It is invariably the case that datasheets published by op-amp manufacturers are intended to help sell opamps. The "audio performance" claims appear to be added as after-thoughts to lure the audio designer.
IMO Mark Alexander completely fails to make a case for his design on the grounds of improved performance. He does not explain why VF topologies are necessarily worse than CF. He claims they are but does not make a convincing case, jumping from various conjectures, finally to the vague statement "clearly, a global VF scheme may not necessarily be the optimum choice for ultrahigh performance audio power amplifiers". Clearly? What is clear to me is that he wouldn't know an ultrahigh performance audio amplifier if it bit him.
He then offers a circuit that achieves high slew rate by placing the compensation capacitors, of 94pF combined value, at a point where there is 30mA available to charge them. BUT he achieves this by introducing no less than a high current op-amp, two pass transistors and two current mirrors into the loop.
If the compensation capacitor peak charging current is the issue he was trying to solve then there are many better ways, not least simply to have a FET LTP with 30mA tail current driving the typical 47pF miller cap.
But that wouldn't need an AD op-amp, of course.
IMO Mark Alexander completely fails to make a case for his design on the grounds of improved performance. He does not explain why VF topologies are necessarily worse than CF. He claims they are but does not make a convincing case, jumping from various conjectures, finally to the vague statement "clearly, a global VF scheme may not necessarily be the optimum choice for ultrahigh performance audio power amplifiers". Clearly? What is clear to me is that he wouldn't know an ultrahigh performance audio amplifier if it bit him.
He then offers a circuit that achieves high slew rate by placing the compensation capacitors, of 94pF combined value, at a point where there is 30mA available to charge them. BUT he achieves this by introducing no less than a high current op-amp, two pass transistors and two current mirrors into the loop.
If the compensation capacitor peak charging current is the issue he was trying to solve then there are many better ways, not least simply to have a FET LTP with 30mA tail current driving the typical 47pF miller cap.
But that wouldn't need an AD op-amp, of course.
peranders said:Christer and everybody else, I'll suggest that you take a peek at Walt Jung's book. Mike, I think you too should take a look.
Page 26 and 115
http://www.analog.com/library/analogDialogue/archives/39-05/Web_Ch1_final_R.pdf
Jung starts by saying:
This is manifest nonsense, and nullifies the subsequent analysis...
For a correct analysis of the so-called 'CF' arrangement see cherry...
http://www.diyaudio.com/forums/showthread.php?postid=422698#post422698
http://www.diyaudio.com/forums/showthread.php?postid=422698#post422698
I think I would have a more constructive discussion with Eliza
http://www.manifestation.com/neurotoys/eliza.php3
Oh wait, maybe I am already talking with Eliza but modified with an EE database instead of the original one?
http://www.manifestation.com/neurotoys/eliza.php3
Oh wait, maybe I am already talking with Eliza but modified with an EE database instead of the original one?
Current feedback settling time and thermal tails
Hi,
"Although current feedback amplifiers have very fast rise times, many data sheets will only show settling times to 0.1%, because of thermal settling tails- a major contributor to lack of settling precision." http://www.analog.com/library/analogDialogue/archives/30-3/ask.html
http://www.edn.com/archives/1994/101394/21df5.htm
Are thermal tails significant to audio other than infront of A/D chips ?
This paper :- http://www.utdallas.edu/~hellums/docs/JournalPapers/SmithCFBamp.pdf
on page 9 mentions that at the investigated frequency that the current feedback input stage had lower 3rd harmonic than VFA.
Hi,
"Although current feedback amplifiers have very fast rise times, many data sheets will only show settling times to 0.1%, because of thermal settling tails- a major contributor to lack of settling precision." http://www.analog.com/library/analogDialogue/archives/30-3/ask.html
http://www.edn.com/archives/1994/101394/21df5.htm
Are thermal tails significant to audio other than infront of A/D chips ?
This paper :- http://www.utdallas.edu/~hellums/docs/JournalPapers/SmithCFBamp.pdf
on page 9 mentions that at the investigated frequency that the current feedback input stage had lower 3rd harmonic than VFA.
Re: Current feedback settling time and thermal tails
This is plausible in principal because a properly balanced diff. stage suppresses even harmonics at the expense of odd ones....
Be that as it may, this is essentially academic, as the relative quantities of ALL harmonics are (or should be) negligible in well-designed front ends...
ash_dac said:
This is plausible in principal because a properly balanced diff. stage suppresses even harmonics at the expense of odd ones....
Be that as it may, this is essentially academic, as the relative quantities of ALL harmonics are (or should be) negligible in well-designed front ends...
4fun said:
I already did. I said that not even feedback itself exist, remember?
Still, feedback is a very useful concept.
I realize such abstract and, perhaps, philosophical claims can seem very confusing and pointless to many. However, my observation from many forum discussions is that many EEs, however clever and experienced, sometimes talk past each other and end up in unnecessay and even hostile debates because they are so stuck at thinking on the component level that they don't see the forest because of all the trees, and thus misunderstand each other. It is sometimes useful to step up a level or or two in abstraction to get a fresh viewpoint and see things in a broader context. Then you can go down to the component level again, hopefully somewhat enlighetened by new insights. I know this sounds strange and pointless to many and that most people never had much training in abstraction. I am computer scientist and we are by necessity trained to think on a more abstract level, since computer programs are so complex that they become totally unmanageble if one doesn't make abstractions. Ah yes, and then there is the training in logic too, and I even tought logic for about ten years to CS students. Same thing there. Many debates derail unecessarily because of logical errors in the arguments, which was actually what made me step into this thread from the start, but I obviously failed to get the message through there.
This doesn't make me a better EE than most of you. It doesn't even make me an EE. However, it does make me see things from a somewhat different point of view that most of you do, and it is almost always useful to have people from another discipline step in and provide their insights. Same thing with computer science. People with another background can sometimes see things that we computer scientists miss. Crossdisciplinary thinking is usually a good thing that everybody benefits from. At my department at university they made a deliberate move long ago to actively recruit a number of people with various backgrounds not in CS in order to get new insights and a broader perspective on things.
Christer said:
I already did. I said that not even feedback itself exist, remember?
Still, feedback is a very useful concept.
I realize such abstract and, perhaps, philosophical claims can seem very confusing and pointless to many. However, my observation from many forum discussions is that many EEs, however clever and experienced, sometimes talk past each other and end up in unnecessay and even hostile debates because they are so stuck at thinking on the component level that they don't see the forest because of all the trees, and thus misunderstand each other. It is sometimes useful to step up a level or or two in abstraction to get a fresh viewpoint and see things in a broader context. Then you can go down to the component level again, hopefully somewhat enlighetened by new insights. I know this sounds strange and pointless to many and that most people never had much training in abstraction. I am computer scientist and we are by necessity trained to think on a more abstract level, since computer programs are so complex that they become totally unmanageble if one doesn't make abstractions. Ah yes, and then there is the training in logic too, and I even tought logic for about ten years to CS students. Same thing there. Many debates derail unecessarily because of logical errors in the arguments, which was actually what made me step into this thread from the start, but I obviously failed to get the message through there.
This doesn't make me a better EE than most of you. It doesn't even make me an EE. However, it does make me see things from a somewhat different point of view that most of you do, and it is almost always useful to have people from another discipline step in and provide their insights. Same thing with computer science. People with another background can sometimes see things that we computer scientists miss. Crossdisciplinary thinking is usually a good thing that everybody benefits from. At my department at university they made a deliberate move long ago to actively recruit a number of people with various backgrounds not in CS in order to get new insights and a broader perspective on things.
Oh well, I just tried to lighten up the heated discussion a bit.
Ok, my humble abstract opinion on witch i think this discussion is about:
A current source is not entirely a current source but also to a lot lesser extent voltage source.
A voltage source is not entirely a voltage source but also to a lot lesser extent a current source.
Therefore both can described be their opposite nomenclature.
The same thing can be said by current feedback and voltage feedback.
4fun said:
I quite agree, as long as we are talking real implementations. It is not true for the ideal voltage/current sources we use in theory though, although we often augment them with further components to behave like physical implementations. In that case we can transform them into each other using the well known theorems.
Exactly what I said previously. Yet it is useful to distinguish the two theoretical extreme cases, and also the real implementations according to which of the theoretical models they are intended to approximate.
- Status
- Not open for further replies.