"opamp" is a generic term
lots of people think "opamp" means "IC." this is not true at all. an opamp can be discrete or IC - it just refers to a particular circuit function. in theory you could call any amplifier with differential inputs, a single output, and open loop gain an opamp. so yeah, what we're talking about still applies to discrete designs.
lots of people think "opamp" means "IC." this is not true at all. an opamp can be discrete or IC - it just refers to a particular circuit function. in theory you could call any amplifier with differential inputs, a single output, and open loop gain an opamp. so yeah, what we're talking about still applies to discrete designs.
dorkus,
True, but most people usually mean op amp chips when they
say op amp with no further specification. It is also clear from
many posts in the thread that op amp chips are assumed in
these posts.
Anyway, you answer my question and verify my suggestion
that the problem should be inherent in all similar topologies,
whether discrete or integrated. The interesting point then is
that it also applies to most power amps, which are also
op amps topologically, although perhaps not considered as
such usually. It seems then it would perhaps be a good idea
to rethink how the feedback is done in power amps, and
make the whole amp inverting instead. Having an inverting
output on the preamp would settle things straight again.
True, but most people usually mean op amp chips when they
say op amp with no further specification. It is also clear from
many posts in the thread that op amp chips are assumed in
these posts.
Anyway, you answer my question and verify my suggestion
that the problem should be inherent in all similar topologies,
whether discrete or integrated. The interesting point then is
that it also applies to most power amps, which are also
op amps topologically, although perhaps not considered as
such usually. It seems then it would perhaps be a good idea
to rethink how the feedback is done in power amps, and
make the whole amp inverting instead. Having an inverting
output on the preamp would settle things straight again.
Invert convert
"It seems then it would perhaps be a good idea
to rethink how the feedback is done in power amps, and
make the whole amp inverting instead."
Yes! For example, John Curl's design of about 25 years ago.
http://marklev.com/marklev/JC3/JC3-from-Electronics-World.jpg
There is nothing new under the sun,
Fred
"It seems then it would perhaps be a good idea
to rethink how the feedback is done in power amps, and
make the whole amp inverting instead."
Yes! For example, John Curl's design of about 25 years ago.
http://marklev.com/marklev/JC3/JC3-from-Electronics-World.jpg
There is nothing new under the sun,
Fred
Interesting, but so it would have been quite surprising if
noone had tried it. Rather, it is perhaps surprising that it
is so uncommon. The question in the JC3 article remains
unanswered: Why didn't it catch on?
BTW was there an inverting preamp to go with it, or didn't
Curl care about the phase?
noone had tried it. Rather, it is perhaps surprising that it
is so uncommon. The question in the JC3 article remains
unanswered: Why didn't it catch on?
BTW was there an inverting preamp to go with it, or didn't
Curl care about the phase?
wow
what an interesting design! john curl certainly was ahead of his time. the double feedback loop is especially curious. is two really better than one? i can see how the shorter loop would sort of "preempt" the feedback of the global loop, thus reducing the amount of correction needed in the global loop and maybe improving the "speed" of the amp. i think today with "no global feedback" designs it is more common to just have smaller local loops though, e.g. around the output and driver stages, and no global loop, yes? and of course many people here (e.g. Pass fans) would argue that no feedback at all is better still.
p.s. christer, yes, most people think IC when they hear "opamp," but i'm not one of them. in terms of operational theory there is no distinction between IC and discrete... at least for the purposes of our argument. there are some factors that are specific to discrete vs. IC, e.g. thermal modulation, device matching, etc. but i don't think any of the issues we've raised deal with those.
what an interesting design! john curl certainly was ahead of his time. the double feedback loop is especially curious. is two really better than one? i can see how the shorter loop would sort of "preempt" the feedback of the global loop, thus reducing the amount of correction needed in the global loop and maybe improving the "speed" of the amp. i think today with "no global feedback" designs it is more common to just have smaller local loops though, e.g. around the output and driver stages, and no global loop, yes? and of course many people here (e.g. Pass fans) would argue that no feedback at all is better still.
p.s. christer, yes, most people think IC when they hear "opamp," but i'm not one of them.
in terms of operational theory there is no distinction between IC and discrete... at least for the purposes of our argument. there are some factors that are specific to discrete vs. IC, e.g. thermal modulation, device matching, etc. but i don't think any of the issues we've raised deal with those.? and of course many people here (e.g. Pass fans)
Well I guess they would would other than the fact the most of Mr. Pass's designs use feedback. In fact the Aleph X topology is a stated evolution sonicaly over the X amp series that did not use global feedback.
"and of course many people here (e.g. Pass fans) would argue that no feedback at all is better still."
Yes, and without a clue why. I use both feed back designs and non feedback designs without fearing any need for apologies for the particular choice taken.
Feedback is sort of like the weather and life is like a box of chocolates in that.
Everybody talks about the weather, but nobody does anything about it. - Mark Twain
Utinam logica falsa tuam philosophiam totam suffodiant,
Fred
P.S. The Aleph X is an inverting feedback topology as are the Zen series of designs.
Well I guess they would would other than the fact the most of Mr. Pass's designs use feedback. In fact the Aleph X topology is a stated evolution sonicaly over the X amp series that did not use global feedback.
"and of course many people here (e.g. Pass fans) would argue that no feedback at all is better still."
Yes, and without a clue why. I use both feed back designs and non feedback designs without fearing any need for apologies for the particular choice taken.
Feedback is sort of like the weather and life is like a box of chocolates in that.
Everybody talks about the weather, but nobody does anything about it. - Mark Twain
Utinam logica falsa tuam philosophiam totam suffodiant,
Fred
P.S. The Aleph X is an inverting feedback topology as are the Zen series of designs.
ML-2
"Rolv-Karsten R¿nningstad says:
Regarding JC-3 and ML-2 power amplifiers:
I do not think the JC-3 was ever released under that name. My personal guess is that John Curl left Levinson before the JC-3 project was finished. The chief engineer Tom Colangelo is mentioned as designer of the ML-2 I think, and I guess he took over the project and finished it, and the amplifier was Christened the ML-2.
Among other things changed on the ML-2 there is an active J505 current source for the input JFET differential stage. There is a balanced input, and the minus-input is buffered with a simple n-channel/p-channel JFET pair. Further the second stage is a cascode connection. The JC-3 shows four output transistors: The ML-2 has four 2N5884 and four 2N5886. The ML-2 also has an active power supply. The heat sinks closest to the front on each side carries two 2N5884 and two 2N5886 respectively."
http://marklev.com/marklev/JC3/index.html
"Rolv-Karsten R¿nningstad says:
Regarding JC-3 and ML-2 power amplifiers:
I do not think the JC-3 was ever released under that name. My personal guess is that John Curl left Levinson before the JC-3 project was finished. The chief engineer Tom Colangelo is mentioned as designer of the ML-2 I think, and I guess he took over the project and finished it, and the amplifier was Christened the ML-2.
Among other things changed on the ML-2 there is an active J505 current source for the input JFET differential stage. There is a balanced input, and the minus-input is buffered with a simple n-channel/p-channel JFET pair. Further the second stage is a cascode connection. The JC-3 shows four output transistors: The ML-2 has four 2N5884 and four 2N5886. The ML-2 also has an active power supply. The heat sinks closest to the front on each side carries two 2N5884 and two 2N5886 respectively."
http://marklev.com/marklev/JC3/index.html
My thoughts exactly. Have a look at this small portion of an Aleph 3 cct which in this part is similar to many other designs. Notice that the signal is applied to the left side fet gate and the feedback is applied to the left side fet source via the long tailed pair. This is because the signal and feedback are in-phase so they must be applied to legs of the fet that have opposite phase properties to each other so the signals will subtract.Christer said:
Now comes the important bit... Notice that to get to the left fet source the nfb has to go through the right side fet gate-source!!! The point is, the nfb signal at the gate of the right fet is a virtually perfect but scaled down version of the output signal, but the nfb signal that is supplied to the "inverting summing junction" of the left fet has suffered the indignities of passing through the right fet, and so any distortion added by this fet has made the nfb signal invalid because it is now not a perfect representation of what is coming out of the amplifier. The summing junction is making incorrect compensations because it is based on false imformation because there is an active (and not distortion-free) component in the feedback path.
The alternative is to have the left fet gate tied to ground, and feed the signal in through the bottom of the 220uF cap that is normally earthed. This makes the whole cct inverting just as in the opamp discussion, and what's more the summing junction is not now a semiconductor but rather the junction of two resistors, the 10k and the 1k. If they together are not several orders of magnitude more linear than a gate-source junction then I'll eat my hat.
Attachments
Maybe it is already said and have I missed it in this long discussion:
If you use the inverting topology for audio the op-amp MUST be unity gain stable! Why? Well you don’t now one forehand by what impedance it is driven and the driving impedance is part of the feedback circuit. Let’s say the input cable is disconnected: Then the closed loop gain is 1 or unity. If the op-amp is not unity gain stable the amp will for sure turn into an oscillator.
The LM3875 is only stable for closed loop gains of 20 dB and higher. When in inverted topology and without sufficient low driving impedance (particularly at high frequencies, just before the 0 dB open-loop crossing point) it will for sure oscillate and possibly damage your tweeters.
You can get rid of this by placing a sufficient low resistor at the input, according to the attached schematic, to keep closed loop gain above 20 dB under all input conditions.
If you use the inverting topology for audio the op-amp MUST be unity gain stable! Why? Well you don’t now one forehand by what impedance it is driven and the driving impedance is part of the feedback circuit. Let’s say the input cable is disconnected: Then the closed loop gain is 1 or unity. If the op-amp is not unity gain stable the amp will for sure turn into an oscillator.
The LM3875 is only stable for closed loop gains of 20 dB and higher. When in inverted topology and without sufficient low driving impedance (particularly at high frequencies, just before the 0 dB open-loop crossing point) it will for sure oscillate and possibly damage your tweeters.
You can get rid of this by placing a sufficient low resistor at the input, according to the attached schematic, to keep closed loop gain above 20 dB under all input conditions.
Attachments
Hat entree
Circlotron,
In non-inverting mode there is no "inverting summing junction" as such. You see in this case the term "summing junction" is really to do with summing current, not voltage. In the non-inverting differential amp the differential output current is related to the difference in gate voltages divided by the sum of the transconductances of each device. It is the summing of the transconductances that benefits distortion. In this mode is is not correct to think of one FET acting as an unwanted gain stage - both devices need to be considered in tandem. Even in the inverting configuration the differential FETs are still in circuit: the output current is related to the "summing junction" voltage divided by the sum of the transconductances. So perhaps the critical question is whether the "summing junction" voltage is equal to the (scaled) difference between input and output, and I think Pjotr's point applies to the answer.
BAM
Circlotron,
In non-inverting mode there is no "inverting summing junction" as such. You see in this case the term "summing junction" is really to do with summing current, not voltage. In the non-inverting differential amp the differential output current is related to the difference in gate voltages divided by the sum of the transconductances of each device. It is the summing of the transconductances that benefits distortion. In this mode is is not correct to think of one FET acting as an unwanted gain stage - both devices need to be considered in tandem. Even in the inverting configuration the differential FETs are still in circuit: the output current is related to the "summing junction" voltage divided by the sum of the transconductances. So perhaps the critical question is whether the "summing junction" voltage is equal to the (scaled) difference between input and output, and I think Pjotr's point applies to the answer.
BAM
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.