
On Thu, 17 Mar 2011 22:27:28 -0700 (PDT), Patrick Turner wrote: On Mar 18, 3:05*am, flipper wrote: On Thu, 17 Mar 2011 05:19:15 -0700 (PDT), Patrick Turner wrote: snip [color=green][i][color=darkred][i] Hey Patrick, I won't comment on your most recent except to day that Flipper is correct in his post that it is important to attack distortion in the hilevel stages first. If need be, find gain somewhere else, a low level stage perhaps if there is a shortfall. The short equations I gave predict the worst case of +ve FB since we assume the triode gain to be simply equal to mu while bootstrapped. In reality, while bootstrapped the triode gain is somewhere between the grounded cathode value A1 & mu. So the +ve FB is always less than predicted. In the McIntosh the local NFB in the OP stage is 15-20 db, depending on loading, Etc. So lots of room for a couple of db +ve FB. The rp of the sample circuit given turns out to be about 10K since the cathode current in each triode is 4.3 mA. That translates to a worst case +ve FB of 2.74. The equations can be further reduced as follows- A1 = ( mu*Rl ) / (rp + Rl ) and A2 = mu Gain change is A2 / A1 So Gain change becomes ( mu ) / ( mu*Rl ) / (rp + Rl ) That reduces to Gain Change A2 /A1 = (rp + Rl ) / Rl Very easy to apply to various conditions. I made some real measurements on working systems 8 or 10 years ago but can't find the results. As I recall, all less than 2 db I think. When I get some time, or if there is rain or snow I will make some measurements. I've got a Circlotron, the 6AS7/6080 Amp & two different twin coupled amps I built for publication back in the day. Yesterday was 8K on the cross country skis, then 45K on the bicycle. Not bad for an old fart. Cheers, John |
#6
Posted to rec.audio.tubes
|
|||
|
|||
Determination of Maximum Positive Feedback in Bootstrapped Driver
Delete unecessary argy bargy.
I remain correct about what I said, No and I've proved it by quoting RDH4 itself. OK, I agree to disagree. I leave the outside world to decide who's right. RDH4 is open to interpretation. Delete more stuff that leaves nobody wiser. I just ain't got time to arhue about what seems true for me. Patrick Turner. |
#7
Posted to rec.audio.tubes
|
|||
|
|||
6LU8 SET or SEUL Amp
On Mar 19, 12:56*am, John L Stewart wrote
Hey Patrick, I won't comment on your most recent except to day that Flipper is correct in his post that it is important to attack distortion in the hilevel stages first. If need be, find gain somewhere else, a low level stage perhaps if there is a shortfall. I was merely quoting RDH4 where they measured an amp with NFB and **with** PFB and **without** .. RDH4 had reason to believe the overall effect of adding PFB was a reduction of IMD. The PFB raises open loop gain which allows more global FB to be applied and the NFB reduced the IMD much more than any increase of N&D because of PFB. All PFB increases N&D locally at least, within a circuit, and I have no argument with the text books. But that does not mean anyone can conclude PFB will always increase overall N&D in the presence of NFB. Electronic behaviour can rarely be described by any simple generalisation and in 10 times out of 20 every generalisation is subject to conditions of operation, behaviour A happens if condition B, C, and D are present and if condition E. F are not present, etc. It was the overall effect of combined PFB and NFB and not the local which the real issue, at it was in your OP. I also believe its best to reduce the OP stage N&D locally as much as possible without causing the driver voltage to be come so high that it begins to contribute more N&D to the total N&D outcome than the OP stage. A McIntosh OP stage in plain beam tet mode with no 50% CFB windings and biased for largely class B op with low value RLa-a would have much more N&D than with the 50% CFB. Even with the 50% CFB, the outcome without other additional loop FB is rather poor compared to a class A UL amp or triode amp. To get 50W from 6L6 or 75W from 6550 means means high N&D. But the McI driver has to give over half the Vac between anode and cathode to each grid, so driver N&D is high. The global NFB enclosing all stages reduces the total N&D of all stages about equally to an acceptable figure which in McI's case is better than many other amps of its day. The McI Rout is especially low because of the large total amount of all loops of NFB and despite the inclusion of PFB in the bootstrapping. EAR509 follow a similar idea of having lots of voltage gain which allows lots of FB, and aided by bootstrapping, carefully done to avoid instability to keep Nyquist happy with resistance loads. But EAR failed the Nyquist if a pure C load was used. Paravicini must have thought nodody was ever going to use a 0.47uF speaker load. I rebuilt a pair of EAR509 and was able to make them unconditinally stable while retaining the existing high OLG and all NFB loops, and retaining a 50kHz BW. Sometimes I think master designers likeParavicini have **** for brains. The short equations I gave predict the worst case of +ve FB since we assume the triode gain to be simply equal to mu while bootstrapped. In reality, while bootstrapped the triode gain is somewhere between the grounded cathode value A1 & mu. So the +ve FB is always less than predicted. In the McIntosh the local NFB in the OP stage is 15-20 db, depending on loading, Etc. So lots of room for a couple of db +ve FB. The rp of the sample circuit given turns out to be about 10K since the cathode current in each triode is 4.3 mA. That translates to a worst case +ve FB of 2.74. The equations can be further reduced as follows- A1 = ( mu*Rl ) / (rp + Rl ) and A2 = mu Gain change is A2 / A1 So Gain change becomes ( mu ) / ( mu*Rl ) / (rp + Rl ) That reduces to Gain Change A2 /A1 = (rp + Rl ) / Rl Very easy to apply to various conditions. I made some real measurements on working systems 8 or 10 years ago but can't find the results. As I recall, all less than 2 db I think. When I get some time, or if there is rain or snow I will make some measurements. I've got a Circlotron, the 6AS7/6080 Amp & two different twin coupled amps I built for publication back in the day. Yesterday was 8K on the cross country skis, then 45K on the bicycle. Not bad for an old fart. Don't overdo it. It can easily dull the brain. I wish Flipper would go for a long bike ride sometime but he's excused if he fears USA motorists. Maybe along swim would suit him, he has the right name for it. I have to work. I'm not a retired old fart. But I can fart well when I have to though. I'm presently doing a second attemt to make a pair of 100W Ming Da amps behave. The Chinese were way out of their brainular cababilities when they designed these POS amps. They are great smoke producers, and an owner's nightmare with regard to biasing.. Anyway, I have retained the tube line up, 6SN7 SET input stage to make about 5Vrms, 6SN7 LTP phase inverter to make two 30Vrms phases, then 300B balanced amp each with Ia 7mA and each making 120Vrms to power grids of a pair of 845 with Ea = 1,200Vdc and Ia 43mAdc each. Initially N&D was 5 times worse than how I have it now, stability was poor, and the list of Chinese errors from their technical incompetence would fill several screen fulls of description. There is no need for bootstrapping because I have revised the +420V rail and created a -190Vdc rail for the 300B balanced driver amp. This 300B driver stage is laughable use of 300B, but the Chinese think westerner buyers like lots of spectacular tubes. In fact the 300B with Ia only 7mA works fine. But Ia was less when they were expected to work from only a 400V rail originally. OK, the amps go back to the buyer who bought them second hand from someone who just wanted to get rid of them. He's very happy for a month before the attacks of N&D start again, and he returns the amps to me. This time I find an anode resistor of a drive tube had gone open and I have to suspect it has the Chinese Carbon Film Resistor Disease, CCFRD where the resistor sure looks nice, and like something ARC Corp might use, (except the paint job is worse), and the resistor has gone open without any sign of heat stress. Well, There are about 20 of these potentially horrible Chinese made resistors in each monobloc and during the first re-engineering attempt I assumed they'd be OK, but they are NOT OK. I have seen several examples where Chinese 1W carbon film resistors just go open. Usually its because there is more than 200Vdc across a carbon film type. Something strange happens in the resistor. The MingDa resistors look like 2.5W rated and I thought they would be fine. WRONG!, not OK. II am replacing all 20 R throughout each monobloc amp with metal film types with heavy watt ratings. Where they have 4 x 470uF caps in series for +1,200Vdc rails, there is 300Vdc across 400V rated cap bypassed with 220k and this looks very prone to failure. I can't bootstrap the 50k anode loads of each 300B tube from the OPT because there are no available taps on OPT which are suitable. I could have taken anode R loads to the +1,200V rail after making yet another series C RC filter, but the split rails of over 700Vdc is enough to get the swing needed for drivng 845, and I avoid HV bothers. But all hell breaks loose if this amp is driven well into clipping; the 300B and 845 become badly affected by temperature changes in grids and the biasing goes very wobbly and strange, and the best cure is to switch off the amps, wait a minute for things to cool before switching back on. All of which doesn't happen/isn't needed in normal OP tubes like 6550. To force the amp to behave when over driven I will try to make a limiting circuit where a high resistance plus low resistance diivider is cap coupled of V1 output and if V1 voltage rises above a threshold due to clipping, two transistor, npn and pnp will be turns on and there collectors will shunt the input signal preventing any V1 output increase. Such a scheme is fine up to clipping because the transistors have extremely high collector resistance when fully turned off. And if someone turns up the volume with a shorted speaker it prevents the OP tubes being turned on fully into a shortcircuit. I rarely cannot proceed from the start of a day to its finish without addressing the incompetence of others I am forced to deal with in my day to day work life. But the Turner-Ming amps do sound very well indeed when they work. There is no doubt the shuguang Chinese made 300B and type B 845 are superb. But I cross my fingers about the Chinese OPTs when I think about them. I don't expect them to last a long time. The leads coming out of the OPT box have insulation far too thin, nowhere as good as what you'd find on any Hammond OPT and the the Chinese saw fit to use green wire with yellow stripes for the CT lead to B+. This color is normally used for chassis earth wiring. I hate to think of the Feng Shui karma issues surrounding chinese OPT designs. Patrick Turner. |
#8
Posted to rec.audio.tubes
|
|||
|
|||
Determination of Maximum Positive Feedback in Bootstrapped Driver
Hope this all mkes sense. Cheers to all, John Maybe I'm not thinking straight but I don't see that answering the question because it depends on what appears to be an arbitrary claim that gain cannot exceed mu. I mean, it begs the question by declaring a limit, which is the essence of the question. Its very easy to think bently instead of straightly. Are you presuming that the PFB is designed to produce a unity voltage feedback (relative to the anode signal swing) so that Rl 'appears' infinite? The RL which is infinite is a constant current source or constant current sink. But the bootstrapping can be arranged to make RL effectively increase in ohms value positively until it indeed becomes infinite and no current change occurs in the tube. But from there the load can become negative. On a load line graph with load sloping down as you move from right to left, the bootstrapping will tip the load line to flat until it is a CCS which is a horizontal line. Then the load can be made to be a negative load value which slopes up as you more from right to left across the Ra curves. Try plotting the lines and current changes while considering different vallues of bootstrap taps on the OPT. As the amount of bootstrapping volage increases, the gain rises above µ, but it takes some doing, and my experiments showed trying to boost gain above µ was completely pointless because of such rapdidly increasing N&D and instabilities. THes s. Usually, applying PFB through an anode is the most useless way to apply positive FB to boost gain in order to increase open loop gain of an amp without much increasing open loop N&D, because the latter cannot be achieved. But McIntosh found bootstrapping to be a real boon for achieving low N&D and low Rout, and methinks whatever McIntosh did agrees with RDH4 conclusion that PFB can be used for an overall outcome which gives less IMD with PFB than without PFB. Bootstrapping is a *small amount* of PFB, and if the OP stage is already fairly linear as is the case with a class A triode or UL or CFB amp then the amount of THD fed back to increase itself is not huge. One has to analyse just what happens to an existing measurable distortion voltage at the OPT tap as it is fed back. Forget about other signal voltages - just think about the distortion voltages and what MUST happen with them. If that's the case then it's trivial to show from the gain equation that gain approaches, and does not exceed, mu but it's not intuitively obvious from the schematic that's the amount of feedback being applied. If the amount of voltage feedback exceeds plate swing then gain can, indeed, be above mu. Indeed. PFB is used all over the joint in complex filter circuits. I have used it myself to boost the Q of a narrow band bandpass filter. Some times the N&D just does not matter. The wien bridge oscillator relies on much PFB and slightly less NFB so that it will continue oscillating, and in a recent revision of my 1kHz oscillator I was able to get 0.0044% with just one opamp with just the right lamp. To illustrate I spiced a simple circuit with a 250V source having 100Vrms on it (simulating an arbitrary transformer feedback signal) applied to a 6SN7 plate through a 27k and cathode to ground through an unbypassed 470 ohm. Opposite phase 1Vrms (for visual 'instant math' convenience) was then applied to the grid. Plate swing was 52Vrms for a gain of 52.- Hide quoted text - - Show quoted text - Google prevents me seeing any further... But try to build real cicrcuits with PFB.... Patrick Turner. |
#9
Posted to rec.audio.tubes
|
|||
|
|||
Determination of Maximum Positive Feedback in Bootstrapped Driver
On Mar 19, 8:39*pm, flipper wrote:
On Fri, 18 Mar 2011 16:10:52 -0700 (PDT), Patrick Turner wrote: Delete unecessary argy bargy. I remain correct about what I said, No and I've proved it by quoting RDH4 itself. OK, I agree to disagree. I leave the outside world to decide who's right. RDH4 is open to interpretation. Delete more stuff that leaves nobody wiser. I just ain't got time to arhue about what seems true for me. Sorry Pat but electronics is not a Van Gogh to be freely 'interpreted' by the observer. I didn't mention Van Gogh. Only you did. The confusion is coming from using PFB to enable 'something else' that is beneficial, such as enabling GNFB in the example you cited. PFB enables it but GNFB is what 'does the good thing'. Youv'e got it. That's exactly what I was trying to tell you. And RDH4 gave the figures to back up their claim about the schematic they show. Another example is John's boostrap, but it's 'something else' is something else. John's circuit has PFB with NFB. There's just enough PFB to effectively increase the RL value seen by the 6SN7 so its THD should reduce maybe 10dB. I say that because with a plain 27k you'd have 1% THD at 10Vrms, but with CCS you'd have 1% THD at 50Vrms, depending on the sample. But the bootstrapping feeds back THD created in the OP tubes plus 6SN7 back through the R divider formed by Ra and 27k so that a fraction of D appears at the OP grid to be re-amplified and add to D already there. some second order products are formed on this merrygoround. But the THD measured with bootstrapping is the results after the merrygoround; the artifacts don't keep on increasing infinitely. If you look at my reply to John you'll see I simulated a 6SN7 plate PVFB circuit where the gain was greater than mu so I went back and measured the distortion. Gain with no PVFB was 11.97 with 2'nd harmonic at -54.6dB Gain with PVFB was 52.2 with 2'nd harmonic at -42.9dB PFB increases distortion. Sure, and you have applied far too much PFB. But you need to measure an amp built like John's or a Mcintosh, and vary the amount of PFB from between no bootstrapping to where gain is say double µ. Do this at a low level of output and plot the THD for same output voltage while adjusting input voltage to keep Vo constant. Maybe the experiment might change your mind. Minds were changed in the McIntosh lab well before 1950...... However, we can use the PFB to enable 'something else'; namely to make Rl 'appear' as if it's infinite (bootstrap). If I do that, by selecting the feedback ratio to match the expected plate swing, which we can estimate by mu since an infinite load should make gain equal to mu, then Gain with selected PFB is indeed 20 and 2'nd harmonic becomes a glorious -131dB. (an unrealistic number indicating the limits of the spice simulation) Huh, you jest, surely? It's the special case (the 'something else') of Rl appearing to be (near) *infinite* that lowers the distortion, not the PFB (which was illustrated by the first example). The increase in RL via bootstrapping value lowers 6SN7 THD but creates a PFB path to make OP stage THD greater. Its about that simple. By "not the PFB" I mean one could, in theory at least, put a 'real' infinite resistor, with an infinite B+ to supply it, there and get even better distortion numbers, because PFB increases distortion over what the equivalent circuit has (and note than an infinite load is not the same circuit as one with a 27k load), but that's not terribly practical while PFB is. I've done all this and have avoided bootstrapping wherever I can because I wanted the full 5 steps forward benefit of driver linearity without the 2 steps backward which PFB brings. But bootstrapping can give you 3 steps forward. One can easily arrange each side of a balanced amp to be a pair of triodes in a µ follower config to make the DC carrying RL far larger than 27k. Or one may use a solid state CCS. The cap coupled output load is the biasing resistor, and its value becomes the main load for the gain triode. I found the use of a CT choke plus series RL at each end gave best results. http://www.turneraudio.com.au/300w-1...tput-jan06.htm But people hate chokes, so they bootstrap instead !!! We could also achieve the same result with a constant current source on the plate (which also simulates a near 'infinite R') so it's clearly not PFB that 'does the good thing'. It's the artificially created (near) 'infinite load' that lowers distortion. Indeed. PFB enables use to do (in special cases) *'something else' (apply GNFB, simulate an 'infinite R', etc) that would be more difficult to achieve by other means. In real life it won't be that good because my sim uses a 'pure' sine for PFB and not a distorted OPT feedback. Using any useful amount of PFB in mass market amps has been avoided like the plague by all manufacturers. Its just too difficult to get right. Its SO EASY to just arrange a suitable amount of OLG and apply only NFB. Patrick Turner. Patrick Turner.- Hide quoted text - - Show quoted text -- Hide quoted text - - Show quoted text - |
#10
|
|||
|
|||
Quote:
Hi Flipper- Sounds intriging & I am curious as usual. Could you post a JPG of your simulation so I can try it here? Far as I know the only way to get gain more tham mu is thru a transformer as the ancients did it before pentodes (or tetrodes) or some kind of PFB applied to the grid or cathode. And the kind you don't want by way of parasitics. I did some quick tests on the first of the Twin Coupled Amps I had built a few years back. The final in this thing runs 6LU8s in PP, the triode sections cathode coupled to the OP grids. At clipping it is good for 37 watts. IMD is down more than 50 db at one db below max OP. It was nice that the caps didn't blow since it has not been run in a few years. Tests are done using a HV Diff Probe into a PICO Scope ADC216. Does scope, SA & RMS volts with 16 bit resolution. With the 6SN7 27K plate resistors returned to the 400 volt regulated supply rather than thru the PFB circuit the gain was 11.4. Then reconnected to the PFB circuit the SN7 gain was 16. If you do the math, turns out to be 3 db, somewhat higher than I recall. I may have done that kind of test on one of the other bootstrapped amps I mentioned in a previous post. Oddly I found I had used a tail resistor of 33K to the -150 volt supply on this 6SN7. Usually I've put in more like 18K. That would account for some of the differance since the SN7 is working at a higher plate resistance here. If you can't get that JPG on here then email to me direct at Cheers, John |
#11
|
|||
|
|||
[But I cross my fingers about the Chinese OPTs when I think about them.
I don't expect them to last a long time. The leads coming out of the OPT box have insulation far too thin, nowhere as good as what you'd find on any Hammond OPT and the the Chinese saw fit to use green wire with yellow stripes for the CT lead to B+. This color is normally used for chassis earth wiring. I hate to think of the Feng Shui karma issues surrounding chinese OPT designs. Patrick Turner.[/quote] Ya, I've had my share of crap right out of the box too. One I remember around 1960 in particular at U of T Physics was a TEK 540 Series scope I used a lot doing research on core memory, before SS memory looked like a sure thing. The one volt p-p calibrator looked OK on the top half of the screen but intermittely showed less than a volt on the bottom half. This fault would come & go at random. Finally I traced it down to one of the tubes in the vertical amplifier. Heater connexions on 9-pin tubes for 6.3 volts are usually 4 & 5. Pin 4 was very well soldered in with the TEK silver doped solder but the wire just stuck thru Pin 5. So the fault came & went in a random way. Last summer thought I had a alternator failure on my Kubota tractor. The charge indicator was on all the time, but oddly only half lit. I put a scope on the 12 volts & saw what seemed to keep time with the engine RPM. Figured a busted stator, a 3-phase thing with a full bridge of rectifiers so $450 dollars later I got another on. No change. Turns out after a lot of tracing I found the wire from the on-off switch (diesels don't have electrical ignition) to the voltage regulator broke off inside the insulation. I guess the guy stripping the wire at the factory squeezed to hard on his cutters as he stripped the wire. Not sure what will fail next. Got a new charger on order for my 2 year old Dell laptop on order now. Cheers, John |
#12
Posted to rec.audio.tubes
|
|||
|
|||
6LU8 SET or SEUL Amp
Flipper mentioned here ....
Of course, one of the problems with bootstraps in general is they fall apart faster than you can say "where did my voltage go?" at the frequency extremes because the bootstrap condition fails Indeed you make an excellent point. However, in the case of McIntosh amps where they have 50% CFB in the OP stage the McI OPTs have exceptionally wide bandwidth with low phase shift. If one didn't have a wide BW OPT then the driver gain will sag a bit too close to the AF band. But nevertheless gain shelving at both ends of the AF band in many amps including most I build is a very important technique to ensure unconditional stability will exist when the the wanted response is -3dB at 10Hz to 65kHz. If the amp OLGain is -3dB at say 20Hz and 8kHz, and this is not unusual in many pentode or tetrode amps with pentode input stages then the NFB will flatten everything out so you get 10Hz to 65kHz. The amount of NFB at 10Hz or 65kHz is very much reduced below what it is at say 400Hz where NFB may be at a maximum. The mu-follower two triode gain stage is my favourite gain block with good gain, wide BW, and high Z in and low Z out. It sounds and measures excellently. If one uses a 6CG7 with 10k0 between bottom anode and top cathode and OL gain of top triode is say 16, then that bootstrapped 10k is made to look like a load of 170k to the bottom triode. Because the top triode is functioning nearly as a pure cathode follower its BW is very high and its cathode is an ideal place from which to connect the bootstrapped 10k anode load of the bottom triode. Hence it is easily possible to get 0.1% THD at 10Vrms, as opposed to having no bootstrapping and a much lower R load for the bottom triode thus increasing THD to perhaps 0.4% at 10Vrms. The top triode sees a cathode load or 160k, and its a win-win outcome for all triodes concerned. You may well state what you think are "general" tendencies but then I find "exceptions to the rule". The gain block which is technically better measuring is the normal common cathode gain triode with CCS anode load and directly coupled to a CF with CCS cathode load. But how much better measuring is this? probably 3dB at most. I found a 12AU7 will sound magnificent with the dual CCS but then so does the Mu- follower - hard to say which sounds best really. The other benefit of the mu-follower which was called a bootstrapped follower is that 5mA of anode current for 2 triodes in series is all you need. With CCS loading for both triodes, you need 10mA. Patrick Turner. |
#13
Posted to rec.audio.tubes
|
|||
|
|||
6LU8 SET or SEUL Amp
On 03/22/11 04:40, Patrick Turner so witilly quipped:
If the amp OLGain is -3dB at say 20Hz and 8kHz, and this is not unusual in many pentode or tetrode amps with pentode input stages then the NFB will flatten everything out so you get 10Hz to 65kHz. worthy of note: NFB also significantly improves THD, _especially_ if you're using pentodes as pre-amps, since pentodes have non-linear characteristics. /me prefers triodes for pre-amps anyway - 12AX7 or 12AT7 preferred and the triode pair within a single tube can give you combined circuit gains as good as (or even better than) a tweeked up pentode pre-amp, with better sound quality (theoretical mu gain of two 12AX7 triodes in series would be around 10,000). Two 12AX7's (4 triodes) driving a beam pentode P-P Class AB with plenty of NFB will typically give you outstanding THD and IM behavior and a nice flat response, even if you use semi-crappy output transformers, but the clipping power level at the edges of frequency response might become a LOT lower [so you'll either need adequate amplifier headroom or better OPTs]. Maybe that's why the circuit config I just described is used so often. That being said, using single-ended 'triode config' power output tubes would likely improve the THD and IM that much more, maybe requiring less NFB to achieve quality results (at the expense of output power and efficiency). |