This input by Bonsai is also important, in addition to what I have already pointed. Just my opinion.
I will test THD at 20kHz. Maybe at 1W and 10W. 15W is too close at clipping.
Gain margin I don't know about.
AC Analysis shows the bandwidth frequencies and phase for the same.
I will also put the input RC before the potentiometer.
Gain margin I don't know about.
AC Analysis shows the bandwidth frequencies and phase for the same.
I will also put the input RC before the potentiometer.
One way to determine the value of the series inductor via sims is to add zero ohm resistors in drain or source of the mosfets to plot the current through them. The criteria are that the amp must remain stable at maximum expected capacitive load, but does not exceed maximum current of the devices, or trigger the overload shutdown. Indeed is 1 µH in most cases enough for stability, and 1 ohm in the case of 16A output devices.
Worse sounding than THD is IM, so you might do that as well. The "hard" test is with 19 & 23 KHz which will also display high order products at a 1 KHz interval.
No. Just to be clear. The bases are seeing very different impedances.
Other than the benefit of well behaved DC offset, having the bases see the same impedance makes for a better harmonic profile at low frequencies, believe it or not. Ratios between harmonics and their overall levels. At least I remember seeing this in the simulator a while ago.
I think this is one of the reasons why many people prefer a fet differential (with jfets the input impedance is much higher and you no longer need to worry about balance). I'm not saying this is the only advantage of jfet diff, and I'm not saying it has no disadvantages.
You see, at the end of the day this stuff is a bit more complex than what most folks are simulating here, and the only real answer is finding out in practice what is best, for each listener. That's my opinion.
Loop Gain Analysis for determining gain and phase margin are absolutely critical values that need to be assessed. There's been several posts by me, minek123 and Bonsai attesting to this. LTSPice will often happily run a circuit without observable problems even if it is inherently unstable.
I use the Tian method. But I know there are other method to assess it, each with their own strengths and weaknesses.
How can you say the bases sees different impedances?
Both see 22kOhm as far as I see.
What are the mechanisms that affects the harmonic profile particularly at low frequencies, and how low frequencies are we talking about?
Besides the DC offset which is a given, are these claims in the above quotation based on simulations only?
With µcap it's almost impossible to run transient analysis without seeing instability: the exception is when leaving maximum timestep = 0 (default) which displays a sinewave as a ragged mountain ridge. With timestep <=10n any instable audio amp will show oscillation, and contrary to LT, oscillators will start just as in a physical circuit. Sometimes that takes way too much time (Xtal oscillators) and in that case it's often enough to lower the Q from 10,000 to 100 or less).
All I can say is that I have observed odd behaviour in the low frequency hamonic profile. IIRC I would start with 50 Hz and simulate, then double the frequency, simulate, then double it again, simulate, and so forth. Perhaps you fellows should try this and see if you can spot something that strikes you as odd. I certainly did. It vanished with balanced impedances or jfet diff.
Microcap by default will give you a harmonic profile for various levels of input in a single screen (but not for various input frequencies, of course). IIRC this also helped me spot anomalies, along with looking at various low frequencies, things that are not being looked at by most folks here.
IF this is something important then I would not expect that a highly knowledgeable person will come in and say "this is it, you need to do this"! I just think there may be something to it, more than static DC offset. Therefore my opinion about balancing the impedances.
I would try to increase, just for simulation purposes, C2 to 470uF and especially C1, to 6800uF. MicroCap distortion simulations at low frequencies are often unrealistic and wrong, if coupling capacitors are used. Or make it DC coupled, just for simulation.
That wakes up memories where I observed such issues. Often the LTP "sees" different voltages and loads at the collectors. Changing the circuit in a way to minimize the differences lowers distortion considerably.
When I tested LM3886 back in 2009, it had rising distortion at low frequencies. I think it is because of the thermal feedback on the chip. But I thought you were talking about lineup's amplifier, post #1, which is the topic of the thread.
BTW, why that Borbely amplifier is shown?
Last edited:
I posted two examples off the top of my mind where the bases of the differential see equal impedances at all frequencies.
In my opinion this is important, and I wanted to make people think about it (I really think there is something to this that hasn't been discussed to this date, I only saw hints about it in this forum).
I am not suggesting that anything of that complexity be implemented here, this is supposed to be a simple build. I just think that the impedances should be approximately equal for AC, including low frequencies. Therefore, here is one more example for Lineup, with approximately equal impedances (at the very least he is learning something new here because he didn't get that his impedances are imbalanced).
In my opinion this is important, and I wanted to make people think about it (I really think there is something to this that hasn't been discussed to this date, I only saw hints about it in this forum).
I am not suggesting that anything of that complexity be implemented here, this is supposed to be a simple build. I just think that the impedances should be approximately equal for AC, including low frequencies. Therefore, here is one more example for Lineup, with approximately equal impedances (at the very least he is learning something new here because he didn't get that his impedances are imbalanced).
Last edited: