Class i discussed here by Kendall Caster-Perry:-
The Class i low-distortion audio output stage (Part 2) | EDN
Main benefit is a 'purely resistive' output impeadance. Includes a formal analysis of the circuit - might be an interesting one to explore in a future design, but I dont know it it is suitable for really high powers.
The Class i low-distortion audio output stage (Part 2) | EDN
Main benefit is a 'purely resistive' output impeadance. Includes a formal analysis of the circuit - might be an interesting one to explore in a future design, but I dont know it it is suitable for really high powers.
The article goes into the theory but does not present much by way of an actual circuit.
Part one does present the following as a starting point:
http://www.edn.com/design/consumer/...roduction--the-problem-with-push-pull-outputs
Putting the theory into practice would be a significant effort.When do we get started? 😀
Part one does present the following as a starting point:
An externally hosted image should be here but it was not working when we last tested it.
http://www.edn.com/design/consumer/...roduction--the-problem-with-push-pull-outputs
Putting the theory into practice would be a significant effort.When do we get started? 😀
Last edited:
There is other thread about Class i and siblings http://www.diyaudio.com/forums/solid-state/202684-class-i-siblings.html
lots of promises in the 1st part write up but I'm not seeing yet how this isn't just another local feedback error correction scheme
mixing output stage local linearizing feedback with a "non-switching" bias scheme has also been done
mixing output stage local linearizing feedback with a "non-switching" bias scheme has also been done
Hi Guys
Although one benefit is a resistive output impedance, I believe the "main" benefit is elimination of crossover distortion and related low-level nonlinearities.
Have fun
Kevin O'Connor
Although one benefit is a resistive output impedance, I believe the "main" benefit is elimination of crossover distortion and related low-level nonlinearities.
Have fun
Kevin O'Connor
Main benefit ???
Hi Andrew,
It's not as 'purely resistive' as you might think, rather it is Zo + Re, where Zo is about the output impedance of any other (regular) EC OPS. Moreover, we have two Re's, the top and the bottom one. If they aren't exactly equal they create another source of distortion (except in case of class-A operation, of course). I wouldn't see this as 'beneficial' feature of class-I.
See: http://www.diyaudio.com/forums/solid-state/202684-class-i-siblings.html#post2826537
and also: http://www.diyaudio.com/forums/solid-state/202684-class-i-siblings-14.html#post2852220
Cheers,
E.
Hi Andrew,
It's not as 'purely resistive' as you might think, rather it is Zo + Re, where Zo is about the output impedance of any other (regular) EC OPS. Moreover, we have two Re's, the top and the bottom one. If they aren't exactly equal they create another source of distortion (except in case of class-A operation, of course). I wouldn't see this as 'beneficial' feature of class-I.
See: http://www.diyaudio.com/forums/solid-state/202684-class-i-siblings.html#post2826537
and also: http://www.diyaudio.com/forums/solid-state/202684-class-i-siblings-14.html#post2852220
Cheers,
E.
Member
Joined 2009
Paid Member
We've seen very low distortion available from ClassAB for a long time now - why would Class i be of any interest other than intellectual ?
Member
Joined 2009
Paid Member
yes, sorry, sounds like I was pouring cold water on it. what would be nice is some idea of whether it sounds good ? subjective of course, but we are emotional animals 😀
No need for apologies. You were right: for a long time there are already very low distortions amps available.
Cheers,
E.
Cheers,
E.
I like how he graphed open loop stability over many levels of static offset.
Always wondered how best to proof stability through the crossing handoff.
Always wondered how best to proof stability through the crossing handoff.
The article claims low distortion before closing the loop. Is it attractive?
Hi Guys
Anything that reduces let alone eliminates crossover distortion is worth investigation.
However, a traditional output stage with a simple bias regulator that monitors output current is a simpler solution, requiring just a two-BJT regulator. Idle current is rock solid and fairly easy to set by selecting Re for the outputs. The downside with this simple approach is that Vq must be one Vbe drop - or two - where class-i and Doug Self's class-A bias servo (three BJTs plus two current sources) both allow a Vq that is quite low, accommodating low Re values.
If you follow the links to the article and read the three parts available, you'll see that device matching is important in class-i.
Have fun
Kevin O'Connor
Anything that reduces let alone eliminates crossover distortion is worth investigation.
However, a traditional output stage with a simple bias regulator that monitors output current is a simpler solution, requiring just a two-BJT regulator. Idle current is rock solid and fairly easy to set by selecting Re for the outputs. The downside with this simple approach is that Vq must be one Vbe drop - or two - where class-i and Doug Self's class-A bias servo (three BJTs plus two current sources) both allow a Vq that is quite low, accommodating low Re values.
If you follow the links to the article and read the three parts available, you'll see that device matching is important in class-i.
Have fun
Kevin O'Connor
Just for fun look up the Musical Fidelity A1, namely the output stage... so this has been done before, quite a long time ago, actually. There is also a better way of supplying the bias voltage, by injecting a small surrent into the feedback node. Then the input side of the differentials can be simply connected together (A1 connects only for AC using one coupling cap per side but this can be simplified to a straight-through DC connection).
Hi Guys
The A1 was clever but not in a good way. Overall THD was not that good.
Have fun
Kevin O'Connor
The A1 was clever but not in a good way. Overall THD was not that good.
Have fun
Kevin O'Connor
Yes, for a number of reasons, one of which was deliberately low diff stage gain. But, if you look at it, it's exactly the same approach as here, except the bias offset voltage is added on the feedback end.
Hi Guys
Where the class-i circuit is strictly a unity-gain output stage, the A1 was a complete amplifier. Diff gain and out device gain both worked towards linearity and bias control. Something is always compromised that way.
Have fun
Kevin O'Connor
Where the class-i circuit is strictly a unity-gain output stage, the A1 was a complete amplifier. Diff gain and out device gain both worked towards linearity and bias control. Something is always compromised that way.
Have fun
Kevin O'Connor
They did squeeze as much gain as possible from the input stage
by keeping it undegenerated but the following stage has only
a symbolic gain to add to the chain , hence about 54dB OLG
at most , although with a 100Khz bandwith.
- Status
- Not open for further replies.