Hi,
I've built a PP amp which sounds OK. It has a 6SL7 LTP splitter as the first stage, with a 6AU6 CCS in the tail. This is followed by 6SN7 differential driver and PP EL34 triode-strapped output stage.
There's no global feedback loop but I've used partial feedback (a) from the 6SN7 plates to the 6SL7 plates and (b) from the EL34 plates to the 6SN7 plates. My OPT is a locally made Indonesian product of unknown quality and doesn't have UL taps. This is why the EL34s are triode-strapped and why I'm not keen to use global feedback (because of the potential instability).
I'm thinking of trying the EL34s in class AB1 pentode connection, to try a different sound with more power. However, I'm aware that such a move will lower the damping factor considerably. In commercial designs, this is usually overcome by using global feedback but I'm wondering if partial fb can be sufficiently effective to give a reasonable damping factor?
I've built a PP amp which sounds OK. It has a 6SL7 LTP splitter as the first stage, with a 6AU6 CCS in the tail. This is followed by 6SN7 differential driver and PP EL34 triode-strapped output stage.
There's no global feedback loop but I've used partial feedback (a) from the 6SN7 plates to the 6SL7 plates and (b) from the EL34 plates to the 6SN7 plates. My OPT is a locally made Indonesian product of unknown quality and doesn't have UL taps. This is why the EL34s are triode-strapped and why I'm not keen to use global feedback (because of the potential instability).
I'm thinking of trying the EL34s in class AB1 pentode connection, to try a different sound with more power. However, I'm aware that such a move will lower the damping factor considerably. In commercial designs, this is usually overcome by using global feedback but I'm wondering if partial fb can be sufficiently effective to give a reasonable damping factor?
I don't believe local FB does anything but straighten out linearity. Global FB is what lowers output Z and hence, DF goes up.
Your option may be to change the output transformers. Can you get Hammond's where you are? They usually have the best overall bang-for-the-buck.
FB is not my strong point in amps though, so if more knowledgeable folks can help....
Your option may be to change the output transformers. Can you get Hammond's where you are? They usually have the best overall bang-for-the-buck.
FB is not my strong point in amps though, so if more knowledgeable folks can help....
Theoretically yes but in practice no.
Voltage feedback over the output tubes i.e from anod to grid lowers the output impedance as much as the gain is decreased but the problem is that drive voltage increase the same amount, therefore it is not practical to use a high amount of local feedback. With global feedback mainly the gain of the first stage is affected and the input voltage increases but it doesn't matter as the first stage can handle higher input voltage due to the feedback.
The amount of local feedback is thus limited by the highest input voltage that can be allowed to that stage, (what the driver can give).
The only thing that limits global feedback is stability and that can be handled by phase correcting networks of different kinds.
I would advice you to learn about feedback stability and apply global feedback if you feel you need a high DF, I personally enjoy the maths involved and applying that in realkity, I measure phas and gain up to 20MHz and design my feedback networks after that.
Regards Hans
If the transformer were perfect, shunt FB (plate-grid as you describe) would control things effectively. However, loss across the transformer, including phase change, frequency response and DCR, is independent of that. It's outside the loop.
If your transformer sounds good, no doubt it will be suitable for NFB. I've wrapped some crappy transformers with NFB before... just don't expect much below 50Hz if your transformer has low inductance.
BTW, I'd remove the local FB to provide more global gain, so you can correct the output stage errors without needing a massive input signal.
Tim
If your transformer sounds good, no doubt it will be suitable for NFB. I've wrapped some crappy transformers with NFB before... just don't expect much below 50Hz if your transformer has low inductance.
BTW, I'd remove the local FB to provide more global gain, so you can correct the output stage errors without needing a massive input signal.
Tim
It's a juggling act. You need quite a lot of feedback to lower output resistance to the levels required by most modern loudspeakers, and that usually means global feedback. However, cathode feedback at the output valves can be surprisingly effective at reducing distortion (more effective than theory would suggest) - but you need plenty of gain to begin with, so it works best with pentodes and beam tetrodes.
If you can go to an active crossover, that removes the inductor in series with the bass driver, typically removing an Ohm of series resistance, reducing the necessity to strive for really low output resistance in the amplifier...
If you can go to an active crossover, that removes the inductor in series with the bass driver, typically removing an Ohm of series resistance, reducing the necessity to strive for really low output resistance in the amplifier...
Cathode feedback is indeed a good idea for reducing distortion but the drawback is that it increases output impedance so DF in this case would go down, quite the opposite of what was asked for.
It must be a very bad transformer to affect the damping factor of the amp in any major way, (but maybe in the case with an output transformer with very high winding resistance but then you would have more serious problems than affected DF) in any case local feedback is not realistic as a method to increase DF, 20dB of global feedback is not that difficult to apply and would increase DF 10 times, that amount of local feedback is clearly unrealistic.
Regards Hans
EC8010, You are quite right but I didn't realise that was what you meant by " cathode feedback at the output valves can be surprisingly effective at reducing distortion" Sorry if I misunderstood what you intended to describe, for me simply cathode feedback is cathode degeneration by unbypassed cathode resistors which still have a surprisingly good effect on linearity.
Regards Hans
Regards Hans
Konnichiwa,
You can use NFB that excludes the output transformer and that excludes the first stage quite effectively. The one thing to remmeber is that any DCR related losses of the Transformer are now outside the conrol of the feedback loop.
My modifications for the Shanling SP-80 include to use a NFB loop from the Anodes of the output stage (EL-34/KT-88/6550 switchable triode/ultralinear/pentode) to the grids of the driver stage (6SN7) only. There is no NFB from the transformer secondary and the first stage (6SL7) is also outside the NFB loop.
I actually DC couple the first and second stage as well, the Feedback loop is DC coupled so between the inputs I have only one pair of coupling capacitors to muck up things and a pair of very low capacitance PTFE/Silver (from micro coax) adjustable capacitors to trim the squarewaveresponse.
The output of the mostly DC coupled balanced amplifier then goes to the output transformer which matches the Speaker to the Amp.
I apply around 6 - 12db MORE NFB that way than in the original Amplifier without risking problems and thus I drive the output transformer from a very low source impedance (which helps keeping transformer distortion down) and attain a similar output impedance as in the original Amp with a long feedback loop.
Of course, there are many other changes too, so I cannot attribute all the sonic changes to the shorter NFB loop.
The sonics are very good BTW.
Well, if you compare the frequency response for the same transformer driven from 10K (PP Pentode pair), 2K (PP Triode Pair) and 160 R (PP Pentode pair with shuntfeedback) you may find that the HF and LF response of the transformer are much improved for the shuntfeedback circuit over the others.... ;-)
The DCR remains outside the loop of course.
Of course, in the other two circuits we can still apply added NFb, which we cannot do in our circuit which already has internal NFB without adding more gain, so the followup NFB can correct some of the problems introduced by driving the OPT from a high source impedance.
And as a further bonus you now have much less phaseshift and other BS inside the feedback loop, so you can apply tons more of feedback with comparably small sonic penalties.
In fact, I'd go as far as saying that to apply loop feedback from the secondary of an output transformer must be counted as monumentally stupid, compared to leaving the output transformer outside the loop....
Sayonara
ray_moth said:
You can use NFB that excludes the output transformer and that excludes the first stage quite effectively. The one thing to remmeber is that any DCR related losses of the Transformer are now outside the conrol of the feedback loop.
My modifications for the Shanling SP-80 include to use a NFB loop from the Anodes of the output stage (EL-34/KT-88/6550 switchable triode/ultralinear/pentode) to the grids of the driver stage (6SN7) only. There is no NFB from the transformer secondary and the first stage (6SL7) is also outside the NFB loop.
I actually DC couple the first and second stage as well, the Feedback loop is DC coupled so between the inputs I have only one pair of coupling capacitors to muck up things and a pair of very low capacitance PTFE/Silver (from micro coax) adjustable capacitors to trim the squarewaveresponse.
The output of the mostly DC coupled balanced amplifier then goes to the output transformer which matches the Speaker to the Amp.
I apply around 6 - 12db MORE NFB that way than in the original Amplifier without risking problems and thus I drive the output transformer from a very low source impedance (which helps keeping transformer distortion down) and attain a similar output impedance as in the original Amp with a long feedback loop.
Of course, there are many other changes too, so I cannot attribute all the sonic changes to the shorter NFB loop.
The sonics are very good BTW.
Sch3mat1c said:
Well, if you compare the frequency response for the same transformer driven from 10K (PP Pentode pair), 2K (PP Triode Pair) and 160 R (PP Pentode pair with shuntfeedback) you may find that the HF and LF response of the transformer are much improved for the shuntfeedback circuit over the others.... ;-)
The DCR remains outside the loop of course.
Of course, in the other two circuits we can still apply added NFb, which we cannot do in our circuit which already has internal NFB without adding more gain, so the followup NFB can correct some of the problems introduced by driving the OPT from a high source impedance.
And as a further bonus you now have much less phaseshift and other BS inside the feedback loop, so you can apply tons more of feedback with comparably small sonic penalties.
In fact, I'd go as far as saying that to apply loop feedback from the secondary of an output transformer must be counted as monumentally stupid, compared to leaving the output transformer outside the loop....
Sayonara
You have just declared that most of the commercially designed tube amps that have existed are designed by incompetent and stupid designers, do you really mean that?
There are obvious advantages, (and disadvantages) of including the OT inside the loop and most designers have made the choice to keep it inside the loop where feedback can more effectively reduce output impedance, straigthen out frequency response and lower distortion.
To mix local and global feedback has been done before, (e.g Knapp, Kiebert and others) and have many advantages but I have never seen any classic design where the OT has been left outside the loop.
Just because there exist many tube amps, (and SS amps) that has been designed by people who doesn't understand the criterias for applying global feedback successfully doesn't really mean that there is something wrong with the basic concept.
I don't doubt for a moment that you could improve the performance of the Shanling SP-80 by your methods but i am equally certain that a mix of local feedback as you applied + global feedback including first stage and OT could improve performance even more.
Regards Hans
Re: Re: Can partial feedback be as effective as global fb for decent damping factor?
In many cases, especially where high-quality transformers are used, this may hold true. However, where there are anomolies in the circuit due to the type of tube used or due to supply, the transformers are less than optimal, including the transformer in the loop can be of great benefit.
It can also be beneficial to include it in the loop to tailor sound, especially in guitar amps.
Unless you have reasons differing, please enlighten us as to your view.
Kuei Yang Wang said:
In many cases, especially where high-quality transformers are used, this may hold true. However, where there are anomolies in the circuit due to the type of tube used or due to supply, the transformers are less than optimal, including the transformer in the loop can be of great benefit.
It can also be beneficial to include it in the loop to tailor sound, especially in guitar amps.
Unless you have reasons differing, please enlighten us as to your view.
My reservation about including the transformer in the fb loop is that I'm not sure how good my pair of non-brand Indonesian output transformers are. Maybe I'll get some instability if I include the OPT in the loop, making things worse insead of better.
My OPTs have quite big cores, so I guess their inductance must be reasonably high. They have 8 ohm secondaries with 4 ohm taps and, according to the turns ratio, they give 3500K plate-to-plate load. This is in accordance with data I've seen for Class AB1 pentode operation of EL34s. However, since I'm using triode-strapped EL34s, I have my 8 ohm speakers connected to the 4 ohm taps, to get plate loads of 7k p-p.
My OPTs have quite big cores, so I guess their inductance must be reasonably high. They have 8 ohm secondaries with 4 ohm taps and, according to the turns ratio, they give 3500K plate-to-plate load. This is in accordance with data I've seen for Class AB1 pentode operation of EL34s. However, since I'm using triode-strapped EL34s, I have my 8 ohm speakers connected to the 4 ohm taps, to get plate loads of 7k p-p.
Konnichiwa,
Your logic is inescapable and completely agrees with Sturgons law (as expected).
For "intelligent" use of NFB look at the HK Citation II Amp or at certain ARC Amplifiers, as well as variuous of Norman Crowhurst's, Kieberts and similar design, many TAB & Klangfilm Amplifiers and a number of US Pro-Audio Valve Amp's from WE, Altec and Co. Some do include a "long" loop but all have most NFB in shorter loops.
I suspect that merely replicated the common schematics from the Valve Amplifier Makers and a few People published, rather than from a concious choice. Allmost all true "high performance" amplifiers had only very low amounts of NFB in the Outer loop. Commercailly sold amplifiers tended to rely just on outer loop as it made the amplifiers simpler and cheaper to make, but quality invariably was compromised).
I have seen many, but the best known are the Siemens/TAB V69 and the WE91A/B.
You CAN apply NFB around the OPT, but most OPT's simply lack the bandwidth to apply much NFB that way without severe sonic penalties, so best leave the Transformer out of the loop and if neccesary compensate the LF and HF rolloff before the transformer to get the widest bandwidth available.
You may be certain, but in my case you are certainly wrong (because I tried that too).
In fact, I tried partial cathode feedback with inner & outer Loop (NDFG) and just inner & outer loop (NDFG) and Inner lopp and partial cathode feedback (NG) and found the best choice was to optimise the inner loop and to pre-equalise the HF rolloff of the OPT to get a response flat to 50KHz, very nice squarewaves, practically unconditional stability into capacitive loads and on to of that a sound that is rather good (and the best by far for the Amplifier).
The problem with putting the Transformer inside ANY feedback loop is that in most transformers we have too many problems on the band edges and they are made worse the higher the source imepdance. Thus by putting all available NFB ahead of the Transformer you minimise it's non-ideal behaviour and gain a huge load of phasemargin to keep things stable, both traits highly relevant in the real world.
You can try, just look at square waves with a range of capacitive loads. I'll suggest you have a good look at the HK Citation II Poweramp. It has many good features and Ideas you can crib off.
Sayonara
tubetvr said:
Your logic is inescapable and completely agrees with Sturgons law (as expected).
For "intelligent" use of NFB look at the HK Citation II Amp or at certain ARC Amplifiers, as well as variuous of Norman Crowhurst's, Kieberts and similar design, many TAB & Klangfilm Amplifiers and a number of US Pro-Audio Valve Amp's from WE, Altec and Co. Some do include a "long" loop but all have most NFB in shorter loops.
tubetvr said:
I suspect that merely replicated the common schematics from the Valve Amplifier Makers and a few People published, rather than from a concious choice. Allmost all true "high performance" amplifiers had only very low amounts of NFB in the Outer loop. Commercailly sold amplifiers tended to rely just on outer loop as it made the amplifiers simpler and cheaper to make, but quality invariably was compromised).
tubetvr said:
I have seen many, but the best known are the Siemens/TAB V69 and the WE91A/B.
tubetvr said:
You CAN apply NFB around the OPT, but most OPT's simply lack the bandwidth to apply much NFB that way without severe sonic penalties, so best leave the Transformer out of the loop and if neccesary compensate the LF and HF rolloff before the transformer to get the widest bandwidth available.
tubetvr said:
You may be certain, but in my case you are certainly wrong (because I tried that too).
In fact, I tried partial cathode feedback with inner & outer Loop (NDFG) and just inner & outer loop (NDFG) and Inner lopp and partial cathode feedback (NG) and found the best choice was to optimise the inner loop and to pre-equalise the HF rolloff of the OPT to get a response flat to 50KHz, very nice squarewaves, practically unconditional stability into capacitive loads and on to of that a sound that is rather good (and the best by far for the Amplifier).
The problem with putting the Transformer inside ANY feedback loop is that in most transformers we have too many problems on the band edges and they are made worse the higher the source imepdance. Thus by putting all available NFB ahead of the Transformer you minimise it's non-ideal behaviour and gain a huge load of phasemargin to keep things stable, both traits highly relevant in the real world.
ray_moth said:
You can try, just look at square waves with a range of capacitive loads. I'll suggest you have a good look at the HK Citation II Poweramp. It has many good features and Ideas you can crib off.
Sayonara
Re: Re: Re: Can partial feedback be as effective as global fb for decent damping factor?
Konnichiwa,
I care to disagree, for the case of Amplifiers intended for high fidelity reproduction.
Guitar Amplifers are intended to PRODUCE sound. That means all is "fair game" for use.
High Fidelity Amplifiers are intended to REPRODUCE (or better re-create) what was produced elsewhere. Hence while much remains "fair game" certain excessive changes to the reproduced recording are not acceptable.
Unless I had reasons differing, do you think I would take the stance I have and implement things the way I do?
I think we need to be clear as to WHY we use Negative feedback and what we want to achieve by using it. Once it is clear WHY we are doing WHAT it becomes easy to see that there is one optimal solution to the problems and many suboptimal ones.
If we intend to reduce the non-ideal behaviour of the output transformer we can apply a number of methodes.
If we intend to reduce the crossove distortion in class AB amplifiers we can apply a number of methodes.
If we intend to increase the damping factor of an amplifier we can use a number of methodes.
If we intend to reduce general harmonic and intermodulation distortion we can use a number of methodes.
And so on.
Sayonara
Konnichiwa,
Geek said:
I care to disagree, for the case of Amplifiers intended for high fidelity reproduction.
Geek said:
Guitar Amplifers are intended to PRODUCE sound. That means all is "fair game" for use.
High Fidelity Amplifiers are intended to REPRODUCE (or better re-create) what was produced elsewhere. Hence while much remains "fair game" certain excessive changes to the reproduced recording are not acceptable.
Geek said:
Unless I had reasons differing, do you think I would take the stance I have and implement things the way I do?
I think we need to be clear as to WHY we use Negative feedback and what we want to achieve by using it. Once it is clear WHY we are doing WHAT it becomes easy to see that there is one optimal solution to the problems and many suboptimal ones.
If we intend to reduce the non-ideal behaviour of the output transformer we can apply a number of methodes.
If we intend to reduce the crossove distortion in class AB amplifiers we can apply a number of methodes.
If we intend to increase the damping factor of an amplifier we can use a number of methodes.
If we intend to reduce general harmonic and intermodulation distortion we can use a number of methodes.
And so on.
Sayonara
ray_moth said:
Oh noes, big whoop. Instability means either LF rise (low inductance) or HF rise (high leakage inductance). If these are too great, it will motorboat or oscillate, respectively. Capacitive loads will aggravate the former to no end, and can even affect the other, depending on which is greater I guess.
Think of it this way: you have a scale of open-loop amplification vs. frequency. At the left end, gain is low because series capacitors and parallel inductors reduce gain. In the middle (maybe 100Hz to 10kHz), gain equals the calculated value. On the right, gain increases slightly due to leakage inductance becoming parallel-resonant with the primary's winding capacitance, this might peak at 10kHz for truely bad OPTs to 50kHz for a normal type to 300kHz for one utterly drowned in overengineering. Above this, gain drops at a normal rate (-12dB/octave or so).
Now apply dry, uncorrected NFB. The graph of gain vs. frequency has been cut down - it still has about the same shape, but the +/-3dB ends are farther out. Think of the previous situation, then draw a horizontal line between the extremes of gain. You've just cut off that much above the line. In reality it's a bit wider, depending on how much phase shift there is. It's also rounded, as the gain drops off as a percentage - but since overall gain is lower, what used to be -6dB is now only -1.5dB. And now you have to tune it for that reason: phase shifts want to oscillate, so you have to either a. correct them, or b. restrict bandwidth to exclude those nasty areas. This can include reducing coupling cap values, adding R+C snubbers (to shunt HF), and so forth. A big help comes from putting a cap across the NFB resistor - this "speeds up" the feedback, extending HF response a bit. This is tuned with squarewave: too little and you get an ugly peak (undercorrection). Too much and the leading edge is rounded and slow (speeding up NEGATIVE feedback SLOWS DOWN the amplifier). You may also notice ringing on the square wave, where the sharp rise and fall creates harmonics which excite HF resonances in the OPT, etc. As mentioned, HF instability causes a rise - this manifests itself as ringing. An R+C snubber (zobel) at the OPT and other stages cancels this nicely.
What NFB does is extend the "flat" range, by reducing gain with itself. But what happens at frequencies where the gain is already lower? Final gain is also lower. It isn't as low as before, so you get a flattened response, as mentioned. However, this also means NFB is lower -- so distortion is higher. At some frequency outside of the band of interest, NFB will be near zero and the distortion level will be the same as before NFB.
To sum up finally in response to your question... you can't hear instability unless a. it is really ruining your frequency response, b. ruining distortion at LF (since HF distortion is ultrasonic..), or at worst, c. oscillating. NFB is a scope-and-signal-generator thing, not a burn-for-half-an-hour-testing-the-sound thing.
Tim
IT transformer
Originally posted by tubetvr
There are obvious advantages, (and disadvantages) of including the OT inside the loop and most designers have made the choice to keep it inside the loop where feedback can more effectively reduce output impedance, straigthen out frequency response and lower distortion.
What about if cct. includes an IT, then all gets more complicated because you include two irons in the NFB,deal with two phase shifts,more maths is involved and the end result is more uncertain.So finally everything depends on the cct.Opinions?
Regards,
Yugovitz
Originally posted by tubetvr
There are obvious advantages, (and disadvantages) of including the OT inside the loop and most designers have made the choice to keep it inside the loop where feedback can more effectively reduce output impedance, straigthen out frequency response and lower distortion.
What about if cct. includes an IT, then all gets more complicated because you include two irons in the NFB,deal with two phase shifts,more maths is involved and the end result is more uncertain.So finally everything depends on the cct.Opinions?
Regards,
Yugovitz
Konnichiwa,
Possibly, but still, is it a good idea?
Okay, lets have a look at an amplifier and work our way backwards.
We have our speaker. Be it sensible or not, modern "High Fidelity" speakers expect a voltage drive and one would likely argue that having an amplifier with low general THD, and IMD is not a "bad thing" (It's not reliably a "good thing", but as basic design aim it will not do harm unless we screw up something).
Okay, next is our Output transformer.
What do we know about how an output transformer behaves?
We can charaterise it very much as 1st order bandpass filter in the semi-ideal sense, at high frequencies various resonances can screw us up. We also know that distortion happens, mainly at low frequencies though. Also, the primary inductance of the transformer is very much level and frequency dependent.
Now with a zero ohm source impedance most transformers will show this semi idealised behaviour. The HF rolloff is determined by the leakage inductance compared to the nominal primary impedance and there is theoretically no LF rolloff (practically the limited core size reduces coupling at very low frequencies). The distortion is very low even though it will eventually rise at low frequencies and for high(ish) levels.
What happens if we drive the transformer from a significant source impedance?
I can only take a few measured examples. I once measured a certain line transformer, with a 33:1 stepdown and 10K nominal primary impedance and essentially unloaded secondary.
Driven from the 50 Ohm output of the Generator showed a virtually flat response from << 4Hz to above 3MHz (where the generator ended).
Now changing the Source impedance to 1K dropped the bandwidth notably, giving only around 4Hz - 150KHz. Going up to 10K source impedance and we got 8Hz-60KHz -3db.
Equally, on a different transformer I measured the distortion to go up 6db for every 1:10 increase in impedance at fairly low drive impedance levels. I would expect that if the drive impedance goes up further than I did (I stopped at a 1:20 ratio between source and nominal impedance) the distortion would go up notably.
So, from the above, no matter bad the transformer is, it will show less distortion and a wider bandwidth with a given load if driven from a low to near zero Impedance. The higher the drive impedance the worse things become.
Next our output stage. For expediency we assume a PP Output Stage in Class AB (a pure Class A SE or PP Amp will have fewer issues as will become clear).
We know that in general the distortion will go up the lower the load impedance is. For pentode stages the reverse actually holds true, but for triode stages and any partial feedback circuits the above holds true.
What happens when the output stage transits from Class A to Class B? Our load impedance seen by the valve remaining active drops to 1/2 of that seen during Class A operation. At the same time the effective source impedance driving our transformer goes up by the ratio.
This introduce a large shift in distortion signature for output stage and output transformer between Class A output conditions and Class B output conditions. The difference is also sonically significant and one of the reasons why Class AB PP Amplifiers often seem to change sonic charater disproprtionatly much between low and high output levels.
Working our way forward from the output stage we now require a Drive Voltage to the grids of the output valves which should be free from distortion (ideally), now a significantly "good" behaviour driverstage is in most cases a trivial issue. Problems only arise if we use unity coupled (MacIntosh) Output stages and the like, which we shall ignore in the context. So let's assume a linear driver stage to the output valve grids and if required enough excess gain to apply as much NFB as we require.
Looking at the above the question places itself, are we better of including the output transformer in the feedback loop or not?
If we drive the output transformer from a very low source impedance any of it's non-ideal properties will be strongly attenuated and hence there will be very little left to require any correction.
The key source of problems and like the one to swamp any and all transformer problems REMAINING if we drive the output transformer from low impedance source is clearly the output stage.
It will have under open loop conditions a high output impedance which will CAUSE problems to appear in the transformer and it's distortion behaviour when dropping into class B is not nice.
So, if we apply our feedback ONLY around the output stage excluding the OPT we have minimal issues with phasemargin and we make a source for the output transformer to be driven from that is low impedance and shows only a small degree of signal level dependent impedance modulation.
If we apply our feedback to include the output transformer we find that at the edges of the passband our gain used to correct the output stages nonlinearity is rolled off. At the same time the high source impedance adds a lot of transformer distortion in that band and the transformer limits severely the amount of feedback we can apply safely without exceeding the phase margin safety.
So, instead of insulating each problem and solving it as it occurs perfably by the most approriate methode (which is a short feedback loop between Output Stage and Driver IMHO) we have maximised each problem to appear on the secondary of the output transformer and after having royally screwed up the signal we loop some negative feedback in to see if can somehow pull out the lumped mess and straighten it again.
Am I the only one whom that strikes as monumentally stupid?
Sayonara
Yvesm said:
Possibly, but still, is it a good idea?
Okay, lets have a look at an amplifier and work our way backwards.
We have our speaker. Be it sensible or not, modern "High Fidelity" speakers expect a voltage drive and one would likely argue that having an amplifier with low general THD, and IMD is not a "bad thing" (It's not reliably a "good thing", but as basic design aim it will not do harm unless we screw up something).
Okay, next is our Output transformer.
What do we know about how an output transformer behaves?
We can charaterise it very much as 1st order bandpass filter in the semi-ideal sense, at high frequencies various resonances can screw us up. We also know that distortion happens, mainly at low frequencies though. Also, the primary inductance of the transformer is very much level and frequency dependent.
Now with a zero ohm source impedance most transformers will show this semi idealised behaviour. The HF rolloff is determined by the leakage inductance compared to the nominal primary impedance and there is theoretically no LF rolloff (practically the limited core size reduces coupling at very low frequencies). The distortion is very low even though it will eventually rise at low frequencies and for high(ish) levels.
What happens if we drive the transformer from a significant source impedance?
I can only take a few measured examples. I once measured a certain line transformer, with a 33:1 stepdown and 10K nominal primary impedance and essentially unloaded secondary.
Driven from the 50 Ohm output of the Generator showed a virtually flat response from << 4Hz to above 3MHz (where the generator ended).
Now changing the Source impedance to 1K dropped the bandwidth notably, giving only around 4Hz - 150KHz. Going up to 10K source impedance and we got 8Hz-60KHz -3db.
Equally, on a different transformer I measured the distortion to go up 6db for every 1:10 increase in impedance at fairly low drive impedance levels. I would expect that if the drive impedance goes up further than I did (I stopped at a 1:20 ratio between source and nominal impedance) the distortion would go up notably.
So, from the above, no matter bad the transformer is, it will show less distortion and a wider bandwidth with a given load if driven from a low to near zero Impedance. The higher the drive impedance the worse things become.
Next our output stage. For expediency we assume a PP Output Stage in Class AB (a pure Class A SE or PP Amp will have fewer issues as will become clear).
We know that in general the distortion will go up the lower the load impedance is. For pentode stages the reverse actually holds true, but for triode stages and any partial feedback circuits the above holds true.
What happens when the output stage transits from Class A to Class B? Our load impedance seen by the valve remaining active drops to 1/2 of that seen during Class A operation. At the same time the effective source impedance driving our transformer goes up by the ratio.
This introduce a large shift in distortion signature for output stage and output transformer between Class A output conditions and Class B output conditions. The difference is also sonically significant and one of the reasons why Class AB PP Amplifiers often seem to change sonic charater disproprtionatly much between low and high output levels.
Working our way forward from the output stage we now require a Drive Voltage to the grids of the output valves which should be free from distortion (ideally), now a significantly "good" behaviour driverstage is in most cases a trivial issue. Problems only arise if we use unity coupled (MacIntosh) Output stages and the like, which we shall ignore in the context. So let's assume a linear driver stage to the output valve grids and if required enough excess gain to apply as much NFB as we require.
Looking at the above the question places itself, are we better of including the output transformer in the feedback loop or not?
If we drive the output transformer from a very low source impedance any of it's non-ideal properties will be strongly attenuated and hence there will be very little left to require any correction.
The key source of problems and like the one to swamp any and all transformer problems REMAINING if we drive the output transformer from low impedance source is clearly the output stage.
It will have under open loop conditions a high output impedance which will CAUSE problems to appear in the transformer and it's distortion behaviour when dropping into class B is not nice.
So, if we apply our feedback ONLY around the output stage excluding the OPT we have minimal issues with phasemargin and we make a source for the output transformer to be driven from that is low impedance and shows only a small degree of signal level dependent impedance modulation.
If we apply our feedback to include the output transformer we find that at the edges of the passband our gain used to correct the output stages nonlinearity is rolled off. At the same time the high source impedance adds a lot of transformer distortion in that band and the transformer limits severely the amount of feedback we can apply safely without exceeding the phase margin safety.
So, instead of insulating each problem and solving it as it occurs perfably by the most approriate methode (which is a short feedback loop between Output Stage and Driver IMHO) we have maximised each problem to appear on the secondary of the output transformer and after having royally screwed up the signal we loop some negative feedback in to see if can somehow pull out the lumped mess and straighten it again.
Am I the only one whom that strikes as monumentally stupid?
Sayonara
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.