I thought the effect was fully explained on pages 145 and 146 of my power amplifier book. (6th edition) It says quite clearly that the base currents drawn by the input transistors are distorted and so create distortion in the voltage domain on passing through an impedance.
About seven years ago, I achieve THD 0,006% at 1kHz 100W/8Ohm without simulation. THD meter resolution is 0,003% minimum. Audio just an hobby, at that time I was video tape recorder product's engineer. But I did not care about slew rate and THD at 20kHz at that time. The full power bandwidth is 120kHz.
I learn simulation at this forum. Specially, from Dadod, Bonsai, and Keantoken.
I learn simulation at this forum. Specially, from Dadod, Bonsai, and Keantoken.
Can you explain a bit about the cause of these distorted currents? What is the physics behind it?
Hi Douglas,
Of course the base currents are distorted, otherwise no distortion across the input impedance (rather obvious, I would say). But where does this nonlinearity stem from? Is it the nonlinear Miller capacitance, the Early effect and/or a nonlinear loading of the current mirror or VAS?
Happy New Year,
Edmond.
Building real amplifiers with JFET inputs is not "hard". Many fine power amplifiers are built with JFET inputs. Each type of device has its plusses and minuses. Good engineers know how to mitigate the shortcomings and exploit the advantages of any given technology. Restricting one's self to only BJT inputs seems a bit close-minded, especially for one who is teaching power amplifier design to a large population.
Are JFET input stages a bit more expensive? Maybe, especially if one chooses to use a monolithic matched pair. Is saving a couple of dollars on an input stage always a priority? No. Is saving a little bit of money by paying pennies for 5532 op amps always a priority? Certainly not. John Curl and the folks at Ayre Acoustics would certainly not agree with ignoring JFET inputs, nor would the folks at Parasound and many others.
Cheers,
Bob
It is certainly true that distortion from SPICE simulations will often be worse than that of the real world, for a host of reasons, not the least of which include poor or simplistic transistor models, failure of the SPICE schematic to capture all real-world capacitive and inductive parasitics, and failure to capture coupling effects.
However, the distortion percentage quoted here is on the order of 0.0001%, or 120 dB down. It is almost certainly for a 1kHz measurement. Although I cannot speak to the given circuit, achieving THD this low at 1kHz is not terribly difficult. The MOSFET power amplifier I built and measured in the early 1980's measured less than 0.001% THD+N at 20kHz full power, and its THD by itself was nearly un-measurable at the time - well less than 0.0003%.
An amplifier with conventional 6 dB/octave loop gain roll-off will often have about 20X as much feedback at 1kHz as at 20kHz. So if that amplifier measures 0.001% at 20kHz, it may not be shocking if it measures 0.00005% at 1kHz if magnetic coupling of nonlinear currents and the like are well-managed.
If SPICE'd THD is 2X to 3X measured THD on a real amplifier with all of its real-world warts, I think that is no big deal, and I would commend SPICE for getting that close. If I simulated an amplifier and got 0.00006% and it measured 0.0002%, I would not be especially disappointed. I guess we always have to manage our expectations. YMMV in the real world always applies.
Cheers,
Bob
That is an obvious possibility, but unless you measured that base current's degree of nonlinearity, it is speculative. More importantly, if we accept that it is nonlinearity in the base current, you still failed to explain the source of that nonlinearity and whether the magnitude of the postulated source of the nonlinearity was sufficient to cause the observed results.
Once again, it is easy to cite Ccb nonlinearity or beta nonlinearity as known possible sources, but it is a bigger step to determine if those sources of nonlinearity have a chance of being sufficient to cause the observed increase in distortion.
Ccb nonlinearity is reasonably well modeled in SPICE. You can even do a thought experiment with a little bit of hand-waiving. How much might Ccb change when Vcb is about 40V and it is modulated by only 1V of common-mode signal? Is that change enough to cause enough non-linear current flow in the source resistance to account for the distortion increase?
Beta nonlinearity may not be wonderfully modeled in SPICE at high and low extremes of collector current, but when the devices are run in their broad sweet spot of collector current, and those currents vary little with signal (as in an IPS), then that nonlinearity may be too small to account for the observed measurements. Ditto for Early effect, since the common-mode signal amplitude is quite small compared to the nominal collector-base voltage for the IPS transistors.
When you observed the increased distortion with increased source impedance, did you run a SPICE simulation to see if it showed up there?
Cheers,
Bob
This is the best that I've come across explaining transistor non linearity due to early effect and to an emitter junction effect, to a lesser degree, on Ic.
Bipolar Junction Transistors
Bipolar Junction Transistors
Hi Edmond
If you look at the BJT hybrid pi model with the Miller Theorem it should be easy to see that if you look at the input Z the "input" Miller C plays an important role in the BJT nonlinear C and thus input distortion.
Cheers
Reodor
While this is very true, you have to look at the numbers to see if it is big enough to matter in this specific case. Of course, SPICE should properly take account of any Miller effect.
Cheers,
Bob
That's not much to write home about.
BTW: It is not an EC (AKA: Error Correction) circuit
Have nice new year
Reodor
Your language is easy to understand for non native English speaking like me. I'm waiting your book.
Your language is easy to understand for non native English speaking like me. I'm waiting your book.
Thank you for your kind compliment. Happy New Year!
Cheers,
Bob
Hi Bob,
How well would you expect the performance to be if the circuit was laid out and built by a non-professional? How about the average builder? If you recall the threads on the SymAsym amplifier, you'll also recall the increased performance due to a layout change. Then there were the offshoots from this project. Not surprising, given the amplifier was a member driven co-operative effort from several members. When you pursue 0.0002% THD performance, all kinds of things now matter. A person given the schematic and PCB (completed) may never reach that target depending on parts selection and the real bug-a-boo ... the matching of certain parts.
I also think it's fair to say that both you, Bob, and Doug Self, avoid many pitfalls subconsciously without having to actually think about it. The same can be said for advanced hobbyists who have done a lot of work in this field. The average person won't even know what situations to avoid. Not completely anyway.
-Chris
True, but then we can consider you an expert in the same way that Doug Self is an expert. I am not surprised that you can put together an amplifier that performs that well. But this isn't the question.
How well would you expect the performance to be if the circuit was laid out and built by a non-professional? How about the average builder? If you recall the threads on the SymAsym amplifier, you'll also recall the increased performance due to a layout change. Then there were the offshoots from this project. Not surprising, given the amplifier was a member driven co-operative effort from several members. When you pursue 0.0002% THD performance, all kinds of things now matter. A person given the schematic and PCB (completed) may never reach that target depending on parts selection and the real bug-a-boo ... the matching of certain parts.
I also think it's fair to say that both you, Bob, and Doug Self, avoid many pitfalls subconsciously without having to actually think about it. The same can be said for advanced hobbyists who have done a lot of work in this field. The average person won't even know what situations to avoid. Not completely anyway.
-Chris