I worked in product development in that plant for 41 years. Much of my early work was on the audio sections of two way radio products used by public safety personal (police and fire) where often life on the line. These are called "mission critical products." The engineering requirements are different, speech intelligibility is paramount, audio fidelity is not wanted. I could probably design and build a "blameless" tube audio amp, IF we, or anyone else could define it.
I have learned that whenever you get more than one engineer in a room to discuss a design (any design) you will have more than one opinion about how that circuit should be designed. Often we have strict criteria or goals about circuit performance, cost, size, recommended components and vendors, and quality metrics. There are requirements dictated by manufacturing, mechanical, and environmental testing teams, as well as real world feedback on our current products. User experience IS important to product development. What good is "blameless" if nobody will buy it. Motorola's later phones were far better technically than the iPhone, but which one won in the marketplace?
Even with strict requirements, design by committee doesn't work. After 20 years and several stalled designs, we transitioned to subject matter experts. The subject matter expert had the power to gather all input from engineering and decide the choice of design for a given sub-circuit. Then there were section teams, with a leader. A radio or phone would have a RF team lead, a Digital team leader, and a Mechanical team leader. Usually each development program would have a manager with ultimate authority. Even with this hierarchy, sometimes it could take months to define a future product before the circuit design could start.
I have been a subject matter expert (audio, receiver front end, and transmit RF) and a team leader (RF) and the "ultimate authority" on pre-production and "proof of concept" radio and phone designs. There are a few other members on this forum with the electrical engineering skills needed to produce a "blameless" tube amp design. I doubt that their blameless amp would look anything like mine.
Granted a tube amp doesn't have these mission critical requirements, since it only needs to satisfy the end user. But who could define the "blameless" amp. Doug set some initial boundaries. HIS preference was for bipolar parts, and he has valid engineering reasons for choosing them. Having accurate spice models IS a valid engineering reason. It is a REQUIREMENT for IC design, so this has led to a much better understanding of bipolar and MOS device models. Much of the spice work in the last 10 years has been centered on sub threshold CMOS and IC process characterization since this is where the money is today. Any advancement in vacuum tube models will come from this hobby.
I have found that some of the spice models found via this forum are actually good enough to produce designs that measure close to real world built and tested performance, although the FFT simulations are still a bit optimistic. We still deal with imperfect components, some of which will never be perfectly characterized, like tubes and OPT's, but we can get closer that the designers of the 50's and 60's did.
We must accept that the STATIC measurements made on audio amplifiers today (THD, IMD, TIM, PIM....) do not always correlate with a listening experience under DYNAMIC music conditions with a real world DYNAMICALLY changing complex (electrical magnitude and phase) load like a speaker. We must work toward dynamic measurements.
What is "blameless". As discussed earlier, it is about minimizing all CURRENTLY KNOWN sources of error. This could be done for any given choice of technology, but the amp produced might become unnecessarily large, expensive and complex. Still, what tech do we choose, triode, pentode or both.....SE or P-P? Do we allow the use of semiconductor devices where the solid ENGINEERING principals dictate, or do we impose a no sand rule for the sake of "purity"? Do we allow the use of a microprocessor to keep our drifting vacuum tubes biased at the optimum point. Do we design a "pure engineering" effort to minimize all distortion with maximum power output, or do we factor in cost, size, manufacturability and efficiency constraints. After all we must be able to afford, build, and use this creation.
Looking back through the enormous body of technical literature we can find some excellent vacuum tube engineering of the past. McDonalds patents (1957) on "augmented cathode followers" were the subject of at least two long threads here, leading me to simulate and build some amplifiers. They led me to a complex DSP controlled vacuum tube design with a modulated SMPS that kept the output tube operating at a constant voltage and current to avoid the Mu and Gm variations with operating point. This removes a lot of the 2nd harmonic distortion inherent in a triode.
So, which "blameless" design does this large committee want? I will pick one and build it someday. Chances are it won't be the same one you or any other engineer wants, but it will be mine. Right now my lab is all packed up and waiting for my new house to be built, so maybe next year!
I made a hot rod every 10 years or so. Around 2000 the land got too valuable and the drag strip closed. I sold my last hot rod project unfinished when I had to pack up and move 1200 miles. I doubt I will do another.
True, the transmitting equipment could do well beyond 40 KHz to accommodate all the subcarriers in use, but the TX audio was limited to 15KHz to avoid interference with the stereo subcarrier, even on mono stations. In any case many of the early FM stations in the US were limited by the Studio to Transmitter Link phone lines. It wasn't until stereo, SCA broadcasting, and data services that microwave STL's became cost effective.
I will guess that the middle ground feedback people adjusted the knob until they found something that sounded like what they had. In a pool of 1000 engineers and a few thousand other technical people, there were about a dozen or so people who had, or had heard a tube amp. Those all tried my amp, and it was one of them who bought it. They probably already had tube friendly speakers that worked good with a low damping factor.
I have several tube amps from a 2 WPC SET to a 125 WPC P-P sweep tube amp. I have several speaker systems, one of which (Yamaha NS-10M) was designed for a SS amp. In all cases I prefer ZERO GNFB, but often use local feedback around the output tube to tighten up the bass and remove boom.
I had not conquered local feedback at the time I made the adjustable feedback amp. I was building mostly guitar amps at that time......I did some similar experiments with guitar amps, but found even more variation due to speaker differences and playing styles.
The Presence control made popular by Marshall is a feedback control that works over a limited frequency range. Want some wild effects, put a volume - bass - treble - tone control stack in place of the presence control....but that is another experiment for another forum.
Well,
Some very knowledgeable interventions in this TD, make myself to clear up my thoughts about this subject. Long interventions from very experienced, technically educated people, helps to put black over white things, that were in a very elusive and confused form in some dark corner of your own mind.
Maybe then, there isn't a blameless standard for tube amplifiers, and never will be, because the very use of tubes and output transformers goes against the "blameless" concept itself, as intended by D. Self ; and we are building them on the basis of a subjectively perceived "niceness" than cannot be supported or defended in the field of a rigorous, and strictly technical performance analysis.
Still, can be interesting to put together some standard of performance goals, for a three stages PP amplifier, featuring moderate drive-ability for medium efficiency loudspeakers that most people can reach to own, and correctly place, room wise, at their homes.
Some very knowledgeable interventions in this TD, make myself to clear up my thoughts about this subject. Long interventions from very experienced, technically educated people, helps to put black over white things, that were in a very elusive and confused form in some dark corner of your own mind.
Maybe then, there isn't a blameless standard for tube amplifiers, and never will be, because the very use of tubes and output transformers goes against the "blameless" concept itself, as intended by D. Self ; and we are building them on the basis of a subjectively perceived "niceness" than cannot be supported or defended in the field of a rigorous, and strictly technical performance analysis.
Still, can be interesting to put together some standard of performance goals, for a three stages PP amplifier, featuring moderate drive-ability for medium efficiency loudspeakers that most people can reach to own, and correctly place, room wise, at their homes.
You think 40 ohms on the primary is significant? a KT88 anode resistance is around 12,000 ohms.
Global feedback does improve damping yes, but not as much as local feedback will. The reason is that you can apply far more local feedback, way beyond the point that your GNFB amp has turned into a radio transmitter or ultrasonic beacon.
Perhaps we need to go right back to amplifier 101?
If you insist on driving the OPT incorrectly (from a high impedance source like a tube without local feedback) the signal is guaranteed to be distorted. You'll then add your (newly distorted) negative feedback in a vain attempt to correct it.
But you talk about damping factor so lets look at what happens in the bass region where the OPT is rolling off: Oh yes - you'll have much reduced open loop gain so you'll get a very small amount of (distorted!) GNFB.
Which as I'm sure you know gives you the worst case for multiplying the harmonic distortions (because you are using feedback) and failing to correct them (because you haven't used enough).
I guess the trend for measuring distortion at 1kHz has blinded people to what goes on in the treble and bass - I'm sure GNFB Williamson style will be absolutely the best solution for all those 1kHz tunes in your library. It reminds me of the people who design amplifiers for silent signals - ignoring effects to the waveform at the extremes - e.g PSRR which affects the signal asymmetrically.
Pure global feedback works MUCH better than you're giving it credit for. Competent valve designs with just global feedback (if you ignore UL) can achieve a nearly ruler flat output impedance from 20Hz to 20Khz, of well under an ohm, without much of a rise in distortion at the frequency extremes.
Morgan Jones' Crystal Palace is another high power and low distortion amp that might possibly be called Blameless, except it could easily be blamed for a hernia or back trouble.
I can personally attest to both.
I can hear Morgan wincing from here.
Once you get the amp good enough that its output cannot be audibly distinguished from its input, it's blameless and you're done. That's really not terribly difficult to do. The CP does indeed do that, and so do many other good tube designs. Their shortcomings are in the deepest bass, but that only accounts for a tiny fraction of recorded music and loudspeakers with that capability.
I'm not saying that GNFB doesn't work at all, I'm saying that however well it works, local feedback will work better.
The basis of this claim is:
1) A more linear amplifier is better than a less linear one with feedback.
2) Local feedback can be applied _way_ harder than GNFB.
3) At transformer roll-off you are still benefitting from all the local feedback
Sure you can make a decent sound using GNFB as long as you avoid the oscillations/motor boating, but using those exact same components you'll create a better sound with local feedback due to the above reasons.
I always feel that the concept of adding phase lags into a feedback circuit goes against the grain too
Yes, it could well be they looked for something they already had or liked.
It might have been that some tried it at one extreme, and lazily thought to themselves, well, better do the right thing by Tubelab, lets' give this knob a twist. And turned it right round opposite to where it was before. And said to themselves "Hmm, that does sound very slightly better (or worse, as the case may be)".
Maybe some actually tried to carefully find what to them was the best position, trying it with various types of music, video sound tracks, or whatever.
Remember, some folk are so cloth eared a very ordinary 5-tube radio sounds good.
I haven't done a survey with a tube amp like you have, but I have been involved in survey of customers on other things. One of the things you learn in the survey business, either the hard way or when discussing what questions to ask with the expert from the survey consultants, is that too much resolution without precision leads you astray.
Survey questions like the following:-
How to you rate the realism of this amplifier (score between 0 and 9; 0=wost, 9=best)?
will give a result that is full of random human factors and all over the place.
A question worded like this:-
Please choose a number that represents the realism of this amplifier
(0 = unacceptable, won't try it again;
1 = not as good as my existing amp;
2 = about equal to my existing amp;
3 = better than my existing amp)
will, especially if you know what their existing amp is, give you coarse grained data, but data you can rely on.
It may have been that for at least some people, that the distortion reduction or improvement in speaker damping, or both, was perceptable at zero feedback, but reduced below what those individuals could detect at only moderate feedback. With such individuals, what knob position they reported as best, if not 'zero', could be just random. For other individuals, maybe they had better ears, and could tell the diffrence between midling and full feddabck.
What was the amount of feedback at the maximum position of the knob anyway? One man's "lots" is another man's "little" If you are a long time tube man, 20 dB of neg feedback is large. To an ss op-amp man, 20 dB is not much at all.
Last edited:
Remember most of my test subjects are from the iPOD / MP3 generation. A good chunk of them don't know how to listen. If it is loud, and the subwoofer shakes the house, it must be good, right?
I don't remember exactly, but it was near 20 db. The line stage in front of the amp was a typical RC coupled 6CG7 which makes about 20 db of gain. The magic knob reduced this to unity as the feedback was reduced to zero. At full gain the feedback level was adjusted to match the gain of the line stage resulting in the same volume from the amp as the knob was turned.
This is a matter of personal preference on the low end, and limited on the high end by the quality of the OPT and how many reactive elements you have in the forward path.
Ideally the phase of the feedback is 180 degrees from the input signal resulting in cancellation. Any reactive elements (OPT, IT, coupling caps, cathode bypass caps, Miller capacitance.......) in the forward path or the feedback path will cause a phase shift at one end of the audio frequency range, If the sum of all phase shifts cause the feedback signal to arrive at the input in phase at any frequency where the open loop gain is greater than 1 the amp will oscillate. Even a phase shift less than needed for oscillation will make an amp unstable.
I tend to use no more than 6 to 10 db of GNFB if I use any at all.
Lol, the most realistic voice I have ever heard came from a tube table-top radio. It was tuned into BBC R4 and it was spooky, it had us looking around for the person speaking
. It wasn't just me listening either. It was an interesting moment!
Lol, the most realistic voice I have ever heard came from a tube table-top radio. It was tuned into BBC R4 and it was spooky, it had us looking around for the person speaking
It wasn't just me listening either. It was an interesting moment!
It doesn't surprise me. There were tube radios and there were tube radios... And despite receiving complaints about technical quality from time to time, and seemingly quite ordinary technical specs on their equipment, BBC engineers knew what they were doing. It helps that there is just about zero energy below about 70 Hz in any human voice, not much above 3.5 kHz, and practically nothing above about 8 to 10 kHz.
Last edited:
I'm not saying that GNFB doesn't work at all, I'm saying that however well it works, local feedback will work better.
The basis of this claim is:
1) A more linear amplifier is better than a less linear one with feedback.
2) Local feedback can be applied _way_ harder than GNFB.
3) At transformer roll-off you are still benefitting from all the local feedback
Sure you can make a decent sound using GNFB as long as you avoid the oscillations/motor boating, but using those exact same components you'll create a better sound with local feedback due to the above reasons.
I always feel that the concept of adding phase lags into a feedback circuit goes against the grain too
You're still at risk from oscillation and motorboating with local feedback. The transformer primary is still by no means an ideal load either so you still need to be careful how the feedback is applied.
The low frequency roll-off still appears in much the same way regardless of global or local feedback - the OPT primary only has finite inductance so you're still fighting an uphill battle as frequency drops. The feedback loop is still fighting this regardless of being referenced to the primary or secondary.
Local feedback is a valid stragety and does in fact work very well (I have experimented), but it's not a panacea, in my opinion.
I had one of those realism moments with music played over FM radio to my factory Ford (Mach 460) car stereo.
There was a pirate radio station broadcasting from a house in our neighborhood. I used to routinely turn on his online guitar lesson show, set my amp in an open window, crank it and play along. He would put the chords and tab on his web page, show the tricks needed to play a song, and then play it through a few times.
The station had enough transmitting power to cover my drive to work (about 4 miles) so I listened often. I got in my car to drive home from work one evening and turned on the stereo. It seemed that there was a live female singer right in front of me on the windshield and an acoustic guitar player
in the passenger seat.
The illusion lasted all the way home, but I stopped at the pirate station, walked up and knocked on the door (I had been there before to play along with the guitar program). The owners wife made the motion for me to be quiet and ushered me into the garage where the guitar player and singer were performing live into microphones that were plugged directly into the computer's sound card. The entire pirate station ran from the PC and usually played pre recorded shows until night time.
There was minimal signal processing in the entire signal chain and ....not a single vacuum tube!
The pirate station was on the air for 7 years. It went off the air when the owner died, but bits and pieces remain on the internet. The old guitar lesion show was called the "G spot" and no longer exists since it was one of Jack's shows. He played a Danelectro.
http://www.jackandjillradio.com/
Last edited:
No, you can't do more with local feedback than with global feedback. You cannot apply far more local feedback because it just isn't possible without significant penalties.
Lets take an example: An EL34/6CA7 in Class A, operating at the datasheet conditions for maximum power output, which is 11 W at 10% THD.
Operating conditions are:
Vht .. 250V
Ia .... 100 mA
Transformer primary 2000 ohms
Transfer gain (gm) 11 mA/V.
For an output of P watts, and a primary impedance Rl the AC anode signal current is Ias = (P/Rl)^0.5
The signal voltage at the grid Vi is then Ias/gm = (P/Rl)^.5 / gm
For an 8 ohm load, the loudpeaker voltage is Vl = (P x Rspeak)^0.5
The maximum possible feedback is when ALL of of the loudspeaker voltage is fed to the cathode opposing the input to the grid. Thus the maximum possible feedback is 20 log[(Vi + Vl)/Vi] = 20 log[1 + (Rl Rspeak)^0.5 gm]
which equals 7.6 dB.
7.6 dB is not a lot of feedback. In high quality tube amplifiers, much more global feedback is used, 20 dB or more.
One can use the ultralinear method of local negative feedback. Haffler & Keroes and others have shown that minimum distortion typically occurs with 43% tappings, which corresponds to quite a bit less than the maximum possible feedback calculated above.
One can also use swinging feedback by means of a resistor from anode to grid. In this case the maxium possible local feedback is easily calculated from 20 log[1 + gmRl] and for the EL34/6CA7 is 27 dB.
Leaving for the moment that 27 dB applied locally in this way means an equivalent input voltage for full output of 53 V RMS (160 V pk-pk), which would be imposible for any pre-satge to deliver without considerable distortion, nobody with any sense would do this because:-
a) The local feeback would force the tube to hold its anode voltage constant and thus any hum on the HT rail will be applied in full to the transformer primary. Without any feedback, a pentode's dynamic anode resistance (15 Kohm for EL34) usefully attenuates hum, and feedback taken from the speaker winding attenuates it still more. For an EL34 the difference is a very useful 25 dB;
b) The output transformer is not inside the feedback loop and so will introduce distortion, expecially near the bass roll off.
One can also increase the local feedback by having a 3rd winding on the transformer, with more turns than for the speaker winding. This is sometimes done, but the extra winding in itself means more cost, and to avoid problems caused by leakage inductance it should be bifilar would with the speaker winding, which adds still more cost. And the more feedback you add this way, the more signal voltage is demanded from the preceding tube, increasing its' distortion. What you gain wrt distortion in extra local feedback you lose in having more distortion in the preceding stage.
Other common output tubes are even less favourable for local feedback.
Last edited:
Far less risk however because you have removed a laggy reactive load with a limited bandwidth from the loop. You have to try quite hard to make a local feedback loop lose stability, with GNFB you often get it for free
No, the way is very different.
With GNFB you have frequency variable feedback, with feedback disappearing when it's most needed.
With local feedback you have (for a significantly larger bandwidth) constant levels of feedback that give the transformer a better bandwidth.
I.e. GNFB squashes the mid-range gain to match the open loop gain at the frequency extremes. LF actually boosts the OPT bandwidth in the first place. Quite the opposite action.
Your GNFB is fighting it. Quite a messy fight too if you look within the loop on a scope. Local feedback isn't, it's merely applying an accurate signal to the primary and letting the transformer do it's job.
The concepts are entirely different.
Perhaps I'm using the wrong term, by 'Local Feedback' (which would be largest on a single tube for a high gain follower).
I'm actually referring to any amount of feedback that doesn't involve a transformer. It could even be an op-amp driving a FET follower - basically my argument here is one of removing the OPT from the loop, not one of making the loop as small as possible.
Of course I also think that any more than 2 or 3 amplifying devices in a loop is excessive, but my argument here is against the perverse practice of including the OPT in a loop.
I was hoping to provoke someone into demonstrating an advantage of doing that but so far nothing except tradition and 'that's the way it's always been done'
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.