Wouldn't that be just the plain normal gm nonlinearity? You mean, given an operating point of 1A at 5V Vgs and then changing the Vgs +-delta would not give the same delta of currents? If you want that to happen you'd need to run two opposing stages in D2S mode, which has the goal of constant gm vs drive, over a given range.ionomolo said:
And, I've shown (gate oder gate-driver) currents in the plots, not gate voltages nor drain currents.
Ah, ok,ionomolo said:
that would mean some memory mechanisms. At LF I could only imagine thermal effects to play a role (that's why I had the idea to let the FETs run in constant power, which needs a modulated cascode, in my D2S study). But I'm not a device-level expert on this details, wrt to linearity issues, so maybe there is more...
Lumba Ogir said:
That capacitance is exactly what makes the amp stable. I have double checked it and i found that i was over-estimating it, that was making me predicting the pole at lower frequency, now everything is closer to reality. Any attempt at removing it (cascoding) mandates the addition of an extra pole to keep it stable, or displacing the cutoff frequency by also cascoding the driver.
I will buy some of the renesas devices because they will make a different amp, lower gain, higher cutoff frequency, more linearity, less degeneration... Actually building and listening to it is the only way to know what's better.
EDIT:
In a feedback loop, the gain has to fall below unity at 180º phase shift. There are four parameteres to touch:
· Pole frequency
· Open Loop voltage gain
· Slope
· Open-loop phase shift.
I find this to be an excellent amp for learning as the baseline circuit has a simplicity and performance that is very hard to match, but it has all the parameters that a more complicated amp would have. There are four devices contributing to rolloff and phase shift: The driver ccs, the driver, the mosfet and the output ccs. Simulations allow to fix the ccs to have an ideal behaviour to isolate the phenomena involving the active transistors. Choke-filtering the ccs's is another option, more strange buit i could lead to interesting results.
Anybody has experience about sound differences between high feedback, low cutoff frequency and low feedback, high cutoff frequency?
ionomolo,
slowly we are getting closer to the truth. The effect of the reverse transfer capacitance is especially apparent here as we are dealing with a tight closed-loop, speed is essential, the difference is considerable (10pF against 160pF). However,it cannot be removed not even decreased just stabilized since capacitances have voltage dependency.
slowly we are getting closer to the truth. The effect of the reverse transfer capacitance is especially apparent here as we are dealing with a tight closed-loop, speed is essential, the difference is considerable (10pF against 160pF). However,it cannot be removed not even decreased just stabilized since capacitances have voltage dependency.
Lumba Ogir said:
I'm not sure if stabilizing the capacitance would not ruin a very nice phenomenon that the amp has. When the current is big the transconductance is big, and the open-loop voltage gain is also bigger, then the voltage at the mosfet is low and the capacitance gets higher, so it somewhat compensates for the increased open-loop gain. Very strange but it seems that this nonlinearity could dynamically adjust the compensation. The idea is to minimize it to only the needed amount. This could be done by cascoding and adding a resistance at the emmiter of the common-base amplifier so it had a selectable impedance and thus a selectable magnification factor for the Cgs.
What i would do is try to make the amp faster by replacing the input transistor or doing something to it (degenerating, cascoding, etc).
ionomolo,
we have been talking about the significance of low and stable reverse transfer capacitance for a long time. We do not want it to vary decreasing the bandwidth thus causing various dynamic distortions, very bothersome to the ear. High bandwidth is needed for low distortion too, not just for stability. Vds must be firmly locked.
we have been talking about the significance of low and stable reverse transfer capacitance for a long time. We do not want it to vary decreasing the bandwidth thus causing various dynamic distortions, very bothersome to the ear. High bandwidth is needed for low distortion too, not just for stability. Vds must be firmly locked.
If you lock the capacitance you will have to lock it to the highest value. You don't have to see it as a good value going to worse but rather as the minimum admissible value going lower when it is possible to go lower.
I have a quantum mechanics exam tomorrow and i can't work more on this, you should not take this as gospel, but rather as a new idea to explore, an amp whose "cdom" goes lower when there is a circumstance that allows higher phase margins.
This deserves ultra detailed analisys and i will do it and post the results, but i can't do it now.
EDIT: Studying nonlinear things needs to be done in time domain and this makes things much harder, but i wonder what would do an amp with a varicap as cdom.
EDIT2: Sorry for expressing things in such a terrible way. The idea is that at high currents the transconductance increases, this means that less voltage increase will be requiered to get the same output, but if capacitance increases, then more current will be needed to get the same voltage increase, so the effects will partially cancel and lead to a more linear result. This is not straightforward as distortion cancelation usually leads to a more complicated distortion pattern (higher order harmonics) so perhaps it would sound "subjectively" worse. I repeat, my belief is that the circuit is so simple that it allows for the exploration of an incredibly broad range of things, as there are few parts to explore. Since the time needed to explore each one is not big as there are not many parameters involved, and also assuming that the original design is good enough to live with it even if the final one takes an eternity to arrive, there is no reason not to try everything, even if it seems silly at first sight.
I have a quantum mechanics exam tomorrow and i can't work more on this, you should not take this as gospel, but rather as a new idea to explore, an amp whose "cdom" goes lower when there is a circumstance that allows higher phase margins.
This deserves ultra detailed analisys and i will do it and post the results, but i can't do it now.
EDIT: Studying nonlinear things needs to be done in time domain and this makes things much harder, but i wonder what would do an amp with a varicap as cdom.
EDIT2: Sorry for expressing things in such a terrible way. The idea is that at high currents the transconductance increases, this means that less voltage increase will be requiered to get the same output, but if capacitance increases, then more current will be needed to get the same voltage increase, so the effects will partially cancel and lead to a more linear result. This is not straightforward as distortion cancelation usually leads to a more complicated distortion pattern (higher order harmonics) so perhaps it would sound "subjectively" worse. I repeat, my belief is that the circuit is so simple that it allows for the exploration of an incredibly broad range of things, as there are few parts to explore. Since the time needed to explore each one is not big as there are not many parameters involved, and also assuming that the original design is good enough to live with it even if the final one takes an eternity to arrive, there is no reason not to try everything, even if it seems silly at first sight.
Cdom needs to be ultrastable, and have very low dielectric absorption otherwise it will adversely affect sound quality.
I don't believe a varicap here is either needed nor desirable.
Good luck with your exam, Ion. When you are through please answer my latest emails with your thoughts.
Ciao,
Hugh
I don't believe a varicap here is either needed nor desirable.
Good luck with your exam, Ion. When you are through please answer my latest emails with your thoughts.
Ciao,
Hugh
Lumba Ogir said:ionomolo,
accoding to Albert, solving problems requires a higher level of consciousness than that created them.
I would define my method as "adaptive montecarlo". It is extremely powerful. Of course trying lower Cgs mosfets is a top priority and i have a low current power supply that will do a great job at feeding the ccs, so you will get the results you want. These are also much more linear so there is no point in having a variable cgs and i will probably try to cascode them for even faster results. Anyways, Cds will be of a similar magnitude to Cds so i need to see if this actually gives an improovement that is worth the complication. Cascoding forces to use higher voltages, and even if the dissipation at each transistor can be reduced, the drivers will have to be higher rated and this will restrict the choice. Everything is a compromise and the only way to know what's the best choice is to try everything.
Anyways i have to do other things before starting with this. There is a 3.5 MHz small component in all of my amps, in some notable cases it is there even when the amps are switched off and the supply caps shorted so it's not the fault of the amps. I have to find what's causing this, then tidy the lab, and finally experiment with this.
ionomolo,
reading your posts I get the impression that you are careful with the details. This design does require carefulness and patience. Besides the CFP, even the cascodes are prone to cause oscillation. Hopefully, signal integrity can be maintained without any deteriorating compensation and other muddying parts. Please remember, cascodes have multiple tasks, are not only there to hold Vds. Again, high bandwidth is necessary to reduce distortions, there`s not much talk about them being hard to measure but it does not mean that they are not present. They are indeed present to the ear and the ear dislikes them. Those distortions ruin the desirable soft musical sound, which actually makes the difference. It is not easily achievable.
reading your posts I get the impression that you are careful with the details. This design does require carefulness and patience. Besides the CFP, even the cascodes are prone to cause oscillation. Hopefully, signal integrity can be maintained without any deteriorating compensation and other muddying parts. Please remember, cascodes have multiple tasks, are not only there to hold Vds. Again, high bandwidth is necessary to reduce distortions, there`s not much talk about them being hard to measure but it does not mean that they are not present. They are indeed present to the ear and the ear dislikes them. Those distortions ruin the desirable soft musical sound, which actually makes the difference. It is not easily achievable.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.