Two steps forward, one step back
I dragged out my Chinese "Music Angel" EL34 and started modding for niece Kate.
Original Schematic is on this thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=59070&highlight=
Did:
Diodes to Ultrafast Soft Recovery
Final Filter caps to 470uF/250V Epcos (giving 235uF effective).
Separate Bias adjust for each output tube etc.
Then ditched global feedback and modified to "Baby Huey" - ish circuit.
I'm now at the absolute limit of the local feedback I can apply using this method and Zout is stiill too high - 4.5 Ohms from the 4 Ohm tap. (Anymore local feedback and I start to overload the diff amp).
Wondering where to go from here:
I WILL have to add some global feedback - note I swapped the diff amp inputs over so that I can use global feedback to the diff amp input rather than go back to the in put stage.
Should I do something about the common screen resistor to the output tubes before trying anything else?
Current schematic - appologies for pdf, I couldn't seem to scale the jpg version to be both readable abd meet file and image sizes.:
Cheers,
Ian
Circuit is now:
I dragged out my Chinese "Music Angel" EL34 and started modding for niece Kate.
Original Schematic is on this thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=59070&highlight=
Did:
Diodes to Ultrafast Soft Recovery
Final Filter caps to 470uF/250V Epcos (giving 235uF effective).
Separate Bias adjust for each output tube etc.
Then ditched global feedback and modified to "Baby Huey" - ish circuit.
I'm now at the absolute limit of the local feedback I can apply using this method and Zout is stiill too high - 4.5 Ohms from the 4 Ohm tap. (Anymore local feedback and I start to overload the diff amp).
Wondering where to go from here:
I WILL have to add some global feedback - note I swapped the diff amp inputs over so that I can use global feedback to the diff amp input rather than go back to the in put stage.
Should I do something about the common screen resistor to the output tubes before trying anything else?
Current schematic - appologies for pdf, I couldn't seem to scale the jpg version to be both readable abd meet file and image sizes.:
Cheers,
Ian
Circuit is now:
Gingertube,
I am not sure why there is a screen resistor - you do not have a too high Vg2.
But regarding your main problem, you will have to ease up on the local feedback otherwise you will overload the input stage. This is the problem with a triode there - not enough swing and before that 2nd harmonic generation. An input pentode will be able to provide more, also a higher h.t. If stability allows, decrease the decoupling 36K. (Not to start other debates, but I cannot see the often encountered resistance to input pentodes. Yes, there is more noise - on paper. Audible? In my experience not unless you stick your head into the loudspeaker.)
I also perceive some resistance to global feedback; again I hope old wives' tales did not find their way in through your kitchen door. Circuits like this have been used with up to 28 dB of stable global feedback (well, sans mosfets, but I cannot see a problem there as they are followers). But you will not get a high damping factor with safe feedback from a pentode output stage (but then you also would have picked up that anything beyond 8 is not really noticable, other things being equal).
Regards.
I am not sure why there is a screen resistor - you do not have a too high Vg2.
But regarding your main problem, you will have to ease up on the local feedback otherwise you will overload the input stage. This is the problem with a triode there - not enough swing and before that 2nd harmonic generation. An input pentode will be able to provide more, also a higher h.t. If stability allows, decrease the decoupling 36K. (Not to start other debates, but I cannot see the often encountered resistance to input pentodes. Yes, there is more noise - on paper. Audible? In my experience not unless you stick your head into the loudspeaker.)
I also perceive some resistance to global feedback; again I hope old wives' tales did not find their way in through your kitchen door. Circuits like this have been used with up to 28 dB of stable global feedback (well, sans mosfets, but I cannot see a problem there as they are followers). But you will not get a high damping factor with safe feedback from a pentode output stage (but then you also would have picked up that anything beyond 8 is not really noticable, other things being equal).
Regards.
Johan, I'm sympathetic to this argument and I do use feedback on many of my amps. Most non-ignorant objections to feedback center on what happens at overload, not what happens when all is right. You can indeed make 28dB of feedback stable as long as the amp isn't clipping, but it's well-known that the feedback greatly complicates the recovery issue.
Now, here's another place where pentodes can be an advantage- antiblocking measures can be much simpler than with triodes.
Now, here's another place where pentodes can be an advantage- antiblocking measures can be much simpler than with triodes.
Guys - a bit of a rave but here goes!!!
Yes, I do dislike global feedback - my (limited) experience is that even small amounts compromise stereo imaging. In any case, my control theory says that in any feedback system, you make it as linear as possible before applying feedback, so the exersize thus far has been worthwhile.
All that local feedback has certainly shifted the output tube/output transformer high frequency pole higher and the low frequency zero lower, which should allow more extended bandwidth with global feedback applied.
I've seen a lot of posts saying that pentode mode requires good screen supply regulation - that was why I ws wondering about deleteing the common supply to the output grids (which is at approx 400V at idle) and perhaps running separate screen resistors back to the zener regulated 360V supply or a new supply, with an extra RC off the main +440V.
SY,
I also noted you saying on one design posted recently, that the global feedback resistor from the output tranny needs to be kept relatively low in value - is this to prevent further phase shift in conjunction with the miller capacitance of the input it feeds? In this case I can expect that miller capacitance to be quite high (say 100pF at a guess) because of the high gain in the 6SL7. This tells me that the global feedback resistor would not want to be any larger than about 10KOhms (for 160kHz pole).
The main loop must have a dominant pole somewhere for stability. This is likely to be in the output tube/output transformer interaction at present. Should I try to shift this pole to say the the output tube inputs by pushing up the grid stopper resistors. This would seem to be contrary to my feedback theory understanding, which says to keep the 2 most dominant poles as far apart as you can manage - any insights?
Cheers,
Ian
Yes, I do dislike global feedback - my (limited) experience is that even small amounts compromise stereo imaging. In any case, my control theory says that in any feedback system, you make it as linear as possible before applying feedback, so the exersize thus far has been worthwhile.
All that local feedback has certainly shifted the output tube/output transformer high frequency pole higher and the low frequency zero lower, which should allow more extended bandwidth with global feedback applied.
I've seen a lot of posts saying that pentode mode requires good screen supply regulation - that was why I ws wondering about deleteing the common supply to the output grids (which is at approx 400V at idle) and perhaps running separate screen resistors back to the zener regulated 360V supply or a new supply, with an extra RC off the main +440V.
SY,
I also noted you saying on one design posted recently, that the global feedback resistor from the output tranny needs to be kept relatively low in value - is this to prevent further phase shift in conjunction with the miller capacitance of the input it feeds? In this case I can expect that miller capacitance to be quite high (say 100pF at a guess) because of the high gain in the 6SL7. This tells me that the global feedback resistor would not want to be any larger than about 10KOhms (for 160kHz pole).
The main loop must have a dominant pole somewhere for stability. This is likely to be in the output tube/output transformer interaction at present. Should I try to shift this pole to say the the output tube inputs by pushing up the grid stopper resistors. This would seem to be contrary to my feedback theory understanding, which says to keep the 2 most dominant poles as far apart as you can manage - any insights?
Cheers,
Ian
I'm uncomfortable using the OPT to set the dominant pole. Usually, I'll set one a decade lower in one of the early (class A) stages. For example, in my little EL84 amp, the driver is a grounded cathode voltage amp direct coupled to a split load. The dominant pole is set in the first stage by a series RC connected across the plate resistor. My own prejudice is to avoid, where possible, a compensation cap across the feedback resistor; rather, I'm willing to sacrifice a little bandwidth at the high end to just use the dominant pole to stabilize things.
An overly large feedback resistor is a common error and makes the amp much harder to stabilize. Less of a problem with the kind of driver I used (feedback return to the cathode), more of a problem with diff amps (feedback returned to the grid), and even more of a problem with diff amps using high mu tubes.
Since the OPT is driving an 8 or 16 ohm load anyway, might as well drop that feedback resistor to 1k just to completely take the unintended zero oout another decade.
An overly large feedback resistor is a common error and makes the amp much harder to stabilize. Less of a problem with the kind of driver I used (feedback return to the cathode), more of a problem with diff amps (feedback returned to the grid), and even more of a problem with diff amps using high mu tubes.
Since the OPT is driving an 8 or 16 ohm load anyway, might as well drop that feedback resistor to 1k just to completely take the unintended zero oout another decade.
That's clever, I never thought of that. I've seen schematics with quite high global NFB resistors, usually because they take the NFB back to the whole of the (unbypassed) cathode bias resistor.
With global NFB to the IP stage's cathode, I usually use a resistor of 47 ohms to 100 ohms in the cathode circuit and NFB resistor from the OPT secondary in the region of 2.2k to 6.8k. I guess this is small enough not to cause a problem?
With global NFB to the grid of a LTP input stage, I have a 220 ohm resistor from the grid to ground and NFB resistor of around 10k to 22k. I hope this is small enough to avoid problems too.
With global NFB to the IP stage's cathode, I usually use a resistor of 47 ohms to 100 ohms in the cathode circuit and NFB resistor from the OPT secondary in the region of 2.2k to 6.8k. I guess this is small enough not to cause a problem?
With global NFB to the grid of a LTP input stage, I have a 220 ohm resistor from the grid to ground and NFB resistor of around 10k to 22k. I hope this is small enough to avoid problems too.
Yep - that makes sense.
I did too big a mod to the MAEL34.
The "Baby Huey" ish shunt feedback was worth doing BUT I should NOT have changed the diff amp to ground referenced grids, current source to -ve bias supply and AC coupling the 1st stage to the diff amp. All that could have been left as was and just put current source returned to 0V in the diff amp cathodes (there is about 100 to 110 Volts to work with from the DC coupling from the input stage). Then the existing global feedback connection could have been used - the best laid plans of mice and men!!!
Weekend is tomorrow - I can see I'm going to be busy.
Cheers,
Ian
I did too big a mod to the MAEL34.
The "Baby Huey" ish shunt feedback was worth doing BUT I should NOT have changed the diff amp to ground referenced grids, current source to -ve bias supply and AC coupling the 1st stage to the diff amp. All that could have been left as was and just put current source returned to 0V in the diff amp cathodes (there is about 100 to 110 Volts to work with from the DC coupling from the input stage). Then the existing global feedback connection could have been used - the best laid plans of mice and men!!!
Weekend is tomorrow - I can see I'm going to be busy.
Cheers,
Ian
One type of local NFB that appeals to me is from the OP tube plates to driver cathodes. This arrangement includes both driver and OP tube in the loop and is useful, for example, if you can't use cross-coupled NFB back to the splitter plates (e.g. in a Williamson-type circuit). It means that you can't use a differential driver; the driver has to be PP with unbypassed cathode resistors but that's not a big disadvantage IMHO.
I don't like Crowhurst's capacitor-coupled NFB loop, though, preferring instead to use direct coupling. By adding a resistor ~ 330k from each driver cathode to its grid, in addition to the usual 470k-1Meg resistor from grid to ground, you can set up the bias nicely for the drivers. Direct coupled NFB arranged like this does have the disadvantage of incurring a large voltage across the NFB reistors (400 volts or more), with quite high heat dissipation. It's best, probably, to use two or more resistors in series for the NFB resistors, to distribute the dissipation and to stay within the maximum voltage rating of the resistors.
I don't like Crowhurst's capacitor-coupled NFB loop, though, preferring instead to use direct coupling. By adding a resistor ~ 330k from each driver cathode to its grid, in addition to the usual 470k-1Meg resistor from grid to ground, you can set up the bias nicely for the drivers. Direct coupled NFB arranged like this does have the disadvantage of incurring a large voltage across the NFB reistors (400 volts or more), with quite high heat dissipation. It's best, probably, to use two or more resistors in series for the NFB resistors, to distribute the dissipation and to stay within the maximum voltage rating of the resistors.
Ray, have you seen the schematics of the Berning amps? Particularly the EA230 and EA2100/2101? It sort of neatly ties together your musings about local feedback (OP plate to driver cathode) and the one on sweep tubes...
The schematics are on the Berning site, www.davidberning.com , in the "support" section.
The schematics are on the Berning site, www.davidberning.com , in the "support" section.
Anyone tried boosted triode mode? http://www.edn.com/article/CA302240.html
The perfect marriage of pentode and triode perhaps?
The perfect marriage of pentode and triode perhaps?
SY
About overload (post #23), point taken and in full agreement. But I never had too much problems here in the tube area - semiconductors YES, and that is why many "good" designs fail the listening test - but that is another topic.
I found passive NFB mixing (feeding both input and NFB signals to input through a resistor network e.g. as in an inverting op-amp) less prone to cutting things off under overload conditions. I use this in my better amps, even at the cost of a slightly lower noise figure.
I have difficulty in choosing the no-feedback option for this reason only; kind of having some distortion some of the time vs. having some all the time?
Another important point: The dominant pole; again I support you. Sorry if I repeat, but this is my main reason for not using an ECC83-type input stage. It can force a C on me (without the option to include a serie R) barely outside the audio range. Some output transformers also have low leakage reactance at the expense of internal C (too many secondary sections). I have measured some (repeating again!) with -3 dB point as result of leakage alone at 200 KHz, but with real response hardly over 30 KHz as result of C. Maybe I am too much of a chicken, but I try to keep basic open loop response as wide as possible so that I can have one R.C somewhere under my control.
I will state carefully that I do find a C over the feedback resistor (sometimes an R.C) as phase-lead compensation useful, but then with proper checking - not just a value slapped on.
Then low feedback resistors, yes. It is convenient to have some limiting resistor over the output just to keep tube impedance under some control. (I like to have an amplifier stable under no-load conditions.) I forfeit some 2% of output power by using 470 - 680 ohm across 8 ohm output terminals, and make this my feedback resistor at the same time. If this necessitates an unconveniently low ground link (one must watch out for wiring inductance), I use a double step.
About overload (post #23), point taken and in full agreement. But I never had too much problems here in the tube area - semiconductors YES, and that is why many "good" designs fail the listening test - but that is another topic.
I found passive NFB mixing (feeding both input and NFB signals to input through a resistor network e.g. as in an inverting op-amp) less prone to cutting things off under overload conditions. I use this in my better amps, even at the cost of a slightly lower noise figure.
I have difficulty in choosing the no-feedback option for this reason only; kind of having some distortion some of the time vs. having some all the time?
Another important point: The dominant pole; again I support you. Sorry if I repeat, but this is my main reason for not using an ECC83-type input stage. It can force a C on me (without the option to include a serie R) barely outside the audio range. Some output transformers also have low leakage reactance at the expense of internal C (too many secondary sections). I have measured some (repeating again!) with -3 dB point as result of leakage alone at 200 KHz, but with real response hardly over 30 KHz as result of C. Maybe I am too much of a chicken, but I try to keep basic open loop response as wide as possible so that I can have one R.C somewhere under my control.
I will state carefully that I do find a C over the feedback resistor (sometimes an R.C) as phase-lead compensation useful, but then with proper checking - not just a value slapped on.
Then low feedback resistors, yes. It is convenient to have some limiting resistor over the output just to keep tube impedance under some control. (I like to have an amplifier stable under no-load conditions.) I forfeit some 2% of output power by using 470 - 680 ohm across 8 ohm output terminals, and make this my feedback resistor at the same time. If this necessitates an unconveniently low ground link (one must watch out for wiring inductance), I use a double step.
Oooooh, 2% power loss! Fatal!!!
I'm with you 100% on the desire to make the amp stable with no load. The lowish feedback resistor, of course, obviates the question.
Where you really see the virtues of the feedback approach that you and I agree on is when you make the load capacitive. I can stick any cap up to 4uF across my amp's output with no ill effect other than a small increase in overshoot. Perhaps higher, but 4 is as big as I test with.
I'm with you 100% on the desire to make the amp stable with no load. The lowish feedback resistor, of course, obviates the question.
Where you really see the virtues of the feedback approach that you and I agree on is when you make the load capacitive. I can stick any cap up to 4uF across my amp's output with no ill effect other than a small increase in overshoot. Perhaps higher, but 4 is as big as I test with.
No, SY, I haven't seen any Berning schematics. I only know, from Spice modelling, that it seems possible to use quite a lot of two-stage NFB, with OP tube plate to driver cathode feedback, so that gloal NFB can be reduced and still attain the desired DF.
The other thing I'm playing with is a step network between the driver and OP stage, instead of simple capacitor coupling. That arrangement, in conjunction with the OP tube plate to driver cathode NFB, should hopefully minimise phase shift, so as to provide a stable basis for adding global NFB.
Incidentally, SY, does your dislike of bypassing the global NFB resistor with a capacitor have anything to do with avoiding the risk of injecting RF, picked up by your speaker leads, into the first stage? Norman Koren states, in his articles about NFB, that this is a real danger, yet almost all global NFB designs use it.
Aha, here's one.
http://web.mit.edu/cheever/www/crow4.htm
And a tip o' the hat to Dan Cheever, whom I've certainly disagreed with strongly in the past. But anyone who puts up Crowhurst articles is OK in my book.
http://web.mit.edu/cheever/www/crow4.htm
And a tip o' the hat to Dan Cheever, whom I've certainly disagreed with strongly in the past. But anyone who puts up Crowhurst articles is OK in my book.
Yes, I've seen what Crowhurst said: basically, that using the NFB resistor bypass cap could be masking problems inside the loop. My own preference for dealing with HF instability is like yours, namely, to deal with it in an realy stage. However, for a little fine tuning to tame ringing on square waves, a small bypass cap across the NFB resistor can still be useful.
Well, I think that's the key- early and hard. Get the bandwidth of everything at least a decade above the OPT's limit, then put a dominant pole in early. If the open loop gain is a bit on the short side, there will be a bit of HF rolloff, but I'd rather have that than marginal dynamic stability. I can always pick up some gain somewhere if this bothers me.
As I understand it, with pentode mode there is a very real need to have some resistance in the order of 1k-2k permanently connected across the speaker terminals, whether that be part of the NFB loop or not. Without such a precaution, disconnecting the speaker can cause destructively high voltages across the OPT primary.
Hi ray_moth,
Generally it is a good idea to just slab a 150 ohms wirewound or something across the speaker terminals just for the reasons you mentioned. It does no harm whatsoever, but for a few Cents one can prevent a possible dammage up to almost a total loss of an amp (OPTs and power tubes).
I mean, great deal/value for a few added Cents, isn´t it?
Tom
ray_moth said:
Generally it is a good idea to just slab a 150 ohms wirewound or something across the speaker terminals just for the reasons you mentioned. It does no harm whatsoever, but for a few Cents one can prevent a possible dammage up to almost a total loss of an amp (OPTs and power tubes).
I mean, great deal/value for a few added Cents, isn´t it?
Tom
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.