Do you recall how well the simulations correlated with reality for the old version? That could give you some hints about the actual performance of the new version.
Either way. I'm looking forward to the measurements. The simulation results certainly suggest that it has potential.
Tom
The pre-driver and driver stages run in class A, and so their distortion can be made arbitrarily low. I think the difference is that I am running DC simulations based on Ebers-Moll and the square-law. The results are here. The theory says BJTs have lower distortion than MOSFETs at high power, but real devices may vary.
Ed
Ed
That's right these are A class,, but also works with huge current and voltage swings, so many factors influences their gain. Again, you can't ignore output load resistance, so you also can't ignore source impedance. That's my point.
I've never made simulations for older version. It was designed in trial-and-error way, looking at results on oscolloscope.
My simulation is all done in Blameless topology with same input and VAS stages. That eliminates the guess work around the driving impedance. IRFP140/IRFP9140 is one pair of the best of mosfets in simulation (better than Lateral type). However, they are still about 10dB worse THD than MJL3281/MJL1302. BTW, 2SC5200/2SC1943 performs as good as MJL3281/MJL1302.
I haven't verified that with real devices.
PS: I feel that BJT works very well when voltage driven (low impedance), such as a darlington VAS with miller cap.
I haven't verified that with real devices.
PS: I feel that BJT works very well when voltage driven (low impedance), such as a darlington VAS with miller cap.
Last edited:
Every time the bootstrap supply kicks in jacking up or down the rails voltage causing a half-wave rectified supply full of harmonics, that voltage supply is also fed to the input stage, is that ok?
Sure, but the results won't be directly applicable to anything but the boring old blameless/Lin/whatever you call it configuration.
From current source side (positive rail) input stage is suppressed very well. And because (sum of) emitter currents are fixed, also collector current can't change significantly, and even if some change occurs, it is still in this same direction on both collectors so do not changes balance of VAS and is further supressed. Aditionally fast frequency changes on rails are filtered by 10nF capacitors, NFB can easily get rid of even smallest impact. As you can see from THD results and from overdrive graph it has no impact on stability and harmonics.
Interesting approach, I like that . Will have to try it.As I wrote older version was built and measured (with sound card only), this mod is only simulation yet.
Hi, Borys, if you want to compare amplifiers, look at this:
It has comparable parameters: THD=0.000122% at 50W 10kHz 8Ohm.
I've compared both buffered and not-buffered VAS and decided to use non-buffered.
You are talking about DC offset or AC imbalance, because you mention impedance. not resistance.
Why are You trying the get as much open loop gain as possible ? Let me say You get 200dB OLG at start, is this the best way to get a good results ?
In other words, let me say we have amp A and amp B. Both have very similar parameters. Amp A has humongous amount of OLG and amp B has more linearized stages but less OLG.
What are advantages of amplifier A over amplifier B ?
Regards
P
It is important to keep gain high enough in audio range (at least in range, where it is most audible, let say 500Hz -7kHz) to reduce THD. At very low frequencies high gain is not necessary, but on the other hand it do not impacts frequency compensation (there is a lot of octaves to go down with amplitude even with single pole). From higher frequency side too much gain may produce higher level of high order harmonics, and makes amplifier harder to get enough phase margin at unity gain.
Going back to your question: there is no special advantage of having high gain at very low frequencies.
Btw. Frankly speaking I've set gain so high, just to show that it is possible, even with quite simple design 😉
Going back to your question: there is no special advantage of having high gain at very low frequencies.
Btw. Frankly speaking I've set gain so high, just to show that it is possible, even with quite simple design 😉
Keep in mind that class AB cannot be linearized completely without turning it into class A.
Ed
Interesting idea, that class A can be COMPLETELY linearized. Or do you mean only that class A is free from one kind of non-linearity? Could you explain your point in details?
We are not talking about "more", we are talking about "completely". 10dB doesn't matter, it can be easily compensated by increased gain.
For conventional 3-stage amps, 140dB open loop gain is about the highest that you can get.
To get 200dB, you need to add another gain stage. There is no elegant way that I know. You still have the restriction that the gain has to roll off to 1 somewhere to keep it stable.
To get 200dB, you need to add another gain stage. There is no elegant way that I know. You still have the restriction that the gain has to roll off to 1 somewhere to keep it stable.
I know that class A has lower distortions, you don't need to convince me about it. My point is that class A always has distortions, an we cannot eliminate it COMPLETELY. Moving from class AB into A we may only REDUCE distortions, not remove.
MOSFETs gradually go from a more or less quadratic region (strong inversion) to a roughly exponential region (weak inversion) when you reduce the current, the part in between is known as moderate inversion.