Yes it is..
I did it as in your second sim. Injecting current to the output of the complete amp. My simplification/approximation was to use DC supply with series resistance to the supply caps, with resistance high enough to be greater than the impedance of the PS caps at LF. This should be relevant to the actual impedance as I see it, since the rectifier 'pwm duty' is usually pretty low at normal listening levels. Anyway, I was just trying to find the 'connection' to what I was hearing, and I think there is something to it.
On bootstrapping (you mentioned op-amp bootstrap before), the classic bootstrap to VAS that is normally taken from the output of the amp, seems to benefit from being connected to the emitter instead (other side of emitter resistor, for those who have emitter resistors 
). That decreases the output impedance of the amp significantly in sim.
). That decreases the output impedance of the amp significantly in sim.
Across the audio band the PSRR of this amp is pretty low too, about 60dB, it is mostly due to supply voltage influencing the output devices, not the frontend. Maybe this has something to do with your feelings about bass.
On the positive side, simulation with cascode MOSFETs show excellent output stage PSRR (that's kinda expected) so that's another advantage of this circuit. It's like having a regulated power supply for each channel, but cheaper and without the headroom penalty of a linear regulated PSU.
On the positive side, simulation with cascode MOSFETs show excellent output stage PSRR (that's kinda expected) so that's another advantage of this circuit. It's like having a regulated power supply for each channel, but cheaper and without the headroom penalty of a linear regulated PSU.
I'm a bit curious about the output impedance with vs without cascode, have you tried simulating it? (just thinking a bode plot of output voltage while feeding AC current to the output). From my sims vs listening experiments, I seem to like it best <10mV @1A AC at LF. I know this is only sim numbers, and reality will be something else..
Just checked. With cascodes bootstrapped from the output voltage, the sim says the output impedance increases by 2% (from 17 mOhm to 19mOhm) using your amp. In other words, it makes no difference to the output impedance, this one is all about loop gain. Previous measurements I made confirm this, basically no difference to Zout.
Hmm... I haven't even simmed the output impedance. As long as it's lower than the wires, it's alright. 10 mOhms isn't going to change the frequency response of the speakers anyway. What matters is that the output impedance remains linear (ie, constant) with output current, otherwise it introduces current-dependent distortion. For the output stage, open loop Zout is 1/gm, it's just two sides of the same coin, so all the gm plots I've posted in the thread are also output impedance plots if you look at them upside down... so "constant output impedance vs current" is the same as "get a flat gm plot" which has been the goal of the thread.
It's interesting actually.
You could ABX it with a DPDT relay, one side sending the amp output straight to the speaker, the other through a 10-20 mOhm resistor.
I'm gonna bet the difference doesn't come from the resistance itself changing the frequency response, rather from the nonlinearity of the amp's output impedance depending on current. In this case, the ABX test would give a null result if the resistor is linear.
Another possible explanation would be the distorted current drawn by the woofer (huge amount of THD in the current, last time I measured one) creating harmonics across the amp's output impedance, which are then played back by the whole speaker. That's one possible explanation of why bi-wiring sounds better. That would give a positive ABX result.
Could also measure frequency response and THD with a real loudspeaker (not a resistive load)...
You could ABX it with a DPDT relay, one side sending the amp output straight to the speaker, the other through a 10-20 mOhm resistor.
I'm gonna bet the difference doesn't come from the resistance itself changing the frequency response, rather from the nonlinearity of the amp's output impedance depending on current. In this case, the ABX test would give a null result if the resistor is linear.
Another possible explanation would be the distorted current drawn by the woofer (huge amount of THD in the current, last time I measured one) creating harmonics across the amp's output impedance, which are then played back by the whole speaker. That's one possible explanation of why bi-wiring sounds better. That would give a positive ABX result.
Could also measure frequency response and THD with a real loudspeaker (not a resistive load)...
I'm thinking there is something with how the amp handles the back-emf from the woofer, I usually hear the improvement in the low end, possibly in the range where the woofer has it's resonance, and/or tuning frequency of the BR. Could be because there is a lot of phase variations of the current going on in those areas.
When it comes to mids/treble, I actually prefer a higher output impedance (/less loop gain) like a JLH class A etc. I have confirmed this preference with tweaking a lot of different amps, and harmonic profiles etc. That could actually have something to do with the current distortion..? But then again, it's common to have resistors in the XO to match mid & tweeter to the woofer..
It's all a mystery to me
When it comes to mids/treble, I actually prefer a higher output impedance (/less loop gain) like a JLH class A etc. I have confirmed this preference with tweaking a lot of different amps, and harmonic profiles etc. That could actually have something to do with the current distortion..? But then again, it's common to have resistors in the XO to match mid & tweeter to the woofer..
It's all a mystery to me
Maybe a conductance amplifier would be something for you. Very easy to implement with Mosfets.
That is my thinking as well. Small amounts (mOhms) of series impedance impacts the current phase. I've measured very-low-Z regulators which actually had -180º output Zphase, and 10mOhms straightens it right out! When the impedance is very low, it doesn't take much series R to swamp the reactive element.
Oh yeah, you gave me an idea: put a bootstrapped opamp in the output stage which amplifies the error voltage between the input and the output x1000, and outputs that, to be measured with an instrumentation amp or transformer. That can be done in any amp. The amplification is meant to make it easier to acquire it on top of the large common mode voltage. I've added it to the layout.
Capacitor 470-680nF in parallel with resistor in 68R solves the problem. It will be even better if you increase the current through the driver and put a resistor to 47R.
20R base resistors on MJL transistors it's a lot.
Last edited: