I have started to have a rethink on amplifier circuitry.
One of the issues is if it is possible to improve the open loop linearity if the OPS is current driven rather than driven by a V source.
Note that I don't mean current drive of the speaker, just the way the OPS is driven by the previous section.
Self briefly dismisses the idea, Ed Cherry[1] has an analysis of the different impacts of Beta (current) nonlinearity versus transconductance nonlinearity.
That implies different response to current source or V source so it looks relevant but I don't find it that clear.
Anyone here have any ideas on this?
David
[1] Feedback, Sensitivity and Stability of Audio Power Amplifiers, JAES, May 1982
Also some of his later work is relevant, references if anybody is interested.
One of the issues is if it is possible to improve the open loop linearity if the OPS is current driven rather than driven by a V source.
Note that I don't mean current drive of the speaker, just the way the OPS is driven by the previous section.
Self briefly dismisses the idea, Ed Cherry[1] has an analysis of the different impacts of Beta (current) nonlinearity versus transconductance nonlinearity.
That implies different response to current source or V source so it looks relevant but I don't find it that clear.
Anyone here have any ideas on this?
David
[1] Feedback, Sensitivity and Stability of Audio Power Amplifiers, JAES, May 1982
Also some of his later work is relevant, references if anybody is interested.
Well, in a classical topology amp, that's already more or less what takes place, at least for low frequencies, and if there isn't a local feedback: a traditional VAS is essentially a bipolar, high compliance current source, but I think what you envision is more something like common emitter OP devices driven by common emitter CCS's.
I have already made preliminary explorations of this idea, mainly because it can solve the bias/thermal compensations issues encountered in a classical OPS: the beta does vary with temperature, but it is a far cry from the exponential dependency of the collector current vs. compensation voltage mismatch.
What halted me is the beta non-linearity of usual power devices, but otherwise, it could be a worthwhile solution
I have already made preliminary explorations of this idea, mainly because it can solve the bias/thermal compensations issues encountered in a classical OPS: the beta does vary with temperature, but it is a far cry from the exponential dependency of the collector current vs. compensation voltage mismatch.
What halted me is the beta non-linearity of usual power devices, but otherwise, it could be a worthwhile solution
Current drive
Hi David,
I started thinking in this direction as well.
It's already more than a year since I'm using current drive for the front-end (voltage amplification) designs.
Attached are a few pictures to explain the approach.
The biggest source of non-linearity is transconductance stage at the input - although there may be some better implementations - trying to keep it simple here. All the further operations with current are excellent, including the transimpedance stage - simple linear conversion, subtracting the current through a single resistor.
The whole thing demonstrates very good phase response, requiring very light and simple compensation (a few pF lead and 10-20pF shunt at the output), as well as rather low open loop distortion levels.
I've got many working design options now - driving the output CCSs at both shoulders and even driving them at 4 points - that's my favorite X4 I'll show you later.
Coming back to the topic - I'm thinking about using a similar kind of approach for the output stage - you're just giving me more inspiration now 😀
What do you think?
Cheers,
Valery
Hi David,
I started thinking in this direction as well.
It's already more than a year since I'm using current drive for the front-end (voltage amplification) designs.
Attached are a few pictures to explain the approach.
The biggest source of non-linearity is transconductance stage at the input - although there may be some better implementations - trying to keep it simple here. All the further operations with current are excellent, including the transimpedance stage - simple linear conversion, subtracting the current through a single resistor.
The whole thing demonstrates very good phase response, requiring very light and simple compensation (a few pF lead and 10-20pF shunt at the output), as well as rather low open loop distortion levels.
I've got many working design options now - driving the output CCSs at both shoulders and even driving them at 4 points - that's my favorite X4 I'll show you later.
Coming back to the topic - I'm thinking about using a similar kind of approach for the output stage - you're just giving me more inspiration now 😀
What do you think?
Cheers,
Valery
Attachments
I find basic theory pushes ideas that don't look like a classical "Blameless" circuit.
My interest is not to create a different circuit just for the sake of novelty but the easiness of a Blameless copy has perhaps meant other options have not been studied carefully.
I am interested in what an optimized amp would be if we started with a blank paper.
The usual Miller compensation around the VAS/TIS lowers the output impedance.
The comparative nonlinearity of the Vbe/Ic curve compared to the Ib/Ic curve, that you noted, would imply that current drive is more linear and the lower impedance may not be helpful.
Still just an idea I want to mull over, at this point.
Best wishes
David
My interest is not to create a different circuit just for the sake of novelty but the easiness of a Blameless copy has perhaps meant other options have not been studied carefully.
I am interested in what an optimized amp would be if we started with a blank paper.
The usual Miller compensation around the VAS/TIS lowers the output impedance.
The comparative nonlinearity of the Vbe/Ic curve compared to the Ib/Ic curve, that you noted, would imply that current drive is more linear and the lower impedance may not be helpful.
Still just an idea I want to mull over, at this point.
Best wishes
David
Valery, can you explain (for beginners like me... 🙂) whats the main difference when the
signal/substraction is kept in a form of voltage and when in form of current?
I mean there is voltage and current in every solution? What does it mean that the signal is rather a current?
For example in your schematic: Q15 is a standard common E amplifier stage,
when can one say that it's output is a voltage or a current one? Thanks!
signal/substraction is kept in a form of voltage and when in form of current?
I mean there is voltage and current in every solution? What does it mean that the signal is rather a current?
For example in your schematic: Q15 is a standard common E amplifier stage,
when can one say that it's output is a voltage or a current one? Thanks!
It's a matter of impedance.
Low output Z, high input Z = voltage transfer
High output Z, low input Z = current transfer
And for the sake of completeness:
Output Z equals input Z = power transfer
In the case of Q15, I'd say we're talking at least a few kOhms of Z_out vs. <100 ohms of Z_in at R3 || Q3_E.
Low output Z, high input Z = voltage transfer
High output Z, low input Z = current transfer
And for the sake of completeness:
Output Z equals input Z = power transfer
In the case of Q15, I'd say we're talking at least a few kOhms of Z_out vs. <100 ohms of Z_in at R3 || Q3_E.
Thanks sgrossklass!
"And for the sake of completeness" what about these combinations..?
low output Z + low input Z: ?
high output Z + high input Z: ?
"And for the sake of completeness" what about these combinations..?
low output Z + low input Z: ?
high output Z + high input Z: ?
Smart deduction😉
In fact I became interested in CE output devices both for feedback optimization and also for efficiency.
And once I started to consider a CE OPS then the conventional wisdom of EF OPS is not much use so I started to consider how best to drive such a CE OPS.
Yes, exactly so.
Still better than the exponential Vbe dependency.
Best wishes
David
Sgrossklass = thank you 🙂
Cortez - just a bit more of explanation - Q15 is NOT a standard common E amplifier stage here. Because of D7, that makes a big difference for this arrangement.
See attached. Circuit 1 is a standard simple current mirror. Circuit 2 is its equivalent, where Q1 is replaced with the diode (D1). This is a simple current mirror as well. Now - we can make it multiplying current, if we use emitter resistors with different values. Providing a rather linear current gain.
David - simplified schematic, attached in post #3 is just for principle demonstration. The key thing I'm sure - the amplifier is going to be organized, as outlined on picture 1 in that post (with the boxes).
Practical implementation, thinking from scratch, most likely would be:
- Complimentary pair as a transconductance stage and NFB junction point (probably a jFET pair);
- Sophisticated highly linear current conveyor;
- Highly linear transimpedance stage at the output.
Circuit 3, attached here, is a sophisticated current conveyor component, I have designed for one of my recent low-distortion amplifiers. It's an enhanced Wilson mirror, combined with an improved Hawksford cascode.
Precise, fast, highly linear, also demonstrating excellent overload behavior (soft clipping).
Now, the question is - do we need a buffer at the output for having the output impedance low enough? Probably yes. However, assuming certain amount of global NFB, its configuration requires more thinking and test simulations.
Cheers,
Valery
Cortez - just a bit more of explanation - Q15 is NOT a standard common E amplifier stage here. Because of D7, that makes a big difference for this arrangement.
See attached. Circuit 1 is a standard simple current mirror. Circuit 2 is its equivalent, where Q1 is replaced with the diode (D1). This is a simple current mirror as well. Now - we can make it multiplying current, if we use emitter resistors with different values. Providing a rather linear current gain.
David - simplified schematic, attached in post #3 is just for principle demonstration. The key thing I'm sure - the amplifier is going to be organized, as outlined on picture 1 in that post (with the boxes).
Practical implementation, thinking from scratch, most likely would be:
- Complimentary pair as a transconductance stage and NFB junction point (probably a jFET pair);
- Sophisticated highly linear current conveyor;
- Highly linear transimpedance stage at the output.
Circuit 3, attached here, is a sophisticated current conveyor component, I have designed for one of my recent low-distortion amplifiers. It's an enhanced Wilson mirror, combined with an improved Hawksford cascode.
Precise, fast, highly linear, also demonstrating excellent overload behavior (soft clipping).
Now, the question is - do we need a buffer at the output for having the output impedance low enough? Probably yes. However, assuming certain amount of global NFB, its configuration requires more thinking and test simulations.
Cheers,
Valery
Attachments
Hi Valery
Surely the phase is directly determined by the amplitude response, in other words, the system is minimum phase?
So the phase response can't be any better than any other similar circuit with the same "frequency response".
Best wishes
David
Last edited:
A bit late here so I will respond in more detail tomorrow.
But one possibility is to build a two section amplifier, with an IPS and a "Power VAS" (for want of a better term) and drive the speakers from that final CE.
Kind of a power amp version of a rail-to-rail op-amp, similar to the way the Blameless is a power version of an old op-amp.
I have already had a few ideas for a complementary JFET front end, looks excellent.
So my interest is how to connect that front end to a power second section that is basically complementary rail referenced CEs, preceded by EF buffers.
Best wishes
David
Last edited:
Right - rail-to-rail op-amp "on steroids" is pretty much in line with what I have in mind. You're right - output stage is the most interesting part here.
Certainly an interesting area. It just so happens that Linear Audio Vol 13 will have an article by Ian Hegglun on exactly this subject. You'll just have to wait a bit 😉
BTW There is a well-known design, published in Wireless World at the time, by Peter Blomley IIRC. He split the input signal in the two polarities and each polarity drove the output stage in current mode. Probably on Doug Selfs' website.
Edit: It's on Keith Snook's site,
http://www.keith-snook.info/wireles.../New Approach to class B Amplifier Design.pdf
Jan
BTW There is a well-known design, published in Wireless World at the time, by Peter Blomley IIRC. He split the input signal in the two polarities and each polarity drove the output stage in current mode. Probably on Doug Selfs' website.
Edit: It's on Keith Snook's site,
http://www.keith-snook.info/wireles.../New Approach to class B Amplifier Design.pdf
Jan
Last edited:
Jan, thank you for the link - right to the point.
Although I'm thinking about even more bold approach 🙂
I'll be back after a bit of more thinking.
Although I'm thinking about even more bold approach 🙂
I'll be back after a bit of more thinking.
I look forward to it.
Thanks for the link, unfortunately it doesn't work for me, just dumps me back at Keith Snook home.
I can find it on the menu, just doesn't open properly...hmm.
Best wishes
David
David, it doesn't work for me either, if I click the link.
But if I right-click and push "Download link" - it works.
I'm using Chrome browser.
But if I right-click and push "Download link" - it works.
I'm using Chrome browser.
I also use Chrome but when I R-click I don't see a "Download Link" option.
Best wishes
David
- Status
- Not open for further replies.