I was thinking that even if there is a mismatch and say one half of the amp would have a few % more gain than the other half, that it would be unimportant. Since you mentioned a 'possible improvement' I thought I'd ask. I couldn't think of anything off the top of my head, I thought maybe you could.
jd
It is obvious from direct measurement, that 2nd harmonic drops when using balanced out operation, although not to zero, or unmeasurable. To get the best 2'nd harmonic cancellation, one would expect that uneven outputs would not be as effective in effecting harmonic cancellation. That is why the trimpot might be useful.
This design approach is what we like to call 'elegant' rather than 'sophisticated'. There are thousands of different designs, especially associated with IC's that use lots of cheap parts and design complexity.
We try for 'perfection' with really good parts, well matched, working together to make the most linear transfer function that we can make, especially avoiding higher order nonlinearities that can be generated by more complex schemes, such as tanh, etc.
This design approach is what we like to call 'elegant' rather than 'sophisticated'. There are thousands of different designs, especially associated with IC's that use lots of cheap parts and design complexity.
We try for 'perfection' with really good parts, well matched, working together to make the most linear transfer function that we can make, especially avoiding higher order nonlinearities that can be generated by more complex schemes, such as tanh, etc.
I would appreciate if you or anyone else would stop associating me with the practise of cloning, because I have never build any of the circuits presented in this or any other BT thread. Take credit for any circuit you want if it feeds your ego, but please leave me out of it. Thank you!!
Courage, good to see you here. I don't know about you in particular, but some have cloned the design to the best of their ability. In fact, most of my designs do tend to be cloned, sooner or later, it seems. That is why I resist putting up exact schematics of my designs. I think that this is what Justcallmedad was really saying, at least I hope so.
Perfect cancellation not only depends on perfect balance between two halves, but also that each half has exactly the same harmonic level and phase as the other. If you have perfect gain but unequal harmonics, it still doesn't 'perfect cancel' and then you're better of with some non-perfect balance. And I guess that would be reality: the two halves never have exactly the same level of harmonics, so best cancellation is with slightly UNbalanced outputs. With a pot, yes, to slightly UNbalance the system to get max harmonic cancellation. So it doesn't really matter whether R19 and R20 are matched, 1% should do.
So it seems that contrary to popular belief, best cancellation results from 'uneven outputs' although you can't predict how uneven.
jd
Both matching and selection of proper components - output MOSFETs do make a big difference.
I do have an access to sensitivity analysis. But I am sorry, I cannot spend much of my time with this. Or you wanted to hire me?
Of course. Look at the gain in http://www.diyaudio.com/forums/showthread.php?p=1930934 it's proportional with the MOSFET transconductance.
This is just another reason why this design is a PITA. High transconductance MOSFET devices are very hard to exactly match at, say, 1% AND over a wide range of drain current. Not to mention the gain which is also proportional to the load impedance (which is, to add insult to injury, never purely resistive). Not to mention the MOSFET AC parameters matching.
Last edited:
You forgot about non-linear input capacitances of output MOSFETs that cause signal rectification and dependence of their idle currents on frequency, hence dynamic distortions. Nulls will drift with frequency as well, that would be reflected by 2'nd harmonic distortion dependence on frequency. PS sag, ripples, noise, and non-linearity will be injected in sources, so such simplicity requires PS complexity.
I would not call it elegant, Sorry John and Pavel. It is primitive.
- Status
- Not open for further replies.