Yes, thanks for bringing up than Crowhurst article (post #35) SY.
The general reason for many problems is often that the basics are not done right. Fact: If there is ringing, there is an oscillatory arrangement somewhere in the loop. While a reactive situation can never be avoided, one can arrange for the loop gain to be low enough in those regions - basic Nyquist/Bode stuff. But as Crowhurst illustrated, often overlooked. This is exactly why I mentioned judicious application of a C over the feedback resistor. (Also, when looking into this stage by stage, one needs to use a low C (high Z) scope probe, otherwise it could diminish ringing especially when applied to anode points. I have been flummoxed by that one! - I now also keep one scope trace permanently on the output to monitor measurement influences.)
Practical conditions act somewhat to our advantage. Although loudspeaker reactance is inductive at high frequencies, the LS "generator" impedance also rises. Thus it looks into a comparative progressively lower impedance load the higher the frequency (also as a result of OPT capacitance), damping induction of h.v. transients. But the danger of this very fact as a cause of ringing must of course be addressed, toward which purpose loading the output with a resistor is wise.
Another point illustrated by the Crowhurst article is the danger of achieving a low output Z mainly by NFB, instead of making the output stage low-Z in itself to begin with (in this respect triodes score!). There was an article by Otala on this ("Intermodulation at the amplifier-loudspeaker interface"; Wireless World, Nov/Dec 1980) written mainly for semiconductors, but illustrating this important design consideration generally.
I must admit to neglecting proper low frequency investigations myself. As long as it seems OK when overloading the amp temporarily..... It would be illuminating to check this with a 0,5 - 1 Hz square wave, which I will do in future.
Regards.
The general reason for many problems is often that the basics are not done right. Fact: If there is ringing, there is an oscillatory arrangement somewhere in the loop. While a reactive situation can never be avoided, one can arrange for the loop gain to be low enough in those regions - basic Nyquist/Bode stuff. But as Crowhurst illustrated, often overlooked. This is exactly why I mentioned judicious application of a C over the feedback resistor. (Also, when looking into this stage by stage, one needs to use a low C (high Z) scope probe, otherwise it could diminish ringing especially when applied to anode points. I have been flummoxed by that one! - I now also keep one scope trace permanently on the output to monitor measurement influences.)
Practical conditions act somewhat to our advantage. Although loudspeaker reactance is inductive at high frequencies, the LS "generator" impedance also rises. Thus it looks into a comparative progressively lower impedance load the higher the frequency (also as a result of OPT capacitance), damping induction of h.v. transients. But the danger of this very fact as a cause of ringing must of course be addressed, toward which purpose loading the output with a resistor is wise.
Another point illustrated by the Crowhurst article is the danger of achieving a low output Z mainly by NFB, instead of making the output stage low-Z in itself to begin with (in this respect triodes score!). There was an article by Otala on this ("Intermodulation at the amplifier-loudspeaker interface"; Wireless World, Nov/Dec 1980) written mainly for semiconductors, but illustrating this important design consideration generally.
I must admit to neglecting proper low frequency investigations myself. As long as it seems OK when overloading the amp temporarily..... It would be illuminating to check this with a 0,5 - 1 Hz square wave, which I will do in future.
Regards.
On second thoughts, my last paragraph may not be so revealing as switching a sine wave input of say 1 KHz temporarily to overload amplitude and back to normal, and view the result on a slow scope sweep (a digital scope is handy here).
The degree of, shall we say, catastrophic overload would be related to the duration of the overload causing signal. It will involve power supply filter time constants for longer durations - not such a simple test, in the end! We are getting into the matter of automatic limiting just before overload as with semiconductor amps.
This is going a little off-thread, but what about it? I have never seen such measures used in tube amplifiers, but it would cure some basic NFB woes.
Regards.
The degree of, shall we say, catastrophic overload would be related to the duration of the overload causing signal. It will involve power supply filter time constants for longer durations - not such a simple test, in the end! We are getting into the matter of automatic limiting just before overload as with semiconductor amps.
This is going a little off-thread, but what about it? I have never seen such measures used in tube amplifiers, but it would cure some basic NFB woes.
Regards.
Johan, that's exactly how I test my amps. I have a piece of software that turns my sound card to a signal generator. You can have a sine wave switch between two amplitudes at any rep rate you choose. I set it for 6dB overload alternating with 3dB under. Watching this on a scope is very, very educational and diagnostic of bad recovery behavior.
Here is another informative Crowhurst on triodes vs. pentodes:
www.aikenamps.com/TI_Other.htm and go to "Puzzled about amplifiers?" (I could find only pdf)
A little off-thread, but since the door has been opened: Ringing on a square wave input...... Crowhurst is sort-of right in the previous article regarding trickery, but there will always be more ringing percentage-wise in early stages because usually power stages have the poorer bandwidth, unless a very superior (and expensive) OPT is used. Think of it; if one wanted to start a cut with the usual RC (anode first stage) not to deep into the audio band, then the rest must be OK up to 100KHz or more for simple stability (say 20 dB of NFB). To my mind such ringing should not overload previous stages up to the signal level at maximum output power, not necessarily be less as a percentage-of-signal than at the output. That was not stated directly.
I did a check this afternoon on an amp to confirm that. (I also blew a resistor when I did not take care to keep the unshielded probe tip away from sensitive circuitry - contrary to my own earlier advice. Permission to laugh.)
Then we usually test with a fast rise-time square wave input (I do too). Wrong! Taking that input components much higher than 20 KHz are not useful, such ringing (at say 80 KHz or higher) will not necessarily be excited in the first place. (.... and this a a very good reason for including a 25 KHz low-pass filter in a pre-amp). I think Crowhurst indicated that one should let such ringing be, provided it is not serious enough to push the amp into sporadic oscillation. In this regard a frequency sweep to determine the exact nature of any h.f. peaks is more realistic than a square wave test, though the latter remains a quick test.
I apologise if this is old hat to those with experience (as several of you are), but it could be informative to less enlightened members. It is also a reminder to me to resist temptations to get fancy.
Regards.
[PS: Regarding the enhanced triode operation reference previously posted: To my mind the necessary screen-anode power supply will pose a real problem capacity-wise, apart from other objections. UL comes close enough to pushing the lower limit of anode swing down - why use something else?.]
www.aikenamps.com/TI_Other.htm and go to "Puzzled about amplifiers?" (I could find only pdf)
A little off-thread, but since the door has been opened: Ringing on a square wave input...... Crowhurst is sort-of right in the previous article regarding trickery, but there will always be more ringing percentage-wise in early stages because usually power stages have the poorer bandwidth, unless a very superior (and expensive) OPT is used. Think of it; if one wanted to start a cut with the usual RC (anode first stage) not to deep into the audio band, then the rest must be OK up to 100KHz or more for simple stability (say 20 dB of NFB). To my mind such ringing should not overload previous stages up to the signal level at maximum output power, not necessarily be less as a percentage-of-signal than at the output. That was not stated directly.
I did a check this afternoon on an amp to confirm that. (I also blew a resistor when I did not take care to keep the unshielded probe tip away from sensitive circuitry - contrary to my own earlier advice. Permission to laugh.)
Then we usually test with a fast rise-time square wave input (I do too). Wrong! Taking that input components much higher than 20 KHz are not useful, such ringing (at say 80 KHz or higher) will not necessarily be excited in the first place. (.... and this a a very good reason for including a 25 KHz low-pass filter in a pre-amp). I think Crowhurst indicated that one should let such ringing be, provided it is not serious enough to push the amp into sporadic oscillation. In this regard a frequency sweep to determine the exact nature of any h.f. peaks is more realistic than a square wave test, though the latter remains a quick test.
I apologise if this is old hat to those with experience (as several of you are), but it could be informative to less enlightened members. It is also a reminder to me to resist temptations to get fancy.
Regards.
[PS: Regarding the enhanced triode operation reference previously posted: To my mind the necessary screen-anode power supply will pose a real problem capacity-wise, apart from other objections. UL comes close enough to pushing the lower limit of anode swing down - why use something else?.]
RE: Crowhurst
I've read three of Crowhurst's books, and numerous articles - including those on feedback and compensation - and I have to point out that he is fairly inconsistant in his concepts.
He does say square wave tests are meaningless in one of his Gernsback books, and then discusses nothing but square wave tests in MANY articles written for Audio and other magazines. The same for the use of the phase compensation capacitor. And, having read it many times, I have to say that his reasoning behind leaving off this capacitor is spotty and hard to follow.
The square wave is a great diagnosis tool, and tells you many things about an amp at a glace. Saying they are "useless" for testing audio amps is as silly and misguided as saying they are the only important thing.
Knowing what is going on above 20kHz and below 20Hz is vital. You simply cannot design a good feedback amp without understanding its full response.
As for phase compensation: the capacitor across the feedback resistor performs a useful job, and is NOT the cause of wild internal oscillations "flying around the feedback loop" as Crowhurst seems to imply.
Also, people seem to confuse "ringing" with an oscillitory response. They are not the same thing. All amplifiers with output transformers will ring when pulsed. They only vary in duration and amplitude. An oscillitory response is caused by the feedback loop no longer being able to limit the gain of the amplifier below unity due to excessive phase shift in the sampled output. The phase shift is found at high frequency due to all the capacitances in the circuit. The feedback signal thus begins to lag behind the input signal more and more, until it actually lags so much it becomes positive feeback - overloading the input while we're at it - causing the spike in response you see. So, by adding the phase compensation capacitor across Rfb, we can "speed up" the feedback signal, overcome the lag and hopefully - if we chose it correctly - keep the % of signal fed back to the input relatively constant. It is elegant.
Of course this assumes that we start with a well-behaved amp, because we don't want the feedback loop to have to clamp down on too large of a signal. That's tough with an output transformer, and multiple stages! But the answer is to add step-response RC coupling circuits for the HF and LF. You can try to model these, but I find experimentation to work the best.
How do you get to the fabled "6dB of stability margin" without breaking out the math books and getting a headache??? You plot the response (in dB) with the feedback loop disconnected, then plot the response with the feedback loop connected. If you never have more than a 6dB rise in gain at any frequency with the loop closed, then you have a stable amp.
I hope this made some sense.
Joel
I've read three of Crowhurst's books, and numerous articles - including those on feedback and compensation - and I have to point out that he is fairly inconsistant in his concepts.
He does say square wave tests are meaningless in one of his Gernsback books, and then discusses nothing but square wave tests in MANY articles written for Audio and other magazines. The same for the use of the phase compensation capacitor. And, having read it many times, I have to say that his reasoning behind leaving off this capacitor is spotty and hard to follow.
The square wave is a great diagnosis tool, and tells you many things about an amp at a glace. Saying they are "useless" for testing audio amps is as silly and misguided as saying they are the only important thing.
Knowing what is going on above 20kHz and below 20Hz is vital. You simply cannot design a good feedback amp without understanding its full response.
As for phase compensation: the capacitor across the feedback resistor performs a useful job, and is NOT the cause of wild internal oscillations "flying around the feedback loop" as Crowhurst seems to imply.
Also, people seem to confuse "ringing" with an oscillitory response. They are not the same thing. All amplifiers with output transformers will ring when pulsed. They only vary in duration and amplitude. An oscillitory response is caused by the feedback loop no longer being able to limit the gain of the amplifier below unity due to excessive phase shift in the sampled output. The phase shift is found at high frequency due to all the capacitances in the circuit. The feedback signal thus begins to lag behind the input signal more and more, until it actually lags so much it becomes positive feeback - overloading the input while we're at it - causing the spike in response you see. So, by adding the phase compensation capacitor across Rfb, we can "speed up" the feedback signal, overcome the lag and hopefully - if we chose it correctly - keep the % of signal fed back to the input relatively constant. It is elegant.
Of course this assumes that we start with a well-behaved amp, because we don't want the feedback loop to have to clamp down on too large of a signal. That's tough with an output transformer, and multiple stages! But the answer is to add step-response RC coupling circuits for the HF and LF. You can try to model these, but I find experimentation to work the best.
How do you get to the fabled "6dB of stability margin" without breaking out the math books and getting a headache??? You plot the response (in dB) with the feedback loop disconnected, then plot the response with the feedback loop connected. If you never have more than a 6dB rise in gain at any frequency with the loop closed, then you have a stable amp.
I hope this made some sense.
Joel
I must admit that I have had difficulty understanding some of Crowhurst's articles. He doesn't always explain things very clearly, IMHO.
I cannot understand Menno Van der Veen's published design for a PP EL34 amp that can be connected in triode, UL or pentode mode, wherein he avoids NFB altogether because he doesn't like it. I'm not questioning his taste, just the practicality of pentode PP without feedback.
It's not off-thread as far as I'm concerned, Johan, it's highly relevant. I think pentode output is inextricably linked with NFB and the stability problems thereof.
I cannot understand Menno Van der Veen's published design for a PP EL34 amp that can be connected in triode, UL or pentode mode, wherein he avoids NFB altogether because he doesn't like it. I'm not questioning his taste, just the practicality of pentode PP without feedback.
So do I.
With all respect, all that might be a stiff helping of window-dressing. I think the same gentleman invented the concept of a "quality factor" for output transformers, which had little to do with its practical performance.
That sort of thing smacks too much of a vending machine for my liking - you press the button and out comes whatever sweetness of beverage or chocolate you might want (if the machine works at all, of course).
With all respect, all that might be a stiff helping of window-dressing. I think the same gentleman invented the concept of a "quality factor" for output transformers, which had little to do with its practical performance.
That sort of thing smacks too much of a vending machine for my liking - you press the button and out comes whatever sweetness of beverage or chocolate you might want (if the machine works at all, of course).
I just stumbled across this thread and am a little surprised no one has mentioned Poindexter's(Audiotropic) fantastic 6V6 "Musical Machine". Poinz, as he's known, makes commerical products, but is also good enough to provide complete DIY info for his amps.
I built a fully balanced version of this amp from vintage Sansui 1000A OPTs and power transformers I had laying around. It is unbelievably pure and clean, but retains the balls of the 7591A's it uses.
You really can't get more simple than Poinz's circuit and he provides a complete parts list and where to get them. Don't let the bargain barrell components fool you, this amp is easily as good as my high $$$ commerical amps.
BTW, here's a 6Moons review of his commerical version of this amp. Don't overlook his Moebius preamp, either. Here's the 6Moons review for the commerical version of it, too.
Signal Circuit
Power Supply
I built a fully balanced version of this amp from vintage Sansui 1000A OPTs and power transformers I had laying around. It is unbelievably pure and clean, but retains the balls of the 7591A's it uses.
You really can't get more simple than Poinz's circuit and he provides a complete parts list and where to get them. Don't let the bargain barrell components fool you, this amp is easily as good as my high $$$ commerical amps.
BTW, here's a 6Moons review of his commerical version of this amp. Don't overlook his Moebius preamp, either. Here's the 6Moons review for the commerical version of it, too.
Signal Circuit
Power Supply
darkmoebius said:
No one mentioned that because it's not a pentode amp. The 6V6s are trioded.
An externally hosted image should be here but it was not working when we last tested it.
Active Ultralinear Pentode Output
I recently purchased some custom Edcor xfmrs made with 20% CFB windings and I plan on trying out active ultralinear CFB:
http://www.diyaudio.com/forums/showthread.php?postid=590565#post590565
I am using some cheap Horiz. deflection tubes for output tubes in 20% CFB configuration, 12HL7s for the linear corrector differential pair. DN2540s for buffered output screen drives from the 12HL7 plates (no UL taps on the xfmr required due to the CFB setup). This should make for a very linear and low output impedance pentode output stage.
Don
I recently purchased some custom Edcor xfmrs made with 20% CFB windings and I plan on trying out active ultralinear CFB:
http://www.diyaudio.com/forums/showthread.php?postid=590565#post590565
I am using some cheap Horiz. deflection tubes for output tubes in 20% CFB configuration, 12HL7s for the linear corrector differential pair. DN2540s for buffered output screen drives from the 12HL7 plates (no UL taps on the xfmr required due to the CFB setup). This should make for a very linear and low output impedance pentode output stage.
Don
It should indeed! Let us know how it works out, please.
On a different point of interest, I'm wondering about screen stoppers and whether these detract to any significant extent from screen regulation. I tend to think that stoppers are a good idea, for both the control grid and the screen, as a safeguard aginst parasitic oscillations. However, assuming one has a regulated supply for the screens, stoppers would seem to be acting against the regulation. Perhaps there's a happy medium, i.e. high enough resistance to be effective in preventing parasitics but also low enough not to have an adverse effect on the regulation?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.