20, a much better idea is to grab a copy of Horowitz and Hill and read the first few chapters; it's easy to find and even available in most libraries. Do the problems. Have fun with the Bad Circuits. With a solid grounding in the basics, the answers you're looking for will be obvious to you.
Not a silly question, but one which people still argue about. Low THD does not guarantee good sound, but too high THD does guarantee poor sound (except for those who prefer or can tolerate distortion). The main problem is that high THD guarantees high intermodulation distortion, as they both arise from the same mechanism, and intermodulation almost always sounds horrible although it is less noticeable with some types of music.
The signal injected at the input cathode will be smaller than the input. It is not usually injected at the input grid as this would lower input impedance, which is likely to increase distortion in the previous stage (typically output from source).
The input cathode resistor is sometimes constrained by bias requirements, or the need to avoid too much local feedback in the input stage - it provides local degeneration as well as defining the feedback point. Once the cathode resistor is set, the feedback resistor follows from the desired feedback ratio which sets the closed-loop gain.
There is a snag with feedback to the cathode. That is that the signal and feedback are not applied to the same point, so there is not pure subtraction. The common-mode gain and distortion of the input stage has to be considered (but, I suspect, is often overlooked). This issue gets worse with high levels of feedback.
The input cathode resistor is sometimes constrained by bias requirements, or the need to avoid too much local feedback in the input stage - it provides local degeneration as well as defining the feedback point. Once the cathode resistor is set, the feedback resistor follows from the desired feedback ratio which sets the closed-loop gain.
There is a snag with feedback to the cathode. That is that the signal and feedback are not applied to the same point, so there is not pure subtraction. The common-mode gain and distortion of the input stage has to be considered (but, I suspect, is often overlooked). This issue gets worse with high levels of feedback.
Again, with the books! BOOKS and BOOKS! What is Books??? (Meant to be an entertaining paraphrased quote from the original ''Star Trek" episode, 'Spock's Brain," in which his brain was removed to serve as the civilization controller by a group of beautiful Amazon alien women. When asked by Kirk where was Spock's brain they basically said they didn't know what he was talking about.
Seriously, the reference to the reference is appreciated. I will put in on my to read list right behind "Servicing Superheterodynes" by John F. Rider.
Last edited:
I tend to agree with DF96.
On purely practical basis I have never listened to or got a decent amp with small amount of loop feedback. Also, in many cases rather than getting increased level of higher harmonics I just could see that feedback only affected low order harmonics and left unaffected higher harmonics. This both with high quality PP OPT (100 KHz -3dB and first resonance well beyond) and good quality OPT.
For me the only practical way to find the right amount of FB is just doing it.
There are tube amps with FB that are as good as the best no-FB amps if not better.....
45
On purely practical basis I have never listened to or got a decent amp with small amount of loop feedback. Also, in many cases rather than getting increased level of higher harmonics I just could see that feedback only affected low order harmonics and left unaffected higher harmonics. This both with high quality PP OPT (100 KHz -3dB and first resonance well beyond) and good quality OPT.
For me the only practical way to find the right amount of FB is just doing it.
There are tube amps with FB that are as good as the best no-FB amps if not better.....
45
Keep in mind that you listen to a system, which comprises a source to transform a storage medium (or live mikes) to an amplitude signal, amplifiers, speakers, room, your ears, your brain, your preferences.
Note also that rarely is it possible to listen to two systems which only vary in amplifier THD. Perception of sound preferences may have nothing to do with changes in amplifier THD. If you change between a no feedback tube amp with highish output impedance and a high feedback amp with low output impedance, the frequency response of typical speaker crossovers will be affected - just one example.
Sheldon
Also, I have noticed that best results for me when using FB always come with plain resistive feedback and no compensations of any kind.
I don't think the OPT is the only to blame. General thinking is that it is the main cause but I don't think so.
It is true that for us common people the choice of tubes is somewhat limited and nobody can do anything about them while the OPT can be made according to the output device.
However I suspect that in some cases the OPT is just stretched it to its limit in terms of specifications, even using the best materials and best techniques.
Typical examples are big tubes like 211, GM70, 845 in SE. The OPT is very much compromised whatever one does in comparison to other solutions and I am not surprised at all about the fact that it is very difficult, or impossible, to get better results by applying FB. In this case the device is the problem for me. These tubes look desirable on chart but simply demand too much in practice....
45
I don't think the OPT is the only to blame. General thinking is that it is the main cause but I don't think so.
It is true that for us common people the choice of tubes is somewhat limited and nobody can do anything about them while the OPT can be made according to the output device.
However I suspect that in some cases the OPT is just stretched it to its limit in terms of specifications, even using the best materials and best techniques.
Typical examples are big tubes like 211, GM70, 845 in SE. The OPT is very much compromised whatever one does in comparison to other solutions and I am not surprised at all about the fact that it is very difficult, or impossible, to get better results by applying FB. In this case the device is the problem for me. These tubes look desirable on chart but simply demand too much in practice....
45
That was just a point about what I have seen and what I would take as sign of FB working as expected or not. As an indication rather than conclusion.
Then for the sound it is not just about harmonics, of course.
I can tell that recently I have been playing with a 6DR7 PP amp which produces so low distortion without FB, excellent frequency response, low IMD, no blocking distortion, better damping factor than most tube amps than one could think that FB would just make it worse. Instead with right amount of FB (12 dB in this case) it is so much better in just about everything it does.
It just sounds better.
45
Last edited:
Another thing I think is very important is the effective use of the amp.
I agree to a certain extent on the idea that the first watt is probably the most important but I also think that headroom is important. With normal speakers (say, 87-90 dB efficiency in normal room) I think 7-10W is really the minimum Pout required to the amp with FB to work properly.
If I have a 2W amp it's logical for me that it will not work very well with feedback just because of the way it clips (abrupt change in character) when it approaches its limited power.
I consider the above mentioned 6DR7 amp as a 5W pure class A amp in the way I use it although clipping starts at 10W! I can't see any problem in that because a 2A3 SE amp is still less efficient, more expensive and more critical for certain aspects.
Has someone ever made an EL84 PP amp using an approach departing from the usual low cost amp? If so, I think he would perfectly understand what I am trying to tell.
45
I agree to a certain extent on the idea that the first watt is probably the most important but I also think that headroom is important. With normal speakers (say, 87-90 dB efficiency in normal room) I think 7-10W is really the minimum Pout required to the amp with FB to work properly.
If I have a 2W amp it's logical for me that it will not work very well with feedback just because of the way it clips (abrupt change in character) when it approaches its limited power.
I consider the above mentioned 6DR7 amp as a 5W pure class A amp in the way I use it although clipping starts at 10W! I can't see any problem in that because a 2A3 SE amp is still less efficient, more expensive and more critical for certain aspects.
Has someone ever made an EL84 PP amp using an approach departing from the usual low cost amp? If so, I think he would perfectly understand what I am trying to tell.
45
Not this nonsense again.
BJTs have such low impedances that it is quite possible to build "no tune" RF amps that can operate across the ham bands from 10M to 180M. There is, however, a problem. When perusing BJT spec sheets, you'll come across a rating something like: "Current gain bandwidth product", and it might be something like 250MHz. That sounds quite impressive until you realize that it is always, always, always the alpha cutoff frequency, and applies only to the grounded base topology. That's because it looks more impressive. The beta cutoff will be a good deal lower, and applies to the much more common grounded emitter topology. If Hfe= 70, then a= 70 / 71= 0.986.
Fb= Fa X (1 - a)= 250 X (1 - 0.986)= 3.52MHz
Once you drop below the 80M band, your no-tune RF amp starts to roll off at the usual -6.0db per octave. This is not a good thing. How do you compensate for that frequency roll-off? You use good ol' gNFB! You tailor the frequency response of your feedback loop to roll off above 3.52MHz, so as to equalize the gain at the high end of the 10M band, and on up into the 160M band.
Remember: we're talking about amps that operate at frequencies that are orders in magnitude higher than audio frequencies, and gNFB still works just fine. These "no-tune" RF amps are in wide use, and the GD things are causing good quality, air variable capacitors to become unobtanium. (I've gone to ham fests, bought and dragged home some non-functional boat anchor just to recover its air variable capacitors.)
As for how gNFB relates to audio, this has been very well characterized a long time ago by investigators like Norman Crowhurst. What they discovered, both from mathematical analysis and empirical investigation is that gNFB serves to greatly reduce the near harmonics (h2, h3 mainly) but at the price of generating lots of low level h4 and up. What does that do to the sound? Here, you're back to listener interpretation, governed by all sorts of variables like personal preference, preferences in program material, preferences shaped by prior experience, listener expectation, and other psycho-acoustical factors. So we're right back to that old problem: if we could design that Ultimate Amp that everyone likes, we could just do it and shut down the whole forum for good. Not. Gonna. Happen. Still, gNFB has just esssssss-loads of advantages: it reduces Zo, makes the circuit performance less dependent on active and passive devices, makes for lower distortion and noise.
Even if you could design that absolutely perfect "wire with gain" that had no distortion whatsoever, I guaran-CUSS-tee you that someone, somewhere (actually a lot of "someone's") wouldn't like it at all.
As for my results: solid state designs (especially BJT-based) like lots of gNFB, so pour it on! I've done such designs, and they definitely sound much better than any "Big Box" offerings I've heard. What's important here is to keep the finals well away from the rails, so that your device capacitance doesn't go all squirrelly on you. If you're gonna use MOSFETs or IGBTs, then don't copy topologies originated for BJTs. Those designs work for BJTs, but other active devices are a whole 'nother story, and just don't work as well. After all, "complimentary" N-Channel and P-Channel devices are a good deal less "complimentary" than complementary NPN/PNP pairs. Avoid them!
Hollow state designs don't like nearly so much. 20db(v) of gNFB sounds positively hideous, and will give your hollow state amp the "great transistor sound" (not good). 12db(v) of gNFB might be a useful compromise value, though it's definitely tending towards that solid statey sound. That much gNFB tends to lend genres like Metal, Techno, and New Wave a "subdued" sound. (Though that is less obvious with softer Rock and Classical.) 6.0db(v) -- 7.0db(v) sounds about right with Metal, as it lends enough "edge" to bring that type of music to life.
The best way to do a design is to make sure your open loop design is a good one. (Before finalizing a design, I like to spend at least a week listening open loop before deciding on what it'll require so far as NFB is concerned. Some finals need more help (e.g. 807, 6L6) than others (6V6, 6BQ6GA / GTB)) If that doesn't work right, better redesign it than add essssss-loads of gNFB to sweep your mistakes under the carpet. If you do that, you also sweep away a lot of the vitality of the music. Then, what's the point? You can always find an amp that sounds like that at any "Mostest Goodest Buy" or "Circuit Village" or some place like that.
Well,
cos(sin(x))
Chained trig functions are notoriously annoying for analysis, and certainly do not arise in the study of phase shifts.
If you want to look at the ratio of sine to cosine of a particular frequency (i.e., using the trigonometric Fourier series), that's fine. "Cosine of sine" is something very different, and is not fine.
Because 1. that's usually the positive input, so it wouldn't be NFB, 2. that's also the signal input, so the feedback ratio depends on source impedance (being shunt feedback), and 3. because tubes are expensive, traditional amplifiers are single ended as far as possible. Therefore, the only reasonable opportunity for feedback is the cathode, which is a fine input because the feedback is low impedance.
Err, input voltage is GRID-TO-CATHODE voltage. Kirchoff says it's subtraction.
Much as audiophools like to think otherwise, they haven't succeeded in contradicting Kirchoff's Laws.
Tim
Last edited:
If you apply the same voltage change to grid and cathode, you will get some signal at the anode even though the grid-cathode voltage difference remains unchanged. This is because a triode has finite anode impedance. It is common-mode gain. As I said, it is often ignored. Thanks for confirming this for me!
Consider ball-park figures: An ECC81/12AT7 triode as input stage. Valve mu is 50, gm about 5mA/V, Ra about 12K, anode load 50K. Assume 20dB feedback. 1V input signal at the grid would be balanced against 0.9V feedback signal at the cathode. Assume a 100R cathode resistor to apply the feedback. Then the effective gm is about 3mA/V due to degeneration, and Ra goes up to about 20K, mu stays at 50. We have 0.1V differential mode signal, which becomes 5V due to valve gain. We have 0.95V common-mode signal. Both then get reduced at the anode by the same fraction due to Ra - the ratio is 5/7. So the unwanted common-mode signal is about 19% of the wanted differential mode signal. That alone means that the amp will have different gain from what we calculated. In addition, the common-mode signal will have different distortion characteristics as it suffers directly from the non-linear anode impedance, while the differential signal will be amplified more linearly as it gets the benefit of the usual triode action. Maybe not a huge effect, but something is 19% different from what we may naively expect. Low-mu valves, much favoured by audiophiles, would have a bigger effect.
You can apply negative feedback to the input grid, just by swapping the polarity of the OPT secondary.
Consider ball-park figures: An ECC81/12AT7 triode as input stage. Valve mu is 50, gm about 5mA/V, Ra about 12K, anode load 50K. Assume 20dB feedback. 1V input signal at the grid would be balanced against 0.9V feedback signal at the cathode. Assume a 100R cathode resistor to apply the feedback. Then the effective gm is about 3mA/V due to degeneration, and Ra goes up to about 20K, mu stays at 50. We have 0.1V differential mode signal, which becomes 5V due to valve gain. We have 0.95V common-mode signal. Both then get reduced at the anode by the same fraction due to Ra - the ratio is 5/7. So the unwanted common-mode signal is about 19% of the wanted differential mode signal. That alone means that the amp will have different gain from what we calculated. In addition, the common-mode signal will have different distortion characteristics as it suffers directly from the non-linear anode impedance, while the differential signal will be amplified more linearly as it gets the benefit of the usual triode action. Maybe not a huge effect, but something is 19% different from what we may naively expect. Low-mu valves, much favoured by audiophiles, would have a bigger effect.
You can apply negative feedback to the input grid, just by swapping the polarity of the OPT secondary.
Tim, from the context it should be clear that the poster didn't mean the literal cos(sin(x)) but he meant the 90-degree shifted version of sin(x) which looks like cos(x).
jan didden
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.