Hi, respect to the G2 resistors (wen you go UL), I was not talking about a substantial voltage difference, but I sometimes saw in old UL schematics G2 a few volts over the plate, I think because there is the same B+ supply, but the path to G2 travels less copper, hence you have, say 300V on G2 , and 285V at the plate. This, with reason or not , makes me feel uncomfortable.
😉 😕 at what # post it was decided to degrade it to the lounge?
The Lounge A place to talk about almost anything but politics and religion.(and technical matters)
Oh! God! #142, this is my bust stop! , see you next time..
The Lounge A place to talk about almost anything but politics and religion.(and technical matters)
Oh! God! #142, this is my bust stop! , see you next time..
Last edited:
I think as soon as the "negative feedback going round and round" meme got busted out. It's not a time delay folks, it's a phase shift! If you don't understand the difference, you're not qualified to use global feedback anyway. 😀
A 15V drop at the plates wrt screens, assuming a 50% tap, means the HT must be 315V. Thus the power loss in the transformer DC resistance is 10% of the total delivered from the HT power rail. This is excessive and not representative of good engineering. One might not be surpised at this order in a cheap radio. But if the amplifier is good enough to be push pull and ultralinear, it's good enough to spend a bit more on the transformer. Otherwise on your example, you've got for each 10 watts delivered by the power rail, 1 watt is lost before it even gets to the anodes, which can only then raise about 4 watts of audio. Of this 4 watts, with this much primary resistance, and therefore a proportional amount in the secondary, only about 3 watts of audio gets delivered to the speaker. The transformer robs you of power getting the DC to the anodes, and robs you again passing the AC signal on to the speaker. Worse than a west african government official!
However, from the point of view of tube operating conditions, a 15 V difference plate to screen is not of any consequence. Screen current is roughly proportional to screen voltage at this level, so the variation in screen current is less than that due to manufacturing tolerances. Screen current and anode current is substantially unaffected by changes in anode voltage, providing the change is not totally inappropriate (as in when the anode winding is faulty & open circuit - you do get excessive screen current & glowing screens in tetrodes then).
The electrons from the cathode are accelerated radially outwards by the screen. By the time they get to the screen, most are going too fast to divert to the actual screen wires - by their inertia they just keep going right on past the wires. Their speed is directly proportional to the square root of the screen voltage, so if the screen voltage is 5% high, they are going about 2.5% faster. No big deal.
Last edited:
All these blameful feelings...
I said the joint was rocking Goin' round and round Yeah, reeling and a rocking What a crazy sound And they never stopped rocking Til the moon went down Well it sounds so sweet I had to take me chance Rose out of my seat I just had to dance Started moving my feet Whoa to clapping my hands...
I said the joint was rocking Goin' round and round Yeah, reeling and a rocking What a crazy sound And they never stopped rocking Til the moon went down Well it sounds so sweet I had to take me chance Rose out of my seat I just had to dance Started moving my feet Whoa to clapping my hands...
I built a SE amp using the soviet 6P41S tetrode with a 33K resistor bypassed with a 2.2uF film cap from the UL tap of the transformer to a 470R screen (G2)resistor. This kept the screen well below the anode, but still allowed the full ac signal to feed to it.
Probably not a 'blameless' design, but it worked quite well and sounds very neutral.
Probably not a 'blameless' design, but it worked quite well and sounds very neutral.
I tend to avoid all those cat fights over whether to include OPTs in gNFB loops or to exclude them (FWIW: In my designs, I include the OPT and take the NFB from the secondary.) Local NFB v. gNFB (Mathematically, makes no difference: 7.0dbv lNFB + 13.0dbv gNFB is the same as 20dbv of gNFB. I decide whether to use lNFB on a case-by-case basis.) It's mostly personal preference anyway, not worth dueling e-penises over this.
This, however, is pure nonsense. Feedback signals don't "go round and round". The NFB is applied to an inverting summing node, and it's the difference signal that is amplified. Remember, NFB is applied to RF amps that're implemented with BJTs because BJTs are inherently low frequency devices, and unless it's a long wave amp, they will be operating above their Beta-cutoff where the output falls off at a 6.0dbv/octave rate. NFB is one means to flatten the frequency response of a wide band, RF, BJT amp. If you don't see these "echos" while o'scoping such an RF amp operating at frequencies decades above audio, then they're not happening at audio frequencies either. Any time delay around the feedback loop translates into phase shift. That can play hell with stability, as there will be some frequency where the total phase shift adds up to 180deg, and if that happens when the gain is still above unity, you have an oscillator, not an amplifier. (Nyquist Stability Criterion).
"You don't see them SY - this isn't television, you hear them!!" -- if you think you hear something, but can't show its existence, well, psycho-acoustic effects. You can trick yourself into hearing just about anything. That's what keeps audiophool snake oil salesmen in business.
Nothing new, and all Lynn Olsen has demonstrated is:
You don't abuse NFB by using it to cover up your open loop mistakes. Fix your open loop design first, then add the NFB to make a good sounding design sound even better. All too many failed to do this, even back in the GAoHiFi. The end result, then and today, is "Big Box" audio systems that all sound alike, and mediocre. They get away with it because most buyers don't care. They haven't sat and just listened in years -- if ever -- and just play them for background noise while they do something else.
This, however, is pure nonsense. Feedback signals don't "go round and round". The NFB is applied to an inverting summing node, and it's the difference signal that is amplified. Remember, NFB is applied to RF amps that're implemented with BJTs because BJTs are inherently low frequency devices, and unless it's a long wave amp, they will be operating above their Beta-cutoff where the output falls off at a 6.0dbv/octave rate. NFB is one means to flatten the frequency response of a wide band, RF, BJT amp. If you don't see these "echos" while o'scoping such an RF amp operating at frequencies decades above audio, then they're not happening at audio frequencies either. Any time delay around the feedback loop translates into phase shift. That can play hell with stability, as there will be some frequency where the total phase shift adds up to 180deg, and if that happens when the gain is still above unity, you have an oscillator, not an amplifier. (Nyquist Stability Criterion).
"You don't see them SY - this isn't television, you hear them!!" -- if you think you hear something, but can't show its existence, well, psycho-acoustic effects. You can trick yourself into hearing just about anything. That's what keeps audiophool snake oil salesmen in business.
Nothing new, and all Lynn Olsen has demonstrated is:
You don't abuse NFB by using it to cover up your open loop mistakes. Fix your open loop design first, then add the NFB to make a good sounding design sound even better. All too many failed to do this, even back in the GAoHiFi. The end result, then and today, is "Big Box" audio systems that all sound alike, and mediocre. They get away with it because most buyers don't care. They haven't sat and just listened in years -- if ever -- and just play them for background noise while they do something else.
Digging back in the corner of my brain I remember something from that linear systems class in college......If the open loop gain is sufficiently large then the system takes on the (inverse) characteristics of the feedback path.
The feedback path in most of our audio amps is a passive voltage divider with a compensation cap to make up for OPT phase shift. This means that all global feedback amps with sufficient open loop gain will sound the same regardless of what's in the forward path.
We know that for the most part this isn't true, and this is because many tube amps are variations of older designs that did not have enough open loop gain to satisfy this criteria. Gain was expensive in the early days.
Back in the discussion about the use of solid state components in tube amps, someone mentioned the use of an opamp front end in a tube amp. Ever try this. I did and it sounded similar to the Carver / Phase Linear system that I had in the early 90's and wound up giving away.
The feedback path in most of our audio amps is a passive voltage divider with a compensation cap to make up for OPT phase shift. This means that all global feedback amps with sufficient open loop gain will sound the same regardless of what's in the forward path.
We know that for the most part this isn't true, and this is because many tube amps are variations of older designs that did not have enough open loop gain to satisfy this criteria. Gain was expensive in the early days.
Back in the discussion about the use of solid state components in tube amps, someone mentioned the use of an opamp front end in a tube amp. Ever try this. I did and it sounded similar to the Carver / Phase Linear system that I had in the early 90's and wound up giving away.
Further to George's (Tublab) comments above.
I do airborne laser scanners and stabiized platforms for the day job.
1 or 2 of the "Rules of Thumb" for Control Systems translate directly to audio stuff, such as "Make it linear as possible BEFORE applying fedback".
Other "Rules of Thumb" are near impossible to implement, such as "The ratio of the 2 most dominant poles should be greater than the the loop gain value". What that means is that if the amp has a loop gain of 20 (typical) and the output tranny (most dominant pole) is at say 60kHz then the next most dominant pole would need to be greater than 20 x 60kHz = 1.2MHz. Since that is usually impractical then we need to introduce methods to manipulate the frequency response whilst leaving the phase response largely intact. That is what those resistor + capacitor "stability" networks across the anode load resistor do.
For beginners at this stability stuff may I recommend that you ignore Nyquist diagrams and use Bode Plots instead. MUCH easier to understand and work with.
When I did "Control Systems" as a Postgrad Diploma subject (quite a few years ago now) they were just starting to teach the Bode Plot method in lieu of Nyquist.
@Gimp,
DC coupling will normally only affect the low frequency end, a low frquency roll off is what the control theory guys call a "Zero" where a "Pole" is the high frequency roll off. off course you can have stability problems at both extremes of the frequency range but the high frequency end normally causes most grief, why? because the low frequency roll off in the output tranny is formed by the Primary Inductance and the Driving Impedance which is a single roll off, whereas at the high frequency end yopu have 2 roll offs, 1 from leakage inductance plus driving inpedance and one from shunt capacitance plus drive impedance , these are additive and result in double the phase shift with increasing frequency. Note that reducing the drive impedance (by local feedback at the output tubes for example, or triode strapping or Ultralinear connection etc.) makes the low frequncy roll off lower and both high frequency roll offs higher. This then helps stabilize the amp when global feedback is connected.
Cheers,
Ian
I do airborne laser scanners and stabiized platforms for the day job.
1 or 2 of the "Rules of Thumb" for Control Systems translate directly to audio stuff, such as "Make it linear as possible BEFORE applying fedback".
Other "Rules of Thumb" are near impossible to implement, such as "The ratio of the 2 most dominant poles should be greater than the the loop gain value". What that means is that if the amp has a loop gain of 20 (typical) and the output tranny (most dominant pole) is at say 60kHz then the next most dominant pole would need to be greater than 20 x 60kHz = 1.2MHz. Since that is usually impractical then we need to introduce methods to manipulate the frequency response whilst leaving the phase response largely intact. That is what those resistor + capacitor "stability" networks across the anode load resistor do.
For beginners at this stability stuff may I recommend that you ignore Nyquist diagrams and use Bode Plots instead. MUCH easier to understand and work with.
When I did "Control Systems" as a Postgrad Diploma subject (quite a few years ago now) they were just starting to teach the Bode Plot method in lieu of Nyquist.
@Gimp,
DC coupling will normally only affect the low frequency end, a low frquency roll off is what the control theory guys call a "Zero" where a "Pole" is the high frequency roll off. off course you can have stability problems at both extremes of the frequency range but the high frequency end normally causes most grief, why? because the low frequency roll off in the output tranny is formed by the Primary Inductance and the Driving Impedance which is a single roll off, whereas at the high frequency end yopu have 2 roll offs, 1 from leakage inductance plus driving inpedance and one from shunt capacitance plus drive impedance , these are additive and result in double the phase shift with increasing frequency. Note that reducing the drive impedance (by local feedback at the output tubes for example, or triode strapping or Ultralinear connection etc.) makes the low frequncy roll off lower and both high frequency roll offs higher. This then helps stabilize the amp when global feedback is connected.
Cheers,
Ian
Last edited:
No.
Direct coupling is not required. One can insert resistor+capacitor pairs in approprite parts of the circuit so that the phase shift is corrected and held within limits until the forward gain is so low it doesn't matter.
A approach often used in audio is to size the transformer inductance and coupling capacitor sizes so that the roll off in gain with low frequencies is dominated by a single component, ideally one of the coupling capacitors. Then the amplifier will be unconditionally stable at the low end regardless of how much global feedback is applied. At the high frequency end a "dominant pole" capacitor is added to make the amplifier unconditionally stable at the high frequency end. High frequency phase correction networks called Zobel networks are often seen across the anode load of early stages.
Loudpeaker loads are reactive and can complicate the picture a little, but the extra compensation required is generally easy.
More sophisticated approaches are possible to make amplifiers unconditionally stable, with AC coupling, with high levels of feedback. When I was at university (a 4 year course in electronics engineering), we studied neg feedback / control systems (the theory of neg feedback is applicable to lots of things - not just audio amplification) in one subject in 2nd year. 2 more intensive subjects in 3rd and 4th year. A total of nearly 400 hours lecture time plus 300 hours laboratory time. And about the same time in doing calculations and design work before each lab. So it amuses me to see folk woffle on with all manner of nonsense (echoes, "sucking the life out the music", "better to apply lots of local..." etc etc) about neg feedback when they clearly couldn't hadle the first week of the feedback course.
Last edited:
Steve Bench mentions poles and zeroes and their applications to tube amps...
i wish there are more discussion about those here...i would like to learn....😉
i wish there are more discussion about those here...i would like to learn....😉
You are giving engineers to much credit and not enough to modern tools and resources.
By laying out requirements and metrics for performance any project could be worked on by any number of people. In fact the guidelines were already laid out by Douglas self, we just need to apply them to a tube amp. Instead of just one person pioneering this work would treat it like any other open source project. I'm not sure you have much experience with how open source software and development works. You still need some gatekeepers to decide what makes it into the final release but that is relatively trivial. You might want to look into opensource hardware development to see how this is becoming a regular way of creating new projects collaboratively. It's really not that hard to start a git-hub.
True 20 years ago access to tools needed to measure the performance of audio circuits were hard to come by. Now anyone with a computer and a halfway decent sound card has everything they need. For the most part everyone on this forum has access to the tools needed to measure circuit performance. I will admit that most do not use these tools but it's not like they don't have access.
We also have a forum filled with very knowledgeable people whose collective knowledge of audio far outweighs and individual.
The real question is if there are enough people interested in working on such a project. I'm not sure that there is.
Even a decent literature review would be good. As this thread clearly shows (again) that the existing knowledge isn't well known - reading up to this point of the thread I have run across at least half a dozen red flag postings.
One I've not read yet which sounds like it might be part way towards this goal is this:
Rethinking Distortion: Towards a Theory of 'sonic Signatures'.
By Bernd Gottinger ProQuest, 2007 - 471 pages.
Your first sentence: I don't think so. It's true that anyone can use a sound card to test frequency response, harmonic distortion, and more. But not too many can do it as well as professional gear.
You still need a brain. Anyone can go to a medical supply house and buy scapels, cardiopulmonary monitors, and anything else you can think of. But only people who have done the necessary 6 to 8 years of university and 2+ years of working under supervision can be a sucessful surgeon. It's to a large extent similar in electronics engineering.
I see lots of folk conjure up something in LTSpice and think they have the right answer. It's not that simple. And I have been using commercial versions of SPICE for over 40 years. That experience gives me a set of basic rules: If it doesn't work in SPICE, it is quite unlikely to work in a real implementation. If it does work in SPICE, it might work in a real implementation. If you built one, and it works to specification, that proves THAT ONE works to spec.
Almost all the contributions I see in diyAudio are tweakers and hobbyists. As I said, there's nothing wrong with that. It's a very good hobby.
One thing that Doug Self seemed to be driven to dispair quite a bit was folk who said they built to his design but it did not perform to spec. Each time he was able to investigate, it turned out they changed one or more things - used different transistors, changed a component value by error or deliberately, etc. Do folk on this forum have the necessary discipline?
But I can be convinced otherwise. Let someone step forward.
I do think you are right in your last sentence. I don't think there is much interest at all. Not surprising - it would be a lot of work and not a lot of fun.
- Status
- Not open for further replies.