Gabevee said:7N7,
Well, I guess that's another two textbooks I need to throw away.
That might apply with solid state op amps without internal compensation due to phase shift.
But tube amps? If there is such severe phase shift that NFB causes high frequency instability... it is a bad design to begin with, no disrespect to Williamson intended.
BTW, I have seen (and have on my hard drive) the original Williamson schematic... what stepped network??? Just a simple resistor to the cathode of the first pentode voltage amp tube.
Otherwise, as for voltage versus power amplification... that is why I said I was not comparing apples to apples. Then I compared it to the 300B. The 7189 can indeed and has been used as a straight voltage amplifier.
http://www.triodeel.com/achpwill.gif
http://www.triodeel.com/dynawil1.gif
http://www.triodeel.com/images/ehpa100.gif
http://www.triodeel.com/images/utc20w.gif
http://www.xs4all.nl/~ideas/amps/
(no, this is not my hard drive, but where I got them from)
Note the first three use a bypass capacitor for the compensation... though generally not necessary. I have used it once and found that my amp oscillated at 100kHz with it. So I don't use it. Also note the last one is of the original Williamson!
If five Williamson amp schematics with the simple basic NFB that virtually all tube amps use isn't convincing empirical evidence enough... I give up!
Gabe
I'm afraid I do not have time to read all those liks that you kindly attached.
The step network I referred to is on the anode side of the input stage in the Williamson.
It is well known that the Williamson is marginal on stability - this despite its negative feedback
Best of luck
7N7
A bit more
First I should say that on this subject I am already on the outer limits of my knowledge; however I believe what I have written to be true.
I have never applied feedback successfully: I cannot face the mathematics for feedback for a SE stage and my attempts with balanced feedback (I do only differential, balanced amplifiers) have been a total failure.
This is a nuisance because I am not opposed to feedback on principle - only all the additional grief it causes.
I would love to use feedback to adjust the gain of my amplifiers; a good engineering solution and it would permit me a wider choice of valves. But I cannot do it so it's no-go area for me.
As for valve amplifiers being inherently unstable, I am not sure about this. It is true that wherever there is a combination of gain, capacitance and inductance it is reasonable to expect that there might be some trouble; but surely this applies to solid state amplifiers too? (I know less about solid state amplifiers than I so about feedback!).
7N7
First I should say that on this subject I am already on the outer limits of my knowledge; however I believe what I have written to be true.
I have never applied feedback successfully: I cannot face the mathematics for feedback for a SE stage and my attempts with balanced feedback (I do only differential, balanced amplifiers) have been a total failure.
This is a nuisance because I am not opposed to feedback on principle - only all the additional grief it causes.
I would love to use feedback to adjust the gain of my amplifiers; a good engineering solution and it would permit me a wider choice of valves. But I cannot do it so it's no-go area for me.
As for valve amplifiers being inherently unstable, I am not sure about this. It is true that wherever there is a combination of gain, capacitance and inductance it is reasonable to expect that there might be some trouble; but surely this applies to solid state amplifiers too? (I know less about solid state amplifiers than I so about feedback!).
7N7
As 7N7 has pointed out the Williamson amplifier has it's corrective poles within the forward path of the amplifier rather than as part of the 'feedback' side of the loop. In general this is a better way of applying the correction necessary to stabilise a feedback amplifier as it is designed to ensure that those parts of the signal that would trigger instability around the loop are not passed 'interstage'. It is not terribly fashionable in ss circles (nor often taught!) as a 'two port' and Op-amp model with external stabilisation poles approach is used for ease of modelling whilst applying the poles succesfully within the forward path requires real understanding of how each stage works...
Valve amplifiers are not inherently unstable - in practice some are and some aren't. Much like ss amplifiers. Nyquist applies to all amplifiers, and not just electronic ones... I've had some 'fun' stabilising data systems that had positive feedback applied in an unstable manner...
On gain, when designing an amplifier I like to engineer each stages gain to fit into a gain scheme for the system it is to work in using local feedback (or stage design) to acheive the desired objective. Occasionally this might extend to two stages with an interstage transformer than itself contributes gain or loses it... The point is that irrespective of the inherent gain of the valve - the stage has a particular gain objective that it must hit (to be succesful!). Local feedback is an important part of this engineering and I bet 7N7 uses it!
I know how to use feedback and I don't like it trans-stage in audio amplifiers. Ever!
ciao
James
Valve amplifiers are not inherently unstable - in practice some are and some aren't. Much like ss amplifiers. Nyquist applies to all amplifiers, and not just electronic ones... I've had some 'fun' stabilising data systems that had positive feedback applied in an unstable manner...
On gain, when designing an amplifier I like to engineer each stages gain to fit into a gain scheme for the system it is to work in using local feedback (or stage design) to acheive the desired objective. Occasionally this might extend to two stages with an interstage transformer than itself contributes gain or loses it... The point is that irrespective of the inherent gain of the valve - the stage has a particular gain objective that it must hit (to be succesful!). Local feedback is an important part of this engineering and I bet 7N7 uses it!
I know how to use feedback and I don't like it trans-stage in audio amplifiers. Ever!
ciao
James
Gabevee said:
SS amps don't have a great big reactive output transformer in their feedback loops, so depending on the circuit, it can be very easy to make a tube amp oscillate.
The Williamson amp when first tested and published used the best components available at the time, like Partridge OPTs and KT66's and was still marginal in stability from all the reports I've read. So making a clone with a lesser quality of parts, particularly the OPT may make for much worse results, stability and SQ wise. I've seen quite a few Williamson type designs burst into oscillation (often intermittently, following some signal conditions) on the bench.
The Crowhurst papers available from the aX site explains feedback functioning as it applies to tubeamps in much more detail.
James D. said:
Er.... not knowingly.
Of course my cathode followers work at 100% feedback. The only time I have successfully used a type of feedback (previously I was referring of course, to global feedback) was on my 813 p-p amplifier.
Here, I connected the wiper of the hum balance pot on each 813 (i.e. the "cathode" terminal) to the ends of the secondary of the output transformer, grounding the centre-tap of the secondaries.
This was very effective and improved the mid-range performance, which suggests that it lowered the output resistance a little.
7N7
Transformers are only as reactive as their internal capacitances/inductances/resistances will allow. Since they are usuallt large in those bulky masses of iron called output transformers... frequency of oscillation will be low... if any. As such the surrounding circuitry should dampen any spurious oscillations. I think that the natural roll off of the transformer would prevent oscillation at any significant frequencies.
A similar philosophy is done with the so-called internal compensation of op-amps. Basically, they're designed to give an open loop gain of a million, from 0Hz to 1 or 2MHz. However, phase shift causes oscillation with NFB. So a capacitor is placed shunted somewhere in the circuit so overall gain lowers something like 6dB per octave until it reaches a unity gain at 1mHz. This way any phase shift will be equally met with reduced gain. This in turn reduces the chance of positive feedback along the NFB loop.
What tube power amp has an open loop gain of 100, let alone a million?
The only time I have every had a tube amp become unstable was when I removed the NFB. Then a phenomenon called motorboating occurred.
This is actually due to power supply deficiencies. And no wonder, since the philosophy in tube amp power supplies is to use 10 or 20 µF. One doesn't seem to realize that audio can get into the power supply, so one needs to choose capacitance such that it filters out audio information also, as low as the amplifier can provide. So a capacitor value of 80 to 100 should be used to prevent motorboating without NFB. This is especially true for stereo amps. This also prevents crosstalk.
As for Williamson amps... I did not read that in the original article (written by D.T.N. Williamson and David Hafler, if memory serves). Of course... that may be due to the fact that most people tend to cover over their failures. Yet, I cannot imagine that the manufacturers who designed the schematics I posted would continue to make amps based on an inherently unstable design. Just doesn't fly with me.
What anode feedback??? The 150 ohm resistors I am seeing going from the anode to the screens on the KT66's are to make them "triodes", no?
So OK, technically that is feedback. But I would be willling to wager that the tube in its inherent mode (as beam power tubes) would work just as well, but not sound quite the same.
Can you provide a link please?
Thanks!
Gabe
Gabevee said:
I was referring to the anodes of the input stage ( 2 x L63 or 1 6SN7).
The input anode has its load resistor (47k) by-passed by 4.7k plus 200pF.
Small power supply capacitors were mandatory in old valve amplifiers: first, large ones of sensible size and cost were not available; second, no rectifier valve can supply the ripple current that a big capacitor requires. The best was the GZ34 which could manage 60uF
7N7
7N7,
I just went over that article. I see what you mean there.
First off... it is somewhat of an op-amp inasmuch as overall voltage gain is somewhere on the order of 10,000 (the 2700 of the first two stages times the driver/power output). The first two stages (the 12AX7) is essentially a preamp (sensitivity of 28mV after NFB. Yikes! One can almost literally hook a magnetic phono cartridge directly to it! Well... save for the RIAA curve).
Most of us make power amps beginning where the 12AU7 is and apply NFB from the secondary to the 12AU7, or driver/inverter stage. As for the reasons the anode feedback is used... seems reasonable to me. I haven't done it myself. I have toyed with the idea, though. Looks to me like 6 of one and half dozen of the other.
As for the larger capacitance... I agree. But of course, I wouldn't put an 80-100µF cap right at the rectifier, but after the choke. Although, I seem to recall that in the back of the RCA tube manual c.1958 there is an amp that uses 80 µF caps in the rectifier circuit, both before and after the choke using a 5U4 (or was it a 5Y3?). So... it seems that it was available at least in the fifties.
Otherwise, that huge gain certainly would be unstable left open loop.
Also, I find it kind of amusing that he says that designing an amp with good response may lose the good qualities by use of NFB, but then goes on to design an amp with good response and low distortion before NFB.
My experience, FWIW, has been that if the design is good, NFB won't help much, if at all, except to reduce gain, increase bandwidth, and level the sweep response (making all frequencies equal in amplitude). But, I haven't gone to extremes with it either.
fdegrove,
Thanks for the link!
Gabe
I just went over that article. I see what you mean there.
First off... it is somewhat of an op-amp inasmuch as overall voltage gain is somewhere on the order of 10,000 (the 2700 of the first two stages times the driver/power output). The first two stages (the 12AX7) is essentially a preamp (sensitivity of 28mV after NFB. Yikes! One can almost literally hook a magnetic phono cartridge directly to it! Well... save for the RIAA curve).
Most of us make power amps beginning where the 12AU7 is and apply NFB from the secondary to the 12AU7, or driver/inverter stage. As for the reasons the anode feedback is used... seems reasonable to me. I haven't done it myself. I have toyed with the idea, though. Looks to me like 6 of one and half dozen of the other.
As for the larger capacitance... I agree. But of course, I wouldn't put an 80-100µF cap right at the rectifier, but after the choke. Although, I seem to recall that in the back of the RCA tube manual c.1958 there is an amp that uses 80 µF caps in the rectifier circuit, both before and after the choke using a 5U4 (or was it a 5Y3?). So... it seems that it was available at least in the fifties.
Otherwise, that huge gain certainly would be unstable left open loop.
Also, I find it kind of amusing that he says that designing an amp with good response may lose the good qualities by use of NFB, but then goes on to design an amp with good response and low distortion before NFB.
My experience, FWIW, has been that if the design is good, NFB won't help much, if at all, except to reduce gain, increase bandwidth, and level the sweep response (making all frequencies equal in amplitude). But, I haven't gone to extremes with it either.
fdegrove,
Thanks for the link!
Gabe
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.